Xiphophorus (Poeciliidae) from the Isthmus of Tehuantepec, Oaxaca, Mexico, with a Discussion of the Distribution of the X

Home , Coatzacoalcos River, Green swordtail, Poeciliidae, Xiphophorus

PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, NY 10024 Number 3441, 34 pp., 7 ®gures, 4 tables June 2, 2004

Two New Species of Xiphophorus (Poeciliidae) from the Isthmus of Tehuantepec, Oaxaca, Mexico, with a Discussion of the Distribution of the X. clemenciae Clade

KLAUS D. KALLMAN,1 RONALD B. WALTER,2 DONALD C. MORIZOT,3 AND STEVEN KAZIANIS4

ABSTRACT The swordtail, Xiphophorus clemenciae (Poeciliidae), has been considered a species of special concern because of its apparent limited range. Although described in 1959, it is of®cially still known only from three locations in the Rio Coatzacoalcos basin, Mexico. Zoogeographic studies have now shown that this species is widespread and abundant but restricted to the uplands of the Rio Coatzacolacos basin where it replaces in many areas the common swordtail, X. helleri. Two new swordtail taxa, X. mixei and X. monticolus, are described from headwater streams of the Rio Jaltepec, a major Rio Coatzacoalcos tributary, Oaxaca, Mexico. The new forms are sympatric in part of their range and replace both X. clemenciae and X. helleri. Morphometric and molecular analyses revealed that X. clemenciae and the two new species constitute a monophyletic clade that exhibits a closer evolutionary af®nity to the ``northern'' swordtails and the ``platy®sh'' group of the genus rather than to X. helleri and the other ``southern'' swordtails. The evolutionary relationships of these taxa are discussed.

1 Division of Vertebrate Zoology (Ichthyology), American Museum of Natural History ([email protected]). 2 Texas State University, Department of Chemistry and Biochemistry, Room 419 Centennial Hall, 601 University Drive, San Marcos, TX 78666 ([email protected]). 3 UT MD Anderson Cancer Center, Science Park±Research Division, Department of Carcinogenesis, P.O. Box 389, Smithville, TX 78957 (angel®[email protected]). 4 The Wistar Institute, 3601 Spruce Street, Philadelphia, PA 19104 ([email protected]).

INTRODUCTION ported from only four additional localities in the Isthmus of Tehuantepec, Rio Coatzaco- The genus Xiphophorus (Poeciliidae) com- alcos basin, Oaxaca, Mexico (Rosen, 1960; prises 24 recognized species of freshwater Ramirez, 1999). Although we and other re- ®shes of which 20 are restricted to the At- searchers have looked for this form in the lantic versant of the Sierra Madre of Mexico isthmus region during the ensuing 30 years, and adjacent parts of Central America. Most no additional locations had been discovered of the species (17) occur within 400 km to and the distribution of X. clemenciae had re- the north and south of the Trans-Mexican mained an enigma. Because of its apparent Volcanic Belt, suggesting that the genus restricted range and rarity it has been cate- evolved in this region. Only two species have gorized as a ``species of special concern'' a wide distribution: X. maculatus, which is (Groombridge, 1993). Three years later X. restricted to the coastal plain, occurs from clemenciae was listed as a taxon for which Veracruz to Guatemala (about 1000 km); and inadequate information was available to X. helleri, which is found in the front ranges make an assessment of its risk of extinction and coastal plain, extending from central Ve- (Baillie and Groombridge, 1996). We reini- racruz to Honduras (1100 km). tiated a search for this species in the upper Traditionally, the genus had been divided Rio Coatzacoalcos basin after 1993 in order into swordtails, taxa with a caudal appendage to determine its true status. We report here or ``sword'', and platy®shes without such a on the results of our exploration. structure, although the two terms have never been rigorously de®ned. As more species be- MATERIALS AND METHODS came known during the last 50 years, it be- came apparent that there exist ``swordtails'' SAMPLE COLLECTION,DEPOSITION, AND without a sword and ``platy®shes'' with a ZOOGEOGRAPHY: Specimens were collected in small caudal appendage. Therefore, the ver- the ®eld with a 10 ϫ 3 foot long seine with nacular and scienti®c usage of the two terms a 0.125 inch mesh. Fishes were usually pre- to describe and classify groups of Xiphopho- served in 10% buffered formalin and after rus can be misleading and confusing. A arrival in the laboratory transferred to 70% group of nine species, many with swords, in- ethyl alcohol, except for ®sh used in DNA habiting the Rio Panuco and Tuxpan systems studies (see below). At a number of locations (the ``northern'' swordtails) were shown to the ®sh were examined in the ®eld and then be monophyletic and clearly distinct from the returned to the stream, because permission to ®ve species with a caudal appendage (the collect was not obtained from the local au- ``southern'' swordtails) south of the Trans- thorities. Mexican Volcanic Belt (Rauchenberger et Names of rivers at ®sh localities (appendix al., 1990). These authors thought that one of 1) were taken from Carta Topogra®ca map the southern swordtails, X. clemenciae, could sheets (1:50,000 and 1:250,000 scale; Insti- be the closest relative of the northern sword- tuto Nacional De Estadistica Geogra®a e In- tail. The same report also suggested that the formatica; INEGI). It is noteworthy that river sword may have evolved more than once in names and their exact spellings are not al- the genus. Three reports (Meyer et al., 1994; ways consistent between different issues of Morizot and Siciliano, 1982; Meyer and the same map sheet or between different Schartl, 2002) raised the possibility that the sheets. On the 1988 edition of Carta Topo- ``southern'' swordtails may not be monophy- gra®ca, Donaji E15C43, the name Rio To- letic and that X. clemenciae is apparently un- losita is used for that section of the river related to the other members of the group downstream from the con¯uence of the Rio which we shall refer to as the helleri clade Jumeapa and Rio Tortuguero; however, on (X. helleri, X. alvarezi, X. signum, X. mayae). Carta Topogra®ca, Minatitlan E15±7, 1998 Xiphophorus clemenciae was described in edition, the river is referred to as Rio Juna- 1959 from small tributaries to the Rio Sara- pan along its entire length. One of the major bia near Rancho San Carlos, Oaxaca, Mexico headwater streams of the Rio Jaltepec is re- (Alvarez, 1959). Since then it has been referred to as Rio Ahuacatengo, San Juan Ma- 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 3 zatlan sheet E15C42, 1:50,000, 1984 edition. then digitally photographed with a Nikon The same river is spelled Rio Aguacatenco Coolpix 950 along with a ruler and a label on the Matias Romero sheet E15C53, 1988 depicting collection locality, individual edition, and Rio Aguacatengo on the Mina- numbers and such. All ®shes were measured titlan sheet E15±7, 1:250,000, 1998 edition. on both ¯anks. Fish images produced for There are numerous other examples. We ®gure 1 were obtained by ®lm-based pho- have endeavored to use the names and spell- tography using live ®shes and techniques ings of the most recent map issues. All dis- described elsewhere (www.xiphophorus.org/ tances given in the text are straight line mea- photography.htm). Morphometric measure- sures. Coordinate measurements were ob- ments were performed digitally using TPSdig tained by handheld units from Garmin software (F. James Rohlf, State University of (GPS38, GPS50, and GPS III Plus). Collec- New York at Stony Brook). Morphometric tion points were imported from Excel (Micro- values were computed in Excel (Microsoft) soft) spreadsheets using ArcView GIS (ver- and are average measurements for both sides sion 3.2, ESRI) with a background map de- based on 23 (21 for females) landmark data rived from the Digital Chart of the World se- points for each individual ®sh (®g. 2). Only ries (1:1,000,000 scale; www.gisdatadepot. fully mature males were photographed, as in- com). This map was extensively modi®ed us- dicated by the developmental stage of the ing ArcView GIS. gonopodium, a modi®ed anal ®n, and sec- Specimens were deposited at Universidad ondary sex character (Kallman and Schreib- Nacional AutoÂnoma de Nuevo Leon (UNL; man, 1973). De®nitions of morphometric San Nicolas de los Garza, NL, Mexico), variables are as previously published (Miller, Universidad Nacional AutoÂnoma de Mexico 1948) except for measurements of the sword, (UNAM; Mexico City, Mexico), and the a secondary male characteristic that results American Museum of Natural History in an extension of the ventral caudal ®n in (AMNH). We thank the Government of Mex- some members of the genus. Total sword ico for permission to collect specimens (Per- length is de®ned as the linear distance be- miso de Pesca de Fomento 200597±213±03, tween the ventral insertion of the caudal ®n and 150598±213±03 issued to Dr. Alexandra and the distal tip of the longest ray. Extended L. Basolo, and 200302.613.03.0147 issued to sword length is de®ned as the linear mea- Dr. Steven Kazianis). surement between the origin of the sword as MORPHOMETRIC ANALYSES: Morphometric it extends from the caudal ®n proper to the measurements were made on 105 preserved tip of the longest ray at the posterior end of female and male X. helleri (5 females, 6 the sword. All indexed values were comput- males, Rio Junapan; 7 females, 7 males, Rio ed by dividing raw measurements by stan- Sarabia; 5 males Rio Romay), X. clemenciae dard length. Average basic measurements, (7 females, 11 males, Rio La Polvora, trib- indexed values, and standard deviations are utary to Rio del Sol; 2 females, 3 males, presented in table 1 for females, and in table headwater of Rio Chalchijapan), X. mixei (6 2 for males. The averaged and indexed mea- females, 14 males, tributary to Rio del Sol surements were compared in a pairwise man- at Platanillo; 6 females, 5 males, Arroyo Cu- ner by use of the Tukey HSD (honestly sig- chara, tributary to Rio Aguacatengo), and X. ni®cantly different) ANOVA test as well as monticolus (3 females, 5 males, tributary to for standard length. Rio del Sol at Chahuitepec; 2 females, 3 GONOPODIAL STRUCTURE: Gonopodia were males, Rio Aguacatengo near Loma Sta. examined with the aid of a Zeiss Stemi Cruz; 4 females, 4 males, Rio San Miguel, 2000c microscope at 50ϫ magni®cation. In- tributary to Rio San Andres). The Rio La dividual gonopodia were excised near the in- Polvora site of X. clemenciae is the one clos- sertion point in the body wall. Because gon- est to the area of the Rio del Sol occupied opodia are three-dimensional structures with by X. mixei and X. monticolus. relatively intricate features, they were placed Fishes were mounted on white backing on glass slides and manipulated using for- paper with pins to enable accurate measure- ceps under the microscope. Particular atten- ments of ®n origins and termini, and were tion was placed on the terminal region which 4 AMERICAN MUSEUM NOVITATES NO. 3441

Fig. 1. Xiphophorus ®shes of the Coatzacoalcos River System. A, B. X. mixei female and male. C, D. X. monticolus female and male. E, F. X. clemenciae female and male. G, H. X. helleri female and male. I, J. X. maculatus female and male. 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 5

Fig. 2. Data points recorded for morphometric measurements. Drawing of a male X. clemenciae with morphological landmarks indicated (see text). includes bony elements derived from rays 3, lecular Probes, Eugene, OR). Working DNA 4a, 4p, and 5a, as these have been extensive- stocks were diluted to a concentration of 50 ly utilized in several Xiphophorus taxonomic ng/␮l. studies (e.g., Rosen, 1960; Rauchenberger et Two distinct nuclear genes were partially al., 1990; Meyer and Schartl, 2002). Figure ampli®ed by polymerase chain reaction 3 depicts some of the anatomical features and (PCR) and sequenced: RAB27 and CCND1 unique nomenclature used for this structure (also known as Cyclin D1). For RAB27, we (Gordon and Rosen, 1951). For photographic ampli®ed an 851-bp genomic region (as de- purposes, gonopodia were ®rst positioned termined in X. maculatus strain Jp 163 A; and then pressed carefully using a glass data not shown) including part of the second block. A Zeiss Axiovert 35 microscope with (primer RAB-EX2F; gctcctggcgctcgggcactc) Hoffman contrast ®lters and overhead light- and fourth exons (primer RAB-EX4R.1; ing was utilized as well a C-mount Minolta gctgattggtcaagtcgaaca). Twenty-nine individ- Maxxum 9xi camera and Kodak TechPan uals were utilized in the analysis resulting in ®lm. Exposure and processing parameters 29,679-bp of sequence. CCND1 was ampli- were optimized for high contrast. Gonopodia ®ed across the ®rst two exons and was rep- were examined from eight X. clemenciae, resented by a 1554-bp amplicon (as deter- seven X. mixei, and seven X. monticolus mined in X. maculatus). Forward primer males (see table 3 and ®g. 4). Individuals of (primer D1F9; atggaggagcagctgctgtgc) was X. helleri (®g. 3) and laboratory-raised X. coupled with the reverse primer (D1R5; maculatus and X. montezumae were also ex- cctggtggcctccactga). A total of 42,254 bp amined for comparison. representing CCND1 were sequenced from DNA ISOLATION,AMPLIFICATION, AND SE- the same 29 ®shes ampli®ed for RAB27. This QUENCING: Field-collected ®shes were ®rst overall strategy produced genomic sequences over-anesthetized in 0.06% MS-222 (tricane that could be unambiguously considered ho- methane sulfate) and then ®xed in 95% eth- mologous among taxa since conservation in anol within 15 ml plastic tubes (Falcon; Bec- exonic regions was high. Simultaneously, ton Dickinson, Franklin Lakes, NJ). This so- these sequences also yielded the needed di- lution was discarded and replenished after 8± vergence in intronic regions. PCR was per- 12 hours to ensure proper preservation of formed using an Expand Long Template DNA. High-molecular-weight DNA was later PCR kit (Roche, Basel, Switzerland), which extracted from these samples using the Pur- also provided additional proofreading (cal- egene isolation kit (Gentra systems, Minne- culated by the manufacturer at ϳ1 misincor- apolis, MN). Samples were resuspended in poration per 4.8 ϫ 106 bases). PCR was per- 1ϫ Tris-EDTA (TE; pH ϭ 8.0) and then formed using Hybaid PCR-Express thermal quantitated using a FLx800 (Bio-Tek, Wi- cyclers (Thermo-Hybaid, Middlesex, UK) nooski, VT) ¯uorometer and PicoGreen (Mo- and 50 ng of genomic DNA. Identical cy- 6 AMERICAN MUSEUM NOVITATES NO. 3441

TABLE 1 Raw (mm) and Indexed Morphometric Measurements for Female Individuals of Four Xiphophorus Species

cling parameters were used for both studied quencing) ¯uorescent thermal-cycle sequenc- loci as follows: 92Њ/2Ј followed by 10 cycles ing utilizing internal oligonucleotides (RAB- of 92Њ/10Љ,60Њ/30Љ,68Њ/2Ј, and 20 cycles of INT2R.3 [cgttcctccaggctctaactt] and RAB- 92Њ/10Љ,60Њ/30Љ,68Њ/2Ј (with a 20Љ increment INT2FA [ttaccatctacaactctttaa] for RAB27, added per cycle). Samples were incubated at and D1F14 [cgtgccagaagcctattac] and D1R12 68Њ for an additional 7Ј and then kept at 4Њ [gatttagttcgtagtcagttc] for CCND1) to gener- until further utilized. PCR amplicons were ate contiguous sequences with high accuracy. excised from 1.0% ethidium bromide-stained Alignments for each taxon were performed agarose gels (CCND1) or directly puri®ed using Sequencher software (version 4.1, (RAB27) using Geneclean-Spin (Q-Biogene, Gene Codes, Ann Arbor, MI) which enables Carlsbad, CA) and quantitated using the ¯uo- visualization and overlay of electrophero- rometric approach detailed above. These grams for more accurate base calling. All se- samples were sent for commercial (Davis Se- quences have been deposited into Genbank 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 7

TABLE 2 Ray (mm) and Indexed Morphometric Measurements for Male Individuals of Four Xiphophorus Species

under the following accession number spanning its entire experimentally deter- series: AY211356±AY211384 (RAB27) and mined east±west range, six X. monticolus AY211327±AY211355 (CCND1). from three locations and ®ve X. mixei indi- SAMPLE SELECTION AND PHYLOGENETIC RE- viduals from three sites. To ascertain the re- CONSTRUCTION: Nuclear DNA samples were lationships of X. clemenciae and the two extracted, ampli®ed, and analyzed from sev- newly described taxa to other Xiphophorus, en X. clemenciae collected at seven locations especially to the other ``southern'' sword- 8 AMERICAN MUSEUM NOVITATES NO. 3441

Fig. 3. A. Photomicrograph of the distal part of a gonopodium of a X. helleri (Rio Robalo). Scale bar is 0.5 mm. B. Close-up of gonopodium shown in panel A, showing anatomical nomenclature used in gonopodial descriptions (after Gordon and Rosen, 1951). Scale bar is 0.16 mm. 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 9

TABLE 3 Gonopodial Characteristics of the clemenciae Clade

tails, DNA samples were also analyzed from strapped (1000 replicates) neighbor-joining four X. helleri (four locations) and from one and parsimony methods as well as likelihood specimen each of X. signum, X. alvarezi, X. (100 replicates) techniques. Similar topologic andersi, X. couchianus, X. maculatus, and X. outcomes were obtained in all cases (as dis- montezumae (stocks derived from the Xiph- cussed below). Results derived from the ophorus Genetic Stock Center, San Marcos, pooled dataset are provided in ®gure 5. TX). Priapella intermedia was included as an outgroup and has been utilized as such in RESULTS several other studies (Marcus and McCune, Xiphophorus mixei, new species 1999; Meyer et al., 1994; Rosen, 1979). De- (Mixe swordtail) tailed information as to the geographical origin of these ®sh is included in the ®gure Figure 1A, 1B legend accompanying the provided pheno- DIAGNOSIS: A small to medium-sized gram (®g. 5). swordtail with a short sword (sword index Derived DNA sequences were aligned us- 0.06), one or two orange lateral stripes and ing the ClustalW program (Thompson et al., black spots in the base of the caudal ®n, 1994) built into the MacVector software suite closely allied to the two species below but (version 7.0, Accelrys, San Diego, CA) and differing from them by a shorter sword. exported into PAUP (version 4.0b10, Apple TYPE SPECIMENS: The holotype, Macintosh edition; Swofford, 2002). The da- UANL15253, is a mature adult male, 39.0 tasets derived from both studied loci were mm SL, 6 mm sword, collected on the 16th examined before and after pooling, by boot- of March, 2002, by Kazianis and Kallman in 10 AMERICAN MUSEUM NOVITATES NO. 3441

Fig. 4. Photomicrographs of the distal part of gonopodia of clemenciae-clade ®shes. A. Gonopodium of a X. clemenciae. B. Gonopodium derived from a X. mixei. C. Gonopodium excised from a X. monticolus. Scale bar for X. mixei is 0.5 mm and applies to all three images. Arrowheads depict the most distal spine of ray 3, which differs among taxa (see text). 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 11 a tributary to the Rio Cuchara (16Њ58.42ЈN, The next one above begins slightly more pos- 95Њ17.31ЈW), Rio Aguacatengo drainage, teriorly and may not extend as far posteriorly Oaxaca Mexico. The allotype, UANL15254, as the caudal ®n. A third still more dorsal is a mature female, taken together with the stripe is still shorter and fainter in coloration. holotype. The paratypes (IBUNAM 13100, A fourth stripe below the midlateral one is three females, seven males; UANL15255, only weakly developed. In females the orange one female) were collected on the 16th of pigmentation is even more subdued and the March 2002, by Kallman and Kazianis in a caudal ®n spots are absent. small stream at Platanillo (16Њ57.29ЈN, ETYMOLOGY: Named after the native in- 95Њ15.35ЈW), Rio del Sol drainage, Oaxaca, habitants, the Mixe, which live in villages Mexico. along the Rio del Sol and the Rio Jaltepec DESCRIPTION: Males with rather short headwaters (Lipp, 1991). The Mixe previ- swords projecting from 1 to 6 mm beyond ously occupied large sections of southern the posterior margin of the caudal ®n. The Mexico, speaking a language of the Zoquean short stubby sword tapers rapidly and is up- linguistic group, but by the mid-18th century turned. Ventral margin of the sword edged have become restricted on the Isthmus of Te- with black pigment cells. The central portion huantepec within areas close to and including or ®eld of the sword has an orange colora- the town of Guichicovi (Thomas and Swan- tion. The short free distal, dorsal margin of ton, 1911). the sword is usually unpigmented. The Grave REMARKS: A strain of X. mixei derived spot (Rosen and Kallman, 1969) is always from the upper Rio del Sol drainage at Sta. well developed, but does not obscure the ®n Cruz Chahuitepec has been maintained in the rays. Males always exhibit a well-developed Xiphophorus stock center for several gener- primary axillary stripe composed of mela- ations. The pigmentary and morphological nophores that runs from the base of the pec- traits that distinguish this species from the toral ®n past the gonopodium into the caudal other two forms have remained constant. peduncle. Some of the males exhibit a fainter Males (34.1 mean SL, range 24±42, n ϭ 70) second stripe that runs one scale row above of the laboratory stock were maintained for the primary one. One male has been seen that up to a year after attaining sexual maturation. had a third stripe below the primary one that Their swords (3.6 mm, range 1±8 mm) were ended in front of the gonopodium. Four to just as short as those of wild-caught males. nine black spots develop in the caudal ®n DISTRIBUTION: Headwaters of Rio del Sol, base. Rio Sarabia system and tributaries to Rio The terminal segment of ray 3 of the gon- Aguacatengo (Rio Jaltepec drainage) west of opodium is differentiated into a crescent- or Platanillo, Oaxaca, Mexico, Rio Coatzacoal- sickle-shaped hook and the blade is blunt dis- cos basin (®g. 6B; appendix 1). tally. The more distal serrae of ray 3 are an- MATERIAL: From Oaxaca: tributaries to Rio gled proximally (®g. 4B). The ramus (ray 4a) Aguacatengo UANL14341 (4), UANL14342 of the gonopodium ends bluntly in most cas- (4), UANL15255 (1), AMNH233383 (8), es; only rarely is the last segment of ray 4a AMNH211352 (2); from Rio del Sol IBUN- recurved over the blade. Proximal serrae on AM13100 (10) paratypes, AMNH211353 (4). ray 4p are rather thick near their base. Ter- minal segment of ray 5a produced into a Xiphophorus monticolus, new species claw which is about the same height as the (Southern Mountain Swordtail) distal serrae on the terminal segments of ray Figure 1C, D 4p. Counts for gonopodial features in rays 3, 4, and 5 are provided in table 3. DIAGNOSIS: This species is closely allied to The males always have from one to two the form above and can be distinguished subdued orange- to salmon-colored stripes from it by a longer caudal appendage (sword along the ¯ank, sometimes up to four. The one index 0.25), fainter orange stripes that fade along the midlateral line is best developed and in older individuals, and by the pronounced runs from the eye to the base of the caudal melanophore pigmentation that edges the ®n, where it connects with the Grave spot. dorsal and ventral sword margins (in males). 12 AMERICAN MUSEUM NOVITATES NO. 3441 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 13

TYPES: The holotype, UANL15256, is a (®g. 4C). Counts for gonopodial features in mature male, 50.0 mm SL, taken on the 16th rays 3, 4, and 5 are provided in table 3. of March 2002 by Kallman and Kazianis in ETYMOLOGY: The name is derived from the the Rio Cahuitepec (upper Rio del Sol). The Latin for a dweller in the mountains which in allotype, UANL15257, is a mature female, reference to the restricted range of the species 64.0 mm SL, taken with the male at the same to headwater streams inside the Sierra Madre. location. REMARKS: This species occurs sympatri- DESCRIPTION: A medium to large-sized cally with X. mixei in the Rio del Sol. An- swordtail with heavy black pigmentation. ecdotal evidence from the Rio del Sol sug- Usually two axillary stripes are present; the gests that it prefers deeper pools. We have more ventral one is always more strongly ex- witnessed a phalanx of nine native women, pressed. Beyond the gonopodium the stripe each holding each other's outer apron (spe- is not continuous, but may be produced in a cially designed for seining), and then diving series of spots. The midventral line is pig- to scour the bottom of the 2±3m deep river mented by melanophores and this pigmen- and coming up against the steep-sided bank. tation is especially heavy near the base of the The ®sh in their ``apronseine'' were mostly caudal ®n. From four to nine black spots are present in the base of the caudal ®n. The X. monticolus. In the Rio Aguacatengo near sword is usually well developed and thickly Loma Santa Cruz, which is a rapidly ¯owing edged with black pigment cells along ventral river with boulders and rif¯es, we have taken and dorsal margins. It is also straight and this species only. Only X. monticolus was faintly orange in color. The Grave spot is found in the Rio San Miguel drainage. very dark and obscures ®n rays. One to two A strain of this species derived from the faint orange lateral stripes may be present upper Rio del Sol has been maintained in the along the ¯ank but are absent in older indi- Xiphophorus Genetic Stock Center (South- viduals, resulting in an overall steel, blue- west Texas State University, San Marcos, gray appearance. The structure of the gono- TX) for many generations. The differences podium is similar to that of the preceding that separate this species from the other two species, including the most distal spines of ray forms have remained constant. Sword length 3, which also tend to be angled proximally of the laboratory strain (36.0 mm mean SL,

← Fig. 5. Phylogenetic tree of the clemenciae clade. Bootstrap values (%) are provided on the illustrated tree, which was derived from the parsimony approach (50% majority-rule consensus tree; consistency index 0.91, retention index 0.95, rescaled consistency index 0.87). Neighbor-joining bootsrap values are included ®rst at each node, followed by parsimony- and maximum likelihood-derived values. Unresolved nodes showing possible polytomies (bootstrap values under 50%) are shown with an asterisk. Source of specimens: P. intermedia: tributary of the Rio Polvora, Oaxaca, Mexico; X. signum: Rio Chajmaic, Alta Verapaz, Guatemala; X. alvarezi: Rio Dolores, Alta Verapaz, Guatemala (XGSC stock); X. helleri: no 1, Rio Sarabia stock, Oaxaca, Mexico (XGSC); no. 2, Jalapa stock, Rio Chachalaca system, Veracruz (XGSC); no. 3, Lancetilla stock, Honduras (XGSC); no. 4, tributary to Rio Aguacatengo at Santiago Tutla Nuevo, Oaxaca, Mexico; X. montezumae: Rascon, Rio Gallinas, Rio Panuco system, San Luis Potosi, Mexico (XGSC); X. maculatus: Jp 163 A strain, Rio Jamapa, Veracruz, Mexico (XGSC); X. andersi: Rio Atoyac, Veracruz, Mexico (XGSC); X. couchianus: Xc stock, Huasteca Canyon, Nuevo Leon, Mexico (XGSC); X. clemenciae: no. 1, Rio Verde, tributary to Rio Uxpanapa; no. 2, tributary to Rio Uxpanapa, 4 km NW of Los Amarillos; no. 3, Rio Romay, Rio Coachapa system; no. 4, tributary to Rio Sarabia at Rancho San Carlos; no. 5, tributary to Rio La Polvora, 2.5 km NW of San Juanita, Rio Sarabia system; no. 6, Arroyo El Boqueron, Rio Jaltepec system; no. 7, tributary to Rio Aguacatengo near Santiago Tutla Nuevo, Rio Jaltepec system; X. mixei: nos. 1±3 tributary at Platanillo, Rio Sarabia system; no. 4, Arroyo Cuchara, tributary to Rio Aguacatengo, Rio Jaltepec; no. 5, Rio Cuautepec at Santa Cruz Chahutepec, Rio Sarabia system; X. monticolus: nos. 1±3, three locations from the Rio San Miguel, Rio Jaltepec system; nos. 4,5, Rio Cuautepec at Santa Cruz Chahutepec, Rio Sarabia system; no. 6, Rio Aguacatengo at Loma Santa Cruz, Rio Jaltepec system. 14 AMERICAN MUSEUM NOVITATES NO. 3441 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 15 n ϭ 40) was 10.5 mm (range 5±18 mm), sig- cation near Rancho San Carlos, we have not ni®cantly larger than that of X. mixei. seen a single individual with this coloration. DISTRIBUTION: Upper Rio del Sol (Rio Sar- Nor have any ®sh with this coloration ap- abia drainage); upper Rio Aguacatengo near peared in the X. clemenciae maintained in the Loma Santa Cruz and Rio San Miguel (Rio laboratory for more than 25 years. The ver- Jaltepec drainage); all in Rio Coatzacoalcos nacular name is inappropriate, in our opin- basin, Oaxaca, Mexico (®g. 6B; appendix 1). ion. The distinguishing traits of this species MATERIAL: From Oaxaca: Rio San Miguel, have remained unchanged during this period. AMNH211358 (8); Rio Aguacatengo, The species is especially common in small UANL15258 (6); Rio del Sol, IBUN- streams and brooks with light current and AM13101 (5 paratypes). rockpools in areas with extensive limestone outcroppings, and it is relatively uncommon Xiphophorus clemenciae Alvarez, 1959 in fast-¯owing streams. Thousands of these (Yellow swordtail) ®sh were present in the small streams at La Figure 1E, F Guadalupe, Arroyo El Boqueron, Arroyo San DIAGNOSIS: This species has been previ- Antonio, Arroyo Pita, and an arroyo tributary ously described (Alvarez, 1959) and has been to the Rio Negro. Isolated pools with hun- utilized in several evolutionary studies (Mor- dreds of X. clemenciae were encountered izot and Siciliano, 1982; Borowsky et al., along the Rio Uxpanapa, Rio Grande, and 1995; Rosen, 1960; Meyer et al., 1994; Mey- Rio Sarabia and in the rock pool near San er and Schartl, 2002). Jose Morelos. The ®sh were uncommon in DESCRIPTION: Males display two to four in- the Rio Paquital, Rio Grande, and Rio Es- tense orange to red lateral stripes along the colapa. We do not think that these differences ¯ank. In males, six to nine deep orange spots re¯ect degrees of dif®culty in seining the are always present in the base of the caudal ®sh, because many X. monticolus and X. mix- ®n. These spots develop at the time of sexual ei were collected in fast ¯owing rivers. maturation. Coloration of the sword is either DISTRIBUTION: Rio Uxpanapa drainage, green or orange and edged with black pig- south of 17Њ17ЈN latitude; tributaries to Rio ment. The sword of mature males is well de- Coachapa at Vidal Diaz Munoz and Hidalgo veloped (sword index 0.25) and slightly up- Amayac, Veracruz; Rio Las Cuevas and Rio turned. Dorsal and ventral margins of the sword are edged with black. The gonopo- Paquital and tributaries near La Guadalupe, dium is similar to those of the preceding spe- Veracruz, Rio Chalchijapan drainage; Arroyo cies except the more distal spines on ray 3 El Boqueron, south of La Victoria, Oaxaca, point anteriorly, as opposed to being angled rill on west bank of Rio Aguacatengo just proximally (®g. 4A). Counts for gonopodial north of a bridge, near Santiago Tutla Nuevo, features in rays 3, 4, and 5 are provided in Oaxaca, Rio Jaltepec drainage; tributaries to table 3. lower Rio Junapan, from Santa Maria Mo- REMARKS: We cannot con®rm Alvarez's relos, Veracruz, north, and headwater streams description of these ®sh as having a deep yel- descending C. Pico de Aguila; tributaries to low coloration in addition to the brilliant or- Rio Sarabia from con¯uence with Rio El ange-red stripes. Among the hundreds of Corte to Rio Polvora and upstream along Rio adult males sampled by us over the entire Polvora, Oaxaca; Rio Verde, Rio Escolapa range of this species, including the type lo- and tributaries, Oaxaca; tributaries to Rio

← Fig. 6. Drainage map of the Rio Coatzacoalcos basin in the Isthmus of Tehuantepec, Oaxaca, Mex- ico. A. Overview of the study area within Mexico. B. Distribution of the clemenciae clade. Squares: X. clemenciae; triangles: X. mixei; stars: X. monticolus. The clade ranges from the continental divide as far north as the mountain front bordering the coastal plain. Note its absence from the southcentral part of the isthmus. Solid bars indicate the observed existence of waterfalls. 16 AMERICAN MUSEUM NOVITATES NO. 3441

Negro, Oaxaca, all above stations in Rio Co- ticolus possesses a rather narrow body and atzacoalcos basin (®g. 6B; appendix 1). caudal peduncle; X. clemenciae exhibits the MATERIAL: From Oaxaca: Rio Escolapa deepest body shapes while X. mixei is inter- drainage, UANL14333 (4), UANL14334 (4); mediate. Dorsal ®n base length is signi®- Rio Negro drainage, UANL14335 (4); Rio cantly shorter in X. mixei and X. monticolus Sarabia drainage, UANL14337 (4), than in X. clemenciae. Predorsal length is UANL15250 (10) AMNH211355 (8); from longest in X. mixei and differs signi®cantly Veracruz: Rio Junapan drainage UANL14336 from the two other species in males (only (4), UANL-14339 (4); from Rio Chalchijapan from X. clemenciae in females). The two drainage, UANL14338 (4), AMNH211354 (5). newly described species differ signi®cantly from X. helleri in body depth, length of dor- Xiphophorus helleri Heckel sal ®n base, and eye to ventral distance. Ad- (1848; Green Swordtail) ditional signi®cant differences exist in caudal Figure 1G, H peduncle depth and mandibular length indices in males and head length and orbital di- DIAGNOSIS: A medium to large swordtail ameter indices in females. with a long straight caudal appendage. Mid- The sword of X. mixei is signi®cantly lateral stripe may be dusky or brownish shorter than those of the other taxa. The (northern populations) or red. Two additional sword develops at the time of sexual matu- reddish stripes may be present above midla- ration and continues to grow for several more teral line and one beneath. Terminal segment months thereafter. The age of the mature of gonopodial ray 3 produced into a crescent- males caught in the ®eld is not known. It shaped hook and blade pointed distally (®g. could be argued that all the mature X. mixei 3). Ray 4a curves strongly backward over the males had just attained maturity and that, blade at an angle greater than 90Њ. Distal ser- therefore, the swords had not yet attained rae of ray 4p reduced in size and number and their ®nal size. This is unlikely because proximal serrae rather slender. Terminal seg- many additional X. mixei males that were ment of ray 5a produced into a claw, several caught in the ®eld and released all had short times larger than the distal serrae of ray 4p swords. Moreover, a stock of X. mixei has (®g. 3). been bred in the Xiphophorus Genetic Stock DISTRIBUTION: In coastal plain and front Center for several generations and males ranges. From Rio Nautla, Veracruz, Mexico, were maintained for at least one year past to northern Honduras (Rosen, [1960: ®g. 7A, maturity. All males developed short swords, appendix 1] provided data for the upper Rio adding con®dence to our conclusion from Coatzacolacos drainage). data derived from ®eld collections. The MATERIAL EXAMINED: From Oaxaca: Rio sword of X. clemenciae is also signi®cantly del Sol, IBUNAM 13102 (2); Rio Junapan, broader at its base than that of the other spe- IBUNAM 13103 (3), IBUNAM 31104 (8) cies, recalling its greater body and caudal pe- UANL15251 (3), AMNH 211356 (14); from duncle depth. Veracruz: Rio Coachapa, UANL15252. The two newly described taxa and X. clemenciae possess gonopodia that are similar COMMENTS in structure. These similarities are present in Within both sexes of the four species, sig- studied structures within rays 4a, 4p, and 5a ni®cant pairwise differences were discovered (table 3). There are, however, consistent dif- in indexed measures of body and caudal pe- ferences that concern the spines on ray 3. duncle depth, dorsal ®n base, predorsal The more distal spines in X. clemenciae are length, eye to snout and eye to ventral dis- facing anteriorly, whereas in all 14 speci- tance, head length, orbital diameter, and anal mens of X. mixei and X. monticolus the distal ®n length (tables 1, 2). Several of these char- spines are angled toward the more proximal. acters exclusively distinguished the three Because the most proximal spines of ray 3 taxa of the clemenciae clade. Body depth and are thick and angled toward the medial caudal peduncle depth indices are signi®- spines, in the two newly described taxa the cantly different for all three forms. X. mon- spines are produced into a somewhat trian- 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 17 gular structure whereas in X. clemenciae they present are adult males of the northern X. form a broader, more comblike feature. Ray helleri populations (Rio Nautla, Rio Chach- 4a (also called the ramus) usually ends blunt- alaca, Rio Antigua, Rio Jampa systems; Ro- ly (11 out of 14 gonopodia), but in three gon- sen, 1960; Meek, 1904). In the Xiphophorus opodia (one from each taxon) the small ter- Genetic Stock Center, a strain derived from minal segment angled sharply anteriorly, a the Rio Chachalaca shows an axillary stripe condition previously illustrated for X. cle- that is produced as a row of spots rather than menciae (Rosen, 1960). The last segment of a continuous band, and it does not extend ray 5a of all three species is differentiated into the caudal peduncle. The same condition into a small claw that is approximately of the is exhibited in a male from the Rio Jamapa same size as the distal serrae on ray 4p. as illustrated in ®gure 21A by Rosen (1960), and ®gure 22 by Zander (1967). Because PHYLOGENETIC RELATIONSHIPS none of our other data reveals a close phylogenetic relationship between the clemen- Several traits unite X. clemenciae, X. mon- ciae clade and X. helleri, we consider the ticolus, and X. mixei and set them apart from trait(s) as being independently derived or, al- the other swordtails. The orange lateral ternatively, the resemblance may be super®- stripes of the three species are produced by cial. According to our observations on the carotenoid pigment, whereas the pigment re- Rio Chachalaca stock, in which two kinds of sponsible for the red stripes of the helleri males occur, some with and some without clade is drosopterin (Kallman and Bao, this pattern (Zander, 1967), presence or ab- 1987). sence of this trait appears to be controlled by A second diagnostic trait for the three spe- a recessive sex-linked allele. Because no var- cies concerns the spots which develop in the iation is present within the clemenciae clade, proximal portion of the caudal ®n. Every no genetic analysis can be easily performed. adult male exhibits red (X. clemenciae)or F1 hybrids between X. clemenciae and X. hel- black (X. mixei, X. monticolus) spots, a trait leri (Sarabia stock) do not exhibit an axillary not seen in any other Xiphophorus. The black stripe. In many backcross hybrids to clemen- spots in the proximal portion of the caudal ciae the axillary stripe is poorly expressed ®n of X. mixei and X. monticolus develop (Kallman, unpubl. ms.), suggesting that the from the reticulum, bands of melanophores genetic control of the axillary stripe in the that border each scale pocket. When the two species may not be the same. Such a males are exposed to agents that cause me- situation would be analogous to the presence lanosomal (pigmentary) aggregation, orange of the Sc (spotted caudal) pattern in X. var- spots are revealed. Thus with respect to this iatus, a platy®sh, and X. cortezi, a northern character the difference between X. mixei and swordtail (Rauchenberger et al., 1990). X. monticolus on the one hand and X. clemenciae on the other is that in the two former MOLECULAR PHYLOGENY species the melanosomes are invariably in the dispersed state and obscure the orange A total of 71,933 total bases of genomic pigment cells, whereas in X. clemenciae the DNA sequences were derived from two nu- situation is reversed. clear genes (RAB27 and CCND1) in 29 ®sh- A third pigmentary trait is more dif®cult es. This approach yielded 2359 nucleotide to evaluate. It concerns the axillary stripe, a positions/characters that were conserved be- band of melanophores extending from below tween all taxa. In addition, 153 were variable the base of the pectoral ®n to beyond the characters and, of these, 57 were informative base of the gonopodium and into the caudal for parsimony analyses. Data analysis re- peduncle, one scale row above the midven- vealed that X. clemenciae as well as the new- tral line. It is present in all adult males of the ly described species X. mixei and X. monti- clade and expressed as a continuous zigzag colus are unambiguously monophyletic. This line (at least in front of the gonopodium) fol- conclusion is based on examination of phen- lowing closely the reticulum. The only other ograms/dendrograms derived from phenetic Xiphophorus in which a similar pattern is (neighbor-joining), parsimony, and maxi- 18 AMERICAN MUSEUM NOVITATES NO. 3441

TABLE 4 Jukes and Cantor Corrected Distance Matrix Between Exemplative Individuals

mum likelihood methods of phylogenetic re- newly described taxa, are similar to distances construction. A phylogenetic tree is provided between X. montezumae and X. maculatus (®g. 5) which shows bootstrap values derived for example, taxa of ``northern'' swordtail from each of the three algorithmic methods. and platy®sh clades, respectively (table 4). Using the parsimony method, strict consen- The position of X. clemenciae itself within sus of 1000 reconstructed trees revealed ab- phylogenetic trees involving members of the solute support for monophyly between X. genus has been the topic of debate in the lit- clemenciae and the newly described taxa erature (Morizot and Siciliano, 1982; Rosen, (data not shown). X. clemenciae individuals 1979; Meyer et al., 1994; Borowsky et al., utilized in the analysis were speci®cally cho- 1995; Meyer and Schartl, 2002). Although sen to provide coverage of its entire geo- this study does not completely address the graphic range. It is noteworthy that X. cle- exact taxonomic relationship of X. clemen- menciae individuals show more divergence ciae and the two newly described species from each other than do those of X. mixei within the genus Xiphophorus, it clearly and X. monticolus. This phenomenon is prob- shows with statistical con®dence that these ably associated with the geographic ranges of species are not closely related to the helleri these two newly identi®ed species, which are clade. Undoubtedly, further inclusion of oth- relatively restricted compared to that of X. er Xiphophorus taxa and further data gener- clemenciae, although this assumption is not ation will address this research topic. proven by the data. Support for monophyly of X. helleri, X. The clemenciae clade is also clearly more alvarezi, and X. signum is strongly supported related to those included platy®sh (X. ma- by the data. The recently described X. mayae culatus, X. couchianus, and X. andersi) spe- was not available and not included in the cies and the included northern swordtail (X. analysis, but presumably would also show montezumae) taxon than to X. helleri, X. al- close phylogenetic relationship to the varezi, and X. signum, as evidenced by boot- ``helleri clade'' taxa (Meyer and Schartl, strap values (100%, 99%, and 97%, depend- 2002). The common green swordtail, X. hel- ing on the method of reconstruction; see ®g. leri, has the widest distribution of any Xiph- 5). Evidence for X. mixei and X. monticolus ophorus. The three other members of the hel- being distinct from other Xiphophorus taxa leri clade are found in Guatemala, suggesting is supported by the phylogenetic reconstruc- that some populations in the eastern part of tion depicted in ®gure 5, which shows strong its range diverged into new forms. The sim- support for monophyly between all included ilarity of the DNA sequences from both gene individuals, regardless of geographic posi- loci of X. alvarezi and the Lancetilla stock tion. The genetic distance between species in of X. helleri argues for a recent divergence the clemenciae clade, which includes the from a common X. alvarezi/X. helleri-like 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 19 ancestor. Moreover, the apparent closer phy- swordtails (®g. 4). In the clemenciae group logenetic relationship of X. alvarezi, and per- the ramus (ray 4a), found just posterior to the haps also X. signum, to some (eastern) X. hel- hook and in direct contact to it, is usually leri populations than to others (Jalapa) has blunt at the distal tip or barely curved back- interesting taxonomic implications. If a sim- ward, a condition closely resembling that in ilar relationship should be revealed by anal- the northern swordtails. The distal section of yses of additional characters, then what is ray 4a of all three studied species is com- now known as X. helleri is really a paraphy- prised of segments that usually show little or letic assemblage of populations. This recalls no taper anteriorly. In contrast, the ramus in an earlier description of the Jalapa popula- the helleri clade is always strongly curved tion of swordtails as a separate species backward distally at an angle of greater than (Meek, 1904). In addition, extremely high 90Њ and this portion consists of from three to levels of allozyme diversity within and six robust segments (®g. 3 shows one such among the helleri clade was previously re- gonopodium). In addition, ray 4a distally ported (Morizot and Siciliano, 1982), provid- shows a wavy convex taper in the helleri ing additional evidence for genetic differen- clade that is not observed elsewhere in the tiation within this group of ®shes. genus (®g. 3). In all examined individuals of Undoubtedly, further examination of the X. clemenciae, X. mixei, and X. monticolus, clemenciae and helleri clades and all Xiph- the last segment of ray 5a is produced into a ophorus taxa will be required to fully address claw, but there was never a pronounced size their exact evolutionary positioning. This difference between it and the distal serrae of study, however, does provide robust support ray 4p. In most instances, the claw failed to that X. clemenciae and the two closely relat- protrude above the distal serrae altogether. ed species, X. monticolus and X. mixei,do This condition again resembles that in some not belong to the evolutionary grouping that members of the northern swordtail group. By includes X. helleri, which supports the con- contrast, the claw in the helleri clade is ro- clusions of other studies (Morizot and Sici- bust (®g. 3) and several times larger than the liano, 1982; Rosen, 1979; Meyer et al., 1994; distal serrae of ray 4p which are also reduced Borowsky et al., 1995; Meyer and Schartl, in number (Rosen, 1960). This condition is 2002). not present in any other Xiphophorus taxon. Of the nine species of Xiphophorus conven- DISCUSSION tionally called platy®sh, only one exhibits a small claw and a second species, X. macu- Xiphophorus mixei and X. monticolus are latus, is polymorphic for presence or absence more similar to each other in gonopodial of the claw in one population (Rosen, 1960). structure than to X. clemenciae, but the overall similarly in the shape of the gonopodia of TOPOGRAPHIC SETTING the three species is striking and sets them apart from all other Xiphophorus. Such over- The Isthmus of Tehuantepec is the narrow all similarities in gonopodial structure be- waist of Mexico where the distance from the tween closely related Xiphophorus species Gulf of Mexico in the north to the Paci®c have been previously reported (Rosen, 1960; Coast to the south is a mere 190 km. The Rauchenberger et al., 1990). As an example, continental divide is located 35 km north of the nine species comprising the monophylet- the Paci®c Ocean. The very center of the ic northern swordtail group predominantly isthmus, roughly followed by Mexican Route differ only in the size of the claw derived 185 (trans-isthmian highway) and by the Rio from ray 5a and the size of the distal serrae Coatzacoalcos and Rio El Corte, is also an of ray 4p (Rauchenberger et al., 1990). The area of low elevation throughout. Altitude distal segment of ray 3 in the swordtails of immediately north of the divide is 220 m the helleri clade is produced into a crescent- (plain of Chivela). There are rather few plac- shaped hook (®g. 3), while in the three spe- es in the center of the isthmus that reach an cies of the clemenciae group it is a more elevation of 300 m above sea level. The bar- elongated structure resembling the northern rier ranges at the low section of the conti- 20 AMERICAN MUSEUM NOVITATES NO. 3441 nental divide extend approximately 30 km south and east. Shortly before these rivers east to west and 10 km north to south and reach their base level and assume an easterly they vary from 300 to 500 m in altitude, but course toward the Rio Coatzacoalcos and Rio there are several low passes through them. El Corte, they descend over major waterfalls Towards the east and west-northwest the di- and cascades (Rio Malatengo just south of vide gains rapidly in altitude with some Mogone; Rio del Sol downstream from peaks reaching 2000 m and above. About Guasamann; Rio Junapan downstream from 100 km to the east, the Sierra Espinazo del Monte Aguila). Diablo, a steep mountain range exceeding There are a few areas of karst outcrop- 1000 m in elevation and composed of cre- pings west of the Rio Coatzacoalcos. They taceous limestone, runs north-south and sep- occur along the base of the high ground that arates the Rio Coatzacoalcos basin from the separates the Rio El Corte and Rio Junapan, Rio Grijalva drainage. on the north side of the mountain front that On the west the low section of the isthmus faces the Rio Jaltepec (there is also a row of is bordered by the eastern limb of the Sierra karst towers on top of this plateau) and near Madre del Sur which rises steeply approxi- the con¯uence of the Rio San Andres and mately 7 km west of the trans-isthmian high- Rio Aguacatengo to form the Rio Jaltepec way and extends in a north-northwesterly di- and in the intramontane valley of the Rio San rection for about 75 km as far as the Rio Miguel, just north of the continental divide Jaltepec. There are several high points well at Guevea de Humboldt. This last river me- north of the divide. Cerro el Borrego, 10 km anders in a northwestern direction for 17 km southwest of Matias Romero, peaks at 1150 until, according to the map, it disappears into m and the Cerro Pico Del Aguila-Cerro the ground at El Boqueron (elev. 360 m). Verde massif, 30 km northwest of Matias From there a narrow gorge at 400 m eleva- Romero, ranges from 800 to 1120 m. The tion between Cerro El Centinela (elev. 1300 mountains gradually descend to the 100±150 m) and Cerro El Otate (elev. 750 m) leads to m level, roughly along a line westward from the Rio San Andres (elev. 350 m) about 1.5 Palomares near the Rio Sarabia to the Rio km west of El Boqueron. Although we have Jaltepec, but then the land rises again to 300 not visited this site, most likely there was a m before dropping off sharply to the Rio Jal- direct connection between the two rivers in tepec, about 25 km southwest of Matias the past. Romero. This northernmost rise may be more The eastern part of the isthmus (east of the orogenically related to the ridge composed of Rio El Corte) from the divide north for about cretaceous slate between the Rio Junapan 75 km consists of a series of folded ridges and Rio Coatzacoalcos toward the east. It is and valleys that trend northeasterly, then lined up with the mountain front bordering easterly and ultimately southwesterly. A the coastal plain to the east, but not with the prominent massive Paleozoic granite intru- front of the Sierra Madre del Sur to the west. sion, the jagged Sierra Tres Picos, exceeds Its maximum height of slightly above 300 m 1000 m, but most ridges are considerably corresponds well with similar maximum el- lower. North of the Sierra Tres Picos the ridg- evations to the east (see below). es are formed of Oligocene and Eocene sand- The Sierra Madre del Sur is being dis- stone and are in general higher toward the sected by the Rio Malatengo, Rio Sarabia, south and east. A trellislike network of small Rio Aguacatengo and Rio San Andres and rivers dissects and erodes the ridges since the their tributaries. The rivers ¯ow from the Pliocene south of the Sierra Tres Picos and continental divide in a north-northeasterly di- since Quaternary times to the north (Geissert rection in steepsided valleys, 100±400 m Kientz, 1999). Note that the Rio Uxpanapa, above their base level, except for the Rio Rio Verde, and Rio El Corte and several Malatengo, which hugs the eastern front of smaller streams ¯ow west or east for consid- the sierra. The Rio Junapan originates from erable distances before turning north. The Si- streams that descend from Cerro Pico Del erra Espinazo Del Diablo, 3±7 km northeast Aguila to the east and from the plateaulike and east of the Rio Uxpanapa, forms the east- mountain front along the Rio Jaltepec to the ern limit of the Rio Coatzacoalcos basin. 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 21

The mountain front meets the Atlantic ocks between 60 and 80 m in elevation above coastal plain approximately 80 km south of sea level. All the major streams run at an the town of Coatzacoalcos or roughly be- elevation of 40 m. The coastal plain gently tween 17Њ22ЈN and 17Њ29ЈN. Along most of slopes toward the Gulf. As far as 45 km in- the front the difference in elevation between land from the Gulf coast there are extensive coastal plain and upland is not very pro- swamps, lagoons, and oxbows along the Rios nounced, a mere 100±150 m. Only few plac- Coatzacoalcos and Uxpanapa. es near the front exceed 300 m in altitude Since Cretaceous times there was never a which then gradually slope down toward the sea passage across the Isthmus, although the south. Such high points are found just south emergent land mass must have been exceed- of the Rio Jaltepec and also south of the ingly narrow until the Pliocene (Durham et junction of the Rio Junapan and Rio El Corte al., 1955). Neocene marine deposits have (a north±south ridge of cretaceous slate sur- been found south along route 185 as far as rounded by Miocene sandstone), near the Mogone along the Rio Malatengo (Durham source of the Rio Cahuapan and just west of et al., 1955). Marine ostracods of mid-Plio- the Rio Uxpanapa (Cerros La Numeracion). cene age have been reported from a locality The elevated terrain along the mountain 15 km north of Jesus Carranza (Michain- front descends gently toward the south for Castillo, 1985), marine foraminifera of lower about 20 km as far as the Valle de Uxpanapa, Pliocene age near Jesus Carranza (Akers, contrary to the overall prevailing dip in the 1984), and late Miocene marine fossils from isthmus, which is from south to north (Dur- a site 3 km south of the Rio Jaltepec (Dur- ham et al., 1955). Several small streams de- ham et al., 1955). It appears that until the scend from the high ground along the moun- Pliocene the Gulf of Mexico extended as far tain front in a southerly direction until they at the present day mountain front. join the major rivers. The term ``Valle de Uxpanapa'' does not refer to the valley of the DISTRIBUTION Rio Uxpanapa. This term is applied to the 5± 10-km-wide level valley that extends at an Xiphophorus clemenciae is the most wide- elevation of 100 m from the Rio El Corte at spread and common Xiphophorus in the up- Cuauhtemoc for 60 km eastward to the Rio lands of the Rio Coatzacoalcos basin (®g. 6). Grande. It then continues southwesterly for It is absent from the coastal plain. Like vir- 30 km along the Rio Uxpanapa at an eleva- tually all other Xiphophorus, it is a species tion of 50 m. The Valle de Uxpanapa shows restricted to the Sierra Madre. Between the the same trend as the ridges: northeasterly Rio El Corte and the Rio Uxpanapa, the then southwesterly. northern limit of its distribution coincides Wherever the sandstone ridges have been closely with the mountain front facing the eroded away between the mountain front and Gulf coastal plain. This species was not tak- the Sierra Tres Picos, cretaceous limestone is en in any of the small streams that descend exposed, often present as karst towers and the mountain front to the north, but it was sinkholes (Geissert Kientz, 1999). The Valle abundantly present in every stream sampled de Uxpanapa has a solid ¯oor of limestone south of it. Conversely, X. helleri ranges over covered by a thin layer of soil. The major the coastal plain (see locality records listed rivers in this region have cut down close to by Rosen, 1960) as far as the mountain front the base level. Elevation of the Rio Uxpan- but not beyond (®g. 7A). Only in the Rio apa at its junction with the Rio Grande is 40 Romay and Rio Cuahapan, streams that arise m, and further upstream near 17Њ03ЈN it runs a short distance south of the front and ¯ow at 100 m. Along latitude 17Њ10ЈNÐroughly north, did the two species occur sympatri- along the Valle de UxpanapaÐelevation of cally. This area has been severely eroded and the Rio El Corte is 40 m, the Rio Chalchi- can be considered as a southward extension japan is 60 m, and the Rio Solosuchil is 70 of the coastal plain. m. West of the Rio Coatzacoalcos the distri- The coastal plain just north of the moun- bution of X. clemenciae is rather spotty. In tain front consists of uncountable low hill- the very center of the isthmus, X. clemenciae 22 AMERICAN MUSEUM NOVITATES NO. 3441 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 23 is found from near the mountain front to m) near Santiago Tutla Nuevo, not far from about 25 km south of it along some of the where the Rio San Andres and Rio Aguaca- tributaries to the Rio Sarabia and Rio Juna- tengo join. These two clemenciae sites are 25 pan. At each station it occurs sympatrically km apart. A straight line drawn along the Rio with X. helleri. Yet farther south in the low Sarabia where it enters the Sierra, and con- center of the isthmusÐthe area drained by tinued to the Tutla Nuevo site, will pass the Rio Malatengo, Rio Grande, and Rio through all clemenciae locations in the upper ChichihuaÐX. clemenciae is absent (19 sites Rio Sarabia and Rio Junapan. sampled), but X. helleri is ubiquitous. X. cle- Xiphophorus monticolus and X. mixei re- menciae is found in small tributaries of the place X. clemenciae at higher elevation in the Rio Sarabia as far upstream as the Rio del interior section of the Sierra Madre del Sur. Sol and Rio La Polvora, but it does not reach They occupy headwater streams of the Rio above the waterfall and cascades downstream Jaltepec (Rio Aguacatengo, Rio San Miguel) from Guasamann. At all but one location, it and Rio Sarabia (Rio del Sol) drainages. X. occurred together with X. helleri, but there mixei is abundant in the Rio del Sol, which were many stations where only X. helleri was is a fast-¯owing river with rif¯es and boul- found. ders, and in all tributaries that ¯ow into it Along the lower Rio Junapan we found X. from the mountain spur that separates the Rio clemenciae present in three tributary streams del Sol from the Rio Malatengo. The new (®g. 6) and Ramirez (Ramirez, 1999) listed forms of Xiphophorus are also present in an additional station in the Rio Tortuguero, small tributaries on both sides of the low where we caught only X. helleri. From then Sol±Aguacatengo divide near Platanillo and on X. clemenciae was not encountered in this one Rio del Sol tributary east of the divide. drainage until the headwaters (Rio El Roba- Elevation of collecting sites ranged from 500 lo) that descend from Cerro Pico del Aguila. to 380 m. One station is approximately 2.5 km NW of We have not found the two species in a the X. clemenciae site in the Polvora drain- tributary near Guasamann, just before the age across a narrow ridge. A member of the Rio del Sol spills over a waterfall and de- clemenciae clade also occurs in the Rio Jun- scends 160 m over a distance of 5 km to join apan at Monte Aguila, but because only the Rio Polvora. This cascade appears to be small immature ®sh were collected, we could the barrier between X. monticolus and X. not identify the species. Both streams were mixei in the uplands and X. helleri and X. also sampled repeatedly further downstream clemenciae in the downstream section of the near route 147 over a period of nearly 10 Rio del Sol/Rio Sarabia drainage. Note also years, but only X. helleri was found. that three additional genera of poeciliid ®sh- Only 2 of 27 sites in the Rio Jaltepec es (Priapella, Poecilia, and Poeciliopsis)do drainage yielded X. clemenciae. It was abun- not occur above the waterfall, although they dant throughout the Arroyo El Boqueron (not are common below (®g. 7B; see appendix 1). to be confused with El Boqueron in the Rio There appears to be a narrow belt in the re- San Miguel drainage near the continental di- gion between the upper and lower Rio del vide) which descends the mountain front to Sol where no Xiphophorus is present. In ad- the Rio Jaltepec, and in a small rill (elev. 100 dition to the tributary at Guasamann, no

← Fig. 7. Drainage map of the Rio Coatzacoalcos basin in the Isthmus of Tehuantepec, Oaxaca., Mexico. A. Collection sites of X. helleri (circle) and X. maculatus (cross). X. helleri occurs in the coastal plain (distribution not shown) as far south as the mountain front, except in the central part of the isthmus, where its range extends to the continental divide. X. maculatus occurs in the coastal plain as far as the mountain front. Note that the distributions of X. helleri and the clemenciae clade (®g. 5B) are complementary to each other. B. Collecting sites of P. intermedia (circles). The ``lightning'' symbol represents sites where no Xiphophorus was collected. The distribution of P. intermedia coincides with that of X. helleri (®g. 6A) and X. clemenciae (®g. 5B). Solid bars indicate the observed existence of waterfalls. 24 AMERICAN MUSEUM NOVITATES NO. 3441

Xiphophorus is present in the Rio San Juan helleri is common, but members of the cle- Viejo, which drops 400 m over a distance of menciae clade are absent. The only stream 5 km and joins the Rio del Sol just below where X. helleri occurs at a higher elevation the waterfall. The Rio Guajalote (Rio Mala- is a tributary (elev. 500 m) of the Rio Mal- tengo tributary), which ¯ows in the opposite atengo at Zacatal that descends the steep direction 3 km east of San Juan Viejo also eastern limb of the Sierra Madre. lacks Xiphophorus. A similar situation may In our interpretation, the clemenciae-clade exist in the Rio Aguacatengo and Rio San ancestor occupied the old uplands of the Co- Andres area, but here our knowledge is less atzacoalcos basin and was the only Xipho- complete because of the dif®culty of access phorus in this region. Since then there ap- to the region. In the foothills of the Sierra pears to have been renewed uplift of the Si- Madre del Sur near the area where the Rio erra Madre, as indicated by the steep valleys Aguacatengo and Rio San Andres join, X. and major waterfalls which isolated the ®sh helleri is common. However, a short distance in the headwaters from those farther down- inside the Sierra two tributaries to the Rio stream. At present the X. clemenciae popu- San Andres, Rio Coyol (sampled at 450 m lation of the Rio Polvora near its junction and 320 m) and Rio Ovo (380 m) near San with the Rio del Sol is the closest one to X. Juan Mazatlan, lacked Xiphophorus. These mixei and X. monticolus. The two species streams descend the divide that separates the very likely evolved in different upland rivers Rio Aguacatengo to the east and the Rio San which subsequently were joined by stream Andres to the west. As in the Rio Sarabia± capture. Northeast of Gueva, headwater Rio del Sol drainage, Priapella, Poeciliopsis, streams of the Rio del Sol, Rio San Miguel and Poecilia are not present in the Rio Agua- and Rio Aguacatengo come in close prox- catengo and Rio San Andres headwaters. imity on the same mountain spur. The Rio In the uplands of the Rio Coatzacoalcos del Sol could have been a tributary to the basin, X. helleri is replaced by members of Rio Aguacatengo west of Platanillo where the X. clemenciae clade. This recalls the re- the ridge between the two rivers is rather lationship of X. helleri with X. signum and low. Subsequently, the Rio del Sol could X. alvarezi (and perhaps also X. mayae)in have been captured by a major Rio del Sol Guatemala where X. helleri is found in the tributary (called Rio Cuchara, but not to be coastal plain and along the front ranges, confused with the stream of the same name whereas the other two forms inhabit interior that joins the Rio Aguacatengo) which de- drainages, many of which are isolated (Ro- scends Cerro de Pico Aguila to the south and sen, 1979). East of the Rio El Corte the joins the river upstream from Guasamann. X. mountain front separates X. helleri from X. helleri arrived at a later time after the emer- clemenciae. X. helleri occupies the streams gence of the coastal plain, and its dispersal along the northern edge of the front (®g. 7A), was along the mountain front and coastal and according to the locality records listed plain. Because the mountain front east of the by Rosen (1960) this is the only swordtail Rio El Corte rises abruptly from the coastal species that occurs in the coastal plain of the plain and dips south and few rivers pass Rio Coatzacoalcos basin. through it, it was a barrier to the southward In the western range of the Rio Coatza- dispersal of the green swordtail. It penetrated coalcos basin, the distribution of the two spe- farther south in the low part of the isthmus cies is more complex. X. helleri is common along the major rivers Rio Junapan, Rio Sar- in the tributaries of the Rio Jaltepec, Rio Jun- abia, Rio Malatengo, and Rio El Corte. This apan, and Rio Sarabia that descend the foot- geologic scenario leaves unanswered the hills, but it is absent in the interior valleys. questions as to why X. clemenciae is absent X. clemenciae, which shows a spotty distri- from the central southern section of the isth- bution along the three rivers, is always sym- mus and why X. helleri did not penetrate far- patric with X. helleri except in Arroyo El Bo- ther south in the Rio Uxpanapa. During Pli- queron. In the low southern central part of ocene and Quaternary times stream systems the isthmus, which is drained by the Rio in this highly tectonically active region may Malatengo and its two major tributaries, X. have been considerably different. 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 25

In view of the distribution of X. clemen- divide (300 m) between the Rio Chichihua ciae, the failure of earlier attempts to deter- and a tributary to the Rio Escolapa, X. helleri mine the true status of this form now be- is present in a swampy area, whereas 2.5 km comes apparent. Ichthyologists including away in the Arroyo San Antonio (250 m) X. ourselves traveled with vehicles along the clemenciae is common. Tributaries of the Rio Trans-isthmian Highway and collected ®shes Malatengo that drain the eastern limb of the in streams close by. With the exception of Sierra Madre southwest (Arroyo La Canada, three locations, however, the highway passes 300 m) and northwest (Rio Los Nanches at through the only region of the isthmus where Zacatal, 500 m) of Matias Romero are oc- the species is absent. The collectors failed to cupied by X. helleri, but in the streams that go into the backcountry to the east and west. drain 5±7 km to the west into the Rio del Sol Xiphophorus clemenciae is widespread only X. mixei and X. monticolus are found. and abundant and can be removed from the The sharp separation of X. helleri from mem- list of endangered and threatened Xiphopho- bers of the clemenciae clade over a wide area rus. However, X. monticolus and X. mixei oc- suggests that this distributional pattern is of cupy limited ranges; as long as the headwa- long standing. ters of the Rio Sarabia and Rio Jaltepec north The complementary distribution of X. hel- of the continental divide remain intact, the leri north and of X. clemenciae south of the two species will not be threatened. Other mountain front east of the Rio El Corte might species of Xiphophorus that are in danger of also suggest that the two species occupy sim- becoming extinct are the three northernmost ilar niches and cannot exist together. We re- species (X. meyeri, X. gordoni, X. couchi- ject this idea, because in the Rio Sarabia and anus) in the Chihuahuan desert. These are Rio Tolosita drainages the two species are forms that are restricted to a few small iso- broadly sympatric. It is also possible that the lated spring pools. X. couchianus is currently two forms prefer different habitats. We found considered extinct in nature, but this deter- X. clemenciae usually associated with lime- mination is pending further exploration of stone habitats. This may be merely a re¯ec- proximal localities to the type collections. A tion of the geological condition of this spe- fourth species that we think is in danger is cies' distribution and not its cause, but on a X. andersi, which is restricted to the small very local level we found a few sites where watershed of the Rio Atoyac above the cataracts at Atoyac, Veracruz. Although located X. clemenciae seemed to prefer limestone in a region of high annual rainfall, diversion habitats and X. helleri did not. The streams of water for agricultural purposes and indus- (16 sites sampled) that drain the high ground trial pollution from sugar mills seriously de- 20 km southwest of Matias Romero along the grades the habitat. Similarly, X. continens,a Rio Jaltepec are occupied by X. helleri with species of very narrow distribution could the single exception of the Arroyo El Bo- also be considered threatened, requiring fur- queron. This is the only stream in this region ther study. with massive limestone outcroppings, and here X. helleri was absent and X. clemenciae abundant. A second site where we found ECOLOGY such a separation was in two small streams In many places X. helleri and members of that ¯ow into the Rio Sarabia southeast of the clemenciae clade exist in close proximity, Palomares. One stream with muddy bottom but in different watersheds across rather mi- and decaying leaves sampled repeatedly nor barriers. A 10±km wide section of lime- since 1982 contains only X. helleri whereas stone country separates X. helleri in the Rio the other stream, one km distant and 20 m Poza Azul from X. clemenciae in the Rio lower in a ravine with limestone bottom and Verde (Rio Uxpanapa system). No swordtail decaying leaves, is occupied only by X. cle- is present in the headwaters of the Rio Za- menciae. A third such site is the ravine catal, but 6.5 km distant in a small stream southeast of Jose Maria Morelos. In the small that ¯ows south, X. clemenciae occurs by the stream at the mouth of the ravine we took thousands (Chalchijapan drainage). On the only X. helleri, but farther inside where the 26 AMERICAN MUSEUM NOVITATES NO. 3441 large rockpools were located, both species ney S. Nairn, Earlanda L. Williams, Dominic were common. Isla, Richard L. Borowky, William E. Wag- ner Jr., and Alexandra L. Basolo for their as- SUMMARY sistance involving this project. Jay Kumar, Our data clearly show that X. clemenciae Dennis A. Johnston, Caitlin R. Gabor, and is an abundant and widespread inhabitant of Margaret B. Ptacek greatly assisted us with the uplands of the Rio Coatzacoalcos system, the morphometric studies and are acknowl- Veracruz and Oaxaca, Mexico. The species edged. We also thank Angelica Narvaez, can be removed from the list of endangered Gloria Pedroza, Diego Fernando Acosta-Al- or threatened species. Based on morpholog- vear, and Luis Espinasa-Perena for assistance ical and molecular evidence derived from and collaboration with the ®eld studies. This two distinct loci, two new taxa, X. montico- work was partially supported by a grant from lus and X. mixei, have been described that the Mitte Foundation and by several federal are closely allied to X. clemenciae. These grants NIH-NCRR (RR017072), and NIH- new forms are sympatric in parts of their NCI (CA75137, CA55245, CA44303, range that comprises headwater streams of CA39729, CA28909). This project is dedi- the Rio Jaltepec and Rio Sarabia drainages cated to the citizens of Mexico, especially the inside the Sierra Madre del Sur, Oaxaca, indigenous Mixe, who shared a wealth of in- Mexico. According to our dataset and meth- formation and helped us in every way; we ods of phylogenetic reconstruction, the three sincerely thank them. species of the clemenciae clade are monophyletic and are more closely related to the REFERENCES platy®shes and the northern swordtail group of the Panuco river basin, rather than to the Akers, W.H. 1984. Planktic foraminifera and cal- helleri swordtail clade in the south. The dis- careous nannoplankton. Biostratigraphy of the tribution of the swordtails is closely corre- Neogene of Mexico. Part II. Lower Pliocene. lated with geological features in the Rio Co- Tulane Studies in Geology and Paleontology 18: 21±36. atzacoalcos basin. East of the Rio El Corte, Alvarez, J. 1959. Nuevas especies de Xiphopho- X. clemenciae is the only swordtail between rus e Hyporamphus procedentes del Rio Coat- the mountain front bordering the Gulf coastal zacoalcos. Ciencia 19: 69±73. plain and the continental divide. It is absent Baillie, J., and B. Groombridge. 1996. 1996 from the coastal plain. X. helleri occurs in IUCN red list of threatened animals. Gland, the coastal plain but not south of the moun- Switzerland: International Union for Conser- tain front. West of the Rio El Corte, X. cle- vation of Nature and Natural Resources Press menciae and X. helleri are sympatric in trib- 368. utaries of major riversÐRio Jaltepec, Rio Borowsky, R.L., M. McClelland, R. Cheng, and Junapan, and Rio SarabiaÐ, that drain the J. Welsh. 1995. Arbitrarily primed DNA ®n- gerprinting for phylogenetic reconstruction in foothills. Inside the Sierra Madre above wa- vertebrates: the Xiphophorus model. Molecular terfalls and cataracts, X. mixei and X. mon- Biology and Evolution 12: 1022±1032. ticolus replace X. clemenciae and X. helleri. Durham, J.W., A.R.V. Arellano, and J.H. Peck, Jr. These complex zoogeographic patterns sug- 1955. Evidence for no cenozoic Isthmus of Te- gest even more complex relationships de- huantepec seaways. Geological Society of rived from the combination of historic geo- America Bulletin 66: 977±992. logic events, populational isolations, and Geissert Kientz, D. 1999. Regionalizacion geo- subsequent divergence. These same events morfologica del estado de Veracruz. Investiga- theoretically contributed to probable differ- ciones Geogra®cas 40: 23±47. ences in microhabitat preference, morpho- Gordon, M., and D.E. Rosen. 1951. Genetics of metric change/®xation, and possible behav- species differences in the morphology of the ioral divergence. male genitalia of xiphophorin ®shes. Bulletin of the American Museum of Natural History 95: ACKNOWLEDGMENTS 409±464. Groombridge, B. 1993. The 1994 IUCN red list We acknowledge Anthony Pedroza, Leona of threatened animals compiled by the World Hazlewood, Lori Hollek, David Trono, Rod- Conservation Monitoring Centre. Gland, Swit- 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 27

zerland: International Union for Conservation and genetic distances among ®shes of the genus of Nature and Natural Resources Press. Xiphophorus (Poeciliidae). Genetics 102: 539± Kallman, K.D., and I.Y. Bao. 1987. Female het- 556. erogamety in the swordtail, Xiphophorus alvar- Ramirez, E.M. 1999. Taxonomia y zoogeogra®a ezi Rosen (Pisces, Poeciliidae), with comments de la ictiofauna dulceacuicola del Estado de on a natural polymorphism affecting sword col- Oaxaca, Mexico. Ph.D. thesis. Departamento de oration. Journal of Experimental Zoology 243: Ecologia, Universidad de Barcelona. 93±102. Rauchenberger, M., K.D. Kallman, and D.C. Mor- Kallman, K.D., and M.P. Schreibman. 1973. A izot. 1990. Monophyly and geography of the sex-linked gene controlling gonadotrop differ- Rio Panuco basin swordtails (genus Xiphopho- entiation and its signi®cance in determining the rus) with descriptions of four new species. age of sexual maturation and size of the plat- American Museum Novitates 2975: 1±41. y®sh, Xiphophorus maculatus. General and Rosen, D.E. 1960. Middle-American poeciliid Comparative Endocrinology 21: 287±304. ®shes of the genus Xiphophorus. Bulletin of the Lipp, F.J. 1991. The Mixe of Oaxaca. Religion, Florida State Museum 5: 57±242. ritual, and healing. Austin: University of Texas Rosen, D.E. 1979. Fishes from the uplands and Press. intermontane basins of Guatemala: revisionary Marcus, J.M., and A.R. McCune. 1999. Ontogeny studies and comparative geography. Bulletin of and phylogeny in the northern swordtail clade American Museum of Natural History 162: of Xiphophorus. Systematic Biology 48: 491± 271±375. 522. Rosen, D.E., and K.D. Kallman. 1969. A new ®sh Meek, S.E. 1904. The fresh-water ®shes of Mex- of the genus Xiphophorus from Guatemala, ico north of the Isthmus of Tehuantepec. Pub- with remarks on the taxonomy of endemic lications of the Field Columbian Museum Zoo- forms. American Museum Novitates 2379: 1± logical Series 5: 1±252. 29. Meyer, A., J.M. Morrissey, and M. Schartl. 1994. Swofford, D.L. 2002. PAUP*. Phylogenetic anal- Molecular phylogeny of ®shes of the genus ysis using parsimony (* and other methods). Xiphophorus suggests repeated evolution of a Version 4. Sunderland, MA: Sinauer Associ- sexually selected trait. Nature 368: 539±541. Meyer, M.K., and M. Schartl. 2002. Xiphophorus ates. mayae, a new species of swordtail from Gua- Thomas, C., and J.R. Swanton. 1911. Indian lan- temala (Teleostei: Poeciliidae). Ichthyological guages of Mexico and Central America and Exploration of Freshwaters 13: 59±66. their geographical distribution: Washington Michain-Castillo, M.L. 1985. Ostracode biostra- DC: U.S. Government Printing Of®ce. tigraphy and paleoecology of the Pliocene of Thompson, J.D., D.G. Higgins, and T.J. Gibson. the isthmian saltbasin, Veracruz, Mexico. Tu- 1994. CLUSTAL W: improving the sensitivity lane Studies in Geology and Paleontology. 19: of progressive multiple sequence alignment 123±139. through sequence weighting, positions-speci®c Miller, R.R. 1948. The cyprinidont ®shes of the gap penalties and weight matrix choice. Nu- Death Valley system of eastern California and cleic Acids Research 22: 4673±4680. southwestern Nevada. Miscellaneous Publica- Zander, C.D. 1967. Okologische und morpholo- tions Museum of Zoology University of Mich- gische Beitrage zur Systematik und geogra- igan 68: 1±155. phischen Verbreitung der Gattung Xiphophorus Morizot, D.C., and M.J. Siciliano. 1982. Protein (Pisces). Mitteilungen aus dem Hamburgischen polymorphisms, segregation in genetic crosses Zoologischen Museum und Institut 64: 87±125. 28 AMERICAN MUSEUM NOVITATES NO. 3441

APPENDIX 1 ICHTHYO-FAUNAL SAMPLING OF THE RIO COATZACOALCOS DRAINAGE SYSTEM 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 29

APPENDIX 1 (Continued) 30 AMERICAN MUSEUM NOVITATES NO. 3441

APPENDIX 1 (Continued) 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 31

APPENDIX 1 (Continued) 32 AMERICAN MUSEUM NOVITATES NO. 3441

APPENDIX 1 (Continued) 2004 KALLMAN ET AL.: NEW SWORDTAILS FROM OAXACA, MEXICO 33

APPENDIX 1 (Continued) 34 AMERICAN MUSEUM NOVITATES NO. 3441

APPENDIX 1 (Continued)