Molecular Phylogenetics and Evolution 55 (2010) 631-639

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

ELSEVIER journal homepage: www.elsevier.com/locate/ympev

Molecular phylogenetic relationships and the evolution of the placenta in

Poecilia (Micropoecilia) (Poeciliidae: Cyprinodontiformes)
Robert W. Meredith, Marcelo N. Pires, David N. Reznick, Mark S. Springer *
Department of Biology, University of California, Riverside, CA 92521, USA

ARTICLE INFO ABSTRACT

Article history: Poeciliids are one of the most intensively studied groups within Cyprinodontiformes owing to their use as
Received 12 August 2009 model organisms for experimental studies on natural and sexual selection, and comparative studies of
Revised 6 November 2009 life-history evolution. Life-history studies have demonstrated multiple origins of placentotrophy and
Accepted 9 November 2009 superfetation in poeciliids, including the recent description of placentotrophy in three species of Poecilia
Available online 14 November 2009
(Micropoecilia): P. bifurca, P. branneri, and P. parae. Here, we use a concatenation of seven nuclear gene
Keywords:
segments and two mitochondrial segments to examine relationships within Micropoecilia and between
Cyprinodontiformes this subgenus and other subgenera in Poecilia (Mollienesia, Limia, Pamphorichthys, Acanthophacelus). The
Poeciliidae combined molecular data set (8668 bp) was analyzed with maximum parsimony, maximum likelihood,
Poecilia. and Bayesian methods. We also employed a relaxed molecular clock method to estimate divergence
Micropoecilia times within Poecilia. All phylogenetic analyses with the combined DNA data set supported the mono-
Acanthophacelus phyly of Poecilia and recovered a basal split between Poecilia (Acanthophacelus) * Poecilia (Micropoecilia)
Placentotrophy and the other three subgenera. Within Micropoecilia,P. bifurca grouped with P. branneri, and these joined
P. parae to the exclusion of P. picta. Ancestral reconstructions based on parsimony and Bayesian methods
suggest that placentotrophy evolved once in Micropoecilia in the common ancestor of P. bifurca, P. bran-
neri, and P. parae. Divergence time estimates suggest that placentotrophy in Micropoecilia evolved in <4
million years.
© 2009 Elsevier Inc. All rights reserved.

1. Introduction et al., 2002, 2007). Most recently, placentotrophy was reported in

all populations of P. parae, P. bifurca and P. branneri that were
The family Poeciliidae (Rosen and Bailey, 1963), equivalent to investigated (Pires, 2007; Pires et al in press). In every case, pla-
the subfamily Poeciliinae of Parenti (1981), is a diverse group of centotrophy was documented by a significant increase in embryo
freshwater fishes in the order Cyprinodontiformes that comprises dry mass during development, and accompanied by a thickened
7220 species in 28 genera (Lucinda, 2003; Lucinda and Reis, maternal follicle and enlarged and externalized pericardial mem-
2005). The group includes well-known aquarium fishes such as brane in the embryo (Pires et al in press).
guppies, mollies, and swordtails (Lucinda and Reis, 2005). Poecili- Micropoecilia was proposed as a new genus by Hubbs (1926),
ids are endemic to the New World and are widely distributed who included M. bifurca, M. parae, M. picta, M. melanozona, and
throughout the Americas and Caribbean. All poeciliids have inter- M. branneri, but also indicated that he was not convinced of the dis-
nal fertilization and all except Tomeurus are viviparous (Regan, tinctness of any of the species and suggested that all may eventu-
1913; Rosen and Gordon, 1953; Rosen and Bailey, 1963). Poeci s ally be united with M. parae. Rosen and Bailey (1963), in their
are characterized by the presence of a gonopodium that is formed classic monograph on the family Poeciliidae (-Poeciliinae of Paren-
by modified anal-fin rays 3, 4, and 5 (Lucinda and Reis, 2005). ti, 1981), described the genus Poecilia and its constituent subgener-
Members of the group are frequently used as model organisms a: Poecilia, Lebistes, Pamphorichthys, and Limia (Table 1). Rosen and
for experimental studies on natural and sexual selection (Endler, Bailey (1963) included Poecilia reticulata (common guppy), the
1983; Houde, 1997; Schluter et al., 1998; Hamilton, 2001), and genus Micropoecilia of Hubbs (1926), Poecilia amazonica, and
comparative studies of life-history evolution that have demon- Poecilia scalpridens in the subgenus Lebistes. Micropoedilia parae,
strated multiple origins of superfetation and placentotrophy M. bifurca, and M. melanzona of Hubbs (1926) were treated as three
(Grove and Wourms, 1991, 1994; Arias and Reznick, 2000; Reznick polymorphs of Poecilia (Lebistes) parae by Rosen and Bailey (1963).
Costa (1991) reallocated Poecilia (Lebistes) scalpridens to the
* Corresponding author. Fax: +1 909 787 4826. subgenus Pamphorichthys. Meyer (1993) reinstated the genus
E-mail address: mark [email protected] (M.S. Springer). Micropoecilia of Hubbs (1926) for Poecilia parae, P. bifurca, P. picta,
1055-7903/$ - see front matter © 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.ympev.2009.11.006.
632 RW. Meredith et aL /Molecular Phylogenetics and Evolution 55 (2010) 631-639
Table 1
Previous classifications of Poecilia species relevant to this study.
Hubbs (1924, 1926) Rosen and Bailey (1963) Meyer, (1993) Poeser et al. (2005) Lucinda and Reis (2005)
Family Poeciliidae Family Poeciliidae Family Poec Family Poeciliidae
Subfamily Poeci Subfamily Poeci Subfamily P. 2 Subfamily Poeci
Tribe Poeciliini Tribe Po; Tribe Poeciliini Tribe Poeciliini
Genus Limia Genus Poecilia Genus Poecilia Genus Poecilia Genus Poecilia
Genus Micropoecilia P. (Poecilia) caucana P. (Lebistes) reticulata P. (Acanthophacelus) Genus Limia
reticulata
M. bifurca P. (Poecilia) latipunctata P. (Pamphorichthys) P. (Acanthophacelus) wingei Genus Micropoecilia
M. picta P. (Lebistes) reticulata Genus Micropoecilia P. (Micropoecilia) parae Genus
Pamphorichthys
M. branneri P. (Lebistes) parae [included M. bifurca P. (Micropoecilia) branneri
Micropoecilia bifurca]
M. parae P. (Lebistes) picta M. picta P. (Micropoecilia) bifurca
M. melanzona P. (Lebistes) branneri M. branneri P. (Micropoecilia) picta
Genus Lebistes P. (Pamphorichthys) minor. M. parae P. (Micropoecilia) minima
L. reticulatus P. (Limia) dominicensis
Genus Mollienisia* P. (Limia) melanogaster
Genus Allopoecilia
A. caucana
Tribe Pamphoriini
Genus
Pamphorichthys
* Hubbs (1924, 1926) used Mollienisia rather than Mollienesia. Bailey and Miller (1950) subsequently provided justification for using Mollienesia instead of Mollienisia.

P. amazonica and P. branneri based on three derived morphological ophacelus rather than Lebistes because the latter name is tied to a
characters: (1) variegated males and females; most specimens molly, not the guppy. and is not available as a subgenus name
with a relatively well-developed dark humeral spot on the side for P. reticulata (Poeser and Isbrücker, 2002). Acanthophacelus
of the body; (2) males with anal-fin rays 6 and 7 and ventral fins Eigenmann, 1907 then becomes the first available subgenus name
extending to the base of the gonopodium palp; and (3) ray 3 of for P. reticulata (Poeser and Isbrücker, 2002). Shories et al. (2009)
the gonopodium with 10-16 rose thorn-like spines. Within Micro- described Poecilia (Acanthophacelus) obscura, which represents a
poecilia, Meyer (1993) also proposed synapomorphic characters for third species in the subgenus Acanthophacelus. Poeser et al.
M. bifurca and M. picta, these two plus M. parae, and these three (2005) also placed all of Meyer's (1993) Micropoecilia species and
plus M. amazonica. Meyer (1993) also defined Poecilia (Lebistes) the newly described Poecilia (Lebistes) minima of Costa and Sarraf
as a monotypic subgenus that included only P. reticulata. Meyer's (1997) into Poecilia (Micropoecilia) (see Table 1).
(1993) M. amazonica is of questionable taxonomic value because In the present study we use a combined mitochondrial and nu-
its type series is composed of a combination of M. branneri and clear DNA data set to examine relationships among species in Poe-
M. parae individuals (CA. Figueiredo, pers. comm. to M. Pires). ser et al.'s (2005) subgenus Micropoecilia (bifurca, branneri, parae,
Subsequent to Meyer (1993), Costa and Sarraf (1997) described picta). Our study includes all of the placentotrophic Micropoecilia
Poecilia (Lebistes) minima and hypothesized a close relationship of species reported by Pires et al. (in press), as well as representatives
this taxon to P. branneri based on five morphological synapomor- of most of the other major subdivisions of Poecilia. This taxon sam-
phies. Costa and Sarraf (1997) also challenged Meyer's (1993) pling provides a basis for (1) evaluating the monophyly of Poecilia
putative synapomorphies for Micropoecilia and argued that (1) (Micropoecilia) (2) determining the number of origins of placento-
the variegated pattern does not occur in both sexes, (2) the hum- trophy in Poecilia (Micropoecilia), and (3) examining relationships
eral spot is not present in all species, and (3) the morphological among the major lineages within the genus Poecilia.
features of the anal-fin rays and third gonopodial ray are also pres-
ent in other groups. Given these differences with Meyer (1993),
Costa and Sarraf (1997) did not concur with Meyer's (1993) rein- 2. Methods and materials
statement of Micropoecilia and questioned the monophyly of this
genus. 2.1. Taxon sampling
Breden et al. (1999) examined relationships among representa-
tives of the major divisions of the genus Poecilia using mitochon- We included all recognized species of Poecilia (Micropoecilia) ex-
drial ND2 sequences. Breden et al. (1999) included two species of cept forP. minima. We also included two representatives each from
Micropoecilia (sensu Meyer, 1993), M. picta and M. parae, which the Poecilia subgenera Acanthophacelus, Limia, Mollienesia, and
grouped together and were the sister taxon to Poecilia reticulata. Pamphorichthys. Two species of Cnesterodon were included as out-
Similarly, Hrbek et al. (2007) and Lucinda and Reis (2005) found groups. A previous molecular study that included both mitochon-
support for an association ofP. reticulata with representative spe- drial and nuclear loci demonstrated that Cnesterodon is the
cies of Micropoecilia based on a multigene concatenation and a ma- closest outgroup to Poecilia (Hrbek et al., 2007). All ingroup and
trix of osteological and soft anatomical characters, respectively. outgroup taxa are listed in Supplementary Information Table 1.
Poeser et al. (2005) described Poecilia wingei (the Endler's live-
bearer, Campoma guppy. or Cumaná guppy) a close relative of Poe- 22. Gene sequences
cilia (Lebistes) reticulata, and placed both species in the subgenus
Acanthophacelus rather than Lebistes. Poeser et al. (2005) argued QUIAGEN DNeasy Tissue extraction kits were used to extract
that apparent autapomorphies of P. reticulata were instead synapo- genomic DNA from the skeletal muscle of 95% ethanol-preserved
morphies ofP. reticulata and P. wingei, and grouped these species in specimens. Two mitochondrial and seven nuclear gene regions
the subgenus Acanthophacelus. Poeser et al. (2005) chose Acanth- were amplified. The first mitochondrial segment included the 3'
 RW. Meredith et al./ Molecular Phylogenetics and Evolution 55 (2010) 631-639 633

end of tRNA^", the complete cytochrome b (cytb) gene, and the 5' (Glyt); TrN+1 (SH3PX3, myh6, Rh, 2nd mitochondrial codon posi-
end of tRNA™. The second segment consisted of the 3' end of tions); GTR+I+I" (Ragl, NADH2, 3rd mitochondrial codon posi-
tRNA*'^, all of tRNAV**, the complete NADH dehydrogenase sub- tions); KS0«I-I^ (X-src exons); TVM+I (X-src introns, 1st and 2nd
unit 2 gene (NADH2), all of tRNAT™, all of tRNA^", and the 5' end nuclear codon positions); TVM++T" (cytb); TrN*IXT^ (exons);
of tRNA^*^, The nuclear regions included two partial exons (8 and TVM+I+I" (mitochondrial protein-coding); K81uf+l+I" (tRNAs);
10), all of exon 9, and two introns (8 and 9) of the tyrosine kinase TVM-T (3rd nuclear codon positions); GTR*T (1st mitochondrial
gene (X-src); exon 1 of myosin, heavy polypeptide 6 (myh6); exon codon positions). Both the bootstrap compatibility method and
2 of ectodermal-neural cortex 1 like protein (ENC1); exon 2 of gly- the partition homogeneity test (four partitions, p - 0.673; eight
cosyltransferase (Glyt); exon 1 of SH3 and PX domain containing 3 partitions, p=0.500; 11 partitions, p - 0.820) suggested that it
(SH3PX3); a portion of the 7 transmembrane receptor region of was appropriate to combine the individual segments into one
rhodopsin (Rh); and exon 3 of recombination activating gene-1 multigene data set.
(Rag1). The nuclear markers Rh, X-src, myh6, ENCI, Glyt, SH3PX3,
and Ragl are located on chromosomes 8, 17(?), 20, 5, 16, 25, and 2.4, Phylogenetic analyses
25, respectively, of the zebrafish Danio rerio.
Sequences for most PCR primers used in this study have previ- PAUP 4.0b10 (Swofford, 2002), RAXML 7.0.4 (Stamatakis, 2006),
ously been published. New internal and external primers were de- and MrBayes V3.1.1 (Huelsenbeck and Ronquist, 2001; Ronquist
signed as necessary to amplify problematic taxa. Sequences for and Huelsenbeck, 2003) were used to perform maximum parsi-
new primers, previously used primers, and primer combinations mony (MP), maximum likelihood (ML), and Bayesian analyses,
used in amplifications are provided in Supplementary Information respectively, on three data sets: (1) the combined data set that in-
Table 2. Previously described primers used in this study include cluded seven nuclear segments and two mitochondrial segments
those for X-src (Meyer and Lydeard, 1993); MYH6, ENC1, Glyt, (length- 8668bp); (2) the combined nuclear data set
and SH3PX3 (Li et al, 2007); Rh (Chen et al 2003); Rag1 and (length= 6159 bp); and (3) the combined mitochondrial data set
NADH2 (Hrbek et al., 2007; Kocher et al., 1995; Ptacek and Breden, (length = 2509 bp). Gaps were treated as missing data in all analy-
1998; Breden et al., 1999); and cytb (Schmidt et al., 1998). Nested ses. Branch and bound analyses were employed to find the shortest
PCRs were carried out following the protocols outlined in Li et al. tree(s) in MP analyses. MP bootstrap analyses (1000 replications)
(2007) for myh6, ENC1, Glyt, and SH3PX3. All other genes were employed heuristic searches with 1000 randomized taxon addition
amplified using the following protocol: initial denaturation at sequences and tree-bisection and reconnection branch swapping.
94 °C for 2 min; 35 cycles of 1 min at 94 °C (denaturation), 1 min ML and Bayesian analyses were performed with the three parti-
at 50 °C (annealing), and 1 min at 72 °C (extension); and a final tioning schemes outlined above with models suggested by Model-
extension of 10 min at 72 °C. In all cases PCR products were run test 3.06 (Posada and Crandall, 1998). ML analyses with RAXML
out on a 1% agarose gel and the product of interest was then ex- employed 500 replicates, randomized MP starting trees, and the
cised and cleaned using QIAGEN QlAquick PCR purification kits. fast hill-climbing algorithm; all other free parameters were esti-
Products were then sequenced in both directions at the University mated. MrBayes v3.1.1 (Huelsenbeck and Ronquist, 2001; Ronquist
of California Riverside's Core Genetics Institute, which uses an and Huelsenbeck, 2003) was used to calculate Bayesian posterior
automated DNA sequencer (ABI 3730xl). Sequencing primers were probabilities MrBayes carries out Metropolis-coupled Markov
designed as necessary. All accession numbers for both the 136 new chain Monte Carlo sampling. Bayesian analyses employed eight
and 17 previously published sequences are given in Supplementary Markov chains (seven hot, one cold), with random starting trees,
Information Table 1. default priors, and chain sampling every 1000 generations. Analy-
ses were terminated once the average standard deviation of split
2.3. DNA alignments and data compatibility frequencies for the simultaneous analyses fell below 0.01. If the
model suggested by Modeltest 3.06 was not available in MrBayes,
We used the program SOAP v1.2a4 (Lóytynoja and Milinkovitch, the next most complex model was used.
2001), with gap opening (11-19) and gap extension (3-11) penal-
ties in steps of two, to identify alignment-ambiguous regions. 2.5. Molecular dating analyses
These settings correspond to those that were used by Gatesy
et al. (1993). Alignment-ambiguous regions were excluded from The likelihood ratio statistic rejected the molecular clock
phylogenetic analyses. The appropriateness of combining the indi- (p « 0.05) for 11 of the 19 possible partitions (cytb, NADH2, mito-
vidual gene segments into a combined data set was tested using (1) chondrial protein-coding genes, ENC1, Ragl, X-src exons, nuclear
the partition homogeneity test (Farris et al., 1994; Swofford, 2002) exons, nuclear 2nd codon positions, nuclear 3rd codon positions,
with each combination of partitions outlined below, 1000 replica- mitochondrial 1st codon positions, mitochondrial 3rd codon posi-
tions, and 100 taxon input orders per replicates, and (2) the boot- tions). Therefore, we employed the relaxed molecular clock meth-
strap compatibility method (de Queiroz, 1993; Teeling et al., 2000) od implemented in BEAST ver 1.4.8 (Drummond et al., 2006;
with 500 bootstrap replicates and a 90* bootstrap support crite- Drummond and Rambaut, 2007), which allows for complex models
rion. ML bootstrap compatibility tests allowed each segment to of evolution and "soft" node constraints (Hedges and Kumar, 2004;
have its own model of sequence evolution as suggested by the Yang and Rannala, 2006). We used the uncorrelated lognormal dis-
Akaike Information Criterion of Modeltest 3.06 (Posada and Crand- tribution (UCLN) model, which draws the rate of each lineage inde-
all, 1998). We used three different schemes to partition the data as pendently from a lognormal distribution. Models of molecular
follows: (1) two nuclear partitions (exons and introns) and two evolution were from Modeltest 3.06 (Posada and Crandall, 1998).
mitochondrial partitions (protein-coding and tRNA); (2) four nu- In cases where the model selected by Modeltest was not available
clear partitions (1st codon positions, 2nd codon positions, 3rd co- in BEAST (i.e., models with three to five categories for substitution
don positions, introns) and four mitochondrial partitions (1st rates), we used the next most general model (i.e., GTR). For each
codon positions, 2nd codon positions, 3rd codon positions, tRNA); partition scheme, three independent analyses were run for ten mil-
and (3) eight nuclear partitions (one for coding sequences from lion generations. The three runs were combined using Log-Com-
each of seven different genes and one for X-src introns) and three biner; Tracer 1.4 (Rambaut and Drummond, 2003) was used to
mitochondrial partitions (NADH2, cytb, tRNA genes). Models for inspect for stationarity/mixing and to confirm that the estimated
individuals partitions were as follows: K80+l (ENC1); HKY«T sample size for each parameter was greater than 200. Given the
634 RW. Meredith et al./ Molecular Phylogenetics and Evolution 55 (2010) 631-639

lack of a fossil record for taxa included in our analyses, we used a parameter prior was specified with x = 1, which specifies an unin-
secondary constraint taken from Hrbek et al. (2007), who reported formative prior with equal prior probabilities. The number of
a minimum of 19.92 mya and a maximum of 24.39 mya for most draws from each prior distribution was set at ten.
recent common ancestor of Pamphorichthys, Limia, and Mollienesia.
We employed a soft constraint, which followed a normal distribu- 3. Results
tion, with 95% of the normal distribution between the specified
minimum and maximum given in Hrbek et al. (2007). 3.1. Sequence alignments and sequence variability
2.6. Ancestral state reconstructions The combined alignment included 8668 bp after excluding
alignment-ambiguous regions of tRNA (6 bp) and X-src (15 bp).
Parsimony and SIMMAP Version 1.0 B2.3.2 (Bollback, 2006) The nuclear and mitochondrial components of the combined data
were used to estimate ancestral states for lecithotrophy versus pla- set were 6159 and 2509 bp in length, respectively. The nuclear
centotrophy. Terminal taxa included all species of Acanthophacelus data included 371 (=6.0%) parsimony-informative sites [52
and Micropoecilia that were included in our phylogenetic analyses. (=2.6%) at 1st codon positions; 24 (1.2%) at 2nd codon positions;
The relative amounts of pre- and post-fertilization maternal provi- 266 (=13.4%) at 3rd codon positions; and 29 (=12.8%) at intronic
sioning to embryos have been quantified for all of these taxa (Pires positions]. The mitochondrial data included 774 (=30.8%) parsi-
et al., in press) and indicate that P. reticulata, P. wingei and P. picta mony-informative sites [164 (=22.5%) at 1st codon positions; 41
embryos lost weight during development and that all or most of (=5.6%) at second codon positions; 538 (=73.8%) at 3rd codon posi-
the resources necessary for embryonic growth were provided by tions; and 32 (=9.8%) at tRNA positions]. Among coding sequences
females in the form of yolk (lecithotrophy); in contrast, P. parae, for individual nuclear genes, parsimony-informative sites ranged
P. bifurca, and P. branneri females exhibited extensive post-fertil- from 4.7% (Glyt and SH3PX3) to 7% (Ragl). Rescaled consistency
ization resource allocation to embryos via a follicular placenta (RC) and retention index (RI) values were both higher for the nu-
(placentotrophy). clear data (0.65, 0.81) than the mitochondrial data (0.30, 0.58). Ta-
SIMMAP implements the procedures of Nielsen (2002) and ble 2 summarizes information on partition length, the number of
Huelsenbeck et al. (2003) for stochastically mapping mutations constant and variable (parsimony-uninformative and parsimony-
on phylogenies. SIMMAP uses a fully Bayesian approach with Mar- informative) characters in each partition, the contribution to the
kov chain Monte Carlo sampling to calculate a posterior probability length of the most-parsimonious tree for the combined data set
distribution that accommodates uncertainty in ancestral states, (see below) that derives from each partition, RC, and RI. The num-
evolutionary rates, and the phylogeny. We used all post burnin ber of changes per site was also calculated for different genes/par-
Bayesian trees (5957 trees) from the combined MrBayes analysis titions and ranged from 0.02 (2nd codon positions of nuclear
(11 partitions). SIMMAP employs a I distribution prior on the genes) to 2.60 (3rd codon positions of mitochondrial genes)
overall substitution rate of the morphological character. The (Table 2).
parameters of the I^ distribution are x and f, and these parameters
describe the mean (z/f)) and variance (/) of the distribution. We 3. Phylogenetic analyses
used 60 discrete categories to describe the T^ distribution of the
rate prior. For two-state morphological characters SIMMAP uses Fig. 1 shows the single most-parsimonious tree (3325 steps)
an additional prior—the bias parameter. The bias parameter is a that was recovered with the combined data set. RAXML analyses
symmetrical beta distribution that is described by a single param- with four, eight, or 11 partitions resulted in ML trees that were
eter, o, and was approximated with 19 categories. Branch length topologically identical to each other (Fig. 2). The ML trees differed
proportionality was maintained by rescaling branch lengths before from the MP tree in grouping Poecilia (Limia) and Poecilia (Pampho-
the priors were applied. Ancestral state reconstructions were per- richthys) to the exclusion of Poecilia ( Mollienesia). Fig. 2 also shows
formed with three different combinations of parameters for the I^ mean Bayesian posterior probabilities (BPP) based on two indepen-
distribution prior: x - 1, f=1; =3, f -2; and x - 5, f = 5. The bias dent runs above the branches, and ML Bootstrap Support Percent-
Table2
Parsimony statistics for different partitions.
Partition Partition Number of Number of variable, Numberof Rescaled —— Retention Number ofsteps on — Number of changes
length —— consant — uninformative informative consistency — index globally most- per site on globally
characters — characters characters — index parsimonious tree — most-parsimonious tree
Nuclear 6159 5565 223 37 065 081 787 013
Nuclear 1st codon 1975 1893 30 52 069 085 106 0.05
Nuclear 2nd codon 1976 1943 9 24 073 088 a1 0.02
Nuclear 3rd codon 1981 1544 171 266 064 0.80 579 029
Introns 227 185 13 29 057 076 61 027
ENC1 847 784 21 a2 0.70 0.86. 82 010
Glyt 886 808 36 42 075 085 91 010
SH3PX3 724 667 23 34 058 0.76 78 011
MYH6 767 692 30 45 068 083 97 013
Ragl 1561 1384 67 110 059 076 241 015
Rh 822 749 24 49 068 0.86 98 012
Xesrc 552 481 22 49 0.60 078 100 018
X-src exon. 325 296 9 20 065 081 39 0.12
Mitochondrial 2509 1543 192 774 030 058 2538 1.01
Mitochondrial 1st codon 729 524 a1 164 036 064 4 0.60
Mitochondrial 2nd codon 729 656 32 a1 054 072 109 0.15
Mitochondrial 3rd codon 729 101 90 538 027 056 1892 2.60
Mitochondrial tRNAs 322 262 29 31 046 066 96 0.30
Cytb 1140 702 7 367 028 057 1210 1.06
NADH2 1047 579 92 376 031 059 1232 118
 RW. Meredith et al./ Molecular Phylogenetics and Evolution 55 (2010) 631-639 635

Cnesterodon decemmaculatus

Cnesterodon hypselurus

Poecilia (Limia) dominicensis

Poecilia (Limia) melanogaster

Poecilia (Mollienesia) caucana

Poecilia (Mollienesia) latipunctata.

Poecilia (Pamphorichthys) araguaiensis

Poecilia (Pamphorichthys) minor

Poecilia (Micropoecilia) bifurca

Poecilia (Micropoecilia) branneri

Poecilia (Micropoecilia) parae French Guyana

l: Poecilia (Micropoecilia) parae Suriname

Poecilia (Micropoecilia) picta Venezuela

L Poecilia (Micropoecilia) picta Trinidad

Poecilia (Acanthophacelus) reticulata

Poecilia (Acanthophacelus) wingei

100 changes [ Poecilia (Acanthophacelus) wingei Venezuela
Fig. 1. The shortest tree (3326 steps) for the combined data set based on analyses with PAUP 4.0b10 (Swofford, 2002).

ages (BSPs) below the branches for analyses with 11 partitions. All cluded Poecilia (Micropoecilia) picta from a clade containing all
of the topological relationships shown in Fig. 2 were also recovered other Poecilia (Micropoecilia) species plus Poecilia (Acanthophace-
in Bayesian analyses of the combined data set with four or eight lus). The mitochondrial MP tree differed from the ML and MPP
partitions. Table 3 summarizes bootstrap support percentages mitochondrial trees in recovering Pamphorichthys as the sister tax-
based on MP and ML analyses, and BPP values based on Bayesian on to all other ingroup taxa.
analyses with four, eight, and 11 partitions. Two synapomorphic indels were discovered. First, a six bp dele-
All species, subgenera, and genera were recovered as monophy- tion in Glyt (3469-3474 in concatenation) is shared by all species
letic (100% BSP and 1.00 BPP). Within the genus Poecilia, there was belonging to Micropoecilia and Acanthophacelus. Second, a seven bp
a basal split between Limia * Mollienesia * Pamphorichthys and deletion (8514-8520 in concatenation) in intron 9 of X-src is
Micropoecilia * Acanthophacelus. Within Micropoecilia, all clades shared by both Poecilia (Micropoecilia) picta individuals.
were strongly supported (100% BSP and 1.00 BPP). Poecilia (Micro-
poecilia) bifurca grouped with P. (M.) branneri, these two were sister 3.3. Molecular dating analyses
to P. (M.) parae, and this clade was the sister-group to P. (M.) picta.
Bayesian trees supported an association of Limia and Pamphorich- A chronogram based on molecular dating analyses with BEAST
thys to the exclusion of Mollienesia, whereas MP and ML bootstrap v1.4.8 is shown in Fig. 3. Point estimates of divergence times, along
analyses favored an association of Limia and Mollienesia to the with 95% highest posterior densities, are given in Supplementary
exclusion of Pamphorichthys. Information Table 5. The base of Poecilia was dated at ~26 mya.
Analyses based on the nuclear and mitochondrial data sets are Within Poecilia, the split between Acanthophacelus and Micropoecil-
generally concordant with the combined data set, but support val- ia was dated at ~21-22 mya.
ues for some nodes are less robust (Supplementary Information The most recent common ancestor of Poecilia ( Micropoecilia)
Figs. 1 and 2; Supplementary Information Tables 3 and 4). In con- was dated at ~18 mya. Within this subgenus, P. parae diverged
trast to analyses with the combined and nuclear data sets, analyses from P. bifurca * P. branneri at ~14-15 mya, and the latter two spe-
with the mitochondrial data set (MPP tree shown in Fig. 4) ex- cies split at ~8-9 mya.
636 RW. Meredith et al. /Molecular Phylogenetics and Evolution 55 (2010) 631-639

Cnesterodon decemmaculatus

Cnesterodon hypselurus

Poecilia (Limia) dominicensis

100
100 Poecilia (Limia) melanogaster
54
17 Poecilia (Pamphorichthys) araguaiensis
100
100 100
Poecilia (Pamphorichthys) minor
100
Poecilia (Mollienesia) caucana
100
100 Poecilia (Mollienesia) latipunctata

Poecilia (Micropoecilia) bifurca

100
100 100
Poecilia (Micropoecilia) branneri
100 100
100
Poecilia (Micropoecilia) parae French Guyana
100
100 100 Poecilia (Micropoecilia) parae Suriname
100
Poecilia (Micropoecilia) picta Venezuela
100
100 100
100
Poecilia (Micropoecilia) picta Trinidad

Poecilia (Acanthophacelus) reticulata.

100
100 Poecilia (Acanthophacelus) wingei
100
100
Poecilia (Acanthophacelus) wingei Venezuela.
0.05 changes
Fig.2. ML phylogram (In likelihood = —27849.8817) obtained with RAXML for the combined data set with 11 partitions. Mean Bayesian posterior probabi ies (BPPs) based
on two independent runs and ML bootstrap support percentages (BSPs) are shown above and below branches, respectively.

3.4. Ancestral state reconstructions for analyses with each combination of priors that were employed.
The three different combinations of priors gave posterior proba-
Fig. 4 shows the posterior probabilities of the reconstructed bilities that differed by no more than 0.0541 at a given node.
ancestral states for lecithotrophy and placentotrophy using rate The ancestral state reconstruction based on parsimony indicates
parameter priors of z-3 and f=2 for the T distribution, and that placentotrophy evolved once in Poecilia (Micropoecilia) in
an uninformative prior on the bias parameter (= 1). Parsimony the common ancestor of P. parae, P. bifurca, and P. branneri. SIM-
ancestral state reconstructions are also shown in Fig. 4. Supple- MAP reconstructions also support a single origin of placentotro-
mentary Information Table 6 gives the posterior probabilities of phy in the common ancestor of these species, but with less
the reconstructed ancestral states (lecithotrophy, placentotrophy) certainty.

Table 3
Bootstrap support percentages (ML and MP) and Bayesian posterior probabilities for analyses with the combined data set. Nodes in Fig. 1 that were supported by 100* bootstrap
percentages in MP and ML analyses, and 1.00 posterior probabilities in Bayesian analyses, are not reported.
Node MP ML Bayesian
Four Eight n Four Eight 1
Paritions — Partitions —— partitions partitions
P. (Pamphorichthys) * P. (Limia)* P. (Mollienesia) 76 99 100 100 1.00
P. (Limia) * P. (Mollienesia) 66 62 61 83 028 004 046
P. (Pamphorichthys) * P. (Limia) 29 38 39 17 072 096 054
P. (Limia) * P. (Mollienesia) * P. (Micropoecilia) + P. (Acanthophacelus) — 175 — 0 02 0 000 000 000
 RW. Meredith et al./ Molecular Phylogenetics and Evolution 55 (2010) 631-639 637

Poecilia (Acanthophacelus) wingei

Poecilia (Acanthophacelus) wingei Venezuela

LL Poecilia (Acanthophacelus) reticulata

i Poecilia (Micropoecilia) picta Venezuela

LL Poecilia (Micropoecilia) picta Trinidad

Poecilia (Micropoecilia) bifurca

1
|
L———— Poecilia (Micropoecilia) branneri

Poecilia (Micropoecilia) parae French Guyana

Poecilia (Micropoecilia) parae Suriname

1! ]
. POCClIa (Limia) dominicensis
|
L [ Poecilia (Limia) melanogaster

_e_._ Poecilia (Pamphorichthys) araguaiensis

! 1
M Poecilia (Pamphorichthys) minor
!I d1
— — Poecilia (Mollienesia) caucana
!
—;—;— Poecilia (Mollienesia) latipunctata
! ]
—e—* Cnesterodon decemmaculatus
1
bt Cnesterodon hypselurus
Oligocene Miocene ] Piio. [ Pt
Palegere. IG
TENOZOIC
30 20 10 0
Million Years Ago
Fig. 3. A timeline in millions of years before present for Poecilia evolution based on analyses with BEAST and the combined data set (11 partitions). Grey bars indicate 95%
highest posterior densities (HPDs). The open circle denotes the node that was used to calibrate divergence times (see text). Plio. Pliocene; Pl. - Pleistocene.

[— — Poecilia (Micropoecilia) bifurca

OO0
@ B Lecithotrophic L Poecilia (Micropoecilia) branneri
Q [ Placentotrophic r—oOL
—— Poecilia (Micropoecilia) parae French Guyana
OO0
r—29. L Poecilia (Micropoecilia) parae Suriname

[ — Poecilia (Micropoecilia) picta Venezuela

L—— Poecilia (Micropoecilia) picta Trinidad

L—— Poecilia (Acanthophacelus) reticulata.

'-— Poecilia (Acanthophacelus) wingei

L — Poecilia (Acanthophacelus) wingei Venezuela

Fig. 4. SIMMAP (circles) and parsimony (squares ) reconstructions for placentotrophy and lecithotrophy in Poecilia (Acanthophacelus) and Poecilia (Micropoecilia).
638 RW. Meredith et aL /Molecular Phylogenetics and Evolution 55 (2010) 631-639

4. Discussion supported this hypothesis. In addition, our results confirm Poeser

et al’s (2005) hypothesis that Poecilia wingei, which was described
4.1. Mitochondrial versus nuclear data after Rosen and Bailey's (1963) publication, is the sister taxon to
Poecilia reticulata.
The mitochondrial data set exhibits much more variation than Finally, our results are consistent with a clade comprising Limia,
the nuclear data set. The average number of substitutions per site Mollienesia, and Pamphorichthys as in Hamilton (2001) and Hrbek
for the mitochondrial data set is 7.8-fold higher than for the nucle- et al. (2007). However, we interpret this result with caution given
ar data set (Table 2). The fastest-evolving partition of the mito- our incomplete taxonomic sampling for these lineages. Our analy-
chondrial data set (3rd codon positions) is 130-fold faster than ses failed to resolve relationships among Pamphorichthys, Limia,
the slowest-evolving partition of the nuclear data set (2nd codon and Mollienesia, whereas Hrbek et al. (2007) reported a high pos-
positions) (Table 2). Whereas the mitochondrial sequences are terior probability (0.97) for a sister-group relationship between Li-
evolving faster and contain a larger percentage of parsimony-infor- mia and Pamphorichthys. In contrast to Hamilton (2001), Hrbek
mative sites than nuclear sequences, the nuclear data exhibit less et al. (2007), and our own molecular results, Lucinda and Reis
homoplasy based on RC and RI values. Even with these differences, (2005) found that Pamphorichthys was more closely related to
trees based on the mitochondrial and nuclear data sets provide ro- Micropoecilia and Acanthophacelus than to Limia based on morpho-
bust support for many of the same clades (e.g, Limia, Mollienesia, logical data. Lucinda and Reis (2005) listed seven characters that
Pamphorichthys, Acanthophacelus). There are also instances of con- diagnose this clade, but also noted that all of these characters were
gruence, but with one of the two data sets providing much stronger “not uniquely derived and/or reversed features" (p. 46).
support than the other data set. Mitochondrial data provide robust
support for the monophyly of Micropoecilia * Acanthophacelus, but 4.3. The evolution of placentotrophy in Micropoecilia
only moderate support for Mollienesia * Pamphorichthys * Limia. In
contrast, nuclear data provide stronger support for Molliene- Within Poeciliidae, placentotrophy has evolved in Heterandria
sia * Pamphorichthys * Limia, but weaker support for Acanthophace- formosa (Grove and Wourms, 1991, 1994), Phalloceros caudimacul-
lus * Micropoecilia. Within Micropoecilia * Acanthophacelus, nuclear atus (Arias and Reznick, 2000), Xenodexia ctenolepis (Reznick et al.,
data provide stronger support for Micropoecilia monophyly than 2007), and three times within the genus Poeciliopsis (Reznick et al.,
mitochondrial data. Together, the mitochondrial and nuclear data 2002). Pires et al. (in press) documented yet another independent
sets complement each other and yield a robust phylogeny for al- origin of placentotrophy in Poecilia (Micropoecilia). Specifically, pla-
most all nodes. The inclusion of gene regions that evolve at differ- centotrophy occurs in P. branneri, P. bifurca, and P. parae, but not in
ent evolutionary rates maximizes the likelihood of obtaining P. picta. Our results provide robust support for the monophyly of P.
phylogenetic resolution at both shallow and deep levels in a phy- branneri, P. bifurca, and P. parae within Micropoecilia. Ancestral
logenetic tree (Hrbek et al., 2007). state reconstructions support the hypothesis that placentotrophy
evolved in the common ancestor of these taxa. The branch on
which placentotrophy evolved extends from ~18 to ~14-15 mya,
4.2. Phylogenetic relationships and establishes a maximum time interval of ~3-4 myr for the evo-
lution of placentotrophy. Reznick et al. (2002) reported maximum
Hubbs (1926) erected the genus Micropoecilia and included five time intervals of 236 and 0.75 million years for the evolution of
species (M. bifurca, M. parae,M. picta,M. melanzona, M. branneri)
in this placentotrophy in two different clades of Poeciliopsis.
genus. Rosen and Bailey (1963) folded Micropoecilia Hubbs, 1926 into
the subgenus Poecilia (Lebistes) along with P. reticulata, P. amazonica, 4.4. Taxonomic recommendations
and P. scalpridens. Micropoecilia was reinstated by Meyer (1993).
Subsequently, the monophyly of this group has been challenged An important debate in poeciliid systematics is whether to treat
(Costa and Sarraf, 1997) and also supported (Figueiredo, 2003; pers. Poecilia as a genus with multiple subgenera (e.g. Rosen and Bailey,
comm.)based on analyses of morphological characters. Most recently, 1963) or to elevate each subgenus to the level of genus. In our view,
Poeser et al. (2005) recognized five species (P. parae, P. branneri, resolution of this debate hinges on maintaining the scientific name
P. bifurca, P. picta, P. minima) in the subgenus Poecilia (Micropoecilia). Poecilia reticulata for the common guppy, which has been and re-
The present study is the most comprehensive molecular phylo- mains a model organism for scientific studies. This stability is main-
genetic study to date on Micropoecilia, both in terms of taxonomic tained if we follow a modified version of Rosen and Bailey's (1963)
representation and gene sampling. We included four of five Micro- taxonomic scheme that recognizes multiple subgenera in the genus
poecilia species recognized by Poeser et al. (2005), all of which Poecilia. Accordingly, we support a taxonomy that recognizes
were originally included in Micropoecilia when it was proposed Acanthophacelus, Micropoecilia, Limia, Pamphorichthys, and Molliene-
as a new genus by Hubbs (1926). Our results corroborate Meyer's sia as subgenera in the genus Poecilia for taxa that were included in
(1993) hypothesis that Micropoecilia is monophyletic. our study. Whereas each of these subgenera has been recognized
Within Micropoecilia, our molecular phylogeny is discordant with in one or more modern classifications [e.g. Rosen and Bailey
Meyer's (1993) morphological phylogeny. Meyer (1993) proposed a (1963) recognized Limia, and Pamphorichthys; Meyer and Radda
sister-group relationship between M. bifurca and M. picta, a sister- (2000) recognized Mollienesia; Poeser et al. (2005) recognized
group relationship between these two and M. parae, and a basal split Acanthophacelus and Micropoecilia], they have never been recog-
between M. branneri and all other species of Micropoecilia. Each of nized as monophyletic subgenera in thesameclassification. Our rec-
these hypotheses is supported by one or more putative gonopodial ognition of these subgenera is based on phylogenetic analyses of the
synapomorphies (Meyer, 1993). In contrast, our results supporta sis- most comprehensive molecular data set yet published for represen-
ter-group relationship between M. bifurca and M. branneri, a sister- tative taxa in the genus Poecilia. Rosen and Bailey (1963) included
group relationship between these two and M. parae, and a basal split mollies and P. vivipara in the subgenus Poecilia rather than Molliene-
between M. picta and all other species of Micropoecilia (Figs. 1 and 2). sia, but molecular studies (Breden et al., 1999; Hamilton, 2001) sug-
Our results also provide support for a sister-group relationship gestthat Rosen and Bailey's (1963) subgenus Poecilia is polyphyletic,
between Micropoecilia and Acanthophacelus and corroborate previ- and that P. vivipara represents a distinct lineage within the genus
ous morphological (Lucinda and Reis, 2005) and molecular (Breden Poecilia thats separate from mollies. We therefore recommend plac-
et al., 1999; Hamilton, 2001; Hrbek et al., 2007) studies that have ing mollies in the subgenus Mollienesia, and retaining the subgenus
 RW. Meredith et al./ Molecular Phylogenetics and Evolution 55 (2010) 631-639 639

Poecilia for Poecilia vivipara. Breden et al. (1999) and Hamilton Hrbek,T, Seckinger, . Meyer,A, 2007. A phylogeneticand biogeographic perspective
(2001) also found that P. vivipara is separate from P. reticulata. on the evolution of poeciliid fishes. Mol. Phylogenet. Evol. 43, 986-998.
Huelsenbeck, |.P., Ronquist, F., 2001. MRBAYES: Bayesian inference of phylogenetic
In contrast to the taxonomic approach that we recommend, which trees. Bioinformatics 17, 754-755.
will maintain taxonomic stability forP. reticulata, elevation of sub- Huelsenbeck, JP, Nielsen, R, Bollback, JP. 2003. Stochastic mapping of
genera to the generic level will necessitate moving Poecilia reticulata morphological characters. Syst. Biol. 52, 131-158.
Kocher,T.D, Conroy, JA., McKaye, KR, Stauffer, ] R, Lockwood, S, 1995. Evolution
toa different genus (i.e Acanthophacelus) becauseP. vivipara repre- of NADH dehydrogenase subunit 2 in East African cichlid fish. Mol. Phylogenet.
sents a distinct lineage based on other molecular studies (Breden Evol. 4, 420-432.
et al, 1999; Hamilton, 2001) and is the type species for the genus Lóytynoja, A., Milinkovitch, M.C, 2001. SOAP, cleaning multiple alignments from
Poecilia. If subgenera are elevated to genera, P. vivipara has priority unstable blocks. Bioinformatics 17, 573-574.
Li, C, Ortí, G, Zhang, G. Lu, G., 2007. A practical approach to phylogenomics: the
and will retain the genus name Poecilia. phylogenyof ray-finned fish Actinopterygii) s a case study. BMCEvol. Biol.7, 44.
Lucinda, P.H.F, 2003. Family Poeciliidae. In: Reis, RE. Kullander, S.O., Ferraris, CJ.
Acknowledgments (Eds.), Check List of the Freshwater Fishes of South and Central America.
EDIPUCRS, Porto Alegre, Brazil., pp. 555-581.
Lucinda, P.H., Reis, R.E, 2005. Systematics of the subfamily Poeciliinae Bonaparte
We thank F. Breden, Heather Alexander, J. de Greef, M. Schartl, (Cyprinodontiformes: Poecilidae) with an emphasis on the tribe
and the Sào Paulo University Museum of Zoology, Brazil (MZUSP), Cnesterodontini Hubbs. Neotrop. Ichthyol. 3, 1-60.
Meyer, A, Lydeard, C, 1993. The evolution of copulatory organs, internal
in particular M. de Pinna, C.R. Moreira, and CA. Figueiredo for pro- erülization, placentae and viviparity in killifishes (Cyprinodontiformes)
viding some of the specimens used in this work. We also thank C. Li inferred from a DNA phylogeny of the tyrosine kinase gene X-src. Proc. R. Soc.
and G. Orti for providing primer sequences for nuclear genes. J. Lond. B 254, 153-162.
Meyer, MK, 1993. Reinstatement of Micropoecilia Hubbs, 1926, witha redescription
Heraty provided helpful comments on nomenclature. This study of M. bifurca (Eigenmann, 1909) from Northeast South America (Teleostei,
was supported by NSF grant DEB0416085 to D.N.R. and M.S.S. Cyprinodontiformes: Poeciliidae). Zool. Abhand. Staatliches Mus. fiir Tierkunde
Dresden 47, 121-130.
Meyer, MK, Radda, A.C, 2000. Notes on the subgenus Mollienesia LeSueur, 1821,
Appendix A. Supplementary data with a description of a new species of Poecilia Bloch & Schneider, 1801
(Cyprinodontiformes: Poeciliidae) from Venezuela. Ann. Naturhist. Mus. Wien
Supplementary data associated with this article can be found, in 1028, 75-81.
Nielsen, R, 2002. Mapping mutations on phylogenies. Syst. Biol. 51, 729-739.
the online version, at doi:10.1016/j.ympev.2009.11.006. Parenti, LR, 1981. A phylogenetic and biogeographic analysis of cyprinodontiform
fishes. Bull. Amer. Mus. Nat. Hist. 168, 335-557.
References Pires, M, 2007. The Evolution of Placentas in Poe i Fishes [Ph.D. Dissertation].
University of California Riverside. pp. 228.
Pires, M, Arendt, J, Reznick DN., in press. The evolution of placentas and
Arías, A, Reznick, D., 2000. Life history of Phalloceros caudiomaculatus: a novel superfetation in the fish genus Poecilia (Cyprinodontiformes: Poeciliidae:
variation on the theme of livebearing in the family Poeciliidae. Copeia, 792-798. subgenera Micropoecilia and Lebistes). Biol. J. Linn. Soc.
Bailey, R.M,, Miller, R R, 1950. Mollienesia versus Mollienisia as the name fora genus Poeser, ., Isbrücker, L.H., 2002. Zum wissenschaftlichen Namen des Guppy. Die
of poeciliid fishes. Copeia, 318. Aquarien-und Terrarien-Zeitschrift 55, 47-49.
Bollback, J.P., 2006. SIMMAP: stochastic character mapping of discrete traits on Poeser, EN, Kempkes, M, Isbrücker, LJH. 2005. Description of Poecilia
phylogenies. BMC Bioinf. 7, 88. (Acanthophacelus) wingei n. sp. from Paría Peninsula, Venezuela, including
Breden, F, Ptacek, M.B., Rashed, M., Taphomn, D., Figueiredo, C.A, 1999. Molecular notes on Acanthophacelus Eigenmann, 1907 and other subgenera of Poecilia
phylogeny of the live-bearing fish genus Poecilia (Cyprinodontiformes: Bloch and Schneider, 1801 (Teleostei, Cyprinodontiformes, Poeciliidae). Contrib.
idae). Mol. Phylogenet. Evol. 12, 95-104. Zool. 74, 97-115.
. Bonillo, C Lecointre, G. 2003. Repeatability of clades as a criterion of Posada, D. Crandall, KA, 1998. MODELTEST: Testing the model of DNA
reliability: a case study for molecular phylogeny of Acanthomorpha (Teleostei) substitution. Bioinformatics 14, 817-818.
with larger number of taxa. Mol. Phylogenet. Evol. 26, 262-288. Ptacek, M.B, Breden, F, 1998. Phylogenetic relationships among the mollies
Costa, W].EM,, 1991. Description d'une nouvelle espéce du genre Pamphorichtlys (Poeciliidae: Poecilia: Mollienesia group) based on mitochondrial DNA
(Cyprinodontiformes: Poeciliidae) du basin de l'Araguaia. Brésil. Rev. fr. sequences. J. Fish Biol. 53, 64-81.
Aquariol. Herpétol. 2, 39-42. Rambaut, A, Drummond, AJ, 2003. Tracer version 1.2 [computer program].
Costa,W.JE.M,,Sarraf, A, 1997. Poecilia (Lebistes) minima, a new species of neotropical Available from: http:]Jevolvezoo.oxac.uk.
poeciliid fish from the Brazilian Amazon. Ichthyol. Explor. Freshwaters 8, 185- Regan, CT., 1913. A revision of the cyprinodont fishes of the subfamily Poeciliinae.
191. Proc. Zool. Soc. Lond. 11, 977-1018.
de Queiroz, A., 1993. For consensus (sometimes). Syst. Biol. 42, 368-372. Reznick, D.N,, Mateos, M., Springer, M.S., 2002. Independent origins and rapid
Drummond, AJ, Rambaut, A, 2007. BEAST: Bayesian evolutionary analysis by evolution of the placenta in the fish genus Poeciliopsis. Science 298, 1018-1020.
sampling trees. BMC Evol. Biol. 7, 214. Reznick, D., Hrbek, T., Caura, S, de Greef, ., Roff, D., 2007. Life history of Xenodexia
Drummond, A J, Ho, S.Y.W. Phillips, M, Rambaut, A, 2006. Relaxed phylogenetics ctenolepis: implications for life history evolution in the family Poecili dae. Biol..
and dating with confidence. PLoS Biol. 4, e88. Linn. Soc. (Lond.) 92, 77-85.
Endler, J.A, 1983. Natural selection on color patterns in poeciliid fishes. Environ. Ronquist, F Huelsenbeck, J.P., 2003. MrBayes 3: Bayesian phylogenetic inference
Biol. Fish. 9, 173-190. under mixed models. Bioinformatics 19, 1572-1574.
Farris, |5, Kallersjo, M. Kluge, AG. Bult, C, 1994. Testing significance of Rosen, D.E, Bailey, RM,, 1963. The poeciliid fishes (Cyprindontiformes), their
incongruence. Cladistics 10, 315-319. structure, zoogeography, and systematics. Bull. Amer. Mus. Nat. Hist. 126, 1-176.
Figueiredo, CA, 2003. Análise ladístia da subfamília Poecilinae Rosen, D.E, Gordon, M, 1953. Functional anatomy and evolution of male genitalia
(Cyprinodontiformes: Poeciliidae) coménfase nas inter-relacóes dos géneros in poeciliid fishes. Zoologia 38, 1-47.
Poecilia, Limia, Pamphorichthys. [Ph.D. Dissertation]. Universidade de S3o Paulo, Schluter, A., Parzefall, J Schlupp, L, 1998. Female preference for symmetrical
Sáo Paulo, SP, Brazil, pp. 274. vertical bars in male sailfin mollies. Anim. Behav. 56, 147-153.
Gatesy, ], Desalle, R., Wheeler, W., 1993. Alignment-ambiguous nucleotide sites and Schmidt, TR, Bielawski, ] Gold, | R., 1998. Molecular phylogenetics and evolution
the exclusion of systematic data. Mol. Phylogenet. Evol. 2, 152-157. of the cytochrome b gene in the cyprinid genus Lythrurus (Actinopteryg
Grove, BD., Wourms, J.P., 1991. The follicular placenta of the viviparous fish, Cypriniformes). Copeia 1998, 14-22.
Heterandria formosa. l. Ultrastructure and development of the embryonic Shories, S, Meyer, M.K, Schartl, M., 2009. Description of Poecilia (Acanthophacelus )
absorptive surface. J. Morph. 209, 265-284. obscura n. sp., (Teleostei: Poecillidae), a new guppy species from western
Grove, BD., Wourms, |P, 1994. Follicular placenta of the viviparous fish, Trinidad, with remarks on P. wingei and the status of the "Endler's guppy".
Heterandria formosa: ll Ultrastructure and development of the follicular Zootaxa 2266, 35-50.
epithelium. J. Morph. 220, 167-184. Stamatakis, A, 2006. RAXMLVIHPC: maximum likelihood-based phylogenetic
Hamilton, A, 2001. Phylogeny of Limia (Teleostei:Poeciliidae) based on NADH analyses with thousands of taxa and mixed models. Bioinformatics 22, 2688-2690.
dehydrogenase subunit 2 sequences. Mol. Phylogenet. Evol. 19, 277-289. Swofford, D.L, 2002. PAUP. Phylogenetic Analysis using Parsimony (+ and Other
Hedges, S-B., Kumar, S., 2004. Precision of molecular time estimates. Trends Genet. Methods) Sinauer Associate, Sunderland, MA.
20, 242-247. Teeling, EC., Scally, M., Kao, D]., Romagnoli, M.L, Springer, M., Stanhope, MJ.,
Houde, AE., 1997. Sex, Color, and Mate Choice in Guppies. Princeton University 2000. Molecular evidence regarding the origin of echolocation and flight in bats.
Press, Princeton, NJ. Nature 403, 188-192.
Hubbs, C.L, 1924. Studies of the fishes of the order Cyprinodontes. Misc. Publ. Mus. Yang,Z, Rannala, B., 2006. Bayesian estimation of species divergence times under a
Univ. Michigan 13, 1-31. molecular clock using multiple calibrations with soft bounds. Mol. Biol. Evol. 23,
Hubbs, C.L, 1926. Studies of the fishes of the order Cyprinodontes VI. Misc. Publ. 212-226.
Mus. Univ. Michigan 16, 1-86.