Molecular Phylogeny of Nemadactylus and Acantholatris (Perciformes: Cirrhitoidea: Cheilodactylidae), with Implications for Taxonomy and Biogeography C

Molecular Phylogenetics and Evolution Vol. 13, No. 1, October, pp. 93–109, 1999 Article ID mpev.1999.0622, available online at http://www.idealibrary.com on

Molecular Phylogeny of Nemadactylus and Acantholatris (Perciformes: Cirrhitoidea: Cheilodactylidae), with Implications for Taxonomy and Biogeography C. P. Burridge School of Zoology, University of Tasmania, GPO Box 252-05, Hobart, Tasmania 7001, Australia

Received July 21, 1998; revised November 6, 1998

There are five recognized species of Nemadactylus. The species of Nemadactylus and Acantholatris are Four of these are restricted to the waters of Australia perciform fishes with representatives in each ocean of and New Zealand, and the remaining species occur the Southern Hemisphere. Mitochondrial DNA se- along the east coast of South America (Fig. 1). One quences were obtained from all five species of Nema- species, commonly called ‘‘king tarakihi,’’ has only dactylus, two of the three Acantholatris species, and recently been proposed (Roberts, 1993; Smith et al., several outgroup taxa. Analysis of cytochrome b se- 1996). The record of N. macropterus at Saint Paul and quences placed A. monodactylus and A. gayi within an Amsterdam Island by Agnot (1951) was a misidentifica- otherwise entirely Nemadactylus clade, suggesting that these genera are synonymous. The Acantholatris se- tion according to Duhamel (1989). Acantholatris is not quences were also very similar to those from three of represented in the waters of Australia or New Zealand. the Nemadactylus species, despite their geographic Instead, its three members occur around isolated is- separation. Analysis of D-loop sequences paralleled lands and seamounts such as Juan Fernańdez and the the cytochrome b results, but provided greater resolu- Desventuradas in the southeastern Pacific, and those tion of species relationships. Nemadactylus sp. and A. which form a loose chain from Tristan da Cunha and gayi are transoceanic sister taxa. Polytypic clades Gough Island in the South Atlantic, to Saint Paul and observed for N. macropterus and A. monodactylus most Amsterdam Islands in the Indian Ocean (Fig. 1). likely reflect incomplete sorting of mitochondrial DNA The taxonomy of these species as based on external lineages. It is proposed that this group dispersed and characteristics has been problematic. Although Allen radiated during the last 0.6–2.6 million years, and the and Heemstra (1976) synonymized Acantholatris with possible mechanisms of this process are Cheilodactylus, the characters described by Gill (1862) discussed. ௠ 1999 Academic Press and Smith (1980), particularly the presence of a relatively long and narrow anal fin, distinguish Acanthola- tris from the cheilodactylid genera Cheilodactylus, Chi- INTRODUCTION rodactylus, and Dactylophora (Lamb, 1990). While the The species of Nemadactylus Richardson, 1839, and remaining cheilodactylid genus Nemadactylus shares Acantholatris Gill, 1862, are marine perciforms which the majority of characteristics which define Acanthola- occur in subtropical to cool-temperate waters through- tris, including the long narrow anal fin, these genera out the Southern Hemisphere (Fig. 1). Juveniles and have been separated primarily on the number of anal adults occur demersally over sand or reef substrates, at fin rays, with the former having 14–19 and the latter depths of 1–350 m (Annala, 1987; Wo¨hler and Sańchez, 10–12 (Gill, 1862; Lamb, 1990). 1994; Andrew et al., 1995). Movements of juveniles and Nemadactylus bergi and A. gayi have been called adults appear limited, with a gradual migration into Cheilodactylus in recent nonsystematic literature (e.g., deeper waters with age (Annala, 1987; Andrew et al., Meleńdez and Villalba, 1992; Wo¨hler and Sańchez, 1995). However, a common trait of these and related 1994) and were also indexed as Acantholatris in the taxa is an offshore pelagic larval phase of 7–12 months systematic work by Greenwood (1995). These species in duration, which implies high dispersal capabilities are obviously not members of Cheilodactylus, given (Annala, 1987; Andrew et al., 1995). These species their relatively long and narrow anal fins, and since mature at 2–6 years and serially spawn large numbers they have anal fin ray counts of 14–15 and 12 they of small eggs (Annala, 1987; Andrew et al., 1995; should be referred to as Nemadactylus and Acanthola- Jordan, 1997); maximum ages exceed 25 years (Andrew tris, respectively (Lamb, 1990). et al., 1995; Jordan, 1997). N. bergi and N. macropterus are morphologically

93 1055-7903/99 $30.00 Copyright ௠ 1999 by Academic Press All rights of reproduction in any form reserved. 94 C. P. BURRIDGE

FIG. 1. Distribution of Nemadactylus and Acantholatris species. Ns, Nemadactylus sp.; Nm, N. macropterus; Nb, N. bergi; Nd, N. douglasii; Nv, N. valenciennesi; Am, A. monodactylus; Av, A. vemae; Ag, A. gayi; ?, possible records of A. gayi (R. Meleńdez C., Museo Nacional de Historia Natural, Santiago, Chile, 1998, pers. comm.). similar, but have been distinguished by the width of the and the external characteristics listed above. The famil- supra-cleithrum relative to the diameter of the eye, the ial allocation of these genera does not influence this relative lengths of their thickened pectoral fin rays, and study, and they are referred to as cheilodactylids pend- lateral line scale counts (Norman, 1937; Lamb, 1990). ing familial revision. However, two of these differences are not consistent In addition to their taxonomic uncertainties, the among the holotype and paratypes of N. bergi (R. W. G. distribution patterns and potentially high dispersal White, University of Tasmania, Australia, 1995, pers. capabilities of Nemadactylus and Acantholatris make comm.), and the separate status of this species requires them interesting subjects for a molecular phylogenetic justification. study. The biogeography and radiation of these and Nemadactylus and Acantholatris are usually allo- similarly distributed fishes has attracted some atten- cated to the Family Cheilodactylidae (Allen and Heem- tion (Eschmeyer and Hureau, 1971; Briggs, 1974; Wil- stra, 1976; Smith, 1980; Lamb, 1990). However, a son and Kaufmann, 1987; Collette and Parin, 1991; recent systematic study based on urohyal morphology Andrew et al., 1995), but no molecular studies have suggested that these genera should be reallocated to a been conducted. Members of the rock lobster genus cirrhitoid sister family, the Latridae (Greenwood, 1995). Jasus have distributions and dispersal capabilities Nemadactylus and Acantholatris are clearly distin- similar to Nemadactylus and Acantholatris, and their guished from all other cheilodactylid and latrid genera radiation, together with that of the related genus by both preliminary molecular data (Burridge, unpubl.) Panulirus, have received much attention (Pollock, MOLECULAR PHYLOGENY OF Nemadactylus AND Acantholatris 95

1990, 1992, 1993; George, 1997), including two molecu- material. Symmetric PCR amplifications were con- lar phylogenetic studies (Brasher et al., 1992; Ovenden ducted in 50-µL volumes, containing 1.0 units of Taq et al., 1997). The theories regarding the radiation of DNA polymerase (Promega), 5 µL of 10ϫ reaction these rock lobsters may well be applicable to much of buffer (Promega), 200 µM dNTPs, 0.5 µM each oligo- the Southern Hemisphere marine fauna with similar nucleotide primer, 1.5–2.5 mM MgCl2, and 25–100 ng species distributions and dispersal capabilities. genomic DNA. The oligonucleotide primers employed The aim of this study was to obtain molecular data were L14841 and H15149 for a 307-bp region of cyto- from species of Nemadactylus and Acantholatris that chrome b (Kocher et al., 1989) and proline-light (5Ј would clarify their taxonomy and provide phylogenetic AACTC TCACC CCTAR CTCCC AAAG 3Ј) and D-loop- information about their radiation with respect to biogeo- heavy (5Ј GGCCC TGAAR TAGGA ACCAR ATG 3Ј) for graphic processes and events. Mitochondrial DNA se- the left domain of the D-loop. Primer L14724 (Paä¨bo, quences were chosen as the type of data to be collected, 1990) was required to obtain clean cytochrome b se- for reasons of resolving power and accessibility (Meyer, quences in some instances. 1994). Thermal cycling conditions for the amplification of cytochrome b were 10 cycles of 94°C/30 s, 45°C/30 s, and MATERIALS AND METHODS 72°C/60 s, followed by 30 cycles of 94°C/30 s, 55°C/30 s, and 72°C/60 s. Conditions for the amplification of Frozen or ethanol-preserved muscle and liver tissues D-loop sequences were identical with the exception of were obtained from all species of Nemadactylus and annealing being conducted at 60°C. An initial denatur- Acantholatris with the exception of A. vemae, for which ation of 94°C/5 min and a final extension of 72°C/10 min only formalin-fixed material was easily accessible (Table were employed for both templates. The results of PCR 1). Material from representatives of the remaining were assessed by 1.0% agarose gel electrophoresis, with cheilodactylid genera and Cirrhitus splendens (Cirrhiti- visualization under UV radiation after ethidium bro- dae) were also obtained for use as outgroups. The mide staining. Cirrhitidae is considered the most plesiomorphic of the Templates were gel purified using the QIAquick Gel five cirrhitoid families (Greenwood, 1995). Extraction Kit (Qiagen). DNA sequences were deter- Total DNA was extracted from approximately 40 mg mined using either the fmol DNA sequencing system of tissue following a standard CTAB phenol-chloroform (Promega) with [␥33P]ATP end-labeled primers, or the procedure (Hillis et al., 1990). The technique of Shed- ABI PRISM BigDye Terminator Cycle Sequencing lock et al. (1997) was employed for the A. vemae Ready Reaction Kit (Applied Biosystems Inc.). Both light and heavy strands were sequenced for each tem- TABLE 1 plate, enabling the verification of character states. D-loop sequences were aligned using CLUSTALW 1.7 Collection Details of Nemadactylus and Acantholatris (Thompson et al., 1994). Individuals Analyzed To test for the presence of certain character states, PCR-amplified D-loop fragments were digested with Indi- Species Collection site Month/year viduals Dpn II and Hinc II restriction endonucleases. Digests were conducted in 10-µL volumes, containing 5 µL of A. monodactylus Tristan Island /1994 3 amplified DNA, 0.5 units of either enzyme, and the A. monodactylus Gough Island /1994 3 appropriate volume of concentrated buffer (New En- A. monodactylus Saint Paul Island 3/1997 9 A. monodactylus Amsterdam Island 2–3/1997 7 gland Biolabs). Digests were incubated at 37°C for 1 h, A. gayi Juan Fernańdez 2–3/1997 3 and the results were visualized as for PCR amplifica- A. vemae Vema Seamounta 6/1966 1 tion, using a 100-bp increment size standard (Promega). N. macropterus Albany, Australia 10/1991–2/1992 32 Maximum parsimony analysis (Cavalli-Sforza and N. macropterus Tasman Island, 9/1991 1 Edwards, 1967) was conducted using PAUP 3.1.1 (Swof- Tasmania N. macropterus New Zealand 4/1992 1 ford, 1993). The cytochrome b and D-loop sequences Nemadactylus sp. Kiama, Australiab 6/1994 1 were analyzed using the Branch and Bound (Hendy Nemadactylus sp. Three Kings 4/1994 2 and Penny, 1982) and Heuristic search algorithms, Islands, N. Z. respectively, with accelerated character state optimiza- N. bergi Mar del Plata, 8/1996 51 tion (ACCTRAN) and MULPARS in effect. The Heuris- Argentina N. douglasii Coffs Harbour, 6/1994 1 tic algorithm employed tree bisection–reconnection Australiab (TBR) branch swapping and 100 replicates of random N. valenciennesi Port Lincoln, 5/1995 1 sequence addition. Three weighting schemes were em- Australia ployed for transition and transversion nucleotide substi- a South African Museum specimen 25030 (formalin-fixed). tutions. These were equal weighting, differential weight- b Australian Museum frozen tissue collection (Nemadactylus sp. ing in accordance with the reciprocal of their observed I.34845-001, N. douglasii I.34844-001). frequencies, or analysis of transversions only. Gaps 96 C. P. BURRIDGE were treated as missing data or analyzed as a separate wise Kimura distances using the Kruskal loss function character state. Consistency (Kluge and Farris, 1969) and monotonic regression, implemented by SYSTAT and retention (Farris, 1989) indices were calculated 5.2. under unweighted conditions using all characters. Non- parametric bootstrap analysis (Felsenstein, 1985a) was performed based on 1000 replicate data sets, retaining RESULTS groups compatible with the 50% majority rule consen- The protocol of Shedlock et al. (1997) was successful sus. For each D-loop bootstrap replicate, 10 rather than in obtaining DNA fragments of up to 450 bp in length 100 random sequence addition replicates were per- from formalin-fixed A. vemae material. However, at- formed with the Heuristic algorithm. tempts at the amplification of cytochrome b and D-loop Neighbor-joining (Saitou and Nei, 1987) and maxi- sequences from this DNA were not successful. mum likelihood (Felsenstein, 1981) analyses were performed using components of the PHYLIP 3.572 package Cytochrome b (Felsenstein, 1993). DNADIST was used to calculate pairwise sequence distances, corrected for multiple With the exception of A. vemae, cytochrome b se- substitutions by the Kimura (1980) two-parameter quences were obtained from one individual of each model with a transition/transversion ratio of 2.0. Clus- Nemadactylus and Acantholatris species and several tering of pairwise distances was conducted using outgroup taxa (GenBank Accession nos. AF067084– NEIGHBOR. Bootstrap analysis was performed on AF067095). Among the species of Nemadactylus and 1000 replicate data sets created by SEQBOOT, and the Acantholatris, substitutions were observed at 27 of the consensus tree was calculated by CONSENSE. Maxi- 307 nucleotide positions analyzed, and the variation mum likelihood analysis was conducted using DNAML was phylogenetically informative at 10 of these. The with a transition/transversion ratio of 2.0, empirical majority of the substitutions represented transitions at base frequencies, one category of substitution rate, third codon positions, and this pattern, combined with global rearrangements, and 10 randomizations of se- the absence of length mutations, suggested that ortholo- quence input order. gous sequences were obtained. The estimated tree An approximate indication of phylogenetic informa- length–frequency distribution for these sequences was tion content was derived from the skewness of tree significantly skewed ( g1 ϭϪ0.61, P Ͻ 0.01), suggest- length–frequency distributions (Hillis and Huelsen- ing the presence of phylogenetic signal. beck, 1992). Estimated g1 values were obtained using Unweighted maximum parsimony analysis recov- the RANDOM TREES (1,000,000 trees) option of PAUP. ered a single most-parsimonious tree of 172 steps, with Likelihood ratio tests were conducted on each data set a consistency index of 0.744 and a retention index of to investigate the presence of clock-like sequence evolu- 0.651 (Fig. 2). The sequences from A. monodactylus and tion (Felsenstein, 1993). The likelihoods compared were A. gayi clustered within an otherwise entirely Nemadac- derived from analyses performed both with (DNAMLK tylus clade. The Acantholatris sequences were very of PHYLIP) and without (DNAML) the constraint of similar to those from Nemadactylus sp., N. macrop- terminal taxa being equidistant from the root. terus, and N. bergi, forming a single clade with a The number of additional steps required to achieve maximum of only three transition substitutions, or alternative topologies was determined using the CON- 0.98% corrected sequence divergence, between any of STRAINTS option of PAUP. Significantly different to- these five taxa (Table 2). The sister clade of this group pologies were identified by the nonparametric two- comprised N. douglasii and N. valenciennesi. The se- tailed Wilcoxon signed ranks test (Templeton, 1983; quences of these two clades differed by 14–18 transi- Felsenstein, 1985b). When n was greater than 20, a tions and two to three transversions, corresponding to normal approximation of the test statistic with correc- 5.39–7.15% sequence divergence (Table 2). Distances tion for ties was calculated by SYSTAT 5.2 (SPSS Inc.). from the Nemadactylus and Acantholatris sequences to The parametric Kishino and Hasegawa (1989) test the nearest cheilodactylid sequence, C. spectabilis, implemented by DNAML was used to identify signifi- ranged between 12.32 and 13.55% (Table 2). High cant differences in the likelihoods of alternative topolo- bootstrap values (Ͼ70%) were observed for the entire gies. Nemadactylus–Acantholatris clade, the N. valencien- D-loop nucleotide diversities (␲) and their standard nesi and N. douglasii clade, and the clade containing deviations were calculated using DnaSP 2.52 (Rozas the five most similar Nemadactylus and Acantholatris and Rozas, 1997), according to equations 10.5 and 10.7 sequences (Fig. 2). The low bootstrap value for the of Nei (1987). D-loop haplotype diversities (h) were relationship between the Nemadactylus-Acantholatris calculated using equation 8.5 of Nei (1987). Two- clade and that containing other cheilodactylids reflects dimensional scaling was performed on the D-loop pair- variation in the composition of the latter. MOLECULAR PHYLOGENY OF Nemadactylus AND Acantholatris 97

FIG. 2. Single most-parsimonious cladogram from the analysis of 307 bp partial mitochondrial cytochrome b DNA sequences from Nemadactylus and Acantholatris species. All character state changes were equally weighted. Cirrhitus splendens (Cirrhitidae) and representatives of the other cheilodactylid genera were included as outgroups. Branch lengths are proportional to the number of substitutions based on ACCTRAN character state optimization, indicated by the numbers below the branches. Bootstrap proportions for the taxa in each clade, as derived from 1000 replicates, are indicated by the numbers above the braches at each node. Tree length ϭ 172 steps, CI ϭ 0.744, RI ϭ 0.651.

Both maximum likelihood and neighbor-joining analy- ent topology, but this was probably due to the small ses produced topologies identical to that from un- number of informative characters. weighted parsimony analysis. Parsimony analysis with Based on maximum likelihood analyses performed increased weighting of transversions over transitions with and without the assumption of a molecular clock, a in accordance with the reciprocal of their observed likelihood ratio test suggested signiﬁcant deviation frequencies also recovered the same topology. The from clock-like sequence evolution (␹2 ϭ 29.22, df ϭ 10, bootstrap values from this weighted parsimony and the P Ͻ 0.02). However, the null hypothesis was not re- neighbor-joining analysis were similar to those from jected when this analysis was repeated with the exclu- unweighted parsimony. Parsimony analysis restricted sion of Cheilodactylus (Goniistius) gibbosus (␹2 ϭ 15.37, to transversion substitutions produced a slightly differ- df ϭ 9, P Ͼ 0.05). 98 C. P. BURRIDGE

TABLE 2

Interspeciﬁc Variation of Partial Cytochrome b and D-Loop Sequences and Measures of Intraspeciﬁc Partial D-Loop Sequence Variation

12345678

Interspeciﬁc variation 1 Acantholatris monodactylus 1113191635 30 22 31 20 90 86 82 2 Acantholatris gayi 0.33 0 0 2 19 16 36 8.76 24 17 23 92 81 75 3 Nemadactylus macropterus 0.33 0.00 0 2 19 16 36 6.36 6.97 28 19 94 88 82 4 Nemadactylus sp. 0.33 0.00 0.00 2 19 16 36 9.10 4.84 8.18 29 91 83 81 5 Nemadactylus bergi 0.98 0.65 0.65 0.65 21 18 36 5.71 6.62 5.44 8.44 87 82 78 6 Nemadactylus douglasii 6.45 6.45 6.45 6.45 7.15 11 38 31.74 32.78 33.51 32.44 30.43 67 101 7 Nemadactylus valenciennesi 5.39 5.39 5.39 5.39 6.08 3.66 36 29.32 27.58 30.19 28.43 27.69 21.43 100 8 Cheilodactylus spectabilis 12.32 12.70 12.70 12.70 12.70 13.55 12.76 28.33 25.68 28.37 27.9 26.74 38.01 36.92 Intraspeciﬁc D-loop variation Haplotypes sampled (n) 63537111 Maximum sequence divergence (%) 10.33 3.38 6.96 2.24 2.8 — — — Minimum sequence divergence (%) 0.83 0.83 2.79 1.39 1.38 — — — Haplotype diversity (h) 1.00 1.00 1.00 1.00 1.00 — — — Nucleotide diversity (␲) 0.054 0.024 0.050 0.018 0.022 — — — Standard deviation of ␲ 0.012 0.008 0.008 0.005 0.003 — — —

Note. Values above the diagonal represent observed substitutions, while those below the diagonal are Kimura (1980) corrected percentage sequence divergences obtained with a transition/transversion ratio of 2.0. For each comparison the upper value represents cytochrome b and the lower value is D-loop. Nucleotide diversities (␲), their standard deviations, and haplotype diversities (h) were calculated using equations 10.5, 10.7, and 8.5 of Nei (1987), respectively.

D-Loop Unweighted maximum parsimony analysis of se- Sequences representing the left domain of the D-loop quences, as aligned in Fig. 3 with gaps treated as were obtained from one to seven individuals per species missing data, produced a single most-parsimonious of Nemadactylus and Acantholatris (A. vemae not se- tree of 311 steps with a consistency index of 0.740 and a quenced) and one individual of the outgroup C. spectabi- retention index of 0.704 (Fig. 4A). As observed from the lis (GenBank Accession nos. AF067096–AF067120, analysis of cytochrome b, A. monodactylus and A. gayi AF072876–AF072877). The sequences analyzed varied sequences clustered within an otherwise entirely Nema- in length between 360 and 366 bp, and the favored dactylus clade. The Acantholatris sequences were again alignment was 373 characters long (Fig. 3). A total of closest to those of Nemadactylus sp., N. macropterus, 150 characters were variable among Nemadactylus and and N. bergi, but the D-loop sequences provided greater Acantholatris, and 91 of these were phylogenetically resolution among these taxa. informative. Transition substitutions were observed at Nemadactylus sp. and A. gayi sequences were struc- 125 of the variable sites, while transversions were tured as sister clades, with predominantly high (Ͼ70%) observed at 62. Gaps were inserted at 19 positions for bootstrap values for their relationships (Fig. 4A). The sequence alignment, with 9 of these positions represent- sequences from N. macropterus, N. bergi, and A. mono- ing otherwise invariant sites. No identical sequences dactylus were structured into four clades. Three of were identified for the D-loop region, and intraspecific these were polytypic, individually containing both N. nucleotide diversities ranged from 0.018 to 0.054 (Table macropterus and A. monodactylus sequences, while the 2). The estimated tree length–frequency distribution of remaining clade was monotypic for N. bergi. The N. all sequences was significantly skewed ( g1 ϭϪ1.25, macropterus (Western Australia) and A. monodactylus P Ͻ 0.01), suggesting the presence of phylogenetic sig- (Amsterdam Island 1) sequence clade was distin- nal. The amplified fragments also contained 28 bp of guished due to its distance from other sequences. the tRNA proline gene and a putative termination- Although the bootstrap values representing three of associated sequence (TAS). These were invariant among these four clades were quite high (Ͼ70%), the inferred taxa, suggesting that orthologous sequences were relationships between these clades received only moder- obtained. ate bootstrap values. The sequences from N. douglasii MOLECULAR PHYLOGENY OF Nemadactylus AND Acantholatris 99 and N. valenciennesi again formed the sister clade to esis was also rejected when the divergent C. spectabilis, the other Nemadactylus and Acantholatris sequences. N. douglasii, and N. valenciennesi sequences were The trees recovered by neighbor joining (Fig. 4B) and removed during the calculation of likelihoods maximum likelihood (not shown) analyses differed (␹2 ϭ 27.36, df ϭ 16, P Ͻ 0.05). However, trees pro- slightly in topology from each other and also in compari- duced with and without the assumption of a molecular son to that of unweighted parsimony. Although the clock differed only in those relationships which varied same polytypic N. macropterus–A. monodactylus and among the analyses described above. Therefore, it does monotypic N. bergi clades were produced by all three not appear that the deviation from clock-like sequence methods, there were differences in the inferred relation- evolution has pronouncedly influenced phylogenetic ships between these clades and also relative to the reconstruction. Nemadactylus sp.–A. gayi clade. It is these same rela- The restriction enzymes Dpn II and Hinc II were tionships which received low bootstrap values from used for the identification of any divergent D-loop parsimony and neighbor-joining analyses. Similar dif- sequences within N. bergi or, alternatively, N. bergi- ferences in topology and low bootstrap values were also like sequences within N. macropterus or A. monodacty- observed during parsimony analysis when including lus. These two enzymes had recognition sites diagnostic gaps as a character state, when differentially weighting for three of the five character states associated with the character state changes, and during the analysis of N. bergi clade (Fig. 3). All of 48 N. bergi individuals alternative sequence alignments (not shown). screened possessed the Dpn II site, but 9 lacked the The tree lengths obtained during the enforcement of Hinc II site. Four of these 9 individuals were sequenced maximum likelihood and neighbor-joining topologies (N. bergi 4–7; Fig. 3), as the Dpn II site may not have during parsimony analysis were only 1 (312 steps, been restricted to N. bergi-like sequences. However, in CI ϭ 0.737, RI ϭ 0.701) and 2 (313 steps, CI ϭ 0.735, each instance at least four of the five defining character RI ϭ 0.697) steps longer than that of unconstrained states of the N. bergi clade were present. Thirty-one parsimony, respectively. These three topologies were individuals of N. macropterus and 16 individuals of A. not significantly different as determined by the Temple- monodactylus were similarly screened for N. bergi-like ton and Kishino–Hasegawa tests (P Ͼ 0.05). Simulta- sequences. Each A. monodactylus individual lacked neous enforcement of N. macropterus and A. monodac- both the Dpn II and the Hinc II restriction sites. All of tylus sequence monophyly during parsimony analysis the N. macropterus individuals also lacked the Hinc II produced 15 trees of 330 steps in length (CI ϭ 0.697, site, but 5 individuals were digested by Dpn II. Two of RI ϭ 0.635), 19 steps longer than the most parsimoni- these individuals were sequenced (Western Australia 2 ous tree. Individual monophyly of either N. macrop- and 3; Fig. 3), but neither possessed N. bergi-like terus (11 trees) or A. monodactylus (5 trees) sequences character states at more than one of the five N. bergi ϭ ϭ required 327 steps (CI 0.703, RI 0.646). These clade-defining positions. enforced monophyly topologies were significantly inferior than the maximum parsimony, neighbor-joining, and maximum likelihood topologies (P Ͻ 0.05). Basal DISCUSSION placement of the N. bergi clade relative to the polytypic N. macropterus–A. monodactylus clades increased tree Phylogenetic analysis of cytochrome b sequences length by 2 steps (313 steps, CI ϭ 0.735, RI ϭ 0.697) clustered A. monodactylus and A. gayi within an other- relative to the most parsimonious topology, but this wise entirely Nemadactylus clade. The Acantholatris tree was not significantly inferior than the maximum sequences were very similar to those from Nemadac- parsimony, neighbor-joining, or maximum likelihood tylus sp., N. macropterus, and N. bergi, forming a single topologies (P Ͼ 0.05). clade with a maximum of only three transition substitu- Two dimensional-scaling of pairwise genetic dis- tions, or 0.98% corrected sequence divergence, between tances provided another representation of the relation- any of these five taxa. The sequences from N. valencien- ships between D-loop sequences (Fig. 5). Nemadactylus nesi and N. douglasii formed the sister clade to these sp. and A. gayi sequences were represented as distinct five taxa. clusters. The N. bergi sequences also formed a tight Phylogenetic analysis of D-loop sequences produced cluster, which was surrounded, but not overlapped, by results similar to that of cytochrome b, but with greater N. macropterus and A. monodactylus sequences. The resolution of the relationships between the five closest intraspecific distances observed for N. macropterus and taxa. The sequences from Nemadactylus sp. and A. gayi A. monodactylus often exceeded interspecific distances clustered as sister clades. Those from N. macropterus, (see also Table 2). N. bergi, and A. monodactylus were structured into four The likelihoods obtained from DNAML and DNAMLK clades, three of which were polytypic, individually reject the null hypothesis of clock-like sequence evolu- containing both N. macropterus and A. monodactylus tion (␹2 ϭ 34.35, df ϭ 19, P Ͻ 0.05). The null hypoth- sequences, while the fourth was monotypic for N. bergi. 100 .P BURRIDGE P. C.

FIG. 3. Partial mitochondrial D-loop sequences obtained from the species of Nemadactylus, Acantholatris, and the outgroup Cheilodacty- lus spectabilis. Sequences are recorded in the 5Ј to 3Ј direction for the light strand. Sequence identity with the reference taxon A. monodactylus (Tristan Island) is indicated by ‘‘.’’, while gaps inserted for alignment are indicated as ‘‘_’’. Sequence alignment was obtained with CLUSTALW 1.7 (Thompson et al., 1994) using the slow/accurate algorithm with a transition weight of 0.5, delayed alignment of sequences less than 40% identical, and gap open and extension penalties of 10 and 6.66, respectively. Potential characters for the identiﬁcation of N. bergi-like haplotypes based on N. bergi 1–3 are highlighted by ‘‘<’’, with a diagnostic restriction enzyme listed if available. The sequences N. bergi 4–7 were obtained after nondigestion by Hinc II (see text). OEUA HLGN OF PHYLOGENY MOLECULAR Nemadactylus AND Acantholatris

FIG. 3—Continued 101 102 .P BURRIDGE P. C.

FIG. 3—Continued MOLECULAR PHYLOGENY OF Nemadactylus AND Acantholatris 103

FIG. 4. (A) Single most-parsimonious cladogram from the analysis of partial mitochondrial D-loop DNA sequences from Acantholatris and Nemadactylus species. All character state changes were equally weighted, and characters containing gaps were ignored. Cheilodactylus spectabilis was the nominated outgroup. Branch lengths are proportional to the number of substitutions based on ACCTRAN character state optimization, as indicated by the numbers below the branches. Tree length ϭ 311 steps, CI ϭ 0.740, RI ϭ 0.704. (B) Corresponding neighbor-joining phenogram. Branch lengths are proportional to Kimura (1980) genetic distance, measured relative to the scale bar. Regions of topological difference between the two trees are indicated by broken lines. In both trees the bootstrap proportions for the taxa in each clade, as derived from 1000 replicates, are indicated by the numbers above the branches at each node. Branches leading to C. spectabilis, N. valenciennesi, and N. douglasii have been omitted from the neighbor-joining phenogram, but do not differ in topology or support from (A).

Consistent relationships between these four clades, the monophyly of N. macropterus and A. monodactylus and relative to the Nemadactylus sp.–A. gayi clade, sequences. The levels of intraspecific D-loop sequence were not recovered, although the topologies obtained divergence for N. macropterus and A. monodactylus were not significantly different. In contrast, signifi- were as large as some of the interspecific sequence cantly inferior topologies were observed when enforcing divergences within Nemadactylus and Acantholatris. 104 C. P. BURRIDGE

Fernańdez (Fig. 4). This accords with their identical anal fin ray counts (Lamb, 1990; Roberts, 1993). Be- cause Roberts (1993) and Smith et al. (1996) compared their specimens of Nemadactylus sp. only with N. macropterus, it is possible that Nemadactylus sp. may simply represent the first record of A. gayi in the waters of Australia and New Zealand; it may not comprise a new species. There are insufficient morphological data available to distinguish these two forms, but the genetic results from this study suggest that separation at some level is warranted (Figs. 4 and 5). However, more material should be analyzed. It is possible that N. bergi may be a junior synonym of N. macropterus, as the characters used to distinguish these taxa (Norman, 1937; Lamb, 1990) are not consistent among the holotype and paratypes of N. bergi (R. W. G. White, University of Tasmania, Australia, 1995, pers. comm.). The D-loop sequences obtained from three N. bergi individuals formed a monotypic FIG. 5. Two-dimensional scaling of Kimura (1980) genetic dis- clade, distinct from the N. macropterus sequences (Fig. tances between Acantholatris and Nemadactylus partial mitochon- 4). However, given the high level of intraspecific diver- drial D-loop sequences, using the Kruskal loss function and mono- gence within N. macropterus (Table 2, Fig. 5), restric- tonic scaling. Circles, A. monodactylus; open diamonds, A. gayi; tion enzyme analysis was employed to test for the closed diamonds, Nemadactylus sp.; triangles, N. macropterus; squares, N. bergi. The polytypic clades observed during phylogenetic presence of N. macropterus-like sequences within addi- analysis are circled. Stress of configuration ϭ 0.088. Nemadactylus tional N. bergi individuals and vice versa, but no such douglasii and N. valenciennesi were excluded from scaling so as to instances were identified. Although the N. bergi clade facilitate maximum resolution of the remaining sequences. was not placed basal to the polytypic N. macropterus–A. monodactylus clades, the enforcement of such a topology required only two more steps than the most parsi- Systematics monious tree and was not significantly inferior than the Analysis of cytochrome b and D-loop sequences placed maximum parsimony, neighbor-joining, or maximum A. monodactylus and A. gayi within an otherwise likelihood topologies. These results suggest that some entirely Nemadactylus clade (Figs. 2 and 4). These distinction of N. bergi is warranted. results suggest that Nemadactylus and Acantholatris The species A. monodactylus, A. gayi, Nemadactylus are synonymous, with the name Nemadactylus having sp., N. macropterus, and N. bergi appear to be very priority. Although there are no molecular data for the closely related, and it is expected that sequence data third species of Acantholatris, A. vemae, it is most likely would place A. vemae close to these taxa as well. In that this species would cluster within Nemadactylus. A. contrast, N. douglasii and N. valenciennesi form a vemae is similar in morphology to A. monodactylus, and divergent sister clade to this group (Figs. 2 and 4). these species are sympatric at Vema Seamount (Pen- These results suggest that some systematic modifica- rith, 1967; Lamb, 1990). tion may be warranted in addition to synonymizing The suggestion of synonymy for Acantholatris with Acantholatris with Nemadactylus. The two divergent Nemadactylus is supported by morphological data. The distinction of Nemadactylus and Acantholatris is based groups could be distinguished by allocation to separate predominantly on the number of anal fin rays, with subgenera within the expanded Nemadactylus. How- Nemadactylus possessing 14–19 and Acantholatris hav- ever, the overall degree of genetic differentiation being 10–12 (Gill, 1862; Lamb, 1990). However, the tween these groups is low compared to that observed recently identified ‘‘king tarakihi’’ species, considered within other cirrhitoid genera (Burridge, unpubl. data). to be Nemadactylus because of its sympatry and almost The very close relationships between at least five identical appearance with N. macropterus, possesses Nemadactylus and Acantholatris species, and in some only 12 anal fin rays (Roberts, 1993). Therefore, conflict cases their questionable separate status, may be better exists between the expected placement of this new emphasized by reallocating these taxa, and probably A. species and the dominant characteristic used to sepa- vemae, as variants of a single species. A morphological rate Nemadactylus and Acantholatris. revision of Nemadactylus is being conducted by C. D. The analysis of D-loop sequences suggests that Nema- Roberts (Museum of New Zealand), which will hope- dactylus sp. is most closely related to A. gayi from Juan fully resolve the taxonomic questions outstanding. MOLECULAR PHYLOGENY OF Nemadactylus AND Acantholatris 105

Divergence Time and the east coast of South America and then along the Molecular clock calibrations derived from cyto- chain of islands and seamounts from Tristan da Cunha chrome b sequences in sharks suggest that the five and Gough Island in the South Atlantic, to Saint Paul most closely related species of Nemadactylus and Acan- and Amsterdam Island in the Indian Ocean. During the tholatris have radiated recently, within the last 0.9 or Pleistocene glaciations this current also flowed faster 2.6 My, when assuming 2.3% synonymous sequence and at slightly lower latitudes (CLIMAP project mem- divergence MyϪ1 (Martin et al., 1992) and 7.5 ϫ 10Ϫ9 bers, 1976; Howard and Prell, 1992), which would have substitutions third codon siteϪ1 yϪ1 (Cantatore et al., made dispersal by this mechanism easier. As the glacial 1994), respectively. Similar estimates of radiation time periods were approximately 10 times longer in duration are obtained when using calibrations derived from than the interglacials, it is the glacial conditions to which other mitochondrial protein coding genes in fish. These most species have had greater exposure (Pollock, 1990). are 0.6 My when assuming 3.3% synonymous diver- Westward dispersal of temperate fishes in the South- gence MyϪ1, derived from eight pairs of perciform ern Hemisphere is most likely to have been facilitated species cytochrome oxidase I sequences (Bermingham by the northern components of anticyclonic current et al., 1997), and 1.2 My when assuming 0.833% total gyres (Kensley, 1981; Collette and Parin, 1991; Pollock, divergence MyϪ1, derived from salmonid NADH dehy- 1993), but such movement of Nemadactylus and Acan- drogenase 3 sequences (McKay et al., 1996). A molecu- tholatris from Australia may have been limited. Recruit- lar clock calibration was not applied to Nemadactylus ment from southwestern Australia into the Southern and Acantholatris D-loop sequences, as there was evi- Indian Ocean Anticyclonic Gyre would have been im- dence for nonclock-like behavior among them. peded during interglacial periods by the south-flowing Estimates of divergence time derived from molecular Leeuwin current and its associated offshore eddies (Fig. clock calibrations must be treated cautiously due to the 6A; CLIMAP project members, 1976; Pollock, 1993). number of assumptions made (Rand, 1994). The calibra- Although recruitment into this gyre would have been tions applied above suggest that the five most closely easier during glacial periods, transport around south- related species of Nemadactylus and Acantholatris ern Africa at these times would have been hindered by have radiated recently, at least within the last 2.6 My the retroflective effects of stronger westerly winds and and possibly during only the last 0.6 My. A similar the shoaling of the Southern Madagascar Ridge (Fig. estimate of divergence time, 0.5 My, has been proposed 6B; CLIMAP project members, 1976; Pollock, 1993). for some members of the rock lobster genus Jasus, Therefore, only Indian Ocean locations may have been which overlap in distribution with Nemadactylus and colonized by westward movement from Australia. If Acantholatris (Ovenden et al., 1997). None of these larvae did pass into the South Atlantic and establish a divergence estimates predate the ages of the oldest population, perhaps at Vema Seamount, dispersal islands and seamounts occupied by each species (Miller, throughout this basin could have been facilitated by the 1964; McDougall and Ollier, 1982; Stuessy et al., 1984). corresponding anticyclonic gyre. However, any dispersal into the Pacific would have been against the Zoogeography prevailing currents and most difficult. The use of these The waters of Australia and New Zealand appear to anticyclonic gyres is also questionable, as it would represent the region from which Nemadactylus and require that larvae could survive the warm tempera- Acantholatris dispersed. Both of the main clades ob- tures at the northern extents of these systems. The served for these species have representatives in this temperatures at these low latitudes would have been region, and those taxa which occur elsewhere exhibited only slightly abated, if at all, during glaciations only limited genetic divergence from Nemadactylus sp. (CLIMAP project members, 1976; Prell et al., 1980). and N. macropterus. Dispersal from Australia or New Given the observed sister taxa relationship between Zealand may have proceeded in either an easterly or a Nemadactylus sp. and A. gayi, and the prevailing westerly direction and was presumably undertaken currents, it appears that dispersal eastward from Aus- during the 7- to 12-month pelagic larval stage of these tralia or New Zealand was responsible for the founding species (Annala, 1987; Andrew et al., 1995). of A. gayi in the southeastern Pacific. However, as Any dispersal in an easterly direction would most Nemadactylus sp. and A. gayi formed a monophyletic likely have been mediated by the West Wind Drift clade, it does not appear that Juan Fernańdez or Current. Dispersal by this current has been proposed Desventuradas populations acted as a source for any for several fish, including cheilodactylids (Eschmeyer movement into the South Atlantic. The north-flowing and Hureau, 1971; Briggs, 1974; Andrew et al., 1995), Humboldt current also prevails around these islands. and also a number of invertebrates (Fell, 1962; New- Therefore, the South Atlantic and Indian Ocean popula- man, 1979; Lutjeharms and Heydorn, 1981; Pollock, tions of Nemadactylus and Acantholatris were founded 1990). In a stepwise manner, the West Wind Drift could by a separate movement east from Australia or New have transported Nemadactylus and Acantholatris from Zealand, a movement west from Australia, or a combi- Australia or New Zealand to the southeastern Pacific nation of both. The phylogenetic information obtained 106 C. P. BURRIDGE

FIG. 6. Approximate oceanographic conditions at present (A) and during a Pleistocene glacial maxima (B). Major differences during glaciations comprise the breakdown of the Leeuwin current and its associated offshore eddies and the retroflective effects in the western Indian Ocean of stronger westerly winds and the shoaling of the Southern Madagascar Ridge. from this study is insufficient to suggest by which isolating mechanism. The widespread distributions of method the populations of N. bergi, A. monodactylus, the latrids Mendosoma lineatum and Latris lineata and A. vemae were founded. However, dispersal east suggest that a similar distribution was possible for from Australia and New Zealand is favored as it was ancestral species of Nemadactylus or Acantholatris, presumably faster and more direct. given the cognate dispersal capabilities of these fami- Although dispersal of Nemadactylus and Acanthola- lies (Andrew et al., 1995). tris from Australia or New Zealand probably proceeded in a stepwise manner, speciation may not have immedi- Polytypic Clades ately accompanied the founding of populations. Gene The presence of polytypic N. macropterus and A. flow between even the most isolated populations may monodactylus clades (Fig. 4) may be explained in terms have been sufficient to prevent speciation, and there- of introgressive hybridization or incomplete lineage fore it is possible that individual ancestral species had sorting. The morphological separation of these taxa widespread distributions in the temperate Southern (Lamb, 1990) does not favor an alternate suggestion Hemisphere. Speciation may have then resulted from a that they may be synonymous. change in dispersal capabilities or oceanographic condi- Although the spawning periods of N. macropterus tions which isolated populations. The transition to and A. monodactylus overlap (Annala, 1987; Andrew et decreased West Wind Drift flow when proceeding from al., 1995), there are no records of sympatry for these glacial to interglacial periods may represent such an two species or natural hybridization between any MOLECULAR PHYLOGENY OF Nemadactylus AND Acantholatris 107 cirrhitoids. Therefore, mitochondrial DNA sequence levels of nucleotide diversity observed for Nemadac- exchange between N. macropterus and A. monodacty- tylus sp. and A. gayi could also be explained in terms of lus appears unlikely. In addition, if introgressive hybrid- population size. ization did take place between these two species, it might be expected that a maximum of two polytypic clades would be observed (introgression in both direc- CONCLUSIONS tions). Analysis of D-loop sequences suggested that The molecular data obtained suggest that Acanthola- three such clades exist (Figs. 4 and 5), although the tris and Nemadactylus should be synonymized, with arbitrary distinction of two of these would be rejected if the latter having priority. Five of these species appear individuals of intermediate relatedness were identified. very closely related and have probably dispersed and Incomplete lineage sorting is the favored explanation radiated throughout the Southern Hemisphere within for the observed polytypic clades. During speciation, the last 0.6–2.6 My, facilitated by their long larval ancestral mitochondrial DNA lineages may not sort durations. Sorting of mitochondrial DNA lineages congruently with respect to species boundaries, and among two of these taxa appears incomplete. Further individuals can possess haplotypes more similar to studies are required to determine some of the dispersal those of nonconspecifics than individuals of the same directions and to resolve questions of specific status. species (Avise, 1986). That is, the gene tree for sister taxa may not be reciprocally monophyletic. The simula- tions of Neigel and Avise (1986) suggest that the ACKNOWLEDGMENTS probability of sister taxa possessing reciprocally monophyletic mitochondrial DNA lineages is only high after The following people are thanked for the provision of fish tissues: 4N generations of genetic isolation, where N is the Tim Andrew and Leslie Ter Morshuizen (J. L. B. Smith Institute of Ichthyology, Grahamstown, South Africa), Guillermo Canete (INIDEP, number of females. Assuming female numbers in ex- Mar del Plata, Argentina), Robert Rigoni (Port Lincoln, Australia), cess of 50,000 individuals and an average generation John Clarke (Smallcraft, Two Rocks, Australia), Patrick Coutin time of 3 years, the minimum divergence time sug- (MAFRI, Queenscliff, Australia), Guy Duhamel (Museúm National gested for N. macropterus and A. monodactylus is too d’Historie Naturelle, Paris, France), Nick Elliott and Peter Grewe recent for there to be a high probability of reciprocal (CSIRO Marine Research Laboratories, Hobart, Australia), Malcolm Francis and Peter Smith (NIWA, Wellington, New Zealand), Michelle lineage monophyly. The female numbers of these spe- van der Merwe (South African Museum, Cape Town, South Africa), cies undoubtably exceed 50,000, given their total an- and German Pequen˜ o (Universidad Austral de Chile, Valdivia, Chile). nual catches and average fish weights (Annala, 1987; Bronwyn Innes and Peter Grewe (CSIRO Marine Laboratories, Andrew et al., 1995; Jordan, 1997), and therefore even Hobart) are also thanked for their operation of an ABI 377 automated the more ancient divergence time estimates may be too sequencer. Drs. Alaistair Richardson, Jon Waters, and Robert White, plus an anonymous referee, made constructive criticisms of the recent for lineage monophyly to have developed. Al- manuscript. At the time of writing the author was supported by an though bottlenecks in effective population size will Australian Postgraduate Award, and the research was funded by a reduce the amount of time required to achieve lineage departmental allocation. monophyly, the high levels of nucleotide and haplotype diversity observed within these two species (see also Grewe et al., 1994) suggest that the number of females REFERENCES has been large and stable (Avise, 1989). Agnot, M. (1951). Observations sur la faune marine et la peˆche aux Despite its similarly recent divergence and large ˆlesı Saint-Paul et Amsterdam. Me´m. Inst. Sci. Madagascar 6: 1–51. female population size, lineage monophyly and a com- Allen, G. R., and Heemstra, P. C. (1976). Cheilodactylus rubrolabia- paratively small level nucleotide diversity were ob- tus, a new species of morwong (Pisces: Cheilodactylidae) from served for N. bergi in comparison with N. macropterus Western Australia, with a key to the cheilodactylid fishes of and A. monodactylus (Fig. 4 and Table 2). These could Australia. Rec. West. Aust. Mus. 4: 311–325. be the result of one or more bottlenecks in N. bergi Andrew, T. G., Hecht, T., Heemstra, P. C., and Lutjeharms, J. R. E. female population size, such as the dramatic decline in (1995). Fishes of the Tristan da Cunha group and Gough Island, South Atlantic Ocean. Ichthyol. Bull. J. L. B. Smith Inst. Ichthyol. the stocks of this species resulting from overfishing 63: 1–43. during the 1960’s (Cotrina, 1971). Similarly, this spe- Annala, J. H. (1987). The biology and fishery of tarakihi, Nemadac- cies may have been founded by only a small range of tylus macropterus, in New Zealand waters. N. Z. Fish. Res. Div. mitochondrial DNA lineages. Because of their linked Occ. Pub. 51: 1–12. inheritance, selection for particular coding sequences Avise, J. C. (1986). Mitochondrial DNA and the evolutionary genetics within the mitochondrial genome can also reduce the of higher animals. Phil. Trans. R. Soc. Lond. 312: 325–342. nucleotide diversity of D-loop sequences (Chenoweth et Avise, J. C. (1989). Gene trees and organismal histories: A phyloge- al., 1998). However, the degree of coding sequence netic approach to population biology. Evolution 43: 1192–1208. Bermingham, E., McCafferty, S. S., and Martin, A. P. (1997). Fish nonsynonymous variation within N. bergi would not be biogeography and molecular clocks: Perspectives from the Panama- large, and therefore is unlikely to have influenced nian Isthmus. In ‘‘Molecular Systematics of Fishes’’ (T. D. Kocher fitness. The lineage monophyly and similarly small and C. A. Stepien, Eds.), pp. 113–128. Academic Press, San Diego. 108 C. P. BURRIDGE

Brasher, D. J., Ovenden, J. R., and White, R. W. G. (1992). Mitochon- Howard, W. R., and Prell, W. L. (1992). Late Quaternary surface drial DNA variation and phylogenetic relationships of Jasus spp. circulation of the southern Indian Ocean and its relationship to (Decapoda: Palinuridae). J. Zool. Lond. 227: 1–16. orbital variations. Paleoceanography 7: 79–117. Briggs, J. C. (1974). ‘‘Marine Zoogeography,’’ McGraw–Hill, New Jordan, A. R. (1997). Demersal trawl surveys of the continental shelf York. of southern and eastern Tasmania 1993–1995. Mar. Res. Div. Dept. Cantatore, P., Roberti, M., Pesole, G., Ludovico, A., Milella, F., Prim. Ind. Fish. Tasmania Tech. Rep. 50: 1–129. Gadaleta, M. N., and Saccone, C. (1994). Evolutionary analysis of Kensley, B. (1981). On the zoogeography of Southern African decapod cytochrome b sequences in some perciformes: Evidence for a slower crustacea, with a distributional checklist. Smith. Contrib. Zool. rate of evolution than in mammals. J. Mol. Evol. 39: 589–597. 338: 1–64. Cavalli-Sforza, L. L., and Edwards, A. W. F. (1967). Phylogenetic Kimura, M. (1980). A simple method for estimating evolutionary rate analysis: Models and estimation procedures. Evolution 32: 550– of base substitutions through comparative studies of nucleotide 570. sequences. J. Mol. Evol. 16: 111–120. Chenoweth, S. F., Hughes, J. M., Keenan, C. P., and Lavery, S. (1998). Kishino, H., and Hasegawa, M. (1989). Evaluation of the maximum When oceans meet: A teleost shows secondary intergradation at an likelihood estimate of the evolutionary tree topologies from DNA Indian–Pacific interface. Proc. R. Soc. Lond. B 265: 415–420. sequence data, and the branching order in Hominoidea. J. Mol. CLIMAP Project Members (1976). The surface of the Ice-age Earth. Evol. 29: 170–179. Science 191: 1131–1137. Kluge, A. G., and Farris, J. S. (1969). Quantitative phyletics and the Collette, B. B., and Parin, N. V. (1991). Shallow-water fishes of evolution of anurans. Syst. Zool. 18: 1–32. Walters Shoals, Madagascar Ridge. Bull. Mar. Sci. 48: 1–22. Kocher, T. D., Thomas, W. K., Meyer, A., Edwards, S. V., Paä¨bo, S., Cotrina, C. P. (1971). La pesca de la castan˜ eta (Cheilodactylus bergi) Villablanca, F. X., and Wilson, A. C. (1989). Dynamics of mitochon- en el a´rea de Mar del Plata. Ana´lisis de las estadı´sticas de captura y drial DNA evolution in animals: Amplification and sequencing with muestreos (1967–1970). Inst. Biol. Mar. Mar del Plata Contrib. conserved primers. Proc. Natl. Acad. Sci. USA 86: 6196–6200. 163: 1–27. Lamb, T. D. (1990). ‘‘The Taxonomy and Phylogeny of Cheilodactylid Duhamel, G. (1989). Ichthyofaune des les Saint-Paul et Amsterdam and Latrid Fishes,’’ B.Sc.(Hons) thesis, Univ. of Tasmania, Hobart, (oceán Indien sud). Me´sogeé Bull. Mus. Hist. Nat. Marseille 49: Australia. 21–47. Lutjeharms, J. R., and Heydorn, A. E. (1981). The rock lobster Jasus Eschmeyer, W. N., and Hureau, J. C. (1971). Sebastes mouchezi, a tristani on Vema Seamount: Drifting buoys suggest a possible senior synonym of Helicolenus tristanensis, with comments on recruiting mechanism. Deep Sea Res. 28: 631–636. Sebastes capensis and zoogeographical considerations. Copeia 1971: Martin, A. P., Naylor, G. J. P., and Palumbi, S. R. (1992). Rates of 576–579. mitochondrial DNA evolution in sharks are slow compared with Farris, J. S. (1989). The retention index and the rescaled consistency mammals. Nature 357: 153–155. index. Cladistics 5: 417–419. McDougall, I., and Ollier, C. D. (1982). Potassium–argon ages for Fell, H. B. (1962). West-Wind-Drift dispersal of echinoderms in the Tristan da Cunha, South Atlantic. Geol. Mag. 119: 87–93. Southern Hemisphere. Nature 193: 759–761. McKay, S. J., Devlin, R. H., and Smith, M. J. (1996). Phylogeny of Felsenstein, J. (1981). Evolutionary trees from DNA sequences: A pacific salmon and trout based on growth hormone type-2 and maximum likelihood approach. J. Mol. Evol. 17: 368–376. mitochondrial NADH dehydrogenase subunit 3 DNA sequences. Felsenstein, J. (1985a). Confidence limits on phylogenies: An ap- Can. J. Fish. Aquat. Sci. 53: 1165–1176. proach using the bootstrap. Evolution 39: 783–791. Meleńdez, R. C., and Villalba, S. C. (1992). New records and biological Felsenstein, J. (1985b). Confidence limits on phylogenies with a data for the icthyofauna from Juan Fernańdez archipelago, Chile. molecular clock. Syst. Zool. 34: 152–161. Estud. Ocean. 11: 3–29. Felsenstein, J. (1993). PHYLIP (Phylogeny Inference Package) ver- Meyer, A. (1994). DNA technology and the phylogeny of fish. In sion 3.5c. Distributed by the author. Department of Genetics, ‘‘Genetics and the Evolution of Aquatic Organisms’’ (A. R. Beau- University of Washington, Seattle. mont, Ed.), pp. 219–249. Chapman & Hall, London. George, R. W. (1997). Tectonic plate movements and the evolution of Miller, J. A. (1964). Appendix II. Age determinations made on Jasus and Panulirus spiny lobsters (Palinuridae). Mar. Fresh. Res. samples of basalt from the Tristan da Cunha group and other parts 48: 1121–1130. of the mid-Atlantic ridge. In ‘‘The Vulcanological Report of the Gill, T. H. (1862). Synopsis of the family of cirrhitoids. Proc. Acad. Royal Society Expedition to Tristan da Cunha, 1962’’ (P. E. Baker, Nat. Sci. Philadelphia 14: 102–124. I. G. Gass, P. G. Harris, and R. W. Le Maitre, Eds.), pp. 565–569. Greenwood, P. H. (1995). A revised familial classification for certain Royal Society of London. cirrhitoid genera (Teleostei, Percoidei, Cirrhitoidea), with com- Nei, M. (1987). ‘‘Molecular Evolutionary Genetics,’’ Columbia Univ. ments on the group’s monophyly and taxonomic ranking. Bull. Nat. Press, New York. Hist. Mus. Lond. 61: 1–10. Neigel, J. E., and Avise, J. C. (1986). Phylogenetic relationships of Grewe, P. M., Smolenski, A. J., and Ward, R. D. (1994). Mitochondrial mitochondrial DNA under various demographic models of specia- DNA diversity in jackass morwong (Nemadactylus macropterus, tion. In ‘‘Evolutionary Processes and Theory’’ (E. Nevo and S. Teleostei) from Australian and New Zealand waters. Can. J. Fish. Karlin, Eds.), pp. 515–534. Academic Press, New York. Aquat. Sci. 51: 1101–1109. Newman, W. A. (1979). On the biogeography of balanomorph bar- Hendy, M. D., and Penny, D. (1982). Branch and bound algorithms to nacles of the Southern Ocean including new balanid taxa; a determine minimal evolution trees. Math. Biosci. 59: 277–290. subfamily, two genera and three species. In ‘‘Proceedings of the Hillis, D. M., and Huelsenbeck, J. P. (1992). Signal, noise, and International Symposium on Marine Biogeography and Evolution reliability in molecular phylogenetic analyses. J. Hered. 83: 189– in the Southern Hemisphere,’’ pp. 279–306. New Zealand Depart- 195. ment of Scientific and Industrial Research, Wellington. Hillis, D. M., Larson, A., Davis, S. K., and Kimmer, E. A. (1990). Norman, J. (1937). Coast fishes part II. The Patagonia region. Disc. Nucleic acids III: Sequencing. In ‘‘Molecular Systematics’’ (D. M. Rep. 16: 1–150. Hillis and C. Moritz, Eds.), pp. 318–370. Sinauer, Sunderland, MA. Ovenden, J. R., Booth, J. D., and Smolenski, A. J. (1997). Mitochon- MOLECULAR PHYLOGENY OF Nemadactylus AND Acantholatris 109

drial DNA phylogeny of red and green rock lobsters (genus Jasus). (1997). Enhanced DNA extraction and PCR amplification of mito- Mar. Fresh. Res. 48: 1131–1136. chondrial genes from formalin-fixed museum specimens. BioTech- Paä¨bo, S. (1990). Amplifying ancient DNA. In ‘‘PCR Protocols: A niques 22: 394–400. Guide to Methods and Applications’’ (M. A. Innes, D. H. Gelfand, Smith, M. M. (1980). A review of the South African cheilodactylid J. J. Sninski, and T. J. White, Eds.), pp. 159–166. Academic Press, fishes (Pisces: Perciformes), with a description of two new species. San Diego. Ichthyol. Bull. J. L. B. Smith Inst. Ichthyol. 42: 1–14. Penrith, M. J. (1967). The fishes of Tristan da Cunha, Gough Island, Smith, P. J., Roberts, C. D., McVeagh, S. M., and Benson, P. G. (1996). and the Vema Seamount. Ann. S. Afr. Mus. 48: 523–548. Genetic evidence for two species of tarakihi (Teleostei: Cheilodac- Pollock, D. E. (1990). Palaeoceanography and speciation in the spiny tylidae: Nemadactylus) in New Zealand waters. N. Z. J. Mar. Fresh. lobster genus Jasus. Bull. Mar. Sci. 46: 387–405. Res. 30: 209–220. Pollock, D. E. (1992). Palaeoceanography and speciation in the spiny Stuessy, T. F., Foland, K. A., Sutter, J. F., Sanders, R. W., and Silva, lobster genus Panulirus in the Indo-Pacific. Bull. Mar. Sci. 51: O. M. (1984). Botanical and geological significance of potassium– 135–146. argon dates from the Juan Fernańdez islands. Science 225: 49–50. Pollock, D. E. (1993). Speciation in spiny lobsters—Clues to climati- Swofford, D. L. (1993). PAUP: phylogenetic analysis using parsimony, cally-induced changes in ocean circulation patterns. Bull. Mar. Sci. 53: 937–944. version 3.1. Computer program distributed by the author, Illinois Natural History Survey, Champaign, Illinios. Prell, W. L., Hutson, W. H., Williams, D. F., Be´, A. W. H., Geitzenauer, K., and Molfino, B. (1980). Surface circulation of the Indian Ocean Templeton, A. R. (1983). Phylogenetic inference from restriction during the last glacial maximum, approximately 18,000 yr B.P. endonuclease cleavage site maps with particular reference to the Quat. Res. 14: 309–336. evolution of humans and the apes. Evolution 37: 221–244. Rand, D. M. (1994). Thermal habit, metabolic rate and the evolution Thompson, J. D., Higgins, D. G., and Gibson, T. J. (1994). CLUSTALW: of mitochondrial DNA. Trends Ecol. Evol. 9: 125–131. Improving the sensitivity of progressive multiple sequence align- Roberts, C. D. (1993). New fishes from the Kermadec Ridge. Seafood ment through sequence weighting, position specific gap penalties N. Z. 4: 102–103. and weight matrix choice. Nucleic Acids Res. 22: 4673–4680. Rozas, J., and Rozas, R. (1997). DnaSP version 2.0: a novel software Wilson, R. R. J., and Kaufmann, R. S. (1987). Seamount biota and package for extensive molecular population genetic analysis. Com- biogeography. In ‘‘Seamounts, Islands, and Atolls’’ (B. H. Keating, put. Applic. Biosci. 13: 307–311. P. Fryer, R. Batiza, and G. W. Boehlert, Eds.), pp. 355–377. Am. Saitou, N., and Nei, M. (1987). The neighbour-joining method: A new Geophys. Union Geophys. Monograph No. 43. method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4: Wo¨hler, O. C., and Sańchez, F. (1994). Feeding ecology of castan˜eta 406–425. (Cheilodactylus bergi; Pisces: Cheilodactylidae) in the south- Shedlock, A. M., Haygood, M. G., Pietsch, T. W., and Bentzen, P. western Atlantic (34–47°S). Aust. J. Mar. Fresh. Res. 45: 507–520.