Evidence for Sympatric Speciation by Host Shift in ... - Semantic Scholar

Current Biology, Vol. 14, 1498–1504, August 24, 2004, 2004 Elsevier Ltd. All rights reserved.

DOI 10.1016/j .c ub . 20 04 . 08 .0 2 9

Evidence for Sympatric Speciation by Host Shift in the Sea Philip L. Munday,1,* Lynne van Herwerden,1 and Christine L. Dudgeon2 1 Centre for Coral Reef Biodiversity School of Marine Biology and Aquaculture James Cook University Townsville, QLD 4811 Australia 2 School of Life Sciences University of Queensland St. Lucia, QLD, 4072 Australia

Summary The genetic divergence and evolution of new species within the geographic range of a single population (sympatric speciation) contrasts with the well-established doctrine that speciation occurs when populations become geographically isolated (allopatric speciation). Although there is considerable theoretical support for sympatric speciation [1, 2], this mode of diversification remains controversial, at least in part because there are few well-supported examples [3]. We use a combination of molecular, ecological, and biogeographical data to build a case for sympatric speciation by host shift in a new species of coraldwelling fish (genus Gobiodon). We propose that competition for preferred coral habitats drives host shifts in Gobiodon and that the high diversity of corals provides the source of novel, unoccupied habitats. Disruptive selection in conjunction with strong host fidelity could promote rapid reproductive isolation and ultimately lead to species divergence. Our hypothesis is analogous to sympatric speciation by host shift in phytophagous insects [4, 5] except that we propose a primary role for intraspecific competition in the process of speciation. The fundamental similarity between these fishes and insects is a specialized and intimate relationship with their hosts that makes them ideal candidates for speciation by host shift. Results and Discussion Habitat specialization, competition for limited resources, and assortative mating according to habitats provide conditions conducive to sympatric speciation by host shift [1, 2, 5]. Species of Gobiodon satisfy all these characteristics. These small (⬍60 mm total length) fishes live among the branches of live corals and rely on their host coral for shelter, breeding sites, and in some cases a source of nutrition [6, 7]. All species of Gobiodon have a specialized pattern of host use and prefer to occupy just a few species of coral over all others available on the reef [6, 8]. Coral hosts are in limited supply and experimental manipulations have *Correspondence: [email protected]

demonstrated that there is intense competition for colonies of preferred coral hosts [9, 10]. Therefore, individuals that shift to an unexploited coral host could benefit from reduced competition and have relatively high fitness even if they were initially poorly adapted to the new habitat. Because of the great diversity of corals present on reefs, there are a range of unexploited coral species at any location that provide the opportunity for host shifts. Finally, there is potential for the development of reproductive isolation among individuals occupying different host species, because adults exhibit extreme fidelity to their host corals [11], and juveniles settle on the same host species as adults [6, 10]. Adults rarely move among coral colonies and most individuals spend their entire reproductive life in the same coral colony (P.L.M., unpublished data). The combination of these key elements in the ecology of coral-dwelling gobies provides the raw material for sympatric speciation by host shift. Here, we describe the evolution of a new species of coral-dwelling goby (Figure 1A) that is known from just one locality in southern Papua New Guinea. This undescribed species is currently referred to as Gobiodon species B [7], but for simplicity it will be referred to here as “new species”. Extensive sampling of coral gobies in Papua New Guinea and on the Great Barrier Reef, Australia by one of the authors (P.L.M.) indicates that this species has a very small geographic range centered around Bootless Bay in southern Papua New Guinea [7, 8]. At that locality, a viable population of several hundred individuals has been observed over a period of eight years (P.L.M., unpublished data). Social groups of up to ten juveniles and adults are found together exclusively inhabiting the coral, Acropora caroliniana. The new species shares a distinctive down-turned lower jaw with G. brochus (Figure 1B, [12]) and another undescribed species, Gobiodon species A (Figure 1C). The bright green and red body coloration of the new species is very similar to that of G. species A and two other species, G. histrio and G. erythrospilus (Figures 1D and 1E) but is unlike that of G. brochus. All these species occur together on reefs in southern Papua New Guinea and with the exception of the new species, have distributions that extend throughout much of the tropical western Pacific [7, 12–14]. Our molecular analysis shows that the new species is closely related to Gobiodon species A (Figure 2) and that the two species diverged between 0.18 and 0.72 million years ago (MYA). The estimated date of divergence is between 0.64 and 0.72 MYA when we use a mutation rate of 3.6% per million years calculated in previous studies where geological events were available to calibrate the rate of D-loop divergence in marine [15] and freshwater fishes [16]. However, considerable variation in D-loop mutation rates has been reported among fishes, and this estimate is likely to be highly conservative. The estimated date of divergence is much more recent, between 0.18 and 0.20 MYA, when we use a D-loop mutation rate of 12.9% per million years that

Sympatric Speciation in the Sea 1499

Figure 1. Species of Coral-Dwelling Goby (Genus Gobiodon) that Occur Together on Reefs in Bootless Bay, Papua New Guinea Gobiodon new species (A), G. brochus (B) G. species A (C), G. histrio (D) and G. erythrospilus (E).

has been reported in other fishes [17] for a comparable region of D-loop to that used here. The ecological and biogeographical data indicate that this new species might have evolved by a host shift in sympatry from the ancestral Gobiodon species A. First, for species in this clade, the ancestral pattern of host use is to inhabit a range of coral species that always includes Acropora tenuis (Figure 2); however, the new species uses just one novel species of coral, A. caroliniana (Figure 2). Speciation by host shift in sympatry requires a well-defined change in host use of the type observed here. In contrast, if speciation occurred in allopatry, the new goby species would be expected to retain the ancestral pattern of host use and at least inhabit A. tenuis. Second, A. caroliniana is not inhabited by any other species of coral-dwelling goby [7, 8] and, therefore, would have been vacant prior to the host shift and provided a refuge from both intra- and interspecific competition. Third, field observations and experiments have demonstrated that host corals are a limited resource for coral-dwelling gobies [6, 8–10] and, therefore, intraspecific competition for space provides a clear mechanism driving host shifts in these species. Finally, the known geographic range of the new species encompasses just a few hundred square km entirely within the much larger range of G. species A [7, 14], which is consistent with a sympatric mode of speciation. Comparisons of host use, geographical distributions, and relative time since divergence are also consistent with a host shift in sympatry. The new species has a very small geographic range [7], which corresponds with the relatively recent divergence from G. species A (Figure 2). The host coral used by the new species, A. caroliniana, is restricted to the Indo-Australian archipelago and is uncommon throughout most of its range, but is locally abundant in Papua New Guinea [18]. The relatively high abundance of A. caroliniana in southern Papua New Guinea would provide the opportunity for a host shift and sufficient habitat to support a viable population. The scarcity of A. caroliniana elsewhere might have then limited the potential for range expansion by the new species. G. species A and G. brochus have much larger geographic ranges than the new species, including much of the Indo-Australian archipelago and tropical western Pacific [7, 12, 14]. These larger geographic ranges correspond with a more ancient divergence (between 0.95 and 4.2 MYA) of G. brochus (Figure 2) and the broad range of their shared host, A. tenuis [18]. G. histrio and G. erythrospilus diverged even earlier and prefer a host coral, A. nasuta, that is widely distributed throughout the Indian and Pacific Oceans [18]. As expected, G. histrio and G. erythrospilus appear to have the largest geographic ranges of the species considered here, extending throughout the Indo-Pacific archipelago and into the Indian and central Pacific oceans [6, 7, 13, 14]. Sympatric speciation is favored among habitat specialists, such as coral gobies, where there are opportunities for some individuals to occupy novel hosts. However, for divergence to occur, the use of a novel host must be accompanied by a mechanism, such as assortative mating, that generates reproductive isolation [1, 2]. Because coral gobies usually breed in the same

Current Biology 1500

Figure 2. Phylogenetic Relationships, Haplotype Diversities, and Patterns of Habitat Use of Gobiodon Species The rooted phylogram of ML phylogenetic analysis, with Paragobiodon xanthosomus as outgroup, shows only interspecies nodes (see Figure 3 for complete phylogram). Also shown are the haplotype network and pattern of host use for each species. Collection location for specimens used in the haplotype digram are the same as for Table 1. n ⫽ number of specimens analyzed for haplotype diagram. Circles depict individual haplotypes, and circle sizes are relative to numbers of individuals sharing a haplotype. The smallest circles represent single individuals, and the largest circle represents eight individuals. Lines with crossbars connect haplotypes in each network to each other. The number of crossbars indicates numbers of base changes (substitutions) between haplotypes. Numbers of unique haplotypes and numbers of substitutions are used to calculate nucleotide and haplotype diversity indices in Table 1. The complete phylogram (Figure 3) indicated that there was no geographic differentiation within species clades; therefore sampling from several locations did not confound the phylogenetic relationships between spp. Patterns of host use exhibited by each species are indicated to the right of the species names. Graphs of host use show percent of different host species inhabited in southern Papua New Guinea [8].The abbreviations are as follows: D, Acropora digitifera; M, A. millepora; N, A. nasuta; V, A. valida; L, A. loripes; T, A. tenuis; C, A. caroliniana; and n, number of gobies sampled to estimate patterns of habitat use.

host species to which they settle from the plankton, any genetic variation that favors a shift in host use by some settling larvae could generate assortative mating. Where the same locus controls host prefence and perfomance in each host, assortative mating occurs as a by-product of specialization on different hosts [19]. This is the simplest genetic model of sympatric speciation, because a change in the expression of just one trait (habitat choice) by some parts of the population will automatically lead to assortative mating of adults in different habitats. Reproductive isolation can also occur where one locus affects host preference and another affects performance on each host. Under strong selection, the preference locus will rapidly become associated with the performance locus, and divergent selection on performance will drive the reproductive isolation of populations using different hosts [20]. We propose that frequency-dependent competition and disruptive selection have been the principal forces favoring a shift in host use and subsequent reproductive isolation of the new population. Intense competition for space is well documented in coral-dwelling gobies [8, 10]. Consequently, individuals that use an unoccupied host species could have relatively high fitness compared to an average individual using the traditional host. Divergent selection of phenotypes occupying each host

would then be driven by a genotype-environment interaction between performance and host use in coraldwelling gobies. Caley and Munday [21] demonstrated that goby species with narrow habitat preferences grow faster in the coral species they prefer to inhabit compared to coral species they inhabit less frequently. Furthermore, reproductive success of coral-dwelling gobies increases with body size of both males and females [22]. Therefore, any trait (physiological, behavioral, or morphological) that increases growth rate will be under strong selection. The genotype-environment interaction will generate disruptive selection for increased performance between the two populations of extreme specialists (the original population using A. tenuis and the new population using A. caroliniana). Disruptive selection could drive rapid adaptation to the new host, so that after a few generations, individuals swapping between host species would have low reproductive success compared to individuals remaining faithful to their host species. Assortative mating within habitats could be reinforced by host recognition mechanisms. Coral gobies lay their eggs on the host coral and the male tends the eggs until they hatch several days later. Imprinting of the embryos or newly hatched larvae on a novel host coral, and subsequent return of settling larvae to that species of coral,


Table 1. Genetic Architecture of Gobiodon Species Investigated in This Study Species

n

nh

h

%␲

G. G. G. G. G.

10 10 13 10 10

10 9 9 5 2

1 0.87 0.83 0.7 0.32

1.93 1.23 0.89 0.41 0.10

erythrospilus histrio brochus species A new species

⫾ ⫾ ⫾ ⫾ ⫾

1.13 0.76 0.56 0.31 0.11

Values shown are the number of samples (n), number of haplotypes (nh), haplotype diversity index (h ), and nucleotide diversity index (␲) shown as a percentage. Samples were from Bootless Bay, PNG: G. brochus (n ⫽ 8), G. new species (n ⫽ 10) and Lizard Island, northern Great Barrier Reef: G. histrio (n ⫽ 10), G. erythrospilus (n ⫽ 10), G. species A and G. brochus (n ⫽ 5).

would promote assortative mating. Imprinting of larvae on their host has been demonstrated in anemone fishes [23], another group of highly host-specific reef fishes, and would accelerate reproductive isolation following a host shift. Furthermore, some coral-dwelling gobies prey on the tissue of their host corals, and preference for the parental habitat could be passed on through chemicals accumulated by consumption of the host. Olfactory cues are known to influence host choices of larval fishes when they settle from the plankton [24], therefore, recognition of host chemistry at the time of settlement might play a role in establishing host fidelity and subsequent assortative mating in coral-dwelling gobies. Sympatric speciation by host shift appears to be the most parsimonious explanation for the evolution of the new species because it requires just two events to produce a new species: 1) a host shift by some settling larvae and 2) subsequent fidelity to the new host. Allopatric speciation by peripheral isolation is also a plausible explanation but is less parsimonious, because it would require four events to produce the new species: 1) the dispersal of a small founder population to an isolated location, 2) addition of A. caroliniana as a host in the new population, but not in the ancestral population, 3) loss of all ancestral hosts in the new population and, 4) subsequent range expansion of the ancestral species to produce the nested geographic ranges we see today. Allopatric speciation by vicariance is even less parsimonious because it would require five events to produce the new species: 1) the generation of a geographic barrier that separates some part of the ancestral population, 2) addition of A. caroliniana as a host in one population, but not the other, 3) loss of all ancestral hosts in that same population, 4) relaxation of the geographic barrier, and 5) range contraction in one species and range expansion in the other species to produce the nested ranges we see today. While both allopatric modes of speciation are plausible, they are less parsimonious than a host shift in sympatry in this instance. Different mechanims of speciation are predicted to leave different signatures in the patterns of genetic diversity of sister taxa [25]. The new species considered here exhibits very low genetic diversity (Figure 2, Table 1); much lower even than its closest relative, G. species A, which also inhabits just one coral host, but has not undergone a host shift. Low genetic diversity indicates

that a strong genetic bottleneck has occurred, as might be expected following divergence of a small founder population. Speciation by peripheral isolation of a small population could explain the low genetic diversity observed in the new species [25]; however, this mode of speciation does not explain why the isolated population made a complete host shift and did not retain the ancestral pattern of host use. Furthermore, the lack of genetic differentiation between individuals of G. brochus and G. histrio (Figure 3) collected from locations separated by over 1000 km (Bootless Bay and Kimbe Bay in Papua New Guinea and Lizard Island on the Great Barrier Reef) demonstrates that larval dispersal of these gobies can maintain gene flow over large spatial scales. The potential for widespread larval dispersal would work against the establishment of a small, genetically isolated (but not geographically isolated) population in close proximity to the parent population. A host shift in sympatry could also explain the low genetic diversity observed in the new species, provided that only a small number of individuals made the host shift and subsequent gene flow between populations was negligible. As described above, reproductive isolation among the populations could evolve rapidly as a result of strong disruptive selection for increased growth rate among individuals inhabiting different host species. Reproductive isolation would be even more rapid if it is reinforced by host imprinting or other host recognition mechanisms. The genetic basis for the genotype-environment interaction that favors disruptive selection may reside in either the nuclear or mitochondrial genome. The non-recombining mt genome (which was used as a marker in this study) is a prime target for disruptive selection, as it is responsible for 90% of the energetic metabolism of the organism and as such is under strong functional selection, with changes in the mtDNA having substantial impacts on fitness [26]. Low genetic diversity following speciation in sympatry could also occur if the genetic variation that favored the host shift was pleiotropically linked to a mutation in the mt genome that influenced performance in the new host coral. If such an adaptive mutation occurs on the mtDNA it may sweep to fixation carrying all linked nucleotide variants along with it [27]. This would eliminate polymorphism, evidenced as a strong genetic bottleneck, and maintain only those individuals in the population having the favored mutation (and other selectively neutral mutations that may be present on the favored mt genome [27], including the neutral section of mt DNA sampled here). Finally, although mitochondrial genomes are almost exclusively maternally inherited in most eukaryotes, this may not be the case for many fishes that are sequential hermaphrodites. In these species all individuals will contribute mt DNA to offspring at some stage of their reproductive life. This effect is likely to be particularly pronounced in coral gobies because they can swap repeatedly between male and female function [28]. Although not formally demonstrated, hermaphroditism may be another means by which maternal preferences (e.g., for laying eggs on specific hosts) could spread rapidly throughout the population, if the preference was linked to mitochondrial type.


Figure 3. Phylogenetic Tree of All Samples Used in the Phylogenetic Analysis of Gobiodon Species Outgroup rooted phylogram of best tree (⫺lnL ⫽ 2791.856) from ML analysis of Gobiodon species with Paragobiodon xanthosomus as outgroup. Numbers above branches are ML bootstrap support values, below branches are majority rule support values for 895,400 tree topologies obtained by Bayesian inference. Sample codes indicate location sampled for each of the species as follows: G. histrio from Lizard Island, GBR ⫽ H; from Kimbe Bay, northern PNG ⫽ HK; from Bootless Bay, southern PNG ⫽ HL; G. erythrospilus were all from Lizard Island, GBR ⫽ E; G. brochus were from Lizard Island ⫽ BL or from Bootless Bay southern PNG ⫽ BG; G. sp. A were all from Lizard Island ⫽ GspA; G. new sp. were all from Bootless Bay, southern PNG ⫽ Gnew.


Although we advocate a role for sympatric speciation in the diversification of coral-dwelling gobies, we recognize that allopatric speciation has also been important in the evolutionary history of these fishes. For example, our analysis shows that G. histrio and G. erythrospilus are closely related but genetically distinct (Figures 2 and 3). Although these two species have identical patterns of host coral use, they rarely co-habit the same coral colony and hybrids have not been observed [7, 10]. The identical patterns of host use for these species indicates that divergence has occurred without a shift in host use. The most parsimonious explanation is divergence in allopatry followed by range expansions and secondary contact. Similarly, G. brochus appears to have diverged from the most recent common ancestor with G. erythrospilus without a host shift and current-day differences in the relative frequency that different hosts are occupied is largely due to the effects of interspecific competition with G. histrio and G. erythrospilus [9]. Sympatric speciation has been largely discounted in marine fishes because most species have a planktonic larval phase with the potential to disperse widely and generate genetic homogeneity over large spatial scales. However, there is increasing evidence that larvae do not always disperse long distances and some return to their natal reefs [29]. Furthermore, habitat selection at settlement and assortative mating can produce reproductive isolation at very fine spatial scales, even in species with the capacity to disperse large distances. Clearly, a pelagic larval phase need not preclude the formation of fine scale genetic structure because behavior can override the potential for genetic mixing [30]. Competition has long been regarded as a potent force for ecological diversification [3] and is an explicit parameter in some mathematical models of sympatric speciation [1]. Wilson and Turelli [31] showed that a stable polymorphism can evolve following the invasion of an empty niche by a species experiencing strong intraspecific competition. Furthermore, their model predicts that heterozygotes will have lowest fitness, and therefore, that the incipient species could become reproductively isolated as a result of disruptive selection. Recent theoretical advances [32, 33] suggest that frequency-dependent competition for a unimodally distributed resource is sufficient to generate disruptive selection and assortative mating. In these models, disruptive selection and assortative mating emerge as a by-product of competition and are not untested assumptions of the model. In addition to these new mathematical models, experiments have confirmed that frequency-dependent competition can generate disruptive selection on important phenotypic traits [34], including those involved in metabolism and growth [35]. The consensus emerging from these theoretical and experimental approaches is that intraspecific competition can play a key role in the divergence of new species in sympatry. Two of the most convincing empirical examples of sympatric speciation are the monophyletic origin of cichlids in crater lakes in Cameroon [36] and the evolution of new host races in the apple maggot fly [37]. In both cases assortative mating of individuals using spatially isolated resources appears to underlie population divergence. Whether intraspecific competition was

the process responsible for niche partitioning has not been tested. Ours is one of the first empirical studies to directly link resource competition and a experimentally tested genotype-environment interaction to a proposed case of sympatric speciation. Sexual selection has also been linked to the maintenance of assortative mating in species that appear to have diverged in sympatry [38, 39]. However, this is unlikely to be the case for the new species of coral-dwelling goby described here, because there is no sexual dimorphism in this clade of fishes [28]. In conclusion, multiple lines of evidence presented here point to the generation of a new species of coraldwelling goby as a result of a host shift in sympatry. Although allopatric speciation by peripheral isolation cannot be discounted, this mechanism does not explain the current day patterns of host use in the new species. Our results support the emphasis given to competition as the process underlying divergent selection and assortative mating in recent models of sympatric speciation. Experimental Procedures We amplified DNA (extracted from caudal fin) using universal fish mitochondrial D-loop primers L15995 5⬘-AACTCACCCCTAGCTCC CAAAG-3⬘ and H16498 5⬘-CCTGAAGTAGGAACCAGATG-3⬘, from all species, except G. brochus, at an annealing temperature of 51⬚C, in the presence of 2.5mM MgCl2. G. brochus samples were amplified using the Universal forward primer (L15995) and a goby specific reverse primer, GdloopR3: 5⬘-CGAAGCTTATTGCTTGCTTG-3⬘ as described above. PCR products were directly sequenced in both directions on an ABI 377 sequencer, using the amplification primers. Aligned sequences (384 bp) were analyzed in PAUP* [40] following identification of the optimal substitution model for the data using likelihood approaches implemented in Modeltest [41], with the following specifications: ML analysis, substitution model GTR ⫹ G, with G ⫽ 0.898, nucleotide frequencies: A ⫽ 0.382, C ⫽ 0.182, G ⫽ 0.146, T ⫽ 0.291, and 100 bootstrap replicates. Trees were rooted with Paragobiodon xanthosomus as an outgroup. The ML tree topology was confirmed with Bayesian inference using the program Mr Bayes [42]. A maximum likelihood model was specified, employing 6 substitution types, with base frequencies estimated from the data, and rate variation across sites modeled using a ␥ distribution, G ⫽ 0.898. The Markov chain Monte Carlo search was run with 4 chains for 1,000,000 generations, with trees sampled every 100 generations. The first 100,000 trees were discarded as “burnin” and a majority rule consensus tree was obtained from the remaining 900,000 trees. The ages of divergence were calculated directly by dividing the % genetic divergence between lineages (obtained from the ML branch length/384 bp of sequence ⫻ 100) by the 3.6% or 12.9% per million years divergence rates for fish D-loop sequences. Nucleotide and haplotype diversity indices (␲ and h respectively) of all Gobiodon species were calculated, because these indices are expected to be less diverse in recently founded populations than in centers of origin [43]. Nucleotide diversity indices were calculated from sequences of all individuals sampled from PNG using the program Arlequin [44]. Haplotype diversity indices were calculated as per Nei [45] using the formula h ⫽ n (1 ⫺ ⌺ xi2)/(n ⫺ 1). Acknowledgments We thank Max Benjamin, Dik Knight, Alex Anderson, Walindi Diving, Loloata Island Resort, Mahonia Na Dari Research and Conservation Centre, and Motupore Island Research Station for excellent logistic support and assistance. We also thank J.H. Choat, S. Connolly, J.B. Crespi, J. Endler, G. Jones, S. Planes, D. Schluter, S. Swearer, W. Rice, R. Warner and three anonymous reviewers for helpful discus-


sions and/or comments. Not all agreed with our conclusions, but all provided insightful and helpful critique. This research was funded by grants from the Australian Research Council and James Cook University to P.L.M. and L.VH. Received: February 22, 2004 Revised: June 24, 2004 Accepted: June 24, 2004 Published: August 24, 2004

22.

23.

24. References 1. Turelli, M., Barton, N.H., and Coyne, J.A. (2001). Theory and speciation. Trends Ecol. Evol. 16, 330–343. 2. Via, S. (2001). Sympatric speciation in animals: the ugly duckling grows up. Trends Ecol. Evol. 16, 381–390. 3. Schluter, D. (2000). The Ecology of Adaptive Radiations, (Oxford: Oxford University Press). 4. Bush, G.L. (1969). Sympatric host race formation and speciation in frugivorous flies of the genus Rhagoletis (Diptera, Tephritidae). Evolution Int. J. Org. Evolution. 23, 237–251. 5. Tauber, C., and Tauber, M. (1989). Sympatric speciation in insects: perception and perspective. In Speciation and its Consequences, D. Otte and J.E. Endler, eds. (Sunderland: Sinauer Associates), pp 307–344. 6. Munday, P.L., Jones, G.P., and Caley, M.J. (1997). Habitat specialisation and the distribution and abundance of coral-dwelling gobies. Mar. Ecol. Prog. Ser. 152, 227–239. 7. Munday, P.L., Harold, A.S., and Winterbottom, R. (1999). Guide to coral-dwelling gobies, genus Gobiodon (Gobiidae) from Papua New Guinea and the Great Barrier Reef. Rev. fr. Aquariol. 26, 49–54. 8. Munday, P.L. (2002). Does habitat availability determine the geographical-scale abundance of coral-dwelling fishes? Coral Reefs 21, 105–116. 9. Munday, P.L., Jones, G.P., and Caley, M.J. (2001). Interspecific competition and coexistence in a guild of coral-dwelling gobies. Ecology 82, 2177–2189. 10. Munday, P.L. (2004). Competition among coral-dwelling fishes: the lottery model revisited. Ecology 85, 623–628. 11. Nilson, G.E., Hobbs, J.P.A., Munday, P.L., and Nilsson, S. (2004). Extreme habitat fidelity through hypoxia tolerance in a coraldwelling goby. J. Exp. Biol. 207, 33–39. 12. Harold, A.S., and Winterbottom, R. (1999). Gobiodon brochus: A new species of Gobiid fish (Teleostei: Gobioidei) from the western South Pacific, with a description of its unique jaw morphology. Copeia 1999, 49–57. 13. Sawada, Y., and Arai, R. (1973). Three species of coral gobies (genus Gobiodon) from the Ryuku Islands. Japan. Bull. Natn. Sci. Mus. Tokyo 16, 585–603. 14. Suzuki, T., Aizawa, M., and Senou, H. (1995). A preliminary review of three species of the Gobiodon rivulatus complex from Japan. I.O.P. Diving News 6, 2–7. 15. Donaldson, K.A., and Wilson, R.R., Jr. (1999). Amphi-Panamic geminates of Snook (Percoidei: Centropomidae) provide a calibration of the divergence rate in the mitochondrial DNA Control region of fishes. Mol. Phylogenet. Evol. 13, 208–213. 16. Waters, J.M., Craw, D., Youngson, J.H., and Wallis, G.P. (2001). Genes meet geology: Fish phylogeographic pattern reflects ancient, rather than modern, drainage connections. Evolution Int. J. Org. Evolution 55, 1844–1851. 17. Alvarado, B.J.R., Baker, A.J., and Mejuto, J. (1995). Mitochondrial DNA control region sequences indicate extensive mixing of swordfish (Xiphias gladius) population in the Atlantic Ocean. Can. J. Fish. Aquat. Sci. 52, 1720–1732. 18. Wallace, C. (1999). Staghorn Corals of the World. (Collingwood: CSIRO). 19. Rice, W.R. (1987). Speciation via habitat specialization: the evolution of reproductive isolation as a correlated character. Evol. Ecol. 1, 301–314. 20. Fry, J.D. (2003). Multilocus models of sympatric speciation: Bush versus Rice versus Felsenstein. Evolution Int. J. Org. Evolution 57, 1735–1746. 21. Caley, M.J., and Munday, P.L. (2003). Growth trades off with

25. 26. 27. 28.

29.

30.

31.

32. 33.

34.

35.

36.

37.

38.

39.

40. 41. 42. 43.

44.

45.

habitat specialization. Proc. Royal Soc. London B (Suppl.) 270, S175–S177. Kuwamura, T., Yogo, Y., and Nakashima, Y. (1993). Sizeassortive monogamy and parental egg care in a coral goby Paragobiodon echinocephalus. Ethology 95, 65–75. Arvedlund, M., and Nielson, L.E. (1996). Do the anemonefish Amphiprion ocellaris (Pisces: Pomacentridae) imprint themselves to their host sea anemone Heteractis magnifica. Ethology 102, 197–211. Elliott, J.K., Elliott, J.M., and Mariscal, R.N. (1995). Host selection, location, and association behaviours of anemonefishes in field settlement experiments. Mar. Biol. 122, 377–389. Mayr, E. (1963). Animal species and evolution. (Cambridge; Harvard University Press). Ballard, J.W.O., and Whitlock, M.C. (2004). The incomplete natural history of mitochondria. Mol. Ecol. 13, 729–744. Rand, D.M. (2001). The units of selection of mitochondrial DNA. Annu. Rev. Ecol. Syst. 32, 415–448. Munday, P.L., Caley, M.J., and Jones, G.P. (1998). Bi-directional sex change in a coral-dwelling goby. Behav. Ecol. Sociobiol. 43, 371–377. Swearer, S.E., Shima, J.S., Hellberg, M.E., Thorrold, S.R., Jones, G.P., Robertson, D.R., Morgan, S.G., Selkoe, K.A., Ruiz, G.M., and Warner, R.R. (2002). Evidence of self-recruitment in demersal marine populations. Bull. Mar. Sci. 70, 251–271. Taylor, M.S., and Hellberg, M.E. (2002). Genetic evidence for local retention of pelagic larvae in a Caribbean reef fish. Science 299, 107–109. Wilson, D.D., and Turelli, M. (1986). Stable underdominance and the evolutionary invasion of empty niches. Am. Nat. 127, 835–850. Diekmann, U., and Doebeli, M. (1999). On the origin of species by sympatric speciation. Nature 400, 354–357. Doebeli, M., and Diekmann, U. (2000). Evolutionary branching and sympatric speciation caused by different types of ecological interactions. Am. Nat. 156, S77–S101. Bolnick, D.L. (2004). Can intraspecific competition drive disruptive selection? An experimental test in natural populations of sticklebacks. Am. Nat. 58, 608–618. Fresen, M.L., Saxer, G., Travisana, M., and Doebeli, M. (2003). Experimental evidence for sympatric ecological diversification due to frequency-dependent competition in Escherichia coli. Evolution Int. J. Org. Evolution 58, 245–260. Schliewen, U., Rassmann, K., Markmann, M., Markert, J., Kocher, T., and Tautz, D. (2001). Genetic ecological divergence of a monophyletic cichlid species pair under fully sympatric conditions in Lake Ejagham, Cameroon. Mol. Ecol. 10, 1471– 1488. Filchak, K.E., Roethele, J.B., and Feder, J.L. (2000). Natural selection and sympatric divergence in the apple maggot. Nature 407, 739–742. Seehausen, O. Explosive speciation rates and unusual species richness in haplochromine cichlid fishes: effects of sexual selection. Adv. Ecol. Res. 31, 237–274. Allender, C.J., Seehausen, O., Knight, M.E., Turner, G.F., and Maclean, N. (2003). Divergent selection during speciation of Lake Malawi cichlid fishes inferred from parallel radiations in nuptial coloration. Proc. Natl. Acad. Sci. USA 100, 14074–14079. Swofford, D.L. (1989). PAUP* Phylogenetic Methods using Parsimony and other Methods. (Sunderland: Sinauer Associates). Posada, D., and Crandall, K.A. (1998). Modeltest: testing the model of DNA substitution. Bioinformatics 14, 817–818. Huelsenbeck, J.P., and Ronquist, F. (2001). MRBAYES: Bayesian inference of phylogeny. Bioinformatics 17, 754–755. Grant, W.S., and Bowen, B.W. (1998). Shallow population histories in deep evolutionary lineages of marine fishes: insights from sardines and anchioves and lessons for conservation. J. Hered. 89, 415–426. Schneider. S., Roessili, D., and Excoffier, L. (2000). Arlequin. Vers 2.000: a software for genetic data analysis. (Geneva, Genetics and Biometry Laboratory, University of Geneva). Nei, M. (1987). Molecular evolutionary genetics. (New York, Columbia University Press).