sign in sign up
<<
Home , Barbulifer, Callogobius, Chriolepis, Cladogenesis, Cleaner fish, Damselfish

Evolution, 57(7), 2003, pp. 1584±1598

RAPID AND ECOLOGICAL DIVERGENCE IN THE AMERICAN SEVEN-SPINED (, GOBIOSOMATINI) INFERRED FROM A MOLECULAR PHYLOGENY

LUKAS RUÈ BER,1,2 JAMES L. VAN TASSELL,3,4 AND RAFAEL ZARDOYA1,5 1Departamento de Biodiversidad y BiologõÂa Evolutiva, Museo Nacional de Ciencias Naturales, Jose GutieÂrrez Abascal 2, 28006 Madrid, Spain 2E-mail: [email protected] 3Department of , Hofstra University, Hempstead, New York 11549 4E-mail: [email protected] 5E-mail: [email protected]

Abstract. The American seven-spined gobies (Gobiidae, Gobiosomatini) are highly diverse both in morphology and ecology with many endemics in the region. We have reconstructed a molecular phylogeny of 54 Gobio- somatini taxa (65 individuals) based on a 1646-bp region that includes the mitochondrial 12S rRNA, tRNA-Val, and 16S rRNA genes. Our results support the of the seven-spined gobies and are in agreement with the existence of two major groups within the tribe, the group and the group. However, they reject the monophyly of some of the Gobiosomatini genera. We use the molecular phylogeny to study the dynamics of speciation in the Gobiosomatini by testing for departures from the constant speciation rate model. We observe a burst of speciation in the early evolutionary history of the group and a subsequent slowdown. Our results show a split among into coastal-estuarian, deep , and tropical reef . Major shifts account for the early signi®cant accel- eration in splitting and speciation rate and the initial divergence of the main Gobiosomatini clades. We found that subsequent diversi®cation is triggered by behavior and niche specializations at least in the reef-associated clades. Overall, our results con®rm that the diversity of Gobiosomatini has arisen during episodes of , and emphasize the importance of ecology in marine speciation.

Key words. Adaptive radiation, Caribbean, diversi®cation rates, marine ®sh, mitochondrial DNA.

Received November 8, 2002. Accepted January 24, 2003.

Adaptive radiation, the evolution of ecological and phe- umenting ®sh adaptive radiations in lacustrine environments notypic diversity within a rapidly multiplying lineage has (e.g. Fryer and Iles 1972; Echelle and Korn®eld 1984; Schlu- engaged biologists for over a century (Osborne 1902; Simp- ter 1996), only a few have focused on marine ®shes (Johns son 1953; Schluter 2000). Adaptive radiations are excellent and Avise 1998; Eastman and McCune 2000). systems in which to study the factors that in¯uence speciation The American seven-spined gobies (Gobiidae: Gobioso- processes. The astonishing biological diversity of these sys- matini; proper spelling of the tribe from Gobiosomini to Go- tems may follow the evolution of key innovations in mor- biosomatini corrected by Smith and Baldwin 1999) represent phological, physiological, or behavioral traits (Stanley 1979; a possible case of a marine adaptive radiation (Hoese and Sanderson and Donoghue 1994). Alternatively, the coloni- Larson 1985). The Gobiosomatini display a great diversity zation of novel habitats is thought to provide new ecological in color, ecology, and behavior; include over 130 in opportunity or release of competition, and hence enable lin- 24 genera; and represent some 40% of the total eages to colonize new adaptive zones (Simpson 1953; Schlu- gobioid genera and perhaps as many as 50% of the species ter 2000). Molecular phylogenies provide a robust framework (Birdsong and Robins 1995). The Gobiosomatini are endemic to study the temporal pattern of speciation and diversi®cation to the New World, and occur in the western Atlantic from in lineages that have undergone adaptive radiation (Johns and Massachusetts to Argentina, and in the eastern Paci®c from Avise 1998; Lovette and Bermingham 1999). The branching the northern Gulf of California to Chile, although most of pattern of a can be used to detect changes the species are found in the Caribbean region. Some genera in speciation rate through time, and to detect asymmetries are either restricted to the Atlantic or to the Paci®c, but most between contemporaneous clades in their number of descen- have representatives in both . Members of the Gobio- dant species (Nee et al. 1994a; Nee et al. 1996; Sanderson somatini display a remarkable diversity in morphology and and Donoghue 1996; Pybus and Harvey 2000). Such infor- are highly selective with regard to habitat (BoÈhlke and Robins mation is needed to distinguish speciation bursts from sto- 1968). Some are found in shallow fresh waters, brackish wa- chastic background rates and to identify historical factors ters, and with mud, shell, gravel, or -covered underlying the emergence of ecological and phenotypic di- substrate. Others are found in or rocky reefs and shelf vergence within a lineage. slopes at depths exceeding 500 meters (Appendix). With the accumulation of phylogenetic data it has become Differences in dentitions indicate specialized diets, al- evident that a considerable amount of biological diversity has though little information for most species (especially the ben- arisen during episodes of adaptive radiation (Schluter 2000). thic genera) is available (BoÈhlke and Robins 1968). Some Approximately 13,000 marine are known, which ac- members of the subgenus are known to remove count for about 61% of the total diversity (Nelson ectoparasites from other ®shes (cleaning behavior) whereas 1994). Nevertheless, in contrast to the wealth of studies doc- others are associated with sponges. One species (Nes longus) 1584 ᭧ 2003 The Society for the Study of Evolution. All rights reserved. EVOLUTION OF AMERICAN SEVEN-SPINED GOBIES 1585

FIG. 1. Map of the Caribbean region with the sampling localities for the Gobiosomatini taxa. Proximate sampling locations have been gathered into a single number (see Appendix for more details). commensally with snapping whereas a close were deposited in the American Museum of Natural History, association with sea urchins is found in Ginsburgellus nov- New York (Appendix; for identi®cation numbers, GenBank emlineatus and multifasciatum (BoÈhlke and Rob- accession numbers, and museum voucher numbers, see the ins 1968). Electronic Appendix, available from the Evolution Editorial Phylogenetic hypotheses of the tribe based on morpholog- Of®ce at [email protected]). Here, we follow BoÈhlke and ical characters are rather inconclusive, perhaps because dif- Robins's (1968) classi®cation of the Gobiosoma into ferent authors have used distinct sets of characters and taxa, the subgenera Austrogobius, Elacatinus, Gobiosoma, Gar- and due to the potential of high levels of homoplasy resulting mannia, and Tigrigobius and use their criteria to assign sub- from body size reduction and/or to extreme hab- generic status to eastern Paci®c species not included in their itats (e.g. soft mud bottom, sponge-dwelling; BoÈhlke and study. Robins 1969; Hoese 1971; Van Tassell 1998). The aim of Total genomic DNA was isolated from white muscle tissue this study was to establish a mtDNA-based molecular phy- or ®n clips by proteinase K/SDS digestion, phenol-chloro- logeny of the American seven-spined gobies to study the form extraction, and ethanol precipitation (Kocher et al. dynamics of speciation in this ecomorphologically and be- 1989). Partial mitochondrial 12S and 16S rRNA genes were haviorally highly diverse group of marine ®shes. A robust polymerase chain reaction (PCR) ampli®ed using the uni- phylogenetic framework will permit us (1) to test for depar- versal primers L1091 and H1478 (Kocher et al. 1989), and tures from the constant speciation rate model and to identify 16Sar-L and 16Sbr-H (Palumbi et al. 1991), respectively. A peak periods of , (2) to detect asymmetries be- PCR product of approximately 1300 bp that connects the tween contemporaneous clades in their number of extant spe- above mentioned gene fragments (including the 3Ј end of the cies, and (3) to map habitat shifts onto the phylogeny and 12S rRNA, the complete tRNA-Val, and the 5Ј end of the test their association with divergence or diversi®cation events 16rRNA genes) was ampli®ed with the primers ®sh-12F1 (5Ј- during the evolutionary history of the Gobiosomatini. TGA AGG AGG ATT TAG CAG TAA G-3Ј) and ®sh-16SR1 (5Ј-AAG TGA TTG CGC TAC CTT CGC AC-3Ј). These MATERIAL AND METHODS primers also work in a variety of other acanthomorph ®sh (L. RuÈber, pers. obs.). The primer ®sh-12SF2 (5Ј-TCT CTG Samples and Sequences TGG CAA AAG AGT-3Ј) was used as an internal sequencing To assess the molecular phylogeny of the American seven- primer. Polymerase chain reaction ampli®cations were con- spined gobies, 65 individuals (plus four outgroups) were col- ducted in 25 ␮l reactions containing 75 mM tris-HCl (pH lected from 17 localities (Fig. 1 and Appendix). Whole ®sh 9.0), 2 mM MgCl2, 0.4 mM of each dNTP, 0.4 ␮M of each were preserved in 70±100% ethanol, and voucher specimens primer, template DNA (10-100 ng), and Taq DNA polymer- 1586 LUKAS RUÈ BER ET AL. ase (one unit, Biotools, Madrid, Spain), using the following (Sanderson 1997) as implemented in TreeEdit version 1.0 program: one cycle of 2 min at 94ЊC, 35 cycles of 60 sec at (Rambaut and Charleston 2001). To roughly estimate diver- 94ЊC, 60 sec at 48±54ЊC, 90 sec at 72ЊC, and ®nally, one gence times between major clades, we applied a molecular cycle of 5 min at 72ЊC. clock using the Tv rate of the 12S and 16S rRNA genes of After PCR puri®cation using an ethanol/sodium acetate 0.14% per million years (Ritchie et al. 1996). precipitation, samples were cycle-sequenced with the ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction Tests of Diversi®cation Rate Kit (V3.0) in 10 ␮l reactions, and following manufacturer's instructions (Applied Biosystems, Foster City, CA), with We investigated whether rates of cladogenesis have 3.25 pmol of primer, 3 ␮l of Terminator Ready Reaction Mix changed through time in the seven-spined gobies using the and 5% dimethyl sulfoxide. The cycling pro®le for the se- constant-rate (CR) test (Pybus and Harvey 2000; Pybus and quencing reaction consisted of 25 cycles of 10 sec at 96ЊC, Rambaut 2002). We also applied the Monte Carlo constant- 5 sec at 50ЊC, and 4 min at 60ЊC. Cycle-sequencing products rates (MCCR) test (Pybus and Harvey 2000) that takes in- were puri®ed using MultiScreen plates (Millipore, Billerica, complete sampling into account using the program MA), and were analyzed on an ABI Prism 3700 DNA An- MCCRTest (Pybus 2000). The CR and MCCR tests are con- alyzer (Applied Biosystems). Sequences have been deposited ducted under the assumption that diversi®cation occurs equal- in GenBank (see Electronic Appendix). ly among lineages. Therefore, we used three approaches to detect nonstochastic differences in species diversity among lineages. First, we performed three-taxon tests as imple- Sequence Alignment and Phylogenetic Analysis mented in LRDIVERSE version 0.8 (Sanderson and Dono- Orthologous DNA sequences were aligned with Clustal X ghue 1996; Sanderson and Wojciechowski 1996). Second, (Thompson et al. 1997) using the default settings, and align- the B1 index was used as a measure of tree imbalance (Kirk- ments were further optimized by eye. Equivocal alignment patrik and Slatkin 1993). Third, the relative cladogenesis sta- positions in gap-rich areas were excluded from the analyses tistic (Nee et al. 1994b, 1995, 1996) as implemented in End- because of uncertainty in assignment. Modeltest Epi version 1.0 (Rambaut et al. 1997) was used to identify version 3.06 (Posada and Crandall 1998) was used to deter- lineages with higher than expected rates of cladogenesis. The mine the substitution model that best ®t the dataset. The relative cladogenesis statistic uses a simple birth- model hierarchical likelihood-ratio tests (LRT) implemented in Mo- to calculate the probability that any ancestral lineage has deltest selected the GTR ϩ I ϩ⌫model (proportion of in- more than a extant tips given k ancestral lineages and n extant variable sites ϭ 0.33; ␣ϭ0.54, empirical base frequencies: tips in total. End-Epi version 1.0 was also used for generating A ϭ 0.33; C ϭ 0.26; G ϭ 0.21; T ϭ 0.20; substitution rates: semilogarithmic lineage through time (LTT) plots. All these A-C ϭ 3.22; A-G ϭ 8.51; A-T ϭ 2.98; C-G ϭ 0.95; C-T ϭ analyses were based on a smaller phylogeny (excluding in- 14.63; G-T ϭ 1.00). These settings were subsequently used traspeci®c data) with 54 ingroup species obtained using Mo- for maximum-likelihood (ML) analyses and to estimate ML deltest, ML, and NPRS as described above. distances for minimum-evolution (ME) analyses. Maximum- parsimony (MP) analyses were conducted with heuristic RESULTS searches (TBR branch swapping, MULTREES option effec- tive, and 10 random stepwise additions of taxa). Transver- Phylogeny of the American Seven-Spined Gobies sions (Tv) were given two times the weight of transitions The alignment of the 12S rRNA, tRNA-Val, and 16S rRNA based on an empirical Ts:Tv ratio of 1.97. Robustness of the gene nucleotide sequences of 65 Gobiosomatini individuals inferred trees was tested using nonparametric bootstrapping with and as outgroups consisted of (Felsenstein 1985) with 1000 and 500 pseudoreplicates for 2195 positions. A total of 1646 positions were analyzed, of the ME and MP analyses, respectively. The quartet puzzling which 782 were invariant and 638 were phylogenetically in- method (Strimmer and von Haeseler 1996) was used to test formative under the parsimony criterion. Pairwise sequence the robustness of the ML analyses using 10,000 puzzling divergence between ingroup taxa varied from zero to 22.6% steps. All above-mentioned phylogenetic analyses were con- (17.8% excluding T. multifasciatum and batrachodes ducted with PAUP* version 4.0b8 (Swofford 2002). that had the longest branches) and substitutions were not A Bayesian inference (BI) of Gobiosomatini phylogeny saturated (as indicated by a Tv:Ts vs. patristic distance plot; was performed with MrBayes version 3.01 (Huelsenbeck and not shown). Ronquist 2001) by Markov chain Monte Carlo (MCMC) sam- The tree reconstructed by the Bayesian phylogenetic anal- pling for one million generations (four simultaneous MC ysis is shown in Fig. 2. Two main lineages corresponding to chains; sample frequency 100; burn-in 100,000 generations; the Microgobius group and the Gobiosoma group of Birdsong chain temperature 0.2) under the GTR ϩ I ϩ⌫model. et al. (1988) were recovered within the Gobiosomatini (Fig. Alternative phylogenetic hypotheses were tested using the 2). The Microgobius group consists of two clades, whereas Shimodaira-Hasegawa test (SH; Shimodaira and Hasegawa at least seven major clades were identi®ed in the Gobiosoma 1999). A LRT (Huelsenbeck and Crandall 1997) was per- group (Fig. 2). Maximum-parsimony phylogenetic analyses formed with ML trees with and without a molecular clock with a 2:1 Tv:Ts weighting scheme resulted in two shortest constraint. To date major cladogenetic events and to perform trees of 6665 steps. Different Tv:Ts weighting schemes (e.g. tests of diversi®cation rates, ultrametric trees were construct- 1:1 and 4:1) resulted in similar and congruent MP trees (not ed using the nonparametric rate smoothing (NPRS) method shown). The ME phylogenetic analysis recovered two trees EVOLUTION OF AMERICAN SEVEN-SPINED GOBIES 1587

FIG. 2. Phylogenetic relationships of the American seven-spined gobies based on a Bayesian phylogenetic analysis of mitochondrial 12S rRNA, tRNA-Val, and 16S rRNA sequence data using the GTR ϩ I ϩ⌫evolutionary model. The number above each node refers to the Bayesian posterior probability (shown as percentage) derived from 9000 Markov chain Monte Carlo sampled trees. Bootstrap values (Ͼ50%) for minimum-evolution (GTR ϩ I ϩ⌫distances) and maximum-parsimony (2Tv:1Ts) phylogenetic analyses are shown below branches (upper and lower values, respectively). The two Gobiosomatini groups (Gobiosoma group and Microgobius group) and the nine Gobiosomatini clades used in the text are indicated. Clades are: ELA, Elacatinus;TIG,Tigrigobius;RIS,Risor; GOB, Gobiosoma; BAR, ; NES, Nes; ABO, Aboma; MIC, Microgobius; BOL, . Outline drawings adapted from Hoese and Larson (1985) and Humann (1994) except Aboma (drawn directly from the specimen). 1588 LUKAS RUÈ BER ET AL.

(score ϭ 3.86). The topology of the ML tree (Ϫln likelihood ogenetic events that occurred in Gobiosomatini history (Fig. ϭ 23,763.45) is shown in Figure 3A. The MP, ME, and ML 3B). The origin of the American seven-spined gobies and the trees showed nearly identical topologies to the BI tree (Fig. initial divergence into the Gobiosoma and Microgobius 2 and 3A). In contrast to MP, ME, and ML phylogenetic groups date back to the late Eocene (around 40 million years analyses where Aboma was resolved as the most ago; Fig. 3B). Divergence into the nine major Gobiosomatini of the Gobiosoma group (although with no bootstrap or quar- clades occurred during the Oligocene (around 35±28 million tet puzzling support), the BI phylogenetic analysis resolved years ago; Fig. 3B). These nine clades have both Paci®c and Aboma after the Nes clade (a topology which was not rejected Atlantic representatives indicating that other factors besides under ML using the SH test: P ϭ 0.95). may have been important in the evolution of Our results reject the monophyly of the genus Gobiosoma the Gobiosomatini. Interestingly, Paci®c lineages are mostly including the subgenera Austrogobius, Elacatinus, Gobioso- resolved in a basal position within each main clade, with the ma, Garmannia, and Tigrigobius (Figs. 2 and 3, SH test: P exception of Bollmannia. Subsequent diversi®cation of the ϭ 0.00) as de®ned by BoÈhlke and Robins (1968). The genus main clades occurred during the -Pliocene (22±3 mil- Enypnias is placed deep within a Gobiosoma clade that in- lion years ago). In the case of the Elacatinus ϩ Tigrigobius cludes the subgenera Austrogobius, Garmannia, and Gobio- ϩ Risor clade, the timing of the main cladogenetic events soma (Figs. 2 and 3A) although the SH test did not reject a correlates with the origin and diversi®cation of the Caribbean topology where the genus Enypnias was placed basal to the fauna (Budd et al. 1995). latter three subgenera (P ϭ 0.06). Furthermore, the subgenus Tigrigobius is not monophyletic (SH test: P ϭ 0.02). Diversi®cation Rate through Time For several species, more than one individual was included A LTT plot of the Gobiosomatini was estimated based on in the phylogenetic analyses. Conspeci®cs of a given species the NPRS tree. The LTT plot is convex (Fig. 5) indicating a were resolved monophyletically with the only exception of burst of speciation early in the history of the group. Convex Elacatinus louisae sampled from and Jamaica. The of E. louisae is supported by results from an extensive phylogeographic study of the genus Elacatinus (M. → S. Taylor, pers. comm.). The highest intraspeci®c sequence FIG. 3. (A) Maximum-likelihood tree based on the GTR ϩ I ϩ⌫ divergence (uncorrected p-distance, 7.6%) was observed be- evolutionary model. Robustness of the nodes was tested with the tween two individuals of Garmannia nudum from El Salvador quartet puzzling method. Nodes with 50±69% (*) or 70±100% (**) and Panama. Differences in scale count are found in north- support are indicated. Atlantic or Paci®c origin of each species is south populations of G. nudum (Hoese 1976). Therefore, both indicated. (B) Ultrametric tree based on the nonparametric rate smoothing analysis. Two datasets were used for the constant-rates/ molecular and morphological data suggest a potential case Monte Carlo constant-rates tests. Taxa that were excluded for the of cryptic species, a situation often encountered in marine 54 ingroup taxa dataset are underlined and the additional two taxa species (Knowlton 1993). Within , identical that were excluded for the 52 ingroup taxa dataset are marked with sequences were found between two individuals from Long a number symbol (#) (see text for details). Thicker lines indicate lineages that have signi®cantly higher than expected diversi®cation Island and Chesapeake Bay. These specimens clustered with rate using the relative cladogenesis statistic (P Ͻ 0.05; 54 ingroup an individual from eastern Florida to the exclusion of two taxa dataset). Black bars indicate major habitat transitions during individuals from western Florida and the . the evolution of the Gobiosomatini: 1, generally deep ocean habitat This intraspeci®c pattern is concordant with the existence of over mud; 2, generally coastal habitats mostly over silt/mud sub- strate; 3, generally coastal-estuarine habitats over diverse sub- zoogeographic boundaries in Florida that are also found in strates; 4, generally on outer edge of coral or rocky reefs over ®ne many coastal maritime in this area (reviewed in Avise or coarse sand substrate with or without rocks; 5, rocky and coral et al. 1987; Avise 2000). reef habitats. Gray bars illustrate some behavioral adaptations in the Elacatinus ϩ Tigrigobius ϩ Risor clade: a, cleaning behavior mostly on moray eels, no apparent cleaning stations (M. S. Taylor, Rates of Evolution pers. comm.); b, in association with sea urchins; c, sponge-dwelling (mostly internal cavity of sponges), with morphological adaptations A likelihood-ratio test with (Ϫln L ϭ 23,988.51) and with- (BoÈhlke and Robins 1969); d, exclusively or occasionally found out (Ϫln L ϭ 23,763.45) the molecular clock enforced re- inside chiton (Choneplax lata) burrows (Taylor and Van Tassell jected overall constancy of rates of evolution in the Gobio- 2002); e, facultative cleaning behavior mostly on moray eels, no somatini (␦ϭ450.12, df ϭ 67, P K 0.001). To further explore apparent cleaning stations, in the Gulf of California mostly in as- such rate heterogeneity, we plotted the mean path lengths sociation with sea urchins (M. S. Taylor, pers. comm.); f, obligate cleaning behavior (generally get food by cleaning but may feed (ϮSD) across the ME tree for each clade (Fig. 4). The Boll- from other sources, maintain cleaning stations), primarily coral- mannia and Microgobius clades showed the lowest rates of dwellers. Sponge-dwelling has occasionally been observed in E. evolution, whereas the Elacatinus, Tigrigobius, Risor, and ®garo (Rocha et al. 2000), E. prochilos (Whiteman and CoÃte 2002), Barbulifer clades showed the highest (Fig. 4). The Risor and and E. oceanops (J. L. Van Tassell, pers. obs.), but cleaning stations have not been observed when species associate with sponges. Oc- Nes clades had large standard deviations due to the extremely casional sponge-dwelling behavior in these species may be ex- long branches of T. multifasciatum and Psilotris, respectively plained by high inter- and intraspeci®c competition of newly settled (Figs. 3A and 4). In the absence of rate constancy we used juveniles for cleaning stations on coral heads. g, sponge-dwelling the NPRS method to construct an ultrametric tree (Fig. 3B) (tube sponges), mostly in sponge cavity (not inside sponge canals) sometimes on outside of sponges. The scale bar below the tree based on the ML tree, which was used for further analyses. shows the time scale resulting from a calibration of the molecular Calibration of a molecular clock with a Tv rate of 0.14% clock (Tv rate of 0,14% per million years) based on the circled per million years allowed us to roughly date the main clad- node. EVOLUTION OF AMERICAN SEVEN-SPINED GOBIES 1589 1590 LUKAS RUÈ BER ET AL.

FIG. 4. Path lengths (average Ϯ SD) based on GTR ϩ I ϩ⌫ distances across the minimum-evolution tree from the most recent ancestor to the members of the nine Gobiosomatini clades are shown. Clade designations as in Figure 2. FIG. 5. Semilogarithmic plot of lineages through time (LTT) using the 54-taxa dataset. The constant-rate (CR) model is rejected under the CR/Monte Carlo constant-rate tests if the internal nodes are LTT plots might also occur due to incomplete taxon sampling signi®cantly closer to the root (convex LTT plot) than expected (Nee et al. 1994a). To rule out the effect of incomplete taxon under the pure birth model (dotted line; Pybus and Harvey 2000). sampling on changes in the rate of cladogenesis through time we used the CR/MCCR tests, which assume that diversi®- cation occurs equally among lineages. To test such an as- on the NPRS tree, the relative cladogenesis statistic identi®ed sumption we performed the three-taxon test, the B1 test, and several events of unusually rapid diversi®cation rates (Fig. the relative cladogenesis statistic. 3B). Signi®cantly higher than expected diversi®cation rates Assuming a hypothetical consisting of one spe- were found at the base of the Elacatinus ϩ Tigrigobius ϩ cies, the three-taxon test indicated a signi®cantly higher di- Risor ϩ Gobiosoma clade, at the base of the Gobiosoma and versi®cation rate at the origin of the Gobiosomatini, but when the Elacatinus clades, and at the base of the two main clades assuming two hypothetical outgroup species, the test results within Elacatinus (Fig. 3B). We removed one species from were nonsigni®cant (Table 1). With this test, Aboma shows both the Elacatinus (E. oceanops) and the Gobiosoma (Gar- a signi®cantly slower diversi®cation rate when it is placed mannia robustum) clades, resulting in a new phylogeny (52 basal to the Gobiosoma group as indicated by the MP, ME, ingroup dataset). This dataset was not rejected under the B1 and ML phylogenetic analyses (Table 1). test (n ϭ 52, B1 ϭ 25.04, P Ͼ 0.05). The B1 test rejected the hypothesis of equal diversi®cation Following Pybus and Harvey (2000), we applied the CR/ rates across lineages (n ϭ 54, B1 ϭ 25.41, P Ͻ 0.05). Based MCCR tests to both the 54 and 52 ingroup datasets. In both

TABLE 1. Results of three-taxon tests of shifts in diversi®cation rates (Sanderson and Donoghue 1994) obtained for selected Gobiosomatini clades. Calculations were performed with LRDIVERSE version 0.8 (written by M. Sanderson) using Monte-Carlo simulations (1000 replicates). Relative age of internal nodes was set to zero and the diversi®cation rate of internal branches was set equal to the average of the three terminal taxa. Asterisk indicates signi®cance at the 5% level.

P-value for Outgroup Ingroup 1 Ingroup 2 null model Unknown 1 sp Microgobius group 37 spp1 Gobiosoma group 94 spp2 0.983*7 Unknown 2 spp Microgobius group 37 spp1 Gobiosoma group 94 spp2 0.926 Microgobius group 37 spp1 Aboma clade 1 sp3 remaining Gobiosoma group 91 spp4 0.971*8 Microgobius group 37 spp1 Nes clade 28 spp3 remaining Gobiosoma group 64 spp4 0.132 Tigrigobius clade 4 spp5 (Paci®c) Elacatinus (Atlantic) 14 spp6 0.913 Elacatinus puncticulatus Elacatinus (sponge-dweller) 6 spp Elacatinus (cleaner) 7 spp 0.837 1 Including and Parella according to Birdsong et al. 1988. 2 Including , Gymneleotris, Ophiogobius, Pariah, Pycnomma, Robinsichthys, and according to Birdsong (1988), Birdsong et al. (1988) and Van Tassell (1998). 3 Assuming that either the Aboma or the Nes clade is the most basal lineage within the Gobiosoma group, respectively. 4 Eleotrica and Ophiogobius cannot be assigned to any clade within the Gobiosoma group, and therefore were left out of these calculations, whereas Gymneleotris, Pariah, Pycnomma, Robinsichthys, and Varicus were assigned to the Nes clade according to Birdsong (1988) and Van Tassell (1998). 5 Including T. zebrella according to Van Tassell (1998). 6 Including E. astronasum, which is a feeder. 7 Higher diversi®cation rate in the two ingroups (P ϭ 0.352). 8 Slower diversi®cation rate in ingroup 1 (P ϭ 0.650). EVOLUTION OF AMERICAN SEVEN-SPINED GOBIES 1591

TABLE 2. Test of the null hypothesis that per-lineage speciation subgenera based on squamation, cephalic pore patterns, or and rates have remained constant through time in the sensory papillae patterns (Ginsburg 1933; BoÈhlke and Robins Gobiosomatini using the constant-rate and Monte Carlo constant- rate tests (Pybus and Harvey 2000). 1968; Hoese 1971; Van Tassell 1998). Furthermore, there is little agreement on the species composition of each subgenus. Number True Our molecular phylogeny provides a robust framework to of number Critical accomplish a taxonomic revision of the genus Gobiosoma sampled of value of species species ␥ (at (sensu BoÈhlke and Robins 1968). Given our molecular re- Phylogeny (x) (y) ␥1 5% level)2 sults, it seems reasonable to elevate some of the former sub- 54 taxa dataset 54 131 Ϫ3.45 (P Ͻ 0.05) Ϫ3.28 genera to a generic rank (e.g. Elacatinus). Future taxonomic 52 taxa dataset 52 131 Ϫ3.58 (P Ͻ 0.05) Ϫ3.30 work will also have to take into account that Tigrigobius is 1 Calculated with Genie version 3.0 (Pybus and Rambaut 2002). not monophyletic, and that the genus Enypnias may be nested 2 Distribution obtained from 10,000 replicates using MCCRTest (Pybus within a clade that includes the subgenera Gobiosoma, Gar- 2000). mannia, and Austrogobius.

Patterns of Diversi®cation cases, the hypothesis of a constant-rate model is rejected (Table 2). Our data, therefore, provide strong phylogenetic With over 130 species, the Gobiosomatini is one of the evidence that after an initial burst, overall speciation rate has most successful groups of tropical marine ®shes in the New decreased through time in the American seven-spined gobies. World, and has been suggested to be an example of an adap- tive radiation (Hoese and Larson 1985). Using the molecular phylogeny as a robust framework, we performed several sta- DISCUSSION tistical tests to characterize the patterns of diversi®cation of and the group, and to test the hypothesis of constant speciation rate in the group. According to the three-taxon test, a higher The phylogeny of the New World Gobiosomatini was re- diversi®cation rate at the origin of the tribe is supported only covered based on mitochondrial 12S and 16S rRNA and by assuming a single hypothetical outgroup species (Table tRNA-Val gene sequence data using representatives of the 1). Unfortunately, the outgroup of the Gobiosomatini is cur- genera Gobius and Vanneaugobius from the tem- rently not known. The possibility that more than one species perate eastern Atlantic-Mediterranean region as outgroups forms the outgroup of the Gobiosomatini needs to be tested. (Figs. 2 and 3). The monophyly of the Gobiosomatini was Several Gobiinae genera, namely Odondebuenia, Corcyro- highly supported and is concordant with the results from gobius, Gorogobius, Egglestonichthys, and have Birdsong (1975), who de®ned the tribe based on the presence been proposed to be closely related to the Gobiosomatini of seven spines in the ®rst dorsal ®n, a dorsal ®n pterygiop- (Miller and Tortonese 1969; Miller 1972; 1978; Miller and hore formula of 3±221110, and a vertebral count of 11 ϩ 16 Wongrat 1979; but see Birdsong 1988; Birdsong et al. 1988; Ϫ 17. Changes in the above-mentioned character states have Van Tassell 1998). Hence, it remains uncertain whether the occurred in Risor and (Birdsong 1975). origin of the Gobiosomatini was associated with an increase Both genera display variability in the dorsal pterygiophore in diversi®cation rate according to the three-taxon test. On pattern and, in the latter, also in vertebral number and ar- the other hand, the relative cladogenesis statistic identi®ed rangement. The unusual morphology of these genera may be unusually rapid diversi®cation rates associated with several associated with their highly evolved, probably obligate, cladogenetic events in the Gobiosoma group (Fig. 3B). sponge-dwelling habits (Beebe 1928; BoÈhlke and Robins Both the LTT plot and the CR/MCCR tests indicate an 1969; Birdsong 1975). The close sister-group relationship of early burst of speciation within the Gobiosomatini followed these two taxa in the molecular phylogeny further suggests by a slowdown (Fig. 5 and Table 2). Such a pattern of change different levels of to sponge dwelling within a in speciation rate is typical of adaptive radiations (Johns and common evolutionary lineage (BoÈhlke and Robins 1969). Our Avise 1998; Lovette and Bermingham 1999) and has been results also support the division of the tribe into the Gobio- documented with similar tests for only two groups of marine soma group and the Microgobius group, which was based on ®shes, ice®shes and rock®shes (Johns and Avise 1998). The the presence of a hypural fusion in the former group (Bird- decrease in speciation rate in such adaptive radiations may song et al. 1988; Figs. 2 and 3). Although a hypural fusion be due to a decrease of speciation opportunities as ecological is missing in Aboma, this genus is placed within the Gobio- and geographical spaces are ®lled (Schluter 2000). soma group with high bootstrap, quartet puzzling, and Bayes- ian support values (Figs. 2 and 3). This phylogenetic place- Factors Underlying Gobiosomatini Evolution ment is concordant with cheek myology and papillae patterns (Van Tassell 1998). Uncertainty remains whether this genus Our molecular phylogeny suggests that ecology rather than is the most basal member of the Gobiosoma group or resolves biogeography has played an important role in the evolution after the Nes clade. Hence, with the current molecular data of the seven-spined gobies. We observed that shifts in ecol- we cannot determine whether the absence of the hypural fu- ogy seem to be associated with speci®c cladogenetic events sion in Aboma represents the plesiomorphic condition or a (shown in Figure 3B). Ecological shifts are either associated secondary loss. with recent speciation events, and thus closely related taxa Gobiosoma, the largest genus within the seven-spined go- exhibit pronounced ecological differences, or instead are as- bies, has been divided into as few as four or as many as seven sociated with the long-term persistence and subsequent ra- 1592 LUKAS RUÈ BER ET AL. diation of lineages, and thus most shifts would be observed cluded in this study). It is likely that similar behavioral and between distantly related lineages, that is, close to the root niche specialization may have been important in the diver- of the tree (Barraclough et al. 1999; Barraclough and Nee si®cation of the other Gobiosomatini clades. However, with 2001). In the phylogeny of the seven-spined gobies, major the exception of striking examples such as Nes longus, that habitat shifts observed conform to the second case. They are lives commensally with snapping shrimp (Appendix), almost located at the origin of the main Gobiosomatini lineages (Fig. no behavioral and ecological data are available from the other 3B). The majority of the Microgobius species studied are clades to further evaluate this hypothesis. found in coastal habitats mostly over mud/silt substrate (Ap- The association of an early burst of speciation with major pendix). In contrast, the species from the Bollmannia clade, habitat shifts, and the subsequent diversi®cation linked to which was resolved as of the Microgobius clade, specializations in behavior and microhabitat, reinforces our are generally found in deeper ocean habitats below 20±30 interpretation of an adaptive radiation in the American seven- meters depth, over mud or coral sand ¯ats; as far as it is spined gobies. The observation of early lineage divergence known species of both groups live in burrows (Appendix). due to major habitat shift in the Gobiosomatini is in agree- The members of the Gobiosoma, Barbulifer, and Aboma ment with increasing data that support the important role of clades are best characterized as shallow coastal or estuarine invasions of novel habitats in adaptive divergence in speci- species that inhabit the continental coasts in both oceans ation (Schluter 2000). To further test this interpretation, more (Appendix). Major ecological shifts from continental coastal- ecomorphological data are required to establish phenotype- estuarine habitats into tropical reef or reef-edge areas are environment correlations (Schluter 2000). associated with the origin of the Elacatinus, Tigrigobius, and Risor clades and the Nes clade, respectively (Fig. 3B and Marine Speciation Appendix). These major habitat shifts are concordant with the burst of speciation early in the evolutionary history of It is challenging to reconcile speciation with the vast ex- the Gobiosomatini detected in our analysis (Fig. 5 and Table tension and apparent absence of physical barriers in marine 2). environments. Under these conditions, organisms with a pe- Major habitat shifts may account for the initial divergence lagic larval stage and the potential for long distance dispersal within the Gobiosomatini, but they cannot alone explain the would be expected to form populations with very little or no subsequent diversi®cation within each clade. We propose that spatial structure (Palumbi 1994; Shulman and Bermingham microhabitat and behavioral specializations may have been 1995; Mora and Sale 2002; Palumbi and Warner 2003), which crucial in the radiation of the clades. For example, members reduces the opportunity for speciation mechanisms to oper- of the three clades Elacatinus, Tigrigobius, and Risor, which ate. However, recent studies indicate that the dispersal of are found in rocky or coral reef habitats, exhibit astonishing pelagic larvae may be restricted by factors such as settlement specializations, that seem to have a single origin (Fig. 3B). preferences and persistence ability of the larvae in new set- È Tigrigobius multifasciatum and Ginsburgellus novemlineatus, tlements (Ohman et al. 1998). Therefore, self-recruitment in two species that are associated with sea urchins, were con- marine populations may be a substantial factor counteracting sistently grouped as sister species (although a long branch gene ¯ow and setting the conditions for local adaptation. The attraction effect cannot be discarded in this case). Risor ruber importance of self-recruitment is suggested by the persistence and Evermannichthys spongicola, two highly specialized of endemic species with pelagic larvae on small isolated is- sponge-dwelling taxa, were grouped together. The Atlantic lands (Robertson 2001), the unexpected genetic subdivision Elacatinus, the neon gobies, are separated into two distinct of many marine populations (Shulman and Bermingham clades, one comprising species that remove ectoparasites 1995; Miya and Nishida 1997; Taylor and Hellberg 2003), from ®shes and maintain cleaning stations (obligate cleaning and new information on the behavior of pelagic larvae en- behavior), the other comprising species that live in associ- hancing near-shore retention (Jones et al. 1999; Swearer et ation with sponges (Fig. 3B). This latter behavior seems to al. 1999; Cowen et al. 2000). be more derived among neon gobies, since the most basal Mechanisms of speciation that are thought to be important member of the group, the Paci®c E. puncticulatus shows fac- in terrestrial and lacustrine systems such as ecological ad- ultative cleaning behavior. Unfortunately, we cannot deter- aptation and have been barely explored in mine the origin of the plankton feeding behavior of E. as- marine systems. However, new data are accumulating that tronasus (Colin 1975) a species not included in this analysis. highlight the importance of these mechanisms on marine spe- It is interesting to note that in the Indo-Paci®c some Labridae ciation (McMillan et al. 1999; Bernardi et al. 2002; Streelman (mostly from the genus ) show cleaning behavior et al. 2002; Taylor and Hellberg 2003). (Grutter 1999), and that these species show striking conver- We have reconstructed the phylogeny of the gobiid New gence in body shape and coloration with Elacatinus from the World tribe Gobiosomatini to investigate the processes driv- Caribbean Sea. The behavioral dichotomy between cleaners ing cladogenesis in this speciose group. We ®nd that major and sponge-dwellers likely played an important role in the habitat shifts seem to be directly related with the origin and diversi®cation of these ®shes that are often found in sym- divergence of the main clades and that their subsequent di- patry. According to the relative cladogenesis statistic, a recent versi®cation seems to be related to behavior and niche spe- radiation event may be associated with the origin of cleaning cializations. Therefore, our results are in agreement with the and sponge-dwelling behaviors (Fig. 3B). Facultative clean- two-phase scenario proposed by Streelman et al. (2002) for ing behavior has evolved independently in two Paci®c spe- the evolution of parrot®shes. In these ®shes an ecomorphol- cies, Tigrigobius limbaughi and T. digueti (species not in- ogical split into reef and seagrass types seems to have played EVOLUTION OF AMERICAN SEVEN-SPINED GOBIES 1593 a signi®cant role in the initial divergence of the major lin- Barraclough, T. G., and S. Nee. 2001. and speciation. eages within the group, whereas color and breeding behavior Trends Ecol. Evol. 16:391±399. Barraclough, T. G., J. E. Hogan, and A. P. Vogler. 1999. Testing seem to have been involved in subsequent diversi®cation whether ecological factors promote cladogenesis in a group of (Streelman et al. 2002). An initial divergence followed by tiger beetles (Coleoptera: Cicindelidae). Proc. R. Soc. Lond. B diversi®cation has also been documented in lacustrine ®shes 266:1061±1067. such as from the East African Great Lakes. This Beebe, W. 1928. Beneath tropic seas, a record of diving among the coral reefs of Haiti. Halcyon House, New York. successful group of ®shes has undergone a rapid adaptive Bernardi, G., S. J. Holbrook, R. J. Schmitt, N. L. Crane, and E. radiation in which ®rst adaptation to major habitats (sand- DeMartini. 2002. Species boundaries, populations and colour dwellers and rock-dwellers) occurred, followed by diversi- morphs in the coral reef three-spot damsel®sh (Dascyllus tri- ®cations through divergent selection on ecomorphological maculatus) . Proc. R. Soc. Lond. B 269:599±605. Birdsong, R. 1975. The osteology of Microgobius signatus POEY traits and mate recognition systems (Albertson et al. 1999; (Pisces: Gobiidae), with comments on other gobiid ®shes. Bull. Danley and Kocher 2001). Fla. State Mus. Biol. Sci. 19:135±187. Speciation in both terrestrial and marine systems may be ÐÐÐ. 1988. Robinsichthys arrowsmithensis, a new genus and spe- driven by similar evolutionary mechanisms, although they cies of deep-dwelling gobiid ®sh from the western Caribbean. may operate at different temporal and spatial scales (Bowen Proc. Biol. Soc. Wash. 101:438±443. Birdsong, R. S., and C. R. Robins. 1995. New genus and species et al. 2001; Colborn et al. 2001). Understanding of speciation of seven-spined (Gobiidae: Gobiosomini) from the of®ng in the marine realm will bene®t from the accumulation of of the Amazon River, . Copeia 1995:676±683. more phylogenetic data combined with an improvement of Birdsong, R., E. Murdy, and F. Pezold. 1988. A study of the ver- our knowledge on the ecology and behavior of marine ®shes. tebral column and median ®n osteology in gobioid ®shes with comments on gobioid relationships. Bull. Mar. Sci. 42:174±214. BoÈhlke, J., and C. Robins. 1968. Western Atlantic seven-spined ACKNOWLEDGMENTS gobies, with descriptions of ten new species and a new genus, and comments on Paci®c relatives. Proc. Acad. Nat. Sci. Phila. We thank M. S. Taylor (Louisiana State University) for 120:45±174. providing samples of T. limbaughi, T. saucrus, and B. com- ÐÐÐ. 1969. Western Atlantic sponge-dwelling gobies of the ge- munis, and helpful information on the biology of Elacatinus nus Evermannichthys: their taxonomy, habits and relationships. and Tigrigobius, as well as S. Villalba for redrawing some Proc. Acad. Nat. Sci. Phila. 121:1±24. Bowen, B., A. , L. Rocha, W. Grant, and D. Robertson. 2001. of the goby outlines. E. P. Lessa, O. Pybus, A. Rambaut, and of the trumpet®shes (Aulostomus): M. Sanderson kindly provided some of the software used. complex on a global scale. Evolution 55:1029±1039. We also thank S. Nee, O. Pybus, and A. Rambaut for pro- Budd, A. F., K. G. Johnson, and J. C. Edwards. 1995. Caribbean viding helpful information on the relative cladogenesis sta- reef coral diversity during the early to middle Miocene: an ex- ample from the Anguilla formation. Coral Reefs 14:109±117. tistic. We are grateful to S. A. Karl, S. Villalba, and four Colborn, J., R. Crabtree, J. Shaklee, E. Pfeiler, and B. Bowen. 2001. anonymous reviewers for helpful comments and suggestions The evolutionary enigma of bone®shes (Albula spp.): cryptic on the manuscript. The following people (and institutions) species and ancient separations in a globally distributed shor- helped arrange local ®eld logistics and aided in the collection e®sh. Evolution 55:807±820. Colin, P. 1975. The neon goby. TFH Publications, Neptune City, of specimens: R. Boyer (Centro de Estudios de Desiertos y NJ. Oceanos), L. Findley, H. Placencia, and A. van der Heiden Cowen, R., K. Lwiza, S. Sponaugle, C. Paris, and D. Olson. 2000. (Centro de Investigacion en Alimentacion y Desarrollo); T. Connectivity of marine populations: open or closed? Science Mathes (Virginia Institute of Marine Science); D. R. Rob- 287:857±859. Danley, P. D., and T. D. Kocher. 2001. Speciation in rapidly di- ertson (Smithsonian Tropical Research Institute); and R. verging systems: lessons from Lake Malawi. Mol. Ecol. 10: Ruiz-Carus (Florida Marine Research Institute). Field col- 1075±1086. lections by JVT have been made possible by funding provided Eastman, J., and A. McCune. 2000. on the Antarctic con- in part by the Smithsonian Tropical Research Station, Pan- tinental shelf: evolution of a marine species ¯ock? J. Biol. ama, and the National Museum of Natural History, Smith- 57(Suppl. A):84±102. Echelle, A. A., and I. Korn®eld. 1984. Evolution of ®sh species sonian Institution. LR was supported by a Swiss Science ¯ocks. Univ. of Maine Press, Orono, ME. Foundation postdoctoral fellowship (823A-061218). Labo- Felsenstein, J. 1985. Con®dence limits on phylogenies: An approach ratory work was ®nanced by a grant from the Janggen-PoÈhn- using the bootstrap. Evolution 39:783±791. Foundation, Switzerland, to LR and the Spanish Ministerio Fryer, G., and T. D. Iles. 1972. The ®shes of the Great Lakes of Africa. Oliver and Boyd, Edinburgh, U.K. de Ciencia y TecnologõÂa (REN2001-1514/GLO) to RZ. Ginsburg, I. 1933. A revision of the genus Gobiosoma ( Go- biidae) with an account of the genus Garmannia. Bull. Bingham LITERATURE CITED Oceanogr. Coll. 4:1±59. Grutter, A. S. 1999. Cleaner ®sh really do clean. Nature 398: Albertson, R. C., J. A. Markert, P. D. Danley, and T. D. Kocher. 672±673. 1999. Phylogeny of a rapidly evolving clade: the cichlid ®shes Hoese, D. F. 1971. A revision of the eastern Paci®c species of the of Lake Malawi, East Africa. Proc. Natl. Acad. Sci. USA 96: gobiid ®sh genus Gobiosoma, with a discussion of relationships 5107±5110. of the genus. Ph.D. diss., Dept. of , Univ. of Calif. San Avise, J. C. 2000. Phylogeography. Harvard Univ. Press, Cam- Diego, CA. bridge, MA. ÐÐÐ. 1976. Variation, synonymy and a redescription of the go- Avise, J. C., J. Arnold, R. M. Ball, E. Bermingham, T. Lamb, J. E. biid ®sh Aruma histrio and a discussion of the related genus Neigel, C. A. Reeb, and N. C. Saunders. 1987. Intraspeci®c Ophiogobius. Copeia 1976:295±306. phylogeography: the mitochondrial DNA bridge between pop- Hoese, D. F., and H. Larson. 1985. Revision of Eastern Paci®c ulation genetics and systematics. Annu. Rev. Ecol. Syst. 18: species of the genus Barbulifer (Pisces: Gobiidae). Copeia 1985: 489±522. 333±339. 1594 LUKAS RUÈ BER ET AL.

Huelsenbeck, J. P., and K. A. Crandall. 1997. Phylogeny estimation Posada, D., and K. Crandall. 1998. Modeltest: testing the model of and hypothesis testing using maximum likelihood. Annu. Rev. DNA substitution. Bioinformatics 14:817±818. Ecol. Syst. 28:437±466. Pybus, O. G. 2000. MCCRTest. Department of Zoology, Univ. of Huelsenbeck, J. P., and F. R. Ronquist. 2001. MrBayes: Bayesian Oxford, Oxford, U.K. inference of phylogeny. Bioinformatics 17:754±755. Pybus, O. G., and P. H. Harvey. 2000. Testing macro-evolutionary Humann, P. 1994. Reef ®sh identi®cation: Florida, Caribbean, Ba- models using incomplete molecular phylogenies. Proc. R. Soc. hamas. New World Publication, Jacksonville, FL. Lond. Biol. B 267:2267±2272. Johns, G. C., and J. C. Avise. 1998. Tests for ancient species ¯ocks Pybus, O. G., and A. Rambaut. 2002. GENIE (Genealogy Interval based on molecular phylogenetic appraisals of Sebastes rock- Explorer). Ver. 3.0 Department of Zoology, Univ. of Oxford, ®shes and other marine ®shes. Evolution 52:1135±1146. Oxford, U.K. Jones, G., M. Milicich, M. Emslie, and C. Lunow. 1999. Self-re- Rambaut, A., and M. Charleston. 2001. TreeEdit. Ver. 1.0 Depart- cruitment in a coral reef ®sh population. Nature 402:802±804. ment of Zoology, Univ. of Oxford, Oxford, U.K. Kirkpatrik, M., and M. Slatkin. 1993. Searching for evolutionary Rambaut, A., P. H. Harvey, and S. Nee. 1997. End-Epi: an appli- patterns in the shape of a phylogenetic tree. Evolution 47: cation for reconstructing phylogenetic and population processes 1171±1181. from molecular sequences. Comput. Appl. Biosci. 13:303±306. Knowlton, N. 1993. Sibling species in the sea. Annu. Rev. Ecol. Ritchie, P. A., L. Bargelloni, A. Meyer, J. A. Taylor, J. A. Mac- Syst. 24:189±216. donald, and D. M. Lambert. 1996. Mitochondrial phylogeny of Kocher, T. D., W. K. Thomas, A. Meyer, S. V. Edwards, S. PaÈaÈbo, trematomid ®shes (Nototheniidae, Perciformes) and the evolu- F. X. Villablanca, and A. C. Wilson. 1989. Dynamics of mito- tion of Antarctic ®sh. Mol. Phylogenet. Evol. 5:383±390. chondrial DNA evolution in animals: ampli®cation and sequenc- Robertson, D. 2001. Population maintenance among tropical ®shes: ing with conserved primers. Proc. Natl. Acad. Sci. USA 86: inference from small-island endemics. Proc. Natl. Acad. Sci. 6196±6200. USA 98:5667±5670. Lovette, I. J., and E. Bermingham. 1999. Explosive speciation in Rocha, L. A., I. L. Rosa, and B. M. Feitoza. 2000. Sponge-dwelling the New World Dentroica warblers. Proc. R. Soc. Lond. Biol. ®shes of northeastern Brazil. Environ. Biol. Fish. 59:453±458. B 266:1629±1636. Sanderson, M. 1997. A nonparametric approach to estimate diver- McMillan, W. O., L. A. Weigt, and S. R. Palumbi. 1999. Color gence times in the absence of rate constancy. Mol. Biol. Evol. pattern evolution, , and genetic differentiation 14:1218±1231. in brightly colored butter¯y®shes (Chaetodontidae). Evolution Sanderson, M., and M. Donoghue. 1994. Shifts in diversi®cation 53:247±260. rate with the origin of angiosperms. Science 264:1590±1593. Miller, P. 1972. Generic status and redescription of the Mediter- ÐÐÐ. 1996. Reconstructing shifts in diversi®cation rates on phy- ranean ®sh Gobius liechtensteini Kolombatovic, 1891 (Teleostei: logenetic trees. Trends Ecol. Evol. 11:15±20. Gobioidae), and its af®nities with certain American and Indo- Sanderson, M., and M. F. Wojciechowski. 1996. Diversi®cation Paci®c gobies. J. Nat. Hist. 6:395±407. rates in a temperate legume clade: Are there ``so many species'' ÐÐÐ. 1978. The status of the West African ®sh Gobius nigri- of Astragalus (Fabaceae)? Am. J. Bot. 83:1488±1502. cinctus with reference to New World autochthones and an Old Schluter, D. 1996. in postglacial ®shes. Phi- World colour-analogue. Zool. J. Linn. Soc. 64:27±39. los. Trans. R. Soc. Lond. B 351:807±814. Miller, P., and E. Tortonese. 1969. Distribution and systematics of ÐÐÐ. 2000. The ecology of adaptive radiation. Oxford Univ. the gobiid ®sh Odondebuenia balearica (Pellegrin Frage). Mus. Press, Oxford, U.K. Civico di Storia Naturale Giacomo Doria. Annali 77:342±359. Shimodaira, H., and H. Hasegawa. 1999. Multiple comparisons of Miller, P., and P. Wongrat. 1979. A new goby (Teleostei: Gobiidae) log-likelihoods with applications to phylogenetic inference. Mol. from the South China Sea and its signi®cance for gobioid clas- Biol. Evol. 16:1114±1116. si®cation. Zool. J. Linn. Soc. 67:239±257. Shulman, M., and E. Bermingham. 1995. Early histories, ocean Miya, M., and M. Nishida. 1997. Speciation in the open ocean. currents, and the of Caribbean reef ®shes. Nature 389:803±804. Evolution 49:897±910. Mora, C., and P. F. Sale. 2002. Are populations of coral reef ®sh Simpson, G. G. 1953. The major features of evolution. Columbia open or closed? Trends Ecol. Evol. 17:422±428. Univ. Press, New York. Nee, S., E. C. Holmes, R. M. May, and P. H. Harvey. 1994a. Ex- Smith, D. G., and C. Baldwin. 1999. Psilotris amblyrhynchus, a new tinction rates can be estimated from molecular phylogenies. Phi- seven-spined goby (Teleostei: Gobiidae) from Belize, with notes los. Trans. R. Soc. Lond. B 344:77±82. on settlement-stage larvae. Proc. Biol. Soc. Wash. 112:433±442. Nee, S., R. May, and P. Harvey. 1994b. The reconstructed evolu- Stanley, S. M. 1979. . Freeman, San Francisco, CA. tionary process. Philos. Trans. R. Soc. Lond. B. 344:305±311. Streelman, J., M. Alfaro, M. Westneat, D. Bellwood, and S. A. Nee, S., E. C. Holmes, R. M. May, and P. H. Harvey. 1995. Esti- Karl. 2002. Evolutionary history of the parrot®shes: biogeog- mating extinction from molecular phylogenies. Pp. 164±182 in raphy, ecomorphology, and comparative diversity. Evolution 56: J. H. Lawton and R. M. May, eds. Extinction rates. Oxford Univ. 961±971. Press, Oxford, U.K. Strimmer, K., and A. von Haeseler. 1996. Quartet puzzling: a quartet Nee, S., T. G. Barraclough, and P. H. Harvey. 1996. Temporal maximum-likelihood method for reconstructing tree topologies. changes in : detecting patterns and identifying caus- Mol. Biol. Evol. 13:964±969. es. Pp. 230±252 in K. J. Gaston, ed. Biodiversity: a biology of Swearer, S., J. Caselle, D. Lea, and R. Warner. 1999. Larval re- numbers and differences. Blackwell Sciences, Oxford, U.K. tention and recruitment in an island population of coral-reef ®sh. Nelson, G. J. 1994. . John Wiley and Sons, New Nature 402:799±802. York. Swofford, D. L. 2002. PAUP*: phylogenetic analysis using parsi- OÈ hman, M. C., P. L. Munday, G. P. Jones, and M. J. Caley. 1998. mony (*and other methods), Ver. 4.0b8. Sinauer Associates, Settlement strategies and distribution patterns of coral-reef ®sh- Sunderland, MA. es. J. Exp. Mar. Biol. Ecol. 225:219±238. Taylor, M. S., and M. E. Hellberg. 2003. Genetic evidence for local Osborne, H. F. 1902. The law of adaptive radiation. Am. Nat. 36: retention of pelagic larvae in a Caribbean reef ®sh. Science 299: 353±363. 107±109. Palumbi, S. R. 1994. , reproductive isolation, Taylor, M. S., and J. L. Van Tassell, 2002. Observations on mi- and marine speciation. Annu. Rev. Ecol. Syst. 25:547±572. crohabitat utilization by three widely distributed Neotropical go- Palumbi, S. R., and R. R. Warner. 2003. Why gobies are like hob- bies of the genus Elacatinus. Copeia 2002:1134±1136. bits. Science 299:51±52. Thompson, J. D., T. J. Gibson, F. Plewniak, J. Jeanmougin, D. G. Palumbi, S., A. Martin, S. Romano, W. McMillan, L. Stice, and G. Higgins. 1997. The Clustal X windows interface: ¯exible strat- Grabowski. 1991. The simple fools guide to PCR. Univ. of Ha- egies for multiple sequence alignment aided by quality analysis waii Press, Honolulu, HI. tools. Nucleic Acids Res. 25:4876±4882. EVOLUTION OF AMERICAN SEVEN-SPINED GOBIES 1595

Van Tassell, J. 1998. Phylogenetic relationships of species within Whiteman, E. A., and I. M. CoÃteÂ. 2002. Cleaning activity of two the gobiid genus Gobiosoma sensu BoÈhlke and Robins (1968) Caribbean cleaning gobies: intra- and interspeci®c comparisons. with comments on their relationships to other genera in the tribe J. Fish. Biol. 60:1443±1458. Gobiosomini. Ph.D. diss. City University of New York, New York. Corresponding Editor: S. Karl 1596 LUKAS RUÈ BER ET AL. rs, and museum voucher generally tube sponges generally tube sponges sponges cleaner ®sh, congregatestations in cleaning sponges in deeper®sh, waters, congregate cleaner intions cleaning sta- cleaner ®sh, congregatestations in cleaning sponges, especially tube sponges mud substrate shell habitat atledges bottom of rock ger head sponges in water column near rock cliffs deeper than 30 meters meters croalgae or grasses croalgae or grasses depths some sand Sponge dweller, medium-depth reefs, Coral heads and large sponges, Sponge dwellerÐin and around Under sea urchins in shallow waters Mud bottoms, ocean Coral sand ¯ats in or near burrows, Shallow waters associated with ma- Shallow waters associated with ma- Coastal estuaries in ®ne mud with les and islands off Venezuela gin Islands, Lesser Antilles to Venezuela Rica to Ecuador nezuela and Central American coast and Central American coast Colombia southern Gulf of California; Costa Editorial Of®ce at [email protected]). PPENDIX A Evolution Locality (locality number in Fig. 1) Distribution Ecology and behavior tad (4) Bahamas, Lee Stocking Island (14)Jamaica, St. Ann's Bay Bahamas (16) to Belize Bahamas to Belize Sponge dweller, medium-depth reefs, Sponge dweller, generally tube specimenGrand Cayman (15)Curacao (17) Brazil Bahamas and Grand Cayman Islands Coral heads and large sponges, Curacao, Bonaire, Netherlands Antil- Coral heads in shallow water and aquarium specimen Bahamas, south to Lesser Antilles Panama, Colon (6)aquarium specimen Panama, Colon-Atlantic Southern Bahamas, Puerto Rico, Vir- Coastal waters in sandy and hard Mexico, San CarlosEl (3) Salvador, Golfo de Fonseca (4)Panama, Panama Panama Bay toPanama, (5) El Panama Salvador BayFlorida (5) Keys, Marathon (10) Gulf of CaliforniaCuracao (17) Gulf of North California Carolina Gulf to to of Colombia Dry Rock California Tortugas reef to habitat Colombia in coarse sand, Coastal, Sponge in dweller sand-rock generally Coastal, habitat inside in log- Rock sand-rock reef habitat areas under small rocks Bahamas, Cayman Islands, Antilles El Salvador, Golfo de Fonseca (4)Gulf of Mexico,El Louisiana Salvador, (8) Departamento Costa La Rica Liber- to northern Peru Gulf of Mexico Mud bottoms, ocean deeper than 20 Mud bottoms, ocean Florida Keys, Marathon (10) South Florida, Greater Antilles, Ve- Florida Keys, Marathon (10) Bahamas, south Florida, to Venezuela Mexico, La PazPanama, (1) Colon (6) Gulf of California Bahamas, south Florida, to Venezuela Shallow waters under cobble rocks Panama, Panama Bay (5) Gulf of California from Sonora to species novum El Salvador, Golfo de Fonseca (4) Panama to Peru Black mud bottom at 20±30 meter Genus Subgenus Species GobiosomaGobiosoma Elacatinus Elacatinus louisae louisae GobiosomaGobiosoma Elacatinus ®garo Gobiosoma Elacatinus genie Elacatinus horsti Gobiosoma Gobiosoma Elacatinus chancei Gobiosoma Austrogobius spilotum EnypniasGenus novumGinsburgellus seminudus species novum novemlineatus Curacao (17) Curacao Deep water rocky habitatsÐschools EnypniasEnypniasEvermannichthys aceras spongicola seminudus zebra BollmanniaBollmannia chlamydes communis Bollmannia marginalis Bollmannia boqueronensis Barbulifer ceuthoecus Aruma histrio Barbulifer ceuthoecus Gobiosomatini Aboma etheostoma Summary of specimens,numbers sample are locations, given distribution, in and the ecological/behavioral Electronic data. Appendix Specimen (currently identi®cation available numbers, from GenBank the accession numbe EVOLUTION OF AMERICAN SEVEN-SPINED GOBIES 1597 Coral ¯ats found onlycement in holes in coral Coral reef, onor vertical coral walls, sponges, heads Estuarine environmentwith in oyster association shell, rocks, etc. mud substrate Estuarine environmentwith in oyster association shell, rocks, etc. Estuarine environmentwith in oyster association shell, rocks, etc. Estuarine environmentwith in oyster association shell, rocks, etc. Estuarine environmentwith in oyster association shell, rocks, etc Estuarine and coastal, mud, sand,habitats weed sand, mud and/or shell substrate sand, mud and/or shell substrate Coastal waters inmud/sand rocky or habitats shell substrate with Coastal waters inmud/sand rocky or habitats shell substrate with Rocky reef areasdepressions, facultative in cleaner small caves and es, especially tube sponges generally tube sponges strate nal, fresh and Coral heads and sponges, cleaner ®sh Coral heads, cleaner ®sh, congregatecleaning in station nezuela Cayman Islands, British Honduras,maica Ja- ern Florida, GulfFlorida of to Mexico, Mexico southern ern Florida, GulfFlorida of to Mexico, Mexico southern Atlantic coast Massachusetts toern south- Florida, GulfFlorida of to Mexico, Mexico southern ern Florida, GulfFlorida of to Mexico, Mexico southern ern Florida, GulfFlorida of to Mexico, Mexico southern lantic coast of Florida of Panama Canal of Panama Canal uador Jamaica, Belize,Gulf of Yucatan, Mexico northern Gulf of Mexico . Continued. PPENDIX A Locality (locality number in Fig. 1) Distribution Ecology and behavior Grand Cayman (15) Bahamas to Lesser Antilles and Ve- Virginia, Chesapeake Bay (12)Honduras, Caribbean New (7) York to Georgia British West Indies, Bahama Islands, Estuarine environment, oyster reef and Virginia, Chesapeake Bay (12) Atlantic coast Massachusetts to south- Florida, St. Petersburg (9) Atlantic coast Massachusetts to south- New York, Long(13) Island, south shore Gulf of Mexico, Louisiana (8) Atlantic coast Massachusetts to south- Panama Bay (5)Panama Bay (5)Florida, Indian River (11) West coast of Mexico to Ecuador West Atlantic coast coast of Massachusetts Mexico to to south- Ecuador Coastal waters in rocky habitats Coastal with waters in rocky habitats with Florida, Indian River (11) Gulf of Mexico, Texas to Florida, At- El Salvador, Golfo de Fonseca (4) Peru to El Salvador and Atlantic coast Grand Cayman (15)Caribbean, Honduras (7)Mexico, Puerto Penasco (2)Panama Canal (5)Panama Canal (5) Florida GulfPanama Keys of Canal and Honduras (5) Jamaica GulfPanama of Bay California (5) Sponge Panama dweller, Canal medium-depth reefs, Panama Canal Sponge Panama dweller, Canal in Intertidal and waters, around rock spong- Peru with to sand El sub- Salvador and Atlantic coast Rock and mud bottom in Panama Fresh Ca- water mud-rock Fresh sediments water mud-rock sediments Mexico, La Paz (1) Gulf of California to Panama and Ec- aquarium specimen Lesser Antilles, Barbados, St. Croix, Florida Keys, Sugar Loaf Key (10) Bahama Islands west to Honduras and species novum Mexico, Puerto Penasco (2) Gulf of California Coastal waters rock with sand substrate species novum Panama Canal (5) Panama Canal Rock and mud bottom in Panama canal Genus Subgenus Species Gobiosoma Tigrigobius gemmatum GobiosomaGobiosomaGobiosoma Gobiosoma Tigrigobius dilepis Gobiosoma Gobiosoma bosc Gobiosoma Gobiosoma bosc Gobiosoma Gobiosoma bosc Gobiosoma Gobiosoma bosc GobiosomaGobiosomaGobiosoma GarmanniaGobiosoma Garmannia paradoxum Garmannia paradoxum Gobiosoma bosc Gobiosoma Garmannia robustum Gobiosoma Garmannia nudum GobiosomaGobiosoma Garmannia Garmannia homochroma nudum GobiosomaGobiosomaGobiosoma ElacatinusGobiosoma Elacatinus lori Gobiosoma Garmannia xanthiprora Garmannia chiquita Garmannia hildebrandi homochroma Gobiosoma Elacatinus puncticulatus Gobiosoma Elacatinus prochilos Gobiosoma 1598 LUKAS RUÈ BER ET AL. 45 meters under rocks on Ͼ Sponge dweller, generally foundin on barrel or sponges Coral reef areassand in bottom small cracks with substrate strate Ocean inmeter silt/mud depths substrate, 15±30 mud habitats shell bottoms habitats, coastal ®ne coral sand, lives commensally with snapping shrimp Ocean inmeter silt/mud depths substrate, 15±30 Beach areas withsand/broken shell silt/mud substrate and some above burrows in the sand substrate in coastal habitats Coral reef,rounded on boulder the surface of large, Coral reefholes on in coral vertical cement walls and in Under sea urchinswaters or rocks in shallow depressions, cleaning behavior terstices of the calcareous rubble sand substrate the Antilles duras, to Colombia Gulf of Californiaama to the Gulf of Pan- ama to Panama gin Islands, and Belize an Sea to Barbados and Curacao . Continued. PPENDIX A Locality (locality number in Fig. 1) Distribution Ecology and behavior Florida Keys, Marathon (10) Bahamas and Florida, south through Spain, Gran Canaria,Spain, Puerto Lanzarote, Rico MalaSpain, Lanzarote, Mala eastern Atlantic and Mediterranean Coastal waters on sand with grass Canary Islands, Madeira and Guinea West Africa, Coastal Canary under small Islands rocks on sand sub- Deep water Spain, Gran Canaria, Puerto Rico eastern Atlantic and Mediterranean Coastal waters in ®ne sand with rock Florida, St. Petersburg (9)Florida Keys, Marathon (10)Florida, Sarasota (9)Florida Keys, Virginia Marathon to (10) Texas Bahamas to YucatanJamaica, St. Ann's Bay (16) Bahamas to Yucatan and Virginia Venezuela to Texas Coral reef Bahamas, areas, Puerto inhabit Inshore Rico, burrows Estuarine Belize, waters in and Hon- fresh over waters, calcareous ®ne silt silt/ Silt/mud substrate with oyster, sponge El Salvador, Departamento La Libertad (4) Florida Keys, Marathon (10)Panama, Bahia Honda (5) Gulf of Mexico to Lesser Antilles Coarse sand substrate near coral Gulf reefs, of California to the Gulf of Pan- Panama Canal (5)Panama Canal (5) Panama Canal Southern Baja and Gulf of California rocky mangrove streams Bahamas, Lee Stocking Island (14) Florida Keys, Bahamas, Jamaica, Vir- Jamaica, St. Ann's Bay (16) Bahamas and throughout the Caribbe- Mexico, Gulf of California (1)Florida Keys, Marathon (10)Bahamas, Lee Stocking Gulf Island of (14) California Bahamas, Cuba, Southern Cayman Florida, Islands south Cuba to Haiti Coral ¯ats under rock ledges and in- Rocky reef areas in small caves and species El Salvador, Golfo de Fonseca (4) Gulf of California to Peru Mostly silt/mud, some broken shell Genus Subgenus Species Risor ruber Vanneaugobius canarensis GobiusVanneaugobius pruvoti niger Gobius xanthocephalus MicrogobiusMicrogobiusMicrogobiusNesPsilotris gulosus microlepis thalassinus longus batrachodes Microgobius erectus MicrogobiusMicrogobius Microgobius carri erectus Microgobius brevispinis birdsongi Gobiosoma Tigrigobius saucrus Gobiosoma Tigrigobius pallens Gobiosoma Tigrigobius limbaughi GobiosomaGobiosoma Tigrigobius Tigrigobius macrodon multifasciatum Outgroups