Evidence of a biogeographic break between populations of a high dispersal starfish: congruent regions within the Indo-west Pacific defined by color morphs, mtDNA, and allozyme data.
An excellent model with which to examine these phenomena is the ubiquitous, tropical starfish Linckia laevigata. This species is continuously distributed throughout the Indo-West Pacific. It is renowned for its brilliant, royal blue color, although it also occurs in various shades of blue, brown, apricot, salmon-orange, gray or purple, sometimes with colors on the aboral surface different from those on the oral surface (Marsh 1974, 1977; Clark and Courtman-Stock 1976; STW, pers. obs.). The royal blue morph of L. laevigata predominates in the Pacific and in reefs off north Western Australia. It is usually the only color morph in West Pacific populations, but in some localities (e.g., Japan, Palau and the Philippines) starfish that are either apricot all over, or apricot on the oral surface and blue on the aboral surface occur in high frequencies. There are many different color morphs which are recorded only from the Indian Ocean, and at most Indian Ocean localities color morphs other than the royal blue morph predominate. For instance, although the royal blue morph occurs, a pale blue morph is more common at Inhaca Island, Mozambique (G. Branch, pers. comm. 1997) and at Aldabra Atoll, Western Indian Ocean (Sloan et al. 1979).
Samples collected for this study from the Indian Ocean at Thailand and South Africa included many different color morphs (vivid salmon-orange, gray-purple, dark purple, gray, brown, slate-blue, or gray-blue) and although the morphs differed at each location, the most common at both locations was a salmon-orange morph (as distinct from the apricot morph). Apart from the Indian Ocean the salmon-orange morph has also been recorded rarely from Japan (K. Okaji, pers. comm. 1996), the Philippines (C. Ablan, pers. comm. 1996), and may occur throughout the Indo-Malay region.
The color in royal blue specimens of L. laevigata arises from the cooccurrence of two carotenoproteins (Zagalsky et al. 1989). Carotenoproteins are frequently responsible for the rich blue, purple, violet, pink, or brown colors of many asteroids, while nonconjugated carotenoids impart yellow, orange, or red colors (Fox 1947). Heritable changes to the factors responsible for the quaternary structure of the protein portion of the pigment, or regulation and/or conjugation of these proteins, may be responsible for the different color morphs in L. laevigata, in which case it is possible that the nonuniform geographic range of color morphs in L. laevigata is evidence that dispersal is not uniform throughout the entire Indo-West Pacific region.
A number of other marine taxa also exhibit different color morphs in the Pacific and Indian Oceans. For instance, the common shore crab Ocypode cerathopthalma is greenish in the Indian Ocean and white with brown markings in the Pacific (R. George, pers. comm. 1996) and the crown-of-thorns starfish Acanthaster planci is gray with dull red markings in the Pacific but iridescent blue or pink in the Indian Ocean (Benzie 1992). Distributions of pairs of sibling and subspecies also show the same geographic segregation between the Indian and Pacific Oceans. For instance, the starfish Culcita novaeguineae is found in the Pacific Ocean and is replaced by sibling species C. schmideliana in the Indian Ocean (Yamaguchi 1977b) and five species of siganid fish have distributions centered on the Pacific Ocean, each with a sibling species occurring in the Indian Ocean and an area of overlap in the Indo-Malay region (Woodland 1983).
While there are abundant distribution data that suggest differences between the Indian and Pacific Oceans, direct evidence for genetic segregation is limited. Two monophyletic species groups of Indo-West Pacific butterfly fishes demonstrate that, at least in some groups, a geographic segregation between the Indian and Pacific Oceans is reflected in a genetic discontinuity (McMillan and Palumbi 1995). Allozyme electrophoresis and mtDNA variation among populations of the coconut crab Birgus latro (Lavery et al. 1995, 1996) and allozyme variation for four species of damselfish (Lacson and Clark 1995) also showed that Pacific populations were strongly differentiated from one Indian Ocean population sampled and allozyme data showed the same pattern for the starfish A. planci (Benzie, unpubl. data).
Allozyme and mtDNA surveys have demonstrated that, despite the presence of some mtDNA structure, gene flow is high over long distances among West Pacific and Western Australian populations of L. laevigata (Williams and Benzie 1993, 1996, 1997), consistent with high dispersal of the larval phase, which lasts 24 to 28 days in laboratory trials (Yamaguchi 1977a). However, these surveys were limited to populations where the royal blue morph was predominant. To determine whether the geographic grouping of different color morphs of L. laevigata is the result of nonuniform gene flow among populations, genetic variation was examined in samples from 10 sites at two localities in the Indian Ocean where the orange morph predominated, using seven allozyme loci and PCR-RFLP variation of the putative control region and flanking sequence, digested with seven restriction endonucleases. These data were then compared with allozyme data previously obtained from 28 sites and RFLP data from 15 sites in Western Australia and the West Pacific Ocean where the royal blue morph is predominant (Williams and Benzie 1993, 1996, 1997).
MATERIALS AND METHODS
Collection and Storage of Samples
A total of 92 individual L. laevigata were sampled from Thailand and South Africa in the Indian Ocean. Individuals sampled were vivid salmon-orange, gray-purple, dark purple, gray, brown, slate-blue, or gray-blue, but not the royal blue sampled previously throughout the West Pacific. In many individuals, the color of the aboral surface differed from that of the oral surface. Thailand samples were collected from Kata Beach, Phuket Marine Station, and Waeo Island. South African samples were collected from seven reefs within the St. Lucia and Maputaland Marine Reserves in South Africa: 2 Mile Reef, Saxon Reef, Reef #25, Reef #13, Reef #32, Reef #21, and Linckia Reef. Starfish were collected by scuba from coral reefs at 4-5 m depth in Thailand and 13-20 m depth in South Africa. Samples of pyloric caeca were collected following the protocol in Williams (1992).
Allozyme variation at seven polymorphic loci was compared with data previously obtained from 1628 individuals sampled from 23 sites within the West Pacific Ocean and five sites within Western Australia (East Indian Ocean) (Williams and Benzie 1993, 1996). Variation was examined at the same seven polymorphic loci as used in previous studies: enolase (ENO), glucose-6-phosphate isomerase (GPI), hexokinase (HK), peptidase using leucylglycylglycine substrate (LGG), peptidase using leucylproline substrate (LP), peptidase using leucyltyrosine (LT), and superoxide dismutase (SOD). Variation observed at LGG* was identical to that at the second locus of LT and has been reported here as LT-2*. Electrophoresis was performed using cellulose acetate gels (Cellogel[TM], Chemetron, Milan) (ENO, GPI) or horizontal starch gels (HK, LT, LP, and SOD) following methods outlined in Williams (1992).
Endonuclease Restriction Analysis
RFLP data were obtained from 29 individuals from South Africa and 15 from Thailand and compared with data previously obtained from 326 individuals from Western Australia (Indian Ocean), the Great Barrier Reef (GBR), Fiji, and the Philippines (all Pacific Ocean) (Williams and Benzie 1997).
A region of approximately 1100 bp of mtDNA (including the 3[prime] end of the 16 S rRNA gene, the putative control region, two transfer RNA genes [tRN[A.sub.thr] and tRN[A.sub.glu]] and a portion of the 12 S rRNA gene] was amplified and PCR products were subjected to endonuclease digestion with each of Dpn II, Dra I, Hinf I, HinP1 I, Rsa I, Msp I, and [Tag.sup.[Alpha]] I in individual reactions following the protocol in Williams and Benzie (1997). The presence or absence of restriction sites was inferred from fragment patterns, and composite haplotypes were assigned to each individual.
Statistical Analyses: Allozymes
Calculations of allele frequencies and population genetic statistics were carried out using the BIOSYS-1 package (Swofford and Selander 1981). Chi-square goodness-of-fit tests were performed to test conformance of populations to Hardy Weinberg equilibrium (HWE), using observed genotype frequencies and those expected under HWE with exact significance probabilities. Measures of genetic variability included the average heterozygosity (direct count) and the average number of alleles per locus. Cavalli-Sforza and Edwards (1967) arc distance and Nei (1978) unbiased genetic distance were also calculated and the neighbor-joining algorithm (Saitou and Nei 1987) in PHYLIP (Felsenstein 1993) was used to produce a unique, unrooted tree from Cavalli-Sforza and Edwards (1967) arc distance data.
Distributions of genotypes between sites within a location were compared using [[Chi].sup.2] heterogeneity analysis applying Yates (1934) correction. Weir and Cockerham (1984) F-statistics were calculated and the equations in Waples (1987) and Workman and Niswander (1970) were used to calculate contingency chi-square tests to test the significance ([H.sub.0]: [F.sub.IS] or [F.sub.ST] = 0) of single-locus values of [F.sub.IS] and [F.sub.ST] using the sequential Bonferroni method to correct for multiple tests. Rare alleles were successively pooled until all expected chi-square values were greater than one, and no more than 20% in a given test were less than five. Tests where these conditions could not be met, when all but the most common allele had been pooled, were excluded from the analysis. The included chi-squared values and their degrees of freedom were summed across loci to test the significance of the average [F.sub.ST] across loci. Again, the sequential Bonferroni procedure was used to correct for the increase in type 1 errors. Unbiased estimates of [F.sub.ST] and 95% confidence limits were calculated by jackknifing over loci using the equation of Johnson et al. (1988). The average number of migrants per generation ([N.sub.e]m) was estimated using the equation for an island model, using the jackknifed estimate of [F.sub.ST].
[F.sub.ST]-values from pairwise comparisons between sites were plotted as a function of geographic distance to test whether genetic structure was the result of isolation by distance. Geographic distance was estimated as the shortest distance by water between two points. All negative [F.sub.ST]-values were converted to zero (negative [F.sub.ST] can be considered to be the biological equivalent of a zero [F.sub.ST]; Weir 1990), and the lowest positive data value (0.001) was added to all data, before taking the log. Regression analyses were performed using EXCEL, and Mantel's normalised statistic Z was calculated using NTSYS-PC (Rohlf 1990). The significance of Z was tested by 10,000 random permutations.
To quantify the changes in genetic differentiation over different spatial scales, a three-level hierarchical analysis of Wright's (1978) [F.sub.ST] was performed using the BIOSYS-1 program. Genetic variance was partitioned accordingly: (1) sites within localities (three sites in Thailand, seven sites in South Africa, 12 sites in East Australia, five sites in Western Australia, two sites in Japan, two sites in the Philippines, two sites in Fiji, two sites in the Solomon Islands, two sites in Guam, and one site in New Caledonia); (2) localities within color morph groups (predominantly blue [East Australia, Western Australia, Japan, Philippines, Fiji, Solomon Islands, Guam, and New Caledonia] or predominantly orange [Thailand and South Africa]); and (3) between color morph groups (predominantly blue or predominantly orange).
Statistical Analyses: RFLP
Measures of diversity within populations were estimated using the DA program in REAP (McElroy et al. 1992). Genetic heterogeneity within populations was estimated by haplotype diversity (h) and nucleotide diversity ([Pi]) within populations, and nucleotide divergence ([d.sub.XY]) among populations (Nei and Tajima 1981). The UPGMA algorithm in PHYLIP (Felsenstein 1993) was used to produce a unique, unrooted tree from nucleotide divergence data among populations.
To test for the presence of population structure, data were analyzed using AMOVA (Excoffier et al. 1992 [in the Arlequin vers. 1.0 package] Schneider et al. 1996), which produces estimates of variance components and F-statistic analogues ([Phi]-statistics), [Phi]-statistics were "haplotypic statistics" (where interhaplotypic distances are presumed to be equal), which are comparable to Weir and Cockerham's (1984) [F.sub.ST] (Excoffier et al. 1992), or "sequence statistics" (having additionally incorporated sequence divergence between haplotypes; Hudson et al. 1992). Hierarchical analysis of the two color morph groups (populations within color morph groups) used sequence statistics. The distance data used in the calculation of [Phi] sequence statistics was Nei and Tajima's (1981) nucleotide divergence between haplotypes, which was calculated using DA in REAR Significance levels of the [Phi]-statistics were determined by comparing observed values with random values calculated from 10,000 random permutations (for more detail see Excoffier et al. 1992). [[Phi].sub.ST] values from pairwise comparisons between sites were then plotted as a function of geographic distance to test whether genetic structure was the result of isolation by distance, using the same method as for allozyme data, except that 0.00001 was added to all haplotypic data before taking the log (instead of 0.001).
The spatial distribution of haplotypes among populations was also examined using the MONTE program in REAP (McElroy et al. 1992), which calculates contingency [[[Chi].sup.2] without pooling of rare haplotypes. Data were analysed among sites within locations, and among locations. Significance levels were determined by the randomisation procedure of Roff and Bentzen (1989). Each dataset tested was randomised 10,000 times.
A character state matrix, with the presence or absence of presumptive sites, was created using the GROUP and GENERATE programs in REAP (McElroy et al. 1992). This matrix was used to construct unrooted mtDNA haplotype phylogenies for L. laevigata using PAUP with the heuristic search setting (Swofford 1990). The 47 haplotypes identified in a previous study were also included in the analysis for completeness. The tree bisection and reconnection (TBR) branch swapping algorithm was used to search for optimal trees within this framework. Restriction sites were treated as relaxed Dollo characters, with gains weighted twice as heavily as losses (McMillan and Bermingham 1996). The search was terminated after reaching a maximum of 1000 trees. A single, consensus tree was obtained by using the 50% majority rule consensus.
Although different color morphs (salmon-orange, gray-purple, slate-blue, gray, gray-blue, or brown) were sampled at each of the new sites in the Indian Ocean, there were no deviations in allozyme allele frequencies within each population from those expected under conditions of HWE. Observed direct-count heterozygosities in each population were within one standard error of those expected under conditions of HWE. Chi-square tests using exact significance probabilities also showed no significant deviations of allozyme genotype frequencies for each of the seven loci from those expected under conditions of HWE. Preliminary analysis of distance data where color morphs at each location were treated [TABULAR DATA FOR TABLE 1 OMITTED] as separate populations, clustered color morphs by geographic location rather than color. RFLP variation also showed no pattern linking any haplotype with a particular color morph. The data suggest that color variation within local populations is representative of intraspecific variation. Data from all color morphs present in Thailand or South Africa were pooled at each site in subsequent analyses.
Measures of allozyme variation indicated high levels of diversity within all populations of L. laevigata, similar to those observed in Pacific populations. The mean number of alleles per locus ranged from 3.9 in South Africa to 4.7 in Thailand, and mean heterozygosities based on seven allozyme loci ranged from 0.361 in Thailand to 0.375 in South Africa (both within one standard error of expected values under conditions of HWE). Both measures of diversity for orange morph populations are similar to those calculated for combined data for blue morph populations (Williams and Benzie 1993, 1996). Slightly lower values for the mean number of alleles in orange morph populations probably reflect differences in sample sizes.
The seven restriction endonucleases identified eight haplotypes in 41 animals. Seven of these eight haplotypes had not previously been identified in Pacific or Western Australian populations (the exception being #6) (Table 1). The two most common haplotypes in the Pacific and Western Australian populations (#1 and #2) were not found in either Thailand or South Africa. Average percent nucleotide diversity was higher in South Africa than Thailand (2.88% and 1.56%, respectively), but average haplotype diversity was lower in South Africa than Thailand (0.63 and 0.85, respectively). Both measures of diversity were within the range observed within blue morph populations (Williams and Benzie 1996).
Both allozyme and RFLP data demonstrated that there was no genetic structure among sites within Thailand. The data from these sites were pooled within each location in all further analyses, and results are presented only for pooled data. Site data were also pooled for South Africa because although the among site differences in RFLP haplotype distribution were significant, there were no significant differences in allozyme data; overall or at individual loci and sample sizes at each site were small in RFLP analyses (ranging from n = 1 at two sites to n = 9).
Allozyme and RFLP data both suggest that Thailand and South Africa are genetically differentiated. Allozyme data suggest that Thailand and South Africa are more closely related to each other than to the blue morph group (Western Australia and West Pacific). [[Phi].sub.ST]-values incorporating sequence data are consistent with this pattern of genetic structure, but spatial variation of RFLP haplotypes can be explained, to a large extent, by isolation by distance. Allozyme and RFLP results are discussed separately below.
Cavalli-Sforza and Edwards (1967) arc distances ([D.sub.C]) and Nei (1978) unbiased genetic distances ([D.sub.N]) between Thailand and South Africa are higher than those observed among locations in the blue morph group. However genetic distances are even higher between orange morph populations (Thailand and South Africa) and blue morph populations (Western Australia, Fiji, GBR, Philippines) (0.276 [less than or equal to] [D.sub.C] [less than or equal to] 0.338, mean = 0.308; 0.078 [less than or equal to] [D.sub.N] [less than or equal to] 0.157, mean = 0.118) (Table 2). Cluster analysis of [D.sub.C] data using the neighbor-joining algorithm (Saitou and Nei 1987) in PHYLIP (Felsenstein 1993) also shows strong grouping of populations by color morph [ILLUSTRATION FOR FIGURE 1 OMITTED].
The most common blue morph allele (LT-1*100) does not occur in any of the orange morph populations, and two other alleles (LT-1*112 and LT-1*110) predominate (Table 3, [ILLUSTRATION FOR FIGURE 2 OMITTED]; Williams and Benzie 1996). South Africa is additionally differentiated from blue morph populations by gene frequencies at ENO* and SOD*, with South African populations showing increased frequencies of SOD*160 and ENO*78 relative to Thailand.
The average [F.sub.ST] between the two orange morph populations (Thailand and South Africa) is significantly different from zero ([F.sub.ST] = 0.047, [N.sub.e]m = 5.1), as are individual locus [F.sub.ST]-values at ENO*, HK*, LP*, and SOD* (Table 4). Chi-square analysis of the distribution of genotypes also shows significant differences between South Africa and Thailand at ENO*, HK*, and SOD* and at the overall total (not shown).
Marked heterogeneity is observed at individual locus [F.sub.ST]-values between orange and blue morph populations, with particularly high [F.sub.ST]-values for LT-1*. However, [F.sub.ST]-values within the Indian ocean (Thailand and South Africa [orange morph] [TABULAR DATA FOR TABLE 2 OMITTED] and Western Australia [blue morph]) are significant at all individual loci (average [F.sub.ST] = 0.073), with the exception of GPI* and LT-2*. The [F.sub.ST] averaged over all loci for the 10 orange and blue populations indicates the presence of significant genetic structure ([F.sub.ST] = 0.030). However, when data from Thailand and South Africa are compared with pooled data for all the blue morph populations (which were not significantly differentiated in Williams and Benzie 1996), the average [F.sub.ST] rises to 0.095 which is slightly higher than that observed among all three Indian Ocean populations (South Africa, Thailand, and Western Australia). The jackknifed estimate of [F.sub.ST] (and 95% confidence limits) among the three differentiated groups (South Africa, Thailand, and blue morph) is 0.082 ([+ or -] 0.009, 95% C.I.) and [N.sub.e]m = 2.8 (2.5-3.2, 95% C.I.).
A three-level hierarchical analysis of F-statistics, performed using Wright's (1978) [F.sub.ST], demonstrated that more than half (56.7%, [F.sub.XY] = 0.056) of the total variance among populations is the result of differences between color morph groups and the remainder is due to differences among sites within locations ([F.sub.XY] = 0.047). However, most of the differences between sites is due to the very small sample sizes from numerous sites in South Africa and Thailand. Repeating the analysis with data pooled for sites (some of which have only one sample) within South Africa and within Thailand, demonstrates that more than two-thirds (69.0%, [F.sub.XY] = 0.017) of the total variance among populations is the result of differences between color morph groups, 26.9% ([F.sub.XY] = 0.007) is due to the variance among localities within color morph groups, and only 4.1% ([F.sub.XY] = 0.0.001) is due to the variance among sites within blue morph localities.
A plot of transformed pairwise [F.sub.ST] values as a function of log (geographic distance [d]) revealed a significant, but weak pattern of isolation by distance ([r.sup.2] = 0.18; Mantel's  normalised statistic Z = 0.424, P = 0.01). A tighter relationship was observed when New Caledonia (which had a small sample size) was removed from the analysis ([r.sup.2] = 0.31; [ILLUSTRATION FOR FIGURE 3 OMITTED]). However the additional significance was related to a regression between two clusters of data points. One cluster was related to pairwise comparisons including either South Africa or Thailand, and the other cluster included all pairwise analyses among Pacific locations. No pattern of isolation by distance was detected within either cluster.
Analysis of geographic heterogeneity in haplotype frequencies using a Monte-Carlo simulation and [Phi]-statistics revealed significant differentiation between Thailand and South Africa, among all six Indo-West Pacific locations, and between color morph groups (Table 5). Regional differentiation in most cases was more apparent when [Phi]-statistics incorporated divergence data between haplotypes, suggesting that molecular distances are larger for pairs of haplotypes drawn from different regions, than from the same region (Excoffier et al. 1992). The exception being a comparison between Thailand and South Africa, which suggests that while haplotypes vary between these populations, the molecular distances between pairs of haplotypes drawn from each population are low.
A pairwise analysis of [[Phi].sub.ST]-values (incorporating divergence data) among locations showed that all combinations were significantly differentiated at the 0.05 level (after a sequential Bonferroni test; Rice 1989; not shown) with three exceptions; there were no significant differences between Western Australia and the Philippines, between Thailand and South Africa, or between Thailand and the Philippines. Differences between any blue morph population and South Africa were similar to those between the same population and Thailand, despite the additional geographic distance separating the population from South Africa.
TABLE 3. Allele frequencies for the seven polymorphic loci screened in Linckia laevigata collected from Thailand (combined data for three sites) and South Africa (combined data for seven sites); n: sample size. Data for Western Australia and West Pacific sites are in Williams and Benzie (1993, 1996). Locus Thailand South Africa GPI* 125 - 0.010 114 0.013 0.010 110 0.038 0.010 100 0.837 0.817 90 0.038 0.096 86 0.050 0.038 80 0.013 0.019 66 0.013 - ENO* 119 0.025 - 100 0.512 0.265 78 0.463 0.735 HK* 113 0.063 0.019 107 0.150 0.394 100 0.663 0.558 92 0.075 0.029 81 0.013 - 79 0.038 - LT-I* 112 0.488 0.625 110 0.475 0.375 106 0.013 - 105 - - 100 - - 98 0.025 - LT-2* 109 0.038 0.058 100 0.925 0.865 89 0.038 0.077 LP* 109 0.013 - 103 0.087 0.048 100 0.762 0.577 94 0.038 0.058 92 0.038 0.192 87 0.063 0.106 85 - 0.010 83 - 0.010 SOD* 160 0.025 0.202 100 0.950 0.798 33 0.013 - 25 0.013 - n 40 52
Pairwise [[Phi].sub.ST] values were log transformed and plotted as a function of log geographic distance (d). Haplotypic data ([[Phi].sub.ST] [hap]) suggest that much of the spatial variation in mtDNA haplotypes could be explained by distance between populations ([r.sup.2] = 0.86, excluding the Philippines and Western Australia) [ILLUSTRATION FOR FIGURE 4A OMITTED]. [[Phi].sub.ST] data incorporating divergence data plotted as a function of geographic distance also showed evidence of isolation by distance, but the relationship was not as tight ([r.sup.2] = 0.34, excluding the Philippines and Western Australia; [ILLUSTRATION FOR FIGURE 4B OMITTED]).
A three-level hierarchical analysis of [Phi]-statistics demonstrated that 28% of the total variance among populations is the result of differences between color morph groups ([[Phi].sub.CT] = 0.281, P = 0.068), 10% to differences among populations within color morph groups ([[Phi].sub.SC] = 0.145, P [less than] 0.0001) with the remainder due to differences among sites ([[Phi].sub.ST] = 0.386).
Pairwise analysis of percent nucleotide divergence among populations ranged from -0.02% (between GBR zones) to 1.64% (between Fiji and Thailand) (Table 6). The highest distances were between the two color morph groups. As was shown for the pairwise sequence [[Phi].sub.ST]-values, differences between any blue morph population and South Africa were similar to, or sometimes lower than, those between the same population and Thailand, despite the additional geographic distance separating the population from South Africa. UPGMA analysis of nucleotide divergence data among populations clustered South Africa and Thailand, separately from the Philippines, Western Australia, GBR, and Fiji [ILLUSTRATION FOR FIGURE 5 OMITTED]. This was not as evident in a neighbor-joining tree (not shown).
The consensus of 1000 trees of equal shortest length (all trees were 74 steps) showed that all 54 haplotypes were closely related with no distinctly differentiated groups and no obvious association with geographic locations (not shown). Figure 6 shows the haplotypes that occurred in each population as darkened circles in a network representation of one possible relationship among haplotypes. Populations within either color morph group share more than one haplotype, and haplotypes differ by a small number of steps. Populations in different color morph groups tend to occupy different (although partially overlapping) parts of the network.
Mitochondrial DNA and allozyme data demonstrated that Indian Ocean populations of L. laevigata, predominated by the salmon-orange color morph (Thailand and South Africa), are genetically differentiated from West Pacific populations, where a royal blue morph predominates. All populations examined were in HWE, notwithstanding the presence of more than one color morph at some locations (e.g., brown, slate-blue, dark purple, and salmon-orange in South Africa). This suggests that genetic differentiation of the two color morph groups (either salmon-orange or royal blue) is not related to color per se, but is instead due to regional differences between Thailand/South Africa and West Pacific/Western Australia.
Marked allozyme data differentiation between orange morph populations and blue morph populations is largely due to the effect of one locus (LT-1*), although all but two loci (GPI* and LT-2*) were significantly differentiated. The shift in allele frequencies observed at LT-1* may be indicative of geographically varying selection rather than restrictions to gene flow. However, the combined weight of several independent sets of circumstantial evidence suggests that, even if selection has influenced structure at LT-1* differences between the two groups are at least in part the result of (past) restrictions to gene flow between the two oceans, rather than the influence of selection alone. For example, while [F.sub.ST]-values were particularly high for LT-1*, four other allozyme loci (ENO*, SOD*, and to a lesser degree HK* and LP*) and mtDNA data show some evidence of differentiation among the same populations. Additionally, the genetic structure at LT-1* is congruent with the distribution of L. laevigata color morphs, both the geographic distribution of sibling species and color morphs in other marine taxa (e.g., Yamaguchi 1977b), and genetic data for other widespread species and two species groups of butterfly fish (Lavery et al. 1995, 1996; Lacson and Clark 1995; McMillan and Palumbi 1995; Benzie, unpubl. data 1997). It also seems unlikely that selective factors would maintain differences between the two oceans, but not change across the environmental gradient over more than 40 degrees of latitude in either ocean.
Only one of the RFLP haplotypes was found in both color morph groups, and the two most common haplotypes in the blue morph group (combined frequency in previous surveys of 0.63) were completely absent from the orange morph group, consistent with the idea of a genetic break. However a significant proportion of the spatial variation in haplotype distribution can be accounted for by geographic distance. This might suggest that there has been recent gene flow, especially among geographically close Indo-Malay populations. This is also supported by the lack of strong phylogeographic pattern in the PCR-RFLP haplotype genealogy, which suggests that either the data are inadequate (too few informative characters or too much homoplasy) or that the population structure observed using this technique is historically recent. However the correlation between geographic distance and [[Phi].sub.ST] is lower when divergence between haplotypes is taken into account.
Pairwise [[Phi].sub.ST] (seq) values between any blue morph population and South Africa, were similar to those between the same population and Thailand, despite the additional geographic distance separating the population from South Africa. [[Phi].sub.ST] (seq) values were higher among Pacific populations than [[Phi].sub.ST] (hap) values suggesting that there was some regionalization of haplotypes within the Pacific, but this was less evident in the Indian Ocean. South Africa and Thailand are significantly differentiated based on haplotype frequencies, but are not significantly differentiated when haplotype divergence data is incorporated, suggesting that the different haplotypes in the two populations are closely related. These combined data also suggest that although there was no strong phylogeographic pattern in the gene genealogy, haplotypes within the orange morph group may be more closely related to each other than to those in the blue morph group.
The genetic structure observed in L. laevigata, particularly at LT-1*, is congruent with genetic patterns observed between Pacific and Indian Ocean populations of another starfish species, [TABULAR DATA FOR TABLE 4 OMITTED] Acanthaster planci (Benzie unpubl. data 1997), the coconut crab, Birgus latro (Lavery et al. 1995, 1996); four species of damselfishes (Lacson and Clark 1995); and two species groups of butterfly fishes (McMillan and Palumbi 1995). Also a number of marine taxa with wide geographic ranges that span both oceans, exhibit different color morphs in each ocean, generally concordant with the distribution of orange and blue color morph groups of L. laevigata (e.g., A. planci; Benzie 1992). Concordant patterns of geographic segregation are also demonstrated by the distribution of many pairs of sibling and subspecies, which have species ranges that abut at or near the Indonesian archipelago (e.g., Yamaguchi 1977b, McMillan and Palumbi 1995).
The occurrence of a congruent biogeographic pattern shared by several different taxa, and several independent genetic markers within a species, is thought to be strong evidence of a vicariant event (Neigel and Avise 1993), consistent with the idea of marine zoogeographic boundary between Indian and Pacific provinces. During Pleistocene glaciations, sea levels fluctuated repeatedly, dropping as low as 200 m below, and occasionally rising above, present levels, changing the seaways between the Indian and Pacific Oceans. [TABULAR DATA FOR TABLE 5 OMITTED] Throughout much of the Pleistocene, Torres Strait was closed and the Indonesian throughflow may have been restricted (Galloway and Kemp 1981). Dispersal is unlikely to have occurred, either in the past or at present, via the southern coast of Australia since the water is probably too cold for tropical larvae to remain viable. Many authors have suggested that these climatological and/or geological processes may have prevented or restricted dispersal of many tropical marine species between the Indian and Pacific Oceans for thousands of years, with the Indonesian archipelago acting as a border between the two biogeographic regions (e.g., Fleminger 1985).
Western Australian populations of L. laevigata, however, although geographically located in the Indian Ocean, are genetically more similar to West Pacific populations and have the same royal blue morph as that observed in the West Pacific. It is not possible to determine whether this is the result of dispersal from the West Pacific to Western Australia subsequent to the end of the last ice age, or the presence of a barrier between a Western Australian population, which persisted throughout the Pleistocene, and the rest of the Indian Ocean. Mitochondrial DNA data demonstrate that recent gene flow is high between the Philippines and Western Australia, and is higher than expected given their geographic separation (Williams and Benzie 1997). The genetic homogeneity may be because the Western Australian population is the result of recent colonization or a large increase in recruits from the Indo-Malay region or the Philippines subsequent to the end of the last ice age.
Alternatively, the Philippines and Western Australian populations may be genetically similar because the many islands and reefs between them facilitate dispersal, so that the region may be thought of as essentially one long archipelago, and therefore genetic homogeneity is the result of continuing dispersal, as opposed to a recent increase in gene flow subsequent to the last ice age. Micropaleontological studies suggest that the northwest of Western Australia has changed very little in extremes of climate in the last 130,000 years (Wells and Wells 1994), and therefore, the reefal fauna associated with the Rowley Shoals and Scott Reef (several hundred kilometers off the present-day Australian coastline) may have persisted throughout Pleistocene glaciations (Marsh and Marshall 1983). In South America and southern Africa, cold upwellings act as barriers to dispersal [TABULAR DATA FOR TABLE 6 OMITTED] of warm water faunas (Morgan and Wells 1991). It is possible that Pleistocene upwellings off the northwest corner of Western Australia (Wells and Wells 1994; Wells et al. 1994) restricted larval dispersal between north Western Australia (Imperieuse Reef and Scott Reef) and the Indian Ocean, while a second region of upwelling below the Indonesian archipelago (Fleminger 1985; B. Opdyke, pers. comm. 1996) restricted direct transfer along the Indonesian coastline (bypassing Western Australia) of Pacific stocks into the Indian Ocean and (vice versa).
A third possibility is that the real situation lies between these two extremes. It may be that Western Australian populations persisted throughout the Pleistocene, but were partially isolated (by the closure of the Torres Strait and the upwellings mentioned above) from the West Pacific, and to a greater degree from the Indian Ocean (either by upwellings or reduced currents off the northwestern shelf of Western Australia). Subsequent to the rise in sea level at the end of the last ice age, dispersal may have been high from the Philippines region to Western Australia (and possibly to some degree in reverse) via the Indo-Malay. This scenario is supported by allozyme data that suggest that Western Australian populations are somewhat differentiated from all other West Pacific populations (Williams and Benzie 1996), while mtDNA suggests that recent gene flow has been high (Williams and Benzie 1997). On the other hand, coastal Western Australian populations (as opposed to the populations in this study, which are several hundred kilometers off the Australian coast) may not fit this scenario, as they are unlikely to have persisted throughout the Pleistocene since water temperatures are thought to have been about 4 [degrees] C lower than present along the coast (Webster and Streten 1978), making it too cold for most tropical species.
If the genetic structure observed in L. laevigata is the result of forces that affected larval dispersal, it would be predicted that a similar pattern should be observed in other widespread marine invertebrates within this region. There is limited opportunity to test this hypothesis, since there are few genetic studies that compare Indian and Pacific Ocean populations, and fewer that include both Western Australian and other Indian Ocean populations. The exception is a study of the starfish A. planci that, like L. laevigata, exhibits different color morphs in the Indian and Pacific Oceans (Benzie 1992) and demonstrates a marked genetic break between the two groups using allozyme data. Populations of A. planci from Thailand, South Africa, Cocos Keeling, and the Maldives are substantially genetically differentiated from Pacific populations, but Western Australian populations are more related to Pacific populations (Benzie, unpubl. data 1997).
A greater degree of differentiation was also observed among Indian Ocean populations of L. laevigata within the orange morph group, than among populations within the blue morph group. This is reflected both by the number of different color morphs present in Thailand and South Africa, and the greater degree of genetic structuring (e.g., [F.sub.ST] between Thailand and South Africa = 0.047, and [F.sub.ST] among 28 sites in West Pacific and Western Australia = 0.002; Williams and Benzie 1996). The lack of genetic uniformity and the nonuniform distribution of color morphs within the Indian Ocean may be due to restricted gene flow. There are fewer reefs and island archipelagos than in the West Pacific to facilitate "island hopping" across vast oceanic distances. Additionally, during the Pleistocene, sea level regressions partially enclosed the Andaman Sea by land (Potts 1983), isolating Thailand completely from the Pacific and partially from the Indian Ocean. At the same time, mesoscale gyres at the southeast corner of Africa may have led to the retention of larvae and prevented their dispersal to other populations (Pollock 1993).
Genetic differences among L. laevigata populations provide support for a degree of isolation of South Africa from other populations surveyed. Haplotype diversity, but not allozyme diversity, was also lower in South Africa than in any other population except Fiji, and may reflect reduced levels of gene flow. This is consistent with a survey of echinoderm fauna in that region. More than three quarters of all species (both tropical and temperate) are endemic (47%) or restricted to the Indian Ocean (37%) (Thandar 1989).
In summary, allozyme data suggest that there has been an historical break in gene flow between orange morph and blue morph populations. [[Phi].sub.ST] data incorporating divergence data between haplotypes is consistent with this scenario. On the other hand, spatial distribution of mtDNA RFLP haplotypes suggests that there may have been recent gene exchange between these populations, following an isolation by distance model. The congruence of genetic structure with the geographic distribution of two main color morph groups in a widespread species like L. laevigata, and the cooccurrence of this pattern in other taxa, is consistent with a biogeographic break between the Indian and Pacific Oceans. Further sampling within the Indo-Malay region and the Indian Ocean of L. laevigata and other widespread species is required to determine more accurately the nature of historical and contemporary influences that have affected population structure, and the extent to which the biogeographical pattern observed in L. laevigata is typical of broadly distributed Indo-Pacific species.
We thank the Natal Parks Board's dive team and R. Taylor for collecting samples from South Africa; the dive team from Phuket Marine Station and M. Phongsuwan for assistance in obtaining collections from Thailand; and C. Ablan for assistance in the field. We also thank R. Alford, E. Bailment, W. Burnett, D. Burrage, H. Lessios, S. Palumbi, and two anonymous reviewers for useful criticisms in review, and T. Simmonds for preparing Figure 2. The Australian Institute of Marine Science and James Cook University both provided funding for research, and the Australian Institute of Marine Science provided a scholarship to STW. This is contribution number 876 from the Australian Institute of Marine Science.
BENZIE, J. A. H. 1992. Review of the genetics, dispersal and recruitment of crown-of-thorns starfish (Acanthaster planci). Aust. J. Mar. Freshwater Res. 43:597-610.
BRIGGS, J. C. 1974. Marine zoogeography. McGraw Hill, New York.
CAVALLI-SFORZA, L. L., AND A. W. F. EDWARDS. 1967. Phylogenetic analysis: models and estimation procedures. Evolution 21:550-570.
CLARK, A. M., AND J. COURTMAN-STOCK. 1976. The Echinoderms of Southern Africa. British Museum (Natural History), London.
EXCOFFIER, L., P. E. SMOUSE, AND J. M. QUATTRO. 1992. Analysis of molecular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics 131:479-491.
FELSENSTEIN, J. 1993. PHYLIP: phylogeny inference package. Vers. 3.5c. Distributed by the author. Department of Genetics, Univ. of Washington, Seattle.
FLEMINGER, A. 1985. The Pleistocene equatorial barrier between the Indian and Pacific Oceans and a likely cause for Wallace's line. UNESCO Tech. Rep. Mar. Sci. 49:84-97.
Fox, D. L. 1947. Carotenoid and indolic biochromes of animals. Annu. Rev. Biochem. 16:443-470.
GALLOWAY, R. W., AND E. M. KEMP. 1981. Late Cainozoic environments in Australia. Pp. 51-80 in A. Keast, ed. Ecological biogeography of Australia. Dr. W. Junk Publishers, The Hague, The Netherlands.
HUDSON, R. R., D. D. Boos, AND N. L. KAPLAN. 1992. A statistical test for detecting geographic subdivision. Mol. Biol. Evol. 9: 138-151.
JOHNSON, M. S., B. CLARKE, AND J. MURRAY. 1988. Discrepancies in the estimation of gene flow in Partula. Genetics 120:233-238.
LACSON, J. M., AND S. CLARK. 1995. Genetic divergence of Maldivian and Micronesian demes of the damselfishes Stegastes nigricans, Chrysiptera biocellata, C. glauca and C. leucopoma (Pomacentridae). Mar. Biol. 121:585-590.
LAVERY, S., C. MORITZ, AND D. R. FIELDER. 1995. Changing patterns of population structure and gene flow at different spatial scales in Birgus latro (the coconut crab). Heredity 74:531-541.
-----. 1996. Indo-Pacific population structure and evolutionary history of the coconut crab Birgus latro. Molec. Ecol. 5:557-570.
MANTEL, N. 1967. The detection of disease clustering and a generalised regression approach. Cancer Res. 27:209-220.
MARSH, L. M. 1974. Shallow-water asterozoans of southeastern Polynesia. I. Asteroidea. Micronesica 10:65-104.
-----. 1977. Coral reef asteroids of Palau, Caroline Islands. Micronesica 13:251-281.
MARSH, L. M., AND J. I. MARSHALL. 1983. Some aspects of the zoogeography of northwestern Australian echinoderms (other than holothurians). Bull. Mar. Sci. 33:671-687.
MCELROY, D., P. MORAN, E. BERMINGHAM, I. KORNFIELD. 1992. REAP: an integrated environment for the manipulation and phylogenetic analysis of restriction data. J. Heredity 83:157-158.
McMILLAN, W. O., AND E. BERMINGHAM. 1996. The phylogeographic pattern of mitochondrial DNA variation in the Dall's porpoise Phocoenoides dalli. Molec. Ecol. 5:47-61.
MCMILLAN, W. O., AND S. R. PALUMBI. 1995. Concordant evolutionary patterns among Indo-West Pacific butterfly fishes. Proc. R. Soc. Lond. 260:229-236.
MORGAN, G. J., AND F. E. WELLS. 1991. Zoogeographic provinces of the Humboldt, Benguela and Leeuwin Current systems. Pp 59-69 in A. F. Pearce and D. I. Walker, eds. The Leeuwin current: an influence on the coastal climate and marine life of Western Australia. Royal Society of Western Australia, Perth.
NEI, M. 1978. Estimation of average heterozygosity and genetic distance from a small number of individuals. Genetics 89:583-590.
NEI, M., AND F. TAJIMA. 1981. DNA polymorphism detectable by restriction endonucleases. Genetics 97:145-163.
NEIGEL, J. E., AND J. C. AVISE. 1993. Application of a random walk model to geographic distributions of animal mitochondrial DNA variation. Genetics 135:1209-1220.
POLLOCK, D. E. 1993. Speciation in spiny lobsters - clues to climatically-induced changes in ocean circulation patterns. Bull. Mar. Sci. 53:937-944.
POTTS, D. C. 1983. Evolutionary disequilibrium among Indo-Pacific corals. Bull. Mar. Sci. 33:619-632.
RICE, W. R. 1989. Analysing tables of statistical tests. Evolution 43:223-225.
ROFF, D. A., AND P. BENTZEN. 1989. The statistical analysis of mitochondrial DNA polymorphisms: [[Chi].sup.2] and the problem of small samples. Mol. Biol. Evol. 6:539-545.
ROHLF, F. J. 1990. NTSYS-pc: numerical taxonomy and multivariate analysis system. Exeter Software, New York.
SAITOU, N., AND M. NEI. 1987. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4:406-425.
SCHNEIDER, S., J. M. KUEFFER, D. ROESSLI, AND L. EXCOFFIER. 1996. Arlequin: a software package for population genetics. Genetics and Biometry Lab, Dept. of Anthropology, Univ. of Geneva, Geneva, Switzerland.
SLOAN, N. A., A.M. CLARK, AND J. D. TAYLOR. 1979. The echinoderms of Aldabra and their habitats. Bull. Br. Mus. Nat. Hist. (Zool) 37:81-128.
SWOFFORD, D. L. 1990. PAUP: phylogenetic analysis using parsimony. Vets. 3.0. Computer programme distributed by the Illinois Natural History Survey, Champaign.
SWOFFORD, D. L., AND R. B. SELANDER. 1981. BIOSYS-1: a fortran program for the comprehensive analysis of electrophoretic data in population genetics and systematics. J. Heredity 72:281-283.
THANDAR, A. S. 1989. Zoogeography of the southern African echinoderm fauna. S. Afr. J. Zool. 24:311-318.
WAPLES, R. S. 1987. A multispecies approach to the analysis of gene flow in marine shorefishes. Evolution 41:385-400.
WEBSTER, P. J., AND N. A. STRETEN. 1978. Late Quaternary ice age climates of tropical Australasia: interpretations and reconstructions. Quat. Res. 10:279-309.
WEIR, B. S. 1990. Genetic data analysis. Sinauer, Sunderland, MA.
WEIR, B. S., AND C. C. COCKERHAM. 1984. Estimating F-statistics for the analysis of population structure. Evolution 38:1358-1370.
WELLS, P. E., AND G. M. WELLS. 1994. Large-scale reorganisation of ocean currents offshore Western Australia during the late Quaternary. Mar. Micropaleontol. 24:157-186.
WELLS, P. E., G. M. WELLS, J. CALI, AND A. CHIVAS. 1994. Response of deep-sea benthic foraminifera to late Quaternary climate changes, southeast Indian Ocean, offshore Western Australia. Mar. Micropaleontol. 23:185-229.
WILLIAMS, S. T. 1992. Methods for analysis of genetic variation in the blue starfish, Linckia laevigata, using allozyme electrophoresis. AIMS, report no. 6. Townsville, Australia.
WILLIAMS, S. T., AND J. A. H. BENZIE. 1993. Genetic consequences of long larval life in the starfish, Linckia laevigata, (Echinodermata: Asteroidea) on the Great Barrier Reef. Mar. Biol. 117:71-77.
-----. 1996. Genetic uniformity of widely separated populations of the coral reef starfish Linckia laevigata from the West Pacific and East Indian Oceans, revealed by allozyme electrophoresis. Mar. Biol. 126:99-108.
-----. 1997. Indo-West Pacific patterns of genetic differentiation in the high dispersal starfish Linckia laevigata. Mol. Ecol. 6:559-573.
WOODLAND, D. J. 1983. Zoogeography of the Siganidae (Pisces): an interpretation of distribution and richness patterns. Bull. Mar. Sci. 33:713-717.
WORKMAN, P. L., AND D. J. NISWANDER. 1970. Population studies on southwestern Indian tribes. II. Local genetic differentiation in the Papago. Am. J. Hum. Genet. 22:24-49.
WRIGHT, S. 1978. Evolution and the genetics of populations. Vol. 4. Variability in and among natural populations. Univ. of Chicago Press, Chicago.
YAMAGUCHI, M. 1977a. Population structure, spawning, and growth of the coral reef asteroid Linckia laevigata (Linnaeus). Pac. Sci. 31:13-30.
-----. 1977b. Larval behaviour and geographic distribution of coral reef asteroids in the Indo-West Pacific. Micronesica 13:283-296.
YATES, F. 1934. Contingency tables involving small numbers and the [[Chi].sup.2] test. J. R. Stat. Soc. Suppl. 1:217-235.
ZAGALSKY, P. F., F. HAXO, S. HERTZBERG, AND S. LIAAEN-JENSEN. 1989. Studies on a blue carotenoprotein, linckiacyanin, isolated from the starfish Linckia laevigata (Echinodermata: Asteroidea). Comp. Biochem. Physiol. 93B:339-353.
|Printer friendly Cite/link Email Feedback|
|Author:||Williams, S.T.; Benzie, J.A.H.|
|Date:||Feb 1, 1998|
|Previous Article:||Genetic variation and polymorphism in the inducible spines of a marine bryozoan.|
|Next Article:||Asymmetric introgression of mitochondrial DNA among European populations of blue mussels (Mytilus spp.).|