|
<br />2748
<br />
<br />Evolution: DeMarais et at.
<br />
<br />MRN, 26, 20, and 3. Samples of G. seminuda used for
<br />mtDNA analyses came from the upper [St. George, Utah (n
<br />= 2)] and lower [Littlefield, Arizona (n = 2)] Virgin River.
<br />For mtDNA, an additional three MRN and 14 G. seminuda
<br />(six from the upper and eight from the lower Virgin River)
<br />were analyzed with two diagnostic enzymes. Morphological
<br />data for 25 artificially produced F 1 hybrids (hereafter referred
<br />to as hybrids) between female G. elegans and male G, r.
<br />robusta are presented for comparative purposes.
<br />Samples of G, elegans were from captive populations at
<br />Dexter National Fish Hatchery, New Mexico. The original
<br />stock was derived from artificial spawning of 11 adults from
<br />Lake Mohave, Arizona-Nevada (20), and the fish were a
<br />composite of F2 progeny from nonassisted spawning in
<br />hatchery ponds. Preserved hybrids were also provided by
<br />Dexter National Fish Hatchery. Specimens of G. seminuda,
<br />MRN, and G. r, robusta were collected from wild popula-
<br />tions; locality data are available from the authors,
<br />Measurements by digital or dial calipers (nearest 0.1 mm)
<br />of 24 morphometric variables were made on preserved spec-
<br />imens housed at Arizona State University. Methods were
<br />those of Hubbs and Lagler (21) as modified by DeMarais (22).
<br />All computations were made on an IBM 3090 using the
<br />Statistical Analysis System (SAS) (23). Variation among
<br />samples was assessed by principal component (PC) analysis
<br />sheared to reduce the etTects of overall size on shape varia-
<br />tion (24). Components were derived from the covariance
<br />matrix of 10glO-transformed morphometric variables and
<br />were sheared by locality.
<br />For analysis of allozymic variation, muscle and liver sam-
<br />ples stored at -80"C were homogenized in distilled water.
<br />Gene products representing 30 presumptive loci (ACP-A,
<br />mAH-A, sAH-A, AK-A, ADH-A, sAP-A, mAAT-A, sAAT-A,
<br />CBP-I, CBP-2, CK-A, EST-I, EST-2, FH-A, mIDHP-A,
<br />sIDHP-A. WH-A, WH-B, mMDH-A, sMDH-A, sMDH-B,
<br />mMEP-A, sMEP-A, PEPB, PEPD, PEPA, PEPS, PGM~A,
<br />PK-A. sSOD-A) were resolved by electrophoresis ofhomoge-
<br />nates through 12% starch gels (25). Locus nomenclature
<br />followed that of Shaklee et al. (26). ButTer conditions, EC
<br />numbers, and tissue sources are ,available from the authors.
<br />mtDNA was isolated from heart, liver, and gonads (when
<br />available) dissected from specimens stored at -80oC. Meth-
<br />ods for isolation and analysis were as described by Dowling
<br />et al. (27). mtDNAs were characterized by digestion with the
<br />following 6-base-recognizing restriction endonucleases:
<br />BamHl, Bel I, BgI II, BstEII, EcoRI, HindIIl, Neo I, Nde I,
<br />Nhe I, Pvu II, Sac I, Xba I, and Xho I.
<br />Relationships among taxa were estimated independently
<br />for each set of characters. For morphology, mean scores on
<br />sheared principal components 2,3, and 4 for each population
<br />were used to calculate average taxonomic distances between
<br />all pairwise combinations of taxa (28). Relationships were
<br />visualized by using the FITCH algorithm of PHYLlP (29), a
<br />least-squares-based method that makes no assumption con-
<br />cerning equality of evolutionary rates. For allozymes, mod-
<br />ified Rogers genetic distances were calculated from allele
<br />frequencies (30), and relationships were visualized by using
<br />the DISTANCE WAGNER algorithm of BIOSYs-1 (31). Estimates
<br />of sequence divergence among mtDNAs were calculated
<br />from fragment comparisons (32), and relationships were
<br />envisioned by FITCH.
<br />
<br />RESULTS
<br />
<br />Morphology. All nominal taxa were readily differentiated
<br />by PC analysis (Fig. 1). The first PC (PCl) and sheared PC2
<br />(H2), respectively, accounted for 93.6% and 4.6% of the total
<br />sample variation. PCl was interpreted as a general size factor
<br />and did not contribute to intersample ditTerentiation. In
<br />contrast, H2 was a size-free shape component that ditTeren-
<br />
<br />Proc. Natl. Acad. Sci. USA 89 (1992)
<br />
<br />f
<br />I
<br />
<br />-0.~0
<br />
<br />.0,35
<br />
<br />e- B:------
<br />
<br />V~/B
<br />y~ ~~
<br />\l s-J-s/s
<br />A~ ~
<br />
<br />~\__:=i
<br />~'-.--- H
<br />H-H/
<br />
<br />-0.40
<br />
<br />-0.45
<br />
<br />~ .0,50
<br />
<br />.0,55
<br />
<br />-0,60
<br />
<br />.0.65
<br />
<br />E/E~
<br />
<br />/ E
<br />
<br />E..E-E/
<br />
<br />-],5 -],0 .6,5 .6,0 -5,5 .5,0
<br />PC1
<br />
<br />FIG. 1. Plot of scores on PCI and sheared component 2 (H2),
<br />Sample designations: G. r. robusta (B, Bill Williams; V. Verde; and
<br />S, Salt), G. seminuda (squares), MRN (triangles), G. elegans (E), (G.
<br />elegans x G. r. robusta}Fl hybrids (H). Only the outermost speci-
<br />mens for each sample are shown, connected by polygons to encom-
<br />pass aU other individuals. Sample sizes are given in the text.
<br />
<br />tiated among taxa (Fig. 1), with length and depth of caudal
<br />peduncle contributing most strongly to separations. Relative
<br />to G. r. robusta,G. elegans had a longer and shallower caudal
<br />peduncle. Hybrids were morphologically intermediate. Al-
<br />though distinct, G. seminudaexhibited H2 scores that over-
<br />lapped those of known hybrids. MRN and populations of G.
<br />r. robusta exhibited considerable overlap, with only the
<br />scores for MRN approaching those of G. seminuda. MRN, G.
<br />seminuda, and hybrids were slightly differentiated from G. r.
<br />robusta and G. elegans on H3 and H4 (data not shown),
<br />components that accounted for <2.0% of total variation.
<br />Clustering of distances generated from mean H2, H3, and
<br />H4 scores summarized morphological relationships among
<br />and between taxa (Fig. 2A). Populations of G. r. robusta and
<br />G. et'egans were most divergent. MRN was most closely
<br />linked to G. r. robusta. Hybrids and G. seminuda were
<br />similar, the distance between them only slightly greater than
<br />those separating populations of G. r. robusta. Hybrids,
<br />however, were more similar to G. elegans, while G, semin-
<br />uda more closely resembled G. r. robusta.
<br />AUozymes. Levels of allozyme variation were low, both
<br />within and between taxa. Maximum genetic distances oc-
<br />curred between G. elegans and G. r. robusta (Fig. 2B),
<br />largely due to fixed or nearly fixed differences at two loci. All
<br />G. elegans were homozygous for a fast allele at both CK-A
<br />and CBP-I. All G. r. robusta (representing three populations)
<br />had only an alternative, slow CK-A allele. Salt River G, r,
<br />robusta were fixed for a slow CBP-I allele, while Bill
<br />Williams and Verde River populations possessed the slow
<br />allele along with the fast (G. elegans-type) allele at low
<br />frequency (12% and 5%, respectively).
<br />MRN and G, seminuda were intermediate to G. r. robusta
<br />and G. elegans (Fig. 2B) due to the presence of both alleles
<br />at CK-A and CBP-I. MRN possessed a higher proportion of
<br />the fast allele at both loci (68% and 60%, respectively),
<br />clustering it with G. elegans, while G. seminuda aligned with
<br />G. r. robusta due to lower proportions (29% and 47%) of these
<br />same alleles. All possible combinations of genotypes (i,e.,
<br />heterozygotes and alternative homozygotes) for the two
<br />marker loci were present. Based on G tests, both loci were in
<br />
|