Genetic variation and phylogeo
n
rri
hio
ral
epte
6 N
Swordtail ?sh have been studied extensively in relation to diverse aspects of biology; however, little attention has been paid to the
Quanti?cation of genetic variation within and among migration of individuals between populations include the
phylogenetic relationships among species have been exam-
ined extensively (Rosen, 1979; Rauchenberger et al., 1990;
Meyer et al., 1994; Borowski et al., 1995; Meyer, 1997;
Morris et al., 2001), little attention has been paid to pat-
terns of genetic variation within and among populations
of swordtails, particularly in a phylogenetic and geographic
* Corresponding author. Present address: Departamento de Biolog??a
Evolutiva, Instituto de Ecolog??a, A.C. Km 2.5 antigua carretera a
Coatepec No. 351, Congregacio?n El Haya Xalapa, Veracruz 91070,
Me?xico. Fax: +52 228 818 7809.
E-mail address: carla.gutierrez@inecol.edu.mx (C. Gutie?rrez-Rodr??-
guez).
Molecular Phylogenetics and Evolutpopulations is crucial to understanding the evolutionary
processes that promote and maintain their biodiversity
(Moritz, 2002) as well as establishing genetic relationships
among populations in order to reconstruct their evolution-
ary history. Within a species, the distribution of genetic
variation within and among populations is in?uenced by
several factors such as gene ?ow and genetic drift. Limited
dispersal of individuals among populations due to the
presence of natural barriers translates into restricted gene
topography of the streams, habitat fragmentation and
patterns of ?ow from one stream to another (Carvalho,
1993).
In this study, we investigate the genetic variation of the
swordtail Xiphophorus cortezi, a freshwater ?sh species
endemic to the Pa?nuco river system of Mexico. Swordtails
and platy?shes (Xiphophorus) have been studied extensively
in relation to their systematics, biogeography, genetics,
oncology, and behavior (reviewed in Me?e and Snelson,
1989) and are considered a model system for studies in
behavioral ecology (Ryan and Rosenthal, 2001). Whilepatterns of genetic variation within and among populations of swordtails. In this study, we sequenced the mtDNA control region from
65 individuals and 10 populations of Xiphophorus cortezi to investigate the genetic variation within and among populations, including
tests for correlations between genetic and geographic distances and tests for species monophyly. We found low gene and nucleotide diver-
sity within populations and high degrees of genetic di?erentiation among populations. Signi?cant and positive correlations between
genetic distance and both river and straight-line geographic distance indicate that genetic di?erentiation among X. cortezi populations
can be explained, to some extent, by an isolation-by-distance model and provide evidence of stream capture. Phylogenetic analyses sug-
gest that X. cortezi is paraphyletic relative to X. malinche, raising questions concerning the status of these taxa as separate species.
 2006 Elsevier Inc. All rights reserved.
Keywords: Genetic di?erentiation; Gene ?ow; Monophyletic; Mitochondrial DNA; Swordtail; Xiphophorus cortezi
1. Introduction ?ow and can result in genetic di?erentiation among popu-
lations. In river systems, factors that can in?uence theXiphophorus cortezi (Cypri
Carla Gutie?rrez-Rodr??guez a,*, Molly R. Mo
a Department of Biological Sciences, O
b Department of Vertebrate Zoology, National Museum of Natu
Received 17 March 2006; revised 27 S
Available online
Abstract1055-7903/$ - see front matter  2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.ympev.2006.10.022graphy of the swordtail ?sh
odontiformes, Poeciliidae)
s a, Natalie S. Dubois a, Kevin de Queiroz b
University, Athens, OH 45701, USA
History, Smithsonian Institution, Washington, DC 20560, USA
mber 2006; accepted 20 October 2006
ovember 2006
www.elsevier.com/locate/ympev
ion 43 (2007) 111?123
hylocontext. Morris et al. (2001) included several populations
of X. cortezi in their phylogenetic study of the northern
swordtails, and in some of their analyses, these populations
did not form a monophyletic group. However, relation-
ships among X. cortezi populations were not speci?cally
addressed or well supported by their allozyme data.
Xiphophorus cortezi is distributed throughout the south-
ern portion of the Pa?nuco river basin of eastern Mexico.
This river enters the Gulf of Mexico at Tampico and drains
the Sierra Madre Oriental. X. cortezi inhabits relatively still
pools in small streams with rocky bottoms but is missing
from larger rivers with sandy bottoms (Rauchenberger
et al., 1990). The physical distribution of X. cortezi in small
streams across three larger river drainages suggests that the
populations might have evolved independently from each
other. Because the rivers that connect the populations
become quite large and therefore are not suitable habitat
for this ?sh, the river habitats downstream may act as
physical and ecological barriers among the populations in
the headwater streams. Therefore, we expect X. cortezi
populations to be genetically di?erentiated. In addition,
some of the populations found in the headwaters of di?er-
ent rivers are geographically closer than the populations
found within the same river system. As the river systems
in this area are known to be relatively instable (Rauchen-
berger et al., 1990), we suspect that stream capture (i.e.,
the process by which a river or stream erodes through a
divide so that its ?ow is diverted into a neighboring drain-
age system) may have played a role in the current distribu-
tion of this species. The purpose of our study is to assess
the distribution of genetic variation within and among pop-
ulations of X. cortezi, establish the geographical relation-
ships among populations, test for cases of stream
capture, and test the hypothesis of monophyly for this
species.
2. Materials and methods
2.1. Samples
We collected 65 individuals from 10 di?erent sites locat-
ed in three drainages of the Pa?nuco River system in Mexi-
co, thereby sampling the known distribution of the species
(Fig. 1 and Table 1). A ?n-clip was collected from the cau-
dal ?n of each ?sh and preserved in salt-saturated 20%
dimethyl sulphoxide solution (Seutin et al., 1991). Fin-clips
from single individuals of swordtail species X. birchmanni,
X. malinche, X. montezumae and X. multilineatus from
Atlapexco, Soyatla, R??o Fr??o, and R??o Coy, respectively,
were included as presumptive outgroups for the phyloge-
netic analyses.
2.2. Ampli?cation, sequencing and alignment
Total DNA was extracted from ?n-clips using the
112 C. Gutie?rrez-Rodr??guez et al. / Molecular PDNeasy tissue kit (Qiagen Inc.) following the manufactur-
er?s instructions. The entire mitochondrial control regionwas ampli?ed in 25 or 50 ll reactions using the polymerase
chain reaction (PCR) and the previously published primers
K (50 AGCTCAGCGCCAGAGCGCCGGTCTTGTAAA
30) and G (50 CGTCGGATCCCATCTTCAGTGTTAT
GCTT 30) (Lee et al., 1995) according to standard methods.
Reactions were performed on a MJ Research PTC-100
thermocycler (MJ Research Inc.) under the following
conditions: 2 min at 94C, followed by 35 cycles at 90 or
94 C for 30 s, 55?63 C for 30 s and 72 C for 45 s, with
a ?nal extension period of 7 min at 72 C. PCR products
were puri?ed using the UltraClean PCR Clean-up
DNA Puri?cation Kit (MoBio Laboratories Inc.) or the
QIAquick Gel Extraction Kit (Qiagen Inc.) and sequenced
in both directions to check the validity of the sequence
data using the BigDye terminator cycle sequencing kit.
Sequences were visualized with an ABI 310 or 3730
automated sequencers (Applied Biosystems) at the DNA
Analysis Facility at Ohio University and the Plant-Microbe
Genomics Facility at Ohio State University respectively.
We aligned the forward and reverse sequences using the
computer software SeqMan II (DNAStar Inc.). Contigu-
ous sequences were initially aligned using Clustal X
(Thompson et al., 1997) with the default settings, followed
by manual alignment with the program Sequence Align-
ment Editor v2.0a11. All sequences were deposited in Gen-
bank (Table 1).
2.3. Population genetics analyses
All population genetic analyses were performed using
ARLEQUIN (Schneider et al., 2000). Levels of intra-pop-
ulation variation were estimated by calculating the mean
gene diversity (h), the probability that two randomly cho-
sen haplotypes are di?erent (Nei, 1987) and the nucleotide
diversity (p), the average number of di?erences between all
pairs of haplotypes (Tajima, 1983; Nei, 1987). An analysis
of molecular variance (AMOVA) (Exco?er et al., 1992)
was used to test whether sequences from previously de?ned
groups were signi?cantly di?erent from each other. We
de?ned groups corresponding to the three drainages (Tam-
peon, Moctezuma and Tempoal) from which the samples
were collected (Table 1). An AMOVA without de?ned
groups was also conducted. We used 16,000 permutations
to test for statistical signi?cance of both AMOVAs.
AMOVA compares the similarity within and among
groups using genetic distance as the measure of similarity.
We used the Tamura-Nei (TrN) model (Tamura and Nei,
1993) with the transition/transvertion (ti/tv) ratio = 3.0232
and the gamma shape parameter (a) = 0.8982 (see below)
to calculate the genetic distance between the di?erent
sequences. HKY, which is the model that best ?tted our
data (see below), is not an option in ARLEQUIN 2.0
and TrN is the closest available approximation to HKY.
We calculated the degree of genetic di?erentiation between
all the sampled populations with pairwise F-statistics (Weir
genetics and Evolution 43 (2007) 111?123and Cockerham, 1984) and their signi?cance by performing
1000 permutations.
hyloC. Gutie?rrez-Rodr??guez et al. / Molecular PTwo Mantel tests with 10,000 permutations were per-
formed to test for a correlation between genetic and geo-
graphic distance using the computer program IBD
(Bohonak, 2002). Genetic distances were calculated using
the method of Nei and Li (1979) and 1000 permutations
in both tests. Nei and Li?s measure of genetic distance
(1979) is an estimate of the number of nucleotide di?erenc-
es between populations that incorporates information on
both haplotype frequencies and the number of substitu-
tional di?erences between haplotypes. Geographic distanc-
es were measured using river distances (the paths along the
water courses connecting two localities) and straight-line
distances (the minimum great circle distance between two
localities). Both distances were calculated using the soft-
ware ExpertGPS 1.3.7 (Topogra?x, 2003).
Fig. 1. Map showing the di?erent collection sitesgenetics and Evolution 43 (2007) 111?123 1132.4. Relationships among haplotypes
We assessed genetic di?erences between pairs of haplo-
types using both the total number of base di?erences and
maximum-likelihood distances (HKY model, ti/tv ratio
and base frequencies estimated from the data, values for
proportion of invariant sites and gamma shape parameter
as obtained from MODELTEST software, see below). Dis-
tance calculations were performed using PAUP* 4.0b10
(Swo?ord, 2002).
We analyzed the phylogenetic relationships among the
haplotypes using a variety of inference methods. Phyloge-
netic trees of the X. cortezi and outgroup haplotypes were
inferred using PAUP* 4.0b10 (Swo?ord, 2002) under
parsimony, maximum-likelihood and least-squares distance
of X. cortezi (d) and four outgroup taxa (m).
dive
ean
hylocriteria. For the parsimony analysis, we conducted a
branch and bound search, treating gaps as a ???fth base??.
For the likelihood analyses, we selected a nucleotide substi-
tution model using hierarchical likelihood ratio tests as
implemented in MODELTEST 3.06 (Posada and Crandall,
1998). We estimated the optimal tree using a successive
approximation approach (Swo?ord et al., 1996; Sullivan
et al., 2005) with PAUP* 4.0b10 (Swo?ord, 2002). This
method consists of a series of successive likelihood analy-
ses, with the parameter values used in a given analysis esti-
mated on the optimal tree from the previous analysis and
reiterated until the topology, the likelihood score for the
tree, and the parameter values match those from the previ-
ous analysis. Each iteration used a heuristic search with
starting trees obtained by random stepwise addition (100
replicates) and TBR branch swapping. For the distance
analysis, we used unweighted least squares as the optimal-
ity criterion and maximum-likelihood (HKY85 + G + I)
distances, with the parameters estimated from the neigh-
Table 1
Populations of X. cortezi sampled in this study, measures of their genetic
Locality Species Abbreviation Drainage
Oxitipa X. cortezi OXI Tampao?n
Caldera X. cortezi CAL Tampao?n
Tambaque X. cortezi TAM Tampao?n
Tanute X. cortezi TAN Tampao?n
La Conchita X. cortezi CON Moctezuma
Amacuzac X. cortezi AMA Moctezuma
San Mart??n X. cortezi SAM Tempoal
Chalpuhuacanita X. cortezi CHA Tempoal
Tecolutlo X. cortezi TEC Tempoal
Xiliatl X. cortezi XIL Tempoal
Atlapexco X. birchmanni ATL Tempoal
Soyatla X. malinche SOY Tempoal
R??o Fr??o X. montezumae FRI Tampao?n
R??o Coy X. multilineatus COY Tampao?n
N = number of sequences (individuals), Nh = number of haplotypes, h = m
because only one sample was collected from the site.
114 C. Gutie?rrez-Rodr??guez et al. / Molecular Pbor-joining tree used in the hierarchical likelihood ratio
tests. Gaps were treated as missing data in a heuristic
search with starting trees obtained by random stepwise
addition (100 replicates) and TBR branch swapping. Nodal
support was assessed under the likelihood criterion using
non-parametric bootstrap resampling (1000 replicates)
and heuristic searches with starting trees obtained by ran-
dom stepwise addition (10 replicates) and TBR branch
swapping.
Relationships among the haplotypes were also inferred
using the statistical parsimony method of Templeton
et al. (1992). Unlike the previous methods, relationships
inferred by the statistical parsimony method are not con-
strained to take the form of a tree (minimally connected
graph), allowing them to take the form of more extensively
connected graphs in which additional connections repre-
sent alternative, equally parsimonious, mutational path-
ways from one haplotype to another. TCS v1.21
(Clement et al., 2000) was used to estimate the statistical
parsimony network (SPN), with the default 0.95 probabil-ity connection limit and treating gaps both as a ?fth state
and as missing data.
Because di?erent outgroups suggested two di?erent root
positions, we conducted three analyses to determine
whether our data could distinguish between the alternative
root placements with con?dence. First, we constrained the
X. cortezi plus X. malinche haplotypes to form a monophy-
letic group and used the Kishino-Hasegawa (KH; Kishino
and Hasegawa, 1989) and Shimodaira-Hasegawa (SH; Shi-
modaira and Hasegawa, 1999) tests to determine whether
the maximum-likelihood trees in the presence versus
absence of this constraint were signi?cantly di?erent in
their ability to explain the data. Second, we searched for
optimal trees (1) using only X. montezumae and X. birch-
manni as outgroups (excluding X. multilineatus) and (2)
using only X. mulitlineatus as the outgroup (excluding X.
montezumae and X. birchmanni). We compared the maxi-
mum-likelihood trees inferred in each of these analyses
(using the successive approximation approach described
rsity and GenBank Accession Nos.
N Nh h p GenBank Accession Nos.
3 1 0.000 0.0000 DQ445674
5 1 0.000 0.0000 DQ445674
5 1 0.000 0.0000 DQ445674
8 1 0.000 0.0000 DQ445674
12 2 0.530 0.0006 DQ445669-70
1 1 ? ? DQ445671
10 1 0.000 0.0000 DQ445671
10 2 0.467 0.0005 DQ445672-73
4 1 0.000 0.0000 DQ445675
7 2 0.476 0.0011 DQ445676-77
1 1 ? ? DQ445678
1 1 ? ? DQ445679
1 1 ? ? DQ445680
1 1 ? ? DQ445681
gene diversity, p = nucleotide diversity, ? = calculations are not possible
genetics and Evolution 43 (2007) 111?123above) with the trees using the same outgroup(s) attached
to the branch identi?ed as the root by the alternative out-
group(s) using the KH and SH tests. For these tests, we
used PAUP* 4.0b10 with the same model as in the maxi-
mum-likelihood tree searches, estimating the transition/
transversion ratios, the nucleotide base frequencies, the
proportion of variable sites and the shape parameter (a),
and using RELL bootstrap resampling (1000 replicates).
To assess the strength of the support for the derivation
of X. malinche from within X. cortezi, we constrained
X. cortezi to be monophyletic and, because of the di?erent
root positions suggested by the di?erent outgroups, we
conducted two separate analyses. We estimated maxi-
mum-likelihood trees using the successive approximations
approach described above with the same model, ?rst using
only X. montezumae and X. birchmanni as outgroups
(excluding X. multilineatus) and then using only X. multi-
lineatus as the outgroup (excluding X. montezumae and
X. birchmanni). We then compared the maximum-likeli-
hood trees inferred in these analyses with those inferred
for the same set of taxa in the absence of the monophyly
constraint using KH and SH tests. For these tests, we used
PAUP* 4.0b10 with the same model as in the maximum-
likelihood tree searches, estimating the transition/transver-
sion ratios, the nucleotide base frequencies, the proportion
of variable sites and the shape parameter (a), and using
RELL bootstrap resampling (1000 replicates).
3. Results
3.1. Sequence variation
We obtained 881 bp sequences for the mitochondrial
DNA control region of 65 X. cortezi individuals from 10
populations as well as of four individuals of closely related
species. For the X. cortezi and four outgroup haplotypes
(69?72) could not be aligned unambiguously, we retained
both the most widespread (four localities) and the most
common (21 individuals). The remainder of the haplotypes
were only shared by individuals from the same locality
(Table 2). Three of the populations (Chalpuhuacanita, La
Conchita, Xiliatl) exhibited two di?erent haplotypes; all
other populations exhibited only a single haplotype. Mean
gene diversity (h) was very low to moderate among X. cor-
tezi populations (Table 1). Nucleotide diversity (p) was
very low in all populations, ranging from 0 to 0.0011
(Table 1) indicating a high degree of similarity of sequences
within populations.
The analysis of variance, in which we speci?ed three dif-
ferent groups based on river drainages (Tampao?n, Moc-
tezuma and Tempoal), showed signi?cant genetic structure
at all hierarchical levels. Most of the variation (53.71%)
was explained by di?erences among the three drainages. Dif-
ferences among the populations within each drainage were
ur p
5
X. cortezi CON 5 7
C. Gutie?rrez-Rodr??guez et al. / Molecular Phylogenetics and Evolution 43 (2007) 111?123 115X. cortezi AMA 1
X. cortezi SAM 10
X. cortezi CHA 7 3
X. cortezi TEC
X. cortezi XIL
X. birchmanni ATL
X. malinche SOY
X. montezumae FRI
X. multilineatus COYthem in our analyses. Three of the haplotypes (H7, H8,
and H9) di?er only in the ambiguously aligned region yet
are clearly di?erent from one another. We thought that it
was preferable to retain the information present in the
identi?cation of three distinct haplotypes even though the
speci?c substitutions separating those haplotypes are
ambiguous. In any case, analyses excluding these positions
yielded qualitatively similar results (not presented).
3.2. Genetic diversity and population structure
Nine haplotypes were identi?ed among the 65 sequenced
X. cortezi individuals and four among the outgroup taxa
(Table 2). Haplotypes H3 and H6 were shared by individ-
uals from di?erent populations, and haplotype H6 was
Table 2
Number and geographic distribution of haplotypes from X. cortezi and fo
Species Location Haplotype
H1 H2 H3 H4 H
X. cortezi OXI
X. cortezi CAL
X. cortezi TAM
X. cortezi TANtogether, 67 variable sites out of the 881 were detected
(7.6%) and for the X. cortezi haplotypes alone, 16 variable
sites were detected (1.8%). The minimum number of muta-
tions for the X. cortezi haplotypes alone was 16 (none of
the variable sites for the nine haplotypes observed in this
species had more than two bases). Although four sitesTotal 5 7 11 7 3also large and explained 42.07% of the variance. Only
4.22% of the variance was attributed to di?erences among
individuals within populations (Table 3). The AMOVA per-
formed without groupings also showed high and signi?cant
degrees of genetic structure (Table 3). Most of the variation
(95.03%) was explained by di?erences among populations,
and only 4.97% by di?erences within populations.
Genetic di?erentiation between pairs of populations is
presented in Table 4. The majority of the FST values were
high (P0.4232) and signi?cantly di?erent from zero. The
main exceptions were the comparisons between pairs of
populations located in the Tampao?n drainage, which were
all zero and thus not signi?cant. FST values between popu-
lations from di?erent drainages were generally higher than
values between populations from the same drainage,
although San Mart??n and Chalpuhuacanita (Tempoal
drainage) exhibited low values in comparison to La Conch-
ita (Moctezuma drainage) relative to other between-drain-
age comparisons and relatively high values in comparison
with Tecolutlo and Xiliatl (Tempoal drainage) relative to
other within-drainage comparisons.
utative outgroup species
H6 H7 H8 H9 H10 H11 H12 H13
3
5
5
8
4
5 2
1
1
1
121 4 5 2 1 1 1 1
uares Variance components Percentage of variation
Among groups 2 99.700 1.82654(Va) 53.71*
Among populations within groups 7 57.428 1.43058(Vb) 42.07*
0.14362(Vc) 4.22*
3.40074
2.74359(Va) 95.03*
0.14362(Vb) 4.97
etic distances (above the diagonal) among X. cortezi populations
CON AMA SAM CHA TEC
19.7771 19.1938 19.1938 22.5261 31.323
19.7771 19.1938 19.1938 22.5261 31.323
19.7771 19.1938 19.1938 22.5261 31.323
19.7771 19.1938 19.1938 22.5261 31.323
2.2833 2.5833 5.3156 18.7771
0.6657 0 3.3323 20.1938
0.8186* 0.0000 3.3323 20.1938
0.7809* 0.6412 0.8206* 22.9261
0.9381* 1.0000 1.0000* 0.9554*
* * * *
hylogenetics and Evolution 43 (2007) 111?123Within populations 55 7.899
Total 64 165.028
AMOVA without grouping
Among populations 9 157.129
Within populations 55 7.899
Total 64 165.028
* Signi?cant values at P < 0.05.
Table 4
Pairwise FST comparisons (below the diagonal) and Nei and Li (1979) gen
OXI CAL TAM TAN
OXI 0 0 0
CAL 0.0000 0 0
TAM 0.0000 0.0000 0
TAN 0.0000 0.0000 0.0000
CON 0.9421* 0.9497* 0.9497* 0.9578*
AMA 1.0000 1.0000 1.0000 1.0000
SAM 1.0000* 1.0000* 1.0000* 1.0000*
CHA 0.9487* 0.9565* 0.9565* 0.9644*
TEC 1.0000* 1.0000* 1.0000* 1.0000*
* * * *Table 3
Hierarchical analysis of molecular variance (AMOVA) for X. cortezi
Source of variation df Sum of sq
AMOVA with drainage grouping
116 C. Gutie?rrez-Rodr??guez et al. / Molecular PMantel tests revealed signi?cant and positive correla-
tions between genetic distance and river distance
(Fig. 2A, r = 0.766, P = 0.0004) as well as between genetic
distance and straight-line geographic distance (Fig. 2B,
r = 0.769, P = 0.0003). All population comparisons fell
within the 95% con?dence intervals for both regressions,
although eight comparisons in the analysis using river dis-
tances were much further than the others from the line of
best ?t. Four of these represent pairs of populations whose
genetic distances are substantially greater than expected
given their geographic distances; they involve comparisons
between San Mart??n and Chalpuhuacanita, and compari-
sons between Tecolutlo and Xiliatl, all from the Tempoal
drainage. The other four comparisons represent pairs of
populations whose genetic distances were substantially
smaller than expected given their geographic distances;
they involve comparisons between San Mart??n and Chal-
puhuacanita from the Tempoal drainage and La Conchita
and Amacuzac from the Moctezuma drainage.
3.3. Relationships among haplotypes
Hierarchical likelihood ratio tests identi?ed
HKY85 + I + G with the following initial parameter esti-
mates as the appropriate model for our data: transition/
transversion (ti/tv) ratio = 3.0323, proportion of invariable
XIL 0.9301 0.9437 0.9437 0.9563 0.9031 0.8727 0.9486 0.9190 0.4232
The signi?cance of the FST values between R??o Amacuzac and the rest of populations is not reported because only a single individual was sampled from the
R??o Amacuzac.
* P < 0.05 signi?cance level.
Fig. 2. Plots of (A) Nei and Li (1979) genetic distance compared to river
distance for all possible pairs of populations, (B) Nei and Li (1979) genetic
distance compared to straight-line geographic distance for the same pairs
of populations. The regression line and the lines that bound the 95%
con?dence intervals are shown and were calculated only for the X. cortezi
data. Comparisons between all X. cortezi populations to the X. malinche
population are included and are denoted with a di?erent symbol.
sites (i) = 0.8015, gamma shape parameter (a) = 0.8982,
and base frequencies A = 0.3091, C = 0.2274, G = 0.1391,
T = 0.3243.
The optimal trees derived from parsimony, distance and
likelihood analyses were generally similar, although the
strict consensus of the maximum-likelihood trees exhibited
greater resolution (8/11 possible nodes) than those derived
from distance (7/11) and parsimony (6/11) analyses. The
maximum-likelihood tree with nodal support (bootstrap)
values is presented in Fig. 3. The only group present in this
tree that is contradicted by the results of analyses based on
other optimality criteria is H3, H4, H6, H10 and H12. In
the parsimony tree H3 and H4 group with H5, while in
the least squares distance tree they group with H1, H2
and H5. None of these con?icting relationships is strongly
supported (bootstrap proportions all 655%). Details of the
results of all three analyses are presented in Appendix A.
In all optimal trees inferred by likelihood and distance
analyses and some of those inferred by parsimony analysis,
the trees cannot be rooted so that the haplotypes found in
X. cortezi (and X. malinche) form a monophyletic group
because di?erent outgroups suggest di?erent root positions
(Fig. 3). The X. mulitlineatus haplotype roots the tree
between the group H7?H9 + H11 (found in X. cortezi
and X. malinche populations from the Tempoal river drain-
age) and the remaining haplotypes (Fig. 3A); whereas the
X. birchmanni plus X. montezumae haplotypes root the tree
Tampao?n drainage) (Fig. 3B). The ?rst root position
(Fig. 3A) is inferred in those parsimony trees that can be
rooted without disrupting monophyly of the X. cortezi
(and X. malinche) haplotypes. Regardless of root position,
the X. malinche haplotype (H11) falls out within those of
X. cortezi and is most closely related to haplotypes from
the Tempoal drainage (H7?9).
In the trees resulting from all of these analyses, the
X. cortezi haplotypes fall into groups that correspond
roughly to the river drainages in which they are found.
All populations from the Tampao?n drainage exhibit a sin-
gle haplotype (H6). The haplotypes from the Moctezuma
drainage (H1?3) form a para- or polyphyletic group rela-
tive to those from the Tempoal drainage, or the Tempoal
and Tampao?n drainages, depending on the position of
the root. Finally, some of the haplotypes from the Tempoal
drainage (H7?9) form a monophyletic group while the oth-
ers (H3?5) are intermixed with haplotypes from the Moc-
tezuma drainage (H1?3). Branch lengths leading to the
various X. cortezi haplotypes are generally short, and only
four groups in the maximum-likelihood tree are supported
by bootstrap values >50% (Fig. 3).
The statistical parsimony networks (Fig. 4) postulate a
number of hypothetical unsampled haplotyes (which di?er
depending on whether gaps are treated as a ?fth state or as
missing data) in addition to the nine X. cortezi and one X.
malinche sampled haplotypes that were connected with
l reg
ite.
C. Gutie?rrez-Rodr??guez et al. / Molecular Phylogenetics and Evolution 43 (2007) 111?123 117on the branch to H6 (found in four populations from the
Fig. 3. Maximum-likelihood bootstrap tree for mitochondrial DNA contro
13). Branch lengths are proportional to the number of substitutions per s
Names of the outgroup species and drainage name for the X. cortezi haplotyp
Rooted with the outgroups X. montezumae and X. birchmanni.probabilitiesP0.95 (ignoring the outgroup X. multilineatus
ion haplotypes from X. cortezi (H1?9) and four putative outgroups (H10?
Numbers above branches indicate bootstrap support if greater than 50%.
es are also indicated. (A) Rooted with the outgroup X. multilineatus. (B)
in the network in which gaps were treated as missing data,
Fig. 4B). Outgroup weights (Castelloe and Templeton,
1994) based on both haplotype frequencies and positions
in the graph (internal versus terminal) suggest that either
haplotype H4, found in the Chalpuhuacanita population,
or H1, found in the La Conchita population, is ancestral
(Fig. 4), depending on the interpretation of gaps.
The X. cortezi haplotypes form three groups of closely
connected haplotypes (1?2 mutations between nearest
neighbors) that are separated from one another by more
distant connections (at least ?ve mutations when treating
gaps as a ?fth state). One contains haplotypes from the
Tampao?n drainage (H6), a second has haplotypes from
the Moctezuma and Tempoal drainages (H1?5), and a
third is formed by haplotypes from the Tempoal drainage
(H7?9). The haplotype from X. malinche (H11) is posi-
tioned in the networks 1 or 2 mutational steps from X. cor-
tezi haplotype (H9) from the Tempoal drainage. In
contrast, the X. birchmanni (H10) and X. montezumae
(H12) haplotypes, and, when gaps are treated as a ?fth
base, that of X. multilineatus (H13), are not linked to the
networks, indicating that the number of nucleotide di?er-
ences between those Xiphophorus species and X. cortezi
exceeds the connection limit of 12 steps, which corresponds
cant di?erences. The maximum-likelihood tree in the
absence of any constraints did not exhibit a signi?cantly
better ?t to the data than the maximum-likelihood tree
constrained so that the X. cortezi plus X. malinche haplo-
types formed a monophyletic group according to both
the KH test (P = 0.552) and the SH test (P = 0.309). This
constrained tree (not shown) was rooted in the position
indicated by X. multilineatus (see Fig. 3A). In addition,
the maximum-likelihood tree including only X. montezu-
mae and X. birchmanni as outgroups did not exhibit a sig-
ni?cantly better ?t to the data than a tree in which these
two outgroups were attached in the position where X. mul-
tilineatus was attached in the analysis including all three
outgroups (KH test, P = 0.532; SH test, P = 0.246). Simi-
larly, the maximum-likelihood tree including only X. multi-
lineatus as the outgroup did not exhibit a signi?cantly
better ?t to the data than a tree in which X. multilineatus
was attached in the position where X. montezumae and
X. birchmanni were attached in the analysis including all
three outgroups (KH test, P = 0.156; SH test, P = 0.097).
Although derivation of X. malinche from within X. cor-
tezi (paraphyly of X. cortezi relative to X. malinche) is
implied by the optimal trees from phylogenetic analyses
under diverse optimality criteria (Fig. 3 and Appendix
mit
rec
118 C. Gutie?rrez-Rodr??guez et al. / Molecular Phylogenetics and Evolution 43 (2007) 111?123to a probability of 0.95 that the number of mutations is
accurately estimated by parsimony. For the X. cortezi hap-
lotypes alone, 19 mutations were estimated by both tree
and network methods that treated gaps as a ?fth state
(20 when the X. malinche haplotype is included).
Tests of the alternative root placements suggested by
di?erent outgroups (Fig. 3A and B) did not reveal signi?-
Fig. 4. Statistical parsimony network based on X. cortezi sequences of the
treating gaps as missing data. Haplotype designations are from Table 1. The
rectangles is proportional to the haplotype?s frequency. Small black circles are h
network. Each of the lines between observed and/or hypothetical haplotypesA), our data do not reject the alternative hypothesis of
monophyly for the X. cortezi haplotypes. Comparison of
the maximum-likelihood trees in the absence of any con-
straints to the maximum-likelihood trees under a con-
straint of X. cortezi monophyly did not reject the
hypothesis of X. cortezi monophyly whether X. montezu-
mae and X. birchmanni were used as outgroups (KH test,
ochondrial DNA control region. (A) Treating gaps as a ?fth state and (B)
tangles represent the inferred ancestral haplotype. The size of the ovals and
ypothetical haplotypes necessary to connect the observed haplotypes in the
represents one mutational step.
hyloP = 0.532; SH test, P = 0.246) or X. multilineatus was used
as the outgroup (KH test, P = 0.783; SH test, P = 0.649).
4. Discussion
4.1. Genetic variation
The results of this study indicate a low degree of genetic
variation in the mtDNA control region of X. cortezi com-
pared to other teleostean ?shes (i.e., Fajen and Breden,
1992; Lee et al., 1995; Salzburger et al., 2003; Stefanni
and Thorley, 2003; Aboim et al., 2005). Out of 881 aligned
positions 16 (1.8%) were variable (the numbers are the
same if the X. malinche sample is included) and out of 65
sequences nine (13.8%) distinct haplotypes were found
(10 if the X. malinche sample is included). Gene diversity
(h) was low to moderate and nucleotide diversity (p) was
very low in X. cortezi populations. Although the number
of genotypes found in each population should be interpret-
ed with caution because of the limited sample sizes, similar
low levels of within population variation in the mtDNA
control region have been found in other poeciliids, such
as the guppy species Poecilia reticulata (Fajen and Breden,
1992; Carvalho et al., 1996; Shaw et al., unpublished data)
and the swordtail species X. birchmanni (Shearer et al.,
unpublished data).
The low genetic diversity within populations may re?ect
founder e?ects, as reductions in genetic polymorphisms
have been observed in introduced populations of guppies
(Shaw et al., 1992; Carvalho et al., 1996). Alternatively,
the low values of nucleotide diversity (p) combined with
moderate haplotype diversity (h) may suggest recent popu-
lation expansion after a bottleneck or a founder event.
However, a Tajima?s D-test (Tajima, 1989) indicated this
is not the case in X. cortezi, as all populations had positive
D values that were not signi?cantly di?erent from zero
(results not shown). These patterns of low genetic diversity
within population are most likely the consequence of con-
sistently small population sizes, which might have resulted
in genetic drift and inbreeding (Nei et al., 1975; Wishard
et al., 1984). Small population size could be a consequence
of human activities, as there are indications of human con-
tamination in these areas that are likely to contribute to
?sh mortality.
In contrast to the low variation within populations, we
found high degrees of genetic di?erentiation among X. cor-
tezi populations. The AMOVA with drainage grouping
showed that the variation among populations within drain-
ages explained a large proportion of the total variance,
indicating genetic structure among populations located in
the same drainage. This was supported by the signi?cance
of the AMOVA performed without groupings and by the
high and signi?cant values of the FST comparisons between
most pairs of populations. High levels of genetic di?erenti-
ation among populations have been reported in other
C. Gutie?rrez-Rodr??guez et al. / Molecular Pfreshwater ?shes (Carvalho et al., 1991; Shaw et al., 1991,
1994; Fajen and Breden, 1992; Alves and Coelho, 1994;Coelho et al., 1997; Ha?n?ing and Brandl, 1998a,b,c; Mesq-
uita et al., 2001) and could be the result of discontinuities in
suitable habitat. As a result, genetic di?erentiation might
re?ect the physical subdivision of populations (Carvalho,
1993). Sections of the streams where X. cortezi and other
swordtails species are distributed dry seasonally and may
be contaminated by human activities, resulting in destruc-
tion of suitable habitat. In addition, X. cortezi only occurs
in small rivers, which are often more subdivided by barriers
to ?sh movement than larger rivers.
The populations from the Tampao?n drainage (Oxitipa,
Caldera, Tambaque and Tanute) constitute an exception
to the pattern of high genetic di?erentiation among popu-
lations from the same drainage. All the populations in
Tampao?n shared the same single haplotype (H6). Such
genetic homogeneity among the sampling sites in this
drainage suggests recent and/or historical migration
throughout the streams of the drainage and could re?ect
the absence of barriers to gene ?ow. Alternatively, it could
re?ect the geographic proximity of the collection sites in
this drainage. The geographic distances between the popu-
lations sampled from the Tampao?n drainage are, in gener-
al, smaller than those between populations in the other
drainages and do not include stretches of larger rivers with
unsuitable X. cortezi habitat. The patterns of low intra-
population genetic diversity combined with high genetic
di?erentiation among populations found in this investiga-
tion have also been reported in studies of other freshwater
?shes based on allozymes (Carvalho et al., 1991; Shaw
et al., 1994; Ha?n?ing and Brandl, 1998b,c) and mtDNA
(Carvalho et al., 1996; Mesquita et al., 2001; Shaw et al.,
unpublished data).
4.2. Geographic patterns of genetic variation and evidence of
stream capture
Our data exhibited signi?cant correlations between
genetic and geographic distances, regardless of whether
geographic distance was measured along watercourses or
as straight lines. These ?ndings suggest that genetic di?er-
entiation among X. cortezi populations can be explained,
to some extent, by an isolation-by-distance model. A
strong geographic e?ect on genetic variation is also implied
by the results of the AMOVA with drainage grouping, in
which a signi?cant amount of the total genetic variation
was attributable to di?erences among groups of popula-
tions from di?erent drainages. On the other hand, there
does not appear to be a one-to-one correspondence
between groups of related haplotypes and drainage groups.
Neither the haplotype trees nor the haplotype networks
exhibit groups that correspond precisely to the three drain-
age groups. In both cases, some haplotypes from the Tem-
poal drainage (H4?5) are closer to those from the
Moctezuma drainage (H1?3), in terms of both common
ancestry and mutational steps, than they are to other hap-
genetics and Evolution 43 (2007) 111?123 119lotypes from the Tempoal drainage (H7?9). Moreover, one
of the haplotypes (H3) is found in both drainages.
hyloSeveral of our results provide evidence of stream cap-
ture. First, the Mantel tests revealed that genetic distance
between populations is more strongly correlated with
straight-line geographic distance than with geographic dis-
tance measured along river courses, suggesting that genetic
di?erentiation is more strongly in?uenced by gene ?ow
between populations in adjacent drainages (as might occur
through stream capture) than by gene ?ow up and down
river courses. Second, in the comparison of genetic distanc-
es versus river distances, the eight most obvious outliers all
represent comparisons involving populations from the
northwestern part of the Tempoal drainage. The popula-
tions in this drainage (represented in our study by San
Mart??n and Chalpuhacanita) exhibit substantially larger
genetic distances to the other populations in the Tempoal
drainage (Tecolutlo and Xiliatl) and substantially smaller
genetic distances to the populations in the Moctezuma
drainage (La Conchita and Amacuzac) than expected given
their river course geographic distances suggesting that
these two streams may have drained into the Moctezuma
in the past. Third, in both the haplotype trees and the
statistical parsimony networks, the haplotypes from San
Mart??n (H3) and Chalpuhuacanita (H4 and H5) from the
Tempoal drainage are more closely related to haplotypes
from the Moctezuma drainage (H1, H2 and H3, which is
shared with San Mart??n) than they are to other haplotypes
from the Tempoal drainage (H7, H8, and H9). The above
?ndings, in conjunction with the fact that the northwestern
branches of the Tempoal drainage are located geographi-
cally between the Moctezuma drainage and the southeast-
ern branches of the Tempoal drainage, are consistent with
a scenario in which the current northwestern branches of
the Tempoal drainage were formerly part of the Moc-
tezuma drainage and were later captured by the Tempoal
drainage. The low elevation (less than 100 m) separating
parts of the drainages also suggests the possibility of stream
capture.
4.3. Root of the haplotype tree and ancestral haplotype
Although tests designed to establish the placement of
the root of the haplotype tree could not conclusively distin-
guish between the alternatives, two lines of evidence favor
the root position indicated by the outgroup X. multilineatus
over that suggested by X. birchmanni and X. montezumae.
First, some of the optimal trees under parsimony as well
as the maximum-likelihood tree under the constraint of
monophyly of the X. cortezi plus X. malinche haplotypes
(trees not shown) were rooted in the position (branch) sug-
gested by X. multilineatus on the maximum-likelihood tree
in the absence of any constraints. Second, trees rooted
using X. birchmanni and X. montezumae as outgroups were
far from approaching signi?cance in distinguishing
between alternative root positions (P = 0.246?0.532), sug-
gesting that these taxa are not particularly informative
120 C. Gutie?rrez-Rodr??guez et al. / Molecular Pfor discriminating between the two alternatives, while trees
rooted using only X. multilineatus as the outgroup werecloser to approaching signi?cance in favoring the root posi-
tion implied by that species over the position suggested by
X. birchmanni and X. montezumae (P = 0.097?0.156).
The statistical parsimony networks indicated that either
haplotype H4 (when gaps treated as ?fth state) or H1
(when gaps treated as missing) is ancestral among the hap-
lotypes from X. cortezi and X. malinche, an inference based
on both the relative haplotype frequencies and numbers of
connections to other haplotypes (Castelloe and Templeton,
1994; Clement et al., 2000). If the root position is assumed
to occur at the node at which X. multilineatus attaches to
the tree (see above), then the optimal phylogenetic tree sug-
gests that the ancestral haplotype is located along the
branch separating X. cortezi haplotypes H7?9 from haplo-
types H1?6, which is only a few mutational steps from the
haplotypes identi?ed as ancestral by the analyses using
TCS, particularly when gaps are treated as missing data
(H1). In the results of that analysis (Fig. 4B) the outgroup
X. multilineatus (H13) connects to the rest of the network
in the same area. In addition to using di?erent kinds of
information to infer the ancestral haplotype (position and
frequency versus outgroup) the minor di?erences between
the results of these analyses may be attributable to the fact
that the TCS analyses only identify observed haplotypes as
ancestral while the phylogenetic analyses can identify an
inferred haplotype as ancestral.
4.4. Monophyly versus paraphyly of X. cortezi haplotypes
and the status of X. malinche
The phylogenetic analyses suggest that X. cortezi is not
monophyletic. Regardless of the root position, X. cortezi
haplotypes from the Tempoal drainage (H7?9) appear to
be more closely related to the X. malinche haplotype (H11)
than they are to otherX. cortezi haplotypes (H1?6). In addi-
tion, the number of substitutions separating the X. malinche
haplotype from the X. cortezi haplotypes are within the
range for comparisons involving the X. cortezi haplotypes
alone. Moreover, our sample of X. malinche (from Soyatla)
and the haplotypes of X. cortezi to which it appears most
closely related (from Xiliatl and Tecolutlo) all come from
the Tempoal drainage. Although our data cannot reject the
alternative hypothesis that the X. cortezi haplotypes form a
monophyletic group exclusive of X. malinche, based on the
phylogenetic relationships and geographic proximity of the
X. malinche haplotypes to the X. cortezi haplotypes from
Tecolutlo and Xiliatl we suspect that these haplotypes
(and the populations in which they occur) are closely related
and that failure to demonstrate this relationship with a high
degree of con?dence is a result of the limited power of the
tests resulting from the small number of mutational
di?erences between the haplotypes in question. The use of
a more polymorphic marker would be necessary to
conclusively resolve these relationships.
The ?nding that X. cortezi may be paraphyletic relative
genetics and Evolution 43 (2007) 111?123to X. malinche raises several issues concerning the status of
these taxa as separate species. One possibility is that the
two taxa are separate species, with X. cortezi ancestral to X.
malinche. Alternatively, X. malinche may not be a separate
species but rather a group of X. cortezi populations that
has evolved some distinctive features. This second possibil-
ity is supported by the fact that the genetic versus geo-
graphic distance comparisons between the X. malinche
population and the various X. cortezi populations lie very
close to (and below rather than above) the best ?tting line
representing the comparisons among the X. cortezi popula-
tions (see Fig. 2). A third possibility is that the taxon cur-
rently recognized as X. cortezi consists of several di?erent
species (from which X. malinche may or may not be specif-
ically distinct). Although a large component of the total
genetic variation is attributable to di?erentiation between
the populations from di?erent drainages, the di?erences
also conform closely to an isolation-by-distance model,
related outgroup (X. montezumae) and the entire set of
haplotypes from X. cortezi and X. malinche. Moreover, in
both the haplotype trees and networks, the X. cortezi and
X. malinche haplotypes form a tight group whose members
are separated by relatively short branches. In contrast, the
X. birchmanni haplotype connects to the trees via a much
longer branch and is so divergent that is not connected
to the statistical parsimony network of X. cortezi and X.
malinche haplotypes.
As noted above, there is ambiguity concerning the root
of the haplotype trees and the relationships of X. birchman-
ni and X. montezumae to the rest of the taxa in our study
(see Section 4 of alternative root positions). However, if
the tree root position indicated by the outgroup X. multi-
lineatus is correct, then the X. malinche and X. birchmanni
haplotypes are most closely related to di?erent X. cortezi
haplotypes (H7?9 versus H6, respectively). This result is
KY
11,
77
42
42
55
11
55
37
C. Gutie?rrez-Rodr??guez et al. / Molecular Phylogenetics and Evolution 43 (2007) 111?123 121suggesting the possibility of ongoing gene ?ow. Moreover,
genetic divergence within X. cortezi (Table 5) is not nearly
as great as divergence between X. cortezi and populations
that clearly represent di?erent species. Nonetheless, it is
possible that sets of populations in the di?erent drainages
are no longer exchanging genes but have not been isolated
long enough for greater di?erentiation to have occurred.
4.5. Relationship of X. cortezi and X. malinche to
X. birchmanni
Contrary to previous ?ndings (Rauchenberger et al.,
1990), which placed X. malinche and X. birchmanni as sister
species, with X. cortezi as their sister group, the results of
our phylogenetic analyses suggest that X. malinche is more
closely related to X. cortezi than to X. birchmanni. In our
phylogenetic trees and networks, the haplotype from X.
malinche (H11) is very closely related to the X. cortezi hap-
lotypes in general, and to those from the southeastern
branches of the Tempoal drainage (H7?9) in particular
(see previous section). In contrast, the X. birchmanni haplo-
type connects to the phylogenetic trees between a distantly
Table 5
Pairwise total base di?erences (above diagonal) and maximum-likelihood (H
in X. cortezi (H1?H10) and four outgroup species (H10, X. birchmanni; H
H1 H2 H3 H4 H5 H6
H1 ? 1 1 2 1 7
H2 0.00115 ? 2 3 2 8
H3 0.00116 0.00235 ? 1 2 6
H4 0.00233 0.00356 0.00114 ? 1 7
H5 0.00114 0.00233 0.00233 0.00116 ? 8
H6 0.00849 0.00988 0.00715 0.00841 0.00979 ?
H7 0.00713 0.00847 0.00848 0.00976 0.00839 0.013
H8 0.00838 0.00975 0.00976 0.01106 0.00966 0.012
H9 0.00590 0.00721 0.00721 0.00848 0.00714 0.012
H10 0.05526 0.05818 0.05253 0.05062 0.05338 0.052
H11 0.00469 0.00597 0.00597 0.00721 0.00590 0.011
H12 0.05214 0.05502 0.04681 0.04495 0.04755 0.047
H13 0.01850 0.02022 0.02025 0.02176 0.02000 0.023Estimates for the number of mutational di?erences between pairs of haplotypes
multiplying the maximum-likelihood distances (number of substitutions/site) bcongruent with the results obtained from several di?erent
analyses of allozyme data for the northern swordtails,
which place X. malinche and X. birchmanni populations
closest to di?erent populations of X. cortezi (Morris
et al., 2001; esp. Fig. 4). Alternatively, if the root position
indicated by the outgroup X. montezumae is correct, then
the X. malinche haplotypes are once again closest to some
of the X. cortezi haplotypes from the Tempoal drainage
(H7?9), but the X. birchmanni haplotype is outside of a
monophyletic group composed of the X. malinche and all
the X. cortezi haplotypes. This result is congruent with
those of a previous analysis of Xiphophorus DNA sequence
data (Meyer et al., 1994), which placed either X. birchmanni
(when analyzed by neighbor-joining) or a group composed
of X. birchmanni and X. pygmaeus (when analyzed by par-
simony) as the sister group to all the rest of the northern
swordtails.
Sequence divergence corroborates the results obtained
from the phylogenetic analyses (Table 5). Maximum-likeli-
hood (HKY + I + G) distances between the X. cortezi and
X. malinche haplotypes (0.0012?.0111) are substantially
+ I + G model) distances (below diagonal) for mitochondrial haplotypes
X. malinche; H12, X. montezumae; H13, X. multilineatus)
H7 H8 H9 H10 H11 H12 H13
6 7 5 34 4 32 14
7 8 6 35 5 33 15
7 8 6 33 5 30 15
8 9 7 32 6 29 16
7 8 6 33 5 30 15
11 10 10 33 9 30 17
? 1 1 31 2 34 14
0.00114 ? 2 32 3 33 13
0.00114 0.00231 ? 30 1 33 13
0.04938 0.05121 0.04746 ? 30 37 38
0.00231 0.00349 0.00115 0.04753 ? 31 12
0.05611 0.05400 0.05410 0.06692 0.04941 ? 36
0.01849 0.01701 0.01704 0.06770 0.01563 0.06414 ?that correct for multiple substitutions at individual sites can be obtained by
y the number of sites (in this case 881).
is the sister to a group composed of X. malinche and X. cor-
(Research Incentive) to MRM. Collecting methods comply
The branch and bound search found 12 optimal trees of
Marked genetic divergence revealed by allozymes among populations
variance inferred from metric distances among DNA haplotypes:
hylolength 86, the strict consensus of which had the following
topology:
(H13,H6,(H4,H5,H3),H1,H2,(((H7,H8),H9),H11),
(H10,H12)).
A.2. Likelihood analyses
Of 881 characters, 49 distinct character patterns
were observed. The successive approximation heuristic
search required three iterations, and although it identi?ed
two trees as maximum-likelihood estimates of the
phylogeny under ?xed parameter values, only one of these
trees (Fig. 2) was identi?ed as optimal when the parameters
were estimated for that tree (ln L = 1632.18538). The
parameter estimates from the ?nal iteration of the
successive approximation analysis on the optimal tree were
ti/tv = 3.663520, i = 0.783835, and a = 0.792371,with the Animal Care Guidelines of Ohio University (Ani-
mal Care and Use approval No. L01-01).
Appendix A. Results of phylogenetic analyses
A.1. Parsimony analyses
Of 881 unordered characters, 814 were constant, 67 were
variable, and 25 were informative (42 uninformative) under
the parsimony criterion with gaps treated as a ?fth state.tezi. Regardless, of whether these three species form a
clade, X. birchmanni seems to be much more distantly relat-
ed to X. cortezi and X. malinche than these two taxa are to
each other.
Acknowledgments
We are grateful to the Mexican government for permis-
sion to collect the ?shes. We would like to thank Oscar
Rios-Cardenas and Abby Darrah for their help in the ?eld.
This research was supported by funding from the National
Science Foundation (IBN 9983561) and Ohio Universitysmaller than those between both the X. cortezi and X.
birchmanni haplotypes (0.0475?0.0582) and the X. malinche
and X. birchmanni haplotypes (0.0475). In contrast, dis-
tances between the X. cortezi and X. malinche haplotypes
and the haplotypes from X. birchmanni (0.0475?0.0582)
are similar to the distances of the former two taxa from
one relatively distantly related outgroup (X. montezumae
0.0450?0.0561), and smaller than those from another (X.
multilineatus 0.0156?0.0234).
Because of limited sampling of other swordtail species
and ambiguities concerning the position of the root, our
results cannot rule out the possibility that X. birchmanni
122 C. Gutie?rrez-Rodr??guez et al. / Molecular Pfrequency of A = 0.308990, C = 0.227496, G = 0.138244,
T = 0.325269. The strict consensus of the two trees identi-application to human mitochondrial DNA restriction data. Genetics
131, 479?491.
Fajen, A., Breden, F., 1992. Mitochondrial DNA sequence variation
among natural populations of the Trinidad guppy, Poecilia reticulata.
Evolution 46, 1457?1465.
Ha?n?ing, B., Brandl, R., 1998a. Genetic and morphological variation in aof the guppy Poecilia reticulata (Poeciliidae), in Trinidad. Biol. J. Linn.
Soc. 57, 219?234.
Carvalho, G.R., 1993. Evolutionary aspects of ?sh distribution: genetic
variability and adaptation. J. Fish Biol. 43, 53?73.
Carvalho, G.R., Shaw, P.W., Hauser, L., 1996. Arti?cial introductions,
evolutionary change and population di?erentiation in Trinidadian
guppies (Poecilia reticulata: Poeciliidae). Biol. J. Linnean Soc. 57, 219?
234.
Castelloe, J., Templeton, A.R., 1994. Root probabilities for intraspeci?c
gene trees under neutral coalescent theory. Mol. Phylogenet. Evol. 3,
102?113.
Clement, M., Posada, D., Crandall, K.A., 2000. TCS: a computer
program to estimate gene genealogies. Mol. Ecol. 9, 1657?1659.
Coelho, M.M., Alves, J., Rodrigues, E., 1997. Patterns of genetic
divergence in Chondrostoma lusitanicum Collares-Pereira, in intermit-
tent Portuguese rivers. Fish Manage. Ecol. 4, 223?232.
Exco?er, L., Smouse, P., Quattro, J., 1992. Analysis of molecular?ed as optimal under ?xed parameter estimates had the fol-
lowing topology:
(H13,(((H6,(H10,H12)),H4,H3),H5,H1,H2),(((H7,H8),
H9),H11)).
A.3. Distance analyses
Using HKY85 + G + I distances (with the same param-
eter estimates used in the ?rst iteration of the likelihood
analysis) and an unweighted least squares optimality crite-
rion, the heuristic search found 12 optimal trees with a
score of 0.00026. The strict consensus of the 12 trees had
the following topology:
(H13,((H6,(H10,H12)),(H4,H5,H1,H2,H3)),((H7,H8),-
H9),H11). In the individual optimal trees, the haplotype
from Xiphophorus malinche (H11) always grouped with
one group (H1?6) or the other (H7?9) of the X. cortezi hap-
lotypes, resulting in an unresolved position in the strict
consensus tree.
References
Aboim, M.A., Menezes, G.M., Schlitt, T., Rogers, A.D., 2005. Genetic
structure and history of populations of the deep-sea ?sh Helicolenus
dactylopterus (Delaroche, 1809) inferred form mtDNA sequence
analysis. Mol. Ecol. 14, 1343?1354.
Alves, J., Coelho, M.M., 1994. Genetic variation and population
subdivision of the endangered Iberian cyprinid Chondrostoma lusitan-
icum. J. Fish Biol. 44, 627?636.
Bohonak, A.J., 2002. IBD (Isolation by Distance): a program for
population genetic analyses of isolation by distance. J. Hered. 93,
153?154.
Borowski, R.L., McClelland, M., Cheng, R., Welsh, J., 1995. Arbitrarily
primed DNA ?ngerprinting for phylogenetic reconstruction in verte-
brates: the Xiphophorus model. Mol. Biol. Evol. 12, 1022?1032.
Carvalho, G.R., Shaw, P.W., Magurran, A.E., Sehers, B.H., 1991.
genetics and Evolution 43 (2007) 111?123common European cryprinid, Leuciscus cepahlus within and across
Central European drainages. J. Fish Biol. 52, 706?715.
Ha?n?ing, B., Brandl, R., 1998b. Genetic variability, population size and
isolation of distinct populations in the freshwater ?sh Cottus gobio L.
Mol. Ecol. 7, 1625?1632.
Ha?n?ing, B., Brandl, R., 1998c. Genetic di?erentiation of the bullhead
Cottus gobio L. across watersheds in Central Europe: evidence for two
taxa. Heredity 80, 110?117.
Kishino, H., Hasegawa, M., 1989. Evaluation of the maximum likelihood
estimate of the evolutionary tree topologies from DNA sequence
data, and the branching order in Hominoidea. J. Mol. Evol. 29, 170?
179.
Lee, W-J., Conroy, J., Howell, W.H., Kocher, T.D., 1995. Structure and
evolution of teleost micochondrial control regions. J. Mol. Evol. 41,
Schneider, S., Roessli, D., Exco?er, L., 2000. ARLEQUIN, Version 2.00:
a Software for Population Genetics Data Analysis. University of
Geneva, Geneva.
Seutin, G., White, B.N., Boag, P.T., 1991. Preservation of avian blood and
tissue samples for DNA analyses. Can. J. Zool. 69, 82?92.
Shaw, P.W., Carvalho, G.R., Magurran, A.E., Seghers, B.H., 1991.
Population di?erentiation in Trinidad guppies (Poecilia reticulata):
patterns and problems. J. Fish Biol. 39 (suppl. A), 203?209.
Shaw, P.W., Carvalho, G.R., Seghers, B.H., Magurran, A.E., 1992.
Genetic consequences of an arti?cial introduction of guppies (Poecilia
reticulata) in N. Trinidad. Proc. R. Soc. B. 248, 111?116.
C. Gutie?rrez-Rodr??guez et al. / Molecular Phylogenetics and Evolution 43 (2007) 111?123 12354?66.
Me?e, G.K., Snelson, F.F., 1989. An ecological overview of poeciliid
?shes. In: Me?e, G.K., Snelson, F.F. (Eds.), Ecology and Evolution of
Livebearing Fishes (Poeciliidae). Prentice Hall, Englewood Cli?s, pp.
13?31.
Mesquita, M., Carvalho, G., Shaw, P., Crespo, E., Coelho, M.M., 2001.
River basin-related genetic structuring in an endangered ?sh species,
Chondrostoma lusitanicum, based on mtDNA sequencing and RFLP
analysis. Heredity 86, 253?264.
Meyer, A., Morrissey, J., Shartl, M., 1994. Recurrent evolution of a
sexually selected trait in Xiphophorus ?shes inferred from a molecular
phylogeny. Nature 368, 539?542.
Meyer, A., 1997. The evolution of sexually selected traits in male swordtail
?shes (Xiphophorus: Poeciliidae). Heredity 79, 329?337.
Moritz, C., 2002. Strategies to protect biological diversity and the
evolutionary process that sustain it. Syst. Biol. 51, 238?254.
Morris, M.R., de Queiroz, K., Morizot, D.C., 2001. Phylogenetic
relationships among populations of northern swordtails (Xiphophor-
ous) as inferred form allozyme data. Copeia 1, 65?81.
Nei, M., Maruyama, T., Chakraborty, R., 1975. The bottleneck e?ect and
genetic variability in populations. Evolution 29, 1?10.
Nei, M., Li, W.H., 1979. Mathematical model for studying genetic
variation in terms of restriction endonucleases. Proc. Natl. Acad. Sci.
USA 76, 5269?5273.
Nei, M., 1987. Molecular Evolutionary Genetics. Columbia University
Press, New York.
Posada, D., Crandall, K.A., 1998. Modeltest: testing the model of DNA
substitution. Bioinformatics 14, 817?818.
Rauchenberger, M., Kallman, K.D., Morizot, D.C., 1990. Monophyly
and geography of the R??o Pa?nuco basin swordtails (Genus Xipho-
phorus) with descriptions of four new species. Am. Mus. Nov. 2975,
1?41.
Rosen, D.E., 1979. Fishes from the uplands and intermontane of
Guatemala: reversionary studies and comparative geography. Bull.
Am. Mus. Nat. Hist. 162, 268?375.
Ryan, M.J., Rosenthal, G.G., 2001. Variation and selection in swordtails.
In: Dugatkin, L.A. (Ed.), Model Systems in Behavioral Ecology.
Princeton University Press, New Jersey, pp. 133?148.
Salzburger, W., Branda?stter, A., Gilles, A., et al., 2003. Phylogeography
of the vairone (Leuciscus sou?a, Risso 1826) in Central Europe. Mol.
Ecol. 12, 2371?2386.Shaw, P.W., Carvalho, G.R., Magurran, A.E., Seghers, B.H., 1994.
Factors a?ecting the distriution of genetic variability in the guppy,
Poecilia reticulata. J. Fish Biol. 45, 875?888.
Shimodaira, H., Hasegawa, M., 1999. Multiple comparisons of log-
likelihoods with applications to phylogenetic inference. Mol. Biol.
Evol. 16, 1114?1116.
Stefanni, S., Thorley, J., 2003. Mitochondrial DNA phylogeography
reveals the existence of an evlutionary signi?cant unit of the sand goby
Pomatoschistus minutus in the Adriatic (Eastern Mediterranean). Mol.
Phylogenet. Evol. 28, 601?609.
Sullivan, J., Abdo, Z., Joyce, P., Swo?ord, D.L., 2005. Evaluating the
performance of a successive approximations approach to parameter
optimization in maximum-likelihood phylogeny estimation. Mol. Biol.
Evol. 22, 1386?1392.
Swo?ord, D.L. 2002. PAUP*. Phylogenetic Analysis Using Parsimony
(*and Other Methods). Version 4b10. Sinauer Associates, Sunderland,
MA.
Swo?ord, D.L., Olsen, G.J., Waddell, P.J., Hillis, D.M., 1996. Phyloge-
netic inference. In: Hillis, D.M., Moritz, C., Mable, B.K. (Eds.),
Molecular Systematics. Sinauer Associates, MA, pp. 407?514.
Tajima, F., 1983. Evolutionary relationship of DNA sequences in ?nite
populations. Genetics 105, 437?460.
Tajima, F., 1989. Statistical method for testing the neutral mutation
hypothesis by DNA polymorphism. Genetics 123, 585?595.
Tamura, K., Nei, M., 1993. Estimation of the number of nucleotide
substitutions in the control region of mitochondrial DNA in humans
and chimpanzees. Mol. Biol. Evol. 10, 512?526.
Templeton, A.R., Crandall, K.A., Sing, C.F., 1992. A cladistic analysis of
phenotypic associations with haplotypes inferred from restriction
endonuclease mapping and DNA sequence data. III. Cladogram
estimation. Genetics 132, 619?633.
Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins,
D.G., 1997. The ClustalX windows interface: ?exible strategies for
multiple sequence alignments aided by quality analysis tools. Nucleic
Acids Res. 24, 4876?4882.
Topogra?x 2003. ExpertGPS, Copyright 2000?2006, TopoGra?x, Stow
MA.
Weir, B.S., Cockerham, C., 1984. Estimating F-statistics for the analysis of
population structure. Evolution 38, 1358?1370.
Wishard, L.N., Seeb, J.E., Utter, F.M., Stefan, D., 1984. A genetic
investigation of suspected redband trout populations. Copeia 1984,
120?132.