American Journal of Botany 91(11): 1757–1766. 2004.

GENETICS AND MORPHOLOGY IN A BORRICHIA

FRUTESCENS AND B. ARBORESCENS (ASTERACEAE)
HYBRID ZONE1

MARIA V. CATTELL2 AND STEPHEN A. KARL3

Department of Biology, SCA 110, University of South Florida, Tampa, Florida 33620 USA

Interspecific plant hybridization is a common and evolutionarily important phenomenon. Here, the results of a study of hybridization
in the Florida Keys between two species of sea oxeye daisy, Borrichia frutescens and B. arborescens, are reported. Nuclear and
chloroplast genetic loci, log-likelihood assignment tests, and maximum likelihood estimates of genealogical class frequencies were
used to identify hybrid and parent genotypes, to investigate the utility of leaf and flower morphology for hybrid identification, and to
study symmetry and degree of introgression between the species. Genetic analyses confirmed the identity of the hybrid and parent
plants that were used for the morphological studies. Together, leaf and flower morphology can be used to identify hybrid and parental
types with moderate accuracy (4% error rate). Population genetic analyses indicate that, in spite of a significant level of hybridization,
pure B. frutescens and B. arborescens are persisting in the hybrid zone. Of the nonparentals, about 18% appear to be F1 hybrids, over
50% F2 hybrids, and the remainder backcrossed individuals but only with the B. frutescens parent. It is postulated that the hybrid zone
in the Florida Keys is being maintained by a combination of positive assortative mating and clonal reproduction.

Key words: Asteraceae; chloroplast DNA; cytonuclear disequilibrium; Florida, USA; hybrid zone; single-copy nuclear DNA.

Interspecific hybridization among plants is a common phe-
nomenon in nature, and hybrid zones are model systems for
the study of evolutionary processes (Barton and Hewitt, 1989;
Rieseberg and Ellstrand, 1993; Arnold et al., 1999). The evo-
lutionary consequences of hybridization, however, are mani-
fold and include increased genetic diversity of one or both
hybridizing parental species through introgression, novel ad-
aptations, breakdown, or reinforcement of reproductive isolat-
ing barriers, novel ecotypes or species, and reticulate specia-
tion from introgressive hybridization (Gallez and Gottlieb,
1982; Arnold et al., 1991; Wendel et al., 1991; Arnold, 1997).
Hybridization also can threaten the persistence of rare species
through gene swamping by a more common, reproductively
compatible species (Rieseberg and Morefield, 1995; Levin et
al., 1996; Rhymer and Simberloff, 1996). Ecological impli-
cations of hybridization include the influences of hybrids on
the organisms with which they interact, such as competing
plant species and herbivorous animals (Strauss, 1994; Fritz et
al., 1999).
For plants, hybridization is thought to play a prominent role
in the evolution of new taxa (Grant, 1981; Rieseberg, 1995,
1997). Much of the evolutionary significance of hybridization,
however, depends on the fate of hybrids, which in turn de-
pends on the fitness of hybrids and their fitness relative to the
parental species. Although direct estimates of fitness are dif-
ficult at best, the fate of hybrids often can, at least in part, be
determined empirically through the assessment of the degree
of introgression following hybridization. If the genes of one
species are introgressing into another, then the fitness of hy-
1
Manuscript received 5 January 2004; revision accepted 2 July 2004.
The authors thank M. Asmussen, S. Bell, B. Cochrane, E. McCoy, J. Nason,
A. Schnabel, P. Stiling, S. Ravdal, M. Roberts, A. Rossi, E. Severance, and
J. T. Streelman for helpful suggestions on experimental design, laboratory
protocols, field work, and data analysis. J. Nason kindly shared analysis pro-
grams. We also thank The Nature Conservancy for providing access to plants
on Little Torch Key. This research was supported in part by a National Science
Foundation Program in Systematics grant (DEB 98-06905) to SAK.
2
Present address: 5497 Gunbarrel Road, Longmont, Colorado 80503 USA.
3
E-mail: karl@mail.cas.usf.edu.
1757
brids must be high enough to allow the hybrids to survive to
maturity and backcross with parental individuals.
The initiation of hybridization between taxonomically valid
species results from a variety of forces but essentially involves
the mixing of gametes of reproductively compatible species.
In many instances, hybridization may be a highly transient
event and of no real evolutionary consequence. Alternatively,
repeated crossing between parental species or stabilization of
hybrid breeding systems can result in the formation of hybrid
swarms, hybrid zones, or new hybrid species (Grant, 1981;
Rieseberg, 1997). If the hybrid fitness is low relative to the
parental species, tension zones may be formed (Barton and
Hewitt, 1985). Severe fitness consequences of hybrid produc-
tion favor the evolution of reproductive incompatibility
through reinforcement. If the fitness of the hybrids is equal to
or greater than the parental species throughout the zone of
hybridization, as well as in the range of the parental species,
then the parental species are likely to merge. Finally, if hybrids
have superior fitness, but only in specific habitats (e.g., eco-
logically intermediate or disturbed habitats), hybrid zones may
be stably maintained (Endler, 1977; Moore and Buchanan,
1985; but see Arnold, 1997 for explanations not involving se-
lection).
When natural selection plays a role in the maintenance of
hybrid zones, the response of genes under selection is expected
to be different from neutral genes. Genes from one species
that are negatively selected in the alternate species are unlikely
to introgress and may only be found in low frequency in the
hybrid zone. Contrariwise, neutral genes are free to cross hy-
brid barriers and introgress into the parental species. Similarly,
genes under positive selection outside of the hybrid zone may
spread into the alternate parental species (Arnold, 1997). The
direction and degree of introgression, however, is influenced
by the frequency of backcrossing between the hybrids and the
parental species, which may be more common than the initial
production of F1 hybrids (Arnold, 1997).
To assess the degree of introgression, the genealogical clas-
ses (parental, F1, F2, backcrosses, etc.) must be accurately dif-
1758 AMERICAN JOURNAL
ferentiated. Intermediacy (i.e., morphological or physiological)
might be used in some species in the early stages of intro-
gression to indicate hybrid status but generally is only of lim-
ited use. In a review of 46 studies of hybrid morphology, Rie-
seberg and Ellstrand (1993) concluded that artificially created
hybrid plants were no more likely to display intermediate char-
acter states than parental ones, and 10% of morphological
characters measured in F1 hybrids and 30% in later generation
hybrids were shown to be novel or extreme relative to the
parental species. The accurate classification of hybrid back-
crosses and pure parental genotypes usually is not possible
using morphological data alone. The use of molecular genetic
markers has circumvented some of the limitations of morpho-
logical or physiological characters. Many of these markers are
codominant and inherited in a Mendelian fashion, making un-
ambiguous identification of hybrids possible. The use of mul-
tiple, independent molecular markers allows the discrimination
of backcrossed individuals and an assessment of the degree of
introgression (Nason et al., 1992; Nason and Ellstrand, 1993).
Further, studying the distribution of genetic variation and ge-
netic associations at both nuclear–nuclear and nuclear–cyto-
plasmic loci also provides information on patterns of mating
and the direction or symmetry of introgression (Cruzan and
Arnold, 1993; Cruzan, 1998).
In this study, we assess morphological and molecular mark-
ers to evaluate the extent of hybridization and introgression in
Borrichia frutescens (Linnaeus) de Candolle, B. arborescens
(Linnaeus, de Candolle: Asteraceae), and their hybrid, B. 3
cubana (Britton and Blake), in sympatry in coastal areas of
the Florida Keys, Florida, USA. Borrichia is a primarily trop-
ical genus of aster that is divided into six species, three of
which occur in North America. Borrichia frutescens is a
woody subshrub that generally has silvery, pubescent leaves.
The plant’s morphology, however, is extremely plastic and has
a range of habitat-associated morphologies (P. Stiling, Univer-
sity of South Florida, personal communication; M. V. Cattell,
personal observation). Borrichia frutescens grows in saline
marshes and estuaries from Maryland to the Florida Keys
(Semple and Semple, 1977; USDA, NRCS, 1999). The con-
gener, B. arborescens, is a larger, woody shrub that can be
found in two forms, either with silvery pubescent or complete-
ly glabrous leaves. Borrichia arborescens also is found along
saline coasts and mangrove marshes, but ranges throughout
the Greater Antilles to Bermuda and the Florida Keys and in
the west central Caribbean from the Yucatan peninsula to Be-
lize (Semple and Semple, 1977). In the continental United
States, only the glabrous leaf form of B. arborescens is found
and it is never found north of Miami, Florida, USA (Semple
and Semple, 1977). Both species are perennial, reproduce sex-
ually and clonally, and individuals may live in excess of five
years (Antlfinger, 1981). In the Florida Keys, where the ranges
of the two species overlap, they are known to hybridize ex-
tensively (Semple and Semple, 1977). The hybrids (taxonom-
ically designated B. 3 cubana) generally are morphologically
intermediate to the parental species; however, like B. frutes-
cens, the hybrids exhibit a range of habitat-associated mor-
phologies. Flowering characteristics such as timing are similar
in all three species.
Here, nuclear and cytoplasmic (chloroplast) DNA variation
was assessed in areas of parapatry. Symmetry of crossing and
introgression were investigated using a cytoplasmic marker in
conjunction with nuclear markers. Leaf, inflorescence, and
OF BOTANY [Vol. 91
Fig. 1. Map of the Florida Keys, Florida, USA, showing hybrid zone
collection sites of Borrichia frutescens, B. arborescens, and B. 3 cubana.
Circles with associated site names indicate the general location of plants used
for both genetic and morphological studies. Unlabeled circles indicate the
general location of plants included in the genetic study only.
stem morphological traits also were measured and examined
for their utility in distinguishing hybrid and parent plants.
MATERIALS AND METHODS
Sample collections—Samples from 27 different locations throughout the
Florida Keys (Fig. 1) were collected between 1996 and 1999. Collecting sites
were primarily chosen for the presence of several individuals of each species
and to thoroughly cover the geographic range of the area of sympatry. Five
to 20 young leaves were collected from single individuals of each of the
putative species at each location. Samples were placed in individually marked
bags and stored either on dry ice or placed directly into silica gel desiccant.
Frozen leaves were kept on dry ice until they were returned to the laboratory
and stored at 2808C. Desiccated samples were kept cool until they were
returned to the laboratory and placed at 2208C. Because these species exten-
sively propagate clonally, each species at each site generally was sampled
only once to minimize inadvertent, multiple sampling of the same individual.
However, when groupings of plants (a patch) were separated by a gap of at
least 3 m, individuals were sampled from multiple patches at a site. A total
of 50 B. frutescens, 53 B. 3 cubana, and 52 B. arborescens individuals were
collected. Samples from putative parental species also were collected from
outside the area of overlap. Seven B. frutescens individuals were sampled at
Upper Tampa Bay Park in Tampa, Florida, two individuals from Ruskin, Flor-
ida, just south of Tampa, and three from individuals on the east coast of
Florida near Jacksonville, Florida. Eight B. arborescens individuals were sam-
pled from the coast of Puerto Rico.
Molecular markers—To obtain single-copy nuclear loci that were diag-
nostic at the species level, a genomic DNA library was constructed and
screened. Total cell DNA from a single individual from Upper Tampa Bay
Park was extracted for library construction using a standard hexadecyltrime-
thylammonium bromide (CTAB) extraction protocol (Milligan, 1992) with
several modifications. One to two grams of leaf tissue were ground to a fine
powder in liquid nitrogen with a mortar and pestle. The macerated tissue was
then added to 8 mL of 658C CTAB extraction buffer and incubated at 658C
for 60 min. After incubation, the solution was extracted twice with 25 : 24 :
1 phenol : chloroform : isoamyl alcohol and once with 24 : 1 chloroform :
isoamyl alcohol. DNA was precipitated by the addition of two-thirds volume
isopropanol and incubation for 30 min at 2208C. Purified DNA was washed
November 2004] CATTELL AND KARL—INTERSPECIFIC
TABLE 1. Polymerase chain reaction annealing conditions and results of restriction enzymes survey of three individuals of Borrichia frutescens
and B. arborescens from Florida USA.
Annealing
Locus temperature (8C) Annealing time (s)
3F6 52 60
2D4 48 60
ITS 52 60
6G3 55 45
Taba-ba 56 45
3B6 62 30
11C3c 50 45
TrnM-rbcLa 58 30
TrnL1-L2a 58 30
TrnH-Ka 58 45
a Chloroplast locus. Primer names refer to those in Taberlet et al. (1991) and Demesure et al. (1995).
b This locus was size polymorphic; however, digestion with HinfI resulted in smaller fragments more clearly indicating the size difference.
c Although this locus contained an apparent interspecific restriction site polymorphism, it was not used in the study because the restriction profile
scoring was unreliable.
with 95% ethanol, dried, and resuspended in 400 mL of a 10 mmol/L tris-CL
and 1 mmol/L ethylenediaminetetraacetic acid (TE) solution.
The library was prepared and screened for single-copy loci according to
Karl and Avise (1993). Briefly, total cell DNA was digested with Sau3A and
fragments from 500 bp to 5000 bp were gel isolated. Size-fractioned DNA
was ligated into the cloning vector pBSSK 1 (Stratagene; La Jolla, California,
USA), transformed into bacteria, and plated on antibiotic media. Recombinant
clones were screened for insert size by restriction endonuclease digestion of
mini-DNA preparations (Ausubel et al., 1993), and genomic copy number
was determined by dot blot analysis using labeled total cell DNA as a probe
(Karl and Avise, 1993). Once single-copy clones were identified, 18 were
arbitrarily chosen from the library for sequencing and primer construction.
Cloned inserts were directly sequenced from mini-prep DNA using either the
dideoxy chain-termination method (Sanger et al., 1977) with a Sequenase
version 2.0 sequencing kit (USB; Cleveland, Ohio, USA) and size sorted on
acrylamide gels or with an ABI Prism automated sequencing kit (Big Dye or
D-rhodamine; Perkin-Elmer, Boston, Massachusetts, USA) and analyzed on
an ABI 310 automated sequencer. Sequences were aligned manually using the
computer program SeqEd (Applied Biosystems; Foster City, California, USA)
and screened for possible polymerase chain reaction (PCR) priming sites using
the computer programs OLIGO (Molecular Biology Insights; Cascade, Col-
orado, USA) and Primer 3 (Rozen and Skaletsky, 1997). Primers to 11 clones
were synthesized by Cybersyn (Lenni, Pennsylvania, USA), and PCR ampli-
fication conditions were optimized using the genomic DNA used in library
construction and miniprep vector DNA from the corresponding clone. Exact
thermal-cycling parameters varied depending on the primers used and were
determined empirically, generally by varying annealing temperature and time
(Table 1).
Primers to three of the 11 single-copy nuclear loci amplified reliably, pro-
ducing a single product of the expected size and were screened for polymor-
phic restriction endonuclease sites. One microliter of genomic DNA from
three individuals from each species was amplified in 50-mL reactions con-
taining 3.0 mmol/L of each dNTP, 2.5 units Taq polymerase (Promega, Mad-
ison, Wisconsin, USA), 2.5 mmol/L MgCl2, 0.10 mmol/L bovine serum al-
bumin (Roche Diagnostics; Indianapolis, Indiana, USA), and 0.41 mmol/L of
each primer. Thermal cycling generally consisted of a 2-min denaturation at
958C, followed by 30 cycles of 1 min at 958C, annealing (Table 1), extension
for 1 min at 728C, and a single final extension step of 7 min at 728C.
To check the amount and fidelity of amplification, ;5 mL of each undi-
gested sample was electrophoresed through a 2.0% agarose gel stained with
ethidium bromide. Successfully amplified DNA was used directly in restric-
tion digestions, without further purification. Eight microliters of each ampli-
fication were digested separately in a 20-mL reaction volume with 10 units
of restriction endonuclease enzyme for a minimum of 2 h, according to the
manufacturer’s recommended reaction conditions (F. Hoffmann-La Roche, Ba-
HYBRIDIZATION IN BORRICHIA 1759
Species-specific
Fragment size (bp) No. enzymes screened enzyme used
312 12 RsaI
298 9 AluI
771 7 HaeIII
941 3 EcoRI
711 3 —b
677 26 —
740 6 —
2866 9 —
472 7 —
1630 6 —
sel, Switzerland). Each locus was digested with up to 26 different enzymes
or until a variable site was found. Digested DNA was electrophoresed in 2.5%
agarose gels stained with ethidium bromide, and restriction fragment profiles
were revealed using short-wave ultraviolet light. Enzymes that produced eas-
ily scored restriction fragment profiles that appeared to be species specific
were used on the whole set of plant samples. As an additional nuclear locus,
the universal PCR primers ITS 4 and 5 (White et al., 1990) were used to
amplify the internal transcribed spacer region between ribosomal genes. The
amplified fragment was digested as described earlier with the same set of
restriction enzymes.
To identify a species-specific cytoplasmic marker, several sets of universal
chloroplast primers were screened (Taberlet et al., 1991; Demesure et al.,
1995). The fragment generated using the ‘‘a’’ and ‘‘b’’ primers of Taberlet et
al. (1991) contained a small, apparently species-specific size difference and
was used to genotype all individuals. Chloroplast fragments were digested
with restriction enzymes as for the nuclear loci. When digested with HinfI,
the restriction endonuclease profile consisted of several smaller fragments,
which eased the scoring of the size difference. To verify the presence and
size of the indel, one individual of each species was amplified, sequenced,
and the sequences aligned as described before.
Individual genotypes—DNA was extracted from leaves using the CTAB
procedure detailed before, with modifications as follows. The extractions were
scaled down so that they could be completed in 1.5-mL microcentrifuge tubes.
For these extractions, a piece of leaf, roughly 1 cm in length (smaller for
dried samples) was ground in 100 mL of 658C CTAB buffer with a small
amount of sterile sand and a sterile, reusable plastic pestle. Once the tissue
was well macerated, 400 mL of warm CTAB buffer was added. The sample
was incubated at 658C for 1–2 h, and the DNA was extracted organically and
purified as for the library construction. The DNA was resuspended in 100 mL
TE. The samples were further purified for PCR amplification by spinning them
through Microcon 100 000 nominal molecular weight limit spin columns. The
final sample volume was adjusted to 50 mL with TE.
Morphological measurements—Leaf samples for genetic analysis were
taken from five stems per plant of each of the three putative species that were
haphazardly collected from each of two individuals within five sites (Fig. 1;
Card Sound, Teatable, Boot Key, Little Torch, and Rockland). Within several
hours of collection; the diameter of the stem tip and the length, width, and
thickness of the third leaf from the meristem were measured with dial calipers
to the nearest one-hundredth of a millimeter. If flowers were present, the
height of the inflorescence was measured from the base of the phyllaries to
the base of the ray flowers. The width of the inflorescence was measured at
the widest point, and the length of the first pair of phyllaries was measured.
These morphological traits were chosen because they appeared to differ be-
1760 AMERICAN JOURNAL OF BOTANY [Vol. 91

TABLE 2. Locus variability revealed by restriction endonuclease digestion.

No. nucleotides Sequence No. sites Nucleotides

Locus No. cut sites surveyed surveyed (%) polymorphic polymorphic (%)

Borrichia frutescens
3F6 5 20 6.41 1 5.00
2D4 5 22 7.38 3 13.64
ITS 5 20 2.59 0 0.00
6G3 2 10 1.06 0 0.00
Taba-ba 3 14 1.97 0 0.00
3B6 10 42 6.20 1 2.38
11C3 4 16 2.16 0 0.00
TrnM-rbcla 42 168 5.86 0 0.00
TrnL1-L2a 6 24 5.09 0 0.00
TrnH-Ka 10 40 2.45 0 0.00
Borrichia arborescens
3F6 4 16 5.13 0 0.00
2D4 4 16 5.37 1 6.25
ITS 6 24 3.11 0 0.00
6G3 4 20 2.13 0 0.00
Taba-ba 3 14 1.97 0 0.00
3B6 11 46 6.79 1 2.17
11C3 2 8 1.08 1 12.50
TrnM-rbcla 42 168 5.86 0 0.00
TrnL1-L2a 6 24 5.09 0 0.00
TrnH-Ka 10 40 2.45 0 0.00
a Chloroplast locus. Primer names refer to those in Taberlet et al. (1991) and Demesure et al. (1995).

tween the parent species and could be measured reliably in the field before
leaves began to wilt.
Data analysis—Estimates of Hardy-Weinberg genotypic frequency equilib-
rium (HWE), linkage disequilibrium, FST, and log-likelihood (LL) assignment
of individual genotypes to species (B. frutescens or B. arborescens) for stan-
dard, genotypic, co-dominant data with gametic phase unknown were calcu-
lated using Arlequin version 2.0 (Schneider et al., 2000). Program default
settings were used for all the genetic analyses. Cytonuclear disequilibrium
(CND; D, D1, D2, and D3) values were estimated using the CND computer
package (Asmussen and Basten, 1994; Basten, 1996). Here, D measures the
nuclear allelic disequilibrium, D1 measures the disequilibrium between one
cytotype and one homozygous genotype, D3 measures the disequilibrium be-
tween the same cytotype and the other homozygous genotype, and D2 mea-
sures the disequilibrium between the cytotype and the heterozygous genotype
(Asmussen et al., 1987; Arnold, 1993; Asmussen and Basten, 1994). Cyto-
nuclear disequilibria were only calculated between the chloroplast and the
nuclear loci ITS and 6G3. The 2D4 and 3F6 loci were not alternately fixed
in the pure parental species making this analysis inappropriate (Arnold, 1993).
Linear regression analysis of the LL genotype scores vs. site of collection for
all individuals was used to indicate geographical localization of parental and
hybrid genotypes.
The occurrence and extent of introgression and hybridization was assessed
following the hierarchical hypothesis testing approach of Nason and Ellstrand
(1993). Individuals were assigned to one of seven genotypic categories based
on their multilocus nuclear genotypes at loci with species-unique alleles. In-
dividuals were classified as pure parental A or B (genealogical classes P1 and
P2 for B. frutescens and B. arborescens, respectively), first filial hybrids H
(F1), second filial hybrids S (F2), or backcrosses A1 or B1 (BP1 and BP2; Nason
and Ellstrand, 1993). A series of hierarchical hypotheses was tested, in which
the presence of nonparental multilocus genotype classes indicates hybridiza-
tion and the presence of second filial or backcross genotype classes indicates
introgression. Maximum likelihood estimates of each class were used to test
for the significance between a reduced model, assuming no backcrossed in-
dividuals, and a full model, with all six classes present. These models were
tested constraining the estimates to values between zero and one and uncon-
strained. A log-likelihood ratio tested with a chi-square statistic indicates sta-
tistical significance between the models (for computational details, see Nason
and Ellstrand, 1993). Standard deviations of the genotype class frequencies
were estimated by bootstrapping with 600 replicates.
Morphological data were tested for significant differences between species
using one-way analysis of variance (ANOVA). Square root or log10 transfor-
mations were used when necessary to fit model assumptions. When assump-
tions could not be met, Kruskal–Wallace nonparametric tests were applied
using species as the grouping variable. Morphological data were tested for
reliability in discriminating plant genotypes using canonical discriminant
function analyses of the means of measurements within plant patches (each
patch is likely a single individual). To help determine which type of data is
most informative, discriminant function analyses were applied to all of the
morphological characters separately, as well as together, for flower and leaf
morphological data sets. Three of the plants did not have flowering stems and
were excluded from analyses. The proportion of the total variance in the
discriminant scores that is not explained by the group differences (Wilks’ l)
was calculated and transformed to approximate a chi-square distribution test
of significance of differences between group centroids. Eigenvalues (ratio of
between-groups sums of squares to the error sum of squares) also were cal-
culated to help estimate the spread of the group centroids. These statistics
were done using SPSS version 7.5 (SPSS, Chicago, Illinois, USA).
RESULTS
Molecular markers—Per locus variation ranged from 0.00
to 13.64% in B. frutescens and from 0.00 to 12.50% in B.
arborescens (Table 2). The enzymes HaeIII, EcoRI, AluI, and
RsaI produced the most reliable and easily scored species-
specific polymorphisms for ITS, 6G3, 2D4, and 3F6, respec-
tively, and were used to genotype all individuals. The chlo-
roplast locus was size polymorphic.
Screening 175 individuals at the four nuclear loci and one
chloroplast locus resolved 37 unique multilocus genotypes
(Table 3). Assessment of putative pure parental species outside
of the Florida Keys (i.e., in parapatry) revealed that the nuclear
loci ITS and 6G3 were essentially fixed for different alleles in
B. frutescens and B. arborescens (Table 3). For convenience,
nuclear alleles that were predominantly B. frutescens are cap-
November 2004] CATTELL AND KARL—INTERSPECIFIC HYBRIDIZATION IN BORRICHIA 1761

TABLE 3. Individual genotypes and log-likelihood assignment probabilities for 175 Borrichia individuals in this study. Capital letters indicate the
typically B. frutescens alleles and lowercase letters indicate the typically B. arborescens alleles. In general, lower-numbered multilocus genotypes
were more B. frutescens-like and the higher-numbered ones more B. arborescens-like. Assignment probabilities were calculated excluding the
chloroplast locus; therefore, individuals with identical nuclear but different chloroplast genotypes have identical log-likelihood values.

Locus Log-likelihood
Geno-
type Number ITS 6G3 2D4 3F6 cp B. frutescens B. arborescens

1 2 II GG DD FF B. fru. 22.842 212.782

2a 27 II GG DD Ff B. fru. 21.368 210.130
3b 34 II GG DD ff B. fru. 21.279 28.865
4 1 II GG Dd FF B. fru. 24.332 211.999
5 2 II GG Dd Ff B. fru. 22.858 29.347
6 1 II GG dd ff B. fru. 25.646 28.685
7c 7 II GG N/Ad FF B. fru. 27.476 223.762
8 1 II Gg DD ff B. fru. 23.906 26.582
9 2 II Gg DD ff B. arb. 23.906 26.582
10 1 II Gg Dd FF B. fru. 26.960 29.717
11 1 II Gg Dd ff B. arb. 25.396 25.799
12 1 II Gg dd ff B. fru. 28.273 26.402
13 1 Ii GG DD ff B. arb. 23.906 26.575
14 1 Ii GG Dd FF B. arb. 26.960 29.710
15 1 Ii GG Dd ff B. fru. 25.396 25.792
16 1 Ii GG Dd ff B. arb. 25.396 25.792
17 1 Ii GG dd Ff B. fru. 28.361 27.661
18 1 Ii GG dd ff B. fru. 28.273 26.395
19 1 Ii Gg DD FF B. arb. 28.097 28.211
20 1 Ii Gg DD Ff B. fru. 26.622 25.559
21 1 Ii Gg DD Ff B. arb 26.622 25.559
22 3 Ii Gg DD ff B. fru. 26.533 24.293
23 1 Ii Gg DD ff N/Ad 26.533 24.293
24 3 Ii Gg Dd FF B. fru. 29.587 27.428
25 3 Ii Gg Dd FF B. arb. 29.587 27.428
26 1 Ii Gg Dd Ff B. fru. 28.112 24.776
27 1 Ii Gg Dd Ff B. arb. 28.112 24.776
28 5 Ii Gg Dd ff B. fru. 28.023 23.510
29 6 Ii Gg Dd ff B. arb. 28.023 23.510
30 2 Ii Gg dd FF B. arb. 212.463 28.031
31 1 Ii Gg dd ff B. fru. 210.900 24.113
32 1 Ii gg Dd ff B. arb. 214.425 26.911
33 1 ii Gg Dd ff B. fru. 212.037 22.607
34 7 ii gg DD ff B. arb 214.560 22.494
35e 8 ii gg Dd ff B. fru. 216.060 21.711
36 28 ii gg Dd ff B. arb. 216.060 21.711
37 15 ii gg dd ff B. arb. 218.926 22.314
a Two individuals of Borrichia frutescens from Ruskin, Florida, USA (just outside of Tampa), had this genotype.
b Three individuals of Borrichia frutescens from Jacksonville, Florida, USA, had this genotype.
c Seven individuals of Borrichia frutescens from Upper Tampa Bay Park, Florida, USA, had this genotype.
d Genotype could not be determined.
e Eight individuals of Borrichia arborescens from Puerto Rico had this genotype.

italized and those that were predominantly B. arborescens are
lowercase throughout the text. Loci 2D4 and 3F6 were not
fixed for alternate alleles in the parent species but did exhibit
significant species-specific allele frequencies (FST,2D4 5 0.448,
P , 0.001; FST,3F6 5 0.334, P , 0.001). Parapatric B. frutes-
cens individuals possessed only one 2D4 allele (D), while an
alternate allele (d) was found in all parapatric and most pu-
tative B. arborescens individuals in the hybrid zone. A similar
situation was true for locus 3F6 with parapatric B. arborescens
individuals fixed for one allele (f), and a second allele (F) was
found in all parapatric and most putative B. frutescens indi-
viduals in the hybrid zone. Both the ‘‘d’’ and ‘‘F’’ alleles were
considered to be species unique. Two chloroplast haplotypes
were identified: a B. frutescens haplotype found in all para-
patric B. frutescens individuals and a B. arborescens haplotype
found in all of the presumed pure B. arborescens individuals
from the Florida Keys. Unexpectedly, the eight putatively pure
B. arborescens individuals from Puerto Rico contained the B.
frutescens chloroplast haplotype.
Population genetic analysis—When all plants (i.e., sym-
patric and parapatric) were lumped together, there were sig-
nificant deviations (P # 0.001) from HWE expectations at
ITS, 6G3, and 3F6. In each case, there were fewer than ex-
pected heterozygotes. Additionally, while the deviation from
HWE expectation was not statistically significant for locus
2D4 (P 5 0.165), the direction was the same as for the other
loci (fewer than expected heterozygotes). Significant linkage
disequilibrium also was found between all possible nuclear
locus pairs (P # 0.001). Both of these results are consistent
with lumping genetically differentiated groups (e.g., pure par-
entals in parapatry). When morphologically classified, B. fru-
tescens and B. arborescens species were analyzed separately
regardless of collection location, no significant HWE devia-
1762 AMERICAN JOURNAL
TABLE 4. Maximum likelihood frequency estimates of genealogical classes in the Florida Keys with (full model) and without (reduced) the
assumption of backcrossing and with and without constraining the estimates between zero and one. Borrichia frutescens is P1 and B. arborescens
is P2.
Model P1 P2
Reduced
Unconstrained 0.3655 0.3157
Constrained 0.3655 0.3157
Full
Unconstrained 0.3479 0.3267
Constrained 0.3281 0.3081
SDa 60.0403 60.0386
a Standard deviations estimated by 600 bootstrap replicates.
tions were indicated (PITS and 6G3 5 1.000 for both species, P2D4
5 0.135 for B. frutescens and 0.071 for B. arborescens, P3F6
5 0.599 for B. frutescens and 1.000 for B. arborescens). For
B. 3 cubana, however, there were significantly more hetero-
zygotes at ITS and 6G3 (P 5 0.004) and significantly fewer
heterozygotes at 3F6 (P , 0.001) than expected, but no sig-
nificant deviation at 2D4 (P 5 1.000). All species pairwise
FST values were fairly large and statistically significant (P #
0.001; B. frutescens vs. B. 3 cubana FST 5 0.188, B. frutes-
cens vs. B. arborescens FST 5 0.788, and B. 3 cubana vs. B.
arborescens FST 5 0.446).
Significant cytoplasmic-nuclear disequilibrium was found
between the B. frutescens chloroplast haplotype and the nu-
clear genotypes at ITS and 6G3 in the Keys individuals. The
D and D1 values were significant (P # 0.05) and positive (DITS
5 0.1780, D6G3 5 0.1813, D1,ITS&6G3 5 0.1841) and D3 was
significant and negative (D3,ITS 5 20.1720 and D3,6G3 5
20.1785). D2 values (20.0121 and 20.0056 for ITS and 6G3,
respectively) were not significantly different from zero. This
analysis indicates that the B. frutescens cytotype is found more
often with homozygous B. frutescens nuclear genotypes (D1)
and less often with homozygous B. arborescens nuclear ge-
notypes (D3) than expected under random mating. There is no
trend for the heterozygous state to be found with either chlo-
Fig. 2. Log-likelihood of each genotype being assigned to Borrichia fru-
tescens or B. arborescens. Assignment analysis is based on nuclear genotypes
only. The numbers in the graph correspond to the numbers assigned to the
genotypes in Table 3. Individuals that are most like B. arborescens are located
in the lower right corner and those most like B. frutescens are in the upper
left. Individuals with log-likelihood in the center of the graph are likely hy-
brids and backcrosses.
OF BOTANY [Vol. 91
F1 F2 BP1 BP2
0.0709 0.2479 0.0000 0.0000
0.0709 0.2479 0.0000 0.0000
0.0718 0.2138 0.1001 20.0603
0.0677 0.2017 0.0944 0.0000
60.0295 60.0842 60.0613 60.0405
roplast type (D2) more often than expected under random mat-
ing.
Following Nason and Ellstrand (1993), the maximum like-
lihood estimates of genotypic class frequencies also indicated
a significant amount of hybridization (Table 4). Both the re-
duced model, for which the backcross frequencies were set to
zero, and the full model, which included all classes, gave sim-
ilar results estimating that about 32% of the Keys population
consisted of nonparental types. Relative to the likelihood of
the reduced model (Ln 5 216.5082), the full model was a
significantly better fit to the data in both the constrained (Ln
5 210.50172, P # 0.01) and unconstrained (212.8930, P #
0.05) analyses. Regardless of the model, most of the nonpar-
ental individuals appeared to be second generation hybrids (F2;
;20 to ;25% of all individuals) with very few classified as
first generation hybrids (F1; ;7%; Table 4). A statistically sig-
nificant number of individuals was estimated to be back-
crossed to B. frutescens (;10%) but not to B. arborescens
(Table 4). The estimated number of hybrids backcrossed to B.
arborescens was negative, but not significantly so.
Assignment tests—Results of the LL genotype assignment
analysis are plotted in Fig. 2 and the likelihood values are in
Table 3. The genotypes at the far extremes of the spread of
the points are most likely pure parental genotypes. In general
in the multilocus genotypes assignment, low genotype num-
bers are more B. frutescens-like and genotypes 1 to 3 and 7
are likely pure B. frutescens individuals. Higher numbered ge-
notypes are more B. arborescens-like and genotypes 34 to 37
are probably of pure B. arborescens individuals. The geno-
types that fall in the center of the distribution of Fig. 2 are
intermediate in LL score and most likely are hybrids and back-
crosses. Log-likely scores were not, however, geographically
assorted, and extreme, as well as intermediate LL scores, were
found at nearly all sites (Fig. 3). A linear regression of LL
genotype score vs. geography for the putatively B. 3 cubana
individuals results in an insignificant slope of 0.0051 with an
r2 5 0.0037 (Fig. 3).
Morphological characteristics—All B. frutescens that were
morphologically measured had either genotype 2 or 3, the B.
3 cubana individuals had genotypes 9, 15, 24, 25, 28, or 29,
and B. arborescens individuals had genotypes 36 or 37 (Table
3). Individuals that were genetically identified as hybrids were
not always intermediate to the parents in the morphological
characters. Significant differences between plant species were
identified for leaf length, leaf thickness, inflorescence height,
first phyllary length, flower number, and stem tip diameter (P
November 2004] CATTELL AND KARL—INTERSPECIFIC HYBRIDIZATION IN BORRICHIA 1763

Fig. 3. Plot of the B. frutescens divided by B. arborescens log-likelihood

genotype assignment scores vs. geographic site of collection for samples from
the Florida Keys only. Circles are individuals with nuclear genotypes identical
to B. frutescens found in parapatry, squares are individuals with nuclear ge-
notypes identical to B. arborescens found in parapatry, and triangles are all
other genotypes (i.e., not pure parental).

# 0.001; data not shown). Putative hybrids were on average

intermediate in measurement for all measured morphological
characters, except leaf length, for which hybrids were the
shortest. No difference was found for leaf and inflorescence
width.
Discriminant function analyses (DFA) indicate that leaf
morphology provides good discrimination along function 1 for
pure B. arborescens vs. pure B. frutescens or B. 3 cubana
(Fig. 4A). Borrichia frutescens and the hybrid were better dif-
ferentiated along function 2; however, six individuals were
misgrouped, three genetically B. frutescens individuals were
grouped with B. 3 cubana, and three genetic hybrids were
grouped with B. frutescens (Fig. 4A). Flower morphology
alone did not discriminate as well between the plant species;
11 individuals were grouped incorrectly. The B. frutescens and
B. arborescens clusters were relatively well defined, but B. 3
cubana individuals clustered with one or the other parental
species and did not form a third grouping (Fig. 4B). The
spread of the group centroids (as indicated by the eigenvalues
for function 1; leaf 5 2.973, flower 5 1.404, combined mor-
phological 5 13.967) and the proportion of the discriminant
scores that are not explained by differences among groups
(Wilks’ l values for function 1 through 2; leaf 5 0.219, flower
5 0.340, combined morphological 5 0.036; P , 0.05) also
indicate that leaf morphological characters are better at dis-
criminating the groups than flower morphology. Discrimina-
tion among the groups was best when all morphological data
were included in the analysis (Fig. 4C). Here, only one of 27
individuals was misgrouped (a B. 3 cubana individual was
classified as B. frutescens). The magnitudes of the eigenvalues
indicate that the spread between the group centroids was great-
est when using all characters. Wilks’ l indicates that 96.4%
of the variance in the DFA scores was explained by among-
group differences.
Fig. 4. Discriminant function analyses of morphological characters for (A)
DISCUSSION leaf morphology only, (B) flower morphology only, and (C) all of the mor-
phological characters combined. Clusters of genotype groups as defined by
The application of a molecular genetic approach to hybrid- the DFAs are enclosed in the lines with a dashed line for B. frutescens, a
ization and introgression in Borrichia proved quite useful. solid line for B. arborescens, and a dotted line for B. 3 cubana. Circles are
Multilocus genotypes based on one cytoplasmic and four nu- B. frutescens individuals, squares are B. arborescens, and triangles are B. 3
clear markers were able to resolve a variety of genotypes and cubana. Stars indicate species cluster centroids.
genotypic classes (e.g., backcrosses). Using an LL assignment
to species resulted in a continuum of LL values ranging from
1764 AMERICAN JOURNAL
very likely B. frutescens and unlikely B. arborescens to very
unlikely B. frutescens and likely B. arborescens. The B. fru-
tescens individuals from outside of the hybrid zone (genotypes
2, 3, and 7) corresponded to large B. frutescens LL values and
small B. arborescens values (Table 3 and Fig. 2), supporting
the assertion that they are pure parental types. Morphologically
putative pure B. frutescens found within the hybrid zone had
genotypes 1 to 3, and large B. frutescens-like LL and small B.
arborescens-like LL. Borrichia arborescens individuals from
outside the hybrid zone (i.e., Puerto Rico) had genotype 35,
and LL scores, indicating clear B. arborescens-like genotypes.
Individuals that were morphologically B. arborescens-like
within the zone had genotypes 34, 36, or 37 and similarly high
LL assignment to B. arborescens. Many genotypes had more
moderate or nearly equal LL scores for both species assign-
ments, indicating that they are likely hybrids (center of the
graph in Fig. 2). These data clearly indicate that a considerable
amount of hybridization must be taking place in the Florida
Keys. It also is clear that not all individuals in the Florida
Keys are hybrids and that pure parental types are persisting
sympatrically with hybrids (Fig. 3).
Not all markers were alternately fixed for species-specific
alleles; however, our approach as outlined by Nason and Ells-
trand (1993) allowed a robust estimation of genealogical clas-
ses from the observed genotypic data. Maximum likelihood
estimates of genealogical class based on observed multilocus
genotypes also indicated that a significant amount of hybrid-
ization and backcrossing is taking place. In the Florida Keys
hybrid zone, most individuals were either pure parental types
or later generation hybrids (i.e., F2 or backcrosses; Table 4).
Only about 7% of the individuals were of the F1 hybrid class.
This low frequency of F1 individuals has been observed often
in other plant hybrid systems and is believed to be character-
istic of a system in which the formation of the initial hybrid
generation (i.e., F1) is a difficult or an infrequent event (Arnold
et al., 1992, 1993; Nason et al., 1992) probably as a result of
pre- and post-fertilization barriers between the parental spe-
cies. Once this compatibility barrier is breached, however, fur-
ther crossings are less difficult and more likely (Arnold, 1997).
Considering the cytonuclear disequilibrium analysis, however,
the insignificant D2 value indicates that there does not seem
to be asymmetry in the direction of hybridization. Apparently
then, the barrier to hybridization is equal regardless of which
species is the maternal or the paternal parent. In addition, the
maximum likelihood estimates of genotype classes indicate
that backcrossing is a significant event in this hybrid zone.
Unlike hybridization, backcrossing does not; however, appear
to be symmetrical with respect to parental species. Of the non-
parental genealogical classes estimated in the constrained full
model, 55.4% of them are BP1 and none are BP2. Clearly,
fertilization between hybrids and B. frutescens is more easily
or frequently accomplished than with B. arborescens. It should
be noted, however, that, given the small sample size (53 hybrid
individuals), while backcrossing with B. arborescens clearly
is less common, we cannot rule it out entirely.
The discriminant function analyses of the morphological
characters suggest that hybrid plants are more morphologically
distinct from B. arborescens than they are from B. frutescens
(Fig. 4). In all of the analyses, the B. arborescens cluster and
group centroid were set farther apart from the B. frutescens
and B. 3 cubana clusters and centroids than the latter two
were from each other. Additionally, only four individuals clus-
tering outside their groups involved B. arborescens. It is im-
OF BOTANY [Vol. 91
portant to note that the misclassifications occurred when con-
sidering flower morphology alone—the least differentiated of
the morphological comparisons. All other misclassifications by
the DFA were between B. frutescens and B. 3 cubana (six
with leaf morphology alone, nine with flower morphology
alone, one with all morphological data). Interestingly, this is
consistent with the hybrid, B. 3 cubana, consisting of a larger
portion of individuals backcrossed to B. frutescens than to B.
arborescens. Finally, although all individuals with high LL
assignment values were of the appropriate morphology, not all
putative species assignments based on morphology were ex-
tremes in LL score. These results, when considered together,
suggest that morphological characters may not provide very
accurate information on the identity of Borrichia species with-
in the region of species overlap. This is particularly true when
differentiating pure B. frutescens from hybrid plants. If mor-
phological assignments are going to be made, these data clear-
ly indicate that several morphological characters must be mea-
sured to accurately assess identity.
There are several factors, such as the distribution of the
parental plants and how long hybridization has been taking
place, that likely contribute to the genetic patterns observed in
the Florida Keys. While it is difficult to determine with cer-
tainty how long the Borrichia species in the Florida Keys have
been hybridizing, it is possible that the plants may have been
there for a long time. It has been estimated that the limestone
of south Florida has been exposed for from 5000 to 8000
years, suggesting the potential for a long period of evolution
in the Florida Keys (Snyder et al., 1990). Thus the habitat may
have been available to these plants for up to 8000 years. If
true, and in the absence of other evolutionary forces, hybrid-
ization and random mating likely would have led to equilib-
rium conditions by now, and few if any pure parental individ-
uals would be expected in the Keys. This is clearly not the
case.
If the hybrid zone is relatively young, historic parental ge-
netic association would still exist and result in the observed
high frequency of parental types. Although there is some in-
dication that these Borrichia species have been sympatric for
some time, it is possible that they have not been hybridizing
for the duration of their sympatry. Semple and Semple (1977)
suggest that hybridization has only recently occurred as a re-
sult of disturbance in the Florida Keys and that less disturbed
habitats tended to be occupied by only one of the parent spe-
cies. Additionally, even if the hybrid zone is old in terms of
the number of years since hybridization began, it may be rel-
atively young in terms of the number of generations that have
passed. These Borrichia species are highly clonal and invest
a relatively small amount into reproductive structures (Semple
and Semple, 1977). Generation time for these plants is un-
known and is generally difficult to establish for clonal plants;
however, vegetatively reproducing genets may persist for hun-
dreds or thousands of years (Wolf et al., 2000). It is therefore
possible that what we are detecting is the early formation of
a hybrid zone. The few putatively F1 hybrid individuals iden-
tified also may be pointing to a young age for the hybrid zone
because at the onset of hybridization, only a minority of the
plants would have hybrid genotypes. This is not supported,
however, by prevalent occurrence of later generation hybrids
and backcrosses. Given this and that suitable habitat likely has
been available for quite some time, we believe that the hybrid
zone is not a recent creation but that other factors are main-
taining pure parentals and hybrids in sympatry.
November 2004] CATTELL AND KARL—INTERSPECIFIC
If this were a geographically simple hybrid zone with a
majority of pure parental genotypes at the extreme ends of the
region of overlap and relatively few parental genotypes in the
center, then selection against hybrids at the extremes and/or
migration of pure parental individuals to the center could ac-
count for the observed genetic patterns. Selection against hy-
brid genotypes in a hybrid zone also could cause a reduction
of heterozygotes and significant disequilibrium. Selection,
however, would have to be acting directly on the loci under
study, or closely linked loci, because neutral alleles can diffuse
through hybrid zones and parental genetic associations should
be disrupted by recombination (Arnold et al., 1987; Hewitt,
1988; Marchant et al., 1988). It is unlikely, however, that se-
lection is acting directly on the restriction site polymorphisms
at the loci used here (see Karl and Avise, 1993, for a discus-
sion of the neutrality of restriction site polymorphisms at sin-
gle-copy nuclear loci). Additionally, they could be linked to
loci under selection; however, this seems unlikely to be the
case simultaneously for four different nuclear loci, as well as
for a cytoplasmic one.
Migration from pure parental populations into a hybrid zone
also can cause continued association of nuclear and cytoplas-
mic genotypes after long periods of hybridization by contin-
ually reintroducing the parental gene combinations into the
zone. In this study, there was no apparent association between
genotype LL score and geography throughout the hybrid zone.
Individuals with genotype LL scores similar to those found in
both species in parapatry (i.e., putative pure parental types)
were observed at all sites along with ones that were clearly
hybrids (Fig. 3). Individuals from throughout the range of LL
scores were found at nearly all sites north to south, and linear
regression analysis of the LL scores of putative hybrid indi-
viduals vs. collection site resulted in an insignificant slope. For
migration to result in the pattern of geographic distribution
and cytonuclear disequilibrium found in this study, migration
rates into the Florida Keys from both pure parent plant pop-
ulations would have to be similar. Equal rates of migration
from pure B. frutescens and pure B. arborescens populations
outside of the zone of hybridization are unlikely from geo-
graphical considerations. The Florida Keys hybrid population
is contiguous with B. frutescens individuals to the north, but
is clearly disjunct from B. arborescens populations in the Ca-
ribbean. The closest B. arborescens individuals are in the Ba-
hamas and Cuba, each of which is separated from the Florida
Keys by at least 80 km of ocean. It seems reasonable, there-
fore, to expect a greater degree of migration and gene flow
from pure B. frutescens populations into the hybrid zone than
from B. arborescens. The genealogical class analysis does in-
dicate a significant amount of backcrossing to B. frutescens
(9.44 6 6.13%) but not to B. arborescens (0.00 6 4.05%),
and this may be due to the presence of the northern reserve
of pure B. frutescens. Notably, although Semple and Semple
(1977) successfully produce viable offspring from artificial
crosses involving B. arborescens and B. 3 cubana, no prog-
eny were produced from B. frutescens and B. 3 cubana cross-
es. Nonetheless, given our sample size, the large standard er-
rors on the maximum likelihood (ML) estimates, and the over-
all low level of backcrossing, it is difficult to conclude true
asymmetry in backcrossing or persistence of parental types in
the hybrid zone due to migration.
We believe that the clonal nature of these species and pre-
dominantly positive assortative mating explain the genetic data
from the Florida Keys. Clonal propagation would have two
HYBRIDIZATION IN BORRICHIA 1765
principle effects. First, if clonal propagation were the dominant
form of reproduction, the opportunity for hybridization
through necessarily sexual means would be reduced and the
resultant merging of the parental species would be slowed.
Second, and more important here, allowing a significant con-
tribution of both sexual and clonal reproduction might simul-
taneously produce frequent opportunities for hybridization
while increasing the longevity of the parental individuals in
the hybrid zone. Clonality also allows for the persistence and
growth of relatively uncommon, presumably difficult-to-create
F1 hybrids, increasing the likelihood of backcrossing and pro-
duction of later generation hybrids. The FST values calculated
indicate that there is genetic differentiation between the three
groups of plant types (pure parents and hybrid). The FST values
also indicate that there is a greater degree of population dif-
ferentiation between the parents than there is between both
parents and the hybrid group, as expected with assortative mat-
ing. For the most part then, in spite of hybridization and back-
crossing, genetic differentiation between the parental species
is being maintained. Random mating between the plants in this
hybrid zone, therefore, seems unlikely. The population genetic
patterns found suggest that the Florida Keys population likely
is exhibiting positive assortative mating of the parental spe-
cies. No data, however, are currently available regarding the
mechanisms of assortative mating. Considerably more infor-
mation is needed on the specifics of the frequency of sexual
reproduction, mechanisms of pollination, longevity of clonal
units, and the geographic distribution and frequency of geno-
types to fully understand the evolutionary history of this com-
plex region of hybridization.
LITERATURE CITED
ANTLFINGER, A. E. 1981. The genetic basis of microdifferentiation in natural
and experimental populations of Borrichia frutescens in relation to sa-
linity. Evolution 35: 1056–1068.
ARNOLD, J. 1993. Cytonuclear disequilibria in hybrid zones. Annual Review
of Ecology and Systematics 24: 521–554.
ARNOLD, M. L. 1997. Natural hybridization and evolution. Oxford University
Press, Oxford, UK.
ARNOLD, M. L., C. M. BUCKNER, AND J. J. ROBINSON. 1991. Pollen mediated
introgression and hybrid speciation in Louisiana irises. Proceedings of
the National Academy of Sciences, USA 88: 1398–1402.
ARNOLD, M. L., M. B. BULGER, J. M. BURKE, A. L. HEMPEL, AND J. H.
WILLIAMS. 1999. Natural hybridization: how low can you go and still
be important? Ecology 80: 371–381.
ARNOLD, M. L., J. L. HAMRICK, AND B. D. BENNETT. 1993. Interspecific
pollen competition and reproductive isolation in Iris. Journal of Heredity
84: 13–16.
ARNOLD, M. L., J. J. ROBINSON, C. M. BUCKNER, AND B. D. BENNETT. 1992.
Pollen dispersal and interspecific gene flow in Louisiana irises. Heredity
68: 399–404.
ARNOLD, M. L., D. D. SHAW, AND N. CONTREARAS. 1987. Ribosomal RNA-
encoding DNA introgression across a narrow hybrid zone between two
subspecies of grasshopper. Proceedings of the National Academy of Sci-
ences, USA 84: 3946–3950.
ASMUSSEN, M. A., J. ARNOLD, AND J. C. AVISE. 1987. Definition and prop-
erties of disequilibrium statistics for associations between nuclear and
cytoplasmic genotypes. Genetics 115: 755–768.
ASMUSSEN, M. A., AND C. J. BASTEN. 1994. Sampling theory for cytonuclear
disequilibria. Genetics 138: 1351–1363.
AUSUBEL, F. M., R. BRENT, R. E. KINGSTON, D. D. MOORE, J. G. SEIDMAN,
J. A. SMITH, K. STRUHL, P. WANG-IVERSON, AND S. G. BONITZ. 1993.
Current protocols in molecular biology. Greene Publishing Associates
and Wiley-Interscience, New York, New York, USA.
BARTON, H. H., AND G. M. HEWITT. 1985. Analysis of hybrid zones. Annual
Review of Ecology and Systematics 16: 113–148.
1766 AMERICAN JOURNAL OF BOTANY [Vol. 91

BARTON, N. H., AND G. M. HEWITT. 1989. Adaptation, speciation and hybrid
zones. Nature 341: 497–503.
BASTEN, C. J. 1996. CND. Program available at website: http://statgen.
ncsu.edu/brcwebsite/softwarepBRC.php.
CRUZAN, M. B. 1998. Genetic markers in plant evolutionary ecology. Ecology
79: 400–412.
CRUZAN, M. G., AND M. L. ARNOLD. 1993. Ecological and genetic associ-
ations in the Iris hybrid zone. Evolution 47: 1432–1445.
DEMESURE, B., N. SODZI, AND R. J. PETIT. 1995. A set of universal primers
for amplification of polymorphic non-coding regions of mitochondrial
and chloroplast DNA in plants. Molecular Ecology 4: 129–131.
ENDLER, J. A. 1977. Geographic variation, speciation and clines. Princeton
University Press, Princeton, New Jersey, USA.
FRITZ, R. S., C. MOULIA, AND G. NEWCOMBE. 1999. Resistance of hybrid
plant and animals to herbivores, pathogens and parasites. Annual Review
of Ecology and Systematics 30: 565–591.
GALLEZ, G. P., AND L. D. GOTTLIEB. 1982. Genetic evidence for the hybrid
origin of the diploid plant Stephanomeria diegensis. Evolution 36: 1158–
1167.
GRANT, V. 1981. Plant speciation. Columbia University Press, New York,
New York, USA.
HEWITT, G. M. 1988. Hybrid zones—natural laboratories for evolutionary
studies. Trends in Ecology and Evolution 3: 158–167.
KARL, S. A., AND J. C. AVISE. 1993. PCR-based assays of Mendelian poly-
morphisms from anonymous single-copy nuclear DNA: techniques and
applications for population genetics. Molecular Biology and Evolution
10: 342–361.
LEVIN, D. A., J. FRANCISCO-ORTEGA, AND R. K. JANSEN. 1996. Hybridization
and the extinction of rare plant species. Conservation Biology 10: 10–
16.
MARCHANT, A. D., M. L. ARNOLD, AND P. WILKINSON. 1988. Gene flow
across a chromosomal tension zone. I. Relicts of ancient hybridization.
Heredity 61: 321–328.
MILLIGAN, B. G. 1992. Plant DNA isolation. In A. R. Hoelzel [ed.], Plant
DNA isolation in molecular genetic analysis of populations: a practical
approach. Oxford University Press, Oxford, UK.
MOORE, W. S., AND D. B. BUCHANAN. 1985. Stability of the northern flicker
hybrid zone in historical times: implications for adaptive speciation the-
ory. Evolution 39: 135–151.
NASON, J. D., AND N. C. ELLSTRAND. 1993. Estimating the frequencies of
genetically distinct classes of individuals in hybridized populations. Jour-
nal of Heredity 84: 1–12.
NASON, J. D., N. C. ELLSTRAND, AND M. L. ARNOLD. 1992. Patterns of
hybridization and introgression in populations of oaks, manzanitas, and
irises. American Journal of Botany 79: 101–111.
RHYMER, J. M., AND D. SIMBERLOFF. 1996. Extinction by hybridization and
introgression. Annual Review of Ecology and Systematics 27: 83–109.
RIESEBERG, L. H. 1995. The role of hybridization in evolution: old wine in
new skins. American Journal of Botany 82: 944–953.
RIESEBERG, L. H. 1997. Hybrid origins of plant species. Annual Review of
Ecology and Systematics 28: 359–389.
RIESEBERG, L. H., AND N. C. ELLSTRAND. 1993. What can molecular and
morphological markers tell us about plant hybridization? Critical Re-
views in Plant Science 12: 213–241.
RIESEBERG, L. H., AND J. D. MOREFIELD. 1995. Character expression, phy-
logenetic reconstruction, and the detection of reticulate evolution. In P.
C. Hoch and A. G. Stephenson [eds.], Experimental and molecular ap-
proaches to plant biosystematics. Monographs in Systematic Botany from
the Missouri Botanical Garden 53: 333–354.
ROZEN, S., AND H. J. SKALETSKY. 1997. Primer3. Program available at web-
site: http://www-genome.wi.mit.edu/genomepsoftware/other/primer3.html.
SANGER, F., S. NICKLEN, AND A. R. COULSON. 1977. DNA sequencing with
chain terminating inhibitors. Proceeding of the National Academy of Sci-
ences, USA 74: 5463–5467.
SCHNEIDER, S., D. ROESSLI, AND L. EXCOFFIER. 2000. Arlequin, version
2.000: a software for population genetics data analysis. Genetics and
Biometry Laboratory, University of Geneva, Geneva, Switzerland.
SEMPLE, J. C., AND K. S. SEMPLE. 1977. Borrichia 3 cubana (B. frutescens
3 B. arborescens): interspecific hybridization in the Florida Keys. Sys-
tematic Botany 2: 292–302.
SNYDER, J. R., A. HERNDON, AND W. B. ROBERTSON, JR. 1990. South Florida
rockland. In R. L. Myers and J. J. Ewel [eds.], Ecosystems of Florida.
University of Central Florida Press, Orlando, Florida, USA.
STRAUSS, S. Y. 1994. Levels of herbivory and parasitism in host hybrid zones.
Trends in Ecology and Evolution 9: 209–214.
TABERLET, P., L. GIELLY, G. PAUTOU, AND J. BOUVET. 1991. Universal prim-
ers for amplification of three non-coding regions chloroplast DNA. Plant
Molecular Biology 17: 1105–1109.
USDA, NATURAL RESOURCES CONSERVATION SERVICE. 1999. The PLANTS
database. National Plant Data Center website: http://plants.usda.gov/
plants.
WENDEL, J. F., J. M. STEWART, AND J. H. RETTIG. 1991. Molecular evidence
for homoploid reticulate evolution among Australian species of Gossy-
pium. Evolution 45: 694–711.
WHITE, T. J., T. BRUNS, S. LEE, AND J. TAYLOR. 1990. Amplification and
direct sequencing of fungal ribosomal RNA genes for phylogenetics. In
M. A. Innis, D. H. Gelfand, J. J. Sninsky, and T. J. White [eds.], PCR
protocols: a guide to methods and opplications. Academic Press, San
Diego, California, USA.
WOLF, A. T., R. W. HOWE, AND J. L. HAMRICK. 2000. Genetic diversity and
population structure of the serpentine endemic Calystegia collina (Con-
volvulaceae) in northern California. American Journal of Botany 87:
1138–1146.

APPENDIX. Primer sequences for nuclear loci screened in this study.

Primer Forward sequence (59–39) Reverse sequence (59–39)

3F6 TTC TTC TTC CTT CCC CGA TT CGA TGT TTT CGT TCA ACA GG
2D4 TGT AGG TGG GCT GAT CTA CT GTA CTG GAC AGA GCC ACC GT
6G3 TGC ACC CTA AAG CTA CCA CA CGT TCA AAG AGG CGA ATC AA
3B6 GCC TTG ATA CGG GAA TGT GT CAT TGG TAT CGT TTG ACC CC
11C3 TAT TCT ACG GCA GGG ATT GG CTG CGG GTT ACA TGT CAA TG