|
<br />B, Af10zymes C, mlDNA
<br /> B;' ~'~] n._]
<br /> Salt robusta V robusta
<br /> erde
<br /> Verde SaIl
<br /> MRN
<br /> seniinuda OW"'"]
<br /> MRN seminuda
<br /> WFD
<br /> elegans elegans
<br />I I I I I I
<br />0,00 0,10 0,20 0.0 1.0 2,0
<br />
<br />1
<br />
<br />Evolution: DeMarais et at.
<br />
<br />A. Morphology
<br />
<br />Bill WilliamJ
<br />
<br />robusta
<br />Verde
<br />Sail
<br />MRN
<br />
<br />
<br />
<br />hybrids
<br />
<br />
<br />elegans
<br />
<br />I -. I I
<br />0,00 0,05 0,10
<br />
<br />Proc. NaIl. Acad. Sci. USA 89 (/992)
<br />
<br />2749
<br />
<br />FIG. 2, Dendrograms (rooted at the midpoint) derived from morphology (A), a1lozymes (8), and mtDNA (C), as described in the text.
<br />
<br />Hardy-Weinberg equilibrium in MRN and G. seminuda, and
<br />multilocus tests failed to detect significant association of
<br />alleles.
<br />mtDNA. Limited mtDNA differentiation (0.0-0.2% se-
<br />quence divergence) was present within and between popu-
<br />lations of G. r. robusla. The mtDNA of G. elegans was
<br />distinct from that of G. r. robusta (3.2-3.3% sequence
<br />divergence) but similar to those of MRN and G. seminuda
<br />(0.2-0.4%). As in G. r. robusta, limited variation was present
<br />within and between samples of G. seminuda and MRN
<br />(0.3-0.5%). The mtDNAs of G. elegans, G. seminuda, and
<br />MRN clustered together, as did those of the various G. r.
<br />robusta (Fig. 2C). mtDNAs were isolated from an additional
<br />14 G. seminuda (from two localities) and 3 MRN to further
<br />test for G. r. robusta-like mtDNA in the Virgin River system.
<br />Cleavage with two restriction enzymes diagnostic for G. r.
<br />robusta andG. elegans mtDNA (Bel I and HindIIl) yielded
<br />only typical G. elegans fragment patterns in all individuals.
<br />
<br />DISCUSSION
<br />
<br />Wagner (33) described three steps in the study of hybrid taxa:
<br />(i) detection of hybrids, (ii) test of postulated hybridity, and
<br />(iii) phylogenetic placement. G. seminuda was recognized
<br />early as a morphological link between G. elegans and G. r.
<br />robusta. Ellis (34) and Miller (15) thought it an intergrade
<br />between G. elegans and G. r. robusta (then considered
<br />subspecies), or an intermediate subspecies. Carl L. Hubbs
<br />speculated in 1938 that Gila from the Virgin River drainage
<br />(specifically the Moapa River) were of hybrid origin (35);
<br />Smith et al. (18) were the first to actually suggest the hybrid
<br />hypothesis in print. Thus, Wagner's step 1 has been welI
<br />established. This paper provides a test of hybrid origin (step
<br />2), while phylogenetic analysis (step 3) remains to be exam-
<br />ined.
<br />_ Morphological,-allozymic, and,mtDNA data are congruent
<br />with respect to the divergence and distinctiveness of the
<br />putative parental species, G. elegans and G, r. robusta,
<br />However, our analyses produced discordant estimates of
<br />relatedness for G. seminuda and MRN relative to G. elegans
<br />and G. r. robusta (Fig. 2). Two alternative hypotheses could
<br />explain these conflicting results: (i) expression of ancestral
<br />polymorphisms, and (ii) hybrid origin,
<br />Ancestral Polymorphism vs. Hybrid Origin. As populations
<br />diverge, relationships may become obscured by the stochas-
<br />tic nature of inheritance. If this were the case in Gila,
<br />discordance among inferred relationships across character
<br />types could reflect insufficient time for ancestral polymor-
<br />ph isms to become fixed (i.e., incomplete lineage sorting) and
<br />for new diagnostic characters to evolve.
<br />Given the magnitude of morphological differences between
<br />G. elegans and G. r. robusta, stochastic inheritance from a
<br />polymorphic ancestor seems unlikely. An ancestral form
<br />
<br />would have to possess the entire range of variation and then
<br />randomly give rise to populations exhibiting different mean
<br />phenotypes with only a fraction of the original variance,
<br />While morphological intermediates such as G. seminuda
<br />might be expected under such a scenario, the existence of a
<br />highly variable ancestor is more difficult to envision because
<br />modem species generalIy do not possess such high levels of
<br />morphological variation.
<br />Patterns of allozyme variation are consistent with differ-
<br />ential sorting of alleles among lineages, which would require
<br />only that an ancestral form be polymorphic at the two
<br />diagnostic loci. Thus, MRN and G. seminuda could have
<br />retained the ancestral condition, while G. r. robusta and G.
<br />elegans became fixed for alternative alleles.
<br />Because of its haploid and strict maternal mode of inher-
<br />itance without recombination, differential sorting of ancestral
<br />polymorphism is especially likely for mtDNA (36). Thus,
<br />similarity of G. seminuda, MRN, and G. elegans mtDNAs
<br />could result from chance inheritance of similar haplotypes
<br />from a polymorphic ancestor that possessed restriction sites
<br />currently diagnostic for G. r. robusta and G. elegans. Our
<br />data, however, are inconsistent with this hypothesis. Esti-
<br />mates of sequence divergence between mtDNAs of G. ele-
<br />gans and G. r. robustawere 8-fold greater than those among
<br />G. seminuda, MRN, and G. elegans (0.4% orless) and 16-fold
<br />greater than those among populations of G. r. robusta (0.2%
<br />or less). Therefore, MRN and G. seminuda apparently share
<br />with G. elegans an mtDNA ancestor not shared with G. r.
<br />robusta.
<br />Consideration of one data set in light of the other empha-
<br />sizes additional inconsistencies. For example, if virtually
<br />identical mtDNAs of G. seminuda, MRN, and G. elegans
<br />reflect true relationships, then morphological convergence of
<br />G. seminuda and especially MRN toward a G. r. robusta
<br />phenotype, or divergenceofG. elegans from a G, r. robusta-
<br />like ancestor, would have to have been recent and extremely
<br />rapid. However, available evidence suggests that this was not
<br />the case. Most modem species of western cyprinoid fishes
<br />were welI differentiated by the Pliocene (37), and late Mi-
<br />ocene fossils in Arizona have been referred to G. r. robusta
<br />(38), Some of these had skeletal features reminiscent of G,
<br />elegans and/or G. cypha, suggesting that morphological
<br />divergence of Colorado River basin Gila was already under
<br />way >6 million years ago, Alternatively, if morphology
<br />accurately reflects among-taxa relationships, then mtDNA
<br />data become difficult to interpret. For example, between the
<br />morphologicalIy similar G. r, robusta and MRN, both rapid
<br />divergence of mtDNAs and unreasonable convergence of
<br />MRN haplotypes toward those of G. elegans would be
<br />necessary to create the observed relationships.
<br />The strongest argument against the ancestral polymor-
<br />phism hypothesis is the general agreement of the three
<br />character sets. It seems unlikely that a stochastic process
<br />
|