Journal of Theoretical Biology 222 (2003) 505–515

Factors affecting binary sex evolution with respect to avoidance of

vertical transmission of deleterious intracellular parasites
Atsushi Yamauchi
Center for Ecological Research, Kyoto University, Kamitanamami Hirano, Otsu 520-2113, Japan
Received 21 October 2002; received in revised form 2 January 2003; accepted 15 January 2003

Abstract

In sexual reproductive systems, the number of sexes is generally binary, viz. male and female. Several theoretical studies have
shown that the evolution of this system is possibly related to cytoplasmic DNA, including deleterious cytoplasmic symbionts. When
organisms are infected by a symbiont that is transmitted vertically to offspring via gametes, the exclusion or degeneration of the
latter may evolve as a characteristic of those organisms. If this necessarily results in the elimination of organelle DNA in gametes, a
reciprocal preference between individuals, one transmitting organelles and the other not, may be favored. In this theoretical study,
factors affecting such an evolutionary process, in which the symbiont is considered as a parasite infecting vertically, horizontally and
naturally, are considered. In addition, host individuals are assumed to recover from the infection to some degree. According to the
analysis, a binary sex system can evolve only when uninfected and infected host individuals co-exist in a single host population. This
condition can be satisfied only if natural infection occurs. Although recovery from infection has both positive and negative effects on
binary sex evolution, the latter is promoted only when natural infection exists. Accordingly, if natural infection does not exist, the
evolution of binary sex system is unlikely with respect to deleterious cytoplasmic symbionts, in absent of heterozogotic advantage in
vertical transmission.
r 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Binary sex evolution; Cytoplasmic DNA; Vertical transmission; Theory

1. Introduction titor DNA inherited from a mating partner. If the

Sexual reproductive systems are adopted by many
organisms belonging to a wide range of taxa, the
number of sexes generally being binary, viz. male and
female. The evolution of the binary sex system includes
two different issues: (1) evolutionary process from the
absence of mating type to binary sex (including
evolution of anisogamy from isogamy), (2) evolution
of two sexes from more mating types. The former has
been considered by many authors, in which two
hypotheses were proposed with respect to selfish
cytoplasmic elements. One hypothesis focused on
intracelluler conflict of cytoplasmic genes (Hoekstra,
1987; Hurst, 1990; 1994–1996; Hurst and Hamilton,
1992; Frank, 1996; reviewed by Partridge and Hurst,
1998; Randerson and Hurst, 2001). When cytoplasmic
DNA is biparentally transmitted to offspring, the latter
may evolve competition ability which destroys compe-
E-mail address: a-yama@ecology.kyoto-u.ac.jp (A. Yamauchi).
competition between cytoplasmic DNA types involves
injuries to nuclear DNA, suppression of cytoplasmic
competition may evolve from the latter. Subsequently, a
reciprocal preference between individuals, one transmit-
ting organelles and the other not, may evolve, implying
a binary sex system. Hurst and Hamilton (1992)
demonstrated the plausibility of this evolutionary
scenario using a mathematical model.
The second hypothesis considered that the evolution
of binary sex is related to intracellular parasites, the
basic idea of this being verbally proposed by Hurst
(1990). When organisms are infected by parasites that
can transmit vertically to offspring via gametes, the
exclusion or degeneration of the latter may evolve as a
characteristic of the former organisms. If this necessarily
results in the elimination of organelle DNA as a side-
effect, mating among individuals possessing such an
ability is disadvantageous, because of the lethal effect on
offspring caused by the absence of organelles. In
addition to this, mating among individuals transmitting

0022-5193/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0022-5193(03)00065-1
506 A. Yamauchi / Journal of Theoretical Biology 222 (2003) 505–515
organelles may also be disadvantageous due to the high
infection probability for offspring. Consequently, a
reciprocal preference between individuals, one transmit-
ting organelles and the other not, may be favored.
Hastings (1992) and Law and Hutson (1992) focused on
the role of intracellular parasites (deleterious symbionts)
in evolution of anisogamy, although those studies did
not emphasize their possible role in the evolution of
mating types.
Hutson and Law (1993) analysed fully the evolution
of binary sex from absence of mating type with respect
to intracellular parasites. In their model, a deleterious
symbiont initially increased and became fixed in the
population. Subsequently, a gene modifying cytoplasmic
inheritance invades, followed by an invasion of self-
incompatibility gene. According to their analysis, in the
initial stage, the deleterious symbiont was ultimately
either included by all members or excluded from the
population. In the former case, even if rare mutants
arise which do not transmit cytoplasm to offspring, their
offspring tends to involve symbiont due to their mating
partner being dominant wildtype. The mutants can be
advantageous only if heterozygotic advantage is intro-
duced to the inheritance of deleterious symbionts; the
fitness of an individual inheriting symbionts from two
parents being lower than that inherited from a single
parent. Such a heterozogotic advantage is necessary for
the successful invasion of an inheritance modifier. When
the heterozygotic advantage is small, evolutions of
uniparental cytoplasmic inheritance and sex are un-
likely.
Randerson and Hurst (1999) proposed a mechanism
by which nuclear modifier eliminating cytoplasm can
evolve even in the population that is occupied by
deleterious cytoplasmic elements. According to their
analysis, the modifier that kills own cytoplasm at
production of gametes cannot invade such a population,
although the modifier that kills partner’s cytoplasm in a
zygote after conjugation can invade and increase in the
population. Their analysis indicated one possible evolu-
tionary process of modifier eliminating cytoplasm.
However, the modifier eliminating cytoplasm from
own gametes may be plausible in anisogamous organ-
isms and conjugation of protists, in which individuals
seem to determine inheritance of cytoplasm by them-
selves. Accordingly, it is worthwhile to consider other
factors promoting evolutions of ‘‘killing own cyto-
plasm’’ modifier and subsequent binary sex system.
If we consider parasites as selfish cytoplasmic
elements, there are various characteristics that possibly
affect binary sex evolution. Some may be able to infect
horizontally or naturally, in addition to vertical
transmission. Other parasites are possibly lost from
the host lineage following host recovery or parasite
vertical transmission failure. Such factors could affect
the evolutionary conditions of ‘‘killing own cytoplasm’’
modifier and a binary sex system. In this paper, the roles
of such factors in the evolution of sex are examined
theoretically, with respect to the avoidance of intra-
cellular parasites.
2. Mathematical model
2.1. Initial conditions
Organisms are considered to reproduce sexually by
random pairing, thereby inheriting cytoplasm biparen-
tally. A parasite considered as infecting this species has
three infection means, vertical, horizontal and natural.
The frequencies of uninfected and infected individuals in
the population at generation t are represented respec-
tively by St and It. Alternation of generations is
considered to include three stages, the first being a
mating and death stage, in which vertical transmission
of parasites and fitness reduction of infected individuals
are included. Regarding the host, if a single or both
parents are infected by an intracellular parasite, their
offspring must also be infected (vertical transmission).
At the immature offspring stage, the parasite reduces
individual survivorship, the survival rate being repre-
sented by a, relative to that of an uninfected individual
(= 1). Considering the mating consequences among St
and It, the genotypic frequencies are modified as
St0 ¼ St2 =W ; ð1aÞ
It0 ¼ aIt ð2St þ It Þ=W ; ð1bÞ
where W is the average fitness, and St0 and It0 ;
respectively, represent genotypic frequencies immedi-
ately after mating. The second stage is an infection stage
in which both horizontal and natural infections occur.
In this stage, uninfected individuals are assumed to be
horizontally and naturally infected by a parasite with
probabilities n and m, respectively. The frequencies
change as
St00 ¼ ð1  m  nIt0 ÞSt0 ; ð2aÞ
It00 ¼ It0 þ mSt0 þ nSt0 It0 : ð2bÞ
where St00 and It00 represent genotypic frequencies
immediately after natural and horizontal infections,
respectively. In the third stage, an infected individual
recovers (no longer infected) with probability r, by
which the frequencies change as
Stþ1 ¼ St00 þ rIt00 ; ð3aÞ
Itþ1 ¼ ð1  rÞIt00 : ð3bÞ
In this formulation, r can also be regarded as the
probability of a gamete not inheriting a parasite, i.e.
incomplete vertical transmission. These are the basic
host–parasite dynamics of the initial population.
 A. Yamauchi / Journal of Theoretical Biology 222 (2003) 505–515 507

2.2. Introduction of a gene suppressing the inheritance of 00

I0;t 0
¼ I0;t 0
þ mS0;t 0
þ nS0;t 0
ðI0;t 0
þ I1;t Þ; ð5bÞ
cytoplasmic DNA n o
00 0 0 0
S1;t ¼ 1  m  nðI0;t þ I1;t Þ S1;t ; ð5cÞ
In the host population, a mutant gene that suppresses
the inheritance of cytoplasmic DNA (including plastids, 00
I1;t 0
¼ I1;t 0
þ mS1;t 0
þ nS1;t 0
ðI0;t 0
þ I1;t Þ; ð5dÞ
mitochondria and parasites) is considered to invade the
equilibrium state of the above initial condition. The where double primes on S and I represent genotypic
wildtype allele is represented by A, the mutant allele frequencies immediately after natural and horizontal
being A: * The latter is assumed to be dominant, implying infections,. In the recovery stage, the frequencies
that A A* individuals do not transmit cytoplasmic DNA change as
to offspring. Consequently, offspring produced from 00 00
S0;tþ1 ¼ S0;t þ rI0;t ; ð6aÞ
mating among A A* individuals do not survive because
00
they inherit neither plastids nor mitochondria from their I0;tþ1 ¼ ð1  rÞI0;t ; ð6bÞ
parents. This also implies that A*A* individuals cannot 00 00
exist in the population. For A A* individuals, it is S1;tþ1 ¼ S1:t þ rI1;t ; ð6cÞ
assumed that the suppression of cytoplasm inheritance 00
I1;tþ1 ¼ ð1  rÞI1;t : ð6dÞ
is accompanied by reduction of fitness, c. In this case,
four types of individuals exist in the population, (i)
uninfected AA individuals, (ii) infected AA individuals, 2.3. Introduction of mating preferences between
(iii) uninfected A A* individuals and (iv) infected A A* AA and A A*
individuals. The frequencies of these individuals are
represented by S0, I0, S1, and I1, respectively. In the In the situation considered here, A A* individuals
mating stage, considering the mating consequences should prefer mating with AA so as to avoid having
summarized in Table 1, the genotypic frequencies are offspring that lack plastid and mitochondrial DNA. On
modified as the other hand, AA individuals may prefer mating with
0

 * since gametes produced by the latter are not
A A;
S0;t ¼ S0;t S0;t þ ðS1;t þ I1;t Þ =W ; ð4aÞ
infected by the parasite. Therefore, evolution of the
0
I0;t


¼ aI0;t ð2S0;t þ I0;t Þ þ ðS1;t þ I1;t Þ =W ; ð4bÞ mating preferences of AA and A A* is considered. Two
quantitative genes p and q are assumed, each condition-
0
S1;t ¼ ð1  cÞS0;t ðS1;t þ I1;t Þ=W ; ð4cÞ ally determining the acceptance probabilities of the
different genotype from itself, depending upon its own
0 cytoplasmic inheritance genotype, AA and A A; * respec-
I1;t ¼ ð1  cÞaI0;t ðS1;t þ I1;t Þ=W ; ð4dÞ
tively. It is known that in many organisms, mating
where W is the average fitness, and single primes on S compatibility (including sex and mating types) is
and I indicate genotypic frequency immediately after determined by few major genes, although such genes
mating. In the subsequent infection stage, the frequen- work as a trigger in the determination of compatibility.
cies change as Their substantial preferences for mating partners are
n o controlled by a lot of genes, including those concerning
00 0 0 0
S0;t ¼ 1  m  nðI0;t þ I1;t Þ S0;t ; ð5aÞ
signal presentations and recognitions. Functions and
efficiencies of those genes can be considered to evolve
gradually. Accordingly, mating preferences are assumed
Table 1 to be qualitative traits in the presented analysis.
Offspring genotypes and infections resulting from various mating Based on a consideration of the proposal by Hurst
combinations
and Hamilton (1992), a two-stage mating model is also
S0 I0 S1 I1 considered here, such partly modifying the original
(uninfected) (infected) (uninfected) (infected) formulation. In the first stage of mating, random pairing
S0 S0 I0 S0 and S1 S0 and S1 occurs among all members of the population. Pairing
(uninfected) (uninfected) (infected) (uninfected) (uninfected) between AA and A A* individuals necessarily proceeds
until mating, although pairing among each of the AA
I0 I0 I0 I0 and I1 I0 and I1
and A A* strains are cancelled out according to certain
(infected) (infected) (infected) (infected) (infected)
probabilities. The pairing of two AA individuals with p1
S1 S0 and S1 I0 and I1 — — and p2 phenotypic values is completed only when both
(uninfected) (uninfected) (infected) (lethal) (lethal) individuals accept each other as partner, the probability
of such being p1p2. On the other hand, the pairing of A A*
I1 S0 and S1 I0 and I1 — —
individuals with q1 and q2 is completed with probability
q1q2. Namely, pairing among both AA and A A* strains
(infected) (uninfected) (infected) (lethal) (lethal)
508 A. Yamauchi / Journal of Theoretical Biology 222 (2003) 505–515

are cancelled out with probability 1  p1 p2 and 1  q1 q2 ; 00

I0;t 0
¼ I0;t 0
þ mS0;t 0
þ nS0;t 0
ðI0;t 0
þ I1;t Þ; ð8bÞ
respectively. In the second stage of mating, re-pairing
n o
occurs randomly among individuals cancelling out the 00
S1;t ¼ 1  m  nðI0;t 0
þ I1;t0 0
Þ S1;t ; ð8cÞ
first stage pairing. Such re-pairing is never cancelled out.
In this two-stage mating system, costs may exist between 00
I1;t 0
¼ I1;t 0
þ mS1;t 0
þ nS1;t 0
ðI0;t 0
þ I1;t Þ; ð8dÞ
the cancellation of the first pairing and re-pairing
processes. A fitness reduction resulting from those where double primse on S and I represent genotypic
processes is represented by coefficient k (o1), with frequency immediately after natural and horizontal
which the fitnesses of re-pairing pairs are weighted. infections. In the recovery stage, the frequencies
Accordingly, the presented model includes five vari- change as
ables. Three variables (S0, I0 and S1) represent the 00 00
S0;tþ1 ¼ S0;t þ rI0;t ; ð9aÞ
frequencies of (i) uninfected AA individuals, (ii) infected
AA individuals, (iii) uninfected A A* individuals, respec- 00
I0;tþ1 ¼ ð1  rÞI0;t ; ð9bÞ
tively. Frequency of infected A A* individuals can be
00 00
represented by I1 ¼ 1  ðS0 þ I0 þ S1 Þ: Two other vari- S1;tþ1 ¼ S1;t þ rI1;t ; ð9cÞ
ables (p% and q) % are the average phenotypic values of (v) 00
I1;tþ1 ¼ ð1  rÞI1;t : ð9dÞ
preference of AA individuals and (vi) preference of A A*
individuals, respectively. In the mating stage, consider- On the other hand, evolution of the average
ing the mating consequences summarized in Table 1, the phenotypic values of p% and q% can be analysed by the
genotypic frequencies are modified as quantitative genetic model of Iwasa et al. (1991). In that
0

 model, the change in average phenotypic value in a
S0;t ¼ S0;t p% 2 S0;t þ ðS1;t þ I1;t Þ

 population between generations is proportional to the
þ ks0;t s0;t þ ðs1;t þ i1;t Þ = selection gradient around that value in the generation

ðs0;t þ i0;t þ s1;t þ i1;t Þ =W ; ð7aÞ concerned. Considering an individual with phenotypic

 values p and q in a population with average phenotypic
0
I0;t ¼ a I0;t p% 2 ð2S0;t þ I0;t Þ þ ðS1;t þ I1;t Þ values p% and q;
% its fitness is represented by

 

þ ki0;t ðs0;t þ 2i0;t Þ þ ðs1;t þ i1;t Þ = % 0;t þ aI0;t Þ þ S1;t þ I1;t
Fðp; qÞ ¼ S0;t ppðS
 
ðs0;t þ i0;t þ s1;t þ i1;t Þ =W ; ð7bÞ % 0;t þ I0;t Þðs0;t þ ai0;t þ s1;t þ i1;t Þ
þkð1  ppÞðS


0
 % 0;t þ I0;t Þa þ ðS1;t þ I1;t Þa
þ I0;t ppðS
S1;t ¼ ð1  cÞ S0;t ðS1;t þ I1;t Þ 
 % 0;t þ I0;t Þa
þkð1  ppÞðS
þ ks0;t ðs1;t þ i1;t Þ=ðs0;t þ i0;t þ s1;t þ i1;t Þ =W ; ð7cÞ 
þ ðS1;t þ I1;t Þ ðS0;t þ aI0;t Þ
0
 
I1;t ¼ ð1  cÞa I0;t ðS1;t þ I1;t Þ % 1;t þ I1;t Þðs0;t þ ai0;t Þ ;
þ kð1  qqÞðS ð10Þ

þ ki0;t ðs1;t þ i1;t Þ=ðs0;t þ i0;t þ s1;t þ i1;t Þ =W ; ð7dÞ which indicates the relative number of gene copies that
where W is the average fitness, and single primes on S survive until the next generation. The focused gene is
and I indicate genotypic frequency immediately after possibly involved in various genotypic individuals; (i)
mating. In this formulation, s0,t, i0,t, s1,t and i1,t represent uninfected AA individuals, (ii) infected AA individuals,
the genotypic frequencies of individuals cancelling out (iii) uninfected A A* individuals or (iv) infected A A*
the first stage pairing: individuals. In Eq. (10), the fitnesses in such cases are
summed with the probabilities of the individual belong-
s0;t ¼ ð1  p% 2 ÞS0;t ðS0;t þ I0;t Þ; ð7eÞ ing to the category. Each fitness includes fitness gains
from both the first stage paring and the following re-
i0;t ¼ ð1  p% 2 ÞI0;t ðS0;t þ I0;t Þ; ð7fÞ
pairing. From Eq. (10), selection gradients of p and q
s1;t ¼ ð1  q% 2 ÞS1;t ðS1;t þ I1;t Þ; ð7gÞ values are

@  1 @Fðp; qÞ 1
logFðp; qÞ ¼ ¼
i1;t ¼ ð1  q% 2 ÞI1;t ðS1;t þ I1;t Þ; ð7hÞ @p %
p¼p;q¼q% % qÞ
Fðp; % @p % qÞ
Fðp; %


where p% and q% are average phenotypic values of the  pS % 0;t ðS0;t þ aI0;t Þ  kðS0;t þ I0;t Þ

preferences in the population. Notably, in Eqs. (7a–7d), ðs0;t þ ai0;t þ s1;t þ i0;t Þ
successes gained from re-pairing are adjusted by a 
þpI% 0;t ðS0;t þ I0;t Þð1  kÞa ; ð11aÞ
coefficient term 1=ðs0;t þ i0;t þ s1;t þ i1;t Þ (see the appen-

dix). In the subsequent infection stage, the frequencies @  1 @Fðp; qÞ 1
change as logFðp; qÞ ¼ ¼
n o @q %
p¼p;q¼q% Fð %
p; %
qÞ @q Fð % qÞ
p; % ð11bÞ
00 0 0 0
S0;t ¼ 1  m  nðI0;t þ I1;t Þ S0;t ; ð8aÞ 2
% 1;t þ I1;t Þ kðs0;t þ ai0;t Þ:
ðqÞðS
 A. Yamauchi / Journal of Theoretical Biology 222 (2003) 505–515 509

According to the quantitative genetic model of Iwasa infection nor recovery (n=0, r=0), in which both p% and
et al. (1991), dynamics of p% and q% can be represented by q% converge to 0 in simulations, implying binary sex
0 1 0  1 evolution. According to the figure, when condition (13)
@ 
d p% ! logF is not satisfied, binary sex never evolves. In such a case,
B dt C Gp Bpq B B @p %
p¼p;q¼
C
q% C
B C¼ B  C; ð12Þ since the population comprises infected individuals only,
@ d q% A Bpq Gq @ @  A
logF all offspring are always infected vertically. Accordingly,
dt @q %
p¼p;q¼ q% A A* individuals with suppressed cytoplasmic inheritance
have no advantage, resulting in failure of invasion of the
where Gp and Gq are additive genetic variances of p and
A* allele. On the other hand, if condition (13) holds, by
q, respectively. On the other hand, Bpq is a genetic
which uninfected and infected individuals co-exist,
covariance between p and q. For a convenience of
binary sex can evolve conditionally depending on the
analysis, the additive genetic variances are set to 2, the
fitness coefficient of re-pairing, k. According to the
genetic covariance being 0 (i.e. ignored), in the following
figure, a critical fitness coefficient of re-pairing (kc)
simulations. Covariances between major genes control-
* and these exists, above which a binary sex system can evolve.
ling cytoplasmic inheritance (A and A)
Within the region in which binary sex can evolve, the
quantitative traits (p and q) are also considered
critical fitness coefficient decreases with increasing
negligible, because that these quantitative traits were
natural infection rate m. In other words, when the
assumed approximately uniform in the population based
natural infection probability is high, binary sex tends to
on the model of Iwasa et al. (1991).
evolve even if re-pairing is costly. Furthermore, binary
Based on these equations, five variables (S0, I0, S1, p%
sex tends to evolve when the parasite infection notably
and q) % change dynamically from generation to genera-
reduces individual fitness (small a).
tion, depending on five parameters; (i) the survival rate
If horizontal infection is introduced ðna0Þ; equili-
of infected immature individuals, a, (ii) the natural
brium of the dynamics Eqs. (1)–(3) cannot be derived
infection rate, m, (iii) the horizontal infection rate, n, (iv)
explicitly, the behavior of the system being analysed
the recovery rate, r and (v) the fitness coefficient of re-
only by computer simulations. Fig. 1(b and c) illustrates
pairing, k. In this model, the most important variables
a parameter region in which both p% and q% converge to 1,
are p% and q;
% which determine the acceptance between the
with r=0 and (b) n=0.2 and (c) n=0.5. These indicate
two mating types. If both p% and q% evolve simultaneously
that as the horizontal infection rate increases, evolution
from 1 to 0, this indicates that mating between an
of reciprocal preference between AA and A A* is
individual producing gametes with organelles and
increasingly less likely under c=0 and r=0. The high
another which does not, is favored, thus implying
horizontal infection rate facilitates parasite spread, by
evolution of binary sex. Therefore, changes in p% and q%
which all members of the host population tend to be
values here are followed from an initial condition
infected. Consequently, invasion of the A* allele (that
% qÞ
ðp; % ¼ ð1; 1Þ; using a computer simulation for varying
suppresses cytoplasmic inheritance) fails, preventing the
parameter sets (a, m, n, r, k).
evolution of binary sex. Furthermore, recovery is also
introduced ðra0Þ; which possibly implies incomplete
vertical transmission of the parasite. Fig. 2 illustrates a
Results parameter region in which both p% and q% converge to 1,
with r=0.2 and (a) n=0, (b) n=0.2 and (c) is n=0.5.
Because the dynamics presented are too complicated Comparing Figs.1 and 2, when both a and m are
for deriving general analytic results, a special case relatively large, the recovery (or incompleteness of
neglecting horizontal infection and recovery is first vertical infection) facilitates the evolution of a binary
analysed, which is a helpful simplification for deriving sex system, owing to the expansion of a parameter
some results. If both horizontal infection and recovery region in which uninfected and infected host individuals
are absent (n=0, r=0), the basic dynamics Eqs. (1)–(3) co-exist, promoting an invasion of the A* allele.
have a stable co-existing equilibrium involving both Next, dynamics including the cost of possession of the
uninfected and infected individuals when nuclear modifier ðca0Þ are analyzed. Figs. 3 and 4
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ao 1  mð2  mÞ =2: ð13Þ illustrate parameter regions in which both p% and q%
converge to 0 under c=0.5, with recovery rates r=0 and
Otherwise, the population will eventually comprise 0.2, respectively. In these figures, horizontal infection
infected individuals only. Further parameter depen- rates are (a) n=0, (b) n=0.2 and (c) n=0.5. Comparing
dences are numerically analysed by using computer Figs. 3 and 4 to Figs. 1 and 2, respectively, it is shown
simulations. that an existence of cost of nuclear modifier suppresses
First, dynamics without a cost of possession of the evolution of reciprocal preference between AA and
nuclear modifier (c=0) are focused on. Fig. 1(a) A A:* This suppression results from a failure of invasion
illustrates a parameter region with neither horizontal of the nuclear modifier. Nevertheless, it is notable that
510 A. Yamauchi / Journal of Theoretical Biology 222 (2003) 505–515

Fig. 1. The parameter region in which reciprocal preference between AA and AA* individuals evolves (i.e. p% ¼ 0 and q% ¼ 0) without cost of nuclear
modifier (c=0) and recovery from infection (r=0). Horizontal transmission rate (n) is (a) 0, (b) 0,2 and (c) 0.5. The additive genetic variances of p and
q are set to 2 (Gp=Gq=2), the genetic covariances between p and q being 0 (Bpq=0).

even if the cost of modifier is relatively high (c=0.5),
reciprocal preference between AA and A A* can evolve in
regions of large m and small a, implying evolution of the
binary sex system.
Discussion
The present study analysed an evolutionary process
from absence of mating type to binary sex system, from
a viewpoint of evolution of ‘‘eliminating own cyto-
plasm’’ modifier. It showed that evolution of a binary
sex system is possibly promoted by an intracellular,
vertically, horizontally and naturally infecting parasite,
without heterozogotic advantage in the inheritance of
deleterious symbionts, as proposed by Hutson and Law
(1993). According to the analysis, when both uninfected
and infected host individuals co-exist, with biparental
cytoplasmic inheritance, an allele that suppresses
cytoplasmic inheritance can spread if the cost of re-
pairing is below a critical value (kc). Subsequently,
reciprocal preference evolves between individuals trans-
mitting and non-transmitting cytoplasm to offspring. In
this evolutionary process, horizontal infection tends to
suppress the evolution of binary sex because it facilitates
parasitic infection of all individuals in the host popula-
tion.
Randerson and Hurst (1999) analysed invasion
conditions of nuclear modifiers that kill either own or
partner’s cytoplasm at reproduction. Their model of
‘‘nuclear modifier killing own cytoplasm in the uni-
cellular case’’ analysed a consistent problem with the
presented study. According to their analysis, if all
members of the population possess selfish cytoplasm, a
 A. Yamauchi / Journal of Theoretical Biology 222 (2003) 505–515 511

Fig. 2. The parameter region in which reciprocal preference between AA and AA* individuals evolves (i.e. p% ¼ 0 and q% ¼ 0) with no cost of nuclear
modifier (c=0) and recovery rate 0.2 (r=0.2). Horizontal transmission rate (n) is (a) 0, (b) 0,2 and (c) 0.5. The additive genetic variances of p and q
are set to 2 (Gp=Gq=2), the genetic covariances between p and q being 0 (Bpq=0).

nuclear modifier killing own cytoplasm cannot invade
into the population. On the other hand, when selfish
cytoplasm co-exists with the non-selfish one, such a
modifier invades and increases in the population,
depending upon costs of possessing selfish cytoplasm
and nuclear modifier. Their analysis indicated an
importance of frequency of selfish cytoplasm in the
evolution of nuclear modifiers killing own cytoplasm.
Nevertheless, in their model, when selfish cytoplasm can
invade, it ultimately occupies the population. The
system was unlikely to result in a coexistence of selfish
and non-selfish cytoplasm, although they did not focus
on factors that resulted in the coexistence. The presented
analysis indicates such factors, considering also the
following evolutionary process of mating type.
Randerson and Hurst (1999) also pointed out that the
nuclear modifier killing own cytoplasm can invade a
population only when the cost of possession of the
modifier was very low, suggesting a difficulty of
evolution of the nuclear modifier. Nevertheless, this
may be possible within a parameter region that they
focused on. They focused on the low cost of selfish
cytoplasm because a high cost prevents both invasion of
the cytoplasm and subsequent invasion of the modifier
in their system. Contrary to this, the presented analysis
shows that a natural infection possibly results in the
coexistence of selfish and non-selfish cytoplasm even if
the cost of the former is high (small a). In such cases, if
the nuclear modifier killing own cytoplasm is accom-
panied by relatively higher cost (large c), it can invade in
512 A. Yamauchi / Journal of Theoretical Biology 222 (2003) 505–515

Fig. 3. The parameter region in which reciprocal preference between AA and A A* individuals evolves (i.e. p% ¼ 0 and q% ¼ 0) with cost of nuclear
modifier 0.5 (c=0.5) and no recovery from infection (r=0). Horizontal transmission rate (n) is (a) 0, (b) 0,2 and (c) 0.5. The additive genetic variances
of p and q are set to 2 (Gp=Gq=2), the genetic covariances between p and q being 0 (Bpq=0).

the population, resulting in the binary sex evolution (see
Figs. 3 and 4).
The analysis presented here indicates that when
natural infection is absent, a binary sex system cannot
evolve under any parameter sets. In such a case, if the
relative fitness of infected individuals is low, the parasite
is completely excluded from the host population.
Otherwise, all host members are infected by the parasite.
Whatever the case, rare mutants that suppress cytoplas-
mic inheritance have no advantage, resulting in the
failure of both anisogamy and binary sex to evolve. Only
when natural infection exists to some degree, can co-
existence of uninfected and infected individuals occur,
resulting in the evolution of binary sex. Nevertheless, a
high natural infection rate also prevents the evolution of
binary sex, because it results in all host individuals being
infected. Consequently, natural infection promotes
binary sex evolution only if its occurrence probability
is at an intermediate level.
Binary sex is likely to evolve when the fitness of
infected individuals is low (small a), which may have
resulted for two reasons. The first reason is illustrated in
Fig. 1(a–c), in which a critical a value as a function of m
exists above which binary sex never evolved with any
fitness coefficient of re-pairing, (k). This indicates that
the low fitness of infected individuals tends to result in
the co-existence of uninfected and infected individuals
by which evolution of binary sex is promoted. The
second effect is illustrated in Fig. 2(a–c), which show
that the critical fitness coefficient of re-pairing becomes
lower (small kc) with decreasing fitness of infected
individuals (small a). This implies that when the parasite
causes serious damage to infected host individuals, a
binary sex system can evolve effectively so as to
 A. Yamauchi / Journal of Theoretical Biology 222 (2003) 505–515 513

Fig. 4. The parameter region in which reciprocal preference between AA and A A* individuals evolves (i.e. p% ¼ 0 and q% ¼ 0) with cost of nuclear
modifier 0.5 (c=0.5) and recovery rate 0.2 (r=0.2). Horizontal transmission rate (n) is (a) 0, (b) 0,2 and (c) 0.5. The additive genetic variances of p
and q are set to 2 (Gp=Gq=2), the genetic covariances between p and q being 0 (Bpq=0).

avoid such disadvantages, despite the higher cost of re-
paring. According to these two processes, binary sex
evolution is promoted by the lower fitness of infected
individuals.
According to the presented analysis, horizontal
infection rate (n) and recovery rate (r) can affect both
positively and negatively the evolution of binary sex,
depending on conditions. If recovery does not exist, an
increase of horizontal infection reduces the parameter
region in which binary sex can evolve (negative effect,
see Figs.1 and 3). Otherwise, it expands the latter
(positive effect, see Fig. 4). If the fitness of infected
individuals is relatively high (large a), recovery from
infection (or incompleteness of vertical infection)
expands the parameter region in which binary sex can
evolve (positive effect, compare large a regions of
Figs. 1–4). Otherwise, recovery increases the critical
value of the fitness coefficient of re-pairing (kc) above
which the binary sex system can evolve, suppressing the
binary sex evolution (negative effect, compare small a
regions of Figs. 1–4). These result from the fact that
both horizontal infection rate (n) and recovery rate (r)
affect the dynamics through two different ways invol-
ving opposite effects. The increment of horizontal
infection and the decrement of recovery tend to result
in the infection of all members in the population,
suppressing evolution of the nuclear modifier and
following mate preference. However, such enhance-
ments of the danger of parasite result in the promotion
of the evolution of the nuclear modifier and preference
as an avoidance mechanism against infection. Accord-
ingly, horizontal infection and recovery can work both
positively and negatively for binary sex evolution,
depending on conditions.
514 A. Yamauchi / Journal of Theoretical Biology 222 (2003) 505–515
The presented analysis shows that a horizontal
infection variously affects the binary sex evolution (see
above). On the other hand, Frank (1996) pointed out
that horizontal transmission can evolve in parasites if its
cost is smaller than that of vertical transmission, by
which symbionts are likely to be monomorphic. He also
considered that such a monomorphism of symbiont
reduced the competitions between symbionts, resulting
in a reduction of parasite virulence. According to the
present analysis, the decrement of virulence (larger a)
tends to suppress the binary sex evolution due to
infected hosts dominating in the population (see Fig. 1).
Consequently, the horizontal infection may have stron-
ger negative effect on binary sex evolution than a
prediction of the present model. This suggests that
effects of characteristics of parasites on binary sex
evolution should be considered simultaneously from two
viewpoints, evolution of parasite virulence (competiton)
and dynamic properties of host population.
The model indicates various factors that can influence
the evolutionary process of a binary sex system with
respect to the avoidance of deleterious intracellular
parasites. Nevertheless, the analysis may have under-
estimated the possibility of binary sex evolution. In the
model, the genetic correlation between p and q is ignored
because of the quantitative genetic model used not
considering any changes in genetic correlation (Iwasa
et al., 1991). Since an AA individual preferring A A* tends
to mate with the latter which has a reciprocal preference
for the former, the genetic correlation between p and q
must increase gradually. If the model includes such a
factor, binary sex may evolve even when the fitness
coefficient of re-pairing is lower (smaller k). Furthermore,
in nature, every organism is exposed to many parasites. If
only one of those parasitic species satisfies the conditions
for binary sex evolution as shown in this study, can the
host species evolve a binary sex system. Once the system
is established, the spread of any new intracellular parasite
may be effectively suppressed, despite their not directly
concerning the evolution of a binary sex system (see
also Law and Hutson, 1992). Therefore, it should be
recognized that binary sex evolution may proceed more
easily than indicated by the consideration presented here.
The study presented here analysed an evolutionary
process from absence of mating type to binary sex
system. However, our alternative process should also be
considered in the analysis of binary sex evolution, i.e.
evolutionary process from many mating types to binary
sex. Iwasa and Sasaki (1987) showed that the multiple
mating type possibly converge to binary mating type,
focusing on mating kinetics. On the other hand, Hurst
(1996) analysed this issue from the viewpoint of the
evolution of selfish cytoplasmic element. Such an
evolutionary process may be affected by infection
patterns of intracellular parasites. Therefore, it should
be analyzed in the future studies.
Acknowledgements
I thank Drs. Y. Natsukari, T. Sunahara and N.
Yamamura, and members of the Faculty of Fisheries,
Nagasaki University, and Center for Ecological Re-
search, Kyoto University, for their encouragement. I
also thank two anonymous referees for their helpful
comments, by which the manuscript was significantly
improved. This study was partly supported by Grant-
in-Aid for Scientific Research (No.13640629) from the
Japanese Ministry of Education, Culture, Sports,
Science and Technology, and the MEXT Grant-in-Aid
for the 21st Century COE Program (Az to Kyoto
University).
Appendix A. Adjustment of success of re-pairing
For simplicity of consideration, re-pairing is assumed
to occur among AA and A A* individuals, of which
frequencies in the population are a1 and a2, respectively.
A fraction 1a1a2 of individuals is assumed to have
already mated. In order to consider the re-paring
process, their frequencies in the re-paring group are
defined as a01 ¼ a1 =ða1 þ a2 Þ and a02 ¼ a2 =ða1 þ a2 Þ:
Using these values, the total reproductive consequences
in the mating group can be represented by w1 a02 1 þ
2w2 a01 þ w3 a02
2 ; where wi indicates the fitness of various
types of offspring. Since mating is undergone among the
a1+a2 fraction of the total population, the real mating
success of re-pairing must be ða1 þ a2 Þðw1 a02 1 þ
2w2 a01 a02 þ w3 a02
2 Þ; which can be rewritten as ðw1 a1 þ
2
2
2w2 a1 a2 þ w3 a2 Þ=ða1 þ a2 Þ: This implies that if mating
occurs among some members of the population, the
mating consequence must include a coefficient for
adjustment, being the inverse of the total frequency
of members participating in mating ð1=ða1 þ a2 ÞÞ: In
Eq. (7), 1=ðs0;t þ i0;t þ s1;t þ i1;t Þ represents such an
adjustment coefficient.
References
Frank, S.A., 1996. Host-symbiont conflict over the mixing of
symbiotic lineages. Proc. Roy. Soc. London B 263, 339–344.
Hastings, I.M., 1992. Population genetic aspects of deleterious
cytoplasmic genomes and their effect on the evolution of sexual
reproduction. Genet. Res. 59, 215–225.
Hoekstra, R.F., 1987. The evolution of sexes. In: Stearns, S.C. (Ed.),
The Evolution of Sex and its Consequences. Birkhuser, . Basel,
pp. 59–91.
Hurst, L.D., 1990. Parasite diversity and the evolution of diploidy,
multicellularity and anisogamy. J theor Biol. 144, 429–443.
Hurst, L.D., 1994. Cytoplasmic genetics under inbreeding and
outbreeding. Proc. Roy. Soc. London B 258, 287–298.
Hurst, L.D., 1995. Selfish genetic elements and their role in evolution:
the evolution of sex and some of what that entails. Philos. trans.
Roy. Soc. London B 349, 321–332.
 A. Yamauchi / Journal of Theoretical Biology 222 (2003) 505–515
Hurst, L.D., 1996. Why are there only two sexes? Proc. Roy. Soc.
London B 263, 415–422.
Hurst, L.D., Hamilton, W.D., 1992. Cytoplasmic fusion and the
nature of sexes. Proc. Roy. Soc. London B 247, 189–194.
Hutson, V., Law, R., 1993. Four steps to two sexes. Proc. Roy. Soc.
London B 253, 43–51.
Iwasa, Y., Pomiankowski, A., Nee, S., 1991. The evolution of costly mate
preferences. II. The handicap principle. Evolution 45, 1431–1442.
Iwasa, Y., Sasaki, A., 1987. Evolution of the number of sexes.
Evolution 41, 49–65.
515
Law, R., Hutson, V., 1992. Intracellular symbionts and the evolution
of uniparental cytoplasmic inheritance. Proc. Roy. Soc. London B
248, 69–77.
Partridge, L., Hurst, L.D., 1998. Sex and conflict. Science 281,
2003–2008.
Randerson, J.P., Hurst, L.D., 1999. Small sperm, uniparental
inheritance and selfish cytoplasmic elements: a comparison of
two models. J. Evol. Biol. 12, 1110–1124.
Randerson, J.P., Hurst, L.D., 2001. The uncertain evolution of the
sexes. Trends Ecol. Evol. 16, 571–579.