Evo-Devo and Phylogenetics

Alessandro Minelli

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Evo-Devo’s Contribution to Phylogenetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Partitioning the Phenotype into Individual Characters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Genes Versus Environment and the Genotype ! Phenotype Map . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Saltational Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Developmental Genes and Phylogenetic Inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Evolving Gene Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
The Phylotypic Stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Phylogenetic Signal from Heterochronic Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Phylogenetics to the Benefit of Evo-Devo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Selecting New Model Species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Abstract
The contribution of evolutionary developmental biology (evo-devo) to phyloge-
netics has two aspects. The first is methodological: how to partition the phenotype
into independent characters, in the light of the evolvability and modularity of
developing systems. Evolvability, the ability to produce heritable phenotypic
variation, has taken central role in explanations of evolutionary change, together
with an increasing appreciation of the complex relationships between genotype
and phenotype, which are characterised by (1) pleiotropy, (2) the involvement of a
large number of genes in controlling single phenotypic traits, (3) the presence of
polyphenism due to the influence of external, nongenetic factors, and (4) the
modular architecture of developing systems. This allows for the occasional

A. Minelli (*)
University of Padova, Padova, Italy
e-mail: alessandro.minelli@unipd.it

# Springer International Publishing Switzerland 2016 1

L. Nuño de la Rosa, G.B. Müller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_40-1
2 A. Minelli

manifestation of saltational evolution. The second contribution of evo-devo to

phylogenetics relates to specific sources of information that can be used in
phylogenetic analysis, as provided by differences in the spatial and temporal
patterns of expression of developmental genes or whole gene regulatory networks
and by heterochronic patterns, especially in the framework of sequence
heterochrony where changes in the temporal sequence of individual developmen-
tal events are considered relative to other events in the ontogeny of the same
organism. In turn, a sound understanding of phylogenetics can benefit evo-devo
in the selection of new model species.

Keywords
Heterochrony • Homology • Model species • Phylogeny • Saltational evolution

Introduction

Evolutionary developmental biology, or evo-devo, emerged as an independent

discipline within the life sciences in the last quarter of the twentieth century.
However, to some extent its historical antecedents can be traced back to the efforts
of Etienne Geoffroy Saint-Hilaire (1772–1844), who systematically looked for
equivalent structural elements in the body plans of animals as different as a verte-
brate, a crayfish, and a squid, and to the long, although discontinuous, tradition of
studies on heterochrony. When Ernst Haeckel (1834–1919) first introduced the latter
term, this was intended to label a deviation from Haeckel’s biogenetic “law”
according to which ontogeny (the development of the individual) recapitulates
phylogeny (the evolutionary history of the species). Heterochrony thus referred to
circumstances where the comparative study of developmental sequences of different
animals cannot be straightforwardly used to infer their evolutionary relationships. In
the twentieth century, these deviations from the biogenetic law were the subject of
Gavin de Beer’s (1899–1972) ground-breaking work. This author established that an
animal’s ontogenetic progression towards sexual maturity does not necessarily
proceed in strict conjunction with the development of its nonreproductive (somatic)
structures. This decoupling allows for the two processes to run at different pace and
eventually to evolve via changes in the relative time of onset or offset of somatic
versus reproductive development or in their relative speed. This way, heterochrony
emerged as a pervasive and variegated developmental basis of evolutionary change
(Gould 1977).
However, when evo-devo eventually took form in the 1980s, this discipline’s
focus was largely divorced from those original ties to phylogenetics. A quarter of
century later, Wiens et al. (2005) could still write that up to the time the overall
contribution of evo-devo to phylogenetics had been quite small. However, a growing
appreciation of the mutual benefits that can derive to both evo-devo and phyloge-
netics from reciprocal interactions has surfaced at last in recent years.
Evo-Devo and Phylogenetics 3

The contribution of evolutionary developmental biology (evo-devo) to phyloge-

netics has two aspects. The first is methodological: how to partition the phenotype
into independent characters, in the light of the evolvability and modularity of
developing systems. Evolvability, the ability to produce heritable phenotypic varia-
tion, has taken central role in explanations of evolutionary change, together with an
increasing appreciation of the complex relationships between genotype and pheno-
type. The second contribution of evo-devo to phylogenetics relates to specific
sources of information that can be used in phylogenetic analysis. In turn, a sound
understanding of phylogenetics can benefit evo-devo in the selection of new model
species.

Evo-Devo’s Contribution to Phylogenetics

The advent of evolutionary developmental biology offers indeed new opportunities

to extract phylogenetic information from a comparison of developmental schedules
of different species (Telford and Budd 2003; Minelli et al. 2007). Evo-devo’s
contribution to phylogenetics has two aspects. The first is methodological: how to
partition the phenotype into independent characters; the second relates to specific
sources of information to be used in phylogenetic analysis, as suggested by
heterochrony or by comparative patterns of expression of developmental genes.

Partitioning the Phenotype into Individual Characters

One of the main steps in a phylogenetic analysis is filling a data matrix: the rows are
the taxa (usually, species) to be compared, the columns are the characters for which
the taxa are compared. A basic requirement is, to include only mutually independent
characters, to avoid giving more weight to those that are instead interdependent. In
practice, however, it is often difficult to determine if two characters are actually
independent or to which extent.
Independence between two characters means that changes in one of them are not
necessarily accompanied by changes in the other. On one side, this lack of correla-
tion can be due to a lack of functional coupling, on the other side it reveals the
autonomy of the developmental processes on which each character depends, a
circumstance that in principle corresponds to the expression of different genes, or
at least to spatial, temporal, or quantitative differences in the expression of the same
genes. Here is an area where evo-devo can positively contribute to a phylogenetic
analysis. To see how, we must refer to two key concepts of evolutionary develop-
mental biology: evolvability and modularity.
With the advent of evolutionary developmental biology, evolvability, i.e., the
ability to produce heritable phenotypic variation (Hendrikse et al. 2007) has taken
central role in explanations of evolutionary change, together with an increasing
appreciation of the complex relationships between genotype and phenotype (the
so-called genotype ! phenotype map; Pigliucci 2010; Wagner and Zhang 2011).
4 A. Minelli

This complexity has many causes, among which (1) pleiotropy, i.e., the fact that
the expression of one gene has commonly an effect on many phenotypic traits,
(2) the involvement of a large number of genes in controlling the developmental
processes culminating in the production of a single phenotypic trait, (3) the influence
of external, nongenetic factors (discussed in a later section of this entry), and (4) the
modular architecture of developing systems.
To some extent at least, a developing organism can be described indeed as a
system of local units, or modules, dominated by specific developmental dynamics,
such as those generating a leaf primordium in a plant or those responsible for the
production of segments in an insect.
Evolutionary changes are also often modular, affecting individual characters that
emerge as hot points of morphological evolution. In many rapid radiations, the
explosion of phenotypes is essentially restricted to large variation in a well-
circumscribed module. This is the case with the copulatory structures of a great
number of insect groups, with the chewing structures (mastax) of rotifers, and with
the stamens or the petals in flowering plants.
Even within a series of homologous parts do individual elements, or group of
elements, often behave as partly independent modules. Examples in the animal
kingdom are found, e.g., among the teeth of mammals (incisors, canines, premolars,
molars) and the segments of arthropods (e.g., in insects, thoracic segments with legs
versus abdominal segments without legs). In plants, examples of developmentally
independent modules are the nectariferous petals of Delphinium (Ranunculaceae)
and the individual petals and stamens of Bauhinia (Fabaceae): in this genus there are
species like B. blakeana, with five petals and three fertile stamens, alongside species
like B. divaricata, with two petals and one functional stamen only.
In phylogenetic analyses, understanding or at least estimating the modularity of
the developmental processes underlying the morphological traits of species or
lineages to be compared is important also when the evolutionary changes in devel-
opmental processes have been systemic, affecting many dimensions and body parts
in integrated way. Systemic, that is, nonmodular change may conceal the actual
relationships between phylogenetically related taxa. This is one of the contexts in
which morphological evidence must be used most cautiously, and we must definitely
acknowledge its subordinate importance in respect to comparative molecular data. A
fitting example is provided by the duckweeds, long considered to form an easily
diagnosable plant family, but eventually reduced to a subfamily of the Araceae
(Henriquez et al. 2014). The systemic evolution the duckweed lineage has under-
gone has completely cancelled the modular architecture of the other Araceae and,
indeed, of the overwhelming majority of plants, only leaving behind a thallus-like
blob of green matter, sometimes (but not always) accompanied by simple roots and
occasionally producing a rudimentary stamen or carpel, all that remains of a typical
flower.
Evo-Devo and Phylogenetics 5

Genes Versus Environment and the Genotype ! Phenotype Map

One of the reasons why an organism’s phenotype cannot be fully predicted from the
genotype is the frequent occurrence of alternative phenotypes in the absence of
genetic differences: this occurs when specific environmental cues are “interpreted”
by the developing system as a switch between alternative pathways. This phenom-
enon is known as phenotypic plasticity and the multiplicity of resulting phenotypes is
described as a case of polyphenism (reviewed in Fusco and Minelli 2010). Environ-
mental influences are often due to differences in the relative length of day and night,
a result of which is the seasonal polyphenism of some butterflies, e.g., the European
Araschnia levana, with two generations per year (a spring and a summer one),
dramatically different in their wing color patterns to the extent that they were
originally described as different species. The temperature at which the embryo is
exposed during incubation is involved instead in the environmental determination of
sex in the American alligator, many turtles, and other reptiles, while in Schistocerca
gregaria and other grasshoppers mechanical stress due to exceedingly frequent
contacts of juveniles with theirs conspecifics results in the production of gregarious
and migratory adults rather than solitary and sedentary ones.
The divide between environmentally controlled polyphenism and genetically
determined polymorphism, however, is not necessarily strong. How easily this
divide can be crossed is shown by the pea aphid (Acyrthosiphon pisum), a species
where males as well as females occur in two different morphs, winged and wingless,
respectively. The mechanisms responsible for the presence versus absence of wings
are different in the two sexes: the male morphs represent a genetic polymorphism,
whereas the female morphs depend on the photoperiod. However, the developmental
pathways leading to these alternative phenotypes are nearly the same in both sexes:
the product of the gene locus (aphicarus) controlling wing development in the male
is also involved in the polyphenic response of the female.
There is growing evidence that populations harbor variable amounts of cryptic
variation, that is, of variation that is not expressed under the environmental condi-
tions under which the population currently lives; a change of external conditions,
however, may uncover this variation and cause the expression of novel phenotypes.
This can be of consequence in phylogenetic analyses. On the one hand, the previ-
ously unobserved phenotypes may wrongly suggest a phylogenetic distance quite
higher than eventually demonstrated by molecular studies; on the other, the newly
expressed phenotypes can offer new targets to selection and thus accelerate evolu-
tionary divergence and perhaps the emergence of evolutionary innovations (Moczek
et al. 2011).

Saltational Evolution

Before the advent of evo-devo, a serious obstacle to reconstructing phylogeny was

the nearly universally (although mostly tacitly) accepted principle, that evolution
necessarily proceeds by progressive accumulation of small changes; as a
6 A. Minelli

consequence, species differing in very conspicuous aspects could hardly be

acknowledged to be phylogenetically close relatives. The strength of this precon-
ception has been strongly reduced by evo-devo. Several lines of evidence concur
indeed in demonstrating that the changes in developmental processes necessary to
obtain a new, strongly divergent phenotype are not necessarily proportional to the
morphological distance between the old and the new phenotypes. On the other hand,
in many evolutionary lineages, some hypothetical phenotypes differing from those
occurring in nature only in minor detail do never occur because of internal con-
straints in the developmental processes by which the phenotype is produced. As a
consequence, the actual distribution of phenotypes within a clade may reflect
developmental constraints (inequalities of evolvability in different directions) rather
than phylogenetic affinities, thus inviting caution in the course of phylogeny recon-
structions. Segment number in centipedes offers a case in point.
All adult specimens of all centipede species have an odd number of leg-bearing
segments. This number is fixed and identical in all species of some subgroups (e.g., it
is always 15 in all of the ca. 1000 species of lithobiomorph centipedes described to
date) but variable in others, often even within a brood issued from the same parents,
for example, between 51 and 59, but limited to the odd values: specimens with
52, 54, 56, or 58 pairs of legs simply do not occur. Thus, moving from a “permitted”
number, e.g., 57, to one of the arithmetically closest values (56 or 58) is not possible.
Nevertheless, a much larger change, i.e., sudden and likely very recent duplication of
the number of leg pairs has been observed in a lineage of scolopendromorph
centipedes. Most of the ca. 700 species belonging to this clade have a fixed number
of 21 pairs of legs, although several have 23, with a single species (Scolopendropsis
bahiensis) including specimens with 21 leg-bearing segments along with others with
23. A duplication of the whole set of trunk segments has been suggested (Minelli
et al. 2009) to account for the origination of the closely related Scolopendropsis
duplicata, a newly discovered species where leg-bearing segments are either 39 or
43. This species has likely diverged from S. bahiensis quite recently, as the dramatic
increase in segment number is the only obvious difference between the two
Scolopendropsis species. It has been hypothesized that the duplication of trunk
segment number, a phenotypically major leap, was very likely the effect of a
minor genetic and developmental change.

Developmental Genes and Phylogenetic Inference

The most conspicuous body of information generated thus far within evolutionary
developmental biology is about the so-called “developmental genes,” i.e., genes
demonstrably involved in the control of specific ontogenetic events or in the shaping
of specific traits of body architecture. Data spans from the mere identification of
these genes and of their nucleotide sequence, to the temporal and spatial patterns of
expression, and the mechanisms by which the expression of these genes is modu-
lated and the way by which, in turn, their products modulate the spatial or temporal
expression of other genes.
Evo-Devo and Phylogenetics 7

These genes can be studied at different levels for their potential phylogenetic
signal.
A first step is to use the gene sequences to reconstruct the phylogeny of the
organisms from which the genes have been isolated. In the case of animals, several
authors have looked at the Hox genes as to privileged genes supposed to carry
important phylogenetic signal because of their roles in controlling key aspects of the
animal’s body architecture such as the orderly sequence of organs along the animal’s
antero-posterior body axis. There are studies, for example, where Hox gene
sequences are used in reconstructing the relationships among the bilaterian phyla
or the major clades within the Arthropoda.
The next step is to search for homologies at the level of gene expression patterns.
This way, largely accepted homologies between body regions of distantly related
arthropod taxa, such as insects and arachnids, have been traced by comparing the
expression patterns of Hox genes.
These patterns are nevertheless subject to evolution, because of gene duplication
followed by functional divergence of the paralogous copies thus obtained, or by
changes in the gene’s regulatory sequences, not to mention gene loss. As a conse-
quence, we observe changes in the spatial extent or in the timing or level of gene
expression.
Examples of the different way in which a morphological trait can be modified by
changes in the regulation of the expression of the same genes are provided by the
pattern of dorsal bristles on the thorax of Diptera, which is controlled by the
expression of the gene scute. Changes in the spatial expression of scute account
for the differences between more distantly related taxa, such as Ceratitis capitata,
Drosophila melanogaster, and Calliphora vicina, whereas differences in the timing
of this gene’s expression underlie the differences in the bristle pattern of Calliphora
vicina compared to another calliphorid, Protophormia terranovae. Changes in the
expression level of just two genes (bone morphogenetic protein 4 and calmodulin)
are instead responsible for the conspicuous differences in beak shape among
Darwin’s finches (Geospiza).
With the rapidly increasing knowledge on gene control cascades, attention has
shifted from individual genes to whole gene regulatory networks. An exceptional
example of the evolvability of developmental gene networks has been revealed in a
comparison of notochord development in the pelagic urochordate Oikopleura and
the ascidian Ciona intestinalis. In the latter, some 50 genes are known to be activated
downstream of Brachyury, but 24 of them do not have a homologue in the small,
very compact genome of Oikopleura. Some of the latter have undergone a lineage-
specific duplication, but less than a half of them are apparently expressed in the
context of notochord formation.
However, the different components of a gene regulatory network do not neces-
sarily evolve at the same pace. For example, within the gene regulatory network
controlling the specification of the endomesoderm in nematodes, evolution is most
rapid for some genes involved in the specification of blastomere identity, as
suggested by a comparison between the genomes of Haemonchus contortus and
Brugia malayi (Maduro 2006).
8 A. Minelli

Evolving Gene Functions

In animals as distantly related as are squids, insects, and vertebrates, the morpho-
genesis of the eye is controlled in part by the lineage-specific homologues of the
same genes. The best known and arguably the most important of these genes is Pax6
(also known as eyeless in Drosophila). The widespread involvement of Pax6/ey
homologues in eye morphogenesis has suggested a common (monophyletic) origin
of all animal eyes (Gehring and Ikeo 1999), despite the gross morphological
differences between ciliary-type eyes, as are those of vertebrates, and
rhabdomeric-type eyes, as are those of insects, a large structural difference that
would instead suggest that eyes originated at least twice independently.
However, its involvement in building eyes is not necessarily the original devel-
opmental role of Pax6. First, the expression of Pax-6 is not restricted to the eyes. For
example, in the squid, Pax6 expression extends to the brain and the arms; in
vertebrates, to a large part of the nervous and sensory systems, including nasal
placodes, diencephalon, latero-ventral hindbrain, and the spinal cord; in Drosophila,
its homolog ey is expressed, other than in the eye, also in the brain and the ventral
nerve cord. Second, Pax6 homologues are also present in eyeless animals, for
example, in the roundworms (nematodes) and in the sea urchins. In the latter, a
Pax6 homolog is expressed in the tube feet. A likely conclusion is that Pax6 was a
patterning gene, originally expressed in the head, which has been co-opted several
times in the regulation of eye development.
Dramatic functional changes have been recorded in the evolution of two members
of the Hox gene family in some arthropod lineages. Hox genes, as mentioned above,
are best known to specify positions along the main body axis of bilaterian animals. In
the arthropods, however, one of these genes (re-named here fushi tarazu) is involved
instead in the segmentation of the trunk and also (limited to the insects) in
neurogenesis. Another Hox gene (zerknüllt, shortly zen) is involved in dorso-ventral
patterning. In the flies (Diptera), a duplication of zen has given rise to a new
functionally divergent gene, bicoid. In Drosophila, bicoid is required for the normal
development of head and thorax, and in another dipteran, the scuttle-fly Megaselia
abdita, it is also required for the development of four abdominal segments.
These examples show that major changes in gene functions do not necessarily
determine an acceleration of morphological evolution. In other terms, homology at
the level of genes, and genes expression patterns, does not necessarily suggest
homology of morphological features and vice versa.

The Phylotypic Stage

The independence of developmental modules is limited by constraints, more evident

at specific times along the ontogeny: specifically, largely invariant stages shared by
(most of) the members of a large group such as vertebrates, or insects, can be often
recognized. These stages do not coincide with the earliest embryonic stages, which
are dramatically affected by conditions such as the amount and spatial distribution of
Evo-Devo and Phylogenetics 9

the yolk in the egg, which is sometimes very different between closely related
species. For example, the sea urchin Heliocidaris tuberculata produces small eggs,
with a modest amount of yolk, which develop into a typical pluteus larva, but the
very closely related species H. erythrogramma, barely distinguishable from
H. tuberculata in the adult stage, produces instead much larger eggs, full of yolk,
from which a juvenile develops directly, bypassing the conventional larval stage.
The initially divergent developmental trajectories converge however towards a later,
much more conserved stage (Richardson 2012). This stage, which is called the
phylotypic stage, is sometimes recognizable as characteristic for a whole phylum,
although it must be acknowledged that often, rather than a point in development,
what is conserved is instead a phylotypic period. This term refers to a more or less
extended segment of the developmental trajectory, within which the traits shared by
the members of a phylum are more or less faithfully conserved among a smaller or
larger number of species. As expected, gene expression is maximally conserved
around the phylotypic stage or period.

Phylogenetic Signal from Heterochronic Patterns

From the old-fashioned perspective of Haeckel’s recapitulation principle (ontogeny

recapitulates phylogeny), heterochrony was nothing but noise obscuring the poten-
tial contribution of a detailed knowledge of ontogeny to the reconstruction of
phylogeny. Subsequent studies, however, have shown that heterochrony per se can
be informative about affinities, that is, in technical language, that heterochrony may
contain phylogenetic signal.
The traditional approach to heterochrony focused on developmental changes in
size and shape relationships: in terms of growth heterochrony, two major patterns
were distinguished, paedomorphosis and peramorphosis, according to whether
maturation is anticipated or delayed and/or the growth period is shortened or
extended, respectively.
Recognizing these evolutionary patterns of change in the developmental
sequences of the species to be compared can be dramatically important to avoid
serious pitfalls in the reconstruction of phylogeny. A good example is provided by
salamanders. Several lineages of salamanders have evolved via paedomorphosis,
that is, they retain throughout their life larval traits such as the presence of external
gills. This has serious consequences on a phylogenetic analysis based on morphol-
ogy. Lineages that have independently evolved by paedomorphosis will likely
cluster together, irrespective of their actual affinities. In some phylogenetic analyses,
most paedomorphic families (Amphiumidae, Dicamptodontidae, Sirenidae,
Proteidae) cluster indeed in a single clade, including also individual paedomorphic
representatives of the Plethodontidae and Ambystomatidae. This obscures the actual
affinities of these lineages. To uncover the latter, it is not necessary to rely on
molecular rather than morphological evidence. The same result is obtained by
excluding from the data matrix of morphological data those traits that are affected
by paedomorphosis, which have been acquired independently by these different
10 A. Minelli

lineages, accompanied by the loss of lineage-specific traits retained instead by their

nonpaedomorphic relatives (Wiens et al. 2005).
An increasing appreciation of the modularity of developmental processes has
fostered a new approach to heterochrony, termed sequence heterochrony (Smith
2001), focused on the changes in the temporal sequence of individual developmental
events relative to other events in the ontogeny of the same organism. Given two
events A and B in a developmental sequence, these can occur in one of the following
orders: (i) A occurs before B, (ii) A and B are simultaneous, or (iii) A occurs after
B. Translated into numerical codes, these timing relationships are assembled in a
matrix which is subsequently subjected to phylogenetic analysis according to the
current methods.
In the flowering plants, heterochronies in the production of individual floral parts
can be responsible for conspicuous and phylogenetically informative differences.
For example, a number of clades in the legume family (Fabaceae) are characterized
by heterochronies such as the anticipation or retardation in the production of a whole
whorl with respect to another (e.g., stamens vs. petals) or of a single organ (e.g., a
sepal or a petal) with respect to the other elements of its whorl.
In vertebrates, the relative times at which the fore and hind limbs differentiate are
characteristic of different clades (Bininda-Emonds et al. 2007). In the primitive
condition, as seen in cartilaginous and bony fishes, forelimbs develop earlier than
the hind limbs. In the frogs (anurans), hind limb development precedes instead the
differentiation of the anterior pair of appendages. In most of the remaining verte-
brates, the development of the two limb pairs is nearly synchronous.
A study of sequence heterochrony involving numerous characters has proved
effective in fixing the phylogenetic position of the turtles in the phylogeny of
amniotes (Werneburg and Sanchez-Villagra 2009).

Phylogenetics to the Benefit of Evo-Devo

Selecting New Model Species

Despite the pervasive comparative attitude that differentiates evo-devo from many
other disciplines in the life sciences, most of the experimental results thus far
contributed by evolutionary developmental biology have been obtained on a small
number of model species. The list of evo-devo’s choice model organisms includes
animals such as mice (Mus musculus), chicken (Gallus gallus), zebrafish (Danio
rerio), the sea squirt (ascidian) Ciona intestinalis, a few sea urchin species such as
Heliocidaris tuberculata and H. erythrogramma, the fruitfly Drosophila
melanogaster, and a tiny nematode worm (Caenorhabditis elegans), plants as the
thale cress (Arabidopsis thaliana), tomato (Lycopersicon esculentum), snapdragon
(Antirrhinum majus), and rice (Oryza sativa) plus the moss Physcomitrella patens.
Except perhaps for Ph. patens, all these organisms were selected only because of
practical advantages, such as short generation time and easy adaptation to artificial
environment. Of recent, however, most suggestions for new entries to be added to
Evo-Devo and Phylogenetics 11

the list of model organisms advocate the phylogenetic position of a species as

anticipating its value in future comparisons (Milinkovitch and Tzika 2007). How-
ever, there are no scientific reasons to expect that the topology of the phylogenetic
tree will inform us unambiguously about historical changes affecting characters
other than those that have been used to build the tree. Unfortunately, there are no
macroevolutionary laws suggesting strong consistent trends in character variation
across the different branches of the tree.
Attention has been primarily targeted towards “basal” representatives of a smaller
or larger branch of the tree of life (i.e., organisms belonging to lineages that are
considered to have branched off early from the common ancestor): the expectation is
that a “basal” branch will be a good proxy for the unknowable common ancestor of a
major clade. However, time (thus, opportunity to change) has run to the same length
for all branches stemming from the same common ancestor, irrespective of their
branching order; in other terms, a “basal” species is not necessarily a more conserved
model of an ancestor, that is, a repository of primitive character states (Jenner 2006;
Minelli and Baedke 2014).
The inadequate concern hitherto demonstrated by evo-devo researchers for com-
parisons within a suitable phylogenetic context is also shown by the fact that many of
the popular model species belong to taxonomically small or very small genera, e.g.,
Arabidopsis (12 species), Ciona (9), Heliocidaris (6), Caenorhabditis (4), Gallus
(4), and Physcomitrella (2). From this perspective, a welcome new entry is the scarab
beetle genus Onthophagus, with about 2000 species, some of which are emerging as
evo-devo models for the study of the developmental genetic basis of evolutionary
novelties, their head and thoracic horns.

Cross-References

▶ Developmental Homology
▶ Developmental Plasticity and Evolution
▶ Evolvability
▶ Genotype-Phenotype Map
▶ Heterochrony
▶ Modularity in Evo-Devo
▶ Palaeo-Evo-Devo
▶ Phylotypic Stage

References
Bininda-Emonds ORP, Jeffery JE, Sánchez-Villagra MR, Hanken J, Colbert M, Pieau C,
Selwood L, Ten Cate C, Raynaud A, Osabutey C, Richardson MK (2007) Forelimb-hindlimb
developmental timing across tetrapods. BMC Evol Biol 7:182
Fusco G, Minelli A (2010) Phenotypic plasticity in development and evolution: facts and concepts.
Introduction. Philos Trans R Soc B Biol Sci 365:547–556
12 A. Minelli

Gehring WJ, Ikeo K (1999) Pax 6: mastering eye morphogenesis and eye evolution. Trends Genet
15:371–377
Gould SJ (1977) Ontogeny and phylogeny. Belknap Press of Harvard University Press, Cambridge,
MA
Hendrikse JL, Parsons TE, Hallgrímsson B (2007) Evolvability as the proper focus of evolutionary
developmental biology. Evol Dev 9:393–401
Henriquez CL, Arias T, Pires JC, Croat TB, Schaal BA (2014) Phylogenomics of the plant family
Araceae. Mol Phylogenet Evol 75:91–102
Jenner RA (2006) Unburdening evo-devo: ancestral attractions, model organisms, and basal
baloney. Dev Genes Evol 216:385–394
Maduro MF (2006) Endomesoderm specification in Caenorhabditis elegans and other nematodes.
Bioessays 28:1010–1022
Milinkovitch MC, Tzika A (2007) Escaping the mouse trap: the selection of new Evo-Devo model
species. J Exp Zool B Mol Dev Evol 308:337–346
Minelli A, Baedke J (2014) Model organisms in evo-devo: promises and pitfalls of the comparative
approach. Hist Philos Life Sci 36:42–59
Minelli A, Negrisolo E, Fusco G (2007) Reconstructing animal phylogeny in the light of evolu-
tionary developmental biology. In: Hodkinson TR, Parnell JAN (eds) Reconstructing the tree of
life: taxonomy and systematics of species rich taxa. Taylor and Francis – CRC Press, Boca
Raton, pp 177–190
Minelli A, Chagas AJ, Edgecombe GD (2009) Saltational evolution of trunk segment number in
centipedes. Evol Dev 11:318–322
Moczek AP, Sultan S, Foster S, Ledon-Rettig C, Dworkin I, Nijhout HF, Abouheif E, Pfennig DW
(2011) The role of developmental plasticity in evolutionary innovation. Proc R Soc B
278:2705–2713
Pigliucci M (2010) Genotype ! phenotype mapping and the end of the ‘genes as blueprint’
metaphor. Philos Trans R Soc B 365:557–566
Richardson MK (2012) A phylotypic stage for all animals? Dev Cell 22:903–904
Smith KK (2001) Heterochrony revisited: the evolution of developmental sequences. Biol J Linn
Soc 73:169–186
Telford MJ, Budd GE (2003) The place of phylogeny and cladistics in Evo-Devo research. Int J Dev
Biol 47:479–490
Wagner GP, Zhang J (2011) The pleiotropic structure of the genotype-phenotype map: the
evolvability of complex organisms. Nat Rev Genet 12:204–213
Werneburg I, Sanchez-Villagra MR (2009) Timing of organogenesis support basal position of
turtles in the amniote tree of life. BMC Evol Biol 9:82
Wiens JJ, Bonett RM, Chippindale PT (2005) Ontogeny discombobulates phylogeny: paedomor-
phosis and higher-level salamander relationships. Syst Biol 54:91–110
Evo-devo of Language and Cognition

Sergio Balari and Guillermo Lorenzo

Contents
Synonyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
The Problem of Ontology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
The Problem of Computation(s) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
The Problem of Representation(s) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
The Problem of Homology/Novelty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
The Problem of Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Abstract
Historically, the task of disentangling the evolutionary origins of language has
been obscured by a number of difficulties that may be diagnosed as the problem of
ontology (what the evolved phenotype is), the problem of computation (what
kinds of cognitive processes subserve linguistic activity), the problem of repre-
sentation (what is the nature of the objects of computation), the problem of
homology/novelty (how language relates with animal cognition at large), and
the problem of selection (how language has been fixed as a species-typical
trait). While assuming that facets of these problems remain as recalcitrant as
ever, this chapter explains how the adoption of the developmental perspective
offers the promise of gaining a degree of explanatory accuracy hitherto unknown
in this field of specialization.

S. Balari (*)
Universitat Autònoma de Barcelona and Centre de Linguística Teòrica, Barcelona, Spain
e-mail: Sergi.Balari@uab.cat
G. Lorenzo
Guillermo Lorenzo: Universidad de Oviedo, Oviedo, Spain
e-mail: glorenzo@uniovi.es

# Springer International Publishing AG 2016 1

L. Nuño de la Rosa, G.B. Müller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_43-1
2 S. Balari and G. Lorenzo

Keywords
Evolution of cognition • Evolutionary linguistics • Computational theory of mind

Synonyms

Computational theory of mind; Evolution of cognition; Evolutionary linguistics

Introduction

The development of evolutionary ideas of mind and language started well before
Darwin, with pre-Lamarckian thinkers and Lamarck himself, who had a strong
influence on Darwin’s own account of the evolution of human mental capacities.
They shared the conviction that the mental capacities of human and nonhuman
animals could be given a naturalistic explanation and that this explanation had a
close connection with another natural phenomenon: the transformation of species.
Within this general framework, however, they faced a paradox that emerged from the
clash between their transformist ideas and their commitment with Locke’s and
Hume’s epistemology.
At least since Aristotle, causal explanations of behavior are based on the assump-
tion that a specific behavior is the product of some proximal, i.e., internal or mental,
cause; mental causes may themselves be caused by some distal (external) cause, but
this need not be necessarily so: the mind has the power to generate causes without
the participation of external stimuli. For any account of cognition along these lines to
work, two basic ingredients have customarily been assumed to be essential: nervous
systems possess (i) some means to represent their environment and (ii) the ability to
deal with manipulative processes in order to produce new representations and,
eventually, cause some behavior. The Lockean/Humean tradition was no exception
to this: ideas represent the world and these are connected through a process of
association. The problem for these earlier evolutionists was that, in line with the
empiricist epistemology they assumed, ideas could only be acquired through the
senses, by direct interaction with the environment. This was difficult to reconcile
with the observation that all animals were capable of producing more or less
complex behaviors immediately at the time of birth, which seemingly supported
the idea of innate instincts rather than that of acquired habits. Erasmus Darwin’s and
Pierre-Jean Cabanis’ way out of this conundrum was quite an ingenious one, as they
appealed to ontogenetic processes to explain the persistence of certain behavioral
traits in organisms of the same species. Such early-acquired sensations and habits,
most of them through the use of specific anatomical structures of each organism,
would later become the basis of new or modified behaviors and anatomical structures
in response to the changing conditions of the environment and could later be
inherited by the organism’s progeny.
Evo-devo of Language and Cognition 3

Darwin’s original account was not too different from the ones developed by his
grandfather Erasmus or by Lamarck. To be sure, Darwin bought wholesale Hume’s
theory of ideas and their associations, which made its way virtually unchanged in the
Descent, with the exception of Darwin’s willingness to accept the existence of innate
inherited social habits (Darwin 1879). The Descent’s first chapter is devoted to
homology, and Darwin’s strategy becomes immediately clear as soon as one reaches
▶ Chaps. 3 and ▶ 4. There, Darwin puts forward a detailed application of the
comparative method but centered on the “mental powers of man and lower animals,”
from basic instincts to moral senses, through attention, memory, tool use, language,
and consciousness, among others. Recall that Darwin redefined Richard Owen’s
original notion of homology by introducing a historical dimension that was never
present in Owen’s homology. In so doing, Darwin simply accommodated homology
to his idea of evolution seen as descent with modification in such a way that any two
traits would be classified as homologous to the extent that it could be shown that they
were modified variants of the same trait in a common ancestor; hence the importance
of innate inherited social habits in Darwin’s model, as they constitute the bedrock on
which a complete evolutionary story of the gradual modification of basic mental/
behavioral traits is constructed, linking basic social habits with the most sophisti-
cated human capacities.
One need not go into the details to identify in Darwin’s account two critical
problems shared with contemporary evolutionary accounts of cognition: the problem
of homology/novelty and the problem of ontology. Both are somehow connected,
since both have to do with the phenotypes one chooses to focus on when investi-
gating the evolution of some cognitive trait. The criteria one applies to determine
whether two phenotypic traits are homologous or not are crucial: should one pay
attention only to form, only to function, or both? The question becomes particularly
vexing when dealing with cognitive phenotypes: must one focus on the functional,
externally observable features of some cognitive trait in order to determine exactly
its phenotypic characteristics, or must one just study its neurophysiological under-
pinnings independently of the observable behaviors/functions these may cause?
Particularly illuminating to appreciate the subtle questions that the problems of
homology/novelty and of ontology pose for any naturalistic account of cognition
is Darwin’s brief treatment of the origins of language in Chap. ▶ 3 of the Descent
(Darwin 1879: 106–114). Right from the outset, Darwin expresses his conviction
that there is a strict functional continuity between the vocal and communicative
behaviors of animals and human speech and that the apparent gap separating humans
from other animals can be explained by the “high development of [the] mental
powers” of the former (Darwin 1879: 108–109). Darwin’s is thus a paradigmatic
example of the typical evolutionary explanation (not uncommon still today) where
functional/behavioral continuity is favored over homological analyses based only on
structural considerations, with the consequence of turning the linguistic phenotype
into a heterogeneous admixture of features, going from brain structures to grammat-
ical properties like agreement, through vocalization organs and multifarious com-
municative behaviors, like gestures and facial expressions, for example.
4 S. Balari and G. Lorenzo

Darwin strived to accommodate his account of the evolution of mind and

behavior within the strict limits imposed by his gradualist vision where the main
causal force was natural selection. But he was forced to make some concessions to
that view, since he found difficulties in explaining the origins of certain traits by
solely appealing to natural selection. Actually, the complementary mechanisms of
community and sexual selection were Darwin’s reaction to a number of objections
raised by some of his closer allies in many other matters—especially Charles Lyell,
Alfred Russel Wallace, and Herbert Spencer—on the power of natural selection as
the adequate mechanism to explain the origins of humans’ higher cognitive capac-
ities. With the obvious differences in emphasis and preference for one or another
alternative explanation, one could fairly contend that the debate on the problem of
selection is still alive today.
In the following sections, the different problems identified in this historical
introduction will be presented, adopting a contemporary stance and suggesting
some possible solutions offered by an evo-devo perspective.

The Problem of Ontology

The task of disentangling the evolutionary origins of language suffers from the lack
of a consensual view about what the evolved linguistic phenotype is supposed to
be. In a nutshell, theoretical positions differ along the following two main
coordinates:

1. Is language an external, socially shared code of sorts, which somehow gets

accommodated within an a priori uncompromised neural substrate in the early
experience of children? Or is language an internal, organ-like component of the
human brain, ready to make sense of certain salient environmental stimuli that
simply shape it into a language-particular condition?
2. Is language a self-contained component of the human brain, showing the signa-
ture of a well-defined form-function unit relatively to other similar brain special-
izations (face recognition, motor planning, early vision, and so on)? Or is
language a composite of different brain specializations, jointly laying down an
emergent functionality (as, say, vision), but which are, individually taken,
unspecific as for their linguistic dedication?

A brief survey of the theoretical elaboration of the concept of “language” reveals

that, as for coordinate (1), a strong shift took place in the mid-twentieth century from
externalist views of language to a predominant internalist stance (Chomsky 1986).
As for coordinate (2), a similar shift is presently on the way toward a composite or
mosaic conception of language when conceived of from such an internalist perspec-
tive (Boeckx 2012), which demotes theses like strong domain specificity or encap-
sulation that were once part and parcel of Chomsky’s internalism. In any event,
evolutionary linguists do not consensually adhere to any of the cardinal positions
Evo-devo of Language and Cognition 5

thus far commented, and many middle-ground positions also exist that complicate
the picture.
A consequence of the situation described above is that evolutionary linguistics
seems doomed to remain anchored to dualist stances that run against the aim of
releasing the evolutionary explanation of language from explanatory recipes with
assorted exceptionalist flavors. Not surprisingly, the field continues to be receptive to
a strong “culture/biology” divide that keeps apart the behavioral/psychological
dimension of language and its biologically proper underpinnings, as if they were
different (somehow connected) realms for which a more integrative ontology
appears to be a mistakenly reductionist project. Besides, in the aftermath of the
influential Hauser et al. (2002) paper, a growing acceptance appears to exist of a
new-wave dualism that keeps apart what is deemed properly biolinguistic (i.e.,
language specific) within the human organism and what is just biological. These
kinds of questions relate to issues that have been the target of recent evo-devo-
oriented philosophical analyses, from which evolutionary linguistics may draw
inspiration.
One obvious point of reference for linguists may be the concept of “developmen-
tal hybrid” currently being applied to accommodate socially shared and traditionally
transmitted practices to the evo-devo agenda (Caporael et al. 2014). The underlying
idea is that organisms are capable of engaging in interactive dynamics with specific
components of the environment that transform their internal constitution in ways that
reinforce the reiteration of such dynamics and pave the way to further, eventually
broader forms of environment-organism interactions. In as much as it makes sense to
say that the organism attunes its own properties in the process to the corresponding
properties of the environment, it also makes sense to conclude that what develops is
a hybrid entity. Hybrids so constructed may act as developmental scaffolds to further
complexifications of such entities. Moreover, the link that obtains between environ-
ment and organism through these hybridization processes favors their endurance and
intergenerational transmissibility, which means that the entities of concern are
exposed to standard evolutionary effects.
Another obvious point of reference for linguistics is the idea of “developmental
modularity,” which looks particularly apt to accommodate a mosaic view on lan-
guage without the strictures of a Fodorian-style view on modularity. Modularity at
the level of the generative underpinnings of a given organic structure does not
necessarily compromise the fate of its component parts in the direction of strict
domain specificity, which is the picture that fits better with a growing body of
research on shared neural patterns of activity in linguistic and nonlinguistic tasks
and on comorbidity of linguistic and nonlinguistic deficits. Moreover, developmen-
tal modularity is an idea particularly congenial with the characterization of language
as a hybrid entity along the lines of the previous paragraph, for concurrently
developing aspects of the hybrid on the way may straightforwardly exert scaffolding
influences on each other.
Some well-documented observations appear to support the idea that the triad
“hybrid-scaffolding-modularity” is particularly fit to offer new ontological founda-
tions to the language faculty. For example, newborns acquire their language-
6 S. Balari and G. Lorenzo

particular phonological competence from general sensory-motor, computational and

statistical abilities. This milestone, together with aspects of social intelligence,
grounds, first, the acquisition of sound-meaning associations and, next, the explo-
sion of syntax (Kuhl 2004). On the other hand, the earliest, most abrupt, and less
resilient maturational effects on language acquisition are actually observed in the
phonological component, which according to the above picture acts as the ultimate
foundation for all the rest (Meisel 2013).

The Problem of Computation(s)

An element of consensus within the cognitive science community is that there is

some causal story to be told about cognition and behavior. The central element of
this causal story is that both cognition and behavior are explained by the action of
some internal (mental) operations that may themselves be caused by some external
stimuli (e.g., through perception), although this latter step is not necessary or
sufficient for a full account of cognitive processes. Of course, the appeal to opera-
tions hardly constitutes an explanation, as what one really needs is a precise
characterization of the operations and, also, of the entities these operations apply
to. Moreover, in the naturalistic context of contemporary cognitive science, one
expects to have an account of operations and entities that agrees with the workings of
the brain. As for the latter, there is a wide consensus that mental operations apply to
ideas or, to use the contemporary terminology, representations. As for operations,
things are much less clear. With the sole exception of Thomas Hobbes and possibly
Leibniz too, who equated mental operations with some form of calculation, only in
the twentieth century will take form the idea that cognitive processes are computa-
tional processes or, in other words, that nervous systems are natural systems of
computation. But what is a natural system of computation?
Gualtiero Piccinini has warned us against loose characterizations of computa-
tional systems—verging into vacuous pancomputationalism—and has enumerated a
number of constraints any system must obey in order for it to be classified as
computational (Piccinini 2015). Along with the constraints of being a functionally
organized input-output system—a property that brains share with other organ
systems––perhaps the most salient constraint Piccinini identifies as the landmark
of computation is the ability to process a special class of physical entities according
to rules that are sensitive to certain formal properties of the said entities. Piccinini
refers to the physical entities that are the objects of computation with the term
vehicles. This terminological move is justified on the grounds that the objects of
computation need not necessarily be discrete entities—like the so-called symbols of
theoretical computer science—or carriers of semantic content—like the traditional
conception of symbol implies. It is clear that brains are natural computational
systems at least in this generic sense: neural processes manipulate voltage changes
in the dendrites, neuronal spikes, neurotransmitters, and hormones, which all qualify
as bona fide vehicles. The question is whether brains are computational systems in
Evo-devo of Language and Cognition 7

other, more specific senses, hopefully coincident with the assumptions of computa-
tional cognitive science.
Computational cognitive science’s version of cognition is epitomized by the
so-called computational theory of mind (CTM), whose locus classicus is the lan-
guage of thought (LOT) hypothesis. In a nutshell, what the CTM+LOT hypothesis
amounts to is that mental computations are performed over data structures made up
of discrete symbols, which moreover have semantic properties. Note that this
conception of computation is much more restrictive than the generic one presented
above in requiring that vehicles be both discrete and carriers of semantic content. It
does in fact correspond to a specific subclass of generic computation that Piccinini
defines as semantic digital computation. Semantic digital computation has tradition-
ally been the most favored one among cognitive scientists because it offers a
framework within which higher cognitive processes, with language being the para-
digmatic case, are easily characterized as computations over complex symbolic data
structures with a syntax and a semantics.
The problem of computation can be formulated in the following terms: is it really
the case that the brain, qua generic system of computation, is also a digital system of
computation? Piccinini and Behar’s (2013:477) conclusion is a rather bleak one:
“current evidence indicates that typical neural signals, such as spike trains, are
graded like continuous signals but are constituted by discrete functional elements
(spikes). Therefore, typical neural signals are neither continuous signals nor strings
of digits: neural computation is sui generis.” The corollary of this conclusion is that a
new kind of cognitive neuroscience is needed. This new perspective is, indeed,
emerging, and its more recent findings suggest that such a pessimistic view might
well be premature.
Until fairly recently, the primary concern of neuroscience was functional neuro-
anatomy, essentially extending the work of Korbinian Brodmann by attaching some
function or other to the cortical areas he described attending mostly to histological
principles. This kind of research, relevant as it is especially on its comparative
aspect, is relatively uninformative when it comes to brain dynamics. The elaboration
of detailed functional brain maps has nevertheless made clear that even more
important than brain regions and their functional specializations are the networks
different areas conform when jointly engaged in some cognitive process—hence the
terms connectomics to refer to the structural description of the brain seen as a
network of nodes and edges and functional connectomics when this description is
enhanced with information concerning brain activity. The inevitable next step in this
chain of events has been the initiation of research on the dynamic aspect of networks:
how and what regions are connected and what signals they convey and how they are
acted upon. This last breakthrough in neuroscience research has been crucial for two
reasons. Firstly, the level of analysis of brain dynamics (or the dynome, as it has
come to be named) defines an intermediate level that smoothly allows for the linking
of the macrolevel of brain regions with the microlevels defined by the neuron and
below (Kopell et al. 2014). Secondly, recent research at the level of brain dynamics
appears to support precisely what Piccinini and Behar (2013) denied in their analysis
of neural oscillations (spikes and spike trains), namely, that neural computation
8 S. Balari and G. Lorenzo

might be of the digital kind after all, with these discrete assembly sequences playing
a central role in different cognitive processes, as well as in language processing. A
relevant datum in this connection is the high preservation of brain oscillations, at
least across all mammalian genera (Buzsáki et al. 2013). So, if neural oscillations are
the key to understand neural computation, a central question is how differences in
cognitive abilities are to be accounted for within this scenario of high preservation.
Clearly, research in the development of cell types, connections, and neural rhythms,
along with their correlates both at the molecular and cognitive levels, should occupy
a central role in the understanding of the evolution of language and cognition.

The Problem of Representation(s)

As stated in the previous section, computations are input-output processes capable of

manipulating certain entities according to a specific set of rules; in other words,
computations can be characterized as functions. Few are those who doubt that
minds/brains compute and that the inputs and outputs of cognitive functions are
what generically may be identified as representations. Epistemological disputes may
arise as to whether representations are innate or learned, or concerning their nature,
their form, and their mode of representation, but ontology appears not to be an issue
in this case: there are representations.
Symbols (in an extended sense of the term capable of also applying to neuro-
physiological states) are the most plausible candidates for being the kind of repre-
sentation format needed by a computational theory of cognition. The problem with
symbols is to explain how they represent and how they acquire representational
content. The usual story with symbols is that they represent by convention, but
conventions are public, social contracts, and mental representations are not public.
What is good for cultural symbols is not necessary good for neurobiological ones.
Locke—neurobiological considerations aside, of course—saw this and tried to
develop a causal theory of representation, the idea that symbols represent through
covariance. Representation by covariance is a cover term for a whole family of
theories that are nonetheless based on the idea that the represented object and its
representation stand in a causal relation (covariance) in the sense that any presenta-
tion of the object causes a corresponding token representation in the mind/brain. In
other words, content is the pair formed by a vehicle and its extension, i.e., the set of
things in the world that would cause its tokening by the brain. Alternatively, one may
assume that representation is not a relation, but just a property of the vehicles of
content, where the latter is seen as intrinsic or perhaps acquired through the vehicles’
roles during cognitive processing—a contention that, according to Egan (2014), runs
into serious difficulties. Crucially, however, neither alternative is necessarily com-
mitted to a notion of representational content that is in turn committed to the
conceptual apparatus typically associated to what is generally associated with the
notion of intentionality. To be sure, there is a growing consensus that the latter is not
a proper part of a naturalistic theory of content (Fodor and Pylyshyn 2015).
Evo-devo of Language and Cognition 9

Now, so far it may seem that the problem of representation(s) has little to do with
either evolution or development, but nothing could be further from the truth. Indeed,
taking language as a paradigmatic case, it is reasonable to suppose that representa-
tions are also subject to developmental processes, which immediately suggests an
analysis in terms of the notion of developmental hybrid like the one sketched in
section “Introduction.” This would clearly favor a conception of representation
along the lines of the covariance model set forth by Fodor and Pylyshyn (2015),
but also would open new avenues of research on the development and evolution of
representations along the lines of models of cultural evolution like the one outlined
in Charbonneau (2015) or those developed within the paradigm of iterated learning
(Kirby 2013).

The Problem of Homology/Novelty

There exists a relative consensus among linguists around the idea that language is a
human-specific endowment with no known equivalents in other nonhuman organ-
isms. Yet experts disagree about the extent to which language is unique, for many
believe that it is a thoroughly innovative capacity (Pinker and Jackendoff 2005) and
others claim that it may be decomposed into shared and unique component parts
(Hauser et al. 2002). The debate on language uniqueness, however, is one in which
discussions are not particularly clarifying, for contenders base their positions on
different implicit, unclear, and intuitive ideas about what makes a particular organic
structure an innovative one. This is therefore an aspect of evolutionary linguistics in
which the injection of concepts and criteria from evo-devo has been claimed to be
particularly urgent (Balari and Lorenzo 2015). The diversity of current evo-devo
approaches becomes a real hindrace in this particular subject matter, however; as
pointed out by Benítez-Burraco and Longa (2010), such diversity often favors the
more relaxed gene-centric versions of evo-devo over other, more promising ones.
For one, Chomsky’s (2010) forays into evo-devo, for example, continue to be
anchored in his classical gene-centric view on universal grammar, which has uncrit-
ically led to a strong species-specificity bias, as well as to exceptionalist theories of
language evolution.
Again, the idea of “developmental modularity” may serve as an obvious bridge, at
least in those aspects of the language mosaic for which reliable developmental
information exists at a molecular level of analysis. Key questions like the following
will become more and more accessible with the body of information that will
predictably accumulate in this area of expertise in the near future:

1. Attending to independently established developmental criteria, can we homolo-

gize different components of human language with different candidate aspects of
the cognitive makeup of other nonhuman organisms? Is the homologation
exhaustive, or residues remain of language that resist the test?
10 S. Balari and G. Lorenzo

2. Were language to be exhaustively put into correspondence with other nonhuman

homologues, how could we still make sense of the fact that no other organisms
learn, use, and can take advantage of languages in the way humans do?

One may advance toward a preliminarily take on these questions by briefly

examining the case of the aspect of language that most students have pinpointed
as the locus of the radical innovativeness of this allegedly human-unique capacity,
namely, the computational procedure capable of recursively putting pieces together
and thus of endowing languages with their never-ending expressive potential
(Hauser et al. 2002). The standard specification of this procedure is that it works
pair-wise, so it takes two units X and Y (two lexical items or two previously
constructed syntactic objects), and it outputs a single unified syntactic unit {X, Y}.
Moreover, the operation creates structure, so at each step, it must also somehow
signal an inner asymmetry within each newly created unit, {X Z, {X X, Y}}, and so
on. There exists a broad consensus around the idea that this operation is human
specific (Berwick and Chomsky 2016). But note that while this may be true if the
operation is focused with an extremely fine-grained lens, it also makes sense
searching for affinities with some nonhuman practices in order to test the idea that
a homological thread exists among them. For example, there exists a large body of
literature on birdsong that offers grounds to the idea that it benefits from the “same”
pairing procedure as language, “save” for the lack of the principle/operation respon-
sible of creating asymmetries. As a consequence, sequences are generated that
display a flatter character relatively to human phrases (Berwick et al. 2011). A
large body of literature seems to give support to these “abstract” parallels from the
point of view of developmental genetics (Pfenning et al. 2014). Are we authorized to
conclude that a bona fide homological relation exists between the corresponding
systems of computation?
In order to answer this question, we need (i) independently well-established
criteria to sanction homological relations and (ii) a good sample of the kinds of
organic materials to which said criteria make reference. These are hard issues that
entail some prior nontrivial questions: do systems of computation qualify as bona
fide organs? And, ultimately, what does qualify as a bona fide organ? Any promis-
sory route leading to a complete answer to this chained collection of questions needs
to start at a solid developmental theory of organ identity, like the one put forward in
Wagner (2014). Its main contention is that “character identity networks” (ChINs)
exist at a molecular level that may offer the basis for organ identity discriminations.
Roughly speaking, ChINs are distinctive patterns of regulatory genes (and related
products) and gene interactions that mediate between signaling information and
realizer genes, ultimately responsible of the attainment of variants of the same
organ thus circumscribed. Evidence of a putative ChIN that appears to justify the
idea that birdsong and human language (at least) share the same organ of computa-
tion may be based on the existence of shared inductive signals (retinoic signaling
pathway), transcription factors (FOXP2), and target genes (SLIT1, NEUROD6, etc.)
(Balari and Lorenzo 2015). Complementarily to this conclusion, one may also
proceed to ask what is it that nevertheless makes language a unique manifestation
Evo-devo of Language and Cognition 11

of that organ (or, for that matter, what is it that makes birdsong a unique ability). The
answer will require a better understanding of the organic basis of the “structured/flat”
character of the corresponding computed strings, as well as of the pattern of interface
relations that systems of computation establish in different organic contexts (Balari
and Lorenzo 2013).

The Problem of Selection

The application of the Darwinian paradigm to the case of language has always been a
contentious issue, mostly due to Chomsky’s anti-selectionist arguments. In outline,
these arguments boil down to the following claims: (1) most uses of language are
organism internal, not environmentally oriented; (2) it seems to be impossible to
identify a particular kind of adaptive pressure to which language could be said to
adjust in the expected lock-and-key manner; (3) many linguistic properties are
actually misadjusted in relation to any particular well-defined purpose (to begin
with, the unbounded expressiveness of languages); and (4) many well-formed
expressions are not usable (e.g., for being too long), and many ill-formed expres-
sions could be used without taxing mutual comprehension (Chomsky 1968).
Within this traditional context, many proposals have however been made about
the role of selection in the evolutionary shaping of language, which differ on its
relevance in the launching of language as an innovative endowment of humans, on
the complexification and fine-tuning of its component parts, and on the kinds of
environmental influences capable of exerting the corresponding selective pressures
(Bickerton 2013). If one takes the vantage point of a full-fledged theory on the origin
and evolution of organs like Wagner’s (2014), the premise may be adopted that the
role of selection is negligible in the initiation of brand new ones but inescapable in
the subsequent processes of radiation of variants. But if one also accepts that the
component parts of language are exhaustively homologizable with other nonhuman
cognitive organs, then the question still arises of how selection has acted in imprint-
ing on language its distinctive characteristics. In this respect, views differ on whether
components added at different stages were specifically selected, fixed for the selec-
tive value of language as a whole, or not selected at all.
The idea of “developmental modularity,” in inviting to break language into
component pieces on developmental grounds, certainly gives some justification to
the possibility that selective effects operate on language in a piecemeal way. But
also, and maybe more importantly, it justifies the necessity of taking apart different
senses or different levels at which such effects may act: first of all, parts must be
proved to be mutually compatible, thus selected in an organismic internal sense, and
then, and only then, they may confront the pressures of a demanding environment,
eventually leading to their fine-tuning and fixation within a population. An important
corollary of this last statement is that aspects of the linguistic hybrid might have
evolved just for their role in facilitating the development of other parts of the hybrid,
i.e., for their being useful for development per se (Minelli 2003: 14).
12 S. Balari and G. Lorenzo

Besides, modules may be intertwined in ways that compromise the developmen-

tal fate of each other, so why and how parts evolve may be due not to pressures
directly exercised by the environment, but to hitchhiking effects of sorts of an
endogenous character. This is how, for example, Balari and Lorenzo (2013: Ch. 6)
try to make sense of the interconnectedness of the architectural components of
languages, the multimodal character of lexical items, and the complexity level of
linguistic computations at once, as due to heterochronic effects after the acceleration
of cortical growth in human evolution. A virtue of these kinds of suggestions is that
they may offer grounds to the idea that language is a “variational modality” of an
ancestral overarching organ (Balari and Lorenzo 2015), in which a set of peculiar
properties are brought together that are not accessible from the developmental
underpinnings of other homologous modalities (Wagner 2014).
Obviously enough, conjectures like the ones put forth to in the previous para-
graphs do not invalidate the putative role of natural selection in particular aspects of
the evolutionary shaping of language, difficult as it may be to establish the partic-
ularities of such a process in the realm of cognition. The idea bumps into troubles, for
example, when one is confronted with Chomsky’s suggestion that the computational
system of languages is a fixed, uniform component of human brains, contrary to the
expectations of the theory of natural selection (Chomsky 2001). Others have argued,
however, that more variation than expected may be found in this aspect of language
(Hurford 2012), a fact that encourages exploring the application of the selectionist
paradigm even to such a refractory arena.

Acknowledgments This work has been partially supported by the Generalitat de Catalunya
through grant 2014-SGR-1013 to the Centre de Lingüística Teòrica of the Universitat Autònoma
de Barcelona (SB). We wish to thank Víctor M. Longa for helpful comments to an earlier draft of
this paper. Any remaining errors are our own.

Cross-References

▶ Developmental Homology
▶ Evo-devo and Cognitive Science
▶ Evo-devo and Culture
▶ Heterochrony
▶ Modularity
▶ Novelty and Innovation

References
Balari S, Lorenzo G (2013) Computational phenotypes. Towards an evolutionary developmental
biolinguistics. Oxford University Press, Oxford
Balari S, Lorenzo G (2015) It is an organ, it is new, but it is not a new organ. Conceptualizing
language from a homological perspective. Front Ecol Evol 3:58. doi:10.3389/fevo.2015.00058
Evo-devo of Language and Cognition 13

Benitez-Burraco A, Longa VML (2010) Evo-devo—of course, but which one? Some comments on
Chomsky’s analogies between the biolinguistic approach an evo-devo. Biolinguistics
4:308–323
Berwick RC, Chomsky N (2016) Why only us language and evolution. The MIT Press, Cambridge,
MA
Berwick RC, Okanoya K, Bechers GJL, Bolhuis JJ (2011) Songs to syntax: the linguistics of
birdsong. Trends Cogn Sci 15:113–121
Bickerton D (2013) Language and natural selection. In: Boeckx C, Grohmann KK (eds) The
cambridge handbook of biolinguistics. Cambridge University Press, Cambridge, pp 478–488
Boeckx C (2012) The I-language mosaic. In: Boeckx C, Horno-Chéliz MC, Mendívil-Giró JL (eds)
Language, from a biological point of view. Cambridge Scholars Publishing, Newcastle upon
Tyne, pp 23–51
Buzsáki G, Logothetis N, Singer W (2013) Scaling brain size, keeping timing: evolutionary
preservation of brain rhythms. Neuron 80:751–764
Caporael LR, Griessemer JR, Wimsatt WC (eds) (2014) Developing scaffolds in evolution, culture,
and cognition. The MIT Press, Cambridge, MA
Charbonneau M (2015) Mapping complex social transmission: technical constraints on the evolu-
tion of cultures. Biol Philos 30:527–546
Chosmky N (1968) Language and mind. Harcourt Brace Jovanovich, New York
Chomsky N (1986) Knowledge of language. Its nature, origin and use. Praeger, New York
Chomsky N (2001) Derivation by phase. In: Kenstowicz K (ed) Ken Hale: a life in language. The
MIT Press, Cambridge, MA, pp 1–52
Chomsky N (2010) Some simple evo devo theses: how true might they be for language. In: Larson
RK, Déprez V, Yamakido H (eds) The evolution of language. Biolinguistic perspectives.
Cambridge University Press, Cambridge, MA, pp 45–62
Darwin C (1879) The Descent of man, and selection in relation to sex, 2nd edn. John Murray,
London
Egan F (2014) How to think about mental content. Philos Stud 170(1):115–135
Fodor JA, Pylyshyn ZW (2015) Minds without meanings. The MIT Press, Cambridge, MA
Hauser MD, Chomsky N, Fitch WT (2002) The faculty of language: what is it? Who has it? How
did it evolve? Science 298:1569–1579
Hurford JR (2012) The origins of grammar. Language in the light of evolution. Oxford University
Press, Oxford
Kirby S (2013) Language, culture, and computation: an adaptive systems approach to biolinguistics.
In: Boeckx C, Grohmann KK (eds) The Cambridge handbook of biolinguistics. Cambridge
University Press, Cambridge, MA, pp 460–477
Kopell NJ, Gritton HJ, Whittington MA, Kramer MA (2014) Beyond the connectome: the dynome.
Neuron 83:1319–1328
Kuhl PK (2004) Early language acquisition: cracking the speech code. Nat Rev Neurosci 5:831–843
Meisel JM (2013) Sensitive phases in successive language acquisition: the critical period hypothesis
revisited. In: Boeckx C, Grohmann KK (eds) The Cambridge handbook of biolinguistics.
Cambridge University Press, Cambridge, MA, pp 69–85
Minelli A (2003) The development of animal form. Ontogeny, morphology, and evolution. Cam-
bridge University Press, Cambridge, MA
Pfenning AR, Hara E, Whitney O, Rivas MV, Wang R, Roulhac PL, Howard JT, Wirthlin M, Lovell
PV, Ganapathy G, Mouncastle J, Moseley MT, Thompson JW, Soderblom EJ, Iriki A, Kato M,
Gilbert MT, Zhang G, Bakken T, Bongaarts A, Bernard A, Lein E, Mello CV, Hartemink AJ,
Jarvis ED (2014) Convergent transcriptional specializations in the brains of humans and song-
learning birds. Science 346:1333–1346
Piccinini G (2015) Physical computation A mechanistic account. Oxford University Press, Oxford
14 S. Balari and G. Lorenzo

Piccinini G, Bahar S (2013) Neural computation and the computational theory of cognition. Cogn
Sci 34:453–488
Pinker S, Jackendoff R (2005) The faculty of language: What’s special about it. Cognition
95:201–236
Wagner GP (2014) Homology, genes, and evolutionary innovation. Princeton University Press,
Princeton and Oxford
Evo-devo and Cognitive Science

Annemie Ploeger and Frietson Galis

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Developmental Cognitive Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Evolutionary Cognitive Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Evo-devo and Cognitive Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Evolutionary Developmental Cognitive Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Developmental Systems Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Genes X Environment Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Epigenetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Evolutionary Developmental Comparative Cognitive Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Evolutionary Developmental Cognitive Neuroscience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Evolutionary Developmental Biology and Cognitive Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Abstract
Evo-devo is an approach that integrates knowledge on evolution and develop-
ment. Cognitive science is a research field that tries to unravel the functioning of
the mind and the underlying processes. In this chapter, the main subfields within
cognitive science that have contributed to a better understanding of the evolution
and development of the mind are discussed. Highlighted are the subfields of
evolutionary cognitive science, developmental systems theory, genes  environ-
ment interaction research, epigenetics, comparative cognitive science, and

A. Ploeger (*)
Department of Psychology, University of Amsterdam, Amsterdam, The Netherlands
e-mail: a.ploeger@uva.nl
F. Galis (*)
Naturalis Biodiversity Center, Leiden, The Netherlands
e-mail: frietson.galis@naturalis.nl

# Springer International Publishing Switzerland 2016 1

L. Nuño de la Rosa, G.B. Müller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_44-1
2 A. Ploeger and F. Galis

cognitive neuroscience. Finally, the question what cognitive scientist can learn
from research in evolutionary developmental biology is addressed. Many
evo-devo biologists study morphogenesis, which is relevant for cognitive devel-
opment, but it is not always straightforward how to apply their knowledge to
cognitive science research. Interdisciplinary research is strongly recommended,
so scholars from different fields such as morphology, genetics, neuroscience,
primatology, and psychology can learn from each other and contribute to the
unraveling of the working of the mind.

Keywords
Evo-devo • Cognitive science • GE interactions • Epigenetics • Developmental
systems

Introduction

Cognitive science is a broad, interdisciplinary research field that tries to unravel

the functioning of the mind and the underlying processes. It includes the study of
perception, motor control, attention, consciousness, learning, memory, represen-
tation of knowledge, language, problem-solving, creativity, decision-making, rea-
soning, and intelligence (e.g., Newell 1990). The emergence of cognitive science
has been called the cognitive revolution, which started in 1959. Linguist Noam
Chomsky argued that language acquisition cannot be explained by simple
stimulus–response associations proposed by behaviorism, which was the dominant
paradigm in psychology at that time. Behaviorists, amongst others their famous
proponents Ivan Pavlov, Edward Thorndike, John B. Watson, and B. F. Skinner,
argued that mental processes cannot be scientifically studied and that psycholo-
gists should restrict their research to observable behavior. Chomsky’s discussion
preluded the rise of cognitive science.
Cognitive science has become a successful field, partly because of the develop-
ment of new research tools, such as brain imaging techniques and computer simu-
lations. Cognitive neuroscience is a rapidly growing subfield in which concepts such
as attention and memory are linked to specific brain areas and neural activity.
Artificial intelligence is another subfield that uses insights from cognitive research
to create computer models of the mind. In huge programs such as the Human Brain
Project, headed by physiologist Henry Markram (see Markram et al. 2011), knowl-
edge on cognitive architectures, brain simulations, high-performance computing,
neuroinformatics, neurorobotics, and other disciplines is combined with the aim to
simulate the whole human brain. Results from this kind of projects show the
fruitfulness of cognitive science and also how important technological progress
has been.
The first aim of this chapter is to provide an introduction of the history of the
attempts of cognitive scientists to integrate their work with developmental and
evolutionary approaches. After the cognitive revolution, the study of cognitive
Evo-devo and Cognitive Science 3

development started to grow. Jean Piaget, with his stage theory of cognitive devel-
opment, became the major proponent of this field. Later, in the 1990s, the study of
the evolution of cognition arose under the flag of evolutionary psychology. Major
proponents of this field in the 1990s were cognitive psychologist Leda Cosmides,
anthropologist John Tooby, linguist Steven Pinker, and social psychologist David
Buss. In the first years of the new millennium, the first attempts were made to
integrate developmental and evolutionary approaches to cognitive science by devel-
opmental psychologists David Bjorklund and David Geary, among others. Only
lately, attempts have been made to integrate evo-devo biology and cognitive science.
The second aim is to show the importance of evo-devo research for cognitive
science. Evo-devo is an approach that integrates knowledge on evolution and
development (see chapter on the “▶ History of Evo-devo”). Evolutionary biologists
study evolutionary change of organisms over generations; developmental biologists
study the development of organisms within a single lifetime. Evo-devo researchers
try to unravel the interaction between these two processes – evolution and develop-
ment, to obtain a fuller understanding of each of these processes.
Before the contribution of evo-devo research to cognitive science is described,
first the contributions of developmental cognitive science and evolutionary cognitive
science are explained separately.

Developmental Cognitive Science

Developmental cognitive science is the study of cognitive development, from

prenatal development to cognitive aging. The study of prenatal cognitive develop-
ment is limited, because it is hard to study the cognition of fetuses in the womb. Most
of this research is focused on the senses, especially the visual and auditory system.
Vision research with premature infants revealed that they can distinguish between
light and dark at 28 weeks after conception, and that they can distinguish different
patterns at 30 weeks. Considering the auditory system, fetuses start to react with
movement to acoustic stimulation between 23 and 25 weeks. It is also well known
that newborns remember sounds they heard during their last month in the womb. In
addition, newborns show a preference for their mother’s voice compared to a
stranger, and they can discriminate between their mother’s and a foreign language.
In general, developmental cognitive scientists study the same topics as cognitive
scientists, but they emphasize differences between children and adults by conducting
longitudinal or cross-sectional studies. An example of a topic that recently has
attracted a lot of attention is the development of executive functions. This is an
umbrella term for all processes necessary for cognitive control, such as inhibition,
task flexibility, planning, and working memory. Researchers generally agree that
normal or good development of executive functions is required to be able to deal
with daily-life problems in our complex society. Many common psychiatric disor-
ders, such as ADHD, autism, and anxiety, are associated with atypical development
of executive functions.
4 A. Ploeger and F. Galis

As was mentioned in the introduction, the theory on cognitive development

developed by Jean Piaget has been most influential. He argued that cognitive
development proceeds stagewise, with children acquiring knowledge by changing
their cognitive structures by the processes of assimilation and accommodation.
When children encounter new situations, they try to assimilate their experiences
into their existing cognitive structures. Only when they find out that the existing
structures fail to explain the new situation, they will build new cognitive structures
(a process called accommodation). New approaches that arose out of Piaget’s theory
are neuroconstructivism (e.g., Karmiloff-Smith 2006), which combines Piaget’s
theory with recent neuroscientific findings and dynamic systems theory (e.g., Thelen
and Smith 1994). Adherents of the latter approach argue that development can be
best described by differential equations, where development is modeled as a trajec-
tory through state space. Following this approach, children learn by discovering that
available patterns of knowledge are incomplete, leaving them in a state of disequi-
librium, after which a new equilibrium can be reached, when new patterns of
knowledge are formed. The process of breaking down old patterns and establishing
new ones occurs by means of phase transitions. This process is considered to be self-
organized, because there are no control parameters that govern it. This approach has
been specifically successful in explaining motor development, but also other areas of
cognitive development.

Evolutionary Cognitive Science

Evolutionary cognitive science is the study of the evolution of cognition, with the
expectation that knowledge about the evolution of cognition improves our under-
standing of the working of the human mind (Barkow et al. 1992). Evolutionary
cognitive scientists try to discover cognitive adaptations that have been under natural
selection or cognitive fitness indicators that were sexually selected. Well-studied
examples of cognitive adaptations are language, face recognition, color perception,
cheater detection, and spatial abilities that are related to hunter-gatherer skills (see
the chapter on “▶ Evo-devo of Language and Cognition”). Geoffrey Miller (2000)
has argued that many aspects of human cognition are sexually selected, such as art,
music, humor, and science. Empirical support for this hypothesis is provided by
studies that showed an association between these phenomena and measures of
fitness, e.g., health and symmetry, and levels of estradiol (in females) and semen
quality (in males). Other evolutionary cognitive scientists have used a comparative
approach. For example, psychologist and primatologist Michael Tomasello (2014)
has compared the cognition of human children and chimpanzees and concluded that
their physical cognition (e.g., knowledge about quantities) is similar, but that 2-year-
old humans already have a better developed social cognition (e.g., empathy) than
adult chimpanzees.
Evo-devo and Cognitive Science 5

Evo-devo and Cognitive Science

Evo-devo research in cognitive science can be carried out in many different ways.
Several subfields that combine an evolutionary and a developmental approach in the
study of cognition are outlined. First, the work of evolutionary developmental
cognitive scientists is discussed. Second, the work of developmental systems theo-
rists is discussed. Third, studies on the interaction between genes and environment
by cognitive scientists are discussed. Fourth, research on epigenetics in cognitive
science are discussed. Fifth, comparative evo-devo studies by cognitive scientists are
discussed. Sixth, the studies on cognition by evolutionary developmental neurosci-
entists are discussed. Finally, the work of evo-devo biologists extrapolating results
on evolutionary developmental mechanisms to the field of cognition are discussed.

Evolutionary Developmental Cognitive Science

Evolutionary developmental cognitive scientists are usually trained as developmen-

tal psychologists and try to understand the origin of the behavior and cognitive
abilities of children. Evo-devo psychologist David Bjorklund (2009) has studied the
adaptive role of cognitive immaturity. Children are often viewed as “unfinished
adults,” but Bjorklund showed that children’s failures on cognitive tests are some-
times adaptive. For example, young children often overestimate their cognitive
abilities. It was found that this overestimation was associated with better cognitive
performance at a later age, probably because overconfidence leads to more explor-
atory behavior. This suggests that overestimation is functional. Evo-devo psychol-
ogist David Geary (2005) has addressed educational psychology from an evo-devo
perspective. He argued that children learn some abilities naturally, such as language
and simple counting. Also children that are not stimulated by their parents or
formally educated will learn these. He called these primary abilities. Other abilities
are secondary, such as higher-order mathematics, which require formal instructions
to be learned. He argued that children will be better motivated to learn secondary
abilities when they are coupled to primary abilities.

Developmental Systems Theory

Developmental systems theorists argue that organisms do not only inherit the DNA
of their parents but a whole developmental system, which includes the environment
in its full spectrum – at the cellular, tissue, body, family, and ecological level (Oyama
et al. 2001). They argue that the scope of some evolutionary cognitive scientists is
too limited. For example, developmental psychologist Elizabeth Spelke has argued
that human babies are born with core knowledge about objects, social agents,
numbers, and geometry (Spelke and Kinzler 2007). Her conclusions are based on
empirical results that revealed that newborns, even on the first day after birth, show
different responses (e.g., different looking times) to different situations, even though
6 A. Ploeger and F. Galis

they have no prior experience with these situations. Spelke argues that, based on her
data, the inevitable conclusion is that some basic knowledge is innate. Developmen-
tal systems theorists tend not to agree with this conclusion. Nonetheless, they agree
that a functional perspective is necessary to get a better understanding of the mind,
but they highlight the importance of ecologically relevant conditions to study
cognition, in agreement with evolutionary ecologists.
Another illustration of the sometimes limited scope of evolutionary cognitive
scientists was put forward by developmental psychologist Annette Karmiloff-Smith
(2006). She criticized some of the arguments put forward by Steven Pinker in the
discussion on language evolution. Pinker argued that the gene FOXP2 is “the
language gene.” Research revealed that members of a family with severe language
problems showed mutations in FOXP2, suggesting that it is a crucial gene in
language development. Later research revealed that FOXP2 is not specifically
involved in language development but in multiple processes that are associated
with producing sequential movements. Karmiloff-Smith argued that the effects of
many genes are associated with general processes and that specific complex traits,
such as language, are the result of several developmental pathways, with no simple
relationship to a specific DNA sequence.
Some researchers have tried to bridge the gap between evolutionary cognitive
science and developmental systems theory. For example, David Bjorklund (2015)
has proposed the concept of an evolved probabilistic cognitive mechanism. He
argued that it is inevitable that natural selection has selected cognitive adaptations
to deal with problems related to survival and reproduction. However, these adapta-
tions will not develop when individuals are placed in an environment that is not
species typical (e.g., for humans an environment without spoken language). There-
fore, he argued that all evolved cognitive mechanisms are probabilistic, with their
development being dependent on the right environmental input.

Genes X Environment Interactions

The discussion on the interaction between the influence of genes and that of the
environment on the development of phenotypes goes back to the seventeenth century
when philosopher John Locke started the nature-versus-nurture debate. Nowadays
most psychologists and biologists agree that both nature and nurture are important,
with some researchers pointing to the importance of culture (e.g., Richerson and
Boyd 2005; see also the chapter on “▶ Evo-devo and Culture”), next to nature
(genes), and nurture (parenting). Recently, a wide array of new discoveries has
been made in studies that examine the interaction effects between specific genes
and specific environmental input (i.e., GE studies) on specific behavioral outcomes
(e.g., aggression, depression, etc.). The paradigm case is a longitudinal study
performed by Avshalom Caspi and colleagues (2002). They followed a large sample
of boys from birth to adulthood in order to study the development of antisocial
behavior. Data on individual differences at a polymorphism in the promoter of the
MAOA gene and maltreatment were collected. It was found that the interaction
Evo-devo and Cognitive Science 7

between maltreatment and having a genotype associated with low levels of MAOA
expression resulted in a high risk of developing antisocial behavior. This was the first
study that showed a significant GE interaction. Many other studies followed.
Interesting theoretical contributions were made by psychologists Bruce Ellis, Jay
Belsky, and colleagues. Ellis et al. (2011) reviewed several GE studies and
observed that some genotypes are associated with a vulnerability to develop psy-
chopathology in interaction with a negative environment (such as maltreatment), but
that the very same genotype is also associated with positive outcomes in interaction
with a positive environment (such as the absence of maltreatment). They prefer to
call these “plasticity genes,” rather than “vulnerability genes.” An appealing com-
parison with orchids and dandelions has been made. Orchids are beautiful flowers
when they are taken good care off, but nothing but a boring empty stem remains
when they are maltreated. In contrast, dandelions are arguably not that beautiful, but
they can grow everywhere, even at a roadside or a dumping ground. Thus, some
people are more susceptible to specific environmental influences – based on their
genotype – and can be regarded as orchids, whereas other people are less susceptible
to these influences and can be regarded as dandelions (although, naturally, plants that
are more tolerant to environmental conditions are not necessarily less beautiful than
those that are less tolerant). This is called differential susceptibility theory and has
received considerable empirical support in the past decade.
A hypothesis derived from this theory is that orchids benefit more from psycho-
therapy than dandelions. It is well known that the success rate of psychotherapies is
variable – some individuals improve considerably, while others do not. Where do
these large individual differences come from? Part of the answer may be that
individuals with a genotype that makes them more susceptible for environmental
influences – the orchids – benefit more from specific therapies, because they are
more susceptible to both positive and negative environmental influences than dan-
delions. A study on the interaction effect of an intervention and a genotype (the
polymorphism in the promoter of the DRD4 gene) on problem behaviors in children
provided support for this hypothesis (for this and other interesting references, see
Ellis et al. 2011). It was found that the positive effect of the intervention was largest
in the group of children with DRD4 genotypes associated with low levels of
dopamine reception efficiency. This study provided experimental support for the
hypothesis that children are differentially susceptible to intervention effects based on
genetic differences. More studies on different types of differential susceptibility, age
effects, and the relationship with therapy success are necessary.
In sum, this work shows the importance of studying both genetic and environ-
mental differences to explain the development of phenotypical outcomes. The
biological mechanisms underlying these processes remain unknown. Studies on
epigenetics (see next section) should contribute to the understanding of these
mechanisms.
8 A. Ploeger and F. Galis

Epigenetics

Recently, new discoveries have been made under the umbrella of epigenetics.
Epigenetics refers to changes that influence how genes are expressed, other than
changes in the DNA sequence (e.g., Masterpasque 2009). Two major epigenetic
mechanisms are DNA methylation (the attachment of methyl molecules to cytosines,
which switches off the expression of the gene) and histone modification (a change in
the histone proteins around which DNA is wrapped, which causes the expression of
the gene to switch on or off). Research relevant for cognitive science comes from
two main sources: studies on the effects of early caregiving on later development and
studies on the role of epigenetics in psychiatric disorders. Famous work was carried
out on the effects of licking and grooming by the mother on later behavior,
physiology, and epigenetics of newborn rats (for relevant references, see
Masterpasque 2009). They found differences in the reactivity of the hypothalamic-
pituitary-adrenal (HPA) axis of offspring raised by either high licking/grooming or
low licking/grooming mothers. High HPA reactivity is associated with stress
responses and psychiatric disorders in humans. The decrease in HPA reactivity in
high licking/grooming offspring of rats is directly linked to decreased methylation of
glucocorticoid receptor genes in the hippocampus. Cross-fostering experiments
showed that the individual differences in reactivity were the result of licking and
grooming patterns and not of differences in genotype of the mother. Performing
similar experiments in humans is not possible, but postmortem studies with
maltreated versus nonmaltreated humans revealed similar epigenetic patterns as
were found in rats.
It is now well known that epigenetic processes play an important role in the
development of psychiatric disorders. For example, research in mice has revealed
that social stress leads to low levels of brain-derived neurotropic factor (BDNF) in
the hippocampus due to histone modification. Similar patterns have been found in
humans diagnosed with major depression disorder. Considering schizophrenia,
postmortem studies revealed low concentrations of reelin in the brains of patients,
which were associated with hypermethylation of the reelin gene. With regard to
autism, studies were performed on a monozygotic twin, of which one was diagnosed
with autism, while the other was not, despite their identical DNA sequence. Results
revealed that the individual with autism showed methylation-dependent silencing of
the BCL-2 and the RORA gene.
Only a few of the many recent studies on epigenetics that are relevant for
cognitive science have been described here. The studies on epigenetics are at the
heart of evo-devo research: how do inherited DNA sequences interact with devel-
opmental processes that vary under the impact of environmental influences to form
constant or novel phenotypes.
Evo-devo and Cognitive Science 9

Evolutionary Developmental Comparative Cognitive Science

Evolutionary developmental comparative cognitive scientists compare the cognition

of nonhuman primates and human children. The book The Origins of Intelligence
(Parker and McKinney 1999) is a hallmark in this field. In this book, the develop-
ment of different primate species was compared, following the theory on cognitive
development by Jean Piaget, as mentioned above. However, the choice to test
children is usually not made because researchers are interested in a developmental
perspective; often adult nonhuman primates and young human children (2–3-years-
olds) are compared because their cognitive levels are similar. For example, the
Primate Cognition Test Battery has been developed in order to have tests that both
adult nonhuman primates and young human children can perform. Recently, a first
large study was published where young chimpanzees and bonobos (together called
Pan infants) and human children were followed longitudinally for 3 years with a new
test battery, the Comparative Developmental Cognitive Battery (for relevant refer-
ences, see Tomasello 2014). Individuals were tested on social cognition (e.g., gaze
following, imitation, goal understanding), physical cognition (e.g., discriminating
different quantities, tool use, understanding of object permanence), attention, and
motivation. Results revealed that over all tasks, the rate of improvement is slower in
Pan infants than in human children, and that abilities that require cooperative
motivation do not emerge at all in Pan individuals. This is useful research, because
it provides us insight in the differences and similarities between three closely related
species, including two sister species of humans, and thereby also in the cognitive
evolution of humans.

Evolutionary Developmental Cognitive Neuroscience

Evolutionary developmental cognitive neuroscientists study how brains change,

both from an evolutionary and a developmental perspective. Two general models
about brain evolution and development have emerged. One model emphasizes the
relative independence and modularity of different brain structures, assuming that, for
example, the auditory system requires different neural networks from the olfactory
system. From an evolutionary perspective, it is argued that most brain areas are
functionally specialized, and hence selection pressures will differentially affect brain
areas (Barrett 2012). The other model assumes that the entire brain will change in
response to selection pressures, and that architectural and functional constraints
ensure that brain size as a whole will change.
Four types of brain growth have been observed in the evolution of mammals
(Finlay et al. 2001). First, brain growth that is associated with body growth; when the
body grows, the brain grows accordingly. Second, the brain can grow while body
size remains constant; this kind of brain growth is associated with enhanced behav-
ioral and cognitive capabilities in the course of evolution. Third, the limbic system,
the part of the brain associated with emotion, motivation, memory and olfaction,
grows independently of overall brain and body size. Fourth, other individual brain
10 A. Ploeger and F. Galis

parts may vary in size independently of overall brain and body size. For example,
prefrontal gray volumes are 4.8 times larger in humans compared to chimpanzees
(for a review, see Schoenemann 2006). In addition, the relative size of the neocortex
and striatum is positively correlated with tool use, innovation, and social learning.
These four different types of growth indicate that both independent evolution of
different brain structures and size changes of the entire brain have been important for
brain evolution and development.

Evolutionary Developmental Biology and Cognitive Science

The last subject addressed is the question what cognitive scientists can learn from
evo-devo research in biology. Many evo-devo biologists study morphogenesis, and
it is not immediately obvious how to relate their research findings to cognitive
science. This issue has been addressed in an earlier paper (Ploeger and Galis 2011;
relevant references can be found in this paper). A first conclusion was that some of
the main issues in evo-devo biology and cognitive science overlap and that tools can
be used profitably by both types of scientists. For example, modularity is a main
topic in both evo-devo biology and cognitive science (see chapter on “▶ Modularity
in Evo-devo”). Evo-devo biologists study the modularity of developmental and
genetic pathways as well as that of body parts, whereas cognitive scientists study
the modularity of the human mind. Evo-devo biologists have developed tools to
study modularity that have been largely unnoticed by most cognitive scientists. It
was argued that cognitive scientists can benefit from these tools. Another example is
the issue of plasticity. Both evo-devo biologists and cognitive scientists are inter-
ested in the question how plasticity is important in an individual life time and how it
evolved over generations. It was also argued in this case that evo-devo biologists
have developed tools that should benefit cognitive scientists.
A second conclusion was that evo-devo biology research can provide new
insights in the evolution and development of psychiatric disorders. One example is
research on developmental constraints. For example, it was proposed that mutations
that give rise to the positive aspects of the savant syndrome, i.e., the impressive
memory capacity, cannot spread in the population, due to a developmental constraint
that has its roots in low modularity. This developmental constraint is thought to
result from the high interactivity (low modularity) among body parts during early
organogenesis (i.e., the phylotypic stage). The interactivity during this stage involves
all components of the embryo, and as a result mutations that affect one part of the
embryo also affect other parts (pleiotropic effects or side-effects), with almost
inevitably negative effects among them. As a result of the sheer unavoidable
deleterious side-effects, there is strong selection against mutations with an effect
on this stage, presumably leading to the extremely strong conservation of the entire
stage. The low modularity of this embryonic stage has implications for the conser-
vation of many traits of the body plan and is for example at the root of the strong
developmental constraint against changes of the number of cervical vertebrae in
mammals. The same hypothesis was proposed for the savant syndrome. Mutations,
Evo-devo and Cognitive Science 11

which give rise to the development of the positive aspects of the savant syndrome, i.
e., an impressive memory capacity, will virtually always have deleterious side-
effects on the development of other phenotypic traits. The support for such strong
deleterious side-effects that are associated with the savant syndrome (e.g., autism
and/or impaired motor coordination) was discussed. One of the new insights that
were reported is that psychiatric disorders that result from brain deviations usually
appear to start to develop as early as during the phylotypic stage, due to the general
instability and vulnerability of the stage that results from the intense inductive
interactivity. Another example is research on epistatic interactions between the
effects of different genes. It is a paradox why psychiatric disorders, such as schizo-
phrenia and autism, are common, why they are highly heritable, and why they still
persist. Why did natural selection not wipe out these disorders? The answer lies in
the polygenic nature of most psychiatric disorders. When multiple genes are
involved, the effects of these genes will interact during development, sometimes
resulting in positive but sometimes in negative outcomes. Interdisciplinary
approaches in which insights from evo-devo research on morphogenesis have
shown to yield new hypotheses about the evolution and development of psychiatric
disorders.

Conclusion

Evo-devo in cognitive science consists of a wide array of subfields, including

evolutionary developmental cognitive research, developmental systems theory,
genetic research (GE interactions), epigenetic research, comparative research,
neuroscientific research, and applications of evo-devo biology in cognitive science.
Interdisciplinary research is strongly recommended, so that scholars from different
fields such as theoretical biology, morphology, embryology, genetics/genomics,
neuroscience, primatology, and psychology can learn from each other and contribute
to the unraveling of the working of the mind.

Cross-References

▶ Epigenetic Innovation
▶ Evo-devo and Cognitive Science
▶ Evo-devo and Culture
▶ Evo-devo of Language and Cognition
▶ Evo-devo of Social Behavior

References
Barkow JH, Cosmides C, Tooby J (eds) (1992) The adapted mind: evolutionary psychology and the
generation of culture. Oxford University Press, New York
12 A. Ploeger and F. Galis

Barrett HC (2012) A hierarchical model of the evolution of human brain specializations. P Natl
Acad Sci USA 109:10733–10740
Bjorklund DF (2009) Why youth is not wasted on the young: immaturity in human development.
Wiley-Blackwell, Hoboken
Bjorklund DF (2015) Developing adaptations. Dev Rev 38:13–35
Caspi A, McClay J, Moffitt TE, Mill J, Martin J, Craig IW, Taylor A, Poulton R (2002) Role of
genotype in the cycle of violence in maltreated children. Science 297:851–854
Ellis BJ, Boyce WT, Belsky J, Bakermans-Kranenburg MJ, van Ijzendoorn MH (2011) Differential
susceptibility to the environment: an evolutionary–neurodevelopmental theory. Dev
Psychopathol 23:7–28
Finlay BL, Darlington RB, Nicastro N (2001) Developmental structure in brain evolution. Behav
Brain Sci 24:263–308
Geary DC (2005) The origin of mind: evolution of brain, cognition, and general intelligence.
American Psychological Association, Washington, DC
Karmiloff-Smith A (2006) The tortuous route from genes to behavior: a neuroconstructivist
approach. Cogn Affect Behav Neurosci 6:9–17
Markram H et al (2011) Introducing the human brain project. Procedia Comput Sci 7:39–42
Masterpasque F (2009) Psychology and epigenetics. Rev Gen Psychol 13:194–201
Miller G (2000) The mating mind: how sexual choice shaped the evolution of the human mind.
Vintage, London
Newell A (1990) Unified theories of cognition. Harvard University Press, Cambridge, MA
Oyama S, Griffiths PE, Gray RD (eds) (2001) Cycles of contingency: developmental systems and
evolution. MIT Press, Cambridge, MA
Parker ST, McKinney ML (1999) The origins of intelligence: the evolution of cognitive develop-
ment in monkeys, apes, and humans. Johns Hopkins University Press, Baltimore
Ploeger A, Galis F (2011) Evo devo and cognitive science. WIREs Cogn Sci 2:429–440
Richerson PJ, Boyd R (2005) Not by genes alone: how culture transformed human evolution.
University of Chicago Press, Chicago
Schoenemann PT (2006) Evolution of the size and functional areas of the human brain. Annu Rev
Anthropol 35:379–406
Spelke ES, Kinzler KD (2007) Core knowledge. Dev Sci 10:89–96
Thelen E, Smith LB (1994) A dynamic systems approach to the development of cognition and
action. The MIT Press, Cambridge, MA
Tomasello M (2014) A natural history of human thinking. Harvard University Press, Cambridge,
MA
Evo-Devo of Social Behavior

Kate E. Ihle, Gro V. Amdam, and Adam G. Dolezal

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Behavioral Modules of Social Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
From the Individual to High Society . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
A Ground Plan for Social Living . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Food and Society . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Genetic Toolkits Shape Social Life-Histories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Molecular Pathways Associated with Reproduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Molecular Pathways Associated with Nutrition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Insulin/Insulin-Like Growth Factor Pathway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Genomic Approaches to Social Evo-Devo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Epigenetics and Social Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

K.E. Ihle (*)

Department of Evolutionary Biology and Environmental Science, University of Zurich, Zurich,
Switzerland
e-mail: kateihle@gmail.com
G.V. Amdam
School of Life Sciences, Arizona State University, Tempe, AZ, USA
Department of Ecology and Natural Resource Management, Norwegian University of Life
Sciences, Aas, Norway
e-mail: Gro.Amdam@asu.edu
A.G. Dolezal
Department of Ecology, Evolution, and Organismal Biology, Iowa State University, Ames, IA, USA
e-mail: adolezal@iastate.edu

# Springer International Publishing Switzerland 2016 1

L. Nuño de la Rosa, G.B. M€uller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_45-1
2 K.E. Ihle et al.

Abstract
The interdisciplinary field of evolutionary developmental biology (evo-devo)
combined the strengths of developmental biology with insights from molecular
evolutionary biology to rapidly advance our understanding of morphological
differentiation and the evolution of diverse multicellular life-forms. The same
concepts have been applied to the study of the evolution of social behavior to
good effect. By treating correlated behavioral and physiological states as units,
similar to the developmental modules defined by more traditional evo-devo
approaches, evolutionary developmental biologists have identified common
“genetic toolkits” involved in the transition from solitary to social and eusocial
life-histories across independent evolutionary origins of social behavior. Data
from the rapidly expanding library of high-quality genomes from organisms
across the tree of life has enabled direct tests of the predictions of an evo-devo
approach to the evolution of social behavior, demonstrating both the power and
limitations of such an approach to explain the evolution of behavioral traits as
well as generating new insights into the often shared proximate mechanisms
underlying the evolution of social behavior.

Keywords
Social evolution • Phenotypic plasticity • Developmental plasticity

Introduction

The origin of eusocial behavior is one of the major transitions in evolution (Maynard
Smith and Szathmáry 1997), and a critical ongoing challenge for biologists is to
understand the genetic basis of such major transitions in complex traits. The
advances made by the field of evolutionary development in our understanding of
morphological evolution suggest a template for the investigation of the proximate
evolutionary pathways to social behavior (Dolezal et al. 2014; Toth and Robinson
2007). The advances in traditional evolutionary developmental biology were closely
linked to the technical advances in the fields of genetics and molecular biology to
identify how evolution has acted on common molecular networks to give rise to
incredibly diverse body plans (Carroll 2008) Similarly, a rapidly expanding array of
sequenced genomes from organisms with varied social behavior within and between
lineages has begun to yield exciting insights into the shared proximate pathways to
social behavior (see Kapheim et al. 2015 and references therein).
Social behavior is a complex trait and can be found in many forms in many
different taxa (Gadau et al. 2009). In this chapter, we focus on the social behaviors
that take place within permanent associations of conspecific individuals. These
forms of social behavior range from simple communal associations in which inde-
pendently reproducing females share a nest site to the vast advanced eusocial
societies composed of morphologically distinct queens that dominate reproduction
and sterile workers that perform all other tasks required by the colony. With
Evo-Devo of Social Behavior 3

increasing levels of social complexity come associated increasing costs to an indi-

vidual in terms of direct fitness. For example, in communal groups, the interaction
between the size of the nest and number of resident females may constrain the
number of reproductive opportunities for the individual females. In primitively
social and advanced eusocial societies, workers have either very limited or no
possibilities for direct reproduction. The latter, most extreme form of social behavior
is found primarily within the social insects: ants, bees, wasps, and termites, and so
they will be the primary subjects of this chapter.
How does complex social behavior evolve, if it comes with such rising direct
fitness costs? To address this question, the evolutionary biologist William
D. Hamilton proposed the concept of “inclusive fitness” to explain how workers
may gain in overall fitness by giving up direct reproduction (Hamilton 1964).
According to this hypothesis, the fitness of an individual is composed of both direct
fitness, individual reproductive output, and indirect fitness, the reproductive output
of related individuals with shared genetic makeup. Here, workers may increase their
overall fitness by helping their mother or sister produce fertile offspring. Inclusive
fitness theory predicts that altruistic helping behaviors can be favored by selection
when the fitness benefits to helping outweigh the costs of foregoing reproduction.
This hypothesis can be summed up by Hamilton’s Rule: b x r > c where b = the
benefit to the receiver of helping behavior, r = the genetic relatedness between the
helper and the receiver of helping behavior, and c = the cost to the altruistic helper
(Hamilton 1964). Recent work has challenged the importance of inclusive fitness in
the evolution of social behavior, arguing that group rather than kin selection is likely
to be the greater evolutionary force (Nowak et al. 2010). This position remains
controversial, with many scientists defending the foundations of Hamilton’s
hypotheses.
While ultimate explanations for the evolution of social behavior aim to under-
stand why social behavior evolved, they can tell us little about how selection acting
on individual phenotypes gives rise to social behavior. These how questions are
particularly amenable to an evolutionary developmental biology approach. By
identifying the basic modules of behavior that combine to form complex behavioral
phenotypes within and across species, the evolutionary developmental biology of
social behavior uses a comparative approach to identify the physiological and
molecular pathways that were targeted by selection for the regulation of social
phenotypes.

Behavioral Modules of Social Evolution

Eusocial insect societies represent the most complex social organization in the
animal kingdom. A major obstacle to studying the evolution of eusocial behavior
had been the fact that many eusocial species diverge long ago from solitary relatives.
An evo-devo approach to the evolution of social behavior uses a broad comparative
method to identify and characterize behavior modules in the life-cycles of the
4 K.E. Ihle et al.

solitary relatives of social species. These behavioral modules are then hypothesized
to form the building blocks of the complex social behavioral phenotypes that make
up a eusocial society. The guiding principle of an evo-devo approach to the evolution
of social behavior is the understanding that selection can only act on the expressed
phenotypic variance within a population and that of the phenotypic correlations
between physiology and behavior present in extant solitary relatives of eusocial
species is reflective of their shared solitary ancestors.

From the Individual to High Society

Solitary Ancestors
Eusocial behavior evolved from the variation inherent in the presumably solitary
ancestors of today’s “super organisms.” Biologists use a comparative approach to
build hypotheses about the behavior of these ancient insects by studying the extant
solitary relatives of eusocial species. The life-cycle of a typical solitary female bee in
the temperate zone begins in summer, when she either emerges from a winter-long
diapause or ecloses as a new adult, after which she will leave the nest and mate. A
period of adult development then begins, fueled by the carbohydrates in collected
floral nectars. Female bees require a protein rich food source to activate their ovaries
for egg production and will begin foraging for pollen prior to brood cell construction
and provisioning. When a brood cell is complete, the female will hoard pollen to
form a small ball, consisting of pollen mixed with nectar onto which she will lay an
egg before sealing the cell in the case of mass provisioning species. Progressively
provisioning species will continue to feed the developing larvae before sealing the
cell prior to pupation. After the female closes the brood cell, she begins the process
again.

Communal Behavior
Communal behavior occurs when two or more females share a nest site but provision
and care for their own offspring independently (Michener 1974). In communal
species, the behavioral blueprint from solitary species is largely intact. Communally
nesting females need not be related and communal associations can occur when
multiple foundresses share the cost of establishing a nest. These associations are also
often formed when sisters emerge and remain as adults at their natal nest to rear their
own offspring. Communal behavior has been hypothesized as an evolutionary
“stepping-stone” to eusocial behavior. In this scenario, alleles favoring cooperative
behavior would be selected for when communal groups are able to outperform
solitary competitors, likely through benefits to shared nest defense of decreased
costs of nest founding (Hölldobler and Wilson 2009). However, the evidence for this
hypothesis is mixed, with recent studies suggesting that communal behavior is a
stable phenotype and a potential evolutionary alternative strategy to eusocial behav-
ior as it occurs primarily in lineages that lack more complex forms of social
organization (Gadau et al. 2009).
Evo-Devo of Social Behavior 5

Primitive Eusociality
Females in primitively eusocial societies have a reproductive division of labor and
participate in shared rearing of the colony’s offspring (Michener 1974). In these
societies, there is no morphological caste differentiation, but reproduction is monop-
olized by one or a few dominant individuals that function as queens. The behavioral
life cycle of a solitary ancestor may be recognizable as a solitary queen founds a nest.
However, when her offspring become adults, a behavioral caste system emerges in
which the queen primarily specializes on egg production, while the workers perform
tasks such as brood rearing, nest maintenance, and foraging. Unlike the workers
from advanced eusocial societies, primitively eusocial workers are physically capa-
ble of mating and reproduction, and queens typically achieve reproductive skew
through physical domination. Should the dominant queen die or become less
competitive, a worker may usurp her position as the primary reproductive in the
colony.

Advanced Eusociality
Animal social groups must meet three characteristic criteria before they are consid-
ered to be advanced eusocial societies: (1) reproductive division of labor, (2) coop-
erative brood care, and (3) overlap of generations (Wilson 1971). Advanced eusocial
societies are characterized by morphological castes with a near complete reproduc-
tive division of labor between the highly fecund queen and the often sterile workers.
The queens of mature eusocial colonies focus entirely on egg production, while the
workers, who are unable to mate and rarely reproduce, perform the rest of tasks
required to maintain the colony. In many advanced eusocial species, behavioral
divisions of labor between worker groups are also present. These behavior divisions
can be related to morphological differences, as in species with specialized soldier
castes, or they can be age-based, as exemplified by the temporal polyethism of the
honey bee (Winston 1987). Honey bee workers pass through an age-associated
progression of tasks known as beginning with tasks within the brood nest such as
the tending of larvae and cell cleaning. Older workers leave the nest as foragers,
collecting the food resources required by the colony. Despite the often huge mor-
phological differences between the queen and worker castes, these phenotypes
develop from the same genome and are triggered by worker-controlled nutritional
rearing regimes rather than by genetic differences.

A Ground Plan for Social Living

The “ground plan” hypotheses on the evolution of eusocial behavior posits that the
complex behavioral phenotypes that make up a eusocial insect colony are derived
from associations between behavior and reproductive physiology present in the
solitary ancestors of social species (West-Eberhard 1987; Amdam et al. 2004). The
reproductive ground plan hypotheses expanded the methods of traditional evo-devo
by treating the correlations between behavior and reproductive physiology that occur
6 K.E. Ihle et al.

at different life-cycle stages as modules, similar to the developmental modules,

exemplified by the classical Hox genes model, that form the foundations of evolu-
tionary developmental biology. These hypotheses argue that social behavioral phe-
notypes are not the product of newly evolved molecular pathways, but rather the
result of developmental differences in the ancestral phenotype. Selection could then
favor the divergent phenotypes, acting on the mechanisms that regulate the sequen-
tial phases of a solitary reproductive cycle to instead direct the social behaviors of
queens and workers.

The Ovarian Ground Plan Hypothesis (OGPH)

In her work with Polistes paper wasps, Mary Jane West-Eberhard observed that
ovarian development as well as hormone levels in queens as well as workers is
correlated with behavior. The OGPH accounts for reproductive division of labor
between queens and workers, and developmental differences influenced the cycle of
reproduction and provisioning behavior tied to ovary development and reproductive
physiology in solitary species. According to the OGPH, the selection acting on these
developmentally derived differences segregated the reproduction and provisioning
phases of a solitary female life cycle into the queen and worker phenotypes such that
queens had developed ovaries and high titers of gonadotropic juvenile hormone
(JH) while workers have undeveloped ovaries and lower-hormonal titers (Fig. 1).
The OGPH was later expanded to include an explanatory mechanism for “age
polyethism,” a correlation between behavior and chronological age that is seen in
the worker caste of many social insects in which young females remain on the nest
while older workers leave the nest on foraging trips (West-Eberhard 1987, 1996).
Here, behavioral maturation in workers is influenced by ovary development and
JH. Younger workers who remain on the nest have slightly developed ovaries and the
capacity for opportunistic reproduction should the queen fail. Older workers leave
the nest as foragers as their ovaries regress and their potential as replacement
reproductive declines (Fig. 1b, West-Eberhard 1996). The relatively high physical
cost to foraging is then balanced by decreasing potential for direct reproduction.

The Reproductive Ground Plan Hypothesis

The reproductive ground plan hypothesis (RGPH) of Gro V. Amdam and Robert
E. Page Jr. is an extension of the OGPH that relates specifically to the foraging
behavior of workers (Amdam et al. 2004). The RGPH was built upon an extensive
body of research on the high and low pollen-hoarding strains of honey bees
developed by Robert Page and M. Kim Fondrk (Page 2013). The pollen-hoarding
strains were divergently selected for the amount of pollen stored in the hive, a
colony-level trait. This colony-level selection regime also resulted in large pheno-
typic differences between the workers, particularly in food collection behaviors and
reproductive physiology (Table 1). High pollen-hoarding strain workers begin
foraging at earlier ages than workers from the low pollen-hoarding strain. They
also bias their food collection toward protein-rich pollen, while low pollen-hoarding
workers bias their foraging toward nectar, the carbohydrate source for the colony.
Amdam and Page noted that workers from high pollen-hoarding lines had larger
Evo-Devo of Social Behavior 7

a Solitary brood cell

preparation

oviposition

developed ovaries

undeveloped ovaries

nursing/forage/
guard

queen
developed
ovaries
b Eusocial

nectar forager
oviposition
smaller ovaries
pollen forager worker
larger ovaries undeveloped
ovaries

ecolosion
/brood
RGPH cell cleaning

nursing/guard/nest
maintenance

Fig. 1 The ovarian ground plan hypothesis (OGHP, West-Eberhard 1987, 1996) argues that the
social phenotypes of queens and workers are developmentally derived from the ovarian cycle of a
solitary ancestor (specifically, a progressively provisioning species with a long provisioning
period). The reproductive ground plan hypothesis (RGPH, Amdam et al. 2004) extends the
OGPH to explain the foraging division of labor for nectar versus pollen collection observed in
honey bee workers. (a) A linked cycle of reproductive physiology and behavior frames the life cycle
8 K.E. Ihle et al.

Table 1 Many physiological and behavioral traits of individual bees were affected by the colony-
level selection for high and low pollen hoarding. These divergent suites of traits inspired the
reproductive ground plan hypothesis
High pollen hoarding Low pollen hoarding
Pollen stores Larger Smaller
Egg to adult development Slower Faster
Peak Vitellogenin titer Higher Lower
Vitellogenin/juvenile hormone feedback Stronger Weaker
Sucrose sensitivity Higher Lower
Ovary size Larger Smaller
Foraging onset Earlier Later
Foraging loading Pollen bias Nectar bias

ovaries (number of ovarian filaments, called ovarioles) than low-strain workers.

These larger ovaries are accompanied by higher production of the yolk precursor
protein, Vitellogenin. The RGPH argues that the variation in pollen versus nectar
foraging observed in honey bee workers reflects and ancestral association between
reproductive physiology and foraging behavior (Fig. 1b, Amdam et al. 2004).
In solitary insects, reproductive physiology is tied to nutritional input. Mosqui-
toes, for example, forage on nectar during periods of self-maintenance when their
ovaries are inactive. Protein intake, in this case a blood meal, is required for female
mosquitoes to activate their ovaries and fuel egg production. The RGPH posits that
this ancestral association between protein foraging and reproduction is still present
in the functionally sterile honey bee workers. Here, more reproductive-tuned
workers, those with larger ovaries and higher titers of Vitellogenin will bias their
foraging loading toward protein-rich pollen, while less reproductively tuned indi-
viduals will bias their collection toward carbohydrate-dense nectar (Amdam
et al. 2004).

Fig. 1 (continued) of solitary females. As a female prepares a cell in which to deposit an egg, her
ovaries are well developed. After oviposition, her ovaries remain undeveloped as she tends, guards,
and forages for her developing larva. Her ovaries begin to swell with developed oocytes as she
prepares the next brood cell. (b) In eusocial species, the OGPH posits that this cycle has been
interrupted with the oviposition and brood care phases segregated into the queen and worker castes,
respectively. Large, active ovaries support the sometimes enormous reproductive output of queens.
Workers have underdeveloped ovaries and pass first through a stage of brood care, or nursing,
behavior before transitioning to outside of the nest foraging behavior as they age and the likelihood
of direct reproduction decreases. The RGPH argues that reproductive physiology regulates the
division of foraging labor observed between honey bee workers for pollen versus nectar collection.
Here more reproductively tuned workers with larger-ovaries (more ovarioles per ovary) and higher-
peak titers of Vitellogenin bias their foraging efforts toward protein (pollen) collection while
workers that are less reproductively tuned workers bias their foraging toward carbohydrate (nectar)
collection
Evo-Devo of Social Behavior 9

Food and Society

Nutrition status and food collection play an enormous role in shaping the life-history
of any organism, and this is doubly true for the social insects where nutritional inputs
during development determine the caste fate of a developing larva. The role of
nutritional status in the triggering of developmental programs is especially clear in
the advanced eusocial insects. In honey bees, for example, queen-destined larvae
receive more and higher nutrient content food than do worker-destined larvae. In
nature, these nutritional differences trigger discrete developmental programs that
produce the morphologically distinct queen and workers castes. The workers tending
the larvae, known as nurses, must tightly control the amount, quality, and timing of
larval feedings in order to produce the distinct queen and worker phenotypes. In the
lab, researchers routinely produce intercaste females, which have morphological
features of both queens and workers, by varying the feeding regimes (Page 2013).
There are no morphological castes in primitively eusocial species, but nutrition
during development also plays a role in determining behavioral caste-fate in these
species. Primitively eusocial queens are often bigger than workers, and large, newly
emerged females are more likely to leave their natal nest as foundresses than their
smaller sisters. The differences in body size are in part due to differential nutrition
during development.
Nutritional status also plays a role in the pacing of behavioral maturation in social
insects. Foraging is the final behavioral phase of life for the workers of many species.
The foraging behaviors of solitary and social individuals of related species can be
very similar. Bees gather floral nectars and pollens regardless of their social organi-
zation, but while solitary females gather food to increase their own fitness, social
workers forage to increase the fitness of the colony. Nutritional physiology links the
food collection behaviors of solitary and social species. Foragers often have depleted
nutrient reserves relative to other behavioral castes, and experimental work in the
honey bee demonstrated that reduction of lipid reserves in workers speeds their
transition to foraging behavior (reviewed in Toth and Robinson 2007). Interestingly,
it is not only individual nutrient status that can induce early foraging in honey bee
workers. Reduction of the stored food within a colony also causes early foraging.
These results suggest that the physiological mechanisms that regulate the foraging
behaviors of solitary species have been co-opted by selection to regulate temporal
polyethism in social species.

Genetic Toolkits Shape Social Life-Histories

After the discoveries showing that the reproductive and nutritional ground plans of
solitary insects also regulated social insect life-history traits, researchers began using
targeted approaches to understand the molecular underpinnings of social behavioral
modules. Many of these studies initially targeted candidate genes identified from
quantitative trait loci (QTLs), genomic regions statistically associated with a partic-
ular phenotype. While these QTLs were identified in the high and low pollen-
10 K.E. Ihle et al.

hoarding strains (Page 2013), many show shared brain gene expression patterns with
Polistes wasps performing food collection behaviors (Toth et al. 2010).
A traditional evolutionary development approach suggests that changes in the
regulation of genes and their expression, rather than functional changes in protein
coding, is the predominant driver of evolutionary change (Carroll 2008). Amy
L. Toth and Gene E. Robinson (2007) argued that gene expression differences
may be particularly important for the evolution and regulation of social phenotypes.
They proposed a model built from the concept of the eusocial insect colony as a
“super organism,” which could then be broken down analogously to the different cell
types that make up an organism. Toth and Robinson (2007) extended this idea to
hypothesize that, just as changes in the pattern of gene expression distinguish
different cell and tissue types in a single organism, transcriptional differences also
distinguish the behavioral and morphological phenotypes of the members (i.e., the
“cells”) of a social insect colony superorganism. Targeted molecular studies
supported a role for reproductive and nutritional ground plans in the evolution of
social behavior and identified potential “genetic toolkits” that regulate these associ-
ations between behavior and physiology across the spectrum of social complexity
(reviewed in Toth and Robinson 2007; Dolezal and Toth 2014).

Molecular Pathways Associated with Reproduction

The dynamic differences observed in the Vitellogenin titers of the high and low
pollen-hoarding honey bee strains were one factor that inspired the formation of the
RGPH. Subsequent experimental work confirmed a regulatory role for Vitellogenin,
not only in the decision to collect nectar versus pollen but also in the timing of the
transition to foraging behaviors (Amdam and Page 2010), and demonstrated that
Vitellogenin is one of the central players in a reproductive genetic toolkit that has
evolved a regulatory role for non-reproductive social behaviors. Expression of the
vitellogenin gene is higher in young, high pollen-hoarding strain workers, but the
decline in vitellogenin expression with age also occurs earlier and more rapidly in
this strain, corresponding to an earlier age of foraging behavior. Experimental
reduction of vitellogenin expression via RNA-interference (RNAi) caused workers
to collect heavier nectar loads and transition to foraging behavior at younger ages,
confirming a regulatory role for Vitellogenin in both the onset of foraging as well as
food collection decisions (Page 2013). A role for Vitellogenin in the regulation of
social behaviors has since been confirmed in several species. For example, vitello-
genin has recently been shown to be associated with parental care behaviors in
burying beetles, suggesting that Vitellogenin and its link to both reproduction and
provisioning behaviors is a powerful and flexible module in the evolution of
behavioral phenotypes (Roy-Zokan et al. 2015).
The novel functions of vitellogenin in the regulation of social behaviors have
likely evolved through several mechanisms. Most studies to date have focused on
changes in the spatial and temporal expression of vitellogenin as underlying its roles
in the evolution of novel phenotypes, but recent discoveries of gene duplications of
Evo-Devo of Social Behavior 11

vitellogenin and vitellogenin-like genes in several species suggest that different or

parallel mechanisms may be responsible for the novel functions of Vitellogenin.
While Vitellogenin is known to regulate behaviors in many species, its mode of
action is currently unknown. However, a growing body of research suggests that
interactions with the systemic JH and the insulin/insulin-like signaling (IIS) pathway
may mediate the effects of Vitellogenin on behavior.

Molecular Pathways Associated with Nutrition

From the moment a larvae hatches from the egg, nutrition and nutrient-sensing
pathways dictate much of its future physiological and behavioral destiny. Nutrition
during development plays a key role in caste-differentiation between workers and
queens, as well as between morphological worker castes in some species (Dolezal
et al. 2014). Later, these pathways also regulate the behavior of workers, either
directly or through their effects on nutrient storage and metabolism (reviewed in
Dolezal et al. 2014).
Signaling pathways and genes associated with nutrient sensing and balance
comprise a second genetic toolkit that regulates social phenotypes (Toth and Rob-
inson 2007). The nutrient sensing Target of Rapamycin (TOR) pathway plays a key
role in the developmental switch between queen- and worker-destined larvae in
honey bees. In larvae fed a diet that results in development as queens, experimental
suppression of TOR instead channels them into the worker developmental pathway.
Expression of hexamerins, genes encoding amino acid storing proteins differ in
queen versus worker-destined larvae in several species. Additionally, experimental
manipulation of hexamerin expression in termites demonstrated that they have a
regulatory function in the differentiation of soldier versus worker phenotypes in
termites (reviewed in Dolezal et al. 2014). Experimental work in other taxa is needed
to test how universal is hexamerin regulation of social phenotypes.
Early targeted studies used comparisons with Drosophila melanogaster to test
correlations between the expression of candidate genes and behavioral modules in
social insects (reviewed in Toth and Robinson 2007). The Drosophila foraging ( for)
gene encodes the cyclic-GMP-dependent protein kinase G (PKG). Allelic variation
in the for gene produces “sitter” and “rover” fly phenotypes. Rovers have higher
expression of for and range farther to forage than sitters. The sitter and rover
Drosophila phenotypes were reminiscent of the nurse versus forager behavioral
modules in honey bees to Yehuda Ben-Shahar and colleagues, who demonstrated
that foragers have elevated expression of Amfor, the honey bee orthologue of the for
gene. Further, experimental elevation of PKG signaling induces early foraging onset,
demonstrating a regulatory role for PKG in foraging behavior (reviewed in Toth and
Robinson 2007). Intriguingly, this pattern is reversed in the harvester ant
Pogonomyrmex barbatus, highlighting the flexibility of this regulatory module
(reviewed in Dolezal et al. 2014).
The behaviorally associated differences in the expression of for genes are regu-
lated in part by both vitellogenin and JH, highlighting the important connections
12 K.E. Ihle et al.

between reproductive and nutrient-sensing toolkits. The JH/Vitellogenin regulatory

module is involved in many of the physiological changes that accompany the
transition from nursing to foraging behavior in several social species (reviewed in
Dolezal et al. 2014). The age-associated lipid loss in honey bee workers prior to
foraging onset is influenced by both dietary and non-nutrient signaling. Vitellogenin
expression can influence lipid storage and metabolism, independently of dietary
factors (Ament et al. 2011). Vitellogenin has also been implicated in the mobilization
of stored lipids and carbohydrates that fuels foraging behavior via its role as an
upstream regulator of adipokinetic hormone (AKH). AKH is an insect glucagon
equivalent, and the expression of its receptor differs between nurse and forager
honey bee workers, suggesting that it may play a role in the regulation of social
behavioral modules (reviewed in Dolezal et al. 2014).

Insulin/Insulin-Like Growth Factor Pathway

The insulin/insulin-like growth factor pathway (IIS) is a major regulatory pathway

linking the reproductive and nutritional genetic toolkits. IIS is a conserved, central
regulator of life-history with regulator roles in both nutrition and reproduction
(Kenyon et al. 2004). In insects, IIS plays a crucial role in egg development, but it
is best studied in its role as a regulator of resource allocation in adults. The IIS
pathway helps to direct resources (i.e., nutrition) toward reproduction when nutrients
are plentiful and toward self-maintenance when resources are scarce. In this capacity,
the IIS pathway has ancient connections to both JH and Vitellogenin (reviewed in
Toth and Robinson 2007).
These interacting regulatory mechanisms appear to be especially accessible to
selection, both natural and artificial. IIS pathway genes are over-represented in QTLs
for traits in the pollen-hoarding strains of honey bees (Page 2013), suggesting that
the behavioral and physiological phenotypes affected by the artificial selection
regime for the high and low strains are under pleiotropic control of this pathway.
The insulin receptor substrate (IRS) gene encodes a membrane-associated protein
that relays signals from the insulin receptor and was identified as a candidate gene for
foraging behavior. Experimental reduction of IRS expression via RNAi confirmed its
regulatory role in foraging preference, causing worker bees to collect less nectar
(reviewed in Dolezal et al. 2014). IRS may affect behavior via downstream effects on
JH signaling, as it does during honey bee development, but this hypothesis awaits
direct testing. However, IRS can also act as a substrate for other receptors, and it is
possible that some of its effects on behavior may be downstream of other pathways
such as epidermal growth factor.
IIS pathway components are differentially expressed in developing queens versus
workers in several species, suggesting that it may play a critical role in the ontogeny
of social phenotypes. There is growing interest in the IIS pathway as common
pathway for the evolution of social behavior, and future research will identify how
the diverse molecules that form the IIS regulate both behavior and development in
social species.
Evo-Devo of Social Behavior 13

Genomic Approaches to Social Evo-Devo

Until recently, most studies examining the proximate pathways of social evolution
used a targeted approach, examining one or a few candidate genes identified from
QTL studies or comparisons with Drosophila. As more and more high-quality, well-
annotated genomes of social organisms became available, the field is beginning to
move toward whole-genome approaches, using large population studies to identify
genomic regions under selection or a comparative approach using species with a
variety of social structures to identify genomic correlates of social behaviors. Such
studies have provided support for the evolutionary developmental approach to the
study of social behavior (see Kapheim et al. 2015 and references therein). As the
evo-devo model suggests, large-scale comparative genomic studies have found
evidence that complex social behaviors are governed by conserved and pleiotropic
regulatory modules also present in related species with different levels of social
complexity (Woodard et al. 2011).
A prediction of genetic evo-devo is that evolution of gene regulatory mecha-
nisms, rather than changes in protein coding genes, are likely to underlie the
evolution of complex traits, especially at increasing levels of pleiotropy (Carroll
2008). In support of this hypothesis is the finding that with increased levels of social
complexity come an increased genomic capacity for gene regulation (Kapheim
et al. 2015). This increased regulatory capacity comes in several forms, including
increased transcription factor pleiotropy, more transcription factor binding sites in
cis-regulatory regions, and evidence for rapid evolution of signal transduction
pathways (Kapheim et al. 2015). These authors also found evidence for an associ-
ation between social complexity and constraints on protein evolution.
However, it is clear that changes in protein coding sequences are also key to the
evolution of novel traits including social behavior (Jasper et al. 2015 and references
therein), with so-called novel taxonomically restricted genes (TRG) contributing to
the development of novel social tissues such as glands that produce social phero-
mones. The expression of TRG is often both spatially and temporally limited and
regulated by conserved, pleiotropic regulatory pathways upstream (Jasper
et al. 2015). The position of a gene in a molecular network is associated with its
exposure to positive selection. Many recent studies have found that genes at the
periphery of regulatory network modules are more likely to experience positive
selection. Differences in the coding sequences of these genes are less likely to have
large and possibly deleterious pleiotropic effects due to their peripheral positions.
However, these rapidly evolving genes are likely regulated upstream by conserved
regulatory elements with highly pleiotropic functions across tissues and develop-
mental stages.

Epigenetics and Social Behavior

An additional level of gene regulation is at the epigenome. Epigenetic mechanisms

are powerful regulators of phenotype, and evidence is mounting that a particular
14 K.E. Ihle et al.

form of epigenetic regulation, DNA methylation, is critical for regulating social

phenotypes. DNA methylation can effect transcription of genes by affecting the
ability of the transcriptional machinery to reach a methylated segment of DNA
(reviewed in Dolezal et al. 2014). In a large comparative genomic study, Kapheim
and colleagues (2015) found that the number of genes predicted to be methylated
increases with social complexity in bees. In addition to species-level differences,
caste-based methylation patterns are predicted to influence both developmental and
behavioral programs. Developing honey bee queens and workers have different
methylation patterns, and genetic knockdown of methylation enzyme DNMT3
disturbed caste-based development (reviewed in Dolezal et al. 2014). Methylation
patterns also differ in the high and low pollen-hoarding strains of honey bee workers,
demonstrating that DNA methylation can be affected by selection over short-time
scales and suggesting that it may shape worker behavior. These finding make
understanding the role of epigenetics in social behavior an important direction for
future research.

Conclusion

The evolutionary development approach to the study of social behavior has helped
shape the large, recent advances in our understanding of the proximate mechanisms
that gave rise to social phenotypes. The theoretical foundations laid by the ground
plan hypotheses of social evolution have provided a useful framework on which to
build the current large-scale genomic studies. As more and more comparative and
population genomics studies are done, we are able to begin to identify how selection
shapes gene networks to shape complex novel behavioral phenotypes from compar-
atively simple substrates in conjunction with the evolution of novel taxonomically
restricted genes.

Cross-References

▶ Eco-Evo-Devo
▶ Extended Evolutionary Synthesis

References
Ament SA, Queenie WC, Marsha MW, Scott EN, Johnson SP, Sandra LR-Z, Leonard JF, Gene ER
(2011) “Mechanisms of stable lipid loss in a social insect.” Journal of Experimental Biology 214
(22):3808–3821
Amdam GV, Page RE Jr (2010) The developmental genetics and physiology of honeybee societies.
Anim Behav 79:973–980
Amdam GV, Norberg K, Fondrk MK, Page REJ (2004) Reproductive ground plan may mediate
colony-level selection effects on individual foraging behavior in honey bees. Proc Natl Acad Sci
U S A 101:11350–11355
Evo-Devo of Social Behavior 15

Carroll SB (2008) Evo-devo and an expanding evolutionary synthesis: a genetic theory of morpho-
logical evolution. Cell 134:25–36
Dolezal AG, Toth AL (2014) Honey bee sociogenomics: a genome-scale perspective on bee social
behavior and health. Apidologie 45:375–395
Dolezal AG, Flores KB, Traynor KS, Amdam GV (2014) The evolution and development of
Eusocial insect behavior. In: Steelman JT (ed) Advances in evolutionary developmental biology.
Wiley, New York
Gadau J, Fewell J, Wilson EO (2009) Organization of insect societies: from genome to
sociocomplexity. Harvard University Press, Cambridge
Hamilton WD (1964) Genetical evolution of social behaviour I. J Theor Biol 7:1–17
Hölldobler B, Wilson EO (2009) The superorganism: the beauty, elegance, and strangeness of insect
societies. WW Norton, New York
Jasper WC, Linksvayer TA, Atallah J, Friedman D, Chiu JC, Johnson BR (2015) Large-scale
coding sequence change underlies the evolution of postdevelopmental novelty in honey bees.
Mol Biol Evol 32:334–346
Kapheim KM, Pan H, Li C, Salzberg SL, Puiu D, Magoc T, Robertson HM, Hudson ME, Venkat A,
Fischman BJ (2015) Genomic signatures of evolutionary transitions from solitary to group
living. Science 348:1139–1142
Kenyon C, Campisi J, Wallace D (2004) From worms to mammals: the regulation of lifespan by
insulin/IGF-1 signaling. Mol Biol Cell 15:354A
Maynard Smith J, Szathmáry E (1997) The major transitions in evolution. Oxford University Press,
New York
Michener CD (1974) The social behavior of the bees. Belknap, Cambridge
Nowak MA, Tarnita CE, Wilson EO (2010) The evolution of eusociality. Nature 466:1057–1062
Page RE Jr (2013) The spirit of the hive. Harvard University Press, Cambridge
Roy-Zokan EM, Cunningham CB, Hebb LE, McKinney EC, Moore AJ (2015) Vitellogenin and
vitellogenin receptor gene expression is associated with male and female parenting in a
subsocial insect. Proc R Soc B Biol Sci 282:20150787
Toth AL, Robinson GE (2007) Evo-devo and the evolution of social behavior. Trends Genet
23:334–341
Toth AL, Varala K, Henshaw MT, Rodriguez-Zas SL, Hudson ME, Robinson GE (2010) Brain
transcriptomic analysis in paper wasps identifies genes associated with behaviour across social
insect lineages. Proc R Soc B Biol Sci 277:2139–2148
West-Eberhard MJ (1987) Flexible strategy and social evolution. In: Itô Y, Brown JL, Kikkawa J
(eds) Animal societies: theories and fact. Japan Scientific Societies Press, Tokyo
West-Eberhard MJ (1996) Wasp societies as microcosms for the study of development and
evolution. In: Turillazzi S, West-Eberhard MJ (eds) Natural history and evolution of paper-
wasps. Oxford University Press, New York
Wilson EO (1971) The insect societies. Harvard Belknap, Cambridge, MA
Winston ML (1987) The biology of the honey bee. Harvard University Press, Boston
Woodard SH, Fischman BJ, Venkat A, Hudson ME, Varala K, Cameron SA, Clark AG, Robinson
GE (2011) Genes involved in convergent evolution of eusociality in bees. Proc Natl Acad Sci U
S A 108:7472
Evo-Devo and Culture

Mathieu Charbonneau

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
An Evo-Devo of Culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Naturalizing Culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
How Biological Development Shapes Enculturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
How Enculturation Shapes Biological Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Cultural Evo-Devo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Abstract
What does Evo-devo offer for a better understanding of cultural evolution?
Cultural evolutionists with a biological bend typically focus on the relation
between genetic evolution and cultural change, a research program referred to
as gene-culture coevolution. Development of the human organism is usually left
unattended by cultural evolutionists, and so are the processes involved in the
production of cultural phenotypes. Moreover, Evo-devo research has yet to have
any marked impact on the social sciences. Examining how Evo-devo can con-
tribute to the study of cultural evolution means understanding how cultural
evolution and development shape one another. However, it is necessary to first
clarify just what sorts of developmental processes we are interested in. There are
two albeit not mutually exclusive candidate answers to this question. First, we can
be interested in the interactions between cultural evolution and biological devel-
opment – how does the development of human individuals and the cultural
evolutionary process shape one another? Alternatively, we can be interested in

M. Charbonneau (*)
Science Studies Program, Departments of Philosophy and Cognitive Science, Central European
University, Budapest, Hungary
e-mail: mathieu.charbonneau1@gmail.com

# Springer International Publishing Switzerland 2016 1

L. Nuño de la Rosa, G.B. M€uller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_47-1
2 M. Charbonneau

the interactions between cultural evolution and the generative mechanisms

involved in the production of cultural phenotypes. The objective of the present
discussion is to address both understandings of the relations between Evo-devo
and cultural evolution.

Keywords
Cultural evolution • Evo-devo • Cultural development • Enculturation •
Metaplasticity

Introduction

What does Evo-devo offer for a better understanding of cultural evolution? Follow-
ing M€ uller (2007), Evo-devo’s research agenda can be broadly characterized as the
solving of two key problems: how does evolutionary mechanisms generate and
modify organismal developmental processes and how does the structure of devel-
opmental processes shape back the patterns and processes of species evolution? In
order to understand either evolution or development, we need to understand how
they shape one another. Analogously, examining how Evo-devo can contribute to the
study of cultural evolution means understanding how cultural evolution and devel-
opment shape one another. However, it is necessary to first clarify just what sorts of
developmental processes we are interested in (Mesoudi et al. 2006). There are two
albeit not mutually exclusive candidate answers to this question. First, we can be
interested in the interactions between cultural evolution and biological development
– how does the development of human individuals and the cultural evolutionary
process shape one another? Alternatively, we can be interested in the interactions
between cultural evolution and the generative mechanisms involved in the produc-
tion of cultural phenotypes. We will refer to these two projects as an Evo-devo of
culture and a cultural Evo-devo, respectively.
The objective of the present discussion is to address both understandings of the
relations between Evo-devo and cultural evolution. However, the reader should be
aware from the onset that there is no such thing today as an Evo-devo of culture or
a cultural Evo-devo. Cultural evolutionists with a biological bend typically focus on
the relation between genetic evolution and cultural change, a research program
referred to as gene-culture coevolution (Boyd & Richerson 1985). The development
of the human organism is usually left unadressed by cultural evolutionists and so are
the processes involved in the production of cultural phenotypes (Charbonneau
2015a; Wimsatt 1999). Moreover, Evo-devo research has yet to have any marked
impact on the social sciences. However, there is an abundance of existing research
that can bridge these gaps, spanning from developmental psychology, ethnography,
and/or the neurosciences. Concepts from Evo-devo also promise to offer important
insights for the study of cultural processes under a novel, insightful light. So while
this entry does not aim at offering a bird’s eye-view of an actual research program –
there is no such research program to begin with – it will identify key intersections
Evo-Devo and Culture 3

between the Evo-devo framework and its potentially relevant fields of application in
the study of cultural evolution.

An Evo-Devo of Culture

Naturalizing Culture

The human capacity to transmit, maintain, and incrementally modify cultural

traditions across generations has had major impacts on the survival and natural
history of the human species. Think of the many techniques for producing, using,
and improving tools and technologies or of the rich variety of belief systems
observed in extant and lost civilizations. A great many of human behaviors are
neither learned directly from the environment nor are they the product of genetic
inheritance. Rather, such human behavioral phenotypes are acquired and
maintained from one generation to the next by human individuals learning from
one another. In analogy to genetic transmission, which sustains a biological
evolutionary process, social learning – the ability to learn from others – sustains
a cultural evolutionary process with important impacts on the natural history of our
species, a cultural evolutionary process likely intertwined with the biological
evolution of our own species (Boyd and Richerson 1985). Accordingly, cultural
evolutionists understand a socially learned behavior as a cultural trait when the
mental representations (such as beliefs, norms, etc.) or information involved in the
production of the behavior has been acquired from others and is widely distributed
in a population. Teaching, imitation, apprenticeship, etc., are all key social learning
processes, and species devoid of such social learning capabilities will fail to sustain
and improve any cultural tradition.
The research program adopted by contemporary cultural evolutionists is a natu-
ralistic one. Adopting a naturalistic approach to the study of human culture means
first and foremost understanding social and cultural processes and phenomena in
continuity with processes and phenomena of other natural domains, such as cogni-
tive processes, mental states, and biological processes. Accordingly, in addition to
the many social sciences specifically devoted to the study of human cultures, such as
anthropology, archaeology, and history, the scientific study of culture has now grown
into a vast interdisciplinary field involving evolutionary and behavioral biology, the
neurosciences, cognitive psychology, and biological anthropology. This does not
mean that cultural phenomena are nothing more than psychological or biological
processes. Rather, the naturalistic program understands cultural processes as material
processes interacting with and partially composed of biological, psychological, and
social processes. (Sperber 1996)
It is generally agreed upon that cultural transmission shares many key features
with genetic inheritance, features enabling social learning to support an evolution-
ary process of human traditions. As far as it insures the transmission of behaviors
across generations and participates in sustaining a cultural evolutionary process,
4 M. Charbonneau

social learning serves as a nongenetic mechanism of inheritance, one with its very
own (nongenetic) channels of transmission (Boyd and Richerson 1985; Jablonka
and Lamb 2005). Accordingly, cultural evolutionists typically orient their research
towards the transmission patterns of cultural traits and their distribution in human
populations. On the basis of the similarities identified between cultural and genetic
inheritance, many cultural evolutionists have adopted the modeling tools and
strategies of population genetics in order to study the population-level effects of
individual episodes of social learning (Boyd and Richerson 1985). However, there
are enough differences between genetic and cultural transmission so that models of
population genetics cannot be straightforwardly transferred to the study of cultural
change. Whereas cultural evolutionists have borrowed modeling methods and
assumptions from population genetics, they have spent a great deal of effort
specifying how the modeling techniques should be adapted to the idiosyncrasies
of cultural transmission and evolution. Central to cultural evolutionary theory then
are questions such as who learns what from whom, whether cultural transmission is
a high-fidelity replicative process or not (i.e., the rate of cultural “mutation”), and
what sorts of biases are involved in the transmission of culture and how these shape
the distribution of cultural variation. The disanalogies picked-up by cultural
evolutionists thus mainly concern differences in the network and channels of
genetic and cultural information transmission. Cultural evolutionary models con-
sequently involve horizontal and oblique transmission – transmission among peers
of the same generation and to unrelated individuals of the next generation, respec-
tively – and learning biases, i.e., preferences to learn some behaviors instead of
others on the basis of the preferred teachers, of the behavior’s outcomes, how
frequent the behavior is in the population, etc. Consequently, the differences also
justify a nonreductionist approach to cultural evolution as genes and cultures can
evolve relatively independently from one another (contra Lumsden and Wilson
1981).
However, there is an important set of disanalogies that cultural evolutionists
rarely address, namely, the differences between the structure of genetic inheritance
and that of the enculturation process, i.e. the process by which an individual acquires
the typical cultural repertoire of its group. Indeed, one important difference between
genetic inheritance and the enculturation process concerns the specific moment in
the organism’s life-cycle and the duration of the acquisition of genetic and cultural
information. Simply put, we inherit the whole of ours genes at the moment of our
parents’ reproduction. In contrast, we inherit our local culture throughout our
lifetime and do so in a piecemeal and sequential manner (see Fig. 1). Building up
one’s repertoire of cultural traits is a life-long process, taking place not before the
development of the human organism – and thus, somewhat in isolation of it, as it is
with genes
but during development. An individual’s enculturation is thus sensitive
to its biological development. Moreover, a culture’s specific enculturation process is
structured, and its structure is itself socially transmitted and open to evolutionary
change. This seemingly banal fact has important consequences for the evolution and
the study of human cultures.
Evo-Devo and Culture 5

Fig. 1 Whereas organisms typically inherit their genome (G) at the moment of reproduction
(Gx0 !Gx+1) and carry that same genome throughout their lifetime (Gx!Gx0 ), an individual’s
cultural repertoire is constructed sequentially and in a piecemeal fashion (Ta) throughout its life
time (T2) (Adapted from Durham 1991, p. 186)

How Biological Development Shapes Enculturation

The enculturation process is sensitive to the cognitive and morphological develop-

ment of the human individual. Perhaps most striking is the sensitivity to morpholog-
ical development. Cultures adopting a sexual division of labor are nearly ubiquitous.
Not only do females and males generally serve different ecological and social roles
but more importantly they typically learn and transmit different cultural behaviors
(e.g., foraging behaviors, clothing habits, etc.). Such division is based on the recog-
nition by the group that one individual is of a certain gender, a recognition not based
on observing the chromosomes individuals possess but on the perceived phenotype
of the individual (e.g., recognition of sexual attributes), and often the learned and
transmitted behaviors will also depend on the age of the individual (e.g., sexual
maturity). The criteria used to distinguish genres vary from one culture to the next.
Cognitive and brain development of the human individual also constrains what
can be learned at specific ages. An individual’s first language is learned during
childhood with little difficulty, whereas second languages learned at a later age will
typically require much more cognitive efforts on the part of the learner. Additionally,
more complex behaviors and abstract knowledge may only be learned when the
proper cognitive capacities and motor skills have developed (Roux and Bril 2005).
Just as with morphological development, a community may also adopt different
norms for measuring cognitive maturity in order to decide if an individual will be
allowed to learn some special skills (Rogoff 2003). Some behaviors and techniques
also require a long time to master, whereas genes are all acquired quickly and at the
same time. This means that one’s developing expertise can interact and possibly be
scaffolded with what other knowledge or skill one is learning at the same time.
How a community attests that the learner is ready for learning more about the
techniques and who will take over such learning is also variable from one culture to
the next (Ruddle and Chesterfield 1977). For instance, this can mean that the network
of interactions of an individual within the community may also vary according to the
6 M. Charbonneau

perceived maturity and gender of the developing individual (Rogoff 2003). In

Western societies, youngsters will typically learn mostly from their parents but as
individuals gain in age, they will enter the schooling system where oblique transmis-
sion is the rule. In other societies, youngsters may directly learn mostly from their
aunts and uncles as parenting tasks are shared across the kinship and then as they get
older join groups of other kids and learn mostly horizontally (e.g., through play).
Human populations vary in what individuals learn, at what age they learn it, and
from whom they learn from. The cultural relativity of the enculturation process makes
the use of population genetics tools problematic, as population genetics models
typically assume a stable life-cycle, one that is generalizable across the species. In
contrast, the variability of the structure of the enculturation process implies that
models in cultural evolutionary theory may not assume some general cultural life-
cycle as there may not be any stable, cross-cultural patterns of enculturation (Wimsatt
1999). Only by integrating the influence of biological development on cultural
transmission and its impacts on structuring the enculturation process will cultural
evolutionists be capable of articulating a general theory of cultural change.
The varying enculturation patterns are themselves maintained and inherited
through social learning, with neither genetic nor environmental factors being capable
of accounting alone for the patterns’ diversity and stability. For instance, the Greco-
Roman education structure strikingly exemplifies the inheritance of the enculturation
process, with the Romans intentionally copying (with some adjustments) the edu-
cation structure of Ancient Greece. In contrast, although they may lack any school-
ing institutions, traditional societies nevertheless exhibit highly structured, lasting
enculturation patterns (Rogoff 2003; Ruddle and Chesterfield 1977). There are other
studies in the developmental psychology, ethnography, and comparative pedagogy
literatures that deal with the transmission of a culture’s typical enculturation process.
Unfortunately, there is little if any work on these topics that expressly adopt a
cultural evolutionary framework.
Taking seriously the relation between individual development and cultural inher-
itance implies collecting data about how different cultures vary in the specifics of
their enculturation process and examining how these differences enable and con-
strain the general evolution of cultural traditions. Moreover, this means to pay closer
examination of how specific enculturation structures come about and how such
structures can undergo evolutionary change. For instance, one key structuring
constraint of many cultural traditions resides in the fact that for many cultural
phenotypes, to acquire the trait one must already have learned some other cultural
traits beforehand. Many complex cultural traits are in fact composed of simpler ones.
So, for instance, in order to learn calculus, you already need to have somewhat
mastered algebra, which in turns relies on you knowing the basics of arithmetic.
These logical constitutive dependencies translate, in terms of the enculturation
process, as a strict sequence of learning which must be respected if the enculturation
process is to successfully allow the learning and further transmitting of these cultural
traits (Wimsatt 1999). Moreover, each step in the sequential acquisition of a complex
trait may also depend on the specifics of the individual’s cognitive maturity, where
the cognitive capacities required to learn and master each trait in the sequence may
Evo-Devo and Culture 7

not develop synchronously. (See Enquist et al. 2011 for a modeling effort of the
sequentiality of the enculturation process).
The dependence relationships between cultural traits may have different impacts
on the cultural evolutionary process. Wimsatt (1999) argues that enculturation,
analogously to the genome, is subject to generative entrenchment. What one learns
earlier in its life-cycle is less prone to change as any change risks having deleterious
cascading effects on the acquisition of other traits depending on the earlier ones. So
if we change the rules of arithmetic, these may not be coherent anymore with algebra
or with calculus, leading to the failure of further learning the latter. Thus the
stabilization and preservation of arithmetic is necessary for the successful transmis-
sion of algebra and calculus. Clarifying exactly at what level the entrenchment of
cultural traits are located will prove to be an important part of the study of encul-
turation. Whereas possessing a language may be a necessary condition for both the
invention and learning of arithmetic, acquiring any specific natural language is not.
Again, contrary to gene inheritance, where the genome of the organism is typically
inherited as a whole during reproduction, enculturation is a piecemeal, culturally
variable process. This means that there are likely more chances that “pleiotropic”
effects among cultural traits will agglomerate into relatively independent package of
cultural traits (e.g.,. learning mathematics vs. learning fishing) rather than being
distributed on the overall cultural repertoire of a population. A clear theory of the
packaging of cultural traits and of their evolution together remains to be formulated.

How Enculturation Shapes Biological Development

Culture is a special form of phenotypic plasticity. Consider, for instance, the case of
Padaung women, known through their touristic name of “giraffe-necked women,” as
a case of culture interacting with the developmental plasticity of the human organism
(Fig. 2). The Padaung tradition of women wearing heavy brass neck-rings has effects
spanning on many different levels of the human organism’s phenotypic plasticity.
The observed depression of shoulder girdle in Padaung women is due to the heavy
brass rings they traditionally wear around their neck, thus giving the optical illusion
of possessing a longer neck. These morphological changes also result in physiolog-
ical problems such as increased blood pressure, and the older Padaung further risk to
break their necks if the rings were to be removed, making them dependent on an
artificial “exoskeleton.” Perhaps most striking are the morphological effects, but for
the cultural evolutionist, it is the brain’s plastic capacity to learn a great variety of
behaviors that is central. As a form of phenotypic plasticity, culture can be under-
stood as a mechanism of phenotypic response to the behavioral displays of others.
Social learning then is the capacity to reproduce behavioral phenotypes similar to
those observed in other members of its population.
Central to culture, then, is the capacity of the human brain to change in response
to the local culture. Not only do we learn from one another, but the structural and
functional organizations of our brain are greatly influenced by what we learn and the
age at which we learn it. For instance, neuroimaging studies have shown cortical
8 M. Charbonneau

Fig. 2 Social transmission as a multistep process. A demonstrator’s mental representation Mx is

used to produce (production) some cultural phenotype Px. For a tradition to be sustained, the learner
acquires a similar mental representation Mx+1 by observing the demonstrator’s cultural phenotype
Px, and so on and so forth (Adapted from Charbonneau 2015a, p. 531)

reorganization in professional musical skill training which has important impacts on

tactile acuity. Moreover, cognitive changes induced by practice are sensitive to the
age at which the skill is learned, with early learning leading to better expertise and
with an increased ability to learn new tasks (Malafouris 2013). Thus different
enculturation regimes may lead to cognitive differences between cultures without
involving neither genetic nor noncultural environmental differences. In other words,
the brain’s plasticity exploited by culture can be altered with by what one learns. This
plasticity of brain plasticity is generally referred to by the term “metaplasticity,” a
term coined to refer to changes in synaptic plasticity induced by synaptic activity
(Abraham and Bear 1996). In the context of an Evo-devo of culture, we can
understand cultural metaplasticity as the capacity of culture to change the social
learning processes and the cognitive capacities of the human individual. In other
words, the metaplastic capabilities of the human brain makes it a cultural artifact par
excellence (Mithen and Parsons 2008).
The study of the cultural metaplasticity of the human brain promises important
consequences for the study of the relations between cultural evolution and biological
development. Culture may not only serve as a behavioral inheritance system
exploiting the brain’s plasticity for memorization and behavior acquisition. It may
also prove to be an important source of cognitive change altogether. In turn,
culturally induced cognitive change can lead to cultural changes that would not
have been possible otherwise (Malafouris 2009, 2010). For instance, increased
tactile acuity in stone tool manufacture may lead individuals not only to learn
novel, more demanding techniques of stone tool production, but also to discover
novel sophisticated techniques altogether (Roux and Bril 2005). Cognitive change
induced by cultural metaplasticity may lead to new possibilities of cultural innova-
tions, with the novel innovations leading to novel cognitive capabilities, and so on
and so forth. As individuals’ cognitive capacities are partly shaped by culture, these
can in turn impact further cultural change, creating an historical process of brain-
culture coevolution that does not involve any genetic change. In other words, human
cognition may well have, in addition to a phylogeny, a cultural history. Little
research has directly addressed the coevolution of cognition and culture from a
Evo-Devo and Culture 9

cultural evolutionary and historical – rather than a phylogenetic – perspective.

However, some work in neuroarchaeology is addressing how the material culture
has co-developed with cognitive changes in early societies (Malafouris 2009, 2010).
Some developmental psychologists have also addressed how social learning capac-
ities are themselves the result of cultural evolution (Heyes 2012).

Cultural Evo-Devo

Cultural traditions are persisting causal chains of mental representations and public
displays, such as behaviors (Boyd and Richerson 1985; Sperber 1996). The first link
in these chains consists in producing some observable behavior from some mental
representation. Whereas we may not directly access the mental representations of
others, by effectively producing a learned behavior, an individual’s private knowl-
edge becomes publicly available for others to learn from. The second step consists in
another individual perceiving the behavior and acquiring from it (or multiple repe-
titions of it) its very own private mental representation of the behavior. In future
instances, the social learner will be able to reproduce the behavioral phenotype,
which in turn will make it publicly available for another individual to acquire it, thus
sustaining a cultural tradition (Fig. 2).
Cultural evolutionists typically emphasize the acquisition phase of social trans-
mission, studying primarily the maintenance and transformation of the transmitted
information. This practice can be illustrated by the use of terms like “cultural
variants” or “cultural traits,” referring indiscriminately to variant mental representa-
tions (e.g., beliefs, preferences, etc.) or to variant cultural phenotypes (e.g., practices,
shape of artifacts, etc.), or both. Consequently, the production phase is generally
black-boxed and the specific generative processes involved in the production of
cultural phenotypes abstracted away. It is at this point that borrowing concepts from
the Evo-devo framework promises a better understanding of the interplay between
the production of cultural phenotypes and their evolution in what is sometimes
referred to as a cultural Evo-devo (see Mesoudi et al. (2006, p. 367)).
There has been little work investigating this avenue mainly because there lacks a
clear understanding of just what cultural development – in analogy to biological
development – consists of (Mesoudi et al. 2006, p. 367). One possibility is to
understand how variation in the socially transmitted mental representations maps
onto variation in the cultural phenotypes they produce as a cultural analog to the
genotype-phenotype map. Genes do not specify development such that variation of
the phenotypes of organisms reduces to variation in their genetic material. Rather,
genes and developmental processes interact with one another in complex ways such
that the mapping between genotype and phenotype becomes itself a complex affair
(Alberch 1991). In analogy, variation in cultural phenotypes would not reduce to
variation in mental representations as the specific processes involved in producing
the cultural phenotypes may shape the latter’s variation in complex ways. A cultural
Evo-devo would then consist in studying these complexities.
10 M. Charbonneau

Most cultural evolutionists are in fact skeptic about using analogies with biological
processes. However, perhaps there is no need to find a strong cultural analog to
biological development. Indeed, if social learning does in fact serve as a nongenetic
inheritance system, the socially transmitted mental representations taking part in
cultural traditions can be conceptualized as generative factors participating in the
production of behavioral phenotypes, not in analogy to genes, but as alternative
developmental resources to genetic information. A similar logic would apply to other
nongenetic inheritance systems (Jablonka and Lamb 2005). In the context of cultural
evolution, we can thus understand the production phase as a form of cultural develop-
ment, i.e., as the processes involved in the production of cultural phenotypes for
inherited developmental resources, here socially acquired mental representations. A
cultural Evo-devo, then, would examine how the generative processes involved in the
production of cultural phenotypes interact, shape, and are shaped by cultural evolution.
The first step in developing a cultural Evo-devo should be to clarify just what the
generative processes are made of. Following Mesoudi and O’Brien (2008), we can
understand the structure of the generative processes involved in the production of
cultural phenotypes through the concept of a cultural recipe. A cultural recipe is a
hierarchically organized set of actions and decisions leading to the satisfaction of a
specific, intended goal. The hierarchical structure of recipes can be decomposed into
subassemblies of actions serving some subgoal that must be satisfied on the road to
the intended end-product, i.e., the cultural phenotype. A subgoal consists of a
measure of what conditions need to be satisfied and what to do next if the conditions
are perceived as being satisfied and what to do when they are not. These subgoals
can also be nested as intermediary steps in the realization of some other subgoals,
thus generating a potentially complex structure of dependencies between action and
decisions assemblies. Ultimately, all subgoals are ruled by a single master goal, that
of the final intended end-result of the recipe. The hierarchical structure of recipes is
typically depicted as a tree-like structure (see Fig. 3 for an example).
Cultural recipes are themselves transmitted from one generation to the next, and
can vary, which confers them the capacity to evolve. Adequately, adopting explan-
atory concepts and tools from Evo-devo will thus highly depend on whether the
structure of cultural recipes can vary in ways similar to the development of an
organism from its genetic material and environmental context of development.
There have been some suggestions that the production of cultural phenotypes and
the evolution of recipes are fit to adapt parts of the conceptual framework from
Evo-devo. In the remainder, we will discuss two of these – cultural modularity and
that of a cultural genotype-phenotype map – and some of their consequences – such
as cultural evolvability and cultural developmental constraints.
Mesoudi and O’Brien (2008) argue that complex recipes, ones possessing many
levels of actions and decisions subassemblies, are likely to be decomposable into
cultural modules given that recipe subassemblies tend to be more functionally
integrated with one another than they are with the whole recipe. In other words,
similarly to a modular genetic architecture, complex cultural recipes would be nearly
decomposable. Moreover, such functionally modular subassemblies will tend to be
transmitted as units as they can be learned as whole and relatively independently
Evo-Devo and Culture 11

Fig. 3 The hierarchical structure of flake detachment. Early prehistoric stone tools were produced
by detaching flakes off a core stone by hitting it with a hammerstone. The flake detachment behavior
is composed of two main subbehaviors that of selecting a target on the core and that of percussion.
Selecting a target consists in choosing a specific point on a core to hit with the hammerstone.
Percussion consists in appropriately positioning the core, grasping the hammerstone, and striking
the core on its target platform. Specific actions are represented at the lower end of the tree. Decisions
are indicated as nodes (Adapted from Stout 2011, p. 1052)

from one another or from the complex recipes in which they figure. Mesoudi and
O’Brien (2008) models also suggest that the more modular cultural recipes are, the
higher their chances of being transmitted. Modular recipes thus would have greater
evolvability than more holistic (or less-modular) ones, as cultural modules, once
learned, can be used in many different recipes (see also Charbonneau (forthcoming)).
Mesoudi and O’Brien (2008) offers a formal treatment of the structure of cultural
recipes and do not address the different material and productive constraints involved
in the cultural development process (Charbonneau forthcoming). Producing cultural
phenotypes is a causal story starting with an individual’s mental representations and
ending in the public display of a specific cultural phenotype. This means that the
production phase depends on cognitive, bodily, and ecological processes that are not
necessarily involved in the acquisition phase. The production of cultural phenotypes
may thus have its own enabling and constraining effects on social transmission and
consequently on the evolution of cultures. The study of the generative mechanisms
involved in the production of cultural phenotypes will thus be a complex endeavor,
requiring the cultural evolutionist to address multiple mechanisms at different levels.
Charbonneau (2015a) identifies four of these levels:

1. The cognitive processes and biases participating in the generation of public

displays from mental representations (e.g., decision-making processes, mental
imagery, motor control, etc.)
2. The external actions recruited in the production of the public displays (e.g.,
locomotion, prehension, manipulation, pronunciation, etc.), including the
affordances and constraints set by the particular body of the demonstrator (e.g.,
opposable thumb, flexibility, dexterity, body size, mass, etc.)
12 M. Charbonneau

3. The specific tools and materials used to produce the public displays (if any)
4. The ecological processes engaged in the production of the public displays (e.g.,
chemical reactions, percussion effects, sound-wave propagation, etc.)

Charbonneau (2015a) argues that when these generative factors are taken into
account, many assumptions typically adopted by cultural evolutionists may prove
wrong. Once such assumption consist in the metrics used to assess gradual cultural
evolution. Cultural evolutionists typically assume that errors in social transmission
and even intentional transformations of cultural traditions tend to produce relatively
similar cultural phenotypes. In other words, small changes in the transmitted infor-
mation will result in small variations in the cultural phenotype, leading to a process
of gradual cultural evolution that can be studied mainly at the level of the informa-
tion being transmitted. However, taking into account both the complex structure of
cultural recipes and the material processes involved in the production of cultural
phenotypes shows that small modifications in the recipes can lead to large changes in
the phenotype. Inversely, small changes in phenotypes may in fact depend on large
changes in the structure of the recipes. For instance, Charbonneau (2015a) points out
that in order to augment the width of lithic blades from 2.4 cm to 2.6 cm, stone
knappers had to pass from a pressure-flaking technique to the use of the lever as only
the latter could exert enough pressure to detach the wider blades. Whereas the blades
are very similar (they have 0.2 cm of width difference), the underlying techniques
and the set of behaviors and artifacts they depend on are radically different.
A closer look at how the productive processes constrain the variation of cultural
phenotype may reveal further complicated cases of cultural genotype-phenotype
mapping (or mental representation-public display mapping), challenging the typical
cultural evolutionist's assumption of an isomorphic mapping. Moreover, investigat-
ing such mapping and the structure of cultural variational spaces can reveal which
cultural forms of cultural phenotypes are possible and which ones are not
(Charbonneau 2015b), suggesting that observed convergence in form in different
cultures may be the result not so much of similar adaptations but of generative
constraints on the development of cultural phenotypes.
Further work is required to make a more serious case for both an Evo-devo of
culture and a cultural Evo-devo. Research in both directions is only at an embryonic
stage. However, in the future we can expect further advances on the impact of
enculturation and metaplasticity on cultural evolution, and also on issues pertaining
to cultural modularity, evolvability, and constraints of cultural development. An
important part of such work will consist in addressing what nonevolutionary social
sciences already have to say about the cultural process and integrate such work into
an evolutionary framework before any useful contribution from Evo-devo can be
productively harnessed.

Cross-References

▶ Evo-Devo and Cognitive Science

Evo-Devo and Culture 13

▶ Evo-Devo of Language and Cognition

▶ Evo-Devo of Social Behavior
▶ Modularity in Evo-Devo

References
Abraham WC, Bear MF (1996) Metaplasticity: the plasticity of synaptic plasticity. Trends Neurosci
19:126–130
Alberch P (1991) From genes to phenotype: dynamical systems and evolvability. Genetica 84:5–11
Boyd R, Richerson PJ (1985) Culture and the evolutionary process. University of Chicago Press,
Chicago
Charbonneau M (2015a) All innovations are equal, but some more than others: (Re)integrating
modification processes to the origins of cumulative culture. Biol Theory 10(4):322–335
Charbonneau M (2015b) Mapping complex social transmission: technical constraints on the
evolution cultures. Biol Philos 30:527–546
Charbonneau M (forthcoming) Modularity and Recombination in Technological Evolution in.
Philosophy & Technology
Durham WH (1991) Coevolution: genes, culture, and human diversity. Stanford University Press,
Stanford
Enquist M, Ghirlanda S, Eriksson K (2011) Modelling the evolution and diversity of cumulative
culture. Philos Trans R Soc B 366:412–423
Heyes CM (2012) Grist and mills: on the cultural origins of cultural learning. Philos Trans R Soc B
367:2181–2191
Jablonka E, Lamb MJ (2005) Evolution in four dimensions: genetic, epigenetic, behavioral, and
symbolic variation in the history of life. MIT Press, Cambridge, MA
Lumsden CJ, Wilson EO (1981) Genes, mind, and culture. Harvard University Press, Cambridge, MA
Malafouris L (2009) “Neuroarchaeology”: exploring the links between neural and cutlural plastic-
ity. In: Chiao JY (ed) Cultural neuroscience: cultural influences on brain function. Elsevier,
New York, pp 253–261
Malafouris L (2010) Metaplasticity and the human becoming: principles of neuroarchaeology.
J Anthropol Sci 88:49–72
Malafouris L (2013) How things shape the mind: a theory of material engagement. MIT Press,
Cambridge, MA
Mesoudi A, O’Brien MJ (2008) The learning and transmission of hierarchical cultural recipes. Biol
Theory 3:63–72
Mesoudi A, Whiten A, Laland KN (2006) Towards a unified science of cultural evolution. Behav
Brain Sci 29:329–383
Mithen S, Parsons L (2008) The brain as a cultural artefact. Camb Archaeol J 18(3):415–422
M€uller GB (2007) Six Memos for Evo-devo. In: Laubichler MD, Maienschein J (eds) From
embriology to Evo-devo. MIT Press, Cambridge, MA, pp 499–524
Rogoff B (2003) The cultural nature of human development. Oxford University Press, Oxford
Roux V, Bril B (eds) (2005) Stone knapping: the necessary conditions for a uniquely hominin
behaviour. McDonald Institute for Archaeological Research, Cambridge
Ruddle K, Chesterfield R (1977) Education for traditional food procurement in the Orinoco Delta.
University of California Press, Berkeley
Sperber D (1996) Explaining culture: a naturalistic approach. Blackwell Publishers, Oxford
Stout D (2011) Stone toolmaking and the evolution of human culture and cognition. Philos Trans R
Soc B 366:1050–1059
Wimsatt WC (1999) Genes, memes, and cultural heredity. Biol Philos 14:279–310
Pleiotropy and Its Evolution: Connecting
Evo-Devo and Population Genetics

Mihaela Pavličev

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
General Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Challenges in Determining Pleiotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
What Is a Trait? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
The Many Mechanisms of Pleiotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
How Many Traits? Pleiotropy and the Evolution of Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Phenotypic Traits Versus Fitness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Evolution of Pleiotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Future Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Abstract
Pleiotropy, the involvement of a gene in development and variation of multiple
traits, is a concept considerable appreciation in developmentally as well as
statistically oriented fields of evolutionary biology. Here I argue that this feature
makes pleiotropy a particularly suitable guiding topic for connecting the two
branches of evolutionary biology, and integrate evolutionary descriptions from
the molecular, over developmental, to populational mechanisms. I first describe
some of the challenges in defining pleiotropy, and then focus on evolution of
pleiotropic constraints, which I suggest is the centerpiece of the connection.
Finally, I address some of the future challenges.

M. Pavličev (*)
University of Cincinnati, Cincinnati, OH, USA
Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, USA
e-mail: Mihaela.Pavlicev@cchmc.org

# Springer International Publishing Switzerland 2016 1

L. Nuño de la Rosa, G.B. Müller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_52-1
2 M. Pavličev

Keywords
Pleiotropy • Modularity • Epistasis • Evolvability

Introduction

Pleiotropy is a concept deeply rooted both in mechanistically oriented developmen-

tal genetics, as well as in statistical approaches of population genetics. Pleiotropy not
only reflects developmental genetic structure but also importantly mediates the effect
of developmental structure on the future response to selection. The question to what
extent pleiotropy constrains long-term response to selection is particularly important
for the role of pleiotropy in evo-devo. In the past, the focus on microevolution
invited model assumptions, which limited the ability to study long-term dynamics of
pleiotropy and of the genotype-to-phenotype map in general (such as lack of context
dependency of genetic effects; see chapter on “▶ The Role of Epistasis in
Evo-Devo”). Here, some of the general challenges of studying pleiotropy will be
discussed first, followed by the focus on the ability of pleiotropy and associated
constraint, to evolve. By having roots in mechanistic and variational traditions, with
much accumulated knowledge in both, pleiotropy is a particularly central concept
when unifying evolutionary biology.

General Definition

In most general terms, pleiotropy refers to a situation in which mutation in a single

gene causes phenotypic changes in multiple distinct traits. These can be large
phenotypic changes, often deleterious in individuals, such as those observed in
pathological syndromes (e.g., phenylketonuria) in which pleiotropic effects were
first recognized, or they can be minute effects on multiple traits which, when studied
across a population of individuals, manifest as covariation between traits. The
pattern of mutational effects on the phenotypic traits can reveal the developmental
genetic structure, as simultaneous changes in different traits suggest that a portion of
the genetic-developmental basis between the traits is shared.
Even though pleiotropic effects were recognized earlier in medical syndromes
and correlated effects have been acknowledged by Mendel and Darwin, the term
pleiotropy was first coined by developmental geneticist Ludwig Plate in 1910,
whose interest was in the intersection of evolution and development. Many of the
early studies of pleiotropy attempted to explain the physiological basis of pleiotropy
(Stearns 2010). Yet, it is the work on the phenotypic correlations and their conse-
quences for evolutionary change that arguably provided the strongest impulse for the
surge of studies of pleiotropy in evolutionary biology, such as culminated in
morphological integration (see the chapter on “▶ Morphological Integration”). The
study of phenotypic correlations and covariances motivated the quantitative genetic
study of underlying genetic covariances and their evolutionary consequences (Lande
Pleiotropy and Its Evolution: Connecting Evo-Devo and Population Genetics 3

Trait 2
Trait 2

Trait 1 Trait 1

Fig. 1 The distribution of variation in two traits in (a) the presence of strong pleiotropy and (b) in
the absence of pleiotropy when the variation in the two traits is independent from each other. In the
first case, the direction of most variation is shown to be the direction along which both traits change
simultaneously. In (b), any phenotypic direction has equal amount of variation

1979), effects of pleiotropy on standing variation (Turelli 1985), as well as the

relationship between the genetic and phenotypic correlations (Cheverud 1988).
The main evolutionary consequence of pleiotropy is its directing of genetic
variation along particular phenotypic directions and preventing others (Fig. 1).
Evolutionary change is biased toward those phenotypic directions, which exhibit
the greatest heritable phenotypic variance (Schluter 1996). With respect to any
particular trait, variation shared with other traits can either prevent or diminish the
response to selection, or it can cause correlated response in other traits. This is
because when mutation at a locus affects more than one trait, but only the change in a
single trait is advantageous, it depends on the selection regime acting on all affected
traits, whether such variation can be selected upon. When such correlated effects are
mainly deleterious, response in a population will likely not realize, in spite of
apparently sufficient variation in the selected trait (Hansen et al. 2003). This situation
is referred to as evolutionary or pleiotropic constraint (see the chapter on “▶ Devel-
opmental Constraint”). If the correlated mutational effects on other traits are selec-
tively neutral, the advantageous mutation can spread in a population; however, it will
simultaneously change traits that have not been directly selected. One of the most
prominent examples using such pleiotropic side effects is the model by George
C. Williams (1957) that senescence is due to antagonistic pleiotropy. According to
this model, the beneficial effects on individuals early in life that contribute to
individual’s reproductive success and are thus selected for, are associated with the
negative pleiotropic effect on late life stages (i.e., senescence). As there is a strong
selection on reproductive life period, but not on life period after reproduction, such
effects persist in the population (the gain of reproductive fitness early in life out-
weighs its cost). More recently, deleterious side effects have been proposed to
themselves become a source of selective pressure to become compensated by
secondary mutations. A good example here are side effects of mutations conferring
drug resistance in some conditions but are fitness reducing in others (Lenski 1988
and many later). This model suggests enrichment of compensatory mutations in
4 M. Pavličev

traits, which are themselves under stabilizing selection but share genes with traits
under directional selection. The compensatory mutations allow the traits’ basis to
avoid the correlated response, hence, not to adopt a new phenotype, but to maintain a
phenotype. Such phenomenon could underlie developmental systems drift (True and
Haag 2001) – a situation in which developmental processes underlying the same trait
in different taxa are highly modified and appear as if they are drifting while the end
phenotype is under stabilizing selection (Johnson and Porter 2007; Pavlicev and
Wagner 2012) (see the chapter on “▶ Developmental Drift”).

Challenges in Determining Pleiotropy

Many aspects of pleiotropy that are important for its influence on short- and long-
term evolutionary change are being revisited. In the following sections, I will first
explain some of the challenges in defining pleiotropy; next, I will focus on the
evolution of pleiotropy itself; and finally, I will conclude with the open challenges
that need to be addressed in the future to gain full appreciation of the role of
pleiotropy in evolutionary change.

What Is a Trait?

One of the most fundamental problems in determining pleiotropy of a mutation is in

defining the traits, in order to count them (Zhang and Wagner 2013). Whereas
infinitely many measurements can be taken, not every measurement that can be
taken on an organism describes a biologically distinct trait. Different approaches
have been taken to tackle this problem. One is based on a biological process, often
growth. Separate processes are thereby considered to characterize separate traits. For
example, the long bones (the bones in arms and legs) have a clear direction of growth
in the length, due to the growth plates from which the cell proliferation occurs. This
makes long bone length a meaningful biological trait. But such trait choice may be
harder when we talk about the skull, where traits are highly interdependent and
biological dimensions are harder to define. In the skull, traits are often defined as the
directions of the variation, using ordination techniques (e.g., such as PCA) to
determine the number of independent directions in space that together capture
most of the variation between the individuals. Single directions thereby consist of
contribution of multiple measurements. While this approach is extremely useful for
the analysis of dimensionality of phenotypic variation, i.e., in how many indepen-
dent directions a population phenotype may respond to selection, such statistical
dimensions (in particular later PCs) may encompass more than a single biological
trait and may therefore cause difficulties when such phenotypic variation is being
associated with underlying genetic sequence variation (Cheverud 2007). The
approach thus has advantages but also disadvantages in particular when used to
define pleiotropy. For example, the highly correlated traits such as fore and hind
limbs, may appear as a single trait axis in some species, and the interesting biological
Pleiotropy and Its Evolution: Connecting Evo-Devo and Population Genetics 5

problem of their overcoming pleiotropic constraint and obtain the ability to produce
independent variation, would be lost. The field has not reached a general consensus
on the definition of trait that would be equally satisfactory in all its uses.

The Many Mechanisms of Pleiotropy

Pleiotropy describes a variational phenomenon, in which differences in genetic

sequence are associated with differences in multiple phenotypic traits. This associ-
ation can arise by many different mechanisms. Already early on, it was recognized
that single or different gene products from the same locus may underlie pleiotropic
effects on different traits and that the respective gene may play a direct role in the
development or physiology of the particular trait or it may affect a trait through a
cascade of other developmental processes and traits. The classical example is the
joint between two bones: even if only one of the bones is affected by a genetic
mutation, the effect will be reflected by the adjacent bone, because its growth is
regulated in part by the physical presence of the joint. Mutation in this case has an
indirect pleiotropic effect via the interactions between traits. With more detailed
insight into gene function, it became recognized that a product of a single gene may
become modified in many ways (splicing, posttranscriptional modifications, epige-
netic states, differential effect of environment, etc.) and finally that various functions
of a multifunctional gene may be performed by different functional domains,
allowing for separate evolution. Thus, a question arises whether pleiotropy is a
property of a gene or of a mutation within a gene (Stern and Orgogozo 2008). Not
every mutation within a gene affects all functions of a gene. Rather, different
mutations in the same gene may have different degrees of pleiotropy, some affecting
single traits, some all traits that gene is involved in the development of. The
pleiotropy at the level of a gene is thus rather a propensity, a number of traits the
mutations in this gene can affect, whereas the realized pleiotropy refers to the
specific mutation.

How Many Traits? Pleiotropy and the Evolution of Complexity

One of the long-standing but often not explicit discussions about pleiotropy is about
its overall effect across the genome and phenome and therefore its prevalence. While
geneticists agree that most genes are to some degree pleiotropic (i.e., pleiotropy is
ubiquitous), there is a disagreement in the models as to whether all genes affect,
directly or indirectly, all traits (pleiotropy is universal). The implicit assumption of
universal pleiotropy, for example, underlies many models, including the suggestion
that as the organisms become more complex, their potential for successful adaptation
decreases (the cost of complexity (Orr 2000)). This extrapolation is based on the idea
that the increase in organismal complexity involves an increase in the number of
traits. Assuming universal pleiotropy, it follows that each gene affects an increasing
number of traits as the complexity increases. Fisher had previously argued that in the
6 M. Pavličev

phenotypic space with two traits, the proportion of advantageous mutations, out of
all possible mutations, decreases when the size of mutation increases (a.k.a. Fisher’s
geometric model (Fisher 1930)). Smaller mutations are more likely to be advanta-
geous. Orr (2000) has shown that the more traits are affected by the mutation, the
more this relationship is exacerbated. Thus, the advantageous mutational steps
diminish in size as the complexity increases, given that increase in complexity also
means increase in the pleiotropy of the mutation.
Contrasting universal pleiotropy is the model of restricted pleiotropy, or varia-
tional modularity, which asserts that the gene pleiotropy is restricted to the traits with
common development or function (see the chapter on “▶ Modularity in Evo-Devo”).
These traits likely frequently experience simultaneous selection, whereas reducing
pleiotropic association with other traits reduces interference between traits that are
under conflicting selection pressure. So far, empirical work established good evi-
dence for restricted pleiotropy of gene effects in genetic and phenotypic variational
patterns (e.g., Mezey et al. 2000; Su et al. 2010). In addition, the absence of negative
effect of pleiotropy on effect size, and even its potential enhancement, has also been
shown (Wagner et al. 2008; Wang et al. 2010).
An important aspect of restricted pleiotropy has been noted by Hansen (2003).
Hansen pointed out that, given a certain number of genes in the genome, reducing
the interference due to pleiotropic genes comes at the price of reducing the number of
genes affecting each trait. As smaller genetic basis offers lesser mutational potential
for a trait, there is therefore a trade-off between restricting interference and reducing
the potential to generate mutation in the first place. Hansen thus proposed that most
evolvable genotype-phenotype maps would be those with intermediate levels of
pleiotropy.

Phenotypic Traits Versus Fitness

The organisms, which are most suitable for the study of evolution because of their
rapid reproduction, namely, microbes, often do not offer appropriate individually
measurable phenotypic traits. Instead, the effects of mutations are measured directly
on fitness of a clone, as a rate of reproduction. Fitness is only a single variable, and
therefore pleiotropy in this case has been conceptualized differently, namely, as the
effect of mutation on fitness measured in multiple environments or in multiple
genetic backgrounds. This concept of pleiotropy converges with gene-by-environ-
ment interaction, and it should be noted that pleiotropy and gene-by-environment
(or gene-by-gene) interaction indeed are mathematically equivalent (Falconer 1952;
Pavlicev and Cheverud 2015).

Evolution of Pleiotropy

While it is interesting to understand how the above aspects of pleiotropy affect its
role in response to selection, another important characteristic of pleiotropy in the
Pleiotropy and Its Evolution: Connecting Evo-Devo and Population Genetics 7

context of evo-devo is its own evolvability. By funneling genetic variation along

certain directions of phenotypic space, pleiotropy conserves developmental structure
and plays a role in maintaining structural properties such as a bauplans. This aspect
of pleiotropy is crucial in explaining phenotypic inertia and discrete organismal
diversity, rather than continuous distribution of phenotypic traits across species.
Evolutionary developmental biologists have made an important point by emphasiz-
ing the role of constraint (as opposed to ubiquitous variation) in shaping the future
potential to evolve. However, this view should not be taken to mean that pleiotropic
constraints are absolute constraints, to which no change can be introduced. The most
convincing evidence that pleiotropy evolves is the evolutionary divergence of
repeated, or serially homologous, elements, such as segments in arthropods, limbs
in quadrupedal vertebrates, digits, vertebrae, petals of the flowering plants, etc.
These organismal parts share genetic basis, i.e., the same core genes regulate the
development of each repeated part. Yet we know that the degree to which the
repeated elements have diverged in some species is enormous, for example, the
divergence of the forelimb and hind limb in apes and humans, or in birds and bats, or
segments in arthropods such as bees in which some segments retained legs, some
have wings, and yet others are fused into the thorax, head, or abdomen (Fig. 2). The
implication of these observations is that the correlated response due to shared genes
must have been alleviated in these traits, in order for them to individualize and
diverge from each other in an organism. To demonstrate this change, many studies
have focused on comparison of phenotypic intertrait correlations between species
(Marroig and Cheverud 2001). While it appears that for some trait complexes, these
are fairly stable and conserved across many species, other trait covariances have
changed considerably over time, such as the forelimb and hind limb during the
transition from quadrupedal to bipedal primates (Young et al. 2010) or different
functional modules in the skulls of Anolis lizards (Sanger et al. 2012). Yet this
interspecific change itself also has its origin in a population. How does this happen?
One of the mechanisms by which pleiotropic effects can change has been
obscured by our own assumptions about the constancy of genetic effects. At the
level of population, heritable phenotypic diversity can be fairly well described by
considering the effects of substitutions at single loci to be independent of each other
and consequently that they can be added up across the genome. This implicitly
assumes that the effects of gene substitutions are a property of the particular locus at
which they occur, and it follows that the phenotypic changes of a population occur
by the change of frequency of the alleles at these loci. It has been long known that the
effect of a substitution can change when the genetic background changes, that is,
when substitution takes place in the context of a different genotype at another locus
and this can have an effect even within a population. Yet these effects are often small
in interbreeding populations, and thus the short-term additive approximation is fairly
good. But these “interactions” between loci (▶ epistasis), or also across environ-
mental conditions, can contribute the kind of variation in gene effects necessary to
select for the effect sizes in different backgrounds. Important for the evolution of
pleiotropy is that such variability of genetic effect is not limited to the effect size, e.
g., whether the substitution increases a trait only slightly or considerably. Rather, the
8 M. Pavličev

Fig. 2 Individualization of
the body segments (Figure
adapted after: Genetic Science
Learning Center. “Homeotic
Genes and Body Patterns.”
Learn. Genetics. March 1,
2016. Accessed September 2,
2016. http://learn.genetics.
utah.edu/content/basics/
hoxgenes/.). In the millipede
above, most of the segments
have uniform structure,
whereas in the bee, the thorax
segment are integrated and
highly differentiated from the
abdominal segments

pleiotropy of a mutation has been also shown to differ across genetic and environ-
mental backgrounds, e.g., whether one or two traits are affected by the mutation, or
whether they are both affected in the same or different ways. This variation in
pleiotropy introduces the potential for pleiotropy to evolve and with it the degree
to which pleiotropy constraints the evolution of traits. Genetic background-
dependent variation in pleiotropic effects has been detected (Leamy et al. 2009;
Pavlicev et al. 2008) and modeled to show that it can indeed modify the trait
correlations (Pavlicev et al. 2011; Watson et al. 2014).
How can this play out in the change of correlation between traits that are affected
in their development by the same gene, such as the serially homologous forelimb and
hind limb?
A pleiotropic gene interacts during development of any particular trait with other
local genes that are also involved in building this trait, and these local genes differ
across traits the pleiotropic gene is involved in, e.g., between the forelimb and
the hind limb. The different effects of pleiotropic genes can thus be thought as
interactions with different (internal) environments, to stay with limb example; that of
the forelimb and that of the hind limb. As these “environments” change under
divergent selection, the effects of pleiotropic genes on the forelimb and hind limb
diverge.
Taking the potential effect of the context into consideration thus enables us to
extend the small-scale dynamics, taking place within the relatively invariant context
of a homogeneous population (microevolutionary level), to larger-scale evolution at
the interspecific level (macroevolution). While the apparent disparity between the
two levels has caused substantial disputes over what processes are important in
evolutionary change, at least some of that void between the intra- and interspecific
Pleiotropy and Its Evolution: Connecting Evo-Devo and Population Genetics 9

change can be bridged by including the context dependency of genetic substitutions

into the models.

Future Challenges

The genotype-to-phenotype map, of which pleiotropy is an important aspect, is an

abstraction of developmental mechanisms that translate genetic variation into phe-
notypes (see the chapter on “▶ Genotype-Phenotype Map”). So far, it has played a
coarse role in the study of heritable phenotypic pattern and its effects on the short-
term response to selection. Future challenges are twofold. One is the understanding
of various mechanisms that generate similar pleiotropic patterns and determining to
what extent the longer-term responses are influenced by the specific mechanism
involved (e.g., whether pleiotropy is due to the same or different gene products). The
other challenge is the evolution of these mechanisms and pleiotropy themselves.
This would essentially confront the molecular and developmental mechanisms with
population dynamics, as much as provide population dynamics with the much
needed molecular and developmental basis. Pleiotropy is not the only aspect of
genotype-phenotype map that can provide this bridge. The reason that pleiotropy is
particularly attractive in this respect is that it is a central concept in both fields of
evolutionary biology and has a rich history of research in both.

Cross-References

▶ Complex Traits
▶ Developmental Constraints
▶ Developmental Drift
▶ Evolvability
▶ Genotype-Phenotype Map
▶ Modularity in Evo-Devo
▶ Morphological Integration
▶ The Role of Epistasis in Evo-Devo

References
Cheverud JM (1988) A comparison of genetic and phenotypic correlations. Evol Int J Organ Evol
42:958–968
Cheverud JM (2007) The dangers of diagonalization. J Evol Biol 20:15–16; discussion: 39–44
Falconer DS (1952) The problem of environment and selection. Am Nat 86:293–298
Fisher RA (1930) The genetical theory of natural selection. Oxford University Press, Oxford
Hansen TF (2003) Is modularity necessary for evolvability? Remarks on the relationship between
pleiotropy and evolvability. Biosystems 69:83–94
10 M. Pavličev

Hansen TF, Armbruster WS, Carlson ML, Pelabon EC (2003) Evolvability and genetic constraint in
Dalechampia blossoms: genetic correlations and conditional evolvability. J Exp Zool B Mol
Dev Evol 296(1):23–39
Johnson NA, Porter AH (2007) Evolution of branched regulatory genetic pathways: directional
selection on pleiotropic loci accelerates developmental system drift. Genetica 129:57–70
Lande R (1979) Quantitative genetic analysis of multivariate evolution, applied to brain: body size
allometry. Evol Int J Organ Evol 33:402–416
Leamy LJ, Pomp D, Lightfoot JT (2009) Genetic variation in the pleiotropic association between
physical activity and body weight in mice. Genet Sel Evol 41:41
Lenski RE (1988) Experimental studies of pleiotropy and epistasis in Escherichia coli.
II. Compensation for maladaptive effects associated with resistance to virus T4. Evolution
42:433–440
Marroig G, Cheverud JM (2001) A comparison of phenotypic variation and covariation patterns and
the role of phylogeny, ecology, and ontogeny during cranial evolution of new world monkeys.
Evol Int J Organ Evol 55:2576–2600
Mezey JG, Cheverud JM, Wagner GP (2000) Is the genotype-phenotype map modular? A statistical
approach using mouse quantitative trait loci data. Genetics 156:305–311
Orr HA (2000) Adaptation and the cost of complexity. Evol Int J Organ Evol 54:13–20
Pavlicev M, Cheverud JM (2015) Constraints evolve: context dependency of genetic effects allows
evolution of pleiotropy. Annu Rev Ecol Evol Syst 46:413–434
Pavlicev M, Wagner GP (2012) A model of developmental evolution: selection, pleiotropy and
compensation. Trends Ecol Evol 27:316–322
Pavlicev M, Kenney-Hunt JP, Norgard EA, Roseman CC, Wolf JB, Cheverud JM (2008) Genetic
variation in pleiotropy: differential epistasis as a source of variation in the allometric relationship
between long bone lengths and body weight. Evol Int J Organ Evol 62:199–213
Pavlicev M, Cheverud JM, Wagner GP (2011) Evolution of adaptive phenotypic variation patterns
by direct selection for evolvability. Proceed Biol Sci/R Soc 278:1903–1912
Sanger TJ, Mahler DL, Abzhanov A, Losos JB (2012) Roles for modularity and constraint in the
evolution of cranial diversity among Anolis lizards. Evol Int J Organ Evol 66:1525–1542
Schluter D (1996) Adaptive radiation along genetic lines of least resistance. Evol Int J Organ Evol
50:1766–1774
Stearns FW (2010) One hundred years of pleiotropy: a retrospective. Genetics 186:767–773
Stern DL, Orgogozo V (2008) The loci of evolution: how predictable is genetic evolution?
Evolution Int J Organ Evol 62:2155–2177
Su Z, Zeng Y, Gu X (2010) A preliminary analysis of gene pleiotropy estimated from protein
sequences. J Exp Zool B Mol Dev Evol 314:115–122
True JR, Haag ES (2001) Developmental system drift and flexibility in evolutionary trajectories.
Evol Dev 3:109–119
Turelli M (1985) Effects of pleiotropy on predictions concerning mutation-selection balance for
polygenic traits. Genetics 111:165–195
Wagner GP, Kenney-Hunt JP, Pavlicev M, Peck JR, Waxman D, Cheverud JM (2008) Pleiotropic
scaling of gene effects and the ‘cost of complexity’. Nature 452:470–472
Wang Z, Liao BY, Zhang J (2010) Genomic patterns of pleiotropy and the evolution of complexity.
Proc Natl Acad Sci U S A 107:18034–18039
Watson RA, Wagner GP, Pavlicev M, Weinreich DM, Mills R (2014) The evolution of phenotypic
correlations and “developmental memory”. Evol Int J Organ Evol 68:1124–1138
Williams GC (1957) Pleiotropy, natural selection, and the evolution of senescence. Evol Int J Organ
Evol 11:398–411
Young NM, Wagner GP, Hallgrimsson B (2010) Development and the evolvability of human limbs.
Proc Natl Acad Sci U S A 107:3400–3405
Zhang J, Wagner GP (2013) On the definition and measurement of pleiotropy. Trends Genet
29:383–384
Epistasis

Thomas F. Hansen

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Epistasis as a Property of the Genotype-Phenotype Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Statistical Epistasis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
The Statistical Genotype-Phenotype Map of Quantitative Genetics . . . . . . . . . . . . . . . . . . . . . . . . . 5
Epistasis, Inheritance, and Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Statistical and Biological Epistasis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Estimating Epistasis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Epistasis Analysis in Molecular Genetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
The Importance of Epistasis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Summary of Epistasis Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Abstract
Epistasis is broadly synonymous with gene interaction, referring to cases in which
the effects of changing a gene depend on the state of other genes. Beyond this, the
term has acquired a number of different technical and nontechnical meanings,
which has led to confusion and misunderstanding in communication across
disciplines. Clear communication about epistasis is particularly pertinent in
evolutionary developmental biology both because of the relevance of epistasis
to some of its key research questions such as the evolution of evolvability and
canalization, and because evodevo acts as a trading zone for cross-disciplinary
communication.

T.F. Hansen (*)

Department of Biology, CEES and EVOGENE, University of Oslo, Oslo, Norway
e-mail: thomas.hansen@bio.uio.no

# Springer International Publishing Switzerland 2016 1

L. Nuño de la Rosa, G.B. M€uller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_53-1
2 T.F. Hansen

Keywords
Gene interaction • Quantitative genetics • Genotype-phenotype map •
Evolvability • Canalization

Introduction

In genetics, the term “epistatic” was introduced by Bateson and Punnett to describe
deviations from the expected 9:3:3:1 ratio of two independently segregating Men-
delian pairs with dominance. In his influential 1909 book on Mendelian inheritance,
Bateson used the terms “epistatic” and “hypostatic” to refer to cases in which one
factor, the epistatic one, conceals the effects of another, hypostatic, one. Hence, his
choice of the Greek term epistatic with the meaning of “upon” “standing” or
“stopping.” This terminology was in analogy with the contemporary use of the
terms dominant and recessive, when one dominant allelomorph (allele) conceals
the effects of another recessive one on the same pair (locus). Bateson saw the need
for different terms to describe the analogous relationship between alleles at different
loci. Bateson did not seem to intend this strictly. Throughout his book he stressed
that dominance is not a principle but a matter of degree, and this extends to epistasis.
Later, different forms of deviations from the 9:3:3:1 ratio gave name to different
types of epistasis such as dominance, recessive, and compositional epistasis.
Bateson’s usage was soon supplemented by another concept of gene interaction.
In the key 1918 paper unifying Mendelian segregation with the biometric laws of
heredity, Fisher noted that the effects of independently segregating factors need not
add up in a linear manner, and he coined the term “epistacy” for deviations from
statistical additivity. With a century of hindsight it is easy to think that Fisher chose a
slightly different term to underline the difference between his statistical and
Bateson’s biological notion of gene interaction, but Fisher provided no discussion
of the matter, and did not make the same terminological distinction with regard to
dominance. In any case, Fisher’s term epistacy eventually slid out of usage and was
replaced with epistasis.
This terminological conflation of statistical and biological epistasis has been an
obstacle in cross-disciplinary, even within-disciplinary, communication about gene
interaction. While biological measures of gene effects are defined as differences
between specific genotypes without regard to their relative occurrence, the statistical
measures are defined as average deviations of the genotype effects from population
averages over all genotypes in a population. The latter makes statistical gene effects
and epistasis dependent on the composition of a population, so that common
genotypes, for example, tend to have smaller effects than rare genotypes. Within
the field of quantitative genetics the statistical definitions of gene effects proved
convenient in terms of describing similarities among relatives and predicting the
short-term response to artificial selection, but the statistical description of epistasis as
a residual from additive effects averaged out the effects of biological epistasis and
led to the notion that epistasis was uncommon, inert, and inconsequential for
Epistasis 3

selection dynamics at least. This clashed with the intuitions of systems-oriented

biologists that (biological) epistasis was ubiquitous and essentially important for
organismal function and evolution.
Population genetics used a notion of epistasis that is closer to the biological
concept than to the statistical concept of quantitative genetics. In theoretical popu-
lation genetics, the effects of genotypes on fitness are stipulated in advance and not
as statistical averages. This is the basis of most of the standard insights on the effects
of epistasis on evolution as in Wright’s shifting-balance theory, the Bateson-
Dobzhansky-Muller model for the evolution of reproductive isolation, coadapted
gene complexes, and the evolution of sex and recombination.
Molecular genetics stuck to Bateson’s narrow definition of epistasis as a mutation
that masks the effect of another mutation on another gene. This was linked to the idea
that an epistatic mutation would be in a gene that acted downstream to an hypostatic
mutation and that epistasis therefore could be used as a tool to infer position of genes
in genetic pathways.
These different notions of epistasis lived side by side during the development of
the modern synthesis but came in closer contact in the 1980s. The emergence of an
evolutionary quantitative genetics brought the methods and theory of quantitative
genetics into evolutionary biology, and the different notions of epistasis and ideas
about its importance came in conflict. Evolutionary developmental biology accen-
tuated this with its focus on how the genotype-phenotype map affects evolution.
Epistasis is a property of the genotype-phenotype map and plays a crucial role in key
research questions of evodevo such as the evolution of evolvability and canalization.
The interest in gene regulation and gene networks in evodevo and systems biology
also brought the molecular genetics view of epistasis in contact with the epistasis
concepts of evolutionary biology.

Epistasis as a Property of the Genotype-Phenotype Map

Gene products function in complex biochemical pathways and are thus embedded in
networks of molecular interaction. The epistasis concept is not used to describe
interactions at this level. Instead it describes interactions between the phenotypic
effects of genetic changes, i.e., allele substitutions including mutations. Epistasis is
not a property of the gene but a property of two or more gene substitutions that may
be epistatic in relation to each other. This makes epistasis an aspect of the genotype-
phenotype map. The mapping from genotypes to phenotypes is an abstract descrip-
tion of how phenotypic changes relate to genotypic changes. An additive genotype-
phenotype map means that any specific substitution of alleles will have the same
phenotypic effect regardless of the state of other genes (i.e., regardless of the position
in genotype space), so that the cumulative phenotypic effect of several substitutions
equals the sum of their individual phenotypic effects. Every deviation from this
pattern may be termed gene interaction and again divided into dominance and
epistasis depending on whether the composite changes happen at the same or
4 T.F. Hansen

PC
PC

PB
Phenotype

PB

PA
PA

g g g
A B C
Genotype

Fig. 1 A nonlinear genotype-phenotype map. The same genetic change, Δg, will have different
phenotypic effects, ΔP, depending on the genetic background (positions A, B, or C) in which it
happens (Modified from Hansen (2015))

different loci, although an interaction between two subsequent changes of the same
allele is sometimes called intralocus epistasis.
The strategy of modeling the dynamics of single alleles one-by-one was success-
ful in demonstrating the power of natural selection and in elucidating fundamental
principles of microevolution, but it has been less helpful in understanding macro-
evolution, because the additive summation of effects becomes increasingly unreal-
istic with larger changes. In a sense, additivity is a constant-evolvability assumption
that allows little room for genetic constraints to affect evolution.
Epistasis can be conceptualized as nonlinearities in the genotype-phenotype map
(e.g., Rice 1998). As shown in Fig. 1, the same genetic change can have different
phenotypic effects depending on position in the genotype-phenotype map. Moving
from position A to position B, the convexity of the map leads to an increased
phenotypic effect. This is called positive epistasis. Moving from position B to
position C, the concavity of the map leads to a decreased phenotypic effect. This is
called negative epistasis. Moving into the flat areas of the map, genotypic changes are
still possible, but their phenotypic effects vanish. This is called canalization (e.g., Flatt
2005). With the map in Fig. 1, the evolvability is high in the middle region, but moving
from position B out towards the edges shows how negative epistasis leads to canaliza-
tion and reduced evolvability. This constitutes an epistatic constraint on evolution,
because it is not possible to change the phenotype beyond the limits of the map.
Real genotype-phenotype maps need not be shaped as in Fig. 1. The degree and
sign of curvature and the existence and position of absolute limits to phenotypic
Epistasis 5

change are empirical questions. The figure illustrates how epistasis allows the
evolution of evolvability, and how this depends not on epistasis in general but on
particular systematic patterns of epistasis. Positive epistasis in the direction of
selection leads to evolution of increased evolvability, while negative epistasis
leads to the evolution of decreased evolvability (canalization).

Statistical Epistasis

The Statistical Genotype-Phenotype Map of Quantitative Genetics

The statistical model of the genotype-phenotype map initiated by Fisher is at the core
of quantitative genetics. Here genetic effects are defined as statistical deviations from
an average. In its modern form the model starts with defining the average effect of an
allele as the average deviation of its carriers from the population mean (technically
an average excess; the difference between average excess and average effect will be
ignored for simplicity). The additive effect (breeding value) of an individual is the
sum of these effects for all the alleles it carries. The actual phenotype of the
individual may deviate from the breeding value both because of environmental
effects and because its genetic component may deviate from the additive sum due
to dominance or epistasis. For example, the average deviation of individuals carrying
two specific alleles at the same locus may not equal the sum of the average effects of
these two alleles. The average deviation from the sum is then the statistical domi-
nance effect of these two alleles. Similarly, the average deviation of individuals
carrying two specific alleles at different loci may differ from the sum of the average
effects of the alleles, and this difference is a (statistical) epistatic deviation. In
general the epistatic effect of any set of alleles is defined as the average deviation
of the carriers of this set from the prediction given by taking the sum of all the lower-
order effects of these alleles, i.e., the sum of their average effects, dominance effects,
and lower-order epistatic effects (Lynch and Walsh 1998).
One may think of the statistical genotype-phenotype map as a multiple regression
of individual phenotypes on the presence/absence of alleles and sets of alleles.
Dominance and epistasis are interaction effects in this model. The variance
explained by the sum of the average effects (i.e., first-order effects) is the additive
(A) genetic variance, and the variance explained by the interactions between alleles
at the same locus is the dominance (D) variance. There are many different epistatic
variances. The variance explained by interactions between two alleles at different
loci is the additive-by-additive (AA) epistatic variance, the variance explained by
interactions among two alleles at one locus and one at another locus is the additive-
by-dominance (AD) epistatic variance, the variance explained by interactions among
four alleles at two loci is the dominance-by-dominance (DD) epistatic variance, the
variance explained by interactions among three alleles at three different loci is the
additive-by-additive-by-additive (AAA) epistatic variance, etc. The sum of all these
variances is the total genetic variance.
6 T.F. Hansen

This decomposition is useful in describing inheritance and similarity between

relatives. The covariance between phenotypes of two related individuals is a sum of
contributions of all these variance components, each weighted with the probability
that the two relatives share the allele sets in question (Lynch and Walsh 1998). For
example, full sibs share half the additive effects and thus half the additive variance;
they further share one quarter of the dominance effects, one quarter of the AA
epistatic effects, and smaller fractions of higher-order epistasis. Offspring and a
parent share half the additive variance, none of the dominance variance, one quarter
of the AA epistatic variance, and smaller fractions of higher-order AAA types of
epistatic variance.
Significantly, because all the alleles carried by an individual are inherited from its
two parents, all the additive variance in a generation has been inherited from the
previous generation. In contrast, none of the dominance variance and only fractions
of the epistatic variances are normally inherited from the previous generation. This is
because sets of alleles are broken up and recombined into new combinations each
generation.

Epistasis, Inheritance, and Selection

From these considerations, it is clear why the additive effects and the additive
variance play central roles in inheritance and selection. Natural selection acts on
variation, and the additive variance is the heritable component of the phenotypic
variance in a population. Natural selection does not see the difference between
components of variance, but only the effects on the additive component are trans-
ferred to the next generation and contribute to evolution by natural selection. The
smaller fractions of epistatic variance that are inherited, most significantly the one
quarter of the AA epistatic variance, can yield a minor evolutionary effect, but this
effect is transient because the selected allele combinations are continuously being
broken down by recombination. If selection ceases, the gain achieved by selection on
epistatic variance is removed at a geometric rate by recombination.
This has served as a theoretical justification for the focus on additive variance in
quantitative genetics and for the single-gene perspective of population genetics and
most other fields of evolutionary biology. Fisher’s average effect is an elegant device
for capturing the dynamics of individual alleles without in fact assuming that their
effects are biologically additive. In a large population, a specific allele will find itself
in myriads of different combinations with other alleles. The effects of selection on
the allele will depend on its phenotypic effect averaged over all these combinations,
and this is precisely what the average effect is measuring. The definition of statistical
epistasis ensures that the epistatic deviations must sum to zero, and hence that they
do not affect the dynamics of individual allele frequencies. Hence, the focus on
statistical additivity in quantitative genetics is not based on an assumption of
biological additivity but on an identification of the statistical averages that govern
the dynamics of individual alleles in complex systems of biological interaction.
Epistasis 7

PC
Phenotype

PB

PA

A B C
Genotype

Fig. 2 The same levels of molecular genetic variation will generate different levels of variation in
the phenotype depending on the genetic background (positions A, B, or C) (Modified from Hansen
(2015))

Statistical and Biological Epistasis

Even though statistical epistasis and epistatic variances are largely inconsequential
for evolutionary dynamics, this does not extend to biological epistasis. As the
additive effects are averages over genotypes in a population, they will change
when the genetic background is changing, and this change is determined by biolog-
ical epistasis. In Fig. 2, distributions of “molecular” genetic variation on the x-axis
are mapped into distributions of phenotypically expressed genetic variation on the
y-axis. At each point, A, B, and C, the molecular variation is the same, but due to the
epistasis the distributions of phenotypically expressed genetic variation are different.
Over the range of variation at each point, the map is approximately linear, and fitting
a statistical regression would support an approximately additive model at each point,
so that the variation mapped to the phenotype axis would be additive genetic
variation. Moving from point A to point B, the positive epistasis increases the
additive variance, and moving on towards point C, the negative epistasis in this
region would reduce the additive variance, and evolvability would disappear as
complete canalization is approached. At each point during this trajectory the pheno-
typic response to selection could be predicted from the additive genetic variances,
but the long-term dynamics would be determined by the effects of epistasis on the
dynamics of the additive variance.
Even if the range of variation was sufficient to cover nonlinearities as in Fig. 3,
the statistical epistasis would be estimated as deviations from the best-fitting linear
8 T.F. Hansen

PC

Phenotype

PB

g g
B C
Genotype

Fig. 3 Fitting an additive model (straight black line) over a range of genetic variation (dashed-line
distribution along x-axis) captures the average effect, ΔP, of an allele substitution, Δg, over the
range but also constrains the average effects to be constant so that ΔPB = ΔPC. Epistasis causes
residual deviations from the linear model (diamonds), but their variance does not indicate specific
patterns in the map

approximation and fail to describe the specific nonlinearities in the map. Epistatic
variance could be detected, but it would be similar regardless of whether the
biological epistasis was positive, negative, or simply random. The model would
predict constant additive effects and evolvability over the range of the map.
A clear conceptual distinction between biological and statistical epistasis
emerged gradually in the 1990s. In a key paper, Cheverud and Routman (1995)
introduced the concept of “physiological” (= biological) epistasis and showed that it
can influence the additive genetic variance. Hansen and Wagner (2001) developed
this further and showed how “functional” (= biological) epistasis could be
represented in a quantitative genetics framework. Carter et al. (2005) used Hansen
and Wagner’s multilinear representation of epistasis to formally describe the effects
of biological epistasis on selection dynamics. In particular, they described how
positive directional epistasis leads to the evolution of increasing additive variance
and evolvability, while negative directional epistasis has the opposite effect. If the
epistasis is nondirectional without any systematic patterns, the dynamics are almost
indistinguishable from an additive model.
Such systematic effects of biological epistasis on the selection response have
nothing to do with selection on epistatic variance. Selection on the epistatic variance
leads to a buildup of linkage disequilibrium that is transient in the sense that it is
rapidly broken down by recombination. In contrast, the effects of directional
Epistasis 9

epistasis are permanent, because they are mediated through changes in the genetic
background that modify the biological effects of subsequent allele substitutions. If
selection increases the frequency of alleles that, say, increase a trait, and these alleles
have an average positive epistatic interaction with other alleles that have a positive
effect on the trait, then these other alleles will more often find themselves in genetic
backgrounds that elevate their effects. These elevated effects are permanent in the
same sense as changes of allele frequencies are permanent.
Permanent effects of epistasis on the selection response were not captured by
quantitative-genetics theory, because the statistical representation of epistasis as
residuals from a regression constrained it to be nondirectional. The missing concep-
tual distinction between statistical and biological epistasis then led many to the
inference that epistasis in general was unimportant (reviewed in Hansen 2013).
The NOIA model of Álvarez-Castro and Carlborg (2007) provides a general
framework for representing most forms of functional (biological) and statistical
epistasis and for translating between them.

Estimating Epistasis

In classical quantitative genetics, epistasis is estimated either as epistatic variance

components inferred from patterns of resemblance between relatives or from line-
cross analyses (Lynch and Walsh 1998). Line-cross analyses are based on regres-
sions of the mean phenotypes of different crosses (“line-cross derivates”) on the
fraction of genes they have from each parental line and on their level of heterozy-
gosity. For example, a back cross between the F1 and a parental is predicted to have
75 % of its genes from this parental and 25 % from the other and to be 50 %
heterozygotic. This allows the fitting of crude models of interaction between genes
from the two parental lines. In principle, nonlinearities of the form illustrated in
Fig. 1 can be inferred from such data, but classical line-cross analysis has yielded
few insights due to its focus on significance testing rather than estimation and on the
distinction between AA, AD, and DD types of epistasis. In any case, this method is
now largely superseded by marker-assisted approaches.
Quantitative-trait locus (QTL) and genome-wide association studies (GWAS) use
molecular markers to identify positions in the genome with effects on phenotypic
traits. These approaches have been focused on identifying genes and estimating their
individual effects, but it is possible to fit regression models with interactions that can
identify epistasis (Lynch and Walsh 1998; Malmberg and Mauricio 2005). The
detection of epistasis is made difficult by the large number of potential interactions
and the use of significance thresholds to detect individual effects. Strong and
systematic patterns of epistasis may go undetected, because they are spread over
many interactions with individually small effects and there is a danger that signif-
icant interactions may be extremes that are atypical of the general patterns. Evidence
for epistasis often comes from variants of these models in which larger ranges of
phenotypes are studied (e. g., Huang et al. 2013).
10 T.F. Hansen

The empirical study of epistasis has suffered from a lack of connection between
statistical methods and theoretical relevance (Hansen 2015). The classical epistatic
variance components have little evolutionary relevance, and the marker-based esti-
mates are typically constrained to be nondirectional by the use of the standard
statistical regression model. Le Rouzic (2014) reviews modifications and methods
for detecting directional patterns of epistasis. There is also a tradition for studying
theory-relevant patterns of epistasis on fitness, for example, by regressing fitness
against correlates of accumulated mutations to estimate levels of synergistic epistasis
among deleterious mutants. More recently, systematic studies of interactions
between induced mutations on fitness and life-history traits in yeast and bacteria
have been used to elucidate the role of epistasis in adaptation (e.g., Perfeito
et al. 2014).

Epistasis Analysis in Molecular Genetics

In molecular genetics, epistatic interactions between, usually loss-of-function, muta-

tions are used to infer the position of genes in a pathway. Following Bateson an
epistatic mutation is a mutation that masks the effect of another (hypostatic) muta-
tion, and this relationship is taken as evidence that the gene with the epistatic
mutation is coming after the other in a pathway. The validity of this inference
requires a number of auxiliary assumptions including the two mutations being the
only factors affecting the phenotype. Drees et al. (2005) give a general overview of
epistasis analysis.
More generally, the relationship between epistasis and the underlying structure of
metabolic pathways, gene-regulatory networks, or physiological/developmental
interactions is a topic of research in systems biology.

The Importance of Epistasis

The main relevance of epistasis for evodevo, at least, comes from its connection to
the evolution of evolvability and canalization. It has only recently been recognized
that this depends on systematic patterns of gene interaction that are not identifiable
within the models of statistical genetics. Consequently, there is only scattered work
to identify and formally describe how the many possible patterns of interaction and
nonlinearity of the genotype-phenotype map may influence evolution. Beyond the
identification of directional epistasis and convexity as key elements in the evolution
of evolvability (e.g., Rice 1998; Carter et al. 2005), there is a body of work on how
canalization may hide genetic variation that can subsequently be released in an
evolutionary capacitance mechanism (e.g., Hermisson and Wagner 2004).
More generally, epistasis is related to the complexity of the genotype-phenotype
map. It is here useful to distinguish between magnitude and sign epistasis. While
sign epistasis refers to cases where a change in the genetic background would change
the order of the effects of genotypes at a locus, magnitude (or order-preserving)
Epistasis 11

epistasis refers to cases where only the magnitude and not the order of effects are
changed. Specifically, sign epistasis has been defined as a change in the ordering of
fitness values, and this sets up the possibility of complex dynamics with the
possibility of internal equilibria and multistability that may act as strong constraints
on evolution (Weinreich et al. 2005). The existence of complex epistasis creating
multipeaked genotype-fitness relations was a premise of Wright’s view of evolution
as expressed in his shifting-balance theory and contrasts with the Fisherian view of
smooth additive landscapes (e.g., Whitlock et al. 1995). For Wright, evolution
consisted in jumps between such peaks mediated by genetic drift in small sub-
populations. A general model of the interaction between genetic drift and epistasis
can be found in Barton and Turelli (2004).
One important question is whether patterns of epistasis may reflect limits to
evolution. If a trait is selected up towards a limit, we may expect a pattern of
negative epistasis where allele substitutions that increase the trait towards the limit
show increasing canalization or even reversals of effect when the trait approaches the
limit. Such epistatic constraints can in principle be investigated by studying the
relationship between phenotypic trait values and the effects of allele substitutions,
but this has of yet not received systematic attention. On the other hand, the existence
of epistasis may also provide the possibility of breaking constraints by allowing
pleiotropic effects to evolve (Pavlicev and Cheverud 2015).
The influence of epistasis increases with increasing distance in genotype space,
and this makes it important in macroevolution and speciation. This is illustrated by
the Bateson-Dobzhansky-Muller model for the evolution of postzygotic reproduc-
tive isolation. Even without differences in selection regime, isolated populations will
experience different genetic changes due to genetic drift (e.g., systems drift). Such
changes must be compatible with the genetic background in their own population,
but there is no selection for compatibility with the genetic background of a different
population, and hybridization will then generate individuals with untested gene
combinations. Such combinations with deleterious effects on fitness are called
Bateson-Dobzhansky-Muller incompatibilities. These will accumulate at an accel-
erating pace with increasing genetic difference between populations, and virtually
guarantee that complete reproductive isolation will eventually arise as genetic
distance is increasing.
Epistasis is a factor in the evolution of recombination and sexual reproduction.
The costs and benefits of breaking up old and creating new allele combinations
depend on the patterns of epistatic interaction among the alleles. While the breakup
of coadapted gene complexes is unfavorable, it can be favorable to create offspring
with diverse gene combinations to increase the probability that some of them are
well adapted or free from combinations of deleterious alleles. If adaptation requires
individually nonfavorable mutations in several genes, the rate of adaptation may be
greatly elevated by sexual recombination. According to the deterministic-mutation
hypothesis, sex is maintained as an adaptation to reduce the mutation load, but this
works only in the presence of relatively strong synergistic epistasis where the fitness
effects of several deleterious mutations are more severe than the (multiplicative)
effects of the mutations in isolation.
12 T.F. Hansen

Summary of Epistasis Terminology

The key distinction in epistasis terminology is between statistical epistasis, Fisher’s

epistacy, on one side, and what has variously been called biological, functional, or
physiological epistasis on the other.
Statistical epistasis refers to the interaction terms in a least-squares regression on
the presence of alleles. It can be divided into pairwise additive-by-additive (AA) and
higher-order interactions. The variances explained by these interaction terms are the
additive-by-additive epistatic variance, etc.
Hansen and Wagner (2001) defined functional epistasis as a dependency of the
effects of a genetic substitution (on one or multiple loci) on the genetic background
(i.e., the state of other loci in the genotype). This is the essence of the biological
epistasis concepts including Cheverud and Routman’s (1995) physiological epista-
sis, which was defined as a dependence of the difference in genotypic values at one
locus on the state of another locus. The idea behind these concepts was to formally
define epistatic effects independently of the composition of a population. They are
still relative to a reference genotype, however, and specification of the reference
genotype remains essential in all modeling of epistasis. Estimation and modeling of
epistasis may be misleading if implicitly assumed reference genotypes are not made
clear. Tools for translating between different reference genotypes and for relating
biological and statistical epistasis are provided in Hansen and Wagner (2001), Barton
and Turelli (2004), and Álvarez-Castro and Carlborg (2007).
Positive and negative epistasis refer to interactions for which the composite effect
of two or more substitutions are elevated above or depressed below the sum of their
individual effects. This requires a scale, and positive epistasis in one direction equals
negative epistasis in the other. Systematic positive or negative interactions in one
direction are called directional epistasis, while cases in which positive and negative
interactions cancels out are called nondirectional epistasis. Magnitude epistasis or
order-preserving epistasis is used when changes in the genetic background only
cause changes in the magnitude of effects, while sign epistasis or order-breaking
epistasis refer to cases in which the order of effects of the genotypes at a locus are
changed. Multilinear epistasis refers to a pattern in which sets of genotypic effects
are proportionally modified by changes in the genetic background.
The terminology for fitness epistasis is convoluted with positive and negative
epistasis sometimes referring to interactions between beneficial (fitness-increasing)
mutations and sometimes to interactions between deleterious (fitness-decreasing)
mutations. In addition, terms such as synergistic, antagonistic, and diminishing-
returns epistasis are used for positive or negative fitness interactions in either
direction. It is also essential to distinguish between Wrightian fitness where epistasis
is usually defined as deviations on a multiplicative scale and Malthusian fitness
where it is usually defined as deviations on an arithmetic scale (Wagner 2010).
Fitness epistasis may also differ depending on whether the reference genotype is one
with maximal or average fitness. Furthermore, epistasis for fitness must be distin-
guished from epistasis in the traits underlying fitness. Unless the fitness function is
Epistasis 13

linear, these will differ, and with a nonlinear (e.g., stabilizing) fitness function, an
additive genetic architecture in the trait will generate systematic epistasis for fitness.
The widespread relevance of gene interaction has given rise to many context-
dependent terminologies including the Bateson-Dobzhansky-Muller incompatibili-
ties for deleterious fitness interactions between alleles from different populations, the
concept of a modifier where one gene is assumed to change the effect of another
without itself having an effect on the trait, the concept of differential epistasis when
pleiotropic effects are differentially modified by a change in the genetic background,
and the concept of compensatory change where the effect of one substitution is
nullified by another.

Cross-References

▶ Canalization
▶ Evolvability
▶ Genotype-Phenotype Map
▶ Pleiotropic Constraint

References

Álvarez-Castro JM, Carlborg Ö (2007) A unified model for functional and statistical epistasis and
its application in quantitative trait loci analysis. Genetics 176:1151–1167
Barton NH, Turelli M (2004) Effects of genetic drift on variance components under a general model
of epistasis. Evolution 58:2111–2132
Carter AJR, Hermisson J, Hansen TF (2005) The role of epistatic gene interactions in the response
to selection and the evolution of evolvability. Theor Popul Biol 68:179–196
Cheverud JM, Routman EJ (1995) Epistasis and its contribution to genetic variance components.
Genetics 139:1455–1461
Drees BL, Thorsson V, Carter GW, Rives AW, Raymond MZ, Avila-Campillo I, Shannon P, Galitski
T (2005) Derivation of genetic interaction networks from quantitative phenotype data. Genome
Biol 6:R38
Flatt T (2005) The evolutionary genetics of canalization. Q Rev Biol 80:287–316
Hansen TF (2013) Why epistasis is important for selection and adaptation. Evolution 67:3501–3511
Hansen TF (2015) Measuring gene interaction. In: Moore JH, Williams S (eds) Epistasis: methods
and protocols. Methods in molecular biology. Springer, New York, pp 115–142
Hansen TF, Wagner GP (2001) Modeling genetic architecture: a multilinear theory of gene
interaction. Theor Popul Biol 59:61–86
Hermisson J, Wagner GP (2004) The population genetic theory of hidden variation and genetic
robustness. Genetics 168:2271–2284
Huang W, Richards S, Carbone MA, Zhu D, Anholt RRH, Ayroles JF, Duncan L, Jordan KW,
Lawrence F, Magwire MM, Warner CB, Blankenburg K, Han Y, Javaid M, Jayaseelan J,
Jhangiani SN, Muzny D, Ongeri F, Perales L, Wu Y-Q, Zhang Y, Zou X, Stone EA, Gibbs
RA, Mackay TFC (2013) Epistasis dominates the genetic architecture of Drosophila quantita-
tive traits. Proc Natl Acad Sci U S A 109:15553–15559
Le Rouzic A (2014) Estimating directional epistasis. Front Genet 5:198
Lynch M, Walsh B (1998) Genetics and analysis of quantitative characters. Sinauer, Sunderland
14 T.F. Hansen

Malmberg RL, Mauricio R (2005) QTL-based evidence for the role of epistasis in evolution. Genet
Res 86:89–95
Pavlicev M, Cheverud JM (2015) Constraints evolve: context dependency of gene effects allows
evolution of pleiotropy. Annu Rev Ecol Evol Syst 46:413–434
Perfeito L, Sousa A, Bataillon T, Gordo I (2014) Rates of fitness decline and rebound suggest
pervasive epistasis. Evolution 68:150–163
Rice SH (1998) The evolution of canalization and the breaking of von Baer’s laws: modeling the
evolution of development with epistasis. Evolution 52:647–656
Wagner GP (2010) The measurement theory of fitness. Evolution 64:1358–1376
Weinreich DM, Watson RA, Chao L (2005) Sign epistasis and genetic constraint on evolutionary
trajectories. Evolution 59:1165–1174
Whitlock MC, Phillips PC, Moore FB-G, Tonsor SJ (1995) Multiple fitness peaks and epistasis.
Annu Rev Ecol Syst 26:601–629
Alternation of Generations in Plants
and Algae

Simon Bourdareau, Laure Mignerot, Svenja Heesch, Akira F. Peters,

Susana M. Coelho, and J. Mark Cock

Abstract
Photosynthetic organisms are found in most of the branches of the eukaryotic tree
of life, and these organisms have diverse life cycles. There has been a tendency
toward dominance of the diploid phase of the life cycle in the land plant lineage,
and recent analyses suggest a similar trend in the brown algae. A number of
hypotheses have been proposed to explain the evolutionary stability of different
types of life cycle, and in some cases these hypotheses are supported by empirical
studies. Molecular analyses are elucidating the regulatory molecules that control
life cycle progression and are providing insights into the developmental pathways
associated with the construction of each generation of the life cycle.

Keywords
Diploid • Epigenetic • Gametophyte • Haploid • Sporophyte

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
The Diversity of Plant and Algal Life Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
A Relationship Between Life Cycle Type and Degree of Multicellular Complexity . . . . . . . . . . . . 6
Theoretical Advantages and Disadvantages of Different Types of Life Cycle . . . . . . . . . . . . . . . . . . . 6
Genetic Regulation of Life Cycle Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

S. Bourdareau • L. Mignerot • S. Heesch • S.M. Coelho • J.M. Cock (*)

UMR 8227 Integrative Biology of Marine Models, Algal Genetics Group, Station Biologique de
Roscoff, CS 90074, Sorbonne Universités, UPMC University Paris 06, CNRS, Roscoff, France
e-mail: simon.bourdareau@sb-roscoff.fr; laure.mignerot@sb-roscoff.fr;
svenja.heesch@sb-roscoff.fr; coelho@sb-roscoff.fr; cock@sb-roscoff.fr
A.F. Peters
Bezhin Rosko, Santec, France
e-mail: akirapeters@gmail.com

# Springer International Publishing AG 2017 1

L. Nuño de la Rosa, G.B. Müller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_58-1
2 S. Bourdareau et al.

The Origins of Sporophyte and Gametophyte Developmental Programs . . . . . . . . . . . . . . . . . . . . . . . 9

Consequences of Life Cycle Type on Genome Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Introduction

The term “algae” groups together photosynthetic organisms from a broad range of
lineages, with representatives in almost all the supergroups of the eukaryotic tree of
life (Fig. 1, see the glossary for definitions of the terms used). From a strict,
taxonomic point of view, “plants” correspond to the kingdom Plantae (equivalent
to the modern group Archaeplastida; Fig. 1), but this term is often used loosely to
include any macroscopic photosynthetic organism, particularly those in terrestrial
habitats. In any event, plants are therefore a subset of the algae.
While it is preferable to use more taxonomically precise names when discussing
phylogeny, the terms plants and algae are nonetheless extremely useful because they
group together organisms that share many common biological features that stem
from their autotrophic lifestyles based on photosynthesis. The broad taxonomic
distribution of these organisms can be traced back to the various mechanisms
whereby they have acquired the ability to carry out photosynthesis. For algae in
the archaeplastid group (which includes green algae, red algae and glaucophytes;
Fig. 1), photosynthetic capacity arose due to a primary endosymbiotic event, which
involved the engulfment of a cyanobacterium by a common ancestral eukaryotic
cell. The enslaved cyanobacterium became the plastid.
The other eukaryotic lineages acquired photosynthesis by more complex second-
ary (and perhaps even tertiary) endosymbiotic events, in which a photosynthetic
eukaryote (usually a red or green alga) was enslaved by another eukaryotic cell. It is
this process of secondary endosymbiosis that has led to the occurrence of photosyn-
thetic organisms in such a diverse array of eukaryotic supergroups (stramenopiles,
alveolates, rhizarians, haptophytes, cryptophytes, and excavates). Given the com-
plicated evolutionary history of plants and algae, it is not surprising that they exhibit
a high level of diversity with regard to many characters, including their life cycles,
the feature that will be discussed in this chapter.
The basic eukaryote sexual life cycle involves an alternation between two key
processes: meiosis, which allows the chromosome number to be reduced by half, and
syngamy or gamete fusion, which restores the level of ploidy by bringing together
the chromosomes of the fusing gametes in a single nucleus within the zygote (John
1994). Before meiosis the cells are diploid, after meiosis they are haploid, and
syngamy restores the diploid state. Variations on this basic life cycle can be defined
based on the relative importance of these two phases, i.e., whether the organism
grows (undergoes mitotic cell divisions) during the haploid or the diploid phase, or
both (Fig. 2). When growth occurs during the diploid phase, the life cycle is called a
diploid life cycle (the human life cycle is one example). When growth occurs during
the haploid phase, the life cycle is called a haploid life cycle (e.g., that of the green
Alternation of Generations in Plants and Algae 3

Fig. 1 Schematic tree of the eukaryotes showing the positions of algal groups. Green lettering
indicates groups that include photosynthetic organisms (algae). LECA last eukaryotic common
ancestor

microalga Chlamydomonas). Finally, in some organisms, growth occurs during both

the haploid and diploid phases. These organisms are said to have haploid-diploid life
cycles. Examples include angiosperms, where the macroscopic plant is the diploid
phase and microscopic pollen grains and embryo sacs constitute the haploid phase.
The above paragraph applies to both unicellular and multicellular organisms. For
the latter, mitosis serves not only to increase cell number (asexual reproduction) but
is also the process that underlies construction of the multicellular body plan (devel-
opment). Multicellular organisms with haploid-diploid life cycles have two multi-
cellular generations. For plants and algae, these two generations are called the
sporophyte and the gametophyte, i.e., the spore-producing “plant” (where meiosis
occurs to produce spores) and the gamete-producing “plant” (i.e., which generates
the gametes), respectively. For these organisms, the alteration of generations referred
to in the title of this chapter is the repeated cycle of sporophyte and gametophyte
generations produced as a haploid-diploid life cycle progresses.
This chapter will summarize current knowledge about the evolutionary origins
and evolutionary trajectories of plant and algal life cycles. We will provide an
4 S. Bourdareau et al.

Fig. 2 Main types of sexual life cycle found in eukaryotes. Differences between eukaryotic sexual
life cycles depend principally on two key events, meiosis, and gamete fusion (syngamy). The
relative positioning of these events determine if the organism spends the majority of its time in the
diploid phase (i.e., has a diploid life cycle) or in the haploid phase (i.e., has a haploid life cycle).
Many organisms have intermediate life cycles with two generations, one that is haploid and the
other diploid (haploid-diploid life cycles)

overview of the different types of life cycle in the major algal groups, look at
evolutionary trends within each group and will attempt to relate these trends to
theoretical predictions. We will also describe recent advances in understanding how
life cycles are controlled at the genetic and epigenetic levels.

The Diversity of Plant and Algal Life Cycles

The phylogenetic group that includes the green algae and terrestrial plants, the
Viridiplantae, consists of two main taxa, the chlorophytes and the streptophytes
(Leliaert et al. 2012; Fig. 1). Most chlorophytes have haploid life cycles (e.g., the
unicellular alga Chlamydomonas), but some taxa with multicellular members exhibit
haploid-diploid life cycles. The haploid-diploid life cycles of multicellular
chlorophyte algae can either involve an alternation between morphologically similar
generations (i.e., isomorphic life cycles, e.g., Ulva) or the two generations can be
Alternation of Generations in Plants and Algae 5

morphologically dissimilar (i.e., heteromorphic life cycles). Clear evolutionary

trends are difficult to discern within this group because individual sub-taxa can
exhibit highly diverse morphologies and because there is still some uncertainty about
the phylogenetic relationships between groups within the chlorophytes. In contrast,
within the streptophytes, there has been a clear general trend toward increased
multicellular complexity and dominance of the diploid phase of the life cycle. The
common ancestor of the streptophytes was probably a unicellular organism with a
haploid life cycle. The diversification of the charophytes saw an increase in the
complexity of the haploid phase with the emergence of a complex, multicellular
haploid generation. The shift from haploid-dominated to diploid-dominated life
cycles began when the embryophytes emerged from within the charophytes, with
the acquisition of a multicellular diploid generation that tended to increase in
complexity as new taxa emerged through evolutionary time.
The red algae include both unicellular and multicellular species, but sexual cycles
have not been described for the unicellular species. With the exception of some
filamentous species, most multicellular red algae belong to one of the two most
recently evolved classes within the red algae, the Bangiophyceae or the
Florideophyceae. All of the species in these two classes have haploid-diploid life
cycles, although the detailed structure of the life cycle can be quite complicated. The
edible seaweed Pyropia yezoensis (formerly Porphyra yezoensis), a member of the
Bangiophyceae, alternates between a leaflike gametophyte and a microscopic, fila-
mentous sporophyte generation. The majority of florideophytes have complex “tri-
phasic” life cycles, with what can be considered to be two sporophyte generations.
Gamete fusion occurs on the female gametophyte, and the zygote grows to form the
first sporophyte generation (the cystocarp), a small “organism” that grows parasit-
ically on the gametophyte. The cystocarp releases spores that develop into the
second, free-living sporophyte generation (the tetrasporophyte), on which meiosis
occurs when mature. Such triphasic life cycles can be considered to be variants on
the standard haploid-diploid life cycle; the two sporophyte stages serve to multiply
this generation of the life cycle.
The brown algae (Phaeophyceae) have diverse life cycles ranging from haploid-
diploid life cycles (with various levels of dominance of the haploid and diploid
phases) to simple diploid life cycles (Bell 1997; Cock et al. 2013). Basal brown algal
lineages all have haploid-diploid life cycles, suggesting that the last common
ancestor of the brown algae also had a life cycle of this type (Silberfeld et al.
2010). The most developmentally complex brown algae are found in recently
evolved orders such as the Laminariales (kelps) and the Fucales, and there is marked
tendency within these orders for the diploid phase to be the dominant phase of the
life cycle. For example, the large thalli of kelps, which can attain up to 50 meters in
length in some species, correspond to the sporophyte generation while kelp game-
tophytes are microscopic, filamentous organisms. On the other hand, the
Ectocarpales, which are the sister order to the kelps, tend to be less developmentally
complex and exhibit diverse haploid-diploid life cycles that include both haploid-
and diploid-dominant cycles (i.e., cycles with two generations but with one gener-
ation larger than the other). The Fucales, which originated about 52–80 Mya (Kawai
6 S. Bourdareau et al.

et al. 2015) and have relatively large, complex thalli, have diploid life cycles. Hence,
although perhaps not as strongly marked as in the archaeplastid lineage, there
appears to be a tendency within the brown algae for diploid-dominant life cycles
to have been associated with the emergence of developmental complexity.
With the recent availability of well-supported phylogenies for the brown algae
(Silberfeld et al. 2010; Kawai et al. 2015), it has become clear that there has been
considerable switching between life cycle types over the course of the emergence of
this lineage (Cock et al. 2013). The brown algae therefore potentially represent an
interesting group in which to correlate life cycle structure with other parameters such
as environmental and ecological context.
Algae in other eukaryotic supergroups also exhibit various types of life cycle,
with, for example, most dinoflagellates having haploid life cycles and the occurrence
of haploid-diploid life cycles in the chlorarachniophytes, but the sexual life cycles of
many of these algae are unknown.

A Relationship Between Life Cycle Type and Degree

of Multicellular Complexity

If we focus on the two most developmentally complex eukaryotic lineages, the land
plants and the animals, there appears to be a strong correlation between dominance
of the diploid phase of the life cycle and the emergence of developmental complex-
ity. As mentioned above, the emergence of land plants corresponded to a gradual
reduction in the importance of the gametophyte generation and an increase in the
relative importance of the sporophyte. In animals, dominance of the diploid phase
was established very early, with the vast majority of these organisms having diploid
life cycles. There is also some evidence for a similar correlation in brown algae (the
third most complex group of multicellular organisms), with recently evolved, devel-
opmentally complex taxa showing a tendency toward diploid-dominant haploid-
diploid life cycles or diploid life cycles. No clear trend is observed in other
multicellular groups such as the red algae, but this may be because these organisms
exhibit lower levels of developmental complexity.
In order to understand the relationship between life cycle structure and the
evolution of multicellular complexity, it is important to take into account the possible
theoretical advantages and disadvantages of different types of life cycle. These
aspects are discussed in the following section (see Otto and Gerstein 2008 and
Coelho et al. 2007 and references therein for further details).

Theoretical Advantages and Disadvantages of Different Types

of Life Cycle

It has been proposed that diploid genomes may be advantageous in a number of

respects. The presence of two copies of each chromosome can result in masking of
recessive deleterious mutations, reducing the negative effects of mutations. Also,
Alternation of Generations in Plants and Algae 7

more genes are present, increasing the probability of advantageous mutations aris-
ing. Diploidy may also be important for long-lived multicellular organisms that have
to deal with rapidly evolving parasites in that a larger battery of alleles is available to
provide resistance. Also, because cell size is often correlated with ploidy, it may be
advantageous to be diploid if large cells are required (or, conversely, haploid if small
cells are advantageous). On a more mechanistic level, the presence of homologous
chromosomes in diploids provides a template for the repair of double-stranded DNA
breaks. Some of these proposed advantages, such as increased cell size or a possible
increased capacity to resist parasites, may be relevant to the emergence of complex
multicellularity.
As far as haploid genomes are concerned, while masking of deleterious mutations
in diploid genomes may be an advantage in the short term, the more effective
elimination of deleterious mutations from haploid genomes due to the absence of
masking may be advantageous in the long term. Similarly, although advantageous
mutations may have a lower probability of arising in a haploid (because there are
fewer gene copies), recessive advantageous mutations will be immediately benefi-
cial. Haploid genomes could also have an energetic advantage, as less resources are
required to replicate a smaller genome.
While these different advantages and disadvantages may help explain the dom-
inance of either diploid or haploid life cycles, they do not provide any explanation
for the emergence (or evolutionary stability) of haploid-diploid life cycles. A
possible advantage of haploid-diploid life cycles is that they reduce the cost of sex
(because sexual reproduction occurs over a period of two generations rather than
one). However, the cost of sex can also be reduced by increasing the amount of
asexual reproduction. Most attempts to explain the prevalence and stability of
haploid-diploid life cycles have concentrated on ecological considerations. Such a
life cycle may be advantageous, for example, if the two phases are able to exploit
different ecological niches, particularly if environmental conditions are variable.
Here “environmental conditions” can be understood in a broad sense, not only in
terms of the physical environment but also in terms of interactions with other
organisms within the ecosystem. For example, if the two phases of the life cycle
have different levels of susceptibility to a particular pathogen, life cycle alternation
could allow the organism to “escape” from an infection (the so-called Cheshire cat
strategy; Frada et al. 2008). Note that, while these hypotheses may explain the
existence of two generations, they do not explain why the two generations should
have different levels of ploidy. It has been proposed that, in some instances,
alternation between two generations may allow one generation to be optimized for
spore production (favoring dissemination) and the other for gamete production
(favoring gamete fusion) (Bell 1997), but this hypothesis is unlikely to apply to all
cases, particularly for isomorphic life cycles for example. It is possible, however,
that the level of ploidy of each generation is irrelevant and the main role of the life
cycle in these instances is to ensure a cyclic alternation between the two different
generations. In other words, in situations where it is advantageous for an organism to
alternate between two different forms, the pre-existing alternation between haploid
8 S. Bourdareau et al.

and diploid phases, inherent to all life cycles, may provide a good starting point for
the evolution of the two alternating variant forms.
A number of studies have attempted to test the predictions of the various
hypotheses discussed above (Otto and Gerstein 2008). For example, there is evi-
dence for unicellular organisms that masking of deleterious mutations in diploids can
make them better adapted to a mutagenic environment (but see below for multi-
cellular organisms). Similarly, haploid life cycles appear to be advantageous if
population sizes are large because more mutations tend to arise in the population,
but selection is limiting. As far as haploid-diploid life cycles are concerned, the
ecological roles of the two generations have also been studied for a number of taxa
across the different algal groups. For heteromorphic cycles, where the sporophyte
and gametophyte are morphologically different, the differences between the ecolog-
ical roles of each generation can be quite evident. However, even for isomorphic
haploid-diploid life cycles, where the sporophyte and gametophyte are morpholog-
ically similar, there are often subtle differences between the two generations that
result in them being better adapted to different niches.

Genetic Regulation of Life Cycle Transitions

In multicellular organisms it is crucial that the initiation and progress of multicellular

development be coordinated with the life cycle. Indeed, initiation of developmental
processes at the wrong stage of the life cycle could have catastrophic consequences.
The regulatory link between life cycle and development is still poorly understood,
but there have been some important advances over the last decade. When consider-
ing such systems, one obvious starting hypothesis is that the regulation of develop-
ment during the life cycle involves some sort of system that senses the level of ploidy
(DNA content) of the cell. However, there is currently no evidence to support such a
mechanism. For example, it has been shown for several different organisms that
experimental modifications of ploidy, such as the creation of tetraploids, do not
necessarily disrupt coupling between life cycle progression and development. These
observations indicated that the coupling of life cycle and development is more likely
under genetic control. Moreover, given that the different stages of a life cycle are all
produced from the same genome, the genetic components are expected to be
influenced by, and integrated with, epigenetic regulatory processes.
Genetic analyses of several organisms have identified key regulators associated
with syngamy (the step of the life cycle where gametes fuse to create a zygote
leading to a doubling of the chromosome number) (Goodenough and Heitman 2014;
Bowman et al. 2016). The green alga Chlamydomonas reinhardtii, for example,
produces gametes of two different mating types, called plus and minus gametes
(Fig. 3a). Two different three-amino acid length extension (TALE) homeodomain
transcription factors (TALE HD TFs) called gamete-specific plus 1 (GSP1) and
gamete-specific minus 1 (GSM1) are expressed specifically in the plus and minus
gametes, respectively. When a plus and a minus gamete fuse, during syngamy, these
two transcription factors are brought together in the same cell, the zygote. In the
Alternation of Generations in Plants and Algae 9

zygote, GSP1 and GSM1 form a heterodimer, which orchestrates the expression of
processes associated with the diploid phase of the life cycle (Lee et al. 2008).
Therefore, in C. reinhardtii, a simple genetic system allows the cell to detect when
there has been a transition from the haploid to the diploid state.
A similar system has been identified in the moss Physcomitrella patens
(Sakakibara et al. 2013; Horst et al. 2016). The C. reinhardtii proteins GSP1 and
GSM1 are members of the BELL and KNOX2 classes of TALE HD TFs, respec-
tively. Analysis of a P. patens strain carrying mutations in two KNOX2 TALE HD
TF genes, PpMKN1 and PpMKN6, showed that it produced a diploid gametophyte
instead of the sporophyte stage of the life cycle (Sakakibara et al. 2013). Similarly,
overexpression of the BELL TALE HD TF gene PpBELL1 resulted in apogamous
sporophytes (i.e., the production of haploid sporophytes without syngamy) (Horst
et al. 2016).
Interestingly, similar molecular systems have been described in another eukary-
otic supergroup, the fungi. In Cryptococcus neoformans, for example, gametes of the
α and a mating types express two different homeodomain transcription factors,
sex-inducer 1α and sex-inducer 2a, respectively. These transcription factors form a
heterodimer in the zygote and trigger sexual development, including basidium and
meiospore formation (Hull et al. 2005). There is therefore a recurring theme of
association of homeodomain transcription factors with the regulation of key life
cycle transitions across diverse eukaryotic supergroups. It is not clear at present
whether these similarities represent convergent evolution or if the different
homeodomain-based regulatory systems are derived from a common, ancestral
system that would therefore date back to the last eukaryotic common ancestor
(LECA; Fig. 1).
There is also direct genetic evidence for the involvement of epigenetic processes
in life cycle control. In P. patens, for example, knockout experiments indicate that
curly leaf (PpCLF) and fertilization-independent endosperm (PpFIE), which are
components of the chromatin-regulating polycomb repressive complex 2 (PRC2),
downregulate the expression of PpBELL1 during the gametophyte stage by tri-
methylating lysine 27 of histone H3 (H3K27me3) in nucleosomes at the PpBELL1
locus (Pereman et al. 2016). PRC2 proteins are not expressed in the zygote after
syngamy, and upregulation of PpBELL1 leads to the development of the sporophyte
generation, presumably through an interaction with PpMKN1 and PpMKN6 (Okano
et al. 2009; Horst et al. 2016) (Fig. 3b).

The Origins of Sporophyte and Gametophyte Developmental

Programs

To understand the emergence of multicellular complexity, it is often very important

to take into consideration the context of the life cycle. In the land plants, for example,
the increase in developmental complexity over evolutionary time was associated
with a transition from dominance of the haploid phase to dominance of the diploid
phase (Pires and Dolan 2012). There has been considerable debate as to whether the
 10

Fig. 3 Molecular regulators of life cycle progression. (a) The life cycle of the unicellular green alga Chlamydomonas reinhardtii showing the expression of the
S. Bourdareau et al.

TALE homeodomain transcription factors GSP1 and GSM1 in plus and minus gametes, respectively, and the formation of a GSP1/GSM1 heterodimer after
Alternation of Generations in Plants and Algae 11

emergence of the sporophyte generation in this lineage involved de novo evolution

of developmental pathways (the so-called “antithetic” hypothesis), or whether the
developmental plan was an adapted version of the gametophyte program (the
“homologous” hypothesis). Genomic approaches are starting to resolve this ques-
tion, and the emerging picture is that recruitment of regulatory networks from the
gametophyte generation played a very important role in this process, although there
have also been sporophyte-specific innovations such as the employment of TALE
homeodomain transcription factors of the KNOX2 family as developmental
regulators.

Consequences of Life Cycle Type on Genome Evolution

The life cycle of an organism is expected to have consequences for the evolution of
its genome. For example, in organisms with haploid-diploid life cycles, selection
should act more efficiently on genes expressed during the haploid phase because
recessive alleles of genes that are expressed during the diploid phase can be masked
by dominant alleles that are also present in the diploid genome (Otto and Gerstein
2008). There is evidence that this phenomenon of masking occurs in unicellular
organisms, but, surprisingly, it may not play an important role in multicellular
organisms. A recent analysis of two land plant species with haploid-diploid life
cycles, the angiosperm Arabidopsis thaliana and the moss Funaria hygrometrica,
did not find any evidence that diploid phase-specific genes evolved more rapidly
than haploid-phase-specific genes (Szovenyi et al. 2013). In fact, the evolution of life
cycle-regulated genes was found to be influenced more strongly by another factor:
breadth of expression. The strength of selection on a gene sequence is related to its
pattern of expression because a gene that is expressed in multiple tissues and at
multiple stages of development is exposed to selection more sustainedly than a gene
with a very restricted pattern of expression. In land plants at least, this phenomenon
appears to influence the evolution of life cycle-regulated genes more strongly than
the masking effect.

Fig. 3 (continued) gamete fusion. (b) Regulators of the gametophyte-to-sporophyte transition in

the moss Physcomitrella patens. Left panel: The polycomb repressive complex (PRC2) represses
expression of the TALE homeodomain transcription factor BELL1 (and MKN1/MKN6?) during the
gametophyte generation by laying down a repressive chromatin mark. Right panel: BELL1 and
MKN1/MKN6 are required for initiation of the sporophyte program, and this process probably
involves the formation of transcription factor heterodimers. Proteins are indicated by colored
shapes. Genes are indicated by italics
12 S. Bourdareau et al.

Conclusion

In this chapter, we have seen that plant and algal life cycles are highly varied and
often very complex. As far as the emergence of multicellularity is concerned, there
appears to be a correlation between the dominance of the diploid phase and multi-
cellular complexity, at least in the most developmentally complex groups such as
animals, land plants, and brown algae. The diversity of algae provides a rich source
of variation to test theoretical predictions about the relative advantages of different
types of life cycle. Algal systems are also providing exciting new insights into the
molecular mechanisms regulating life cycle progression and the evolutionary pro-
cesses that have led to the emergence of the sporophyte and gametophyte genera-
tions of the life cycle. These various themes illustrate the importance of life cycles as
key processes underlying important evolutionary transitions, including adaptations
to new environments and the evolution of multicellular complexity.

Glossary

Alga Photosynthetic eukaryotes, other than land plants

Diploid Phase of the life cycle with two sets of chromosomes
Epigenetic A change in gene expression that is not due to modifi-
cation of the DNA sequence of the genome
Gametophyte The gamete-producing generation of a plant or algal
life cycle
Generation The organism produced at each stage of a life cycle. We
use generation here to distinguish morphological/func-
tional stages of the life cycle such as the sporophyte
and the gametophyte from the ploidy phases (haploid
and diploid phases)
Haploid Phase of the life cycle with a single set of chromosomes
Haplodiploidy Sometimes used as a synonym for haploid-diploid life
cycles, but this term can lead to confusion because it is
also used to describe Hymenoptera life cycles that
involve development of haploid males from
unfertilized eggs and diploid females from fertilized
eggs (also called arrhenotoky)
Meiosis Cell division process that results in daughter cells that
contain half as many chromosomes as the parent cells.
Recombination between chromosomes during meiosis
generates new combinations of alleles in the chromo-
somes of the daughter cells
Phase Stage of a life cycle with a specific level of ploidy, e.g.,
the diploid or the haploid phase
Plant Macroscopic photosynthetic eukaryote. When used in a
taxonomic sense, this term refers to a member of the
Alternation of Generations in Plants and Algae 13

kingdom Plantae, equivalent to the modern taxonomic

group the Archaeplastida (Fig. 1)
Primary endosymbiosis Capture of a cyanobacterium by a eukaryotic cell and
enslavement to form a plastid
Secondary endosymbiosis Capture and enslavement of a photosynthetic eukaryote
by another eukaryotic cell leading to the production of
a secondary plastid
Sporophyte The spore-producing generation of a plant or algal life
cycle
Syngamy Fusion of gametes leading to doubling of the chromo-
some number in the resulting zygote

References
Bell G (1997) The evolution of the life cycle of brown seaweeds. Biol J Linn Soc 60:21–38
Bowman JL, Sakakibara K, Furumizu C, Dierschke T (2016) Evolution in the cycles of life. Annu
Rev Genet 50:133–154
Cock JM, Godfroy O, Macaisne N et al (2013) Evolution and regulation of complex life cycles: a
brown algal perspective. Curr Opin Plant Biol 17:1–6
Coelho S, Peters AF, Charrier B et al (2007) Complex life cycles of multicellular eukaryotes: new
approaches based on the use of model organisms. Gene 406:152–170
Frada M, Probert I, Allen MJ et al (2008) The “Cheshire cat” escape strategy of the coccolithophore
Emiliania huxleyi in response to viral infection. Proc Natl Acad Sci USA 105:15944–15949
Goodenough U, Heitman J (2014) Origins of eukaryotic sexual reproduction. Cold Spring Harb
Perspect Biol 6:a016154
Horst NA, Katz A, Pereman I et al (2016) A single homeobox gene triggers phase transition,
embryogenesis and asexual reproduction. Nat Plants 2:15209
Hull CM, Boily M-J, Heitman J (2005) Sex-specific homeodomain proteins Sxi1alpha and Sxi2a
coordinately regulate sexual development in Cryptococcus neoformans. Eukaryot Cell
4:526–535
John DM (1994) Alternation of generations in algae: its complexity, maintenance and evolution.
Biol Rev 69:275–291
Kawai H, Hanyuda T, Draisma SGA et al (2015) Molecular phylogeny of two unusual brown algae,
Phaeostrophion irregulare and Platysiphon glacialis, proposal of the Stschapoviales ord. nov.
and Platysiphonaceae fam. nov., and a re-examination of divergence times for brown algal
orders. J Phycol 51:918–928
Lee JH, Lin H, Joo S, Goodenough U (2008) Early sexual origins of homeoprotein hetero-
dimerization and evolution of the plant KNOX/BELL family. Cell 133:829–840
Leliaert F, Smith DR, Moreau H et al (2012) Phylogeny and molecular evolution of the green algae.
Crit Rev Plant Sci 31:1–46
Okano Y, Aono N, Hiwatashi Y et al (2009) A polycomb repressive complex 2 gene regulates
apogamy and gives evolutionary insights into early land plant evolution. Proc Natl Acad Sci
USA 106:16321–16326
Otto SP, Gerstein AC (2008) The evolution of haploidy and diploidy. Curr Biol 18:R1121–R1124
Pereman I, Mosquna A, Katz A et al (2016) The Polycomb group protein CLF emerges as a specific
tri-methylase of H3K27 regulating gene expression and development in Physcomitrella patens.
Biochim Biophys Acta 1859:860–870
14 S. Bourdareau et al.

Pires ND, Dolan L (2012) Morphological evolution in land plants: new designs with old genes.
Philos Trans R Soc Lond Ser B Biol Sci 367:508–518
Sakakibara K, Ando S, Yip HK et al (2013) KNOX2 genes regulate the haploid-to-diploid
morphological transition in land plants. Science 339:1067–1070
Silberfeld T, Leigh JW, Verbruggen H et al (2010) A multi-locus time-calibrated phylogeny of the
brown algae (Heterokonta, Ochrophyta, Phaeophyceae): investigating the evolutionary nature of
the “brown algal crown radiation”. Mol Phylogenet Evol 56:659–674
Szovenyi P, Ricca M, Hock Z et al (2013) Selection is no more efficient in haploid than in diploid
life stages of an angiosperm and a moss. Mol Biol Evol 30:1929–1939
Evolution of Symmetry in Plants

Catherine Damerval, Florian Jabbour, Sophie Nadot, and Hélène L.

Citerne

Abstract
Symmetry provides organisms with an efficient means to cope with physical
constraints and explore three-dimensional space. We describe the diversity and
evolution of symmetry types in the aerial parts of the major group of land plants,
the angiosperms. Two main types of symmetry occur: bilateral symmetry, where
structures can be divided into two mirror halves, and radial symmetry, with
multiple planes of symmetry. Different organ arrangements or phyllotactic pat-
terns produce different types of symmetry, which may vary within a plant’s life
span. Leaves are usually flat bilaterally symmetrical organs with a bifacial
organization resulting from an abaxial-adaxial differentiation associated with
photosynthetic activity. Alterations in the genetic pathway underlying this asym-
metry are thought to play a role in the repeated evolution of unifacial leaves.
Flowers are composed of a series of organs that are considered to be highly
modified leaves on a short compact axis. The symmetry of flowers as a whole is
one of the most studied traits in plant evolutionary developmental genetics.
Bilateral symmetry is derived from radial symmetry, probably from coevolution
with specialized pollinators. Nearly 200 transitions in floral symmetry types have
been recorded over the course of angiosperm evolution. Symmetry can change

C. Damerval (*) • H.L. Citerne

GQE – Le Moulon, INRA, Univ. Paris-Sud, CNRS, AgroParisTech, Université Paris-Saclay, Gif-
sur-Yvette, France
e-mail: damerval@moulon.inra.fr; h_citerne@hotmail.com
F. Jabbour
Muséum national d’Histoire naturelle, Institut de Systématique, Evolution, Biodiversité, UMR
7205 ISYEB MNHN/CNRS/UPMC/EPHE, Sorbonne Universités, Paris, France
e-mail: fjabbour@mnhn.fr
S. Nadot
Laboratoire Ecologie, Systématique et Evolution, UMR 8079 Université Paris-Sud/CNRS/
AgroParisTech, Orsay, France
e-mail: sophie.nadot@u-psud.fr

# Springer International Publishing AG 2017 1

L. Nuño de la Rosa, G.B. Müller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_59-1
2 C. Damerval et al.

during flower development, and the timing of this change can vary between
species. CYCLOIDEA-like transcription factors have been recruited repeatedly
for the control of floral bilateral symmetry in angiosperms. The establishment of
bilateral symmetry in leaves and flowers thus relies on different growth processes
and gene networks.

Keywords
Symmetry • Evo-devo • Leaf • Flower • CYCLOIDEA

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Phyllotaxis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Leaf Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Genetic Bases of Asymmetry Along the Abaxial-Adaxial Axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Symmetry Changes Along the Mediolateral Axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Floral Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Plant-Pollinator Interactions Have Fueled the Evolution of Floral Symmetry . . . . . . . . . . . . . . 10
Evolution of Floral Symmetry Across Angiosperms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Establishment of Floral Symmetry During Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
The Genetic Control of Floral Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Future Prospects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Introduction

The term symmetry comes from two Greek words: σύν (sún, meaning “with”) and
μEτρoν (métron, meaning “measure”) and originally indicated a relation of com-
mensurability. It quickly took on the more general meaning of equilibrium of pro-
portions, qualifying the harmony of the different elements of a unitary whole, thus
becoming closely related to the idea of beauty and regularity. It is a widely used
concept in physics and mathematics. In its current meaning, symmetry is defined in
terms of the invariance of an object under specified groups of rotations and reflec-
tions. Among the various types of symmetry that are mathematically defined, two
types of symmetry are appropriate for describing the phenotypes we consider in this
chapter: radial symmetry, corresponding to the repetition of a same structure around
a single axis of symmetry (n-fold rotational symmetry), and bilateral symmetry,
where a single symmetry plane divides the organism or structure in two mirror
images. An organism or a structure is asymmetrical when neither an axis nor a plane
of symmetry can be defined.
In living organisms, symmetry can theoretically be examined at every level of
complexity from cells to tissues, organs, or whole organisms. Symmetry is present in
most body plans, and may be a convenient and adaptive means for organisms to
better explore the three-dimensional space they live in. In biology, however, sym-
metry is approximate. For instance, bilaterally symmetrical bodies or organs do not
Evolution of Symmetry in Plants 3

have exactly identical mirror halves; indeed, the degree of difference between both
halves is often considered an indicator of the stability of development.
This chapter focuses on angiosperms (flowering plants), a group that represents
90% of extant land plant biodiversity. Angiosperm stem and root axes are usually
radially symmetrical and terminated by meristems (respectively, the shoot apical
meristem (SAM) and the root apical meristem) that ensure continuous growth and
organ formation. Three aspects of symmetry in the aerial parts of this group are
covered here: (1) phyllotaxis, the arrangement of plant organs, (2) leaf symmetry
(independently of leaf form, which is very diverse among angiosperms and beyond
the scope of this review), and (3) flower symmetry, the study of which has produced
the most abundant literature devoted to evolutionary developmental studies of
symmetry in plants.

Phyllotaxis

The SAM generates, in a sequential and regular order, units called phytomers
(Fig. 1a), each composed of a leaf, an axillary bud (which together form the node),
and a portion of the stem (the internode). In the same way, inflorescence meristems
generate flowers, and floral meristems generate floral organs. The resulting pattern of
arrangement is called phyllotaxis. The most common arrangements of leaves are
spiral (Fig. 2a), distichous (Fig. 2b–c), opposite (two leaves at a node), and verticil-
late (more than two leaves at a node). In the flower, the insertion of floral organs is
either spiral or verticillate (i.e., whorled), the latter being the most widespread. When
phyllotaxis is spiral, the divergence angle at which consecutive primordia are
generated by the SAM generally approaches 137.5 (the so-called “golden angle”).
Organs appear inserted on parastichies (i.e., secondary clockwise and anticlockwise
spirals obtained by linking the positions of adjacent organs), and the number of
parastichies matches two consecutive numbers of the Fibonacci series (Fig. 2a,
13 clockwise and 8 anticlockwise spirals). From a top view, the different phyllotactic
patterns can either be described as bilateral (for distichous phyllotaxis, Fig. 2b–c) or
radial (n-fold rotational symmetry with n = the number of parastichies for spiral
phyllotaxis, n = 2 for opposite phyllotaxis, and n = the number of organs for
verticillate/whorled phyllotaxis). The type of phyllotaxis is species specific, but
patterns of organ insertion may change during the life of the plant and with the
type of organ. For example, cotyledon and early leaf phyllotaxis are not necessarily
the same as adult leaf phyllotaxis, and the phyllotaxis of flowers in an inflorescence
is not necessarily in continuity with vegetative phyllotaxis.
Such regular arrangements have long fascinated not only botanists but also
mathematicians and physicists, and various models have been put forward to account
for the phyllotactic patterns found in plants. The hormone auxin has been shown to
be a major player in the establishment of phyllotactic patterns in Arabidopsis, with
local maxima at the shoot apical meristem triggering organ formation. The formation
of these maxima depends on the asymmetrical relocalization on the cell membrane of
efflux carrier proteins of the PIN-formed family. Recent experiments in Arabidopsis
4 C. Damerval et al.

a SAM
axillary bud

adaxial blade

proximal
mediolateral
Phytomer

distal
internode

abaxial

petiole

b
Proximal

Adaxial
AS1
HD-ZIPIII AS2 tasiR-ARF Leaf margin

Distal
WOX1
WOX3
main miR165/166 KANADI ARF YABBY Auxin
axis
Abaxial

c
main
axis
Adaxial

Abaxial
(i) (ii) (iii)

Fig. 1 Schematic representation of vegetative shoot structure and leaf abaxial-adaxial symmetry
types and underlying gene regulatory network of leaf polarity. For details concerning gene names
and interactions, see text. (a) Main components of a vegetative shoot (SAM = shoot apical
meristem). The three symmetry axes of a leaf are shown (in bold). (b) Cross section of the leaf
blade along the proximo-distal axis, showing the main genetic determinants and their interactions
(activation: pointed arrows, repression: T-shaped arrows) involved in abaxial-adaxial polarity and
blade outgrowth in Arabidopsis. (c) Changes in leaf symmetry displayed in cross section along the
mediolateral axis; i) conventional bifacial leaf with abaxial-adaxial differentiation and bilateral
symmetry along the mediolateral axis; (ii) unifacial abaxialized leaf with radial symmetry; (iii)
unifacial abaxialized leaf with bilateral symmetry along the abaxial-adaxial axis
Evolution of Symmetry in Plants 5

Fig. 2 Photos of leaves and flowers, illustrating various types of symmetry and organ arrangements
and interactions between flowers and pollinators. Leaves (a) Sempervirens sp. with a spiral phyllo-
taxis. (b) Celtis occidentalis (Cannabaceae), conventional bifacial bilaterally symmetrical leaf with an
alternate distichous phyllotaxis. (c) Iris pseudacorus (Iridaceae), unifacial bilaterally symmetrical leaf,
with an alternate distichous phyllotaxis. For flowers, radial (d–e) and bilateral symmetry (g–i) are
illustrated in the context of 3-merism, which is the basic organization in monocots (a, f), and 5-merism
which is the basic organization in core eudicots (b, g–i). (d) Aphyllanthes monspeliensis
(Asparagaceae). (e) Geranium maderense (Geraniaceae). (f) Ophrys apifera (Orchidaceae), zygomor-
phic flower with a highly modified ventral petal (labellum). (g) Stachys sylvatica (Lamiaceae),
zygomorphic lip flower (lower lip = 3; upper lip = 2 petals). (h) Pisum sativum subsp. elatius
6 C. Damerval et al.

show that the interplay between auxin distribution and local biomechanical con-
straints is a major determinant of phyllotactic patterns (Sassi and Traas 2015). Even
though genes whose mutations disrupt phyllotaxis have been characterized in
Arabidopsis, the genetic networks downstream of auxin signaling are still poorly
known, and even less is known regarding the bases of phyllotactic pattern diversity.

Leaf Symmetry

Leaves are determinate lateral organs that are generally differentiated at maturity
along three axes: proximo-distal, abaxial-adaxial (dorsoventral), and mediolateral
(Figs. 1a and 2a–c). Leaves are usually asymmetrical along the proximo-distal and
abaxial-adaxial axes and bilaterally symmetrical along the mediolateral axis. In most
monocots, leaves typically consist of a proximal sheath wrapped around the stem
and a distal flat blade; in other angiosperms, classical proximo-distal differentiation
consists of a proximal petiole and a distal flat blade. Flattening generally occurs
along the mediolateral axis and results in a high surface-to-volume ratio. It probably
evolved very early in the evolution of vascular plants as an adaptive trait for
optimizing photosynthesis. Flattening is tightly linked with specialization of the
two faces of the blade, creating bifacial leaves with differentiated adaxial (i.e.,
closest to the SAM) and abaxial sides (Fig. 1a). The adaxial side is the upper side
where light capture is optimized, whereas the abaxial side has a higher stomata
density enabling gas exchange and transpiration. Internal leaf anatomy is differen-
tiated with palisade mesophyll cells and xylem tissue on the adaxial side and spongy
mesophyll cells and phloem on the abaxial side. Microdissection experiments carried
out in the 1950s by Ian Sussex showed that a signal from the SAM is required for the
establishment of abaxial-adaxial leaf polarity. When communication between the
SAM and the young leaf primordium is disrupted, the leaf becomes abaxialized,
meaning that abaxial identity is the default identity.

Genetic Bases of Asymmetry Along the Abaxial-Adaxial Axis

The genetic network involved in abaxial-adaxial leaf polarity has been unraveled in
model species from the core eudicots and the monocots, in Arabidopsis and maize in
particular (Fig. 3; Yamaguchi et al. 2012; Fukushima and Hasebe 2014). Adaxial

Fig. 2 (continued) (Fabaceae), zygomorphic flag flower (flag = upper petal). (i) Tropaeolum majus
(Tropaeolaceae), zygomorphic flower with a spur borne on the dorsal sepal. Interactions between
flowers and insects: (j) Trichius fasciatus (Diptera) feeding on pollen from Angelica sylvestris
(Apiaceae), with the umbel forming a landing platform. (k) Bombus sp. (Hymenoptera) feeding
nectar and/or pollen from a zygomorphic flower of Salvia sp. (l) Pieris rapae (Lepidoptera) feeding
on nectar from a zygomorphic flower of Lavandula angustifolia (Lamiaceae) (Photographs
S. Nadot, C. Damerval and A. Decourcelle)
Evolution of Symmetry in Plants 7

Caryophyllales

Asteridae
Asteraceae
Lamiales - Antirrhinum,
Plantago, Stachys
Lonicera

Berberidopsidales

Santalales
Core Eudicots Saxifragales

Vitales

EUDICOTS

Arabidopsis
Celtis
Rosidae
Geranium
Iberis
Lathyrus
Tropaeolum

Dilleniales

Gunnerales

Buxales
Trochodendrales

Proteales Proteaceae
Solanales
Ranunculaceae:
Ranunculales
Ceratophyllales
Nigella, Delphinieae

Acorus
Allium
Monocots

Aphyllanthes
Commelinaceae
Iris
Juncus
Orchidaceae - Ophrys
Poaceae - Oryza, Zea
Chloranthales
Magnoliids

Aristolochia

Austrobaileyales
Nymphaeales
Amborellales

Fig. 3 Simplified angiosperm phylogeny, indicating all genera discussed in the text
8 C. Damerval et al.

determinants include genes from two families of transcription factors: ARP (ASYM-
METRIC LEAVES1 (AS1), ROUGH SHEATH2 (RS2), PHANTASTICA (PHAN)) and
HD-ZIPIII (class III of homeodomain-leucine zipper). The action of the ARP genes
is probably mediated by interacting partners. The role of PHAN in adaxial identity
has been demonstrated in snapdragon (Antirrhinum majus), but loss-of-function
mutants in Arabidopsis of the PHAN orthologue AS1 do not have a consistently
abaxialized phenotype, while mutants in maize of the PHAN orthologue RS2 do not
have any clear abaxial-adaxial polarity phenotype. In Arabidopsis, AS1 and its
unrelated partner AS2 (ASYMMETRIC LEAVES2) positively regulate HD-ZIPIII
factors. HD-ZIPIII genes are expressed in the adaxial region and are targeted by
regulatory miRNAs (miR165 and miR166) expressed in the abaxial region. Unlike
the ARP genes, the role of HD-ZIPIII genes as necessary and sufficient factors of
adaxial identity appears conserved between Arabidopsis and Poaceae. Abaxial
determinants comprise members of the KANADI family and auxin response factors
(ARF). KANADI transcription factor genes are expressed in a domain complemen-
tary to the HD-ZIPIII genes. They are also necessary and sufficient to promote
abaxial identity and are broadly conserved among angiosperms. Both ETTIN/ARF3
and ARF4 in Arabidopsis are negatively regulated by a small RNA (tasiR-ARF).
Antagonistic interactions involving both transcription factors and noncoding small
RNAs are important for determining and stabilizing a sharp boundary between the
abaxial and adaxial domains, along which the blade will grow (Fig. 1b).
The precise mechanisms acting downstream of the polarity genes and controlling
blade outgrowth are not fully understood yet. It has been shown that YABBY
transcription factor genes, previously thought to be abaxial determinants, play
instead a primary role in blade outgrowth. However, their patterns of expression
can differ between species, suggesting the pathway they control is probably not
conserved among angiosperms. Genes belonging to two classes of another transcrip-
tion factor gene family, the WOX1 and WOX3/PRESSED FLOWER subfamilies,
which are expressed in the leaf margin and along the abaxial-adaxial juxtaposition
region, also play an essential role, at least in Arabidopsis. Auxin also appears to be a
signal promoting blade outgrowth (Fig. 1b).

Symmetry Changes Along the Mediolateral Axis

Bifaciality and bilateral symmetry along the mediolateral axis are thought to be the
ancestral states for the leaf in angiosperms (Harrison et al. 2002). Unifacial leaves,
with no abaxial-adaxial differentiation, have evolved multiple times and are fre-
quently found in monocots (Fig. 3) where they nonetheless retain a bifacial sheath.
The shift to unifaciality of the blade can lead to a change in leaf symmetry from
bilateral to radial, as found in species of Allium or Juncus. In other cases, the blade is
flattened resulting in bilateral symmetry (e.g., in Acorus, Iris (Fig. 2c), and some
species of Juncus). Flattening in unifacial leaves is different from flattening in
bifacial leaves in that it involves important cell proliferation toward the SAM side,
generating a symmetry plane that is perpendicular to the symmetry plane of bifacial
Evolution of Symmetry in Plants 9

leaves (Fig. 1c). Anatomical studies suggest that unifacial leaves have abaxial
characteristics without any differentiation of the epidermal and mesophyll cells
and with vascular bundles arranged in a circle with an inner-outer orientation of
xylem poles and phloem tissues. A comparative study of two species of Juncus with
unifacial leaves and with either radial (J. wallichianus) or bilateral
(J. prismatocarpus) symmetry showed that in both species gene expression in the
blade tissue was typical of the abaxial domain of the bifacial leaf of rice (Yamaguchi
et al. 2010). However, a YABBY gene (DROOPING LEAF) was found to be
specifically expressed in the proliferative zone close to the SAM only in
J. prismatocarpus with bilaterally symmetrical leaves. In rice, this gene is expressed
in the medial part of leaf primordia where it promotes midrib thickening. Later on in
J. prismatocarpus, proliferative activity at the leaf margin appears to be correlated
with the expression of a WOX1 gene. These results suggest that part of the genetic
circuitry involved in blade outgrowth has been conserved and reused for leaf
flattening in Juncus. Whether similar processes are involved in the independent
evolution of flattened unifacial leaves in other groups remains to be determined.

Floral Symmetry

Flowers form a reproductive unit composed of a series of lateral organs that are
considered to be highly modified leaves. Sterile organs surround the fertile organs on
a short axis, following an almost invariable order: stamens, producing the male
gametophytes (pollen grains), surround the carpels, producing the ovules (each
containing a female gametophyte). The sterile organs (frequently a greenish outer
calyx (sepals) and a showy inner corolla (petals)) collectively form the perianth and
provide protection for the developing fertile organs and attractiveness in the case of
animal pollination. The basic number of each type of organ is known as “merism,”
which is one component of the floral ground plan. Symmetry is another component
of floral descriptions and has been used since Theophrastus (371–287 BC). It was
recognized very early on that flowers present an overall symmetry that can be radial
(actinomorphy – Fig. 2d–e) or bilateral (zygomorphy – Fig. 2f–i) whatever the
merism, and that this character could be used, along with other characters, to produce
a classification system of flowering plants. Strictly speaking the terms actinomorphy
and zygomorphy should only apply to flowers with organs inserted in whorls, but in
practice they are extended to flowers with spirally inserted organs. Indeed, as in
whorled flowers, it is possible to identify discrete sets of organ types in spiral
flowers, even though such flowers are more prone to display organs of intermediate
identity. Although symmetry can be defined for each organ whorl (or pseudo-whorl),
in practice it is generally applied to the perianth, especially the corolla, and to the
androecium, which often undergoes a reduction in the number of functional stamens
in zygomorphic flowers. Zygomorphy can be more or less elaborate, involving organ
displacement around the receptacle or strong differentiation along the dorsoventral
axis generating different morphologies within a single whorl (e.g., Fig. 2f). Zygo-
morphic flowers can also vary in the placement of their fertile organs, toward the
10 C. Damerval et al.

upper part of the flower (in lip flowers, for example, Fig. 2g) or the lower part of the
flower (in flag flowers, for example, Fig. 2h) (Endress 1994). Protruding structures
such as spurs or pouches storing nectar can also contribute to zygomorphy (Fig. 2i).
In actinomorphic and zygomorphic flowers, individual perianth organs are usually
bilaterally symmetrical along their mediolateral axis but asymmetry also exists, for
example, in the dorsal and lateral petals of snapdragon or lateral and ventral petals of
zygomorphic Fabaceae (Fig. 2h). As lateral organs, floral organs also exhibit
proximo-distal and abaxial-adaxial asymmetries, an illustration of which are the
conical cells typically found on the adaxial surface of petals that play a major role in
pollinator attraction (Glover and Martin 2002).

Plant-Pollinator Interactions Have Fueled the Evolution of Floral

Symmetry

Pollinators are believed to be major agents in the evolution of zygomorphy. The

fossil record suggests that zygomorphy evolved from actinomorphy around 50 mil-
lion years after the emergence of angiosperms, in coincidence with the diversifica-
tion of specialized insect pollinators (Crepet and Niklas 2009). The theory goes that
zygomorphic flowers coevolved with specialized insects toward a better placement
of their reproductive organs with respect to pollinator body shape, maximizing
pollen transfer and resulting in efficient cross-pollination (Fig. 2j–l). Actinomorphic
flowers are accessible from all sides (in front view) and are typically pollinated by a
wide range of pollinators. By contrast, zygomorphic flowers are more specialized,
even though they can be visited by a range of species depending on the environ-
mental context. The shape and size of the flower have been hypothesized to be under
stabilizing selective pressure within species with zygomorphic flowers, in order to
ensure the best success rate of cross-pollination by specific pollinators and subse-
quent seed set. Consistent with this hypothesis, several studies have shown that
zygomorphic flowers are less variable in size than actinomorphic ones (reviewed in
Citerne et al. 2010). A drawback of narrow specialization in zygomorphic flowers is
that it can increase species vulnerability if environmental factors lead to a mismatch
between blooming period and pollinator activity.

Evolution of Floral Symmetry Across Angiosperms

Multiple lines of evidence suggest that perianth actinomorphy is the ancestral state of
the flower. This view is supported by the fossil record, in which actinomorphic
flowers predate zygomorphic flowers (Crepet and Niklas 2009), as well as by
phylogeny. Increasingly complete and robust molecular phylogenies of angiosperms
(http://www.mobot.org/MOBOT/research/APweb/) now provide a backbone of the
tree, which is almost completely resolved at the family level. This means that
phylogenetic relationships among the 64 orders and 416 families are almost
completely known (Fig. 3) and that the main events of diversification within
Evolution of Symmetry in Plants 11

angiosperms have been dated. The most comprehensive estimate to date found
nearly 200 transitions in floral symmetry across the whole of the angiosperms
(Reyes et al. 2016), almost three times the previous estimate of 70 (Citerne et al.
2010). For instance, in Proteaceae alone, a basal eudicot family of ca. 1700 species,
zygomorphy has been found to have evolved up to 18 times independently from an
actinomorphic ancestor, with four reversals to actinomorphy.
The relative importance of historical constraints and selection pressures on the
evolution of symmetry can be evaluated statistically by analyzing the co-occurrence,
or lack of co-occurrence, of traits in a phylogenetic context. For example, in the
Asteridae clade, there is a negative evolutionary relationship between perianth
zygomorphy and stamen number, as revealed by the very rare occurrence of zygo-
morphic flowers with more than ten stamens (Jabbour et al. 2008). Recently, it has
been shown that zygomorphy, reduced stamen number, and the presence of a corolla
are three character states that act synergistically as a key innovation (O’Meara et al.
2016).
The concept of morphospace provides a novel way (through a mathematical
approach based on the ordination of morphological variables) of apprehending the
evolution of floral symmetry in different morphological contexts (Chartier et al.
2014). The representation of all trait combinations observed in nature within a space
of possible morphologies gives an idea of the possible constraints limiting the
evolution of types of flower symmetry.

Establishment of Floral Symmetry During Development

Studies of flower development describe and analyze the processes of organogenesis

(organ formation) and morphogenesis (shape acquisition) during the time required
for a floral meristem to develop into a fully formed structure ready to bloom. Flower
morphology, including symmetry, is generally described at anthesis, when the flower
has just opened and is considered to be adult.
The meristem is a changing structure and the type of symmetry recorded at any
given developmental stage may be transitory. As a result, symmetry can potentially
change several times during development. The symmetry of the early meristem may
differ from that of the bud after all organs are initiated and from that of the flower at
anthesis. For example, in snapdragon, the early meristem is bisymmetrical (with two
perpendicular symmetry planes); the bud is near actinomorphic at sepal initiation but
after that point becomes and remains zygomorphic (Vincent and Coen 2004). At
anthesis, androecium and corolla are strongly zygomorphic and dorsal and lateral
petals are internally asymmetrical, while the ventral petal is bilaterally symmetrical
(Fig. 4a). The description of flower symmetry therefore depends on the organ or set
of organs considered and on the developmental stage of the flower.
In clades where actinomorphy is ancestral and predominant, it is often observed
that in zygomorphic species, zygomorphy is established relatively late during devel-
opment (Endress 1999). For example, in Ranunculaceae, zygomorphy evolved once
from actinomorphy in the ancestor of tribe Delphinieae. Flower buds are
12 C. Damerval et al.

b
a IM Dorsal
CYC/DICH DIV
dorsoventral axis

RAD DRIFs
dorsal

RAD DIV
Lateral

DRIFs DIV
lateral

DRIFs

ventral Ventral

Fig. 4 Illustration of corolla zygomorphy and regulatory gene network specifying petal identity in
snapdragon (Antirrhinum majus). For details concerning gene names and interactions, see text. (a)
Diagram of individually dissected petals. The inflorescence meristem (IM) is shown, which
determines the dorsoventral axis along which the petals are differentiated. (b) The regulatory
gene network determining the three types of petal identity: dorsal, lateral, and ventral

actinomorphic throughout organogenesis. They become zygomorphic late in devel-

opment, i.e., after carpel initiation. Dorsal petal primordia develop into spurred
petals, whereas the development of ventral petals is arrested shortly after initiation
(Fig. 5). In addition, in inflorescences with mostly actinomorphic flowers, it was
noted that zygomorphy is established relatively late in the peripheral flowers (e.g., in
Asteraceae, Ren and Guo 2015).
Developmental studies provide the basis for formulating evolutionary hypotheses
based on comparative analyses. Studying the establishment of symmetry during
flower development makes it possible to identify the pivotal stages at which sym-
metry changes or is constrained (e.g., by the position of the meristem on the
inflorescence or by the merism/meristem size ratio). This is necessary for identifying
the developmental stages at which molecular investigations should be made to
understand how morphological transitions relate to changes in gene expression.

The Genetic Control of Floral Symmetry

The widespread distribution across flowering plants of peloric (radially symmetrical)

forms of normally zygomorphic species suggests that floral symmetry is controlled
by a few key developmental regulators. Twenty years ago, research groups led by
Enrico Coen in the UK and Jorge Almeida in Portugal unraveled a network of key
developmental genes controlling floral symmetry in snapdragon. In this species,
unequal corolla and stamen development along the dorsoventral axis depends on the
activity of four genes: CYCLOIDEA (CYC), DICHOTOMA (DICH), RADIALIS
Evolution of Symmetry in Plants 13

Nigella damascena

Wild a b c d e
Teratum

f g h i j

Consolida regalis

k l m n o
Development

Fig. 5 Developmental series in the flower of Nigella damascena and Consolida regalis
(Ranunculaceae), showing changes in symmetry according to the stage, organ and level of analysis.
(a–j) Nigella damascena wild-type (a–e) and teratological (f–j) flower morphs. In both morphs,
organ primordia initiate on a spiral (a–b, and f–g), resulting in a pseudo-actinomorphic meristem. In
the wild-type morph, the corolla looks actinomorphic (c), although the petals are spirally inserted on
the meristem (c). In the teratological flower morph without petals (h), many sepal-like organs are
initiated and the phyllotaxis is also spiral. Flower development as investigated with scanning
electron microscopy shows that the pre-anthetic flower is pseudo-actinomorphic. A study of flower
anatomy shows that the wild-type morph is pseudo-actinomorphic, the phyllotaxis of the calyx and
the androecium being clearly spiral (d). The teratological flower morph is internally (from the
anatomical aspect) asymmetrical (spiral phyllotaxis, single bract, three sepals) (i). Wild-type and
teratological anthetic flowers are actinomorphic (e, j). (k–o) Consolida regalis. At early develop-
mental stages, organogenesis is spiral (k–l). The bud is zygomorphic due to the corolla reduced to a
single petal. The latter has a delayed development compared with the stamens (l) and becomes
spurred (m) when organogenesis is completed. The transverse section (n) clearly shows the bilateral
symmetry of the floral bud. Note the hollow spur of the dorsal petal nested in the spur of the dorsal
sepal. Strictly speaking, the floral bud is pseudo-zygomorphic as the perianth organs and the
stamens are spirally initiated. The anthetic flower (o) is zygomorphic

(RAD), and DIVARICATA (DIV). These four genes form two groups that act antag-
onistically to determine regional identities in the floral meristem. The first group
(CYC-DICH-RAD) determines dorsal identity. CYC and DICH are close paralogues
belonging to the TCP gene family of plant-specific transcription factors. Both genes
are expressed in the dorsal region of the floral meristem prior to organogenesis; their
expression becomes restricted to the dorsal petals and staminode (a nonfunctional
and underdeveloped stamen) (CYC) and to the dorsal half of the dorsal petals (DICH)
later in development. In the dorsal staminode, CYC has a negative effect on cell
proliferation by repressing cell cycle genes. The activity of CYC and DICH is
mediated by RAD, a MYB-class transcription factor. CYC and DICH can potentially
14 C. Damerval et al.

bind directly to RAD, inducing its expression in the dorsal region of the developing
flower. Ventral identity is specified by DIV, another MYB transcription factor. DIV
activity is controlled posttranscriptionally in the dorsal and lateral floral organs by
RAD, which outcompetes DIV for MYB-related DRIFs (DIV and RAD interacting
factors) (Fig. 4b).
Whether this network, or some of its components, is conserved in other plant
species has been a major focus of research over the past 20 years. It has been shown
that some of these key molecular players, and in particular CYCLOIDEA-like genes
(hereafter CYC-like genes), have been recruited repeatedly for the evolution of
zygomorphy in diverse lineages. Most studies to date have been conducted in the
large core eudicot clade, to which snapdragon belongs (Fig. 3). In core eudicots, a
pathway mediated by genes from the CYC2 clade, a core eudicot-specific CYC clade
(Howarth and Donoghue 2006), is involved in the control of floral symmetry. In
clades where zygomorphy evolved independently (in both Asteridae and Rosidae),
asymmetrical CYC2 expression has been correlated with unequal floral development
along the dorsoventral axis, a role that has been corroborated by functional studies in
a few model core eudicot species. Although CYC2 genes have been recruited
repeatedly in core eudicots for the control of floral bilateral symmetry, these genes
have distinct evolutionary histories in these different lineages, since they have
undergone independent duplications that can potentially be the raw material for
functional diversification. Independently derived lineage-specific paralogues can
differ in the way their roles are partitioned, in their effect on growth and in their
region of activity. For instance, asymmetrical CYC2 expression in core eudicots is
primarily on the dorsal side of the flower (sometimes expanding into the lateral
region), but there are exceptions where asymmetrical expression is on the ventral
side. The effect on growth is also variable: for instance, in Iberis amara, a close
relative of Arabidopsis, persistent asymmetrical dorsal expression of CYC in devel-
oping flowers reduces petal growth (Busch and Zachgo 2007), unlike in snapdragon
where it promotes growth in the later stages of floral development. Ectopic expres-
sion in Arabidopsis flowers of CYC from I. amara and snapdragon has opposite
effects: petal growth reduction for the former and growth enhancement for the latter,
indicating that protein function has diverged between the two (Busch and Zachgo
2007). It is not clear what the ancestral expression pattern of CYC2 may be, because
radial species not derived from zygomorphic ancestors exhibit either persistent radial
CYC2 expression, or transient early expression that can be asymmetric, as in
Arabidopsis (Cubas et al. 2001). Nevertheless, it appears that repeated spatiotem-
poral shifts in CYC2 expression underlie the evolution of asymmetrical floral
development along the symmetry plane in core eudicot lineages.
The conservation and co-option of RAD and DIV functions in lineages outside of
the core Lamiales (the order to which A. majus belongs) are still unclear even though
asymmetrical expression of one or the other, or both, has been observed in species
from other clades within Asteridae. In core Lamiales, asymmetrical co-expression of
RAD and CYC in the dorsal/lateral regions of the flower may have evolved from a
persistent radial expression of both genes, as found in actinomorphic early-diverging
Lamiales (Zhong and Kellogg 2015). However, in Arabidopsis, endogenous RAD-
Evolution of Symmetry in Plants 15

like genes could not be activated by CYC (Baxter et al. 2007), suggesting that RAD
genes may not be co-opted in the floral symmetry pathway of Rosidae.
Secondary loss of zygomorphy within diverse lineages of core eudicots is asso-
ciated with spatiotemporal changes in CYC2 expression, but not generally with gene
degeneration and loss, suggesting that these genes may have other functions. Two
main types of changes leading to a loss of asymmetrical expression are observed:
absent or downregulated expression (often with residual expression early in devel-
opment) or expansion of the expression domain to all regions of the floral whorl.
However, in core Lamiales, secondary loss of zygomorphy is associated not only
with changes in CYC expression but also with a breakdown of the CYC-RAD-DIV
regulatory pathway. This is found in the wind-pollinated flowers of Plantago where
one CYC paralogue and RAD are lost and where changes in expression of the
remaining single CYC copy (across the flower bud) (Preston et al. 2011) and DIV
(regulating stamen rather than petal development) (Reardon et al. 2014) are corre-
lated with its derived radial symmetry.
Outside of the core eudicots, asymmetrical dorsoventral expression of CYC-like
genes in developing flowers has been described in certain zygomorphic basal
eudicots, monocots, and in the magnoliid genus Aristolochia (Fig. 3). In the latter,
asymmetrical dorsoventral expression is evident only in late floral development,
suggesting that the early establishment of zygomorphy may be under the control of
other genetic factors (Horn et al. 2015). By contrast in orchids, where floral structure
and bilateral symmetry are particularly elaborate (Fig. 2f), CYC-like genes do not
appear to play an important role in the floral symmetry pathway. The development of
distinct organs within the perianth (outer and inner tepals and lip) is linked to
differential expression of B-class MADS-box DEFICIENS-like (DEF-like) genes.
The perianth organs are specified by different combinations, known as the “orchid
code,” and relative expression levels of four DEF-like paralogues (Mondragón-
Palomino and Theissen 2011). The involvement of MADS-box genes in the floral
symmetry pathway may be conserved among monocots, since asymmetrical expres-
sion patterns have been recorded in Commelinaceae and Poaceae. Their interactions
with CYC-like genes, if any, are currently unknown.
CYC-like genes, and in certain cases MADS-box genes, appear to be the central
players for asymmetrical floral development in angiosperms, subdividing organ
identity within a whorl (Fig. 4a). In independent acquisitions of zygomorphy,
novel interactions of these key genes with different targets may have evolved,
providing scope for the morphological variation displayed by zygomorphic flowers.
The factors regulating CYC-like genes and the role of CYC-like genes in actinomor-
phic species are still poorly known. Such knowledge would help us understand what
make these genes prone to repeated recruitment for the control of zygomorphy.
16 C. Damerval et al.

Future Prospects

Angiosperm leaves evolved and diversified from leaves of the ancestral lineage of
seed plants emerging in the Late Devonian-Early Carboniferous. Bilateral symmetry
along the mediolateral axis is ancestral in leaves. Physical factors impacting photo-
synthesis (e.g., light, atmospheric CO2 concentration, heat dissipation) have proba-
bly been instrumental in the evolution of flattening and abaxial-adaxial
differentiation promoting bilateral symmetry. By contrast, bilateral symmetry in
flowers is a derived state within angiosperms, resulting from an ongoing process
of coadaptation between pollinators and flowers and postdated by 50 million years
the origin of angiosperms in the Early Cretaceous.
It is not surprising that the control of symmetry within individual lateral organs,
be it leaves or floral organs, seems to share certain features (auxin fluxes, expression
of similar transcription factors, establishment of boundaries by antagonistic molec-
ular interactions), which coexist with particular determinants of organ identity.
While asymmetry of individual flower organs along the proximo-distal and
abaxial-adaxial axes might be controlled by genetic circuitries similar to those of
leaves, symmetry along the mediolateral axis in petals seems to result from different
processes, involving CYC-like genes but also other factors. The interplay of several
actors, including the activity of key transcription factors, even though they differ in
the two structures, is a common feature of the control of bilateral symmetry in leaves
and flowers. The parallel recruitment of key regulators from the same developmental
program is a feature of biological evolution, and in plants, it has been described for
the independent evolution of leaves, roots, and C4 photosynthesis, for example.
Indeed, present knowledge points to the repeated co-option of CYC-like genes for
bilateral symmetry in flowers. Studies investigating independent shifts to either
radial or bilateral symmetry in unifacial leaves are still too scarce to know if a
similar developmental program has been repeatedly co-opted.
Deciphering gene regulatory networks in a comparative framework has become
the next stage for understanding the origin of morphological diversity. New tech-
nologies will help us progress toward this goal and renew our approach of evolu-
tionary developmental questions in the near future. Developmental studies are
fundamental for identifying the crucial stages where the underlying molecular
changes are predicted to take place. New nondestructive imaging methods such as
the recently developed X-ray micro-computed tomography give access to the three-
dimensional organization of developing flowers, enabling the observation of rare or
difficult material. Additionally, live-imaging techniques such as light sheet fluores-
cence microscopy can be used in some species and organs. With new generation
sequencing techniques, genome and/or transcriptome sequencing has become feasi-
ble for virtually any species, challenging the candidate gene approach and giving
access to potentially novel genes affecting the trait of interest. Techniques such as
virus-induced gene silencing are feasible in a large panel of species, enabling
functional validation. In addition, better resolved phylogenies and new theoretical
and analytical approaches for reconstructing micro- and macroevolutionary patterns
will help uncover the driving forces and constraints in phenotype evolution.
Evolution of Symmetry in Plants 17

Acknowledgments We acknowledge Thierry Deroin for helpful discussions on flower anatomy

and development and careful reading of the manuscript.

Cross-References

▶ Evo-devo of the Origin of Flowering Plants

▶ Evolution of Development in Monocots
▶ Evolution of Floral Organ Identity
▶ Evolution of Photosynthesis: Atmospheric Composition and the Optimisation of
Water Use Efficiency

References
Baxter CE, Costa MM, Coen ES (2007) Diversification and co-option of RAD-like genes in the
evolution of floral asymmetry. Plant J 52:105–113
Busch A, Zachgo S (2007) Control of corolla monosymmetry in the Brassicaceae Iberis amara.
Proc Natl Acad Sci USA 104:16714–16719
Chartier M, Jabbour F, Gerber S, Mitteroecker P, Sauquet H, von Balthazar M, Staedler Y, Crane
PR, Schönenberger J (2014) The floral morphospace – a modern comparative approach to study
angiosperm evolution. New Phytol 204:841–853
Citerne H, Jabbour F, Nadot S, Damerval C (2010) The evolution of floral symmetry. In: Kader JC,
Delseny M (eds) Advances in botanical research, vol 54. Elsevier, London, pp 85–137
Crepet WL, Niklas KJ (2009) Darwin’s second “abominable mystery”: why are there so many
angiosperm species? Am J Bot 96:366–381
Cubas P, Coen E, Zapater JM (2001) Ancient asymmetries in the evolution of flowers. Curr Biol
11:1050–1052
Endress PK (1994) Diversity and evolutionary biology of tropical flowers. Cambridge University
Press, Cambridge, UK
Endress PK (1999) Symmetry in flowers: diversity and evolution. Int J Plant Sci 160:S3–S23
Fukushima K, Hasebe M (2014) Adaxial-abaxial polarity: the developmental basis of leaf shape
diversity. Genesis 52:1–18
Glover BJ, Martin C (2002) Evolution of adaptive petal cell morphology. In: Cronk QCB, Bateman
RM, Hawkins JA (eds) Developmental genetics and plant evolution. Taylor & Francis, London,
pp 160–172
Harrison CJ, Cronk QCB, Hudson A (2002) An overview of seed plant leaf evolution. In: Cronk
QCB, Bateman RM, Hawkins JA (eds) Developmental genetics and plant evolution. Taylor &
Francis, London, pp 395–403
Horn S, Pabón-Mora N, Theuß VS, Busch A, Zachgo S (2015) Analysis of the CYC/TB1 class of
transcription factors in basal angiosperms and magnoliids. Plant J 81:559–571
Howarth DG, Donoghue MJ (2006) Phylogenetic analysis of the “ECE” (CYC/TB1) clade reveals
duplications predating the core eudicots. Proc Natl Acad Sci USA 103:9101–9106
Jabbour F, Damerval C, Nadot S (2008) Evolutionary trends in the flowers of Asteridae: is
polyandry an alternative to zygomorphy? Ann Bot 102:153–165
Mondragón-Palomino M, Theißen G (2011) Conserved differential expression of paralogous
DEFICIENS- and GLOBOSA-like MADS-box genes in the flowers of Orchidaceae: refining
the ‘orchid code’. Plant J 66:1008–1019
O’Meara B, Smith SD, Armbruster WS, Harder LD, Hardy CR, Hileman LC, Hufford L, Litt A,
Magallón S, Smith SA, Stevens PF, Fenster CB, Diggle PK (2016) Non-equilibrium dynamics
and floral trait interactions shape extant angiosperm diversity. Proc R Soc B 283:2015304
18 C. Damerval et al.

Preston JC, Martinez CC, Hileman LC (2011) Gradual disintegration of the floral symmetry gene
network is implicated in the evolution of a wind-pollination syndrome. Proc Natl Acad Sci USA
108:2343–2348
Reardon W, Gallagher P, Nolan KM, Wright H, Cardeñosa-Rubio MC, Bragalini C, Lee C,
Fitzpatrick DA, Corcoran K, Wolff K, Nugent JM (2014) Different outcomes of the MYB floral
symmetry genes DIVARICATA and RADIALIS during the evolution of derived actinomorphy in
Plantago. New Phytol 202:716–725
Reyes E, Sauquet H, Nadot S (2016) Floral symmetry changed at least 199 times in angiosperms.
Taxon 65:945–964
Ren JB, Guo YP (2015) Behind the diversity: ontogenies of radiate, disciform, and discoid capitula
of Chrysanthemum and its allies. J Syst Evol 53:520–528
Sassi M, Traas J (2015) When biochemistry meets mechanics: a systems view of growth control in
plants. Curr Op Plant Biol 28:137–143
Vincent CA, Coen ES (2004) A temporal and morphological framework for flower development in
Antirrhinum majus. Can J Bot 82:681–690
Yamaguchi T, Yano S, Tsukaya H (2010) Genetic framework for flattened leaf blade formation in
unifacial leaves of Juncus prismatocarpus. Plant Cell 22:2141–2155
Yamaguchi T, Nukazuka A, Tsukaya H (2012) Leaf adaxial-abaxial polarity specification and
lamina outgrowth: evolution and development. Plant Cell Physiol 53:1180–1194
Zhong J, Kellogg EA (2015) Stepwise evolution of corolla symmetry in CYCLOIDEA2-like and
RADIALIS-like genes expression patterns in Lamiales. Am J Bot 102:1260–1267
Evolution of Complexity

Daniel W. McShea

Abstract
To study the evolution of complexity in organisms, we need an understanding of
complexity that enables us to measure it. In biology today, organismal complexity
has two main operational senses: (1) a horizontal sense: the number of different
part types at a given hierarchical level (e.g., the number of cell types in a
multicellular individual) and (2) a vertical sense: the number of levels of nestedness
of parts within wholes (e.g., a eukaryotic multicellular individual is one level of
nestedness above a free-living protist). How do horizontal and vertical complex-
ity behave in evolution? For horizontal complexity, an increasing trend is pre-
dicted by current theory, that is, by the zero-force evolutionary law (ZFEL), but at
most hierarchical levels, evidence is lacking and the existence of a trend is
uncertain. For vertical complexity, there is unambiguous evidence for a trend in
the maximum, a rise in the maximum hierarchical level achieved by organisms
over the history of life. However, the underlying mechanism of change and the
forces driving the trend are unknown. Interestingly, there is some evidence that
the rise in vertical complexity, the addition of new levels, is – when it occurs –
accompanied by systematic losses in horizontal complexity at lower levels.

Keywords
Horizontal complexity • Vertical complexity • ZFEL • Complexity drain • Evo-devo

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Horizontal and Vertical, Objects and Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Horizontal Complexity and the Zero-Force Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Horizontal Complexity and the Structure of Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

D.W. McShea (*)

Biology Department, Duke University, Durham, NC, USA
e-mail: dmcshea@duke.edu

# Springer International Publishing AG 2017 1

L. Nuño de la Rosa, G.B. Müller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_123-1
2 D.W. McShea

A Trend in Horizontal Complexity? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Vertical Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
An Evolutionary Syndrome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Introduction

The study of the evolution of organismal complexity is not about progress, perfor-
mance, or perfection in evolution. It is not about genes or about the genetic basis of
sophisticated adaptations like the eye or the brain. It is not an application of
nonlinear equations, information theory, or any subfield of mathematical biology.
At least, it is not these things to start with. At some point, when the basic patterns of
change in complexity in evolution have been discovered, we will certainly want to
investigate the relationship between complexity and adaptedness, to study the
genetic basis of eyes and brains and such, and to consider what mathematical
gadgetry might be deployed to capture and predict these relationships. But until
we know just what it is that is changing, how it changes, and under what circum-
stances it changes, until then, we do not even know what we are trying to investigate.
So the first priority must be to devise a working understanding of complexity, one
that will enable us to measure it in real organisms, in practice, not just in principle.
The second is to actually measure it in organisms over the history of life and to
discover the pattern of change of complexity in evolution. We will be interested in
the whole pattern of change, not just the increases – the emergences of new
spectacularly complex organisms – but also the decreases, the frequent retreats
into simplicity.
The assumption, the hope, is that complexity will turn out to be more than a word
that we use to capture the wonder, awe, and mystery of biology and that it will prove
to be an important causal factor, one that can be – like temperature – measured and
its effects quantified.

Horizontal and Vertical, Objects and Processes

We begin with a simple and intuitive view of complexity as number of different part
types. A fish with 130 cell types has a complexity of 130. An arthropod with 50 cell
types is less complex, more precisely it is 50/130 or 0.38 times as complex. This is
complexity in a horizontal sense, where horizontal refers to the fact that we are
counting part types at a single level of organization, in this case the cell level. And
therefore, technically, when we count part types, we also need to specify the level of
organization. The complexity of a fish at the level of its cells is 130. But its
complexity at a higher level, the level of tissues and organs, is about 90. There is
no contradiction here. Complexity is simply a level-relative concept, with different
values at different levels.
Evolution of Complexity 3

Thus, it is not meaningful to speak about the “real” or “true” complexity of an

organism in a way that implies there is some level-independent sense of the word.
And contrary to conventional intuitions, an organism has no “true” complexity that
is some function of its genes. A pufferfish has 20,000–25,000 genes (about the same
as a human), but that number is just the complexity of the fish at the molecular level,
and it no more represents a “true” complexity of the organism than does the number
of cell types. (Actually, properly speaking, the complexity of the fish at the molec-
ular level would be a count of all of the types of molecule that it contains, a number
far greater than the number of genes.) Organisms – indeed all objects – simply have
different complexity values at different levels.
Complexity as part types is a measure of difference, the number of parts that are
significantly different from each other. It is a discrete measure in that it treats the
differences among parts as discontinuous. In many cases however, especially in
biology, differentiation is continuous, and in those, the appropriate measure is a more
general one, complexity as degree of differentiation among parts. The vertebrae of a
fish are all very similar from one end of the column to the other. In a mammal, the
vertebrae are more differentiated – cervicals are different from thoracics, which in
turn are different from sacrals, and so on. So a mammal column is more complex
than a fish column (at the level of vertebrae). Any of the usual measures of
continuous differentiation – such as standard deviation or variance – can be used
as measures of complexity in such cases.
The notion of “parts” is not well established in biology, but it turns out not to be
very troublesome. Organisms are not as cleanly separable into parts as machines are,
but many parts are – like organs and cells – reasonably discrete objects, sufficiently
so that they can be counted. Philosophical issues associated with the use of the parts
concept have been addressed (McShea and Venit 2001), and solutions to some of the
practical problems involved in counting have been devised (McShea 2002).
We are treating organisms here as objects, ignoring initially the intricacies of their
development and their marvelous adaptations. Those will all become important as
we explore further. In particular, we will want to ask about the relationship between
complexity and development and between complexity and adaptation. But in order
to ask those questions, we need an independent way to measure complexity. That is,
in order to ask, say, whether more complex organisms are better adapted, we need a
way to measure complexity that is independent of adaptation, just as asking about the
relationship between health and happiness requires us to devise a measure of
happiness that is independent of health.
Notice that the independence required here is conceptual. That is, it may turn out
that as an empirical matter, complexity and adaptation are correlated with each other,
and if the two variables are conceptually independent, that is an interesting finding.
But if we adopt a measure of complexity that has the concept of adaptation built into
it, definitionally, the discovered empirical correlation would be at least in part an
autocorrelation, between adaptation and itself, and that would tell us nothing.
Horizontal complexity is just one variety of complexity. There is also vertical
complexity, or the number of levels of nestedness of parts within wholes (see section
on “Vertical Complexity” below). Both have to do with physical structure. Both treat
4 D.W. McShea

organisms as objects. But there is also complexity of processes, the number of

different kinds of interaction among the parts of an organism, either in its physiology
or its development. And there is the irregularity of the arrangement of parts
(or processes) within an organism, in time or space. Even more types can be
imagined (McShea 1996). What they all share is a view of complexity as a physical
property, a property purely of an organism’s structure and independent of function,
one that at least in principle can be operationalized and measured in a straightfor-
ward way. As it happens, only the horizontal and vertical senses have been
operationalized and applied to organisms so far in biology, so those are the senses
I focus on here.
There is nothing necessary or inevitable about this way of thinking about com-
plexity. Other ways can be imagined. But this view has become standard in biology
in the last two decades (Valentine et al. 1994; Doolittle 2012). Also, not long ago, our
best assessments of organismal complexity were entirely impressionistic, or based
surreptitiously on the Great Chain of Being, with the consequence that it was
impossible to give answers to questions about trends in complexity that were both
serious and scientific. But as methods in evolutionary developmental biology have
advanced, and as our knowledge of organismal structure at multiple levels of
organization has grown, the data necessary to investigate trends are for the first
time becoming available. The viewpoint described here offers methods and mea-
sures to transform those data into objective answers.

Horizontal Complexity and the Zero-Force Law

For evolution and horizontal complexity, theory has for the time being outstripped
empirical investigation. We know what horizontal complexity is expected to do in
evolution, under certain ideal conditions – namely, to increase, on average – but we
do not yet know much about whether in fact it has.
The theory that predicts increase is called the zero-force evolutionary law (ZFEL,
McShea and Brandon 2010), which says that in the absence of natural selection
favoring increase or decrease in complexity (and of any constraints favoring increase
or decrease), the number of part types or degree of differentiation among types is
expected to increase (McShea and Brandon 2010). The reason is simply that in
evolution, everything varies, so that parts that are initially identical will tend to
become different from each other and parts that are already somewhat different will
tend to become even more different. Concretely, in the absence of constraint or
contrary selection, the segments in segmented animals will tend to become more
different from each other, from one generation to the next. More generally, in any
system with both heritability and variation, chance variation in parts accumulates,
leading to divergence of parts. This is true only when such divergence is not blocked
by forces (in biology, natural selection) or constraints of various kinds. (For more on
constraint, see the chapter on “▶ Developmental Constraints.”)
The ZFEL is consistent with standard evolutionary theory, which in the absence
of selection predicts divergence due to drift. But the ZFEL goes further, predicting
Evolution of Complexity 5

increasing complexity even when variation is under tight control by natural selection
acting at lower levels. That is, if one segment of a segmented animal – say a segment
near the mouth – is selected for ability to manipulate food and another more distal
segment is selected for ability to walk or swim, the two segments will tend to become
more different from each other. And that is the ZFEL. Calling this the ZFEL would
seem to violate the no-selection requirement, but it does not, because the law predicts
increasing complexity whenever no selection acts directly on the complexity of the
structure in question. In the case here, there is no selection acting on the degree of
differentiation of the segment series as a whole, no selection acting to keep the
segments similar to each other, and none favoring their becoming different. There is
only selection on each one, separately, for a special purpose. The segments are not
drifting, but they are changing randomly in a special sense: randomly with respect to
each other, in other words independently of each other. Thus, the ZFEL predicts
increasing complexity under a wide range of circumstances.
Finally, consider what happens when selection does act on complexity, not just on
individual parts independently but on the degree of differentiation among them. Now
the prediction changes. Selection acting against complexity could overwhelm the
ZFEL tendency, producing morphological stasis or even simplification. On the other
hand, selection favoring complexity is expected to reinforce the ZFEL, producing
complexification at a rate greater than expectation due to passive divergence of part
types. Detecting selection of this sort would require a quantitative version of the
ZFEL (not yet developed), in order parse any given increase into its ZFEL-driven
and selection-driven components.

Horizontal Complexity and the Structure of Development

Besides selection acting on complexity, internal constraints of various kinds could

deflect the increase predicted by the ZFEL. In particular, if organismal development
were structured in such a way as to make a loss of parts more likely than a gain of
parts, the complexity increase predicted by the ZFEL might be thwarted. In the
conceptual scheme of the ZFEL, the understanding in such a case would be that the
accumulation of differences among parts proceeds nevertheless, but that it is simply
overwhelmed by the bias in favor of part loss imposed by the structure of develop-
ment (McShea and Brandon 2010). Is development in fact structured in this way?
There is some reason in theory to think it might be. It has been suggested that part
loss might be easier than part gain in evolution simply because mutations are more
likely to destroy developmental pathways leading to existing parts than to originate
new pathways leading to new ones (discussed in McShea 1996). Destruction is
easier than creation, the argument goes.
This logic of loss is powerful, but it is worth pointing out that, unlike machine
parts, biological materials are what have been called “excitable media” (Goodwin
1996), with the result that removal of a part can lead to a novel contact between
tissues in development, sometimes inducing a new part (Müller and Streicher 1989).
What is more, development is known to be to some degree hierarchically structured,
6 D.W. McShea

to consist of processes organized in cascades of developmental dependency that

make losses extremely difficult. Many metazoans, for example, pass in early devel-
opment through a phylotypic stage in which the basic body plan is laid down (Raff
1996). Later developmental structures are built on top of the phylotypic stage, so to
speak, and therefore depend on it, as a house depends on its foundation. When
development is structured this way, losses of parts affecting the phylotypic stage are
likely to lead to an inviable embryo and thus are expected to be strongly opposed by
selection, while buffering mechanisms which prevent such losses will be highly
favored (i.e., canalization). A similar principle has been documented at a smaller
scale, the molecular level, in recent years. Finnigan et al. (2012) discovered a kind of
complexity ratchet in an oligomeric ring that forms part of an ATPase in fungi. In the
abstract version of this ratchet described by Doolittle (2012), an initially homo-
oligomeric ring, initially consisting of six molecules of type A bound together in a
ring, becomes hetero-oligomeric by neutral drift, coming to consist of As and Bs. In
a second step, A loses by drift the ability to bind to itself, with the result that B is
required for a functional ring to form. In other words, the complexity of the ring
increases by drift (essentially, by the ZFEL), but cannot decrease, because B cannot
be lost without destroying ring functionality.
The hierarchical structure of early development, recognized in Von Baer’s law, is
widely acknowledged, but it has so many exceptions that it would be hazardous to
claim that it results in a general bias toward gain over loss in development. The same
goes for the ratchet described by Doolittle and Finnigan et al. Thus, at present,
despite the huge increase in research on development in particular biological systems
in the past two decades, there is nothing that can be said about the on-average
structure of development, about whether organizations favorable to accumulation are
more prevalent than those favorable to loss. In other words, there is no data on
whether development in general is structured in a way that promotes either part loss
or part gain, or indeed whether it biases complexity generation at all, on average.

A Trend in Horizontal Complexity?

Empirically we know very little about any overall directionality in horizontal

complexity in evolution. Valentine et al. (1994) document an increase in the max-
imum number of cell types over the Phanerozoic in metazoans. Many increasing
trends have been documented within certain metazoan groups at the tissue/organ
level, such as the increase in degree of differentiation among arthropod limb types
(discussed in McShea 1996). But decreasing trends have been demonstrated as well,
such as the decrease in number of types of skull bones in vertebrates (Sidor 2001). At
a lower level of organization, within cells, a loss of part types – organelles and other
structures – has been documented in the transitions to multicellularity in plants and
animals (see below, under “An Evolutionary Syndrome”). And the general tendency
for complexity to decrease in organisms living in constant environments, especially
parasites, has long been acknowledged. In fact, losses of complexity have been
recognized in recent years to be much more pervasive than previously thought
Evolution of Complexity 7

(O’Malley et al. 2016). Still, the overall pattern for horizontal complexity in the
history of life is unknown. The ZFEL predicts increase as the default expectation, but
we do not know whether the default conditions are met, on average, or whether
selection or constraints routinely overcome the ZFEL.
It might seem that the literature on novelty or innovation in evolution is relevant
here (see the chapter on “▶ Novelty and Innovation”). And in fact it could be
relevant, if these terms are understood in a purely structural sense, as number of
part types. But novelty and innovation are often used in a functional sense or in some
hybrid sense combining both structure and function, and in that case, its relationship
to complexity is uncertain. A novel function could arise with the gain of a part, as
occurred in the gain of mitochondria in the lineage leading to eukaryotes. But it
could also arise as a loss of parts, as in the loss of limbs in whales, leading to a fully
aquatic life, an innovative functional mode for that lineage.

Vertical Complexity

Vertical complexity is number of levels, in other words, the depth of nestedness, or

number of tiers of parts within wholes in an organism. A multicellular eukaryotic
individual has greater vertical complexity than a solitary eukaryotic protist, because
a multicellular is – historically speaking – an aggregation of single cells and
therefore one level of nestedness deeper than a solitary protist.
The trend in vertical complexity over the history of life is well known and well
documented (Heim et al. in press; McShea and Changizi 2003). It has four unam-
biguous data points. By convention, the basement level for organisms is set at the
level of the prokaryotic cell. The second level is occupied by the eukaryotic cell,
which arose as an aggregation of prokaryotic cells (initially, presumably, a eubacte-
rial endosymbiont living within an archaebacterial host). The third level is occupied
by the multicellular eukaryotes and the fourth by societies or colonies of multi-
cellular eukaryotes. Figure 1 shows the trajectory of the trend in first occurrences, in
other words, the trend in the maximum.
A scale based on nestedness, parts within wholes, raises some conceptual issues:
(1) The starting point is somewhat arbitrary. If non-organismal systems are included,
the scale could presumably have to extend downward to include entities at the
molecular scale, such as chemical cycles, that is, entities that are the components
of prokaryotes, either presently or historically. (2) Occupation of a level of
nestedness can be understood to be continuous, rather than discrete. A green alga
like Gonium that is a simple aggregation of eukaryotic cells occupies the multi-
cellular level, but an oak tree with more cells and greater internal organization can be
said to occupy that level to a greater degree. In other words, it is more individuated at
that level. McShea and Changizi (2003, and references therein) devised an expanded
vertical complexity scale that interpolates two sublevels, marking degrees of indi-
viduation, between the major levels. (3) Vertical complexity is usually understood
to be a property of the structure of organisms, but there is no reason the concept
could not be extended to include ecological associations of many sorts. Multispecies
8 D.W. McShea

Fig. 1. The trend in vertical

complexity over the history of 4 colony
life, showing first occurrences
of organisms at each
hierarchical level

vertical complexity
multicellular
3 individual

eukaryotic cell
2

1 prokaryotic cell

3000 2000 1000 0

time (millions of years ago)

bacterial associations, including some biofilms, might be considered a level above a

solitary bacterium. And certain trophic webs might reach to levels higher than level
4, societies and colonies. Interestingly, the standard four-level scale for organisms
already implicitly includes ecological associations, since the first eukaryotic cell
arose as a multispecies association. (See Eldredge and Salthe 1984 for a treatment of
hierarchy scales in biology generally.)
The trajectory of the trend in Fig. 1 raises a number of puzzles. One of the most
salient has to do with the cause of the apparent acceleration. The transitions from
prokaryotic cell to eukaryotic cell and from eukaryotic cell to multicellular eukaryote
took about 1.4 billion years each. But the next transition, from multicellular to well-
individuated societies/colonies, took only a few 100 million years. Increasing ver-
tical complexity is a route to larger body size, so the acceleration could be connected
with the general rise in body size among organisms at that time, perhaps triggered by
the rise in atmospheric oxygen (Payne et al. 2011). A second puzzle is the apparent
slowing, or even truncation, of the trend at the level of society/colony. It has been
about 500 million years since the first well-individuated colonies (bryozoans)
appeared in the fossil record, but the next level – colonies of colonies – has not
yet arisen, so far as we know (McShea and Changizi 2003). It is possible that human
societies are on their way to that level, but the degree of individuation of our larger
social units has not been assessed in any rigorous way, and indeed a case could be
made that the human commitment to sociality does not even rise to the bryozoan
level. If that is right, then the slowing of the vertical complexity trend in the past
500 million years is an unsolved puzzle.
A third puzzle has to do with the dynamics of the trend, the pattern of change in
vertical complexity within lineages that accounts for the overall trend over the
Evolution of Complexity 9

history of life. One possibility is that the trend is the result of passive diffusion away
from a lower limit or wall. The first organisms may have been as vertically simple as
possible, so that as diversity increased, vertical complexity could only have
increased as well. With a wall on the left, there is nowhere to go but up. And in
fact, a study of changes in degree of individuation within and among levels revealed
no upward tendency (Marcot and McShea 2007), consistent with a diffusive dy-
namic. But another possibility, advanced by Knoll and Bambach (2000), is that
whenever a new vertical level is achieved (e.g., the eukaryotic cell), change occurs
by diffusion away from a left wall (arising from the improbability of eukaryotic cells
returning a prokaryotic condition), but this diffusion is limited by a wall on the right
(e.g., the biomechanical difficulties presented by multicellularity). Heim et al. (in
press) offer evidence, based on body size distributions, that the history of life can be
understood as a series of expansions away from left walls and repeated scaling of
right walls.
Some excitement about the vertical aspect of complexity has been generated in
biology in the past two decades by Maynard Smith and Szathmáry’s (1995) book on
the “major transitions” in the history of life. They understand major transitions as the
origin of new levels of replication or of significant changes in the way that infor-
mation is transmitted. Their new levels of replication correspond closely with the
levels of complexity addressed here (prokaryotic cell, eukaryotic cell, multicellular
eukaryote, society/colony), but the other major transitions (including the origin of
sexual reproduction and of human language) do not. The relationship between levels
and information, as well the meaning of the term “major transition,” is currently
under scrutiny in biology and the philosophy of biology (Calcott and Sterelny 2011;
see also the chapter on “▶ Major Transitions”), and in the meantime, the relationship
of that literature to vertical complexity remains uncertain.
There is also a fair-sized literature on mechanisms underlying change in vertical
complexity, seeking to understand the forces at work in the transition from solitary
individual to simple aggregate to highly individuated whole. In particular, the
question has been how selection on the whole is able to suppress the tendency of
lower-level individuals to behave selfishly, undermining the interests of the whole
(Maynard Smith and Szathmáry 1995; see also discussion in Simpson 2012). There
has also been some interest in the prior question of how aggregates arise in the first
place, necessarily prior because selection on the whole cannot act until the whole has
arisen (Brandon and Fleming 2015). A related theme in that literature has to do with
the structural features of new higher levels, with the engineering aspects of hierar-
chical structure, as opposed to fitness issues (Calcott 2008). For example, Venit
(2007) investigated the relationship between degree of division of labor among
lower-level entities and the degree of connectedness among them. And he found
that in marine invertebrate species that form highly individuated colonies, lower-
level individuals did not share body cavities (complete connectedness) nor were they
fully walled off from each other (complete isolation), but instead showed interme-
diate levels of connectedness.
10 D.W. McShea

An Evolutionary Syndrome

It is worth reiterating that horizontal and vertical complexity are conceptually

independent, meaning that nothing in the definition of either term necessarily implies
anything about the other. A system can be vertically complex (having many levels)
and horizontally simple (having few part types at some level), vertically complex
and horizontally complex, and the reverse of both of these. In other words, all four
combinations are possible in principle. Still, there is some data to suggest that there is
a connection in fact, more specifically, that as a new higher level emerges, parts are
lost at the next level down, what has been called a complexity drain (McShea 2002;
O’Malley et al. 2016). In the emergence of the eukaryotic cell, the eubacteria that
evolved into mitochondria became simpler, losing molecular components. More
generally, it is thought that in multispecies bacterial associations, one level up
from solitary bacteria, certain species may tend to lose the molecular machinery
associated with certain metabolic functions as other species in the association take
over those functions (also known as the “black queen” effect). Also, in multicellular
organisms, the cells have fewer part types on average than free-living protists
(McShea 2002). To see this, consider an extreme case, human blood cells, which
have essentially no parts at all, compared to free-living protists, which have many.
Finally, there is some reason to think that the same pattern of loss occurred in the
origin of intensely social/colonial organisms: as the individuals in clonal associa-
tions differentiated into multiple types (castes in insects, polymorphs in marine
invertebrates), they lose parts, becoming simpler as they specialized.
These observations suggest a possible pattern at the largest scale, a recurring set
of changes at the scale of life as a whole, in other words a kind of evolutionary
“syndrome” (McShea 2015) with three signature “symptoms.” The first is the trend
in vertical complexity itself, the episodic addition of ever-higher levels of nestedness
in organisms. The second is the increase in horizontal complexity within each level,
that is, the increase in number of part types. And the third is a decrease in horizontal
complexity within those parts, the complexity drain. More research is needed to see
if these patterns are robust and then to investigate possible causes.

Cross-References

▶ Developmental constraints
▶ Macroevolution
▶ Major transitions
▶ Morphological disparity
▶ Novelty and innovation
▶ Origins of multicellularity
Evolution of Complexity 11

References
Brandon R, Fleming L (2015) Why flying dogs are rare: a general theory of luck in evolutionary
transitions. Stud Hist Phil Biol Biomed Sci 49:24–31
Calcott B (2008) The other cooperation problem: generating benefit. Biol Philos 23:179–203
Calcott B, Sterelny K (eds) (2011) The major transitions in evolution revisited. The MIT Press,
Cambridge
Doolittle WF (2012) A ratchet for protein complexity. Nature 481:270–271
Eldredge N, Salthe SN (1984) Hierarchy and evolution. Oxf Surv Evol Biol 1:184–208
Finnigan GC, Hanson-Smith V, Stevens TH, Thornton JW (2012) Evolution of increased complex-
ity in a molecular machine. Nature 481:360–364
Goodwin BC (1996) How the leopard changed its spots. Simon and Schuster, New York
Heim NA, Payne JL, Finnegan S, Knope ML, Kowalewski M, Lyons SK, McShea DW, Novack-
Gottshall PM, Smith FA, Wang SC (in press) Hierarchical complexity and the size limits of life.
Proc R S Lond B Biol Sci
Knoll AH, Bambach RK (2000) Directionality in the history of life: diffusion from the left wall or
repeated scaling of the right? Paleobiology 26:1–14
Marcot J, McShea DW (2007) Increasing hierarchical complexity throughout the history of life:
phylogenetic tests of trend mechanisms. Paleobiology 33:182–200
Maynard Smith J, Szathmáry E (1995) The major transitions in evolution. Oxford University Press,
New York
McShea DW (1996) Metazoan complexity and evolution: is there a trend? Evolution 50:477–492
McShea DW (2002) A complexity drain on cells in the evolution of multicellularity. Evolution
56:441–452
McShea DW (2015) Three trends in the history of life: an evolutionary syndrome. Evol Biol
43:531–542
McShea DW, Brandon RN (2010) Biology’s first law. University of Chicago Press, Chicago
McShea DW, Changizi MA (2003) Three puzzles in hierarchical evolution. Integr Comp Biol
43:74–81
McShea DW, Venit EP (2001) What is a part? In: Wagner GP (ed) The character concept in
evolutionary biology. Academic Press, San Diego, pp 259–284
Müller GB, Streicher J (1989) Ontogeny of the syndesmosis tibiofibularis and the evolution of the
bird hindlimb: a caenogenetic feature triggers phenotypic novelty. Anat Embryol 179:327–339
O’Malley MA, Wideman JG, Ruiz-Trillo I (2016) Losing complexity: the role of simplification in
macroevolution. Trends Ecol Evol 31:608–621
Payne JL, McClain CR, Boyer AG, Brown JH, Finnegan S, Kowalewski M, Krause RA Jr,
Lyons SK, McShea DW, Novack-Gottshall PM, Smith FA, Spaeth P, Stempien JA, Wang SC
(2011) The evolutionary consequences of oxygenic photosynthesis: a body size perspective.
Photosynth Res 107:37–57
Raff RA (1996) The shape of life: genes, development, and the evolution of animal form. University
of Chicago Press, Chicago
Sidor CA (2001) Simplification as a trend in synapsid cranial evolution. Evolution 55:1419–1442
Simpson C (2012) The evolutionary history of division of labour. Proc R Soc B Biol Sci
279:116–121
Valentine JW, Collins AG, Meyer CP (1994) Morphological complexity increase in metazoans.
Paleobiology 20:131–142
Venit EP (2007) Evolutionary Trends in the individuation and polymorphism of colonial marine
invertebrates. PhD thesis, Department of Biology, Duke University
Coevolution and Macroevolution

John N. Thompson, Kari A. Segraves, and David M. Althoff

Abstract
Coevolution is reciprocal evolution of interacting species driven by natural
selection. Selection imposed by interactions between or among species can
cause trait changes that alter ecological outcomes, patterns of local adaptation,
and diversification of lineages. For example, selection can reduce the effect of the
interaction when one species suffers a loss in fitness (antagonistic interactions) or
increase the effect when species benefit from the association (mutualistic inter-
actions). The selected traits may either change the cost of the interaction or the
probability that the interaction occurs at all. These evolutionary changes can lead
to local coadaptation as interacting species adapt and counteradapt to one another
over time. In some cases, one or more of the locally coadapted species may
become reproductively isolated from other populations as local coevolution
decreases the chance of mating among populations. This cessation of gene flow,
coupled with further evolutionary change, could lead to the formation of nascent
species. There is, then, a direct potential connection between local coadaptation
of populations, speciation, and macroevolutionary diversification. Some of the
most challenging questions in coevolutionary biology center on understanding
how coevolving traits change as they are expressed in a diversity of genetic and
environmental backgrounds, how such traits can directly or indirectly lead to
reproductive isolation, and whether these traits are likely to cause recurrent
patterns of speciation that produce macroevolutionary patterns. This article

J.N. Thompson (*)

Department of Ecology and Evolutionary Biology, University of California, Santa Cruz, Santa
Cruz, CA, USA
e-mail: jnthomp@ucsc.edu
K.A. Segraves • D.M. Althoff
Department of Biology, Syracuse University, Syracuse, NY, USA
Archbold Biological Station, Venus, FL, USA
e-mail: ksegrave@syr.edu; dmalthof@syr.edu

# Springer International Publishing AG 2017 1

L. Nuño de la Rosa, G.B. Müller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_125-1
2 J.N. Thompson et al.

considers what is currently known about the steps of this hierarchical process of
evolutionary, and sometimes coevolutionary, diversification of interactions
among species and how shifts in development may play an instrumental role in
diversification.

Keywords
Coadaptation • Coevolution • Diversification • Geographic mosaic theory • Mac-
roevolution • Evo-devo • Reciprocal selection • Speciation

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Adaptive Coevolutionary Divergence and Geographic Mosaics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
From Coevolutionary Geographic Mosaics to Ecological Speciation . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Determining the Role of Coevolution in Patterns of Macroevolutionary Divergence . . . . . . . . . . 6
Unresolved Questions on the Evolutionary Developmental Biology of Coevolution . . . . . . . . . . 10
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Introduction

Throughout evolutionary history, organisms have evolved in response to their

physical environment and to the other species with which they interact. Together,
selection imposed by abiotic and biotic environments continually reshapes the
phenotypes of organisms. Selection imposed by biotic environments (i.e., other
species), though, can be especially strong in shaping trait evolution, because it can
produce an ongoing feedback in the traits of interacting species. That is, changes in
one species can precipitate changes in the second species, and this in turn can cause
further change in the first species. The same process can occur within larger webs of
interaction through a combination of direct and indirect feedbacks. This ongoing
process of reciprocal evolutionary change among interacting species, termed coevo-
lution, is an important mechanism that drives changes in organismal phenotypes.
From the beginning of life on Earth, coevolution was likely an instrumental force
that shaped not only the communities of single-celled microbes, but also perhaps
gave rise to multicellular organisms and the evolution of eukaryotes (Margulis and
Fester 1991). Coevolution of proto-mitochondria, proto-chloroplasts, and heterotro-
phic single-celled organisms spurred the integration of these separate organisms into
a single functioning unit that eventually led to the diversity of life forms we see
today. Given the fact that every organism interacts with a multitude of other species
during its lifetime and the large potential for coevolution to shape these species
interactions, there is no question that coevolution is one of the processes molding the
diversity of life. Indeed, coevolutionary interactions are diverse and have been found
among predators and prey, parasites and hosts, competing species, and mutualistic
species (Thompson 1994). Hence, coevolutionary selection will produce traits that
Coevolution and Macroevolution 3

are advantageous for interacting with different species. Many times, the differences
we observe among closely related species are a direct result of their adaptations for
interacting with different suites of species.
Changes in development are crucial for these adaptations and can be responsible
for fine-tuning traits or initiating larger phenotypic changes. Even so, there has been
surprisingly little integration of coevolutionary theory into developmental biology
and vice versa. The goal of this chapter is to explore how coevolution and develop-
ment could be integral in driving adaptations that eventually lead to speciation and
large-scale macroevolutionary diversification. We start by first describing the coevo-
lutionary process within and among populations and then discuss how this process
could drive speciation and fuel macroevolutionary diversification. Along the way,
we highlight how development has been critical for promoting coevolution and
creating traits that may spur speciation. We end by outlining several major unan-
swered questions about the linkage among coevolution, development, and
macroevolution.

Adaptive Coevolutionary Divergence and Geographic Mosaics

Populations, rather than species, are the unit of evolutionary and coevolutionary
change. The evolution of a species is the composite of all its populations that have
adapted to varying degrees to their local physical environments and to their interac-
tions with other species. The exchange of migrants among these populations is the
connection that maintains the cohesiveness of a species across its geographic range.
During the process of local adaptation and coadaptation with other species, some
populations diverge strongly from one another in ecologically and evolutionarily
important traits. Such strong divergence occasionally provides the basis for ecolog-
ical speciation by reducing gene flow among populations, and even more rarely,
leading to adaptive radiations of new lineages. The microevolutionary processes at
the level of populations can scale up to drive macroevolutionary patterns. During this
process, minor or major changes in the developmental program are sometimes
necessary to modify or produce new traits that are important for interactions. If
changes to the developmental program are large enough, there may be shifts in life
histories or correlated changes in traits important for reproduction that begin the
process of reproductive isolation among populations that are diverging.
For interactions among species, local coadaptation may result in a geographic
mosaic of coevolution. The geographic mosaic theory of coevolution stipulates that
the coevolutionary process is fueled by three sources of variation in interactions
among species (Thompson 1994, 2013). Selection mosaics arise across ecosystems
as natural selection favors different traits or trait combinations in different environ-
ments. Coevolutionary hotspots and coldspots arise as natural selection favors
reciprocal evolutionary change among interacting species in some environments
but not in others. Trait remixing occurs as gene flow, genomic processes, and
metapopulational processes continue to reshuffle the combinations of genes that
4 J.N. Thompson et al.

are locally available for selection to act upon. Hence, the structure of selection, the
strength of reciprocal selection, and the traits available for selection all vary among
environments.
The formal population genetic way of envisioning these sources of coevolution-
ary variation is as a genotype by genotype by environment interaction (GxGxE). If a
genotype of one species interacts with a genotype of a second species in two
different environments, the effect of the interaction on the two species would often
differ between the two environments. An example would be two identical twins
living in different places. Both are infected with the same virus clone, but the virus is
expressed in a more virulent way in one environment than in the other environment.
If now, instead of one genotype, a population of individuals with many different
genotypes interacts with another genetically variable population, the distribution of
outcomes from their interactions is almost bound to vary among environments.
Genotype by genotype by environment variation in outcome is the basis for coevo-
lutionary divergence. The distribution of outcomes can change not just through
selection on mutations with new structural properties but also through the develop-
mental timing of expression of genes important in species interactions.
The combination of selection mosaics, coevolutionary hotspots and coldspots,
and trait remixing therefore produces the first steps in macroevolutionary divergence
driven by coevolution. It does so, though, not just by acting differently on the same
genes or their expression in different environments, but also by acting on different
genes in different environments. Selection may act mostly on morphological traits in
some environments, but on physiological or behavioral traits in other environments.
An interaction between the same two or more species may even evolve to be
antagonistic in some environments, commensalistic (one species benefits with no
effect on the other) in other environments, and mutualistic in yet other environments.
This variation in interaction outcomes and traits provides the raw material for
differences to evolve among populations and, eventually, species.
Examples of geographic mosaics of coevolution are now known for many kinds
of interactions. Plants and fungal pathogens show geographic differences in defense
and counterdefense genes (Laine et al. 2014). Coevolving Taricha newts and garter
snakes differ geographically in the levels of tetrodotoxin in the skin of the newts and
the ability of garter snakes to resist high levels of the toxin after ingesting newts
(Hague et al. 2016). Similarly, wild parsnip plants differ among populations in the
combination of defensive furanocoumarins they employ against specialist parsnip
webworms, and the webworms match those differences with detoxifying P450
enzymes in their gut (Li et al. 2014). These geographic mosaics can also include
differences in the ecological outcomes of interactions. The outcome of interactions
between woodland star (Lithophragma) plants and Greya moths range from antag-
onistic to mutualistic among ecosystems (Thompson and Fernandez 2006). As
interactions change in structure, strength, and outcome across environments, evolu-
tionary changes in developmental pathways, either early on or late in developmental
cascades, can help provide trait variation to fuel adaptation.
The genotype by genotype by environment variation that forms the basis of the
geographic mosaic of coevolution can therefore be driven by geographic differences
Coevolution and Macroevolution 5

in the developmental expression of traits or by differences in the effectiveness of an

expressed trait in different environments. Thus, developmental plasticity may be a
major component of trait evolution. Natural selection can even restrict the expression
of coevolved traits to some environments but not others. For example, some species
of Daphnia water fleas plastically develop hardened “helmets” only in environments
in which they detect the chemical signatures of potential predators in the surrounding
water (Petrusek et al. 2009). Escape from predation has also been suggested as the
basis of developmental differences between wet season and dry season in the wing
colors and eyespots on Bicyclus butterflies in Africa (Brakefield 2010).

From Coevolutionary Geographic Mosaics to Ecological

Speciation

For the geographic mosaic to shape speciation, gene flow among diverging
populations must weaken relative to the strength of divergent selection. Ecological
speciation, in which strong divergent selection directly promotes reproductive iso-
lation among populations, can be a key component of this process. For example,
populations of threespine sticklebacks have diverged repeatedly into locally adapted
populations during and following the Pleistocene as marine populations became
trapped in coastal freshwater lakes and ponds during rising and falling sea levels
(Schluter 2016). Those freshwater populations then diverged repeatedly into benthic
forms (bottom dwelling) and limnetic (open water) forms, driven in part by com-
petitive interactions among individuals and strong selection for specialization in prey
use that favored different traits (e.g., body armor and shape) in different aquatic
microhabitats. These forms also differ in their propensity to mate with one another
based on body size differences that exist between the benthic and limnetic forms
(Conte and Schluter 2013).
Similarly, species of Anolis lizards have diverged in repeatable ways among
Caribbean islands, depending on which Anolis species are present on an island
(Losos 2009). Through competition, lizards have radiated into different morpholog-
ical forms that exploit different microhabitats (e.g., tree trunks vs. small limbs), and
each island has predictable combinations of Anolis species with specific body types
or ecomorphs. These morphological forms differ in a number of traits that correlate
with shifts in body size and shape as well as the substrates on which they forage
(Harmon et al. 2003). Importantly, the ecomorphs do not have a single origin that
then dispersed to different islands. Instead, the same forms have evolved repeatedly
in predictable ways as lizard populations have colonized different islands (Mahler
et al. 2013). These morphological changes have occurred through differences in the
early development of the lizards (Sanger et al. 2012). The adaptive radiation of
Anolis species has therefore been driven at least to some extent by ongoing coevo-
lution of competing species.
Coevolutionary divergence can also be driven by interactions between predators
and prey. Red crossbill birds in North America have diverged into at least nine
ecotypes that specialize on feeding on the seeds of different conifer species. These
6 J.N. Thompson et al.

ecotypes use different calls, vary in the extent to which they are nomadic or
sedentary, and show some genomic divergence from one another (Parchman et al.
2016). More extreme divergence, however, is found in one ecotype called the South
Hills crossbill that specializes on lodgepole pines and shows substantial divergence
from other crossbills at a small number of genetic loci. This ecotype has bills that are
adapted to extracting seeds from closed cones of lodgepole pines, and the pines have
adaptations that make it difficult for the birds to extract the seeds. Both the birds and
the pines differ from other populations in ways that suggest local coadaptation
(Benkman et al. 2009). The traits involved in this divergence involve allometric
changes in size and shape that could readily evolve through small developmental
changes. Unlike many other crossbill populations, these local birds have become
sedentary rather than nomadic and, in the process, have become increasingly repro-
ductively isolated from other crossbill populations over the past 6000 years.
The most famous example of ecological speciation that has led to an adaptive
radiation is Darwin’s finches (Grant and Grant 2014). Within the Galapagos Island
chain are approximately 14 species of Geospiza finches that differ primarily in body
size and bill morphology. These species evolved from a common finch ancestor into
specialist and generalist species that differ in the type (seeds vs. insects) and size of
food they consume. The changes in bill morphology necessary to specialize on
different food types are driven by changes in allometry as well as developmental
changes in individual traits (Foster et al. 2008). In this case, the extent to which
coevolution has directly fueled diversification of the birds is not known, but com-
petitive interactions among these species combined with specialization in food types
suggest that interspecific interactions rather than adaptations to different physical
environments have been the major driver of this adaptive radiation (Grant and Grant
2014). The radiation of Darwin finches highlights the difficulty in disentangling the
relative role of coevolution versus other factors in creating and maintaining diver-
gence among species.

Determining the Role of Coevolution in Patterns

of Macroevolutionary Divergence

Coevolution of two or more lineages has the potential to generate a wide range of
macroevolutionary patterns because the relationships between adaptive divergence
among populations, phylogeographic divergence, and speciation are probably rarely
the same across interacting lineages. Moreover, adaptive divergence of populations
may lead to speciation in both (or all) partners; however, it is also possible that
speciation only occurs on one side of an interaction even if populations on both sides
are coevolving. Despite the lack of general expectations for macroevolutionary
patterns associated with coevolution, a common starting point is to examine phylo-
genetic patterns of speciation between interacting lineages because this may give an
indication when coevolution may have had a strong role in speciation. One caveat of
this approach, however, is that because there are many patterns of speciation that can
be generated between coevolving lineages, from strong congruence of speciation
Coevolution and Macroevolution 7

events to more disconnected ones, macroevolutionary patterns cannot be attributed

to the coevolutionary process without additional forms of evidence (Althoff et al.
2014).
One of the most common popular expectations of coevolutionary divergence is to
observe parallel speciation in two coevolving lineages, but that is probably a very
rare outcome. Part of that expectation arose from a deep misinterpretation of Paul
Ehrlich and Peter Raven’s 1964 paper on macroevolutionary patterns, which is
commonly cited as predicting parallel cladogenesis. The process they describe,
however, cannot, by definition, result in parallel speciation. Ehrlich and Raven
(1964) suggested that, during a coevolving interaction between two lineages, a
mutation in a species in a host lineage would free it from interactions with a parasite
lineage. That would allow the mutant host lineage to undergo speciation and
potentially adaptively radiate in the absence of interactions with the parasites.
Eventually, a mutant in the parasite lineage would be able to colonize the mutant
host lineage and, in turn, radiate across the host lineage. This process does not
require the mutant parasite lineage to colonize the mutant host lineage in ancestral-
descendent order. As a result, any parallelism in speciation in the two coevolving
lineages would occur only at higher taxonomic levels, and not at the species level.
Instead, we would predict to observe starbursts of speciation followed by lag phases
that alternate in timing between the interacting lineages.
The predictions of “escape-and-radiate” coevolution, coupled with what is now
known about the geographic mosaic of coevolution, suggest that parallel speciation
of coevolving species is highly unlikely except under certain circumstances. These
include interactions between hosts and vertically transmitted parasites and interac-
tions between hosts and species that control the movement of gametes among host
individuals. The best examples of the latter situation are interactions between plants
and pollinating floral parasites, such as yuccas and yucca moths, woodland stars and
Greya moths, leafflower plants and leafflower moths, and figs and fig wasps. These
interactions involve insect pollinators that eat the developing seeds in the same
flowers that they pollinate.
Because diversification fueled by coevolution produces no single pattern, infer-
ring the role of coevolution requires a multifaceted approach that combines (1) phy-
logenetic and, when possible, phylogeographic investigation of species
relationships, (2) analysis of how natural selection shapes trait evolution among
the interacting species, and (3) analysis of how selection on traits involved in
interactions may directly or indirectly lead to reproductive isolation. The overall
evaluation therefore requires analyses at both microevolutionary and macroevolu-
tionary levels. Microevolutionary and macroevolutionary analyses of extant species
allow analyses of selection on potentially coevolving traits and the causes of
geographic and interspecific differences in current maintenance of those traits and
ecological outcomes. The fossil record can greatly deepen this understanding by
providing important information on the potential origin of coevolving traits, their
diversification over longer periods of time, and the potential limits on what appears
to be directional selection on traits in short-term studies of extant species. For this
reason, there are still few clear examples in which coevolution has been shown to
8 J.N. Thompson et al.

lead directly to large-scale patterns of speciation within interacting lineages (Althoff

et al. 2014; Hembry et al. 2014).
We also do not understand whether antagonistic or mutualistic interactions are
more likely to influence speciation. And, at a deeper level, we do not know if
different forms of antagonism (e.g., competition, predation) or mutualism (e.g.,
mutualisms between symbionts and hosts, mutualisms between free-living species)
differ in how they shape speciation and macroevolutionary change. For example,
natural selection often acts on competing species to mitigate or dissolve the interac-
tion, whereas in interactions between parasites and hosts, selection acts on the
parasite to increase the interaction and on the host to decrease the interaction or its
negative effects. Whether any form of antagonism or mutualism is more likely to
lead to trait divergence that could drive reproductive isolation is currently unclear.
That said, parasites or mutualists that control host reproduction seem to be the
clearest candidates for a major role of coevolution in shaping diversification.
Interactions between plants and pollinating floral parasites provide some of the
best examples of how control of host reproduction by symbionts may lead to an
extraordinarily high diversification of species. Hundreds of fig species worldwide
are each pollinated by their own one or two fig wasps (Cruaud et al. 2012), and
potentially hundreds of leafflower plants (Glochidion) throughout the tropical
Pacific region are similarly each pollinated by their own one or a few Epicephala
moths (Kawakita 2010). Diversification has also occurred in other insect lineages
that act as pollinating floral parasites and in the plants they pollinate, but the resulting
diversification occurs sometimes not only at the species level but also at higher
taxonomic levels through host shifts. For example, prodoxid moths have colonized a
diverse array of plant families. Many of the interactions between these moths and the
host plants are antagonistic, with moth larvae feeding on seeds, floral, scape, or other
plant tissues. But one lineage of the moths, called yucca moths, has become the sole
pollinators of a lineage of monocots, and another lineage of the moths, called Greya
moths, has become the major pollinators of a group of eudicots. The yucca moths
actively pollinate yuccas as they oviposit into the flowers, whereas the Greya moths
passively pollinate their host flowers during oviposition or nectaring using different
mechanisms (Pellmyr and Leebens-Mack 1999, Thompson et al. 2013).
Developmental shifts have been crucial in shaping the traits important to these
interactions. For example, changes in development played a central role in the
evolution of the specialized tentacular mouthparts used by yucca moths to actively
collect and deposit pollen on yucca flowers (Fig. 1). Similarly, changes in the relative
and absolute lengths of the abdominal segments of Greya moths among species and
populations have affected how they pollinate different host species during oviposi-
tion with pollen adhering to the abdomen (Thompson et al. 2013). The tentacles of
yucca moths are particularly impressive, because they evolved de novo and are not
present in any other insect group. Morphological work suggests that the movement
of the tentacles is linked to the proboscis and likely stem from the same develop-
mental template as the proboscis (Pellmyr and Krenn 2002). The evolution of the
tentacles occurred once at the base of the radiation of yucca moths and was a key trait
for increasing the pollination efficiency of the moths (Pellmyr and Leebens-Mack
Coevolution and Macroevolution 9

Fig. 1 Mutualistic and antagonistic traits in yucca moths (Tegeticula spp.; Prodoxidae). (a) Close-
up of female T. altiplanella mouthparts showing “tentacles” used for pollination (solid arrow) and
proboscis (dotted arrow). (b) Thin, needle-like ovipositor of female T. yuccasella that deposits eggs
deep within yucca pistil. (c) Short, stout ovipositor of T. cassandra female that deposits eggs just
below pistil surface

1999). The tentacles are truly a mutualistic trait, because cheater moth species that
evolved from pollinator species deposit eggs into yucca fruits rather than flowers and
lost functional tentacles (Pellmyr 1999).
There is no question that the evolution of the specialized mouthparts was instru-
mental in the relationship among yuccas and yucca moths. What is less clear,
however, is whether this key trait is directly responsible for the burst of speciation
in the pollinator moths and yuccas or if it sets the stage for other aspects of the
interaction to drive speciation. Among pollinator species, there are two radiations of
moths that differ in the how they place their eggs in yucca flowers. The shift in egg
placement from deep within the flower next to ovules to just on the pistil surface
corresponded with changes in the morphology of the egg-laying structure of the
moths (Pellmyr 1999). Deep ovipositing species have a long, thin, needle-like
ovipositor that damages plant ovules in contrast to shallow ovipositing species that
have short, stout ovipositors (Fig. 1). This shift in oviposition strategy has resulted in
correlated evolution of the male intromittent organ as well, suggesting a means of
mechanical reproductive isolation among moth species that differ in oviposition
strategy (Althoff 2014). The current results from yucca moths suggest that the
10 J.N. Thompson et al.

antagonism might be more directly responsible for the diversification of moths

rather than the mutualism (Althoff 2016). Even so, it is clear that development has
played a major role in creating key traits and modifying mutualistic and antagonistic
traits important in the overall relationship between plant and moth lineages.

Unresolved Questions on the Evolutionary Developmental

Biology of Coevolution

As species continually interact and coevolve, the traits important in interactions will be
under constant selective pressure either to diverge or remain fixed at optimal trait values.
Development influences every aspect of an organism’s biology, and small changes in
developmental pathways can produce extraordinary changes in the phenotypic expres-
sion of traits. In many ways, coevolution has the potential to act strongly on develop-
ment to help shape the traits of interacting species. Because coevolutionary biology has
only begun to grapple with how shifts in development can affect coevolution (and vice
versa) among species and shape the diversification of interacting lineages, there are
major research questions that are ripe for exploration including these:

1. How does coevolution shape interactions throughout development? Individuals

at different ages and sizes are susceptible to attack by different parasites and
predators, and they may depend on different mutualists. For example, some
mutualistic interactions between plants and mycorrhizal fungi may be very
important during plant seedling establishment, but have little or no impact on
adult performance. Many fish species need to avoid gape-limited predators as
juveniles but then escape predation once they reach a size threshold. There is,
therefore, strong potential for developmental changes to fundamentally affect
how individuals within a population interact with other species and how those
interactions affect fitness. In particular, species with morphologically and phys-
iologically distinct developmental stages, such as holometabolous insects,
amphibians, and marine species with pelagic larvae have vastly different interac-
tions between development stages. We know relatively little about how coevolu-
tion with species at one stage of life may affect interactions with other species at
later stages of life.
2. How does the interplay between coevolution and development coordinate
coevolved defenses and counterdefenses across developmental stages? Some
interactions span several life history stages and may require different responses
at different stages. For example, parasitoid wasps and flies use other insects as
hosts for development. Adults must locate hosts, overcome host defenses, and
deposit eggs in or on host insects. In turn, the parasitoid larvae have to overcome
the host’s defenses, especially internal physiological/immunological defenses.
Selection at different parts of the life stages will favor different traits and those
traits must be coordinated to be expressed at the right time during the develop-
ment of the parasitoid. Similarly, the host insects must be able to respond to all the
different stages of attack by the parasitoid. If we add in the other interactions that
Coevolution and Macroevolution 11

both parasitoids and host insects have with other species, we can begin to see the
complexity of traits that need to appear during the right stages of development.
We have little understanding of how the genes responsible for the development of
different traits important for different interactions may act synergistically or as
constraints on evolution.
3. How does developmental plasticity contribute to coevolution and, ultimately,
speciation into new adaptive zones? Development has a dual role in the produc-
tion of trait variation. For many traits, environmental inputs over the course of
development can have large impacts on phenotypic expression. This plasticity
along with underlying genetic variation provides the wealth of phenotypic vari-
ation that is available for natural selection. Development, however, also con-
strains the range of trait variation within certain values of trait space. For
example, within Bicyclus butterfly species, the number and size of eyespots
vary between the wet and dry seasons. Across species, eyespots vary more, but
this variation is constrained to a subset of possible eyespot positions, colors, and
sizes (Brakefield 2010), and eyespot shape is due to major genes and hormonal
gradients that control eyespot development (Oostra et al. 2014). It is also unclear
how often major or minor shifts in development can open up a new adaptive zone
that fuels speciation in interacting lineages. For example, changes in development
can produce de novo phenotypes in mutualistic interactions, such as in the
evolution of yucca moth mouthparts (Pellmyr and Krenn 2002), or in antagonistic
interactions, such as the allometric changes in body size in Anolis lizards that
have led to repeated speciation among competing species (Losos 2009). Each of
these changes has been instrumental in opening up new avenues of speciation
through coevolution with other species.

Conclusions

We are only just beginning to understand how developmental mechanisms shape the
ecological diversification of coevolving species and the macroevolutionary patterns
that result. Changes in development can provide the phenotypic variation required
for natural selection, potentially creating major changes in phenotype from simple
changes in developmental pathways. These changes could cause direct modification
of traits important for mating or produce a cascade of changes via correlated
evolution with other traits that secondarily influence reproductive isolation. The
synergy between development and continued coevolution of interacting species has
the potential to move species to new adaptive zones, increase reproductive isolation,
and spur diversification.

Cross-References

▶ Convergence
▶ Developmental Plasticity
12 J.N. Thompson et al.

▶ Eco-Evo-Devo
▶ Macroevolution
▶ Morphological Disparity

References
Althoff DM (2014) Shift in egg-laying strategy to avoid plant defense leads to reproductive
isolation in mutualistic and cheating yucca moths. Evolution 68:301–307
Althoff DM (2016) Specialization in the yucca-yucca moth obligate pollination mutualism: a role
for antagonism? Am J Bot 103:1803–1809.
Althoff DM, Segraves KA, Johnson MTJ (2014) Testing for coevolutionary diversification: linking
pattern with process. Trends Ecol Evol 29:82–89
Benkman CW, Smith JW, Keenan PC, Parchman TL, Santisteban L (2009) A new species of the red
crossbill (Fringillidae: Loxia) from Idaho. Condor 111:169–176
Brakefield PM (2010) Radiations of mycalesine butterflies and opening up their exploration of
morphospace. Am Nat 176(Suppl 1):S77–S87
Conte GL, Schluter D (2013) Experimental confirmation that body size determines date preference
via phenotype patching in a stickleback species pair. Evolution 67:1477–1484
Cruaud A, Ronsted N, Chantarasuwan B, Chou LS, Clement WL, Couloux A, Cousins B,
Genson G, Harrison RD, Hanson PE, Hossaert-Mckey M, Jabbour-Zahab R, Jousselin E,
Kerdelhue C, Kjellberg F, Lopez-Vaamonde C, Peebles J, Peng YQ, Pereira RAS,
Schramm T, Ubaidillah R, van Noort S, Weiblen GD, Yang DR, Yodpinyanee A, Libeskind-
Hadas R, Cook JM, Rasplus JY, Savolainen V (2012) An extreme case of plant-insect
codiversification: figs and fig-pollinating wasps. Syst Biol 61:1029–1047
Ehrlich P, Raven P (1964) Butterflies and plants: a study in coevolution. Evolution:586–608
Foster DJ, Podos J, Hendry AP (2008) A geometric morphometric appraisal of beak shape in
Darwin’s finches. J Evol Biol 21(1):263–275
Grant PR, Grant R (2014) Forty years of evolution. Princeton University Press Princeton, Princeton
Hague MT, Avila JLA, Hanifin CT, Snedden WA, Stokes AN, Brodie ED (2016) Toxicity and
population structure of the rough-skinned newt (Taricha granulosa) outside the range of an arms
race with resistant predators. Ecol Evol 6:2714–2724
Harmon LJ, Schulte JA II, Larson A, Losos JB (2003) Tempo and mode of evolutionary radiation in
iguanian lizards. Science 301:961–964
Hembry DH, Yoder JB, Goodman KR (2014) Coevolution and the diversification of life. Am Nat
184:425–438
Kawakita A (2010) Evolution of obligate pollination mutualism in the tribe Phyllantheae
(Phyllanthaceae). Plant Species Biol 25:3–19
Laine AL, Burdon JJ, Nemri A, Thrall PH (2014) Host ecotype generates evolutionary and
epidemiological divergence across a pathogen metapopulation. Proc R Soc B Biol Sci
281:20140522
Li W, Zangerl AR, Schuler MA, Berenbaum MR (2014) Characterization and evolution of
furanocoumarin-inducible cytochrome P450s in the parsnip webworm, Depressaria
pastinacella. Insect Mol Biol 13:603–613
Losos JB (2009) Lizards in an evolutionary tree. University of California Press, Berkeley
Mahler DL, Ingram T, Revell LJ, Losos JB (2013) Exceptional convergence on the macroevolu-
tionary landscape in island lizard radiations. Science 341:292–295
Margulis L, Fester R (1991) Symbiosis as a source of evolutionary innovation. MIT Press,
Cambridge, MA
Oostra VA, Mateus RA, van der Burg KRL, Piessens T, van Eijk M, Brakefield PM, Beldade P,
Zwaan BJ (2014) Ecdysteroid hormones link the juvenile environment to alternative adult life
histories in a seasonal insect. Am Nat 184:E79–E92
Coevolution and Macroevolution 13

Parchman TL, Buerkle CA, Soria-Corrasco V, Benkman CW (2016) Genomic divergence and
diversification within a geographic mosaic of coevolution. Mol Ecol 22:5705–5718
Pellmyr O (1999) Systematic revision of the yucca moths in the Tegeticula yuccasella complex
(Lepidoptera: Prodoxidae) north of Mexico. Syst Entomol 24:243–271
Pellmyr O, Krenn HW (2002) Origin of a complex key innovation in an obligate insect-plant
mutualism. Proc Natl Acad Sci U S A 99:5498–5502
Pellmyr O, Leebens-Mack J (1999) Forty million years of mutualism: evidence for eocene origin of
the yucca-yucca moth association. Proc Natl Acad Sci U S A 96:9178–9183
Petrusek A, Tollrian R, Schwenk K, Haas A, Laforsch C (2009) A “crown of thorns” is an inducible
defense that protects Daphnia against an ancient predator. Proc Natl Acad Sci U S A
106:2248–2252
Sanger TJ, Revell LJ, Gibson-Brown JJ, Losos JB (2012) Repeated modification of early limb
morphogenesis programmes underlies the convergence of relative limb length in Anolis lizards.
Proc R Soc Lond Ser B Biol Sci 279:739–748
Schluter D (2016) Speciation, ecological opportunity and latitude. Am Nat 187:1–18
Thompson JN (1994) The coevolutionary process. University of Chicago Press, Chicago
Thompson JN (2013) Relentless evolution. University of Chicago Press, Chicago
Thompson JN, Fernandez CC (2006) Temporal dynamics of antagonism and mutualism in a
geographically variable plant-insect interaction. Ecology 87:103–112
Thompson JN, Schwind C, Guimarães PR Jr, Friberg M (2013) Divergence through multitrait
evolution in coevolving interactions. Proc Natl Acad Sci U S A 110:11487–11492
Morphological Disparity

Melanie J. Hopkins and Sylvain Gerber

Abstract
Morphological disparity, the measure of morphological variation among species
and higher taxa, has been at the core of an important research program in
paleobiology over the last 25 years. Its quantification is based on the construction
and exploration of morphospaces, multidimensional spaces spanned by a set of
morphological descriptors, and benefits from a well-established analytical proto-
col. Two main classes of indices are routinely used to describe the distribution of
taxa in morphospace in terms of their spread and spacing. This unique focus on
the morphological component of clade dynamics has promoted disparity as a
distinct measure of biodiversity complementing traditional taxonomic proxies.
Disparity studies have led to improved understanding of the evolutionary history
of major clades and fostered new research on adaptive radiations, rates of
evolution, and morphological innovation. Currently, active areas of methodolog-
ical development focus on characterizing the geometric properties of
morphospaces, devising indices that describe the structure of disparity, and
incorporating phylogenetic information. There have also been increasing efforts
to identify the determinants of disparity, from developmental to functional and
ecological considerations, leading to conceptual extensions such as allometric
disparity. The importance of trends, extinction, and chance as factors in the
evolution of disparity remains relatively underexplored and needs more attention.

Melanie J. Hopkins and Sylvain Gerber contributed equally to this work.

M.J. Hopkins (*)
Division of Paleontology, American Museum of Natural History, New York, NY, USA
e-mail: mhopkins@amnh.org
S. Gerber
Institut de Systématique, Évolution, Biodiversité, UMR 7205 MNHN-CNRS-EPHE-UPMC-
Sorbonne Universités, Muséum National d’Histoire Naturelle, Paris, France
e-mail: sylvain.gerber@mnhn.fr

# Springer International Publishing AG 2017 1

L. Nuño de la Rosa, G.B. Müller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_132-1
2 M.J. Hopkins and S. Gerber

Keywords
Evo-devo • Morphospace • Morphometrics • Phenotypic evolution • Diversity •
Macroevolution

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Measuring Disparity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Morphological Descriptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Disparity Indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Visualizing Morphospaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Current Issues in Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Explaining the Evolution of Disparity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Developmental Morphospaces and Allometric Disparity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Functional Morphospace and Functional Disparity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Spatial, Environmental, and Temporal Structure of Morphospace . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Extinction and Extinction Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Chance and Historical Contingency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Introduction

Many macroevolutionary patterns in deep time can and have been discussed in
purely taxonomic terms. Significant episodes of radiation and extinction, for
instance, can be detected from the documentation of changes in the number of
species through time. However, even though our ability to distinguish species
usually implies the existence of morphological differences between them, change
in the number of species alone does not convey the magnitude of these morpholog-
ical differences.
This is one of the reasons for the rejuvenation of morphological disparity analyses
as an important research agenda in paleobiology and macroevolution since the
1990s. Morphological disparity is a macroevolutionary measure of morphological
variation. Although it has been used historically in reference to the level of morpho-
logical variation observed among body plans, its current and most widespread use is
as a description of the degree of morphological distinctness within a set of taxa at and
above the species level (morphological diversity within species being intraspecific
variation).
Estimates of morphological disparity have been employed primarily to document
the evolutionary history of particular clades. Fossil data has been of primary
importance for such studies. The fossil record contains examples of extinct mor-
phologies that may not be inferable from modern morphological diversity and
provides a direct record of temporal occurrence of different morphologies. Notable
patterns include the tendency for disparity to peak early in the evolutionary history of
Morphological Disparity 3

a clade (Hughes et al. 2013), and the frequent discordances between changes in
taxonomic diversity and morphological disparity over the evolutionary history of a
clade (Foote 1993a). Important reviews that also describe some of the impact that
disparity studies have had on evolutionary biology, particularly in the area of
adaptive radiations and rates of evolution, include Foote (1997), Wills (2001),
Erwin (2007), and Wagner (2010). In this chapter, we will briefly review common
methods for measuring disparity and then focus on the methodological and concep-
tual developments that occurred over the past decade.

Measuring Disparity

In empirical studies, the quantitative assessment of morphological disparity starts

with the definition of an adequate set of morphological descriptors from which can
be obtained a measure of dissimilarity between morphologies. In so doing, one
establishes a morphospace, the multidimensional state space spanned by these
morphological descriptors. The positioning of taxa relative to one another in the
morphospace reflects their degree of morphological similarity: the closer, the more
morphologically similar.

Morphological Descriptors

Different families of morphological descriptors exist:

1. Traditional morphometric descriptors. These are continuous data measured on a

ratio scale. They may include length or perimeter measurements, angles between
two linear features, estimates of area, or ratios between such measurements.
These descriptors often require some sort of transformation to make them com-
parable in terms of scale or units.
2. Geometric morphometric descriptors. These are sets of two- or three-dimensional
coordinate points whose configuration captures the geometry of the morpholog-
ical feature of interest. Each point is associated with an intersection, junction or
extreme (fixed landmarks), or with a curve (semi-landmarks and outlines).
3. Discrete character descriptors. These are categorical observations, such as the
absence or presence of a trait, or qualitative descriptions of different states of
expression of a trait that is present. They can have two or more states which may
be ordered or not. In order to estimate disparity from discrete character data, the
character-taxon matrix is converted to a pairwise distance matrix, or dissimilarity
matrix (see Lloyd 2016 for review). Most recent disparity studies based on
discrete character descriptors have co-opted character matrices that were origi-
nally constructed for phylogenetic purposes, and thus typically do not include
autapomorphies. Whether this is a problem remains an open issue: there is
currently no consensus on whether autapomorphies should be included in
4 M.J. Hopkins and S. Gerber

character matrices intended for disparity analyses (e.g., Gould 1991; Ruta and
Wills 2016).
4. Model-based descriptors. These are parameters that describe different shapes or
morphological features under a specified model, such as one describing growth.
A classic example is Raup’s 1966 model for shell coiling, which was based on the
geometry of the logarithmic spiral (Gerber 2016).

The use of traditional and geometric morphometric descriptors requires that all
taxa express the traits being measured. Because discrete character data sets can also
include information about the presence/absence of traits, a broader range of mor-
phologies can be accommodated in the analysis compared to morphometric data.
However, this is typically at the expense of a “cruder” description of morphologies.
In general, different types of descriptors capture and emphasize different aspects of
morphology, different levels of trait correlation and redundancy, and different scales
of change. Disparity patterns may or may not be consistent across different types of
descriptors even for the same set of sampled taxa; empirical studies to date are
summarized in Hopkins (2017).
All of these descriptors are conducive to morphospaces as defined above. The
term “morphospace” is thus a very broad designation that includes a great variety of
mathematical spaces. Those can have quite distinct properties and geometries (e.g.,
Gerber 2016), and they may differ in their renditions of given evolutionary patterns.
The investigator should therefore have an understanding of the properties of the
morphospace employed in order to tease apart the biological signal from potential
artifacts associated with a particular methodological approach.

Disparity Indices

Because morphological descriptors define a space, morphological disparity can also

be defined as the quantitative characterization of the spread and spacing of taxa in
this space, that is, the pattern of morphospace occupation. There are two components
to morphospace occupation, the amount and structure of disparity, but most studies
so far have exclusively focused on the former. The most commonly used disparity
indices are based on the standard measures of statistical dispersion and describe the
amount of disparity regardless of its structure.

1. Sum of (univariate) ranges, or total range. This metric represents the spread of the
distribution in morphospace. The sum of ranges is sensitive to sample size and is
thus frequently subjected to rarefaction analysis when comparisons are being
made between groups or samples of different numbers of taxa. The sum of ranges
is also dependent on orientation. This may have a nontrivial impact on compar-
ison of subgroups ordinated in the same morphospace, since the major axis of
variation of the entire group may differ from those of subgroups.
2. Sum of (univariate) variances, or total variance. Computed as the trace of the
covariance matrix, or equivalently, the sum of its eigenvalues, it describes the
Morphological Disparity 5

spacing of taxa in morphospace and is relatively insensitive to sample size. It is

not redundant with the previous index. For a given total range, different values of
total variance indicate a more or less densely occupied region of morphospace
within stable boundaries. Subgroups may contribute differentially to the total
variance; the contribution of the subgroup, or the partial disparity, is computed
from the sum of the squared distances of each member of the subgroup to the
overall centroid (Foote 1993b).

Wills (2001) described these and other indices in extensive detail, and
Ciampaglio et al. (2001) ran a series of simulations to study their behavior for
various types of morphospace patterns typically encountered in empirical case
studies. However, many morphospaces may exhibit an affine rather than a Euclidean
geometry. For such spaces, ratios of generalized variances (determinant of the
covariance matrix, or equivalently the product of the eigenvalues) have been
recommended as affine-invariant measures of disparity (Huttegger and Mitteroecker
2011). In the case of discrete character space, an alternative to total variance is the
average pairwise dissimilarity. One benefit of this index is that it is estimated directly
from the dissimilarity matrix and so does not require the use of ordination methods
(see below).

Visualizing Morphospaces

Unless one is concerned with very simple description of morphologies, the high
dimensionality of most morphospaces prevents the visualization of all their dimen-
sions at once. Getting a visual assessment of the extent and structuring of variation
thus requires the use of multivariate ordination methods such as principal component
analysis, principal coordinates analysis, or nonmetric multidimensional scaling,
which can extract the most relevant and salient features of the variation documented
with fewer dimensions.
Importantly, such a representation is a projection of the morphospace and not the
morphospace itself. This is often an informative and useful depiction but high-
dimensional spaces cannot be displayed as bivariate or trivariate plots without loss
of information, and thus these projections can be misleading. Fortunately, estimating
disparity does not rely on such projections and can be assessed from the entire
dataset (the true morphospace). The disparity indices mentioned above extend to any
number of dimensions, and there is therefore no need to resort to dimension
reduction techniques to measure disparity (e.g., keeping only the set of ordination
components that describe 95% of the original variance), even when low-variance
components contribute very little to the overall disparity.
It has been customary in the morphospace literature to distinguish empirical from
theoretical morphospaces. The latter, constructed from model-based descriptors, are
generally singled out as independent from the empirical sample of specimens studied
and capable of producing nonexistent morphologies, thus revealing areas of
morphospace that have not been occupied through evolution. These features,
6 M.J. Hopkins and S. Gerber

however, do not pertain to theoretical morphospaces only but typify many of the
so-called empirical morphospaces as well. The use of the empirical/theoretical
distinction generally reflects a confusion between the morphospace and its ordina-
tion (e.g., projection on principal components). The addition of new taxa may alter
the ordination but will not alter the relative distances between the previously
measured taxa.

Current Issues in Methods

Recent methodological developments have focused primarily on improving the

description of morphospace patterns and incorporating phylogenetic information in
their exploration.
Fairly distinct patterns of morphospace occupation have been documented and
common disparity indices can sometimes overlook their differences. For example,
equally disparate clades can show drastically different structuring of the amount of
morphological variation they display. Additional indices of morphospace occupation
have been suggested to characterize these distinct patterns, particularly with respect
to the dimensionality and the discontinuity of the distribution of taxa in morphospace
(e.g., clustering, Wills et al. 2012). Efforts to describe morphospace structure have
been complicated by two issues. First, tests for clustering lose statistical power as
dimensionality increases. Second, clustering can occur for artifactual reasons, as
well as biological reasons (see section “Explaining the Evolution of Disparity”
below). For example, regions of morphospace that cannot be occupied may exist
due to the use of character data which includes logically impossible character
combinations.
Recent disparity analyses have tended to focus on clades for which phylogenetic
hypotheses at the level of the OTUs described can be obtained. It is then possible to
map the hypothesized tree onto the morphospace, producing what has come to be
referred to as a phylomorphospace. Such a representation can help determine if
phylogeny is a strong contributor of the structuring of morphological variation in a
clade. Phylogenetic hypotheses have also been suggested as a means to account for
the incomplete fossil record of clades. Models of character evolution such as
parsimony can be used to assign character states to the internal nodes of the tree
(“hypothetical ancestors”), fill in missing data, and correct for ghost-ranges. These
approaches still need careful assessments of their statistical properties and heuristic
values. They do alter the raw disparity signal in ways (e.g., asymmetric adjustment
of stratigraphic range, superimposed model of character evolution) that might
obscure the true nature of the processes underlying the clade dynamics.
Finally, there has also been considerable recent interest in inferring past clade
disparity from the distribution of morphology among extant members of the clade,
usually with the aid of a phylogenetic hypothesis of how clade members are related
to one another evolutionarily. However, the robustness of such inferences has rarely
been tested against the fossil record. One recent attempt found that the morpholog-
ical disparity index was the most reliable for inferring an early burst of disparity in
Morphological Disparity 7

birds, but still failed to recapitulate disparity patterns in the bird fossil record
(Mitchell 2015). The inability to estimate extinct morphologies from extant taxa
alone is the major obstacle facing this sort of approach, especially when the structure
of morphospace occupation has changed, or when large or peripheral regions of
morphospace are no longer occupied.

Explaining the Evolution of Disparity

Over the last 25 years, disparity curves have been constructed for many groups –
although unevenly among vertebrate, invertebrate, and plant clades – following the
methodological framework outlined above. Together with diversity curves, these
disparity patterns have proved extremely useful in offering an expanded description
of the evolutionary history of biodiversity by combining its taxonomic and morpho-
logical components. Comparative surveys and reviews of these studies have focused
on the important task of documenting evolutionary histories patterns and assessing
their relative frequencies. These include for instance the various modes of morpho-
logical diversification (with or without concordance with taxonomic diversification)
and the selectivity versus randomness of extinction with respect to morphology.
The recognition of different patterns of disparity across different clades has also
led to interest in determining the processes that underlie the evolution of disparity
and the often heterogeneous patterning of morphospace. Historically, hypotheses
about the determinants of morphospace occupation have fallen into three categories:
extrinsic factors, biotic and abiotic; intrinsic factors related to development or
growth; and chance. On the theoretical side, a few stochastic and analytical models
have been implemented to explore the expected behavior of disparity in face of
changing mode of morphological transitions, taxonomic turnover rates, and size of
morphospace (e.g., Pie and Weitz 2005; Gerber et al. 2011 and references herein).
These studies have highlighted the difficulties in distinguishing the drivers of
disparity and isolating their signature from sampling error and stochastic variation.
More work is needed in this area to define null models and characterize the expected
signature of the various contributors of disparity.
In parallel, empirical efforts to understand the “mechanistic” bases of the dynam-
ics of disparity have concentrated on functional and developmental aspects of
morphological variation. These approaches have in common the (sometimes
implicit) recognition of morphological phenotypes as being made up of quasi-
independent units of evolutionary transformation, a phenomenon referred to as
evolutionary modularity. Internally, these units are developmentally and functionally
integrated and can be under distinct selective regimes and evolutionary constraints.
Hence, while the disparity signal built from the entire set of morphological descrip-
tors (characterizing the overall body) is relevant for documenting the evolutionary
history of a particular clade (global pattern), this more comprehensive description of
morphologies conflates the (quasi-independent) histories of its constitutive parts.
Relevant subsets of the global library of descriptors can be used to build disparity
signals attached to specific body parts corresponding to plausible modules, and may
8 M.J. Hopkins and S. Gerber

in fact be more informative in identifying the underlying processes driving the

evolution of disparity at the global level.
Accordingly, some authors have used character partitions to derive disparity
curves reflecting differential functional and developmental contributions, such as
external functional biology vs. internal anatomy, or ecological vs. nonecological
characters (e.g., Ciampaglio 2002). More recent works have developed approaches
directly targeting traits of specific functional and developmental relevance to phe-
notypic variation. We review them briefly below.

Developmental Morphospaces and Allometric Disparity

Morphological disparity analyses have traditionally focused on adult variation, but

changes in adult phenotypes are mediated by development, and development there-
fore influences morphospace occupation and disparity dynamics (Gerber 2014).
Fortunately, the morphological data retrievable from the fossil record are by no
means restricted to the adult stage and disparity can be measured at any develop-
mental stage. If the same set of descriptors is used to describe these different stages,
then juvenile and adult morphologies can be displayed within the same
morphospace. Earlier studies incorporating multiple developmental stages focused
on changes in disparity through ontogeny (e.g., Zelditch et al. 2003; Eble 2003).
More recent studies have focused on characterizing and analyzing the developmental
trajectories themselves.
For example, Gerber et al. (2008) used coefficients describing allometric growth
patterns in fossil ammonoids as descriptors for defining a multidimensional space in
which each point represents an ontogenetic trajectory (allometric space). The distri-
bution of points in this space, that is, the allometric disparity, can be quantified with
the same indices as in standard disparity analysis. Allometric disparity can be
compared to morphological disparity and their relative behaviors make it possible
to distinguish different types of change in allometric trajectories, thus linking
changes in adult disparity to specific modes of developmental evolution.

Functional Morphospace and Functional Disparity

Likewise, it is possible to focus on morphological descriptors that are tied to

particular functions in the organisms under study. Commonly used characters relate
to feeding ecology or biomechanics; there have also been developments in the use of
models to derive functionally relevant descriptors (Anderson et al. 2011). Although
functional properties of organisms are related to their morphology, functions that
depend on multiple traits can potentially be met by many different morphologies
(Wainwright 2007). As a result, the functional disparity and morphological disparity
of a particular structure may only be weakly related in some systems. In general, the
variation described by anatomical and functional datasets for the same taxa are likely
Morphological Disparity 9

to be correlated to some degree but not coincident (e.g., Anderson and Friedman
2012).

Spatial, Environmental, and Temporal Structure of Morphospace

Morphologies may be linked with environmental or ecological parameters in other

ways when specific morphology-to-function associations are not known. For exam-
ple, mapping group affiliations – where the group is defined by some ecological
relatedness or habitat affinity – onto taxa in morphospace may reveal clustering
associated with geographic occurrence or niche occupation, as well as the temporal
changes in those associations (e.g., Hopkins 2014).
As should be obvious from any disparity curve (e.g., Foote 1993a), both disparity
and morphospace structure are strongly associated with time as clades evolve into
new areas of morphospace. Although morphological diversification is often concep-
tualized as a diffusive process of volume-filling, increases in disparity may be highly
structured and uneven due to underlying trends (which are often due to develop-
mental or functional constraints). Indeed, some trends may not increase disparity at
all, if areas of morphospace are abandoned as the clade mean shifts (Hopkins 2016).
Most of the literature on trends has been concerned with their documentation and
categorization for univariate traits, such as body size. Categorical schemes are
intended to be indicative of some underlying processes, such as species selection
and constraints in the form of upper or lower bounds, but in general, the relationship
between trends and the evolution of disparity is largely unexplored.

Extinction and Extinction Space

Areas of morphospace can only be abandoned through extinction. As such, extinc-

tion, particularly selective extinction, necessarily alters morphospace occupation and
structure and therefore impacts disparity. For example, extinction selectivity has
been associated with morphological specialization in some (but not all) clades. Korn
et al. (2013) used indices which describe changes in morphospace occupation to
define a multidimensional space for distinguishing between different selectivity
modes during mass extinction events. One advantage of their approach was that it
is not necessarily context-dependent. Because the indices summarize change in
morphospace without recourse to the descriptors defining that morphospace, results
based on disparate morphospaces can be compared to one another.

Chance and Historical Contingency

The importance of historical contingency in shaping patterns of disparity can be

grasped most readily in the context of extinction, where the random culling of taxa
can substantially alter the dynamics of disparity and put the evolutionary history of a
10 M.J. Hopkins and S. Gerber

clade on new tracks. Contingency pervades at all scales however, and extends
beyond the case of random patterns of survivorship. The effect of particular contin-
gent historical events on disparity is nevertheless difficult to apprehend, because
they can be context-dependent and affect only one or a few lineages within a clade,
or be restricted to specific areas of the clade’s geographic distribution. Their effects
can still be significant at the global scale, however, and induce temporal shits in
disparity. Contingent explanations can only emerge from detailed studies and “dis-
sections” of disparity signals (e.g., geographic partitions and subclade components
of global signals) and should not be confounded with general properties and
common “laws” that might underlie all clades’ histories. The difficult task of
disentangling these classes of explanation invites a better characterization of the
expected behavior of disparity, abstracted from the volatility of taxonomic rates, and
the proposal of adequate models of diffusion in morphospace accounting for the
properties of the mechanisms underlying evolutionary change in morphology.

Concluding Remarks

Since Gould’s advocacy for the study of morphological disparity as an important

macroevolutionary quantity (Gould 1991), paleontologists have successfully
unearthed the history of many groups in terms of changes in diversity and changes
in disparity. In so doing, they also have highlighted the frequent decoupling of these
two facets of biodiversity. Over the years, methodological approaches to the study of
disparity have been standardized in some ways, primarily through the maturation of
morphometric techniques and the increasingly popular use of just a few informative
indices of morphospace occupation.
These established analytical routines are powerful and will undoubtedly continue
to be used to document disparity patterns for many clades. It is also clear, however,
that many questions and issues, including some raised in the early years of the
disparity research program, are still unresolved and/or lack appropriate conceptual
and methodological frameworks for their analyses.
For example, recent research into the properties of different kinds of
morphospaces has revealed situations in which classic measures of morphological
distance (and thus disparity) might not be mathematically or evolutionary meaning-
ful. Some features of morphospace occupation can also occur for both biological and
artifactual reasons. All of these can affect measures of disparity and mislead our
descriptions of evolutionary patterns.
Indices that measure the structure of disparity have been neglected compared to
those that measure the amount of disparity, and we therefore know much less about
the evolution of morphospace structuring (e.g., discreteness and dimensionality) as
clades wax and wane, and from a technical viewpoint, the impact of sample size,
taxonomic error, and morphospace dimensionality on such indices.
In terms of data, while cladistic matrices are increasingly used as discrete
character spaces in disparity analyses and offer the possibility to combine
morphospace and phylogenetic approaches, little is known with regard to the validity
Morphological Disparity 11

of their use as morphospaces (in particular with respect to their usually large amount
of missing data, but see Lloyd 2016 for a start).
In parallel to continued empirical research and methodological development,
there is a growing need for mathematical models of disparity and of diffusion in
morphospace. Explicit incorporation of stochastic effects is a major component in
the study of trait evolution within lineages (e.g., “fossil time series”) or across trees;
in fact, in this area, there has been a recent shift away from the use of a stochastic
model as a null hypothesis towards model selection approaches that use some
criterion to select from among a set of models, of which one may represent
stochasticity. Similarly, disparity studies may benefit from a shift towards statistical
inference, with models that include potential determinants such as growth patterns,
modular anatomical organization, functional constraints, selection, extinction, con-
tingency, and chance, and away from post hoc explanations of descriptive patterns.

Cross-References

▶ Evolution of Complexity
▶ Macroevolution
▶ Mass Extinctions
▶ Methods and Practices in Palaeo-Evo-Devo
▶ Morphometrics and Evo-Devo

References
Anderson PSL, Friedman M (2012) Using cladistic characters to predict functional variety: exper-
iments using early gnathostomes. J Vertebr Paleontol 32:1254–1270
Anderson PSL, Bright JA, Gill PG, Palmer C, Rayfield EJ (2011) Models in palaeontological
functional analysis. Biol Lett 8:119–122
Ciampaglio CN (2002) Determining the role that ecological and developmental constraints play in
controlling disparity: examples from the crinoid and blastozoan fossil record. Evol Dev
4:170–188
Ciampaglio CN, Kemp M, McShea DW (2001) Detecting changes in morphospace occupation
patterns in the fossil record: characterization and analysis of measures of disparity. Paleobiology
27(4):695–715
Eble GJ (2003) Developmental morphospaces and evolution. In: Crutchfield JP, Schuster P (eds)
Evolutionary dynamics: exploring the interplay of selection, accident, neutrality, and function.
Oxford University Press, Oxford, pp 33–63
Erwin DH (2007) Disparity: morphological pattern and developmental context. Palaeontology 50
(1):57–73
Foote M (1993a) Discordance and concordance between morphological and taxonomic diversity.
Paleobiology 19(2):185–204
Foote M (1993b) Contributions of individual taxa to overall morphological disparity. Paleobiology
19(2):403–419
Foote M (1997) The evolution of morphological diversity. Annu Rev Ecol Evol Syst 28:129–152
Gerber S (2014) Not all roads can be taken: development induces anisotropic accessibility in
morphospace. Evol Dev 16(6):373–381
12 M.J. Hopkins and S. Gerber

Gerber S (2016) The geometry of morphospaces: lessons from the classic Raup shell coiling model.
Biol Rev. doi:10.1111/brv.12276
Gerber S, Eble GJ, Neige P (2008) Allometric space and allometric disparity: a developmental
perspective in the macroevolutionary analysis of morphological disparity. Evolution 62
(6):1450–1457
Gerber S, Eble GJ, Neige P (2011) Developmental aspects of morphological disparity dynamics: a
simple analytical exploration. Paleobiology 37:237–251
Gould SJ (1991) The disparity of the Burgess Shale arthropod fauna and the limits of cladistic
analysis: why we must strive to quantify morphospace. Paleobiology 17(4):411–423
Hopkins MJ (2014) The environmental structure of trilobite morphological disparity. Paleobiology
40(3):352–373
Hopkins MJ (2016) Magnitude versus direction of change and the contribution of macroevolution-
ary trends to morphological disparity. Biol J Linn Soc 118(1):116–130
Hopkins MJ (2017) How well does a part represent the whole? A comparison of cranidial shape
evolution with exoskeletal character evolution in the trilobite family Pterocephaliidae.
Palaeontology 60:309–318
Hughes M, Gerber S, Wills MA (2013) Clades reach highest morphological disparity early in their
evolution. Proc Natl Acad Sci U S A 110(34):13875–13879
Huttegger SM, Mitteroecker P (2011) Invariance and meaningfulness in phenotypic spaces. Evol
Biol 38:335–351
Korn D, Hopkins MJ, Walton SA (2013) Extinction space – a method for the quantification and
classification of changes in morphospace across extinction boundaries. Evolution 67
(10):2795–2810
Lloyd GT (2016) Estimating morphological diversity and tempo with discrete character-taxon
matrices: implementation, challenges, progress, and future directions. Biol J Linn Soc 118
(1):131–151
Mitchell JS (2015) Extant-only comparative methods fail to recover the disparity preserved in the
bird fossil record. Evolution 69(9):2414–2424
Pie MR, Weitz JS (2005) A null model of morphospace occupation. Am Nat 166:E1–D13
Ruta M, Wills MA (2016) Comparable disparity in the appendicular skeleton across the fish-
tetrapod transition, and the morphological gap between fish and the tetrapod postcrania.
Palaeontology 59:249–267
Wagner PJ (2010) Paleontological perspectives on morphological evolution. In: Bell MA, Futuyma
DJ, Eanes WF, Levinton JS (eds) Evolution since Darwin: the first 150 years. Sinauer Associ-
ates, Sunderland, pp 451–478
Wainwright PC (2007) Functional versus morphological diversity in macroevolution. Annu Rev
Ecol Evol Syst 38:381–401
Wills MA (2001) Morphological disparity: a primer. In: Adrain JM, Edgecombe GD, Lieberman BS
(eds) Fossils, phylogeny, and form: an analytical approach. Topics in geobiology. Kluwer
Academic/Plenum Publishers, New York, pp 55–145
Wills MA, Gerber S, Ruta S, Hughes M (2012) The disparity of priapulid, archaeopriapulid and
palaeoscolecid worms in the light of new data. J Evol Biol 25:2056–2076
Zelditch ML, Sheets HD, Fink WL (2003) The ontogenetic dynamics of shape disparity. Paleobi-
ology 29(1):139–156
The Impact of Atmospheric Composition
on the Evolutionary Development
of Stomatal Control and Biochemistry
of Photosynthesis over the Past 450 Ma

Matthew Haworth, Giovanni Marino, and Mauro Centritto

Abstract
The conversion of carbon dioxide (CO2) and water into glucose and oxygen (O2)
by photosynthesis has been a central component of the atmosphere and climate
system over Earth history. The diffusive uptake of CO2 and its biochemical
assimilation have in turn been strongly affected by atmospheric composition.
Here, we illustrate how declining [CO2] and rising [O2] have exerted selective
pressures to reduce the uptake of O2 (photorespiration) in favor of CO2 (photo-
synthesis) by the enzyme ribulose-1,5-bisphosphate carboxylase/oxygenase
(rubisco). In the last 10 Myr when [CO2] fell to less than 300 ppm, C3 photo-
synthesis became less efficient and mechanisms concentrating CO2 at the rubisco
active site were favored leading to the expansion of C4 photosynthesis. The need
to optimize carbon gain relative to water-loss has acted as a key selective pressure
in the evolutionary development of stomatal function and epidermal patterning, to
not only maximize diffusion of CO2 into the leaf but also regulate excessive
transpirative water-loss. This stomatal control of photosynthesis generally allows
angiosperms to sustain greater levels of stomatal conductance and CO2-uptake
than species with more ancient evolutionary origins. This is particularly evident
in the grasses, where dumb-bell stomata and the allocation of a higher percentage
of the epidermis to gas exchange permit greater rates of stomatal conductance and
photosynthesis than species with kidney-shaped stomata. The diffusive and
biochemical components of photosynthesis have been strongly influenced by
declining CO2:O2 over the past 100 Myr. However, current rising [CO2] may
affect these selective pressures, having implications for future plant growth.

M. Haworth (*) • G. Marino • M. Centritto

Tree and Timber Institute, National Research Council (CNR – IVALSA), Florence, Italy
e-mail: haworth@ivalsa.cnr.it; marino@ivalsa.cnr.it; mauro.centritto@cnr.it

# Springer International Publishing AG 2017 1

L. Nuño de la Rosa, G.B. Müller (eds.), Evolutionary Developmental Biology,
DOI 10.1007/978-3-319-33038-9_171-1
2 M. Haworth et al.

Keywords
Rubisco • Evolution of stomata • Photorespiration • Diffusive limitations • Atmo-
spheric carbon dioxide • Atmospheric oxygen

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Rubisco Specificity and Levels of Atmospheric CO2 and O2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Diffusive Resistances to CO2-Uptake and Stomatal Control of Photosynthesis . . . . . . . . . . . . . . . . 6
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Cross-References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

Introduction

Photosynthesis utilizes light energy to convert carbon dioxide (CO2) and water into
glucose and oxygen (O2). This process has shaped the land, atmosphere, and oceans
over Earth history. In turn, the evolutionary development of photosynthesis has also
been influenced by environmental changes through the selective pressures exerted
by the respective rates of carboxylation and oxygenation, alongside the need to take-
up, transport, and conserve water. These selective pressures can be encapsulated
by the measure “water use efficiency” (WUE) that indicates the amount of CO2
taken-up relative to water-loss (Fig. 1). Ribulose-1,5-bisphosphate carboxylase/
oxygenase (rubisco) is the central enzyme involved in photosynthesis, determining
rates of carboxylation (photosynthesis) and oxygenation (photorespiration) of
ribulose-1,5-bisphosphate. Photosynthetic rubisco likely evolved from similar pro-
teins involved in methane formation and sulphur metabolism in anoxic environments
4.0 to 3.0 Ga. The expansion of photosynthetic microbes led to the oxygenation of
the Earth’s atmosphere 2.4 Ga. The increased abundance of atmospheric oxygen
made the environment less favorable to the previously dominant anaerobic microbial
organisms, and over time led to the development of the ozone layer, shielding the
Earth from high energy UV radiation. The drawdown of atmospheric CO2 associated
with the oxygenation of the atmosphere may also have induced a global glaciation
(early Proterozoic Snowball Earth 2.4 Ga); possibly marking the first occurrence
whereby photosynthesis significantly influenced the Earth’s climate through its
effects on atmospheric composition (Nisbet and Nisbet 2008). However, such
alterations in atmospheric composition and temperature would also influence pho-
tosynthesis via the lower availability of CO2, greater abundance of oxygen, and
effects on enzyme activity. The action of photosynthetic organisms over Earth
history has been proposed to have led to the development of comparative stability
in the levels of atmospheric [CO2] and [O2] that mirror the kinetics of rubisco
(Tolbert et al. 1995; Tcherkez et al. 2006). It is the selective pressures exerted by
this interaction between photosynthetic processes and the environment that we hope
to briefly outline in this chapter on the evolutionary development of the diffusive
characteristics and biochemistry of photosynthesis. We will focus on the
The Impact of Atmospheric Composition on the Evolutionary Development of. . . 3

temperature
wa
ava ter
ilab
ility
:O 2
2
CO

low high

C3 photosynthesis selective pressure C4 photosynthesis

passive stomata to optimise WUE active stomata
low stomatal density high stomatal density

Fig. 1 An illustration of the impact of mean temperature during the growing season, water
availability, and the atmospheric CO2:O2 ratio on selective pressures to optimize water use
efficiency (WUE). High temperatures reduce the specificity of rubisco, decrease the activity of
rubico activase, and increase the solubility of O2 in comparison to CO2. Low water availability
restricts potential for stomatal conductance and CO2 uptake. Low atmospheric CO2:O2 ratios favor
photorespiration over photosynthesis

evolutionary development of photosynthesis and its accompanying metabolic and

regulatory processes in vascular plants over the past 450 million years, as experi-
ments with living plants with ancient evolutionary histories have allowed an under-
standing of the selective pressures that have shaped present day vegetation.
Moreover, it is these same selective pressures that will determine the response of
terrestrial vegetation to current climatic and atmospheric changes.

Rubisco Specificity and Levels of Atmospheric CO2 and O2

Photosynthetic rubisco does not have identical affinities for both CO2 and O2.
Rubisco will preferentially select CO2 over O2; this is known as the specificity of
rubisco. However, there is vastly more O2 in the atmosphere than CO2, with O2
4 M. Haworth et al.

120

photosynthesis (%) 100

80

60

40 C4

C3 - C4 intermediate
20
C3
0
0 5 10 15 20 25 30 35 40
[O2] (%)

Fig. 2 The effect of increasing atmospheric concentration of oxygen on the proportion of photo-
synthesis relative to photorespiration of plant species with C3 photosynthesis (Rosoideae rosa
hybrid tea rose; triangle white fill), C3–C4 intermediate photosynthesis (Moricandia arvensis;
square black fill) and C4 photosynthesis (Zea mays; circle grey fill). Photosynthesis is expressed
as a percentage of its value at 1.5% [O2] where levels of photorespiration are assumed to be
negligible. Error bars indicate one standard error either side of the mean. The concentration of
[CO2] remained constant at 400 ppm at all [O2] levels

measured as a percentage while CO2 is determined on a parts-per-million (ppm: i.e.,

μmol mol 1) basis. This disparity in the concentrations of [CO2] and [O2] in
the atmosphere has acted as a selective pressure on the evolution of plants through-
out the past 450 Ma exerted via the competing processes of photosynthesis and
photorespiration. As the CO2:O2 ratio rises, the proportion of photosynthesis to
photorespiration increases in C3 plants (Fig. 2). Analysis of rubisco specificity in
phylogenetically diverse plants suggests that the atmospheric CO2:O2 ratio over
Earth history (Fig. 3a) has affected the characteristics of rubisco. Those species
originating during periods of comparatively lower CO2:O2 tend toward higher
specificity for CO2, but lower maximum rates of carboxylation. This lower carbox-
ylation capacity is generally compensated by possession of higher concentrations
of rubisco (Galmes et al. 2014). However, rubisco is a comparatively expensive
compound for plants as it requires the investment of nitrogen (the availability of
nitrogen frequently limits plant growth in many terrestrial habitats), and rubisco
often accounts for more than 25% of leaf nitrogen. At [CO2] levels below 200 ppm,
characteristic of atmospheric [CO2] during glacial episodes, the rates of photorespi-
ration begin to approximate those of photosynthesis (Tolbert et al. 1995). The
increase in photorespiration and the decline in photosynthesis at low [CO2] make
C3 photosynthesis less economic (Taylor et al. 2014). This exerted selective pres-
sures favoring adaptations to reduce photorespiration that led to the origination in the
Eocene (or possibly earlier) and expansion 5–7 Ma, of C4 photosynthesis (Sage et al.
2012). C4 photosynthesis reduces photorespiration by effectively concentrating CO2
 a b 40
[O2 ] [CO2 ]
(%) (ppm)

-1
30

-2
35 4500 800

stomata
planate leaf
cycads and
Ginkgoaceae
conifers
angiosperms
C3 grasses
C4 grasses
20
4000 700
30

P N (ɥmol m s )
3500 10 R2 = 0.859
600 F1,27 = 172.2
-13
P = 3.1 x 10
3000 0
500 0 200 400 600 800 1000
-2 -1
25 G s (mmol m s )
2500
400 c 0.8 gymnosperms angiosperms

-1
2000
20 angiosperms 0.6
300

number of species
-1
1500

-2
200 0.4
1000
15
gymnosperms 100 0.2
500 early
G m (mol m s bar )

vascular
plants pteridophytes
10 0 0 0.0
-450 -400 -350 -300 -250 -200 -150 -100 -50 0
fern
cycad
tree

conifer

time (Ma)
eudicot
eudicot
herb
monocot

Fig. 3 (a) Levels of atmospheric [CO2] (blue line) and [O2] (red line) (from Berner 2009) and the diversification of major plant groups (Niklas et al. 1983) over
the past 450 Ma. The origination of plant groups and/or occurrence of morphological/physiological adaptations are marked by dashed vertical lines. (b) The
The Impact of Atmospheric Composition on the Evolutionary Development of. . .

relationship between photosynthesis (PN) and stomatal conductance (Gs) of vascular plants under present atmospheric conditions (ferns = white circle;
cycads = white triangle; Ginkgo biloba = inverted grey triangle; conifers = black diamonds; angiosperms = white squares). The black line indicates best fit and
the two grey lines either side indicate 95% confidence intervals of the mean. (c) Mesophyll conductance to CO2 (Gm) of a fern (Cyrtomium fortunei), cycad
(Lepidozamia perroffskyana), conifer (Agathis australis), eudicot tree (Olea europea), eudicot herb (Gossypium hirsutum), and monocot (Avena sativa). All
5

error bars indicate standard error

6 M. Haworth et al.

within the bundle sheath at the active site of rubisco (Fig. 2). This mechanism is also
favored at higher temperatures where the specificity of rubisco for CO2 declines and
the solubility of O2 relative to CO2 increases (Fig. 1) (Crafts-Brandner and Salvucci
2002). The transition from C3 to C4 is complex, involving a large number of genetic
mutations to produce the morphological (Kranz anatomy) and physiological (prin-
cipally the enzyme phosphoenolpyruvate carboxylase as a CO2 acceptor and nico-
tinamide adenine dinucleotide phosphate-malic enzyme) adaptations responsible for
C4 photosynthesis (Sage et al. 2012). However, this transition likely occurred in
stages of mutations that offered incremental selective advantages. This has resulted
in species with photosynthesis bearing characteristics of both C3 and C4 physiol-
ogies, the so-called C3–C4 intermediates (also known as “C2” plants). These C3–C4
intermediates do not minimize photorespiration to an extent comparable to C4
plants, but rather recapture for photosynthesis a higher percentage of the CO2
released via photorespiration (the development of a so-called photorespiratory
bypass) (Fig. 2). The transition from C3 to C4 has occurred more than 60 times in
different lineages of plants (Sage et al. 2012; Osborne and Beerling 2006),
suggesting that the selective pressures of low [CO2]/high [O2] are highly influential,
and that many C3 plants possess the genetic capability to “switch on” C4 attributes –
this has formed the basis for numerous projects to improve the WUE of C3 staple
crops by modifying their physiologies toward C4 photosynthesis (Gowik and
Westhoff 2011). Plants still require a degree of photorespiration, as it is an important
component of carbon metabolism and plays a protective role against abiotic stresses
such as excess light energy. Mutant plants without the capacity for photorespiration
are unable to tolerate abiotic stress and have low levels of survival. Nonetheless, the
reduction of photorespiratory losses of carbon would imply that C4 plants should
experience a selective advantage under the present atmospheric conditions of com-
paratively low [CO2]/high [O2] in the context of much of the preceding 450 Myr.
However, C4 photosynthesis is energetically more expensive than C3, requiring two
thirds more adenosine triphosphate to fix each molecule of CO2. Indeed, the majority
of C4 plants occur in warm to hot regions where water is limited and temperatures
unfavorable to C3 photosynthesis (some C4 plants also occur in warm humid
nutrient poor environments where C4 photosynthesis ameliorates the impact of
low nutrient availability on CO2-uptake), illustrating the costs and benefits associ-
ated with both photosynthetic physiologies (Fig. 1).

Diffusive Resistances to CO2-Uptake and Stomatal Control

of Photosynthesis

As the availability of CO2 for photosynthesis declines, a frequent response of plants

is to promote diffusion of CO2 into the leaf by opening stomata and increasing
stomatal conductance (Gs). Stomata are the pores covering the leaf that contribute
to the maintenance of leaf homeostasis by regulating the uptake of CO2 for photo-
synthesis against the loss of water via transpiration. A stomatal complex consists of
two guard cells surrounding a stomatal pore through which CO2 is taken-up for
The Impact of Atmospheric Composition on the Evolutionary Development of. . . 7

photosynthesis and water lost via transpiration. Turgor changes in the guard cell
regulate the size of the stomatal aperture. The guard cells may be surrounded by
specialized epidermal cells known as subsidiary cells. The actions of stomata are
closely coordinated with the availability of CO2 and rates of photosynthesis in the
mesophyll driven by the carboxylation activity of rubisco outlined above (Engineer
et al. 2016). Stomata first occur in the fossil record more than 400 Ma and played a
central role in the colonization of the land by early plants. These early stomata are
largely identical to “kidney-shaped” stomata found in most dicots, suggesting that
their structure and function (particularly in preventing desiccation) have remained
largely unchanged (Edwards et al. 1998). However, the earliest stomata-like struc-
tures were probably not involved in gas exchange, but the distribution of spores by
allowing spore-bearing tissues to dry (Duckett et al. 2009). The genes encoding
proteins responsible for the development of stomata in spore-bearing tissues of the
moss Physcomitrella patens are orthologs of the genes responsible for stomatal
development in the model angiosperm Arabidopsis thaliana (Chater et al. 2016).
The evolutionary exaptation of stomata for gas exchange allowed early terrestrial
plants to take-up CO2 for photosynthesis at a faster rate. As [CO2] fell during the late
Devonian, the number of stomata on the leaf surfaces of Devonian plants increased.
The higher rates of transpirative cooling associated with this increased Gs are
proposed to have facilitated the evolutionary development of the planate leaf by
reducing thermal stress; larger leaves intercept more light energy resulting in
increased temperature, and also lose less heat energy via transpiration due to the
presence of a more developed “boundary layer” of still air over the leaf surface. This
increase in leaf area served as a selective advantage, enabling plants to capture
greater amounts of light for photosynthetic carbon gain (Beerling et al. 2001).
The origination of many groups of plants or major physiological/morphological
adaptations has coincided with intervals of low or falling [CO2] and/or high [O2]
(Fig. 3a). This may have favored those species with greater stomatal control that are
able to maximize Gs and potential diffusive uptake of CO2, but also capable of
restricting water-loss under unfavorable growth conditions. When measured under
current ambient [CO2] of 400 ppm, the rates of photosynthesis are closely linked to
Gs in phylogenetically diverse plants (Fig. 3b). Generally, more recently derived
angiosperms exhibit higher photosynthesis and Gs than more ancient groups (ferns,
cycads, Ginkgoales, conifers) that originated in atmospheres of higher CO2:O2.
Moreover, analysis of the conductance of CO2 across the mesophyll layer from the
substomatal air-space to the chloroplast (termed “mesophyll conductance”: Gm)
would suggest such an evolutionary trajectory is replicated, with angiosperms
possessing greater capacity for the uptake of CO2 (Fig. 3c). However, more exten-
sive screening of Gm in different plant phylogenies is necessary before any firm
conclusions may be drawn regarding evolutionary development of CO2 transport
across the mesophyll. Nonetheless, the greater potential for gas exchange observed
in the angiosperms would only constitute a selective advantage if accompanied by
more effective stomatal control to regulate water-loss when photosynthesis is limited
and prevent desiccation during drought or high evapotranspirative demand (Haworth
et al. 2011). Stomatal control is the speed of stomatal aperture adjustment to a change
8 M. Haworth et al.

in conditions (e.g., light, [CO2], water availability) and the tightness of stomatal
closure when photosynthesis is not taking place (i.e., at night or under severe
drought). There are two main modes of stomatal behavior: passive and active. The
guard cells of passive stomata follow changes in leaf water potential. As such,
passive stomatal behavior is generally slower when responding to an external
stimulus. In contrast, the active pumping of osmolytes across the plasma membrane
of guard cells allows species with active stomatal behavior to regulate Gs more
rapidly. This has led to a hypothesis proposing an evolutionary trajectory from
passive to active stomatal behavior, with the stomata of angiosperms exhibiting
increased sensitivity to light, super-ambient [CO2], leaf-to-air vapor pressure deficit
(a measure of evapotranspirative demand placed on the leaf), and the stress hormone
abscisic acid (McAdam and Brodribb 2012). However, genetic analysis of mosses
and lycophytes (Chater et al. 2013) and gas exchange analyzes of Gs in phylogenet-
ically diverse plants (Hasper et al. 2017) indicate that active stomatal behavior
originated in early plant lineages. Nevertheless, the higher potential rates of Gs
and photosynthesis observed in angiosperms would indeed constitute a selective
advantage in a “low [CO2]” world.
Evolutionary developments in epidermal patterning may account for the higher
rates of Gs observed in angiosperms via an increase in the proportion of the cuticle
allocated as stomatal complexes and therefore potentially available for gas exchange
(de Boer et al. 2016). Larger numbers of smaller stomata are thought to be more
efficient in achieving high Gs but also offering greater stomatal control, as smaller
stomata are considered to close more rapidly (Raven 2014). This may account for
observations of generally low densities of large stomata in groups such as the ferns
and cycads in comparison to higher densities of small stomata in angiosperms
(Franks and Beerling 2009). This epidermal patterning in angiosperms may also
allow for more complex vein architecture within the leaf to enable more efficient
transport of water (Roth-Nebelsick et al. 2001). While a eudicot (angiosperm) with a
high density of small kidney-shaped stomata sustains higher Gs than a cycad, when
the rates of stomatal closure are normalized they are broadly consistent. This may
indicate that the individual kidney-shaped stomata of an evolutionarily ancient cycad
are behaving in a manner similar to those of the more recently derived angiosperm
(Fig. 4). However, while the majority of species possess kidney-shaped stomata, the
stomatal complexes of grasses are morphologically different, termed “dumb-bell”
stomata. Dumb-bell stomata are mechanically different to kidney-shaped stomata in
their opening/closing mechanism. As dumb-bell stomata are larger, to fully open
they rely not only on an increase in the turgor of guard cells, but also a greater loss in
turgor pressure in the adjacent subsidiary cells to accommodate full stomatal open-
ing. This permits larger and more rapid stomatal movements in grasses (Fig. 4)
(Franks and Farquhar 2007) and may have contributed to the expansion of grasses
over the past 20 Myr as atmospheric [CO2] has declined (Robinson 1994).
These stomatal modifications are important not only in terms of optimizing
carbon gain relative to water-loss, but also in the protection of water transport
systems. Under conditions of low water availability and high evapotranspirative
demand plants may experience xylem embolism, where air enters the xylem causing
 1000 180
160 kidney shaped
800 large stomata
140 low density

-2
120
600
100
kidney shaped
80 small stomata
400 high density

Gs (mmol m s-1)
60

normalised Gs (%)
200 40
20 dumb-bell shaped
large stomata
0 0 high density
-500 0 500 1000 1500 2000 -500 0 500 1000 1500 2000
time (seconds) time (seconds)

Fig. 4 Absolute and normalized rates of stomatal closure to darkness (indicated by a vertical dashed line and change from yellow to black in the upper
horizontal bar) of a cycad (Cycas siamensis – blue line) and angiosperm (Chenopodium quinoa – green line) with kidney-shaped stomata and an angiosperm
monocot (Arundo donax – red line) with dumb-bell stomatal complexes. Panels on the right provide an illustration of stomatal morphology and stomatal
patterning
The Impact of Atmospheric Composition on the Evolutionary Development of. . .
9
10 M. Haworth et al.

a loss of water movement. Such xylem embolisms may be fatal to plants in a low
[CO2] world (particularly at high altitude where the partial pressure of CO2 is further
reduced) where high Gs is required for sufficient uptake of CO2 for photosynthesis.
Indeed, the shift from xylem tracheids to vessels with higher conductivity and/or
resistance to embolism may be related to increased stomatal control allowing
prevention of high levels of transpiration and development of excessive negative
pressures within the xylem (Meinzer et al. 2009; Sperry et al. 2006).

Conclusion

Over the past 100 Ma, atmospheric composition has been characterized by declining
[CO2] toward the point during quaternary glaciations where the carboxylation and
oxygenation rates of rubisco were broadly similar. The selective pressures resulting
from this low [CO2] have induced the evolutionary development of rubisco with
increased specificity for CO2 in C3 plants, C4 photosynthesis to concentrate CO2 at
the active site of rubisco, more efficient water transport systems, enhanced stomatal
control, and modified epidermal patterning to allocate a greater proportion of the leaf
surface to gas exchange. These changes have likely contributed to the decline of
more ancient plant groups and the expansion and diversification of the angiosperms
(Fig. 3a). However, atmospheric [CO2] has risen at a greater rate over the past
250 years than at any time in the past 65 Myr (with the possible exception of the
Paleocene-Eocene Thermal Maximum 55.8 Ma). This has the potential to alter the
strength of the selective pressures induced by atmospheric composition that have
driven modifications in the biochemistry of photosynthesis, diffusive limitations to
CO2 uptake, and stomatal control of photosynthesis over the past 65 Myr. A greater
understanding of the genetic regulation of photosynthesis, and its integration with
supporting gas exchange and water transport systems, will advance our understand-
ing of not only the evolutionary development of photosynthesis in major plant
groups, but the likely effects of future rising [CO2] on agriculture and natural
vegetation.

Acknowledgments We are grateful to funding from the EU FP7 project 3 to 4 (289582).

Cross-References

▶ Evo-Devo of the Origin of Flowering Plants

▶ Evolution of Development in Monocots

References
Beerling DJ, Osborne CP, Chaloner WG (2001) Evolution of leaf-form in land plants linked to
atmospheric CO2 decline in the late Palaeozoic era. Nature 410(6826):352–354
The Impact of Atmospheric Composition on the Evolutionary Development of. . . 11

Berner RA (2009) Phanerozoic atmospheric oxygen: new results using the Geocarbsulf model. Am
J Sci 309(7):603–606. doi:10.2475/07.2009.03
de Boer HJ, Price CA, Wagner-Cremer F, Dekker SC, Franks PJ, Veneklaas EJ (2016) Optimal
allocation of leaf epidermal area for gas exchange. New Phytol 210(4):1219–1228. doi:10.1111/
nph.13929
Chater C, Gray JE, Beerling DJ (2013) Early evolutionary acquisition of stomatal control and
development gene signalling networks. Curr Opin Plant Biol 16(5):638–646. doi:10.1016/j.
pbi.2013.06.013
Chater CC, Caine RS, Tomek M, Wallace S, Kamisugi Y, Cuming AC, Lang D, MacAlister CA,
Casson S, Bergmann DC (2016) Origin and function of stomata in the moss Physcomitrella
patens. Nat Plant 2:16179
Crafts-Brandner SJ, Salvucci ME (2002) Sensitivity of photosynthesis in a C4 plant, maize, to heat
stress. Plant Physiol 129(4):1773–1780. doi:10.1104/pp.002170
Duckett JG, Pressel S, P’ng KM, Renzaglia KS (2009) Exploding a myth: the capsule dehiscence
mechanism and the function of pseudostomata in sphagnum. New Phytol 183(4):1053–1063.
doi:10.1111/j.1469-8137.2009.02905.x
Edwards D, Kerp H, Hass H (1998) Stomata in early land plants: an anatomical and ecophysio-
logical approach. J Exp Bot 49:255–278
Engineer CB, Hashimoto-Sugimoto M, Negi J, Israelsson-Nordström M, Azoulay-Shemer T,
Rappel W-J, Iba K, Schroeder JI (2016) CO2 sensing and CO2 regulation of stomatal conduc-
tance: advances and open questions. Trends Plant Sci 21(1):16–30. doi:10.1016/j.
tplants.2015.08.014
Franks PJ, Beerling DJ (2009) Maximum leaf conductance driven by CO2 effects on stomatal size
and density over geologic time. Proc Natl Acad Sci U S A 106(25):10343–10347. doi:10.1073/
pnas.0904209106
Franks PJ, Farquhar GD (2007) The mechanical diversity of stomata and its significance in
gas-exchange control. Plant Physiol 143(1):78–87. doi:10.1104/pp.106.089367
Galmes J, Kapralov MV, Andralojc P, Conesa MÀ, Keys AJ, Parry MA, Flexas J (2014) Expanding
knowledge of the Rubisco kinetics variability in plant species: environmental and evolutionary
trends. Plant Cell Environ 37(9):1989–2001
Gowik U, Westhoff P (2011) The path from C3 to C4 photosynthesis. Plant Physiol 155(1):56–63
Hasper TB, Dusenge ME, Breuer F, Uwizeye FK, Wallin G, Uddling J (2017) Stomatal CO2
responsiveness and photosynthetic capacity of tropical woody species in relation to taxonomy
and functional traits. Oecologia 184(1):43–57
Haworth M, Elliott-Kingston C, McElwain JC (2011) Stomatal control as a driver of plant
evolution. J Exp Bot 62(8):2419–2423. doi:10.1093/jxb/err086
McAdam SAM, Brodribb TJ (2012) Stomatal innovation and the rise of seed plants. Ecol Lett
15(1):1–8. doi:10.1111/j.1461-0248.2011.01700.x
Meinzer FC, Johnson DM, Lachenbruch B, McCulloh KA, Woodruff DR (2009) Xylem hydraulic
safety margins in woody plants: coordination of stomatal control of xylem tension with
hydraulic capacitance. Funct Ecol 23(5):922–930. doi:10.1111/j.1365-2435.2009.01577.x
Niklas KJ, Tiffney BH, Knoll AH (1983) Patterns in vascular land plant diversification. Nature
303(5918):614–616
Nisbet EG, Nisbet RER (2008) Methane, oxygen, photosynthesis, rubisco and the regulation of the
air through time. Philos Trans R Soc B 363(1504):2745–2754
Osborne CP, Beerling DJ (2006) Nature’s green revolution: the remarkable evolutionary rise of C4
plants. Philos Trans R Soc B 361(1465):173–194
Raven JA (2014) Speedy small stomata? J Exp Bot 65(6):1415–1424. doi:10.1093/jxb/eru032
Robinson JM (1994) Speculations on carbon dioxide starvation, late tertiary evolution of stomatal
regulation and floristic modernization. Plant Cell Environ 17(4):345–354
Roth-Nebelsick A, Uhl D, Mosbrugger V, Kerp H (2001) Evolution and function of leaf venation
architecture: a review. Ann Bot 87(5):553–566
12 M. Haworth et al.

Sage RF, Sage TL, Kocacinar F (2012) Photorespiration and the evolution of C4 photosynthesis.
Annu Rev Plant Biol 63:19–47
Sperry JS, Hacke UG, Pittermann J (2006) Size and function in conifer tracheids and angiosperm
vessels. Am J Bot 93(10):1490–1500. doi:10.3732/ajb.93.10.1490
Taylor SH, Ripley BS, Martin T, De-Wet LA, Woodward FI, Osborne CP (2014) Physiological
advantages of C4 grasses in the field: a comparative experiment demonstrating the importance
of drought. Glob Chang Biol 20(6):1992–2003
Tcherkez GG, Farquhar GD, Andrews TJ (2006) Despite slow catalysis and confused substrate
specificity, all ribulose bisphosphate carboxylases may be nearly perfectly optimized. Proc Natl
Acad Sci 103(19):7246–7251
Tolbert NE, Benker C, Beck E (1995) The oxygen and carbon dioxide compensation points
of C3 plants: possible role in regulating atmospheric oxygen. Proc Natl Acad Sci USA
92:11230–11233