Monophyly of Heterandriini (Teleostei: Poeciliidae) revisited: a critical review of the data

The systematics and taxonomy of poeciliid fishes (guppies and allies) remain poorly understood despite the relative importance of these species as model systems in the biological sciences. This study focuses on testing the monophyly of the nominal poeciliine tribe Heterandriini and the genus Heterandria, through examination of the morphological characters on which the current classification is based. These characters include aspects of body shape (morphometrics), scale and fin-ray counts (meristics), pigmentation, the cephalic laterosensory system, and osteological features of the neurocranium, oral jaws and suspensorium, branchial basket, pectoral girdle, and the gonopodium and its supports. A Maximum Parsimony analysis was conducted of 150 characters coded for 56 poeciliid and outgroup species, including 22 of 45 heterandriin species (from the accounted in Parenti & Rauchenberger, 1989), or seven of nine heterandriin species (from the accounted in Lucinda & Reis, 2005). Multistate characters were analyzed as both unordered and ordered, and iterative a posteriori weighting was used to improve tree resolution. Tree topologies obtained from these analyses support the monophyly of the Middle American species of “Heterandria,” which based on available phylogenetic information, are herein reassigned to the genus Pseudoxiphophorus. None of the characters used in previous studies to characterize the nominal taxon Heterandriini are found to be unambiguously diagnostic. Some of these characters are shared with species in other poeciliid tribes, and others are reversed within the Heterandriini. These results support the hypothesis that Pseudoxiphophorus is monophyletic, and that this clade is not the closest relative of H. formosa (the type species) from southeastern North America. Available morphological data are not sufficient to assess the phylogenetic relationships of H. formosa with respect to other members of the Heterandriini. The results further suggest that most tribe-level taxa of the Poeciliinae are not monophyletic, and that further work remains to resolve the evolutionary relationships of this group.

The family Poeciliidae comprises three subfamilies that are restricted to fresh and brackish continental waters: Poeciliinae, Procatopodinae, and Aplocheilichthyinae (Ghedotti, 2000;Lucinda, 2003;Hrbek et al., 2007).The Poeciliinae includes at least 228 species distributed across much of the tropical and subtropical portions of the Americas, from the La Plata estuary of northern Argentina to southeastern United States, with species richness reaching a zenith in Middle America and the West Indies (Rosen & Bailey, 1963;Lucinda, 2003;Hrbek et al., 2007;Albert et al., 2011).The Procatopodinae includes at least 78 species in the humid tropical regions of South America (i.e.Fluviphylax) and Africa (Lucinda, 2003), and the Aplocheilichthyinae is represented by about 11 species of Aplocheilichthys from humid tropical regions of Africa (Huber, 1999).
All poeciliine species have internal fertilization achieved by means of an intromittent organ (gonopodium) in males, which is a modification of anal-fin rays 3-5.Poeciliine species bear their young as free-swimming juveniles (without a yolk sac), and exhibit either ovoviviparity (eggs hatch within the uterine cavity), or viviparity (maternal provisioning to the developing larvae while in the uterine cavity).The ovoviviparity condition is thought to be derived within the subfamily (Pollux et al., 2009).One poeciliine species, Tomeurus gracilis from northeastern South America, is facultatively ovoviviparous, releasing eggs or live young depending on environmental conditions (Breder & Rosen, 1966;Wourms, 1981).The Poeciliinae is also characterized by several salient morphological features in addition to the gonopodium; expansion of the fourth epibranchial, absence of exoccipital condyles, and neural arches of the first vertebra open (Rosen & Bailey, 1963;Parenti & Rauchenberger, 1989;Ghedotti, 2000).
The name Heterandriini was introduced by Hubbs (1924) for a group of poeciliins in which the third ray of the gonopodium lacks a pair of curved horn-like projections, and in which the fifth ray always is smooth on the posterior edge.This group originally included nine genera; Allogambusia (now Priapichthys), Alloheterandria (now Priapichthys and Gambusia), Darienichthys (now Priapichthys), Heterandria, Neoheterandria, Panamichthys (now Priapichthys), Priapichthys, Pseudopoecilia, and Pseudoxiphophorus (treated as a subgenus by Rosen, 1979).Hubbs (1926) published diagnoses to identify each of these genera.According to Hubbs (1926), species of the tribe Heterandriini evolved in Central America and northwestern (trans-Andean) South America.Rosen & Bailey (1963) organized the Poeciliidae into five subfamilies, nine tribes, and 52 genera, primarily on the basis of similarities in the gonopodium and gonopodial suspensorium.They included six genera in the tribe Heterandriini; Heterandria, Neoheterandria, Phallichthys Poeciliopsis (which was divided into the subgenera Poeciliopsis and Aulophallus), and Priapichthys.The diagnostic characters of Heterandriini sensu Rosen & Bailey (1963) were also largely taken from the gonopodial complex; i.e., gonopodial length more than one-third standard length, presence of three or more gonapophyses, moderately or welldeveloped ligastyle, and two to four plate-like gonactinosts.However, these traits are shared with species of other nominal poeciliine tribes, and are not in fact unique to species of the Heterandriini.
Several subsequent phylogenetic studies based on morphological (Rodriguez, 1997;Ghedotti, 2000;Lucinda & Reis, 2005) and molecular (Hrbek et al., 2007) data suggest that some of the tribes proposed by Rosen & Bailey (1963) are not monophyletic.These papers also show that some genera not previously placed in the Heterandriini (i.e., Alfaro, Brachyrhaphis, Pseudopoecilia, Neoheterandria) are more closely related to Heterandria formosa Girard (1859), the type species of the genus, than to the Middle American species of Heterandria which were previously assigned to Pseudoxiphophorus (Heckel, 1848).Despite these concerns, the pre-cladistic classification of Rosen & Bailey (1963) remains the most recent comprehensive morphological revision and taxonomic authority for Poeciliidae (Lucinda, 2003;Hrbek et al., 2007).The Poeciliidae of Rosen and Bailey (1963) is equivalent to the Poeciliinae of Parenti (1981).
This study focuses on the systematics of the nominal poeciliine taxon Heterandriini (sensu Lucinda & Reis, 2005) and the type genus Heterandria, using osteological and other morphological characters (morphometrics, meristics, pigmentation, and cephalic sensory system).As currently recognized, the Heterandriini includes five genera (Priapichthys, Neoheterandria, Heterandria, Poeciliopsis, and Phallichthys) with more than 60 species (Parenti & Rauchenberger, 1989;Ghedotti, 2000).The goal of this study is to test the monophyly of the taxa Heterandriini and Heterandria by evaluating all the characters used historically in the higher classification of the group within a formal phylogenetic context.
External characters, morphometrics, and meristics were examined from museum specimens of 66 species preserved in alcohol.Meristics and morphometrics followed the procedure outlined by Miller (1948) for cyprinodont fishes.Counts of scales in the lateral series follow Parenti (1981).All meristics and morphometrics were taken form the left side of specimens when possible using an Olympus SZX-12 dissecting microscope with 6-inch LCD digital caliper (± 0.1 mm error).
Clearing and staining for bones and cartilages followed the procedure of Taylor & van Dyke (1985) with some modifications to account for the small size of the specimens.Oxygen peroxide and xylene were not used to avoid decalcification.Specimens were transferred directly from ethyl alcohol to Alcian Blue to maximize uptake of the hydrophobic stain.The viscera were removed after staining with Alcian blue because the small cartilages around the orbit are more apparent and the damage to the specimens is reduced (Reis, pers. comm., 2009).Prior to digestion in 30% trypsin solution (in sodium borate [NaBO]) specimens were treated one hour in 30% NaBO.
Dissections of cleared-and-stained fishes were made using microdissection tools following the method outlined by Weitzman (1974), and Ghedotti's (2000) modification of removing the branchial basket before the suspensorium.Bones were disarticulated to functional groups (e.g., neurocranium, suspensorium, pectoral girdle), or into individual elements (e.g., maxilla).Specimens from a total of 56 species were dissected and coded for phylogenetic analysis.Outlines and standardized features of individual bones or functional groups were traced from lateral and medial aspects with the aid of a camera lucida mounted on an Olympus SZX-12 dissecting microscope.Images were digitized using an HP Scanjet 5470c (2400dpi, 48-bit color) scanner and edited using Adobe Photoshop 7.0.
Descriptions of 150 characters and their alternative states are provided with characters organized by functional group (Appendix 2).Character states were polarized using conditions observed in outgroup taxa.Question marks were used to indicate when a character state could not be coded due to a lack of available specimens, or when the coding was non-sensical; e.g., state of the ligastyle (a free ossified haemal spine dorsal to the gonopodial suspensorium and close to the vertebral column) when a gonopodium is not present in the species.Whenever possible characters were coded for states observed in multiple specimens (Albert & Crampton, 2005).Due to a lack of male specimens of Quintana and Fluviphylax gonopodial characters from these taxa were coded from literature.In most species, specimens of both sexes were examined, but in sexually dimorphic species only states observed in males were coded.

Results
These phylogenetic results are shown with multistate characters analyzed ordered and unordered, and used iterative weighting on the RC to resolve polytomies.The strict consensus tree of 12 MP trees based on 150 unordered characters is given (Fig. 1; each tree 1430 steps, CI = 0.20, RI = 0.48, RC = 0.10).A single MP tree was obtained from analysis of this same dataset with all multistate characters treated as unordered and followed by a posteriori weighting based on the RC (Fig. 2; 1444 steps, CI = 0.20, RI = 0.47, RC = 0.09).A single MP tree with a slightly different topology (Fig. 3) was obtained when the multistate characters were treated as ordered (1446 steps, CI = 0.20, RI = 0.47, RC = 0.09).The strict consensus tree with all characters weighted equally, and with all multistate characters treated as unordered, is more-well resolved for the tribes Cnesterodontini, Scolichthyini, Gambusiini, and Alfarini (Fig. 1).
This study also suggests that the characters used in previous morphological analyses of Poeciliinae do not support the monophyly of Heterandriini.The monophyly of Heterandriini sensu Lucinda & Reis (2005; see Table 3), including only species of the subgenera Heterandria and Pseudoxiphophorus was not supported by the results of this study.Heterandria formosa (the type species) was not found most closely related to Pseudoxiphophorus (from Mesoamerica) in any of the topologies recovered (Figs.1-3).We therefore recommend restricting Heterandria to H. formosa (following Regan, 1913), and regard Pseudoxiphophorus as the valid name for the Mesoamerican clade (following Hubbs, 1924).

Discussion
Comparisons with previous studies.Some of the phylogenetic results reported above are consistent with the conclusions of previous studies of Poeciliinae.The phylogeny reported in the strict consensus tree (Fig. 1) supports the monophyly of the Cnesterodontini (sensu Lucinda & Reis, 2005), Gambusiini (sensu Rosen & Bailey, 1963), and Tomeurini (sensu Rosen & Bailey, 1963;Parenti & Rauchenberger, 1989;Lucinda & Reis, 2005).The monophyly of the Priapichthyini (sensu Lucinda & Reis, 2005) and Alfarini (sensu Ghedotti, 2000) were not well tested, as each was represented by species in a single genus.
Other results of this study are not consistent with the published taxonomy of Poeciliinae.The Cnesterodontini was not well resolved in the topologies obtained from weighting the unordered multistate characters by the RC (Fig. 2), or from analyzing the data with ordered multistate characters (Fig. 3).Those topologies recovered the Cnesterodontini within Poeciliopsis.The Gambusiini is not monophyletic according to the topology obtained by analyzing the data as ordered-multistate characters (Fig. 3).According to that topology, Brachyrhaphis is not most closely related to the clade Gambusia + Belonesox.
None of the phylogenies recovered in this study support the monophyly of the Poeciliini (sensu Ghedotti 2000;Lucinda & Reis 2005).In the strict consensus tree (Fig. 1), H. formosa is sister to Quintana and is closely related to Limia and Xiphophorus suggesting that these four genera should be included in Heterandriini.However, the clade of Quintana and H. formosa is not well supported by Bremer decay analysis.Further, male specimens of Quintana atrizona were not available for this study, and the data matrix was coded based on published description of this monotypic genus (Hubbs, 1934;Rosen & Bailey, 1963;Lucinda & Reis, 2005).In the topology obtained from a posteriori weighting of the unorderedmultistate characters (Fig. 2), H. formosa is most closely related to other genera of Poeciliini (Micropoecilia, Pamphorichthys, Limia, and Poecilia).
The topology obtained from analyzing the dataset as ordered-multistate characters (Fig. 3) supports H. formosa as closely related to the Brachyrhaphini.To verify the phylogenetic relationship between Brachyrhaphis and Heterandria it will be necessary to examine all (or most) of the 12 valid species of Brachyrhaphis.If H. formosa and Brachyrhaphis are not closely related, the whole Heterandriini could be restricted to H. formosa and Brachyrhaphis could be referred to as Brachyrhaphini sensu Lucinda & Reis (2005).
The strict consensus tree (Fig. 1) suggests that Neoheterandria, Xenophallus, and Scolichthys are closely related.If this relationship is true, Neoheterandria and Xenophallus might be included in the tribe Scolichthyini (rather than the tribe Gambusiini sensu Lucinda & Reis, 2005).This topology is supported by the analysis of unordered multistate characters with a posteriori weighting (Fig. 2), but not by the analysis of ordered-multistate characters (Fig. 3).
Poeciliopsis was found to be monophyletic in the strict consensus tree (Fig. 1), and is well supported by the Bremer decay analysis, as in Mateos et al. (2002).However, in the other two topologies arising from this study Poeciliopsis was not found to be monophyletic (Figs. 2, 3).Even though the relationships and monophyly of Poeciliopsis are uncertain in all of the analyses Poeciliopsis is not closely related to H. formosa and therefore should not be included in Heterandriini.
According to Lucinda & Reis (2005), Phallichthys should be included in the tribe Girardinini, with Girardinus, Poeciliopsis, Neoheterandria, and Phalloptychus.In the present study specimens of Girardinus and other species of Neoheterandria were not examined but Poeciliopsis and Phallichthys were not found to be closely related.The position of Carlhubbsia varied in this study, but it should be noted the species of Carlhubbsia used here (C.stuarti) differed from that (C.kidderi) used by Lucinda & Reis (2005).
Gonopodial characters have been proven effective in structuring aspects of higher level poeciliid classification, and also at the species level.A fully-resolved, generic-level, morphology-based phylogeny of the Poeciliidae awaits complete osteological surveys of the skeletal system for representatives of all the major clades, including taxa missing from the present study (Girardinus Priapella and Xenodexia).Furthermore, it is important to note that the tribe-level designation was originally introduced to differentiate poeciliid groups with characteristic gonopodial structures and distinct geographical distributions (Hubbs, 1924), and not necessarily to diagnose monophyletic groups.Indeed, the traditional tribe-level taxa used in poeciliine systematics do not closely match the results of modern phylogenetic investigations (Rodriguez, 1997;Costa, 1998;Ghedotti, 1998Ghedotti, , 2000;;Lucinda & Reis, 2005;Hrbek et al., 2007).

Taxonomy of Heterandria and Heterandriini
The taxonomic history of Heterandria and Pseudoxiphophorus is complex.Girard (1859) recognized Limia formosa from the coastal plains of southeastern United States (South Carolina to Mexico) on the basis of (among other things) a stout body with low vertebral counts (17+13 = 30), small teeth in the outer row of each jaw, a relatively tall dorsal fin (having long rays) positioned over the anal fin, and a long alimentary canal (longer than body) with numerous convolutions.This species was subsequently designated as the type of Heterandria Agassiz (1853; see Rosen & Bailey, 1963;Rosen, 1979), as a distinct genus from the type species of Limia L. vittata (Guichenot, 1853), from Cuba.
Bleeker (1860) recognized as Pseudoxiphophorus a species previously described as Xiphophorus bimaculatus (Heckel, 1848) collected from the coastal plains of Vera Cruz, Mexico.Pseudoxiphophorus was described on the basis of having a relatively elongate body (with high vertebral counts; 18+14=32), a flat head, large conical teeth in the outer row of both jaws, many and small gill rakers, a low, long dorsal fin positioned in advance of the anal fin, and a short simple alimentary canal (less than length of body).Regan (1913) treated Pseudoxiphophorus as a subgenus of Heterandria while Hubbs (1924) recognized Pseudoxiphophorus and Heterandria as distinct genera.Miller (1974) illustrated the gonopodia, gonopodial suspensorium, and color patterns of these two taxa, and also examined morphometric and meristic data.He concluded that the differences between the gonopodia are sufficient to recognize the two genera.Although Rosen & Bailey (1963) synonymized Pseudoxiphophorus bimaculatus and Heterandria formosa into the genus Heterandria, the description of new Heterandria species caused Rosen (1979) to recognize these taxa as two distinct subgenera.In his review of the genus Heterandria Rosen (1979) added more species to the taxon he referred to as the subgenus Pseudoxiphophorus referring to the assemblage of Heterandria species from Middle America, and limited the subgenus Heterandria to H. formosa.Rosen (1979) also provided a cladogram based on synapomorphies of the described species, area cladograms, and proposed the first identification key for the all species of the subgenus Pseudoxiphophorus.
The monophyly and taxonomic composition of Heterandria has subsequently remained poorly resolved.Whereas Radda (1985) and Ghedotti (2000) treated Pseudoxiphophorus and Heterandria as different genera, Parenti and Rauchenberger (1989) regarded Pseudoxiphophorus as a subgenus of Heterandria.The most recent systematic treatment of Poeciliinae focused on the tribe Cnesterodontini (Lucinda & Reis, 2005) and suggested that the Heterandriini should be limited to species of the genus Heterandria sensu Rosen & Bailey (1963).
More than twenty-seven species have been described since Rosen & Bailey (1963), and most of the characters traditionally used to diagnose tribe and genus-level taxa are now known to be present in different combinations among species within these higher taxa.In particular, many gonopodial characters have not been corroborated as phylogenetically informative at the tribe or generic-level in the face of increased taxon sampling.Despite more than a century of research, the monophyly and interrelationships of Heterandriini and Heterandria remain uncertain, and a formal taxonomic revision is necessary.
According to the present study, the monophyly of Heterandria sensu Rosen & Bailey (1963) is not supported by morphological data.However, by design, the character definitions used here were compiled from previous studies on groups other than Heterandriini (Rauchenberger, 1989;Rodriguez, 1997;Costa, 1998;Ghedotti, 1998Ghedotti, , 2000;;Lucinda & Reis, 2005).It is therefore not surprising that these characters do a relatively poor job of resolving phylogenetic relationships within the Heterandriini.However, the results of this study do show that Heterandria and Pseudoxiphophorus are not closely related, supporting the hypothesis of Regan (1913), Miller (1974), and Radda (1985) that Pseudoxiphophorus should be recognized as a generic-level taxon.The results of the present study also suggest that Heterandriini should be restricted to H. formosa and its closest relatives.The dataset arising from this study is not sufficient to address the phylogenetic relationships of H. formosa to other possible members of Heterandriini.Langerhans et al.(2005) and Hrbek et al.(2007) hypothesized that gonopodial characters are phylogenetically plastic, having evolved multiple times in the Poeciliidae under similar genetic and environmental circumstances.A prediction of such plasticity is that species of Poeciliinae should exhibit many derived changes in the morphology of the gonopodium, a condition that has indeed been observed.Therefore, in light of the high amounts of homoplasy observed in gonopodial characters, accurate tribe-level taxonomic designations of species must await more complete phylogenetic studies at the species level.
Appendix 2. Descriptions of characters and alternative states.Characters are grouped by functional group as follows: jaws, dentary, neurocranium, suspensorium, branchial arches, pectoral fin and girdle, dorsal fin, vertebrae and ribs, caudal skeleton, anal fin and gonopodium, scales and pigmentation, external morphology, and cephalic sensory system.
The distal arm of the maxilla present different shape at the tip, the character was considered square if it had a completely flat bottom, even if the tip was slender.If the shape of the tip was slightly or totally rounded, it was given character state 1.When the tip ended in a peak, even if it had round edges it was given character state 2. (Costa, 1998, ch.3): (0) present, (1) absent.

Notch on the dorsal portion of maxilla
A dorsal notch was present in most of the species analyzed and was shared by most of the Poeciliinae and Profundulidae.This character was a modification of character 7 in Costa (1998), who proposed that the shape of the dorsal process of the maxilla was continuous or curved.In this study, the size of the dorsal process was compared with the size of the ventral process.

Epiotic processes in adults
Ghedotti recognized three states for this character, in the present study was included another state, a medium size state, when the epiotic processes reach the first vertebra but do not go beyond it (referred to as "long" by Lucinda & Reis, 2005).According to Ghedotti (1998), this character was not a synapomorphy for Anablepidae because some species of that family lack epiotic processes.Ghedotti (1998) stated that the epiotic processes in Fluviphylax obscurum were absent, while in this study the specimen of F. pygmaeus examined possessed small epiotic processes (state 2).Lucinda & Reis (2005) stated that long epiotic processes were present in Jenynsia, Aplocheilichthys Phallichthys, Alfaro, Poeciliopsis, Quintana, and Carlhubbsia.According to Lucinda & Reis (2005), Priapichthys, Girardinus, Poecilia, Neoheterandria, Scolichthys, Xiphophorus, and Limia have medium-sized epiotic processes, while short processes are present in Heterandria and Pseudopoecilia.Ghedotti (2000) reported the absence of epiotic processes in Tomeurus and Cnesterodon.Lucinda & Reis (2005) (Ghedotti, 1998, fig. 4), ( 1) short.