Evolution, 34(2), 1980, pp. 259-270
TROPHIC DIFFERENTIATION IN ILYODON, A GENUS OF STREAM-DWELLING GOODEID FISHES: SPECIATION VERSUS ECOLOGICAL POLYMORPHISM
BRUCE J. TURNER' AND DANIEL J. GROSSE Museum of Zoology, University of Michigan, Ann Arbor, Michigan 48109
Received March 15, 1979. Revised September 10, 1979
An impressive diversity of trophic ad tected two trophically divergent, sympat aptations is a hallmark of teleostean fish ric clones of the triploid unisexual fish es, and has been a characteristic of their species Poeciliopsis 2-monacha-lucida. evolution that dates from their first The clones, though obviously closely re appearance in the fossil record. Students lated, were differentiated in dental mor of this diversity have usually regarded a phology and feeding behavior. particular suite of functionally related The implications of the discovery of dis trophic specializations as attributes of a continuous trophic polymorphisms in sin given species that adapt it to a particular gle populations are far-reaching. From the ecological niche and tend to exclude it perspective of ichthyology, the existence from other niches. Generally, though al of such polymorphisms (involving well lowance is made for developmental vari defined "taxonomic" characters) suggests ation or sexual dimorphism, a species has that the number of biological species in been presumed to have only one set (or a volved in several extreme radiations of narrow range) of trophic adaptations. It trophic diversity, most notably those of has been thought extremely unlikely for cichlids in the African Rift lakes (Fryer contemporaneous adult members of the and Iles, 1972) and of cyprinids in Lake same Mendelian population to possess Lanao (Myers, 1960), may have been se radically different (and mutually exclu riously overestimated by morphological sive) arrays of trophic features. Thus, es inferences (see also Kornfield, 1978). The pecially at or near the species level, ich nature of the radiations themselves, as thyologists have tended to equate trophic well as the systematic utility of trophic differentiation with systematic diver characters in general, may have to be gence. The generality of this relationship reevaluated. From the perspective of ecol has recently been seriously challenged. ogy, trophic polymorphism implies that Sage and Selander (1975) have demon "a single species occupies niches that are strated that three trophically specialized as distinctly different as is usually the case sympatric cichlids (Cichlasoma) in the among validly different species" (Hutch Cuatro Cienagas basin of Coahuila, Mex inson, 1978, p. 179), and leads to a host ico, a molluskivore (with molariform pha of questions concerning the ecological re ryngeal teeth and a short gut), an algal lationships of the trophic morphs. From detritus feeder (papilliform pharyngeal the perspective of evolutionary genetics, the teeth and a long gut) and a piscivore-all discontinous trophic variation may well be previously thought to be distinct species the type of niche-specific polymorphism (Taylor and Minckley, 1966)-exhibited that Maynard Smith (1966) and others coordinate geographic variation at several have regarded as a fundamental requisite allozyme loci, and were therefore most for sympatric speciation. likely conspecific. Vrijenhoek (1978) de- The only other known case of trophic polymorphism in teleosts is that postulat , Present address: Department of Biology, Virgin ed in species of the characoid genus Sac ia Polytechnic Institute & State University, Blacks codon by Roberts (1974): up to four dental burg, Virginia 24061. morphs occur in what appear to be con 259 260 B. J. TURNER AND D. ]. GROSSE
FIG. 1. Lateral views of Ilyodon specimens from "dichotomous" populations, males above, females below, in each. a. Type A, Rio Tuxpan drainage (Rio "Terrero"; UMMZ 191681, male is 56 mm Standard Length). b. Type B, collected simultaneously with "A" (UMMZ 191680, male is 57.5 mm S.L.). c. Type A, Rio Armeria drainage (Ilyodonfurcidens; Rio de Comala; UMMZ 189595, male is 59 mm S.L.). d. Type B (Ilyodon xantusi); collected simultaneously with "C" (UMMZ 189594, male is 64 mm S.L.).
specific population samples of these algal rant of the Mexican plateau into the Pa grazing fishes from the mountain streams cific Ocean. of Panama and northwestern South Amer The llyodon populations of certain trib ica. Unfortunately, genetic or other bio utaries of two river drainages, the Rio logical data from the Saccodon dental Armeria and Rio Tuxpan (=Rio Coa morphs are entirely lacking. huayana), can be divided into two very In this paper we present biochemical distinctive morphological (presumably genetic analyses of two sets of sympatric, trophic) "types" (Figs. 2 and 3): trophically specialized, putative species of a little-known Mexican fish genus, Ily Type "A": The head is tapered, with rel odon. This genus is a member of the vi atively small jaws; the gape of viparous fish family Goodeidae, a family the mouth is conspicuously of less than 40 species of cyprinodontoid lateral, and the jaw teeth (bi fishes with unique reproductive adapta fid in both types) tend to be tions (Miller and Fitzsimons, 1971). The short and organized into two family is essentially endemic to the Mex obvious rows on the premax ican plateau, where it has undergone an illae (usually one on the den adaptive radiation into habitats and tary bones). trophic niches more usually occupied by Type "B": The head is very blunt with members of several other fish families. Il relatively massive jaws; the yodon species are small (usually less than gape of the mouth is almost 120 mm total length), laterally com completely transverse (fron pressed, minnow-like fishes (Fig. 1). Un tal), and the jaw teeth tend to like nearly all other goodeids (most of be elongate, posteriorly exca which are typically lake dwellers), llyodon vated, fewer in number, and are fluviatile, like Saccodon, and are al in larger specimens organized most exclusively confined to the montane in but a single row on both rivers that flow off the southwest quad- jaws. TROPHIC DIFFERENTIATION IN ILYODON 261
a c
b
FIG. 2. Enlarged frontal views of specimens shown in Fig. 1; data as in Fig. 1. Note the wider, more massive jaws of the "B" types in both river systems.
The morphology of type"A" suggests that types in the R. Armeria as I lyodon fur of an insectivore or planktivore, that of cidens and Balsadichthys xantusi, respec "B" a substrate feeder, possibly a har tively; "Balsadichthys" was vacated in fa vester of lithophilic algae or small ani vor of Ilyodon by Miller and Fitzsimons mals; however, no detailed dietary data (1971). are yet available. The two trophic types The two trophic types are readily dif breed true in the laboratory, and develop ferentiated by eye (above a size of about their distinctive phenotypes when reared 35 mm in total length), but relative mouth on identical diets. They thus probably do width (ratio of premaxillary width to head not represent "ecophenotypic" or devel length) provides a convenient discrimina opmental modifications of the same array tion (Hubbs and Turner, 1939). The fre of genotypes (see discussion by Kornfield quency distribution of the ratio is essen and Koehn, 1975). With the exception of tially bimodal in samples of populations a small difference in adult male color pat that contain both types (Fig. 3). tern in the R. Armeria (see below), the The two river drainages also contain morphological differences between the Ilyodon populations that are not dichot types are all plausibly associated with di omous, but rather highly variable in vergent trophic adaptations. trophic features. Those in the R. Tuxpan The trophic types are quite distinctive, are otherwise virtually identical to the di and, if presented with Figure 2 few ich chotomous populations. The frequency thyologists (ourselves initially included) distributions of the mouth width:head would hesitate to suggest that they are length ratio in these populations are not different species. In the last major taxo bimodal (Fig. 3C and D) and means are nomic revision of the family, Hubbs and near the antimode of the dichotomous Turner (1939) recognized the A and B populations. In the R. Tuxpan drainage, 262 B. J. TURNER AND D. J. GROSSE
R. TUXPAN -+- lations are encountered almost exclusively in high-gradient tributaries with rock and
TRIBUTARY 16 KM N boulder-strewn substrates and abundant
U A PIHUAMQ("RIO TERRERO ) riffles. The non-dichotomous populations are usually found in low-gradient habitats with pebble or silt substrates. When first
TRIBUTARY 8 KM N observed, in collections from the R. Ar PIHUAMO ("RIO TERRERO"?J meria, Ilyodon of intermediate phenotype were regarded as interspecific (and inter -+ generic) hybrids (Hubbs and Turner, TRIBUTARY AT TECALITLaN 1939). c Knowledge of the relationships of the trophic types in the dichotomous popula -+- SAN RAFAEL BRIDGE tions is obviously central to understanding the evolution of trophic diversity in Ily D odon. Do they represent discrete gene pools, or are they components of the same + R. ARMERIA Mendelian population? In this paper we 20 attempt to answer this question by com
15 R. DE COMALA parison of polymorphic allozyme loci in AT COMALA the trophic types in two dichotomous pop E 10 ulations, one from each of the river streams. The data suggest that at least one of these sympatric "species pairs" may in .50 .55 .60 fact be a single Mendelian population with MOUTH WIDTH / HEAD LENGTH discrete trophic morphs. FIG. 3. Histograms of the frequency distribu tions of relative mouth widths (mouth width/head MATERIALS AND METHODS length) in five Ilyodon samples. Solid black squares Specimens.-Samples were taken by represent individuals scored as Type A by eye, open white squares represent those scored as Type B. poisoning with emulsified rotenone at the Striped (gray) boxes represent individuals which following localities: could not be reliably scored by eye. Vertical bar above each distribution indicates the position of the 1. Rio Tuxpan drainage: Rio "Terrero" mean, horizontal bar the extent of the 99% confi about 1.5 km W of the town of 21 de dence interval of the mean (both calculated by the Noviembre (16 km by highway N of arcsine [pJl/2 transformation). A. Allozyme survey Pihuamo), jalisco, Mexico. September sample, Rio Tuxpan, 1977. B. Same or lower trib utary, approximately 8 km downstream from "A"; 14, 1977. collected 1955 (UMMZ 172154, 201952). Note sim 2. Rio Armeria drainage: Rio de Comala, ilarity of distribution to that of "A." C. Continuous just S of the town of Comala, approx or "nondichotomous" sample; unnamed tributary at imately 7 km N of the city of Colima, Tecalitlan; (UMMZ 189080) preserved material, Mexico. September 13, 1977. collected 1968. D. Continuous sample, Rio Tuxpan at San Rafael bridge; preserved material (UMMZ Specimens were frozen on dry ice in the 127160), collected 1975. E. Allozyme survey sample, -90 1977, Rio de Comala at Comala; Rio Armeria drain field and held at C in the laboratory. age. Unassignable specimens (striped boxes) in "A" Assignment of trophic type.-The and "E" were excluded from the allozyme survey. trophic type of each specimen was deter (See text.) mined by eye; specimens that could not be unequivocally assigned to either type were the unimodal populations appear to be measured but were not analyzed further. more frequent than the bimodal (dichot Visual discrimination was based on the omous) ones. It is not known if the two width and shape of the mouth, features kinds of populations ever occur in the that have a strong allometric relationship same tributary. The dichotomous popu- with body size. In general, visual discrim- TROPHIC DIFFERENTIATION IN ILYODON 263 ination of the trophic types becomes im TABLE 1. Polymorphic allozyme loci in two allopat precise below about 35 mm in standard ric dichotomous Ilyodon populations. length, and discrimination even by mea Relative surement of relative mouth width becomes mobility of protein R. R. virtually impossible below about 25 mm. Locus Allele x 100 Tuxpan Armeria Electrophoretic analysis.-The protein Ada a 87 + products of some 30 presumptive struc b 60 + tural genes were surveyed by horizontal c 74 + + starch gel electrophoresis. Details of meth d 100 + ods are available upon request. Agp a 52 + At each polymorphic locus the genotyp b 100 + ic frequency distribution of each of the c 37 + + trophic types in each geographic sample Cpk-l a 100 + + was compared by contingency table (eval b 83 + + uated by Chi-square test of association, Cpk-2 a 100 + + or by Fisher's exact method) and the b 82 + + allelic frequencies derived from the geno Ldh-3 a 92 + + type data were compared by t test us b 100 + ing a method based on arcsine transfor 86 + mations (Sokal and Rohlf, 1969, p. 607). . Mdh-3 a 100 + + The zygotic frequencies in each of the b 88 + trophic types were compared to those ex Pgd a 89 + + pected from Hardy-Weinberg equilibrium b 100 + by Chi-square test, using Yates' correc Phi-l a 90 + + tion. Finally, the genotype data from both b 100 + c 75 + of the trophic types in each sample were d 78 (+) pooled and their distribution compared to a Hardy-Weinberg equilibrium with alle lic frequencies derived from the pooled nate between the trophic types; had we data. sorted by measurement of relative mouth width, nearly every specimen could have been categorized. RESULTS General allozyme comparisons.-Five Determination of trophic types.-The loci were polymorphic (criterion: p :s;; .99) frequency distributions of relative mouth in each of the geographic samples (Table widths in the two samples are shown in 1); but only two, Cpk-l and Cpk-2, were Figures 3A and 3E; the distributions are polymorphic in both. Each of the samples clearly bimodal. The two figures also contained unique secondary alleles not show relative mouth widths for 36 speci present in the other. For example, the Ada mens (less than 10% of the total) that locus in the Rio Armeria showed a tetra-al could not be unequivocally assigned to lelic polymorphism, but was monomor trophic type by eye, and which were ex phic in Rio Tuxpan. There were no fixed cluded from electrophoretic analysis. allelic differences between the A and B These specimens do not cluster in the an trophic types in either of the samples. tinodes of the distributions; it is therefore Heterozygosity.-Comparison of n, the likely that they were not hybrids or inter average proportion of loci heterozygous mediates. As a group, they were smaller per individual, are given in Table 2; the in size than the rest of the specimens (x = table also includes a contingency X2 com 36.8 ± 1.0 mm versus 41.9 ± 0.5 mm) parison of the numbers of individuals het and our inability to sort them into cate erozygous at zero to four loci in each gories probably reflects a size-related lim trophic type. The only significant differ itation of our ability to visually discrimi- ence that emerges from these comparisons 264 B. J. TURNER AND D. J. GROSSE
TABLE 2. Heterozygosity and frequency distributions of heterozygous loci! in two dichotomous Ilyodon population samples.
Heterozygous loci per individual 0 N fit ± SE
Rio Tuxpan Type A 24 16 5 45 .019 ± .0033 Type B 39 26 10 75 .020 ± .0027 Pooled 63 42 15 120 (x2Types = 0.13, P = .94)
Rio Armeria Type A 44 32 22 6 104 .030 ± .0030 Type B 45 35 28 7 3 118 .035 ± .0030 Pooled 89 67 50 13 3 222 (X2 Types = 3.1, P = .85) (X2 Localities = 13.0, P = .005)
I Determined from a survey of 30 loci. 2. f! = mean proportion of heterozygous loci per individual.
is between the two pooled geographic (N = 45). Consequently, the estimates of samples in the latter comparison. iI val allelic frequency, especially in the "A" ues, in all samples (range .022-.035) are type, are associated with large standard below the average proportional heterozy errors, and small but significant differ gosity per locus for fishes (0.0513) reported ences may have gone undetected. The im by Nevo (1978). portance of this point is enhanced by re Comparisons of individual allozyme ports in the literature of rather subtle loci.-In the Rio Tuxpan sample (Table allozymic differences, some detectable 3), there are no significant differences in only in pooled data from a number of sam genotype or allelic frequencies between ples, in several reproductively isolated types A and B at five polymorphic loci. sympatric sibling fish species (Nyman, The contingency probabilities computed 1972; Allendorf et al., 1976; Henricson for each of the polymorphic loci were used and Nyman, 1976). Note, however, that to generate a combined probability test if the differences are too subtle, their bi (Sokal and Rohlf, 1969, p. 621 ff); the re ological interpretation becomes ambigu sultant probability (Table 3) is .47, well ous. A small difference in allelic or geno above statistical significance. There is no typic frequency at one or a few loci could discernible genetic differentiation between be interpreted either as evidence of repro the trophic types. There are no significant ductive isolation or of niche-specific selec departures from zygotic frequencies pre tion in a single Mendelian population that dicted by Hardy-Weinberg equilibria in is ecologically partitioned (review by Hed either separate or pooled samples. These rick et al., 1976). data offer no support to the hypothesis Even if there is absolute identity of al that the A and B trophic types are distinct lelic frequencies in each of the morphs, species, but are consistent with a system interpretation of the data may be con of ecological polymorphism involving two founded by a combination of relatively re trophic "morphs" in a single Mendelian cent speciation and selection. It is now population. However, while they do ren appreciated that speciation does not nec der it unlikely, for the following reasons essarily involve major genetic restructur the data do not categorically falsify the ing (Lewontin, 1974, p. 173; other evi hypothesis that there is reproductive iso dence reviewed by White, 1978, p. 43). If lation between the morphs. the trophic types are reproductively iso The secondary alleles at four of the five lated descendants of an immediate com polymorphic loci are rare (q ~ .07), and mon ancestor, and the loci in question the sample size of the "A" type is small were polymorphic in the ancestral popu- TROPHIC DIFFERENTIATION IN ILYODON 265
TABLE 3. Genotype and allelic frequencies at polymorphic allozyme loci in a dichotomous Ilyodon population from the R. Tuxpan. drainage (aR. Terrero").
Hardy-Weinberg
Presumptive genotypes Association Allele frequencies Separate Pooled
Locus Morph ala alb bib ale N X' dj. P P q X' P X' p A 24 18 2 45 .74 .24 .02 .7 .87 Agp 1.6 2 0.63 1.2 .76 B 45 21 4 3 73 .78 .20 .02 1.4 .72 A 43 45 .98 .02 .02 .90 Cpk-2 (F)' 0.69 .17 .76 B 78 85 .96 .04 .16 .77 A 44 45 .99 .01 .02 .90 Cpk-3 (F) 0.15 .20 .73 B 74 9 83 .95 .05 .03 .66 A 39 6 45 .93 .07 .23 .71 Mdh-3 (F) 0.16 .21 .72 B 81 4 85 .98 .02 .05 .87
A 41 3 45 .94 .06 1.2 .37 Pgd .009 0.93 .15 .78 B 77 8 85 .95 .05 .2 .73
N Fisher's combined probability test": -2 L In Pi = 9.9, d.f. = 10 ;=1 P = .47
1 "F" = Prob. calculated by Fisher's exact method. a Sakal & Rohlf (1969, p. 621). lation, speciation might have occured The argument that allelic frequency without allelic substitution or fixation. identity implies conspecificity assumes The alleles at all five loci would then per that the alleles at any locus in each of the sist in the two descendant gene pools. morphs are truly identical. Allelic iden There is evidence that the fitness of some tity, as opposed to electrophoretic equiv allozymic variants is correlated with phys alence, has not yet been established in Il ical environmental variables, especially yodon. Electrophoretic equivalence may temperature (Mitton and Koehn, 1975; be misleading: the work of Johnson (1977) other literature reviewed by Nevo, 1978, and Coyne (1976) suggests that seemingly p. 163). If all alleles were present in both identical allozymes may well be differen descendant gene pools, and temperature tiated by additional techniques. If "step or some other physical property of the en models" (Cobbs and Prakash, 1977) of vironment were the main determinant of electrophoretic variation in natural pop the allelic frequencies at the polymorphic ulations have broad applicability, then loci, then identity of allelic frequencies in this consideration becomes especially per the two morphs would be simply a func tinent to any hypothesis of conspecificity tion of their existence in the same habitat. based on allozyme data, for these models This explanation of allelic identity in imply that electrophoretic equivalence of volves a combination of events of low nonidentical proteins may be quite com probability: speciation without substitu mon. The allozymes of Ilyodon have thus tion or fixation of least one of the loci and far been characterized only in routine sur determination of allelic frequencies at all vey systems of moderate resolving power. five of them by physical ecological vari Thus, conspecificity of the trophic types ables. The sequence is therefore unlikely, is the simplest explanation of their allo but it is a possibility that we have not zymic identity, and the one supported by excluded. the vast bulk of the data on the allozymic N 0\ 0\
TABLE 4. Genotype and allelic frequencies at five polymorphic loci in a dichotomous Ilyodon sample from the Rio Armeria (Rio de Comala).
Hardy-Weinberg Presumptive genotypes Association Allele frequencies Separate Pooled to Locus Morph ala alb bib ble bid ale elc old did N X' dj. P P q r s X' P X' P ':-' 102 A 32 35 19 5 2 5 1 3 .52 .39 .06 .02 4.3 .63 >-3 Ada 1.17 3 .77 q B 33 50 18 1 5 1 6 2 1 117 .51 .39 .06 .04 71 <.005 ~ t
I (F) = Probability calculated by Fisher's exact method. TROPHIC DIFFERENTIATION IN ILYODON 267 differentiation of species (Avise, 1974, lation of the trophic types in the Rio Ar 1976; Ayala, 1975), but the possibility that meria is a function of locality, cohort, they are reproductively isolated has not microhabitat, or some other ecological been entirely eliminated. variable. In the Rio Armeria sample (Table 4), A difference in fin colors (first noted by there are statistically significant differ Hubbs and Turner, 1939) may facilitate ences between the genotype frequency dis whatever level of reproductive isolation tributions of the two trophic types at two exists in the Rio Armeria Ilyodon. In a loci, Ldh-3 (P = .022) and Phi-l (P < large fraction of males of the "B" type, .005). There is also a significant difference yellow pigment in the dorsal and/or caudal in allelic frequency atthe Phi-l locus (PA = fins is concentrated into well defined, rel .97, PH = .89, t = 2.44, .02 >P > .01). atively intense, terminal or subterminal There are, in the "B" trophic type only, bands. In the "A" morph, the yellow pig significant departures from the zygotic fre ment, when present at all, is invariably quencies predicted by Hardy-Weinberg diffuse and of reduced intensity. This dif equilibria at two loci: Ada (P < .005) and ference in color pattern could serve as a Phi-l (P = .02). At the Ada locus, 86.3% visual cue in a behavioral isolating mech of the Chi-square value is due to an anism involved in assortative mating. If excess of Phi-e homozygotes. At the this is the case, variability of the bands Phi-l locus, 86.2% of the Chi-square val within the "B" morph may be of signifi ue is due to an excess of Phi-le homozy cance. In our sample, about 70% of the gotes, and 13.1% to a deficiency of "ale" males in the "B" morph had conspicious heterozygotes. At present we have no yellow bands on one or both fins, but ready explanation for the excess of homo roughly 30% lacked the yellow bands en zygotes at the two loci in the B morpho tirely; the mean sizes of the two kinds of Among the potential causes are selection males ("yellow" and "not yellow") could against certain heterozygotes, the presence not be statistically distinguished. of a null allele at one or both loci, inbreed ing, or Wahlund effects. Of these, in DISCUSSION breeding appears intrinsically unlikely in The simplest interpretation of the allo large populations of goodeids, and can zyme data from the Rio Tuxpan (Rio probably be eliminated. Since the hetero "Terrero") sample is that the two types zygote deficiency occurs at two loci, a (morphs) are conspecific and are compo Wahlund effect (produced by subdivision nents of a trophic polymorphism segre of the sample of the B morph with respect gating within a single Mendelian popula to the "e" allele at both loci appears to be tion. In the Rio Armeria (Rio de Comala) the most parsimonious explanation, hut sample, reproductive isolation, or at least we lack data to exclude other factors. some genetic differentiation, probably ex Overall, the allozyme data are consistent ists between the "A" trophic type and at with the hypothesis that there is repro least a portion of the "B" type, but the ductive isolation between the trophic overall genic similarity of the types is types in this geographic sample. However, nonetheless most impressive, and may the virtual allelic identity of the morphs well indicate that their separation was at four of the five polymorphic loci, and very recent. the definite possibility that the "B" type The hypothesis of conspecificity of the itself may be genetically subdivided, sug two trophic types in the Rio Tuxpan, gest caution in the interpretation of the strongly suggested but not proven by our data: a portion of the population with a analysis, is likely to be somewhat contro "B" trophic phenotype might not be re versial. Conspecificity cannot be excluded productively isolated from the "A" type. by the demonstration of additional mor Further work may demonstrate that ge phological differences between the types. netic differentiation or reproductive iso- Unless the number of newly-discovered 268 B. J. TURNER AND D. J. GROSSE
divergent characters is very large, the dif cialized microfauna associated with silt ferences could theoretically be attributed substrates). The selection was almost cer to pleiotropy of the genes encoding the tainly in the direction of increasing trophic trophic features themselves, or to adap specialization. The trophic phenotypes fa tively mediated linkages with those genes. vored by selection were extremes of rela The level of reproductive isolation of the tive mouth (jaw) width, jaw development, types should be assessed directly, and and tooth morphology, number and place preferably by a variety of biological cri ment. Precise combinations of extreme teria. features were involved. (There may have If the trophic types in the Rio Tuxpan been some favorable associations even in are not conspecific, their reproductive iso the original continuous population.) (3.) lation may have evolved by sympatric Since genotypes encoding intermediate speciation involving allelic substitutions at phenotypes or the "wrong" combination but a few critical loci (sensu Tauber and of extreme trophic features were of re Tauber, 1977a, 1977b); their apparent al duced fitness, selection favored the evo lozymic identity and absolute sympatry lution of a polymorphic "system," consist appears much more difficult to explain by ing of alternative arrays of specific (and orthodox geographical or quantum specia mutally exclusive) trophic adaptations. tion (e.g., Dobzhansky et al., 1977). The These polymorphic systems are probably trophic types in the Rio Armeria are prob under the genetic control of "supergenes" ably reproductively isolated, but their (sensu Ford, 1975); each supergene might great overall genic similarity again sug consist of a single, highly pleiotropic gests that their speciation involved mini "principal" locus and modifiers controlling mal allelic substitution, and may well dominance, or of several major loci sep have occurred in sympatry. arately encoding jaw development, tooth The following is a crude model for the number, etc. (4.) In some dichotomous evolution of trophic diversity in Ilyodon. populations the trophic types may have It is suggested by the great genic similarity acquired reproductive isolation, in sym of the sharply distinct trophic types in di patry, via mechanisms similar to those chotomous (bimodal) populations in both discussed in detail by Tauber and Tauber the Rio Tuxpan and Rio Armeria, the vir (1977a). It is not clear whether the acqui tual restriction of these populations to cer sition of reproductive isolation is a nec tain characteristic habitats, and the exis essary consequence of the evolution of tence of "intermediates" (continuous trophic specialization of Ilyodon, or if it populations) in other habitats. (1.) The is simply an alternative (one perhaps ini continuous populations are not interspe tially fostered by selection favoring assor cific hybrids, but are simply trophically tative mating) to a highly developed poly variable Ilyodon populations adapted to morphic system. A central component of low-gradient streams with mud, silt or the model is the hypothesis that trophic pebble substrates and with few large rocks polymorphisms in Ilyodon populations or boulders. (2.) In high-gradient tributar are encoded by supergenes essentially ies with large boulders, disruptive selec analogous to those controlling pigment tion has favored genes encoding extreme and pattern polymorphisms in certain mi trophic phenotypes. (The original Ilyodon metic lepidopterans. If more than one lo colonists of these streams were probably cus is involved in determining the trophic of intermediate phenotype.) The environ phenotypes (and this is very likely the mental basis of the hypothetical disruptive case), this hypothesis is certainly plausible selection are at present unknown, but (see discussions by Ford, 1975; and Turn such habitats may have provided some er, 1967a, 1967b). Its evaluation however, new food sources (e.g., algae or other or will require vastly increased knowledge of ganisms attached to the surfaces of large the genetics underlying the trophic fea rocks) and eliminated others (e.g., spe- tures of both the dichotomous and the TROPHIC DIFFERENTIATION IN ILYODON 269 continuous populations, and may prove types were pooled. Conspecificity of the difficult. Our model does predict that di trophic types appears to be the most chotomous Ilyodon populations should straightforward interpretation of the data: exist in other high-gradient, rock-laden the sharp trophic differences are proba tributaries of both rivers, tributaries not bly a discontinuous polymorphism segre directly connected to the Rio Terrero or gating within a local population of a single Rio de Comala; in addition, they may also biological species. The relationships of the be encountered in equivalent habitats in trophic types in a sample from the Rio de other river systems where Ilyodon occur Comala seem more complex. Of five loci (notably the Rio Balsas). In all such new polymorphic in both types, significant dif cases, the trophic types should be ex ferences between the types were detected tremely similar genetically (whether or not in the allelic frequencies of one locus and reproductive isolation has evolved) and in the genotypic frequencies of a second. should be more closely related to nearby In one of the types, highly significant het continuous populations than to more dis erozygote deficiencies were detected at tant discontinuous ones. two loci. A Wahlund effect (subdivision Polymorphism involving alternative with respect to genotype frequencies at suites of functionally interrelated morpho both loci) is suggested as the best expla logical specializations has now been im nation for simultaneous deficiencies at two plicated in the trophic adaptations of loci. Thus, there is a possibility that one members of three highly divergent groups of the types consists of two or more ge of teleosts: Cichlasoma (Cichlidae, order netically distinct populations. Dichoto Perciformes), Sacco don (Parodontidae, or mous Ilyodon populations may have der Cypriniformes) and Ilyodon (Good evolved from continuous ones in special eidae, order Atheriniformes). This wide habitats by disruptive selection favoring taxonomic distribution strongly suggests extreme trophic phenotypes, followed by that the phenomenon may be a more sig the establishment of clear-cut polymor nificant feature of teleost evolution than phisms probably encoded by supergenes. has been previously appreciated, and may The final step in the process may be the well have been overlooked in many acquisition by the trophic types of repro groups. ductive isolation in sympatry.
SUMMARY ACKNOWLEDGMENTS Certain populations of the fluviatile We thank R. R. Miller for suggesting Mexican fish genus Ilyodon (family Good that a genetic comparison of the Ilyodon eidae), in a few tributaries of two adjacent trophic types might be of interest, and for river systems, are sharply dichotomous placing his extensive personal knowledge with respect to suites of trophic special and collections of the genus at our dispos izations. These trophic types have been al. Dr. Miller collected and studied Ily considered specifically distinct; they are odon populations in the Rio Tuxpan easily differentiated, by eye, as adults and drainage as early as 1955, and our work breed true in the laboratory. Nondicho would have been quite impossible without tomous populations are found in other his pioneering efforts. Thanks are also ex tributaries of the same rivers. In a sample tended to Ms. D. Kingston, now engaged from the Rio Terrero, a tributary of the in a taxonomic revision of the genus; R. Tuxpan (R. Coahuayana), five of 30 though she does not endorse our conclu allozyme loci surveyed were identically sions, Ms. Kingston repeatedly facilitated polymorphic in both trophic types. All ge our studies in the best traditions of profes notype (zygotic) frequencies were in agree sional science. We thank B. L. H. Brett, ment with those predicted by Hardy B. Chernoff, J. Humphries, and P. Yant Weinberg equilibria, and agreement was for helpful discussions, and M. L. Smith maintained when the data from the two and jayantha Wijeyaratne for expert as- 270 B. J. TURNER AND D. J. GROSSE sistance in the field and laboratory, re JOHNSON, G. B. 1977. Assessing electrophoretic spectively. The manuscript profited from similarity. Ann. Rev. Eco!. Syst. 8:309-328. KORNFIELD, I. L. 1978. Evidence for rapid specia editorial suggestions of R. M. Andrews, tion in African cichlid fishes. Experientia 34:335 T. A. Jenssen and J. A. Cranford. 336. The cooperation of the Mexican govern KORNFIELD, I. L., AND R K. KOEHN. 1975. Ge ment, through a collecting permit issued netic variation and speciation in New World by Lie. Alfonso Yanez Ramos, Director cichlids. Evolution 29:427-437. LEWONTIN, R C. 1974. The Genetic Basis of Evo General de Regiones Pesqueras, Instituto lutionary Change. Columbia Univ. Press, N.Y. Nacional de Pesca, is gratefully acknowl MAYNARD SMITH, J. 1966. Sympatric speciation. edged. Supported by NSF grants DEB 76 Amer. Natur. 100:637-650. 20958 (BJT) and DEB 77-17515 (R. R. MILLER, R. R., AND J. M. FITZSIMONS. 1971. Ameca splendens, a new genus and species of Miller). goodeid fish from western Mexico; with remarks on the classification of the Goodeidae. Copeia LITERATURE CITED 1971:1-13. MITTON, J. B., AND R K. KOEHN. 1975. Genetic ALLENDORF, F., N. RYMAN, A. STENNEK, AND G. organization and adaptive response of allozymes STAHL. 1976. Genetic variation in Scandina to ecological variables in Fundulus heteroclitus. vian brown trout (Salmo truua L.): evidence of Genetics 79:97-111. distinct sympatric populations. Hereditas 83:73 MYERS, G. S. 1960. The endemic fish fauna of Lake 82. Lanao, and the evolution of higher taxonomic AVISE, J. 1974. Systematic value of electrophoretic categories. Evolution 14:323-333. data. Syst. Zoo!' 23:465-481. NEVO, E. 1978. Genetic variation in natural pop ---. 1976. Genetic differentiation during specia ulations: patterns and theory. Theoret. Pop. Bio!. tion, p. 106-122. In F. J. Ayala (ed.), Molecular 13:121-177. Evolution. Sinauer, Sunderland. NYMAN, L. 1972. A new approach to the taxonomy AYALA, F. J. 1975. Genetic differentiation during of the "Salvelinus alpinus species complex." Rep. the speciation process. Evo!. Bio!. 8:1-78. Inst. Freshw. Res. Drottningholm No. 52:103 COBBS, G., AND S. PRAKASH. 1977. An experimen 131. tal investigation of the unit charge model of pro ROBERTS, T. R. 1974. Dental polymorphism and tein polymorphism and its relation to the ester systematics in Saccodon, a neotropical genus of ase-S locus of Drosophila pseudoobscura, D. freshwater fishes (Parodontidae, Characoidei). J. persimilis and D. miranda. Genetics 87:712-742. Zoo!' (London) 1973:303-321. COYNE, J. 1976. Lack of genic similarity between SAGE, R D., AND R K. SELANDER. 1975. Trophic two sibling species of Drosophila as revealed by radiation through polymorphism in cichlid fishes. varied techniques. Genetics 84:593-607. Proc. Nat. Acad. Sci. USA 72:4669-4673. DOBZHANSKY, TH., F. J. AYALA, C. L. STEBBINS, SOKAL, R R, AND F. J. ROHLF. 1969. Biometry. ANDJ. W. VALENTINE. 1977. Evolution. W. H. W. H. Freeman and Co., San Francisco. Freeman and Co., San Francisco. TAUBER, C. A., AND M. J. TAUBER. 1977a. A ge FRYER, G., AND T. D. ILES. 1972. The Cichlid netic model for sympatric speciation through Fishes of the Great Lakes of Africa. Oliver and habitat diversification and seasonal isolation. Boyd, Edinburgh. Nature 268:702-705. FORD, E. B. 1975. Ecological Genetics. 4th Ed. ---. 1977b. Sympatric speciation based on al Halstead Press, John Wiley and Sons, N.Y. lelic changes at three loci: evidence from natural HEDRICK, P. W., M. E. GINEVAN, AND E. P. Ew populations in two habitats. Science 197:1298 ING. 1976. Genetic polymorphism in heteroge 1299. neous environments. Ann. Rev. Eco!. Syst. 7:1 TAYLOR, D. W., AND W. L. MINCKLEY. 1966. New 32. world for biologists. Pacific Discovery 19:18-2 2. HENRICSON, J., AND L. NYMAN. 1976. The eco TURNER, J. R. G. 1967a. The evolution of super logical and genetical segregation of two sympatric genes. Amer. Natur. 101:195-221. species of dwarfed char (Salvelinus alpinus (L.) ---. 1967b. Why does the genotype not congeal? species complex). Rep. Inst. Freshw. Res. Drott Evolution 21:645-656. ningholm No. 55:15-37. VRIJENHOEK, R. C. 1978. Coexistence of clones in HUBBS, C. L., AND C. L. TURNER. 1939. Studies a heterogeneous environment. Science 199:549 of the fishes of the order Cyprinodontes. XVI. A 552. revision of the Goodeidae. Misc. Pub!. Mus. WHITE, M. J. D. 1978. Modes of Speciation. W. Zoo!' Univ. Michigan No. 42: 5-80, plates I-V. H. Freeman and Co., San Francisco. HUTCHINSON, G. E. 1978. An Introduction to Pop ulation Ecology. Yale Univ. Press, New Haven. Corresponding Editor: J. R. Powell