446

Mol. Biol. Evol. 19(4):446–455. 2002q 2002 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038

Evidence for Multiple Genetic Forms with Similar Eyeless Phenotypes inthe Blind Cavefish, Astyanax mexicanus

Thomas E. Dowling,* David P. Martasian,† and William R. Jeffery‡*Department of Biology, Arizona State University; †Bodega Marine Laboratory, University of California; and ‡Department ofBiology, University of Maryland

A diverse group of animals has adapted to caves and lost their eyes and pigmentation, but little is known abouthow these animals and their striking phenotypes have evolved. The teleost Astyanax mexicanus consists of an eyedepigean form (surface fish) and at least 29 different populations of eyeless hypogean forms (cavefish). Currentalternative hypotheses suggest that adaptation to cave environments may have occurred either once or multipletimes during the evolutionary history of this species. If the latter is true, the unique phenotypes of different cave-dwelling populations may result from convergence of form, and different genetic changes and developmental pro-cesses may have similar morphological consequences. Here we report an analysis of variation in the mitochondrialNADH dehydrogenase 2 (ND2) gene among different surface fish and cavefish populations. The results identify aminimum of two genetically distinctive cavefish lineages with similar eyeless phenotypes. The distinction betweenthese divergent forms is supported by differences in the number of rib-bearing thoracic vertebrae in their axialskeletons. The geographic distribution of ND2 haplotypes is consistent with roles for multiple founder events andintrogressive hybridization in the evolution of cave-related phenotypes. The existence of multiple genetic lineagesmakes A. mexicanus an excellent model to study convergence and the genes and developmental pathways involvedin the evolution of the eye and pigment degeneration.

Introduction

Many hypogean (subterranean) animals are knownfor their troglomorphic characters, including the reduc-tion of eyes and pigmentation, but the evolution of thesestriking morphologies is still unresolved (Culver, Kane,and Fong 1995; Romero 1985). Major issues in the evo-lution of cave animals include the process of cave col-onization and the mechanisms responsible for the ap-pearance of troglomorphic features. Highly restrictedgeographic ranges and, for the most part, lack of geneticstudies of hypogean species and their epigean (surfacedwelling) ancestors hamper most evolutionary studies ofhypogean animals. A notable exception is the teleostAstyanax mexicanus, which exhibits an epigean formand about 30 different conspecific hypogean forms(Mitchell, Russell, and Elliot 1977). Hypogean A. mex-icanus populations exhibit degenerate eyes, which aresunken into the orbits and covered by a flap of skin,reduction or loss of melanin pigmentation, an expandedgustatory system, and other troglomorphic traits(Schemmel 1967, 1974; Wilkens 1988; Jeffery and Mar-tasian 1998; Jeffery et al. 2000; Jeffery 2001).

The epigean form of A. mexicanus is widely dis-tributed in northeastern Mexico and southern Texas. Thefirst hypogean A. mexicanus populations were discov-ered in La Cueva Chica (Chica cavefish), La Cueva delos Sabinos (Los Sabinos cavefish), and La Cueva de ElPachon (Pachon cavefish) in the Sierra de El Abra (fig.1), a limestone escarpment in Tamaulipas and San LuisPotosı, Mexico (Mitchell, Russell, and Elliot 1977), and

Abbreviation: ND2, NADH dehydrogenase 2.

Key words: Astyanax mexicanus, population structure, mtDNA,ND2.

Address for correspondence and reprints: Thomas E. Dowling,Department of Biology, Arizona State University, Tempe, Arizona85287-1501. E-mail: thomas.dowling@asu.edu.

initially described as three different species. Breeding,electrophoretic, and karyotypic studies now support thecontention that the epigean and hypogean forms are thesame species (Wilkens 1971; Avise and Selander 1972;Kirby, Thompson, and Hubbs 1977). Since the firstcavefish populations were discovered in the Sierra de ElAbra region, 26 additional hypogean populations havebeen reported (e.g., La Cueva de la Curva or Curvacavefish; El Sotano de la Tinaja or Tinaja cavefish), themajority from caves in an extensive valley parallelingthe western slope of the escarpment (fig. 1; Mitchell,Russell, and Elliot 1977). Geographically isolated hy-pogean populations have also been discovered in theSierra de Guatemala to the north and in the Micos regionto the west of the Sierra de El Abra (fig. 1; Wilkens andBurns 1972; Mitchell, Russell, and Elliot 1977). Somehypogean populations, including the Chica cavefish andthe cavefish population from La Cueva del Rıo Subter-raneo (Subterraneo cavefish) in the Micos region (fig.1), contain mixtures of eyed, intermediate, and eyelessindividuals resulting from introgression with epigeanforms (Avise and Selander 1972; Mitchell, Russell, andElliot 1977; Romero 1983). More recently, an additionalhypogean Astyanax population has also been reported inGruta de las Granadas, Guerrero, Mexico, outside therange of A. mexicanus (Espinasa, Rivas-Manzano, andExpinosa Perez 2001). The Guerrero cavefish wereprobably derived from epigean Astyanax aeneus, abroadly distributed epigean species inhabiting southernMexico and Central America.

Two hypotheses have been advanced to explainhow hypogean A. mexicanus populations were estab-lished. There was either a single founder event with sub-sequent dispersal between the cave systems or therewere two or more founder events that resulted in mor-phologically convergent troglomorphic populations de-rived from different epigean ancestors. The single originof cave populations in the Sierra de El Abra is supported

Multiple Genetic Forms of Blind Cavefish 447

FIG. 1.—Sketch map of the Sierra de El Abra region in Tamau-lipas and San Luis Potosi, Mexico, showing locations of caves andhypogean and epigean sampling sites. Filled circles: locations of hy-pogean populations sampled. A. La Cueva Chica (Chica cavefish). B.La Cueva de la Curva (Curva cavefish). C. El Sotano de la Tinaja(Tinaja cavefish). D. La Cueva de los Sabinos (Los Sabinos cavefish).E. La Cueva de El Pachon (Pachon cavefish). F. La Cueva del RıoSubterraneo (Subterraneo cavefish). Open circles: locations of otherknown hypogean populations. Roman numerals: locations of sampledepigean populations. I. Rıo Frio. II. Rıo Comandante. III. Nacimientodel Rıo Mante. IV. Arroyo Lagarto. V. Rıo Puerco. VI. Arroyo LosSabinos. VII. Arroyo San Felipe. VIII. Nacimiento del Rıo Choy. IX.Rıo Tampaon. X. Rıo Naranjo. XI. Arroyo Micos. Additional epigeansampling sites outside this area are not shown. Shaded areas representuplands, whereas unshaded areas represent lowlands and valleys. Ma-jor streams are represented by fine lines. The insert indicates the ap-proximate location of the sketched area in northeastern Mexico. Adapt-ed from Jeffery & Martasian (1998).

by low levels of variability at 17 different allozyme lociin the Chica, Pachon, and Los Sabinos cavefish (Aviseand Selander 1972). These data led Avise and Selander(1972) to conclude that genotypic (and presumably phe-notypic) variation among these hypogean populationsresulted from stochastic variation associated with smallpopulation sizes. Subsequent mark and recapture stud-ies, however, suggested that some cavefish populationsare larger than previously appreciated (Mitchell, Russell,and Elliot 1977), reducing the impact of genetic drift.Regardless, the cavefish populations inhabiting largecontiguous cave systems (e.g., Los Sabinos and Tinajacavefish; Mitchell, Russell, and Elliot 1977) may haveoriginated from a single invasion. More recent studiesusing RAPD markers suggested that the Chica, Pachon,Tinaja, and Curva cavefish populations are more closely

related to each other than to nearby surface fish (Espi-nasa and Borowsky 2001), although bootstrap supportfor this relationship was not robust. The hypothesis ofmultiple, independent origins is supported by geneticcrosses between Los Sabinos and Pachon cavefish whichexhibit complementation of eye phenotypes in the F1generation (Wilkens 1971) and the existence of isolatedhypogean populations where gene flow might be imped-ed by geographic barriers (Wilkens and Burns 1972;Mitchell, Russell, and Elliot 1977).

Given the recent progress in developmental biologyof A. mexicanus (Jeffery and Martasian 1998; Jeffery etal. 2000; Yamamoto and Jeffery 2000; Jeffery 2001;Strickler, Yamamoto, and Jeffery 2001), it has becomemore important than ever to understand the evolution ofthe hypogean forms. Here we describe an analysis ofvariation in the mitochondrial NADH dehydrogenase 2(ND2) gene and morphological studies among hypogeanand epigean populations of A. mexicanus, which identifya minimum of two genetically distinct hypogean line-ages. The results are consistent with roles for severalindependent origins or introgressive hybridization (orboth) in the evolution of hypogean A. mexicanus andtheir troglomorphic phenotypes.

Materials and MethodsAnimals

Epigean and hypogean A. mexicanus were collectedfrom the sites shown in figure 1 using baited traps, hand-held nets, and seines. Additional populations were ob-tained from the following localities (sample labels inparentheses): epigean A. mexicanus, Phantom Cave,Texas (TEX1 and 2) and Cuatro Cienegas, Coahuila,Mexico (CC1); A. aeneus, Rıo Mezcalapa, Tabasco,Mexico (MU1 and 2), Rıo Jamapa, Veracruz, Mexico(RJ1), Rıo Arenal, Costa Rica (RA1–3), and Rıo SanCarlos, Costa Rica (RSC1–3). The Peruvian outgroupspecies Astyanax bimaculatus was purchased from a petstore (That Fish Place, Lancaster, Pa.). Some A. mexi-canus individuals were brought back to the laboratoryto establish breeding populations (under the auspices ofMexican Permit Number 040396-213-03), whereas oth-ers were captured, fin-clipped, and released. A smallportion of the caudal fin was excised and placed in 95%ethanol saturated with EDTA for subsequent DNAextraction.

Characterization and Analysis of DNA Variation

DNA was extracted from tail-fin clips by standardphenol-chloroform extraction procedures (Davis, Dib-ner, and Batty 1986, pp. 320–323). Single-stranded con-formational polymorphisms (SSCPs) (Dowling et al.1996; Sunnucks et al. 2000) were screened on 6% nativepolyacrylamide gels using the primers ND2-Acave (59-CGCCACAATCCTCAACGG-39) and ND2-Ccave (59-TGGCGGTTGATGAGTATG-39). At least one strand ofone representative of each SSCP mobility variant oneach gel was sequenced to verify haplotype, either man-ually (Perkin-Elmer cycle sequencing kit) or using anABI 377 automated sequencer. This procedure resulted

448 Dowling et al.

Table 1Numbers and Distribution of SSCP Haplotypes

Haplotype

SAMPLING LOCALITYa

A B C D E F-Hb F-Ib F-Eb I II III IV V VI VII VIII IX X XI Total

A1 . . . . . . . . .A2 . . . . . . . . .A3 . . . . . . . . .A4 . . . . . . . . .A5 . . . . . . . . .

22 25 2

1

11

4

72

7

8 2

10

14

21

21

3

4

21

111

20

1

2

18

2

2

16

1

2

10

10

1932

333

23A6. . . . . . . . .A7 . . . . . . . . .A8 . . . . . . . . .A9 . . . . . . . . .A10. . . . . . . .

13

19 11

21

213

211

A11. . . . . . . .A12. . . . . . . .A13. . . . . . . .A14. . . . . . . .A15. . . . . . . .A16. . . . . . . .B1 . . . . . . . . .B2 . . . . . . . . .B3 . . . . . . . . .B4 . . . . . . . . .

19 95

1510

10 6 31

111

1

1111

191

195

1519

22 19 29 10 25 10 9 18 10 15 24 24 24 9 25 26 25 20 20 364

a Sampling localities as identified in Figure 1.b H, I, and E refer to hypogean, intermediate, and epigean morphotypes, respectively, in the mixed sample from Subterraneo.

in several sequences from most haplotypes, representingmultiple populations (e.g., the most common haplotype,A1, was sequenced in 17 individuals from 12 popula-tions). SSCP variants were identified by a two-charactercode, with the letter indicating the lineage and the num-ber denoting the allele within that lineage. Allele num-bers were assigned in order of discovery and do notreflect levels of divergence.

The entire ND2 gene was characterized from individ-uals representative of each SSCP allele, A. mexicanus fromTexas and northern Mexico, several samples of A. aeneus,and the outgroup, A. bimaculatus. Sequences were ob-tained from one strand each of two separate amplificationproducts generated with the primer pairs ND2-Bcave (59-AAGCTATTGGGCCCATACCC-39)-ND2-Ccave andND2-Dcave (59-CACCATTTGCCCTTCTCATA-39) andND2-E (59-TTCTACTTAAAGCTTTGAAGGC-39) usingmethods described above. The ND2 sequences have beendeposited in GenBank under accession numbers(AF441132–AF441164).

Population genetic analyses of SSCP alleles wereperformed using Arlequin 2.0 (Schneider, Roessli, andExcoffier 2000). Standard measures of diversity (e.g.,gene and nucleotide diversities, average number of dif-ferences, theta) were calculated for each population (re-viewed in Nei 1987, pp. 254–286) and levels of diver-gence among populations quantified by AMOVA (Ex-coffier, Smouse, and Quattro 1992). The number of al-leles was also tabulated for each sample and correctedby dividing by sample size. Jukes-Cantor distancesamong haplotypes were calculated using MEGA2 (Ku-mar et al. 2001), distances among populations generatedwith REAP (McElroy et al. 1992), and similarities vi-sualized using the Neighbor-Joining method as imple-

mented in MEGA2. Geographic structure of SSCP var-iation was also assessed using nested clade analysis (re-viewed in Templeton 2001). Clade structure was deter-mined using the program TCS 1.13 (Clement, Posada,and Crandall 2000) and significance tested using GeoDis2.0 (Posada, Crandall, and Templeton 2000). Phyloge-netic trees of SSCP alleles were generated by PAUP*(Swofford 1998) through heuristic search with 25 ran-dom addition sequence replicates, with no root specified.Topologies from sequences of the entire ND2 gene wererecovered as above, except that A. bimaculatus was usedas the outgroup. Jukes-Cantor distances were also cal-culated from full gene sequences and clustered using theNeighbor-Joining algorithm as implemented in PAUP*(Swofford 1998). Support for specific nodes of topolo-gies obtained through parsimony and Neighbor-Joininganalyses of complete gene sequences was examined bybootstrap resampling (1,000 replicates for eachapproach).

Staining and Analysis of Axial Skeletons

Larval and adult fishes were fixed in formalin for1–4 days. The specimens were double stained for car-tilage and bone by the Alcian Blue-Alizarin Red method(Wassersug 1976). The number of rib-bearing thoracicvertebrae was counted in cleared whole-mountspecimens.

Results

To characterize the genetic distinctiveness of dif-ferent A. mexicanus forms, we examined sequence var-iation in a mitochondrial gene, ND2, from 364 individ-uals representing 6 hypogean and 11 epigean popula-

Multiple Genetic Forms of Blind Cavefish 449

Table 2Standard Measures of Genetic Variation for Each Population (Labeled as Indicated inFig. 1)

Locality N

Numberof

Haplo-types Theta Gene Diversity

Mean numberof

PairwiseDifferences

NucleotideDiversity

Epigean populationsI . . . . . . .II . . . . . .III . . . . . .IV. . . . . .V . . . . . .VI. . . . . .

1015242424

9

323234

0.0035(0.0023)0.0010(0.0010)0.0018(0.0013)0.0009(0.0009)0.0035(0.0020)0.0072(0.0040)

0.51(0.16)0.53(0.05)0.37(0.12)0.51(0.05)0.24(0.11)0.75(0.11)

0.76(0.61)0.54(0.47)0.71(0.55)0.51(0.46)0.24(0.29)1.68(1.08)

0.0025(0.0022)0.0017(0.0017)0.0023(0.0020)0.0017(0.0016)0.0008(0.0011)0.0055(0.0040)

VII . . . . .VII . . . . .IX. . . . . .X . . . . . .XI. . . . . .

2526252020

56542

0.0035(0.0020)0.0043(0.0023)0.0035(0.0020)0.0028(0.0018)0.0009(0.0009)

0.30(0.12)0.41(0.12)0.48(0.12)0.36(0.13)0.53(0.04)

0.32(0.34)0.46(0.42)0.54(0.46)0.39(0.38)0.53(0.46)

0.0010(0.0012)0.0015(0.0015)0.0018(0.0017)0.0012(0.0014)0.0017(0.0017)

Hypogean populationsA . . . . . .B . . . . . .C . . . . . .D . . . . . .E. . . . . . .F. . . . . . .

221929102537

113113

0(0)0(0)0.0017(0.0012)0(0)0(0)0.0016(0.0012)

0(0)0(0)0.63(0.06)0(0)0(0)0.61(0.05)

0(0)0(0)0.82(0.60)0(0)0(0)0.75(0.57)

0(0)0(0)0.0027(0.0022)0(0)0(0)0.0025(0.0021)

tions (fig. 1, table 1). Analysis of SSCPs in a 306-bpfragment of ND2 identified 26 variable positions and 20haplotypes. Of the variable positions, 20 (77%), 4(15%), and 2 (8%) of the changes occurred in the third,first, and second positions, respectively. All changes butone were transitions, and six resulted in amino acidsubstitutions.

Estimates of population genetic parameters fromSSCP fragments are provided in table 2. On an average,epigean populations exhibited 3.6 haplotypes (range 2–6) and gene diversity of 0.45 (range 0.24–0.75). Esti-mated nucleotide diversity was 0.0020 (range 0.0008–0.0055), and the average pairwise comparison yielded0.61 differences (range 0.24–1.68). These values weregenerally higher than estimates from hypogean popula-tions, with only two of the six hypogean samples (Tinajaand Subterraneo) exhibiting more than one haplotype(table 2). Tests of variability (measured by number ofalleles corrected for sample size and theta) between epi-gean and hypogean populations indicated significantlyreduced variability in hypogean populations relative toepigean samples (SPSS for Windows, version 10.0.7,Mann-Whitney U tests, P 5 0.004 and 0.006,respectively).

Phylogenetic analysis of SSCP variants identifiedtwo distinct lineages (A and B, fig. 2), differing by aminimum of seven substitutions (ca. 3.5% sequence di-vergence). Lineages A and B contained 16 and 4 hap-lotypes, respectively, with most haplotypes within line-ages differing by one or two substitutions. The 11 epi-gean populations sampled exhibited only A lineage hap-lotypes (table 1), with the majority of individualssampled exhibiting A1 (59.9%), A3 (14.9%), A5 (8.1%),or A9 (9.5%). Haplotype A1 was widely distributed,found in all but one epigean sample. Hypogean samplesexhibited a mixture of seven lineage A and B haplo-

types, five of which were unique to cavefish samples.All 47 individuals in the Chica and Pachon samples (lo-calities A and E, respectively) exhibited haplotype A1.In a mixed sample of epigean, intermediate, and hypo-gean morphotypes from La Cueva del Rıo Subterraneo(locality F), all 10 individuals with the cave morphotypepossessed a unique haplotype (A15) that differed fromA1 by one change. The 9 and 18 individuals from thiscave with intermediate and epigean morphotypes exhib-ited haplotypes A1, A15, or A5. Lineage B haplotypes(B1, B2, B3, and B4) were only found in the remainingthree caves (localities B–D, fig. 1), and all individualssampled from these localities exhibited lineage Bhaplotypes.

The geographic distribution of variation was as-sessed by AMOVA, with samples divided into northern(I–IV, E, fig. 1) and southern (V–XI, A–D, fig. 1) trib-utaries. This approach identified significant differencesamong populations (FST 5 0.88, P , 0.001). This resultwas largely attributable to variation among populationswithin tributaries (FSC 5 0.87, P , 0.001), with no sig-nificant difference between these two tributaries (FCT 50.03, P 5 0.303). Clustering of distances (fig. 3) indi-cated that there was little genetic differentiation amongsouthern epigean populations, whereas northern surfacepopulations (especially III) were more divergent. Hy-pogean populations were distinct from geographicallyadjacent epigean populations, especially samples B–D,potentially reducing levels of among-tributary variationdetected. Reanalysis of these data excluding cave pop-ulations resulted in a reduction of variation among pop-ulations (FST 5 0.53, P , 0.001) but an increase in theportion because of differences among tributaries (FCT 50.17, P 5 0.018), identifying moderate restrictions togene exchange among northern and southern tributaries.

450 Dowling et al.

FIG. 2.—One of 12 minimum length topologies (drawn to scale)of SSCP haplotypes in A. mexicanus (length 5 28 steps, consistencyindex [CI] 5 0.857, retention index [RI] 5 0.939, uninformative char-acters excluded from calculations but included in branch lengths offigure). Clades provided by TCS are identified by squares and labeledas described in text. The shaded circle represents an inferred ancestralhaplotype. Labels for alleles as provided in table 1.

Nested clade analysis (reviewed in Templeton2001) was performed to examine the geographic struc-ture of haplotypic variation. Four one-step clades wereidentified (fig. 2). Three of these (1-1, 1-2, and 1-3) areprimarily found in epigean samples and fall in lineageA (defined above) with the fourth (1-4 identical to lin-eage B) found only in three caves. Clades 1-1 and 1-2were represented by a few, rare haplotypes that were notassociated with geography. Analyses of the remainingtwo one-step clades (1-3 and 1-4) indicated that somehaplotypes were associated with specific geographic lo-calities (P , 0.001 for both clades), with patterns ofsignificance resulting from isolation by distance for bothclades. Analysis of the total cladogram also identifiedsignificant association of haplotypes (P , 0.001) withgeographic locations; however, the lack of samplesthroughout the drainage precludes discrimination be-tween fragmentation and isolation by distance as the re-sponsible factor.

To place the level of ND2 haplotype divergence be-tween lineages A and B in phylogenetic perspective, thefollowing samples were examined: A. mexicanus from

northern Mexico (outside the Sierra de El Abra region)and Texas, populations of a closely related species, A.aeneus from Veracruz and Tabasco, Mexico (Obregon-Barboza, Contreras-Balderas, and de Lourdes Lozano-Vi-lano 1994) and Costa Rica (Bussing 1998, pp. 79–85),and a Peruvian outgroup species (A. bimaculatus). Se-quence for the entire ND2 gene was obtained from theseindividuals and representatives of each SSCP haplotype.Of the 1,056 positions, 122 were variable in the ingrouptaxa. Distribution of variation was similar to that of theSSCP fragment with 94 (77%), 21 (17%), and 7 (6%)polymorphic third, first, and second positions,respectively.

Parsimony and Neighbor-Joining analysis (fig. 4)yielded similar results, differing only in the placementof the root. Lineage A haplotypes from the Sierra de ElAbra and Micos regions (Chica, Subterraneo, and Pa-chon cavefish and local epigean populations) formed amonophyletic group (bootstrap value .86%), with epi-gean samples from northern Mexico and Texas, a closesister group (bootstrap value of .90%). The closest rel-atives of this lineage were haplotypes of A. aeneus fromVeracruz and Tabasco, Mexico (bootstrap value of.98%) but exclusive of A. aeneus from Costa Rica. TheND2 haplotypes from lineage B formed another mono-phyletic group (bootstrap value of .99%), which wasdevoid of epigean haplotypes and divergent from theother A. mexicanus lineage (ca. 3.5% sequence diver-gence). This level of divergence was comparable to thatbetween A. aeneus from Costa Rica and the widespreadA. mexicanus lineage, producing a trichotomy betweenthese three lineages.

During the course of our studies we observed dif-ferences in body length between lineage A and B cave-fish, with the latter showing anteroposterior compres-sion. To investigate the morphological basis of bodycompression, we examined axial skeletons in variousepigean and hypogean populations (fig. 5, table 3). Skel-etal preparations showed that all the sampled epigean A.mexicanus, as well as A. aeneus and A. bimaculatus,have 12 rib-bearing thoracic vertebrae. Most Chica andPachon lineage A cavefish also have 12 thoracic verte-brae; however, the three types of lineage B cavefish usu-ally exhibit only 11 thoracic vertebrae. The Subterraneolineage A cavefish exhibited a mixture of axial skeletonswith 11 or 12 rib-bearing thoracic vertebrae. A likeli-hood ratio test indicated that A. mexicanus from lineageA exhibited significantly more rib-bearing vertebraethan those of lineage B (SPSS for Windows, version10.0.7, G 5 99.2, P , 0.001), indicating that mtDNAhaplotype lineage is associated with vertebrae number.

Discussion

Sequence analysis of the mitochondrial gene ND2identified considerable variation in A. mexicanus. Epi-gean populations were generally more variable than hy-pogean populations. The only hypogean populationswith more than one haplotype were the Subterraneo andTinaja cavefish. The existence of more than one haplo-type at La Cueva del Rıo Subterraneo may result from

Multiple Genetic Forms of Blind Cavefish 451

FIG. 3.—Neighbor-Joining topology of epigean and hypogean populations. Labels as provided in figure 1.

ongoing hybridization between epigean and hypogeanforms (e.g., Avise and Selander 1972; Mitchell, Russell,and Elliot 1977; Romero 1983; Langecker, Wilkens, andJunge 1991). El Sotano de la Tinaja hosts the largesthypogean population of those studied (Mitchell, Russell,and Elliot 1977), possibly allowing it to maintain morevariability than other smaller populations.

This outcome is consistent with the results of aprevious allozyme survey. In a survey of 17 loci, Aviseand Selander (1972) found variation to be high in sixepigean samples (average heterozygosity of 11.2%) withthree hypogean samples having considerably lower var-iation (less than 7.7%). They attributed this reduction inlevels of within-population variation to genetic drift insmall cave populations. Genetic distances among sam-ples were also low, leading to the conclusion that hy-pogean and epigean forms are conspecific.

Our examination of the geographic distribution ofmtDNA variation indicated that there was considerablestructure among populations. These differences werelargely caused by isolation by distance of surface andcave populations and could not be attributed to any spe-

cific geographic partition. Phylogenetic analysis of ND2haplotypes identified a minimum of two distinct geneticlineages in A. mexicanus: lineage A, consisting of theChica, Pachon, and Subterraneo cavefish and closely re-lated epigean populations and lineage B, consisting ofthe Los Sabinos, Tinaja, and Curva cavefish populations,with no closely related epigean counterparts. The line-age A and B haplotypes are more divergent from eachother than they are from those of another species, A.aeneus from southern Mexico and Costa Rica. Theseresults indicate that the phenotypes shared by A and Blineage cavefish, including reduction of eyes and pig-mentation, exist within a background of relatively highgenetic divergence.

The ND2 haplotype data are supported by morpho-logical and biochemical differences between lineage Aand B cavefish populations. As shown here, most line-age A cavefish have 12 rib-bearing thoracic vertebrae,apparently the ancestral state in A. mexicanus, whereaslineage B cavefish are compressed along their antero-posterior axes and usually have only 11 rib-bearing tho-racic vertebrae. Reduction of body size has been pro-

452 Dowling et al.

FIG. 4.—One of the nine phylogenetic trees obtained from analysis of sequence from the entire ND2 gene (length 5 166 steps, CI 5 0.687,RI 5 0.891, uninformative characters excluded from calculations but included in branch lengths of figure). Topology is drawn to scale exceptfor branches involving the outgroup, A. bimaculatus. Numbers on branches represent results from bootstrap analysis for both parsimony andNeighbor-Joining analyses (before and after the slash, respectively). Labels for A and B lineages as provided in table 1.

posed as a troglomorphic character in fishes (Romeroand Paulson 2001), but this is the first time that quan-titative changes in the axial skeleton have been linkedto this feature. The Los Sabinos, Tinaja, and Curvacavefish are lightly pigmented because of the presenceof vestigial melanocytes, whereas melanin-producingchromophores are absent in the albinistic Pachon cave-fish (Wilkens 1988; Jeffery 2001). Although all cavefishshow enhanced numbers of gustatory organs relative toepigean fish, the number of taste buds is much greaterin Pachon than in Los Sabinos cavefish (Schemmel

1967; Mitchell, Russell, and Elliot 1977). Finally, theeye regulatory genes Pax6 and Prox1 exhibit slightlydifferent expression patterns in the presumptive opticregions of Pachon and lineage B cavefish embryos (Jef-fery et al. 2000; Strickler, Yamamoto, and Jeffery 2001).These properties suggest that distinct morphological andbiochemical differences are present in lineage A and Bcave populations.

The molecular and morphological data imply thatlineage B cavefish either were colonized by epigean A.mexicanus long ago, permitting accumulation of rela-

Multiple Genetic Forms of Blind Cavefish 453

FIG. 5.—Representative axial skeletons of A. mexicanus lineage A (A) and lineage B (B). The arrowheads indicate the number of thoracic,rib-bearing vertebrae in lineage A (12) and B (11). The numbers indicate every fifth vertebra in the axial skeleton.

Table 3Number of Thoracic Vertebrae in Adult Epigean andHypogean Astyanax Populations

Sample

Number

11 12

A. bimaculatus . . . . . . . . . . . . . . .A. aeneus . . . . . . . . . . . . . . . . . . .A. mexicanus (epigean) . . . . . . . .A. mexicanus (hypogean) . . . . . .

000

95

31

China (A) . . . . . . . . . . . . . . . . . . .Curva (B) . . . . . . . . . . . . . . . . . . .Tinaja (C). . . . . . . . . . . . . . . . . . .Los Sabinos (D) . . . . . . . . . . . . .Pachon (E) . . . . . . . . . . . . . . . . . .Subterraneo (F) . . . . . . . . . . . . . .

19

1414

35

15100

153

tively extensive nucleotide substitutions in the ND2gene, regression of the axial skeleton, and the appear-ance of other troglomorphic features, or were estab-lished more recently by a unique epigean lineage that isextinct or no longer occupies the region. Given the ex-tent of our sampling in surface waters in the Sierra deEl Abra and surrounding regions in northern and south-ern Mexico, the former alternative seems more likely;on the other hand, ND2 haplotypes exhibited by lineageA cavefish are identical (Chica and Pachon) or nearlyidentical (Subterraneo) to adjacent epigean localities,possibly indicating a more recent origin for these hy-

pogean populations. In support of this interpretation, theaxial skeleton of lineage A (table 3) and the eyes andpigmentation of Chica and Subterraneo cavefish are lessregressed than those of other cavefish populations(Mitchell, Russell, and Elliot 1977). However, the highdegree of eye regression and complete absence of bodypigmentation in Pachon cavefish conflicts with this in-terpretation (Mitchell, Russell, and Elliot 1977; Wilkens1988; Jeffery and Martasian 1998; Jeffery 2001). Dis-crimination between these alternatives will require a de-tailed phylogeographic analysis (reviewed in Avise2000) of this complex group, with the latter alternativesupported if more extensive sampling identifies epigeanlineages similar to the unusual lineage B cave haplotype.

It is also possible that lineage A hypogean popu-lations could be old and share a common origin with Blineage cavefish, with their mtDNAs more recentlytransferred from epigean populations through introgres-sive hybridization (e.g., Dowling and Hoeh 1991; Ger-ber, Tibbets, and Dowling 2001). The Chica populationcontains putative hybrid individuals with intermediateeye and pigment morphologies thought to be derived byperiodic introgression with epigean fishes, which enterLa Cueva Chica via a connection with the nearby RıoTampaon (Mitchell, Russell, and Elliot 1977; Romero1983, fig. 1). In contrast to La Cueva Chica, La Cuevade El Pachon is a former spring resurgence isolated from

454 Dowling et al.

surface drainage in the valley below, and with no ob-vious access route for epigean fish (Mitchell, Russell,and Elliot 1977). Avise and Selander (1972) and Mitch-ell, Russell, and Elliot (1977) failed to observe hybridsin La Cueva de El Pachon. However, Langecker, Wilk-ens, and Junge (1991) reported hybrids in this cave in1986–1988. We did not see hybrids in La Cueva de ElPachon in 1996–2000, and every fish we have capturedthere has regressed eyes and is devoid of body pigmen-tation. Introgression with local epigean fishes at La Cue-va de El Pachon would have affected our haplotype re-sults, unless the hybrids observed by earlier investiga-tors had been expunged from the population. If hybrid-ization does not account for our results, then it ispossible that Pachon cavefish have evolved more re-cently than lineage B cavefish and are undergoing trog-lomorphic evolution more rapidly than other cavefishpopulations.

ND2 data from the Micos region also suggest thathybridization may not readily account for the origin ofthe unique A15 haplotype of Subterraneo cavefish (table1). An intermittent stream containing epigean fish sinksinto La Cueva del Rıo Subterraneo during the rainy sea-son, and cave pools near the entrance contain fish ofmixed forms (Wilkens and Burns 1972; Mitchell, Rus-sell, and Elliot 1977). In this cave, hypogean fishes arelocated in pools distant from the entrance and exhibithaplotype A15, which was not found in epigean popu-lations collected in the intermittent stream immediatelyoutside the cave (N 5 40, table 1). Although our sampleof intermediate forms is small, six of nine individualsexhibited the diagnostic haplotype A15, suggesting thathybridization may typically involve hypogean femalesand epigean males, counter to the direction necessaryfor replacement of hypogean mtDNA haplotypes. Thepotential impact of introgressive hybridization from epi-gean forms into lineage A cavefish populations must beprovided by a future examination of variation in nucleargene loci.

Our results are consistent with two scenarios forthe origin of Astyanax cavefish. First, divergent ND2haplotypes present in lineage B cavefish populations andto a lesser degree in the Subterraneo cavefish are con-sistent with multiple independent origins. Second, someof the haplotype data could also be explained by vari-ation in the level of introgressive hybridization amongcertain hypogean populations and their proximate epi-gean populations. In these cases, the influence of allelicvariation from surface populations also would generatedistinctive genetic lineages, creating significant conse-quences for the evolution of troglomorphic features. Theimportance of genetic diversity within different cavefishpopulations was demonstrated by crosses between Pa-chon and Los Sabinos cavefish, which resulted in F1progeny with more extensive optic development thaneither parent, suggesting that mutations in differentgenes are involved in eye degeneration (Wilkens 1971).

Considering the wide distribution of A. mexicanusand the large number of reported hypogean populations(Mitchell, Russell, and Elliot 1977), it is likely thatmany different genetic combinations exist in natural

populations. This diversity of genetic backgrounds andthe ability to routinely propagate and manipulate the em-bryos of this species in the laboratory (Jeffery and Mar-tasian 1998; Yamamoto and Jeffery 2000; Jeffery 2001)make A. mexicanus a valuable model for studying theevolution of eye and pigment degeneration.

Acknowledgments

We thank Catherine Beard, Brenna Bemis, DeidreHeyser, Kelly Hornaday, Clark Hubbs, David Jeffery,Dino Rossi, Allen Strickler, Yoshiyuki Yamamoto, andMeredith Yeager for assistance in fish collecting andDeidre Heyser and Kelly Hornaday for technical assis-tance. Dean Hendrickson provided samples from CuatroCienegas, Mexico and Texas. W. L. Minckley and Al-demaro Romero provided critical comments and helpfuldiscussion. This research was supported by NSF grantsto T.E.D. (DEB 9220683) and W.R.J. (DEB 9726561and IBN-0110275).

LITERATURE CITED

AVISE, J. C. 2000. Phylogeography: the history and formationof species. Harvard University Press, Cambridge, Mass.

AVISE, J. C., and R. K. SELANDER. 1972. Evolutionary geneticsof cave-dwelling fishes of the genus Astyanax. Evolution26:1–19.

BUSSING, W. A. 1998. Freshwater fishes of Costa Rica. Edi-torial de La Univesidad de Costa Rica, San Jose, CostaRica.

CLEMENT, M., D. POSADA, and K. A. CRANDALL. 2000. TCS:a computer program to estimate gene genealogies. Mol.Ecol. 9:1657–1660.

CULVER, D. C., T. C. KANE, and D. W. FONG. 1995. Adaptationand natural selection in caves. Harvard University Press,Cambridge, Mass.

DAVIS, L. G., M. D. DIBNER, and J. F. BATTY. 1986. Basicmethods in molecular biology. Elsevier Press, New York.

DOWLING, T. E., and W. R. HOEH. 1991. The extent of intro-gression outside the hybrid zone between Notropis cornutusand Notropis chrysocephalus (Teleostei: Cyprinidae). Evo-lution 45:944–956.

DOWLING, T. E., C. MORITZ, J. D. PALMER, and L. H. RIESE-BERG. 1996. Nucleic acids IV: analysis of fragments andrestriction sites. Pp. 249–320 in D. M. Hillis, C. Moritz,and B. K. Mable, eds. Molecular systematics, Sinauer As-sociates, Sunderland, Mass.

ESPINASA, L., and R. B. BOROWSKY. 2001. Origins and rela-tionship of cave populations of the blind Mexican tetra, As-tyanax fasciatus, in the Sierra de El Abra. Environ. Biol.Fishes 62:233–237.

ESPINASA, L., P. RIVAS-MANZANO, and H. ESPINOSA PEREZ.2001. A new blind cave fish population of the genus Asty-anax: geography, morphology and behavior. Environ. Biol.Fishes 62:339–344.

EXCOFFIER, L., P. SMOUSE, and J. QUATTRO. 1992. Analysis ofmolecular variance inferred from metric distances amongDNA haplotypes: application to human mitochondrial DNArestriction data. Genetics 131:470–491.

GERBER, A. S., C. A. TIBBETS, and T. E. DOWLING. 2001. Therole of introgressive hybridization in the evolution of theGila robusta complex (Teleostei: Cyprinidae). Evolution55:2028–2039.

JEFFERY, W. R. 2001. Cavefish as a model system in evolu-tionary developmental biology. Dev. Biol. 231:1–12.

Multiple Genetic Forms of Blind Cavefish 455

JEFFERY, W. R., and D. P. MARTASIAN. 1998. Evolution of eyeregression in the cavefish Astyanax: apoptosis and the pax6gene. Am. Zool. 38:685–696.

JEFFERY, W. R., A. E. STRICKLER, S. GUINEY, D. G. HEYSER,and S. I. TOMAREV. 2000. Prox1 in eye degeneration andsensory organ compensation during development and evo-lution of the cavefish Astyanax. Dev. Genes Evol. 210:223–230.

KIRBY, R. F., K. W. THOMPSON, and C. HUBBS. 1977. Karyo-typic similarities between the Mexican and blind tetras.Copeia 1977:578–580.

KUMAR, S., K. TAMURA, I. B. JAKOBSEN, and M. NEI. 2001.MEGA2: molecular evolutionary genetic analysis software.Bioinformatics (in press).

LANGECKER, T. G., H. WILKENS, and P. JUNGE. 1991. Intro-gressive hybridization in the Pachon Cave population ofAstyanax fasciatus (Teleostei Characidae). Ichthyol. Explor.Freshw. 2:209–212.

MCELROY, D., P. MORAN, E. BERMINGHAM, and I. KORNFIELD.1992. REAP: an integrated environment for manipulationand phylogenetic analysis of restriction site data. J. Hered.83:157–158.

MITCHELL, R. W., W. H. RUSSELL, and W. R. ELLIOT. 1977.Mexican eyeless characin fishes, genus Astyanax: environ-ment, distribution, and evolution. Spec. Publ. Mus. TexasTech. Univ. 12:1–89.

NEI, M. 1987. Molecular evolutionary genetics. Columbia Uni-versity Press, New York.

OBREGON-BARBOZA, H., S. CONTRERAS-BALDERAS, and M. DE

LOURDES LOZANO-VILANO. 1994. The fishes of northernand central Veracruz, Mexico. Hydrobiologia 286:79–95.

POSADA, D., K. A. CRANDALL, and A. R. TEMPLETON. 2000.GeoDis: a program for the cladistic nested analysis of thegeographical distribution of genetic haplotypes. Mol. Ecol.9:487–488.

ROMERO, A. 1983. Introgressive hybridization in the Astyanaxfasciatus (Pisces: Characidae) population at La Cueva Chi-ca. Nat. Speleol. Soc. Bull. 45:81–85.

———. 1985. Can evolution regress? Nat. Speleol. Soc. Bull.47:86–88.

ROMERO, A., and K. M. PAULSON. 2001. It’s a wonderful hy-pogean life: a guide to the troglomorphic fishes of theworld. Environ. Biol. Fishes 62:13–41.

SCHEMMEL, C. 1967. Vergleichende Untersuchungen an denHautsinnesorganene ober- und unterirdisch lebender Astya-nax-Formen. Z. Morphol. Tiere 61:255–316.

———. 1974. Genetische Untersuchungen zur Evolution desGeschmacksapparates bei cavernicolen Fischen. Z. Zool.Syst. Evolutionforsch. 12:9–22.

SCHNEIDER, S., D. ROESSLI, and L. EXCOFFIER. 2000. Arlequin.Version 2.000. Software for population genetics data anal-ysis. Website: http://anthropologie.unige.ch/arlequin/.

STRICKLER, A. G., Y. YAMAMOTO, and W. R. JEFFERY. 2001.Early and late changes in Pax6 expression accompany eyedegeneration during cavefish development. Dev. GenesEvol. 211:138–144.

SUNNUCKS, P., A. C. C. WILSON, L. B. BEHEREGARAY, K.ZENGER, J. FRENCH, and A. C. TAYLOR. 2000. SSCP is notso difficult: the application and utility of single-strandedconformation polymorphism in evolutionary biology andmolecular ecology. Mol. Ecol. 9:1699–1710.

SWOFFORD, D. L. 1998. PAUP*: phylogenetic analysis usingparsimony (* and other methods). Version 4.0. Sinauer As-sociates, Sunderland, Mass.

TEMPLETON, A. R. 2001. Using phylogeographic analyses ofgene trees to test species status and processes. Mol. Ecol.10:779–791.

WASSERSUG, R. J. 1976. A procedure for differential stainingof cartilage and bone in whole formalin-fixed vertebrates.Stain Technol. 51:131–134.

WILKENS, H. 1971. Genetic interpretation of regressive evo-lutionary process: studies of hybrid eyes of two Astyanaxcave populations (Characidae, Pisces). Evolution 25:530–544.

———. 1988. Evolution and genetics of epigean and cave As-tyanax fasciatus (Characidae, Pisces). Evol. Biol. 23:271–367.

WILKENS, H., and R. L. BURNS. 1972. A new Anoptichthyscave population (Characidae, Pisces). Ann. Speleol. 27:263–270.

YAMAMOTO, Y., and W. R. JEFFERY. 2000. Central role for thelens in cavefish eye degeneration. Science 289:631–633.

KEITH CRANDALL, reviewing editor

Accepted November 13, 2001