Phylogeographic patterns and cryptic speciation across ...

doi: 10.1111/j.1420-9101.2011.02382.x

Phylogeographic patterns and cryptic speciation across oceanographic barriers in South African intertidal fishes S. V O N D E R HEYDEN*, R. C. K. BOWIE , K. PROCHAZKAà, P. BLOOMER§, N. L. CRANE– & G. BERNARDI** *Evolutionary Genomics Group, Department of Botany and Zoology, Stellenbosch University, Matieland, South Africa Museum of Vertebrate Zoology, Department of Integrative Biology, University of California, Berkeley, CA, USA àDepartment of Agriculture, Forestry, and Fisheries, Branch: Fisheries, Cape Town, South Africa §Department of Genetics, Molecular Ecology and Evolution Program, University of Pretoria, Pretoria, South Africa –Department of Biology, Cabrillo College, Aptos, CA, USA **Department of Ecology and Evolutionary Biology, University of California Santa Cruz, Santa Cruz, CA, USA

Keywords:

Abstract

Cape Agulhas; Cape of Good Hope; Clinus superciliosus; Muraenoclinus dorsalis; phylogeography; South Africa; speciation.

Biogeographic boundaries are the meeting zone of broadly distributed faunas, or the actual cause of a faunal break. In the latter case, closely related sister species should be found across such a boundary. To achieve such a situation, preliminary stages are expected, where phylogeographic breaks followed by genetic cryptic speciation would be observed. Biogeographic boundaries, in the Cape Point ⁄ Cape Agulhas region of southern Africa, offer an ideal system to test such predictions. Here, we studied two intertidal clinid fish species that are endemic to southern Africa, Clinus superciliosus (n = 127) and Muraenoclinus dorsalis (n = 114). Using mitochondrial control region, 16S rRNA, 12S rRNA and NADH2 genes and the nuclear rhodopsin and the first intron of the S7 ribosomal protein gene, we show both phylogeographic breaks and likely cryptic speciation in each species. Pairwise Fst results suggest population genetic structuring for both species, with higher levels for M. dorsalis (Fst = 0.34–0.93) than for C. superciliosus (Fst = 0.1–0.74). Further, we recover two and three distinct lineages within M. dorsalis and C. superciliosus, respectively. Phylogenetic topologies, concordance between nuclear and mitochondrial markers and levels of sequence divergence, which are consistent with closely related sister species pairs, suggest the presence of cryptic species. Our results therefore meet the expectation for reduced gene flow at a biogeographic barrier, which translates into significant genetic breaks and cryptic sister species.

Introduction Biogeographic boundaries that delineate the meeting of geographic ranges of species assemblages result from different mechanisms. Most frequently, biogeographic boundaries correspond to the contact zone of otherwise broadly distributed species (Wallace, 1898; Briggs, 1974; Awad et al., 2002; Irwin, 2002). In some instances Correspondence: Giacomo Bernardi, Department of Ecology and Evolutionary Biology, University of California Santa Cruz, 100 Shaffer Road, Santa Cruz, CA 95060, USA. Tel.: +831 459 5124; fax: +831 459 3383; e-mail: [email protected]

however, the boundary itself has been suggested to cause the distribution disjunction. In this case, populations that are formerly continuous are separated by extrinsic factors such as geographic or geological events, known as vicariant events, which reduce gene flow across boundaries. Interpopulation genetic divergence across the boundary is then predicted. This situation eventually results in what is usually referred to as a phylogeographic break (Avise, 2004; Lima et al., 2005; Taylor & Hellberg, 2005) and often such boundaries act on multiple species that may have very different life histories (Dawson et al., 2002; von der Heyden, 2009; Teske et al., 2011). If this situation is not disrupted by

ª 2011 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 2505–2519 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2011 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY

2505

2506

S.

VON DER

HEYDEN ET AL.

secondary contact, speciation later occurs across the boundary, which then becomes a bona fide biogeographic boundary (Burton, 1998). Such a theoretical prediction is testable, because in this scenario, the cause of the disruption of gene flow produces species assemblages that exhibit a large number of sister species across the boundary (Burton, 1998; Heads, 2005). Interestingly, the intermediate situation between the beginning of the process, the phylogeographic break, and the end result, complete speciation across the boundary, has not been assessed. In the absence of significant shifts in the selection regime across the newly formed biogeographic boundary that could result in morphological change, the intermediate situation should reveal the presence of entities across the boundary that would be genetically different from each other, but morphologically indistinguishable, and these are referred to as genetic cryptic species (e.g. Knowlton, 1993; Edkins et al., 2007; Vrijenhoek, 2009). To determine whether the biogeographic transition in the Cape of Good Hope region of southern Africa follows such biogeographic models, we decided to focus our attention on a group of endemic intertidal organisms, the klipfishes (family Clinidae). Clinids are small benthocryptic fishes that are found in intertidal and shallow subtidal habitats in temperate areas of both hemispheres. All species found in southern Africa are endemic and comprise around 50% of all clinid species globally.

Whereas some clinids are egg layers with some exhibiting parental care (e.g. Coyer, 1982), those from southern Africa, as well as Australia and New Zealand, exhibit internal fertilization and brood their young. Southern African clinid species show a process called ‘superembryonation’, where females gestate several broods of larvae at different developmental stages simultaneously (Prochazka, 1994). It is likely that live-bearing is an adaptation employed by intertidal fishes for retaining eggs and larvae in the near-shore zone (see Gibson, 1982 for review) and for allowing reproduction throughout the year. In a study of the reproductive biology of the South African clinids, Prochazka (1994) demonstrated that larval recruits to tide pools were only marginally larger than the largest embryos found, suggesting a very short period between parturition and recruitment and a consequently highly restricted dispersal window. While life-history characteristics are not always a good predictor of gene flow levels (e.g. Ayre et al., 2009; Shanks, 2009), realized dispersal is likely to be reduced in clinids, and as such, they are ideal candidates to study gene flow across biogeographic boundaries. Oceanographic features of southern Africa are complex. To the west, the southern Atlantic Ocean is influenced by the cold, northwards flowing Benguela Current. In contrast, the eastern shores of southern Africa are influenced by the warm, southward moving Agulhas current (Fig. 1). The oceanographic features of the south

Fig. 1 Map of southern Africa with an emphasis on the major oceanic currents that concern the region studied here. The cold Benguela current flows northward along the west coast, whereas the warm Agulhas current flows southward along the east coast. Currents meet in a transition zone between Cape Point and Point Agulhas.


Phylogeographic patterns and cryptic speciation

coast are dynamic, and it is thus impossible to locate an ‘exact’ point at which the Benguela and Agulhas Currents meet. The widening of the Agulhas Bank causes the Agulhas Current to move away from the coast, which causes the current to retroflect (Lutjeharms, 2006), whereas some eddies (called Agulhas rings) continue westwards and mix with the cooler Benguela waters. This makes the area between Cape of Good Hope and Cape Agulhas an overall area of contact (Fig. 1). The differing current regimes have a profound effect on the coastal regions; at least three main biogeographic regions have been described for the South African coast: the ‘cooltemperate Namaqua Province’, the ‘warm-temperate south coast Agulhas province’ and the ‘subtropical east coast province’ (Griffiths et al., 2010). Most of the clinid diversity is restricted to the cooler waters of the Namaqua and Agulhas provinces, but some do extend their range into the warmer waters of the east coast (Smith, 1995), although they do not reach the levels of abundance of the western and south-western coasts (Prochazka & Griffiths, 1992). Several molecular studies of South African marine species have identified a few localities that appear to reduce gene flow between adjacent populations. These include Cape Point, Cape Agulhas, Algoa Bay, along the Wild coast and at the border between northern South Africa and Mozambique (von der Heyden, 2009; Teske et al., 2011). However, not all taxa studied show concordant genetic breaks and it is not clear whether life history and population genetic structuring are closely linked. Most studies confirm the broad classification of the shores into the three major biogeographic areas. Genetic breaks also reflect historical, rather than contemporary factors and show recent population divergence (60 000 years ago; von der Heyden et al., 2008), as well as dividing distinct species (Edkins et al., 2007). Interestingly, gene flow patterns and life history appear to be correlated. For C. cottoides, a clinid fish with likely reduced dispersal capacity, gene flow utilized an in-shore counter current, whereas for the goby, Caffrogobius caffer, which might have an extended planktonic dispersal phase, asymmetric gene flow with the Agulhas Current has been observed (von der Heyden et al., 2008; Neethling et al., 2008). In this study, we focused on two clinid species, Clinus superciliosus and Muraenoclinus dorsalis. These species are very abundant and broadly occur in the intertidal across the boundaries mentioned above. The super klipfish C. superciliosus is mostly found in the intertidal (and large adults of up to 30 cm in the shallow subtidal) from Namibia (where it is not very abundant) to the Kei River in South Africa (Fig. 1). The nosestripe klipfish M. dorsalis has a more restricted vertical distribution on rocky shores and is often found in the mid- to high intertidal area. Remarkably, M. dorsalis can withstand emersion from water, and individuals and groups of fishes have been observed sheltering among weeds or under boulders

2507

outside of rocky pools. It has a similar geographic range to C. superciliosus and has been found from southern Namibia to southern KwaZulu-Natal, although it is not abundant at the edges of the distribution (S. von der Hyden, pers. obs., Fig. 1). Given the smaller size of M. dorsalis and its closer association with the intertidal zone, this species is probably much more limited in movement than C. superciliosus, and it is therefore likely that these species will exhibit different levels of population genetic structuring and that biogeographic barriers will act on both species differently. Molecular genetic approaches have proven powerful in testing alternative biogeographic scenarios (Heads, 2005; Riddle et al., 2008). The main goal of our study was to test the hypothesis that biogeographic boundaries that are present along the southern African coast have played a role in (i) restricting gene flow and therefore coincide with the presence of strong phylogeographic breaks and (ii) the presence of genetic sister species across such boundaries. Such hypotheses are supported by recent molecular work that has shown genetic structure for a number of marine species studied, including broad-cast spawners (that theoretically have fewer limitations to dispersal than live-bearing clinids; von der Heyden, 2009; Teske et al., 2011). Further, cryptic speciation with sister lineages in adjacent biogeographic boundaries has also been shown (Edkins et al., 2007). In this study, we used nuclear and mitochondrial molecular markers on a broad geographic set of samples of our two focal species. To evaluate our results both intra- and interspecifically, we analysed the data using both population genetic and phylogenetic approaches by adding several other clinid species to the analysis.

Materials and methods Collections and DNA samples Besides the two target species, which were collected throughout most of their distribution range in South Africa and Namibia (Table 1, Fig. 2a,b), we also needed additional species against which the divergence of cryptic species could be compared. For that purpose, an additional six South African clinids were analysed. These included four Clinus species, C. acuminatus, C. cottoides, C. heterodon and C. venustris, and two other clinids, Blennophis anguillaris and Cancelloxus longior. Two North American clinids were used as outgroups, the giant kelpfish Heterostichus rostratus and the crevice kelpfish Gibbonsia elegans (Table 1). After collection, samples were immediately placed in 95% ethanol and stored at ambient temperature in the field and then at 4 C in the laboratory. Muscle or liver tissue was later dissected from these samples. Total genomic DNA was prepared from 75 to 150 mg of muscle or liver tissue by proteinase K digestion in lysis buffer (10 mM Tris, 400 mM NaCl, 2 mM EDTA and 1% SDS)


2508

S.

VON DER

HEYDEN ET AL.

Table 1 Sampled species, numbers and abbreviations of individuals used in this study.

Table 1 (Continued ). Species (population number)

Species (population number) Focal species Clinus superciliosus Clade 1 1 2 3 4 5 6 7 8 9 10 11

12 13

Locality

Abbreviation

Swakopmund Paternoster Silverstroomstrand Mouille Point Sea Point Slangkoppunt Cape of Good Hope Miller’s Point Long Beach Betty’s Bay Gansbaai Agulhas De Hoop Knysna Tsistikamma Port Alfred Gonubie Haga Haga

SWA PAT SIL MOP SPO SLA CGH MIP LON BET GAN AGU DHO KNY TSI PAL GON HAG

4 5 5 1 16 14 16 6 1 10 10 5 3 2 2 8 3 2

MOP

3 3

MIP

1 2

Oudekraal Betty’s Bay De Hoop Tsistikamma

OUD BET DHO TSI

1 1 1 2

Mouille Point Gansbaai

MOP GAN

14 6

Tieties Bay Mouille Point Sea Point

TIE MOP SPO

18 11 5

Aasbank Betty’s Bay Gansbaai Agulhas Herold’s Bay Port Alfred Haga Haga

AAS BET GAN AGU HER PAL HAG

2 16 7 12 3 10 10

MOP

2

MIP

2

MOP

2

Clade 2 Mouille Point Red morph Green morph Miller’s Point Red morph Subtidal mottled Clade 3

Muraenoclinus dorsalis Clade 1

Clade 2a 1 2 3 Clade 2b 4 5 6 7 8 9 10 Other species South Africa Blennophis anguillaris Mouille Point Cancelloxius longior Miller’s Point Clinus acuminatus Mouille Point

Locality

Abbreviation

No.

MOP

2

SPO

1

RB2543

1

GUA RLL

1 1

MIB

2

No. Clinus cottoides Mouille Point Clinus heterodon Sea Point Clinus venustris North America Gibbonsia elegans Guadalupe Island, Mexico Santa Rosalillita, Mexico Heterostichus rostratus Mission Bay, San Diego

overnight at 55 C. This was followed by purification using isopropanol ⁄ chloroform extractions and alcohol precipitation (Sambrook et al., 1989). PCR amplification and sequencing Two sets of data were used in this study. The first set of data was used to fully assess the population structure of our two target species, C. superciliosus and M. dorsalis, by sequencing two mitochondrial markers, the noncoding hypervariable 5¢end of the control region and the protein encoding NADH dehydrogenase subunit 2 (ND2) gene, and one nuclear marker, the rhodopsin gene. Amplification of the mitochondrial loci was carried out using the universal fish primers CR-A and CR-E (Lee et al., 1995) for the control region and the primers C-ND2F and C-ND2R (von der Heyden et al., 2008) for ND2. In both cases, PCR amplifications were carried out with an annealing temperature of 54 C. Amplification of the nuclear rhodopsin gene used the nested amplification protocols of Sevilla et al. (2007) with RHO30F and RHO 319R for the first set of primers and Rho F2x and RhoR4n for the second set of primers. A second data set aimed at placing the population data obtained with the first data set in a broader phylogenetic context. In this case, loci with typically slower substitution rates were used. Two mitochondrial loci, the 12S and 16S rRNAs, and two nuclear loci, the first intron of the S7 ribosomal protein and the rhodopsin gene (with the same primers and protocols as mentioned earlier), were chosen. Universal primers 12SAL and 12SAR and 16SAL and 16SBH were used for the mitochondrial loci (Palumbi et al., 1991). The primers S7RPEX1F and S7RPEX2R (Chow & Hazama, 1998) were used for the S7 locus, with an annealing temperature of 56 C. After purification following the manufacturer’s protocol (ABI, Perkin-Elmer), sequencing was carried out in both directions with the primers used in the PCR amplification on an ABI 3100 automated DNA sequencer (Applied Biosystems, Foster



2509

Fig. 2 (a) Sampling sites of Clinus superciliosus. The region between Paternoster and Tsitsikamma was enlarged for better representation. Geographic distribution of Clade 1, Clade 2 and Clade 3 is represented as solid arcs. (b) Sampling sites for Muraenoclinus dorsalis. Geographic distribution of Clade 1 (with subclades 1a and 1b) and Clade 2 is represented as solid arcs.

City, CA, USA). In the case of nuclear loci, heterozygous individuals were found to be very rare, and when present, the allelic phase was easily resolved, as we did not find individuals with more than one heterozygotic site. These individuals were scored as two alleles, A and B, with the alternate base for each allele.

Phylogenetic analyses We used the computer program MAFFT (Katoh et al., 2002) implemented by the Geneious software package (Drummond et al., 2011) to align the DNA sequences. Phylogenetic relationships were assessed by


2510

S.

VON DER

HEYDEN ET AL.

maximum-likelihood (ML, using GARLI, Zwickl, 2006), maximum parsimony and neighbour-joining (MP and NJ, using PAUP*, Swofford, 2003) methods. For ML topologies, we conducted 10 independent runs in GARLI, using default settings and the automated stopping criterion, terminating the search when the ln score remained constant for 20 000 consecutive generations. The best likelihood of those runs was retained. MP searches included 100 random addition replicates and TBR branch swapping with the Multrees option in effect. NJ reconstructions used distances based on substitution models obtained with Modeltest (Posada & Crandall, 1998). Statistical confidence in nodes was evaluated using 2000 nonparametric bootstrap replicates (Felsenstein, 1985) for MP and NJ searches and 100 replicates for ML searches in GARLI, using the automated stopping criterion set at 10 000 generations. Topological differences were tested using a Shimodaira and Hasegawa test (Shimodaira & Hasegawa, 1999) implemented in PAUP*, based on resampling of estimated log-likelihoods tests (RELL, 1000 replicates). When topological differences were not found to be statistically significant, data sets were combined in the subsequent analyses. Genetic divergence, gene flow and population structure Phylogenetic relationships were compared without molecular clock enforcements, and topological differences were tested using a Shimodaira and Hasegawa test (Shimodaira & Hasegawa, 1999) implemented in PAUP* (4.0b10, Swofford, 2003). Genetic divergence was estimated using distances based on substitution models obtained with Modeltest. To account for polymorphism in each species (or population), divergence was estimated as the average pairwise distance between species (or population) minus the average pairwise distance within a species (or population). Gene flow (Fst and Nem) was estimated using A R L E Q U I N (version 3.11; Excoffier et al., 2005). The measure of Fst (and relatives) has been shown to being vulnerable to large within-population variability, and therefore, the use of a corrected standardized measure of Fst (called F’st) has been suggested (Hedrick, 2005; Meirmans, 2006; Jost, 2008) (but see Ryman & Leimar, 2009; Heller & Siegismund, 2009; Jost, 2009; for discussion). Such standardization, however, applies mostly to situations where within-population variability is very high (particularly in the case of microsatellites) compared to between-population variability, which was not the case here. We therefore present classical Fst so as to allow for easier comparison with other studies. Population structure was estimated by an analysis of molecular variance (A M O V A ; Excoffier et al., 1992) using ARLEQUIN. Populations were grouped in different regions and alternative groupings were tested with an A M O V A to find, without a priori, whether and how data defined

western and eastern regions. Corrections for simultaneous multiple comparisons were applied using the sequential Bonferroni correction (Rice, 1989). To better visualize evolutionary relationships among haplotypes, MP networks were constructed in TCS (Clement et al., 2000). Directional gene flow was further analysed using a coalescent approach. Migration rates were estimated using the software M I G R A T E 1.7.3 (Beerli, 2003), which is a ML estimator based on the coalescent theory. It uses a Markov chain Monte Carlo approach to investigate possible genealogies with migration events. Analysis of each data set was carried out with 10 short Monte Carlo chains of 4000 steps each and five long chains of length 20 000, with a sampling increment of 20. For each estimate, 10 replicates were used to generate a mean value of migration (Nm) and its associated standard deviation.

Results Sequences The mitochondrial control region and ND2 gene were sequenced for 127 Clinus superciliosus and 114 Muraenoclinus dorsalis (Table 1). Among C. superciliosus individuals, the aligned control region sequences were 429 base pairs (bp) long (173 bp variable, 131 bp informative) and showed eight indels that corresponded to a total of 12 bp. The entire ND2 region was recovered; it was 1048 bp long (184 bp variable, 172 bp informative). For M. dorsalis, the control region was 447 bp long (313 bp variable, 88 bp informative) and showed three indels that corresponded to a total of 5 bp. The ND2 region was 1048 bp long (233 bp variable, 222 bp informative). The nuclear-encoded rhodopsin gene was sequenced for 120 Clinus superciliosus and 55 Muraenoclinus dorsalis. Aligned rhodopsin sequences were 430 bp long. For C. superciliosus, 16 bp were variable and 13 bp were informative; for M. dorsalis, 5 bp were variable and 2 bp were informative. For the general phylogenetic analysis of different clinid species, the aligned 12S rRNA and 16S rRNA were 359 bp (107 bp variable, 81 bp informative) and 628 bp (160 bp variable, 143 bp informative) long, respectively. In the 16S rRNA region, three indels, corresponding to a total of 115 bp, were difficult to align and were deleted from the analysis. The first intron of the nuclear S7 ribosomal protein was 671 bp long (256 bp variable, 224 bp informative); the rhodopsin gene was 430 bp long (73 bp variable, 65 bp informative). All sequences have been deposited in GenBank (accession numbers: HQ158800–HQ159391, JN122427–JN122620). Intraspecific phylogenies Intraspecific phylogenetic relationships based on the two mitochondrial loci (control region and ND2) were not



2511

Fig. 3 (a) Phylogenetic relationships of Clinus superciliosus based on combined control region and ND2 sequences, as well as for nuclear rhodopsin sequences. Maximum-likelihood, maximum parsimony and neighbour-joining methods were used and resulted in similar topologies; maximum likelihood is presented here. Inset box presents acronyms that correspond to sampling sites (Fig. 1). Bootstrap values and statistical support are presented next to the nodes. (b) Phylogenetic relationships of Muraenoclinus dorsalis based on combined control region and ND2 sequences, as well as for nuclear rhodopsin sequences. Maximum-likelihood, maximum parsimony and neighbour-joining methods were used and resulted in similar topologies; maximum likelihood is presented here. Inset box presents acronyms that correspond to sampling sites (Fig. 1). Bootstrap values and statistical support are presented next to the nodes.

found to be significantly different (Shimodaira–Hasegawa test, P = 0.25); thus, the two loci data sets were combined and analysed together. In addition, all three search methods (ML, MP and NJ) resulted in topologies that were statistically equivalent (SH test, P = 0.5) (Fig. 3a,b). Clinus superciliosus individuals partitioned into three highly divergent clades (average pairwise combined sequence divergence, 13.0%, Fig. 3a). Clade 1 comprised the vast majority of the samples (114 samples). Samples

from this clade were collected throughout the range of the species. This clade is discussed in more detail in the Population Structure section. Clade 2 included nine individuals that displayed the rare colour morphs red or green. These individuals were collected in two nearby locales, Miller’s Point and Mouille Point. Clade 3 included five individuals that were collected in different locales and did not seem to be morphologically distinguishable from other sampled individuals (Fig. 3a, Table 1).


2512

S.

VON DER

HEYDEN ET AL.

Fig. 3 Continued.

Muraenoclinus dorsalis partitioned into two, highly divergent (20.3%) clades that, similar to C. superciliosus, partitioned geographically (Fig. 3b). Within these clades, each sampling locality of M. dorsalis contained haplotypes almost completely restricted to that locality. Clade 1 comprised two disjunct subclades (labelled 1a and 1b) that were also geographically partitioned (Figs 2b and 3b). Clade 1a included samples from the west coast, and Clade 1b included samples from the central southern coast, as well as the east coast. Clade 2 included samples from the centre of the distribution of the species. Although no sites were shared between the two subclades 1a and 1b, some sampling sites had representatives from either of the two subclades as well as from Clade 2. Such sites were therefore a mixture of Clade 1 and Clade 2 individuals (Fig. 3b). Genetic variation was lower for nuclear markers than for mitochondrial markers, as expected. Intraspecific relationships based on the nuclear marker (rhodopsin

gene) mirrored the results based on mitochondrial markers. Clinus superciliosus individuals partitioned in the same three previously identified mitochondrial clades. Variability of the rhodopsin marker was very low for Muraenoclinus dorsalis, where only two fixed differences were present. Those fixed differences separated individuals that belonged to the two mitochondrial clades, but variation was not sufficient to further separate individuals that belong to the mitochondrial subclades 1a and 1b (Fig. 3b). Population structure Due to the higher resolving power of the mitochondrial markers, population structure analyses were carried out on those markers. As described earlier, phylogenetic analyses showed that C. superciliosus and M. dorsalis partitioned into three or two clades, respectively. When the major clade for each species was analysed (called



Clade 1 in both species), population structure was found to be variable and gene flow was shown to be very limited between populations, with averages of Fst = 0.32, Nm = 1.04 and Fst = 0.62, Nm = 0.30, for C. superciliosus and M. dorsalis, respectively (Table 2a,b). This can clearly be seen in the MP network, which shows limited mixing of haplotypes among geographic areas for both species (Fig. 4). In addition, an analysis of molecular variance (A M O V A ) was performed on populations from both major clades. This analysis considered groups by sequentially adding one population at a time, from west to east, thus shifting a boundary between western and eastern groups by increments of one population at a time, to identify genetic breaks without imposing a priori boundaries. If a genetic break is present, a maximum level of Phict is expected. In the case of C. superciliosus, maximum values were obtained between groups 11 (De Hoop, Knysna and Tsitsikamma) and 12 (Port Alfred), where almost 52% of the total variance could be explained by this genetic break (Table 3). For M. dorsalis, maximum values were obtained between groups 3 and 4 (Sea Point and Betty’s Bay), where 39% of the total variance could be explained. This was expected because this break corresponded to the separation between Clades 1a and 1b. A further analysis of the largest of the two M. dorsalis subclades, Clade 1b, showed that the highest value of

2513

Phict was obtained between groups 5 and 6 (Betty’s Bay and Gansbaai, Table 3), which explained 55.8% of the variance. Across these genetic breaks, gene flow was found to be biased towards an eastward direction for both C. superciliosus and M. dorsalis. In the case of C. superciliosus, we found an average of 39.1 (± 1.82) eastward migrants per generation, as opposed to 11.5 (± 0.6) westward migrants per generation. Similarly, for M. dorsalis, there were 0.64 (± 0.16) eastward migrants per generation against 0.03 (± 0.06) that migrated westward. Interspecific phylogenies To place the intraspecific results presented above in a more general context, we established a phylogeny that included other South African clinid species and North American clinid outgroups. Phylogenetic relationships based on the two mitochondrial loci (12S rRNA and 16S rRNA) and the two nuclear loci (S7 and rhodopsin) were not found to be significantly different (P > 0.1); thus, the data sets were combined. In addition, all three search methods (ML, MP and NJ) resulted in topologies that were statistically equivalent (SH test, P > 0.2, Fig. 5). Representatives of the genus Clinus did not form a natural group (Fig. 5). As the purpose of this work was

Table 2 Population structure in (a) Clinus superciliosus; (b) Muraenoclinus dorsalis. (a) SWA SWA PAT SIL SPO SLA CGH LON BET GAN AGU DHO PAL GON

0.15 0.34 0.41 0.35 0.29 0.28 0.37 0.51 0.19 0.42 0.58 0.73

PAT

SIL

SPO

SLA

CGH

LON

BET

GAN

AGU

DHO

PAL

GON

5.7

1.9 8.8

1.4 3.5 10.1

1.9 4.8 30.4 27.1

2.5 6.3 24.7 40.8 inf.

2.5 6.1 43.7 3.9 13.2 11.9

1.7 5.4 27.3 18.4 18.8 41.4 5.8

1.0 1.8 2.4 1.8 2.2 3.0 2.3 1.8

4.2 8.8 11.2 3.8 4.8 6.4 8.0 4.9 4.1

1.4 1.7 1.5 1.0 1.2 1.4 1.5 1.2 1.2 3.7

0.7 0.8 0.8 0.6 0.7 0.8 0.8 0.6 0.7 1.6 2.7

0.4 0.5 0.4 0.4 0.4 0.6 0.4 0.4 0.3 1.1 3.4 6.4

0.10 0.22 0.17 0.14 0.14 0.15 0.36 0.10 0.37 0.54 0.67

0.09 0.03 0.04 0.02 0.03 0.29 0.08 0.40 0.55 0.72

0.04 0.02 0.20 0.05 0.36 0.21 0.49 0.62 0.71

)0.03 0.07 0.05 0.32 0.17 0.46 0.60 0.70

0.08 0.02 0.25 0.13 0.42 0.55 0.64

0.15 0.31 0.11 0.40 0.56 0.69

0.36 0.17 0.46 0.61 0.74

0.20 0.45 0.58 0.74

0.21 0.38 0.48

0.27 0.22

0.13

(b) AAS AAS BET GAN AGU HER PAL HAG

0.19 0.64 0.62 0.93 0.87 0.92

BET

GAN

AGU

HER

PAL

HAG

4.26

0.56 0.23

0.61 0.31 1.23

0.07 0.07 0.50 23.2

0.14 0.10 0.34 1.96 0.87

0.08 0.07 0.30 1.90 0.52 97.72

0.81 0.76 0.93 0.91 0.93

0.45 0.67 0.74 0.77

0.04 0.34 0.34

0.53 0.66

0.010

Below the diagonal, Fst values; above the diagonal, Nm values. Statistically significant figures are in bold type. See Table 1 for abbreviations. ª 2011 THE AUTHORS. J. EVOL. BIOL. 24 (2011) 2505–2519 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2011 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY

2514

S.

VON DER

HEYDEN ET AL.

Fig. 4 Haplotype network for the combined control region and ND2 mitochondrial DNA data set for M. dorsalis and C. superciliosus. The size of the circles is proportional to the frequency of each haplotype; the smallest circles represent extinct ⁄ unsampled haplotypes. Each line represents one mutational step, unless otherwise shown. Note that geographically closely related groups were combined into one locality, i.e. (Mouille Point + Sea Point + Slangkoppunt = Sea Point; Gonubie + Haga Haga = Haga Haga; Tieties Bay + Paternoster + Jacobsbaai + Silverstroomstrand = Tieties Bay.

not to study the phylogenetic relationships of South African clinids, this will not be discussed in further detail. However, our results suggest that the classification of South African clinids would likely benefit from an in-depth phylogenetic study. The phylogenetic results were consistent with the population studies described elsewhere, where C. superciliosus and M. dorsalis separated into three and two major clades, respectively. These clades were found not only in the combined data set, as presented in Fig. 5, but also for each marker taken independently (not shown). Importantly, those clades 1.

were concordant for both mitochondrial and nuclear markers, 2. were closest relatives and 3. exhibited divergences between them that were comparable to divergences exhibited by other closely related recognized clinid species. Indeed, the average pairwise sequence divergence between individuals in Clade 1 and Clade 2 was 2.5% in C. superciliosus and 3.5% in M. dorsalis, values that are similar to the pairwise sequence divergences between recognized species of other Clinus representatives such as C. acuminatus and C. heterodon (2.1%), and C. acuminatus and C. cottoides (3.0%).



Table 3 Analysis of molecular variance for Clinus superciliosus (Clade 1) and Muraenoclinus dorsalis (Clade 1). Population

Abbreviation

No.

Clinus superciliosus 1 Swakopmund 2 Paternoster 3 Silverstroomstrand Mouille Point 4 Sea Point 5 Slangkoppunt 6 Cape of Good Hope 7 Long Beach Miller’s Point 8 Betty’s Bay 9 Gansbaai 10 Agulhas 11 De Hoop Knysna Tsitsikamma 12 Port Alfred 13 Gonubie Haga Haga

SWA PAT SIL MOP SPO SLA CGH LON MIP BET GAN AGU DHO KNY TSI PAL GON HAG

4 5 5 1 16 14 16 1 6 10 10 5 3 2 2 8 3 2

Muraenoclinus dorsalis 1 Tieties Bay 2 Mouille Point 3 Sea Point 4 Aasbank 5 Betty’s Bay 6 Gansbaai 7 Agulhas 8 Herold’s Bay 9 Port Alfred 10 Haga Haga 4 Aasbank 5 Betty’s Bay 6 Gansbaai 7 Agulhas 8 Herold’s Bay 9 Port Alfred 10 Haga Haga

TIE MOP SPO AAS BET GAN AGU HER PAL HAG AAS BET GAN AGU HER PAL HAG

18 11 5 2 16 7 12 3 10 10 2 16 7 12 3 10 10

Phict

3.38 4.33 0.63 0.00 2.72 11.13 15.45 27.96 40.65 49.89

52.24 46.85

1.65 34.67 39.16 36.14 23.45 20.80 21.84 23.44 0.00 6.70 55.84 50.85 25.85 20.16 0.00

Discussion Molecular methodologies have greatly advanced our understanding of marine biodiversity by adding molecular phylogenies to existing morphological or behavioural data, as well as detecting cryptic species that had previously been unnoticed based on traditional taxonomic approaches (Knowlton, 1993). Marine biodiversity, for example, has been predicted to be vastly underestimated due to the poor evaluation of the number of cryptic species (Bouchet, 2006). In southern Africa, it is likely that at least 25% of endemic marine fishes are undescribed and that these are most likely to be small and found in shallow environments (von der Heyden, 2011), which is particularly pertinent to this study. Yet the processes that lead to the formation of cryptic species have not been thoroughly explored. The goal of this study was to evaluate the role of biogeo-

2515

graphic transitions in structuring populations while promoting cryptic speciation. The early phases of speciation in the sea may result from different processes. For example, the possibility of parapatric speciation along ecological boundaries has been suggested before (e.g. Rocha & Bowen, 2008). In our case, however, little is known about ecological differences in closely related clinid species, thus preventing us to extrapolate on their potential evolutionary significance. Yet when biogeographic boundaries are important components of faunal separations, they should result in (i) genetically differentiated populations and (ii) the presence of sibling cryptic species. Our data illustrate two important results. First, the two species investigated are likely to be a complex of cryptic species; secondly, there appear to be a number of oceanographic as well as biological barriers influencing population genetic structuring of these livebearing fishes. Cryptic speciation in southern African clinid fishes This study tested the hypotheses that genetic sister species occur across biogeographic barriers, given that there are several examples of rocky shore species in southern Africa exhibiting cryptic speciation (Heller and Dempster 1991; Edkins et al., 2007). This hypothesis was supported as both C. superciliosus and M. dorsalis show evidence of likely cryptic speciation. Genetic cryptic species do not exhibit any morphological differences but can be identified genetically. The separation of individuals into reciprocally monophyletic clades for both mitochondrial and nuclear markers, together with sequence divergences similar or higher than between recognized species in closely related taxa (Fig. 5), defines genuine cryptic species by being consistent with modern species concepts (Hausdorf, 2011). This is the case for Muraenoclinus dorsalis, for which only one species has been described to date (without any of the taxonomic conundrums that have dogged C. superciliosus, see below), and yet mitochondrial and nuclear markers separate individuals into two distinct clades. The two clades are sympatric and individuals belonging to the two clades have been collected from adjacent pools on the rocky shore at Sea Point at Cape Town (S. von der Heyden, pers. obs.). However, despite high genetic divergence (20% at the mitochondrial level), no ecological or morphological differences between the two clades have yet been identified. Interestingly, Clades 1 and 2 have a sympatric distribution on the south-west coast, with both found on the southern Atlantic seaboard of Cape Town and at Gansbaai. Despite intensive sampling, Clade 2 has not been found further east that Gansbaai or further north than Cape Town, suggesting that this clade has a restricted distribution on the south-west coast. For Clinus superciliosus, the situation is more complicated because there is a long history of species description


2516

S.

VON DER

HEYDEN ET AL.

Fig. 5 Phylogenetic relationships of a representative panel of South African clinids, including the focal species Clinus superciliosus and Muraenoclinus dorsalis. The phylogenetic tree is based on combined mitochondrial (12S rRNA, 16S rRNA) and nuclear (1st intron of ribosomal protein S7, rhodopsin) sequences. Maximum-likelihood, maximum parsimony and neighbour-joining methods were used and resulted in similar topologies; maximum likelihood is presented here. North American clinids Heterostichus rostratus and Gibbonsia elegans were used as outgroups. Bootstrap values and statistical support are presented next to the nodes.

in this group. Gilchrist & Thompson (1908) described two species that were very similar to the regular Clinus superciliosus, but differed in a number of morphological features, for example in the structure of the orbital cirri, colouration and patterning. These were named Clinus arborescens and Clinus superciliosus var. ornatus (Gilchrist & Thompson, 1908). Penrith (1969) did not recognize these two species and denoted them to be morphological variants of C. superciliosus. Given the morphological variability of C. superciliosus, and the lack of specimens of C. arborescens and C. ornatus, it is understandable that all were grouped as one species. However, given that we recover three clades, including one clade that includes only red and green colour morphs, it is likely that these clades may represent the missing C. superciliosus taxa. Initial morphological analyses support difference in the

shape of the supraorbital tentacles and fin spine numbers between clades (W. Holleman, pers. comm.), but too few specimens have been examined to draw firm conclusions at this point. Population genetic structuring of South African clinid fishes Our second hypothesis, based on previous results of genetically structured species with broadcast spawning and brooding life histories, was also supported by our data, which shows that there are a number of barriers to gene flow resulting in genetically structured populations. These differ somewhat between C. superciliosus and M. dorsalis, but nevertheless some important conclusions can be drawn when comparing them to the population



genetic structure of the clinid C. cottoides (von der Heyden et al., 2008) and other South African marine species (von der Heyden, 2009; Teske et al., 2011). First, the most pronounced break for C. superciliosus, to the west of Port Alfred, mirrors that found for C. cottoides, in which it was shown that no gene flow occurred between Port Alfred and all sampling sites to the west, which corresponds to a break in gene flow for several other marine species in the region (Teske et al., 2011). Interestingly, this does not appear to affect M. dorsalis to the same degree, as this species shows the main barrier to gene flow between Betty’s Bay, Gansbaai and Cape Agulhas. This region has been shown to play a crucial role in limiting gene flow in a number of marine organisms, including abalone, estuarine invertebrates and C. cottoides (Teske et al., 2011), as well as showing profound faunal differences for estuarine fishes (Harrison, 2002). Of all three species of clinid investigated to date, M. dorsalis shows the most structured populations. As Figs 3b and 4 show, very few haplotypes are shared between even geographically close sampling areas (as little as 50 km distance), and there are many apparently private haplotypes for each area sampled. Fst values (excluding those between Aasbank and Betty’s Bay that are geographically only a few kilometres apart) range between 0.34 and 0.93, which is indicative of strong population structuring in this species (Table 2b). This is not surprising as M. dorsalis is predominantly distributed on the mid- to high rocky shore, where opportunities for movement are much more limited compared to C. cottoides or C. superciliosus that are found from the midshore to the shallow subtidal (Prochazka & Griffiths, 1992) and for which Fst are lower (Fst = 0–0.7). The greatest barrier to gene flow for M. dorsalis appears to be around the Cape Peninsula, as evidenced by Clades 1a and 1b that are geographically separated by Cape Point (Figs 3b and 4). This has also been shown to be a significant break in faunal and floral distributions of marine organisms (Emanuel et al., 1992) and clearly separates the west coast from the south-west and south coast populations, as there are no shared haplotypes between sampling localities. Of further interest is that M. dorsalis sampled at Cape Agulhas fall into two well-supported subclades within Clade 1b: one containing fish from Cape Agulhas and Gansbaai (to the west of Cape Agulhas only) and one with fish exclusively from Cape Agulhas and Herold’s Bay (to the east). This suggests a region of overlap between south-west and south coast M. dorsalis in the Cape Agulhas area. Further, it appears that there has been an eastward colonization of M. dorsalis, as the clade containing fishes sampled at Port Alfred and Haga Haga (the east coast sampling areas) arises from within fishes sampled further to the west (Fig. 3b). This is consistent with gene flow patterns that are predominantly from west to east (eastward Nm = 0.64 migrants per generation, against westward movement Nm = 0.03). Such a

2517

trend was also shown for C. superciliosus (eastward Nm = 39.1, westward Nm = 11.4) and C. cottoides (von der Heyden et al., 2008) and may well be further evidenced that some live-bearing fishes utilize eastward flowing Agulhas countercurrents for dispersal rather than the larger westward Agulhas Current, which is typically further offshore and perhaps more likely to transport larvae that have a longer pelagic dispersal phase (Neethling et al., 2008; Teske et al., 2011). Although contemporary environmental and oceanographic variables play an important role in southern African marine species distributions, there are several historical factors that may have strongly influenced the genetic distribution patterns observed in this study. Most importantly are probably the events surrounding the last glaciation. This has been shown to have significant impact on numerous southern African marine species (see von der Heyden et al., 2010), as well as species in many of the other temperate marine regions globally (Lukoschek et al., 2007; Maggs et al., 2008; Larmuseau et al., 2009). It is highly likely that the difference in sea levels (around 130 m in southern Africa; Van Andel, 1989; Ramsay, 1995; Ramsay & Cooper, 2002) and temperature differences (Sachs et al., 2001) combined to significantly influence the marine environment. First, the drop in sea level caused a large decrease in available habitat, which exposed large areas of the Agulhas bank. Notably, because of the very narrow continental shelf on the west and especially east coasts (see Ramsay, 1995 and references therein), it is likely that there was little available habitat in these regions during glacial maxima and it is feasible that especially the east coast populations may have been isolated from those on the south coast. For C. cottoides that has completely isolated populations on the east coast, similar to C. superciliosus, this divergence has been estimated to be at least 60 000 years old (which coincides with glaciation events at marine isotope stage 4; Petit et al., 1999). Although this glaciation was not as severe in southern Africa as the last glacial maximum, it, nevertheless, may have served to completely isolate some populations. Whether these populations have come into secondary contact with the breakdown of potential barriers to dispersal remains unknown. However, there is evidence from other southern African endemic marine species which suggests that genetically isolated lineages are also physiologically differentiated and are unable to establish themselves permanently in neighbouring biogeographic provinces (Teske et al., 2011). Interestingly, in contrast to C. superciliosus and C. cottoides that have isolated populations on the east coast, M. dorsalis appears to show a recent colonization event of its eastern range distribution. Perhaps historical populations were extirpated during the last glaciations, because of a lack of suitable rocky shore habitat, whereas C. cottoides and especially C. superciliosus are probably more generalist in their use of the shallow subtidal


2518

S.

VON DER

HEYDEN ET AL.

environment and populations, although isolated, were able to survive. This hypothesis, however, requires further testing with a number of other southern African marine animals, including other high-shore specialists.

Conclusion When abiotic factors lower gene flow and ultimately drive the creation of a biogeographic barrier, it is predicted that faunal discontinuities are preceded by phylogeographic breaks soon followed by likely cryptic speciation events. We uncovered, in closely related clinid species, the presence of both phylogeographic breaks and cryptic speciation processes, thus fulfilling those predictions. Due to their cryptic nature, it is likely that such occurrences are more widespread than we realize. Targeted searches, such as the one presented here, will help us fully appreciate how common these situations truly are.

Acknowledgments We would like to thank Philippe Chanteur for help in the field. SvdH was supported by National Research Foundation South Africa (NRF-SA) and Claude Leon Postdoctoral Fellowships. Part of this work was supported by NRF-SA (GUN2053548, GUN2073184), National Science Foundation (INT-0117358) and University of California Santa Cruz-Committee on Research.

References Avise, J.C. 2004. Molecular Markers, Natural History and Evolution, 2nd edn. Sinauer Associates, Sunderland, MA. Awad, A.A., Griffiths, C.L. & Turpie, J.K. 2002. Distribution of South African marine benthic invertebrates applied to the selection of priority conservation areas. Divers. Distrib. 8: 129–145. Ayre, D.J., Minchinton, T.E. & Perrin, C. 2009. Does life history predict past and current connectivity for rocky intertidal invertebrates across a marine biogeographic barrier? Mol. Ecol. 18: 1887–1903. Beerli, P. 2003. Migrate – a maximum likelihood program to estimate gene flow using the coalescent, Tallahassee ⁄ Seattle. Available at: http://evolution.gs.washington.edu/lamarc/ migrate/html. Bouchet, P. 2006. The magnitude of marine biodiversity. In: The Exploration of Marine Biodiversity: Scientific and Technological Challenges (C.M. Duarte, ed.), pp. 32–64. Fundacion BBVA, Bilbao. Briggs, J.C. 1974. Marine Zoogeography. McGraw-Hill, New York. Burton, R.S. 1998. Intraspecific phylogeography across the Point Conception biogeographic boundary. Evolution 52: 734–745. Chow, S. & Hazama, K. 1998. Universal PCR primers for S7 ribosomal protein gene introns in fish. Mol. Ecol. 7: 1255–1256. Clement, M., Posada, D. & Crandall, K.A. 2000. TCS: a computer program to estimate gene genealogies. Mol. Ecol. 9: 1657–1660. Coyer, J.A. 1982. Observations of the reproductive behavior of the giant kelpfish, Heterostichus rostratus Girard (Pisces: Clinidae). Copeia 3: 344–350.

Dawson, M.N., Louie, K.D., Barlow, M., Jacobs, D.K. & Swift, C.C. 2002. Comparative phylogeography of sympatric sister species Clevelandia ios and Eucyclogobius mewberryi (Teleostei, Gobiidae) across the California transition zone. Mol. Ecol. 11: 1065–1075. Drummond, A.J., Ashton, B., Buxton, S., Cheung, M., Cooper, A., Duran, C. et al. (2011) Geneious v5.4. Available at: http:// www.geneious.com/. Edkins, M.T., Teske, P.R., Papadopoulos, I. & Griffiths, C.L. 2007. Morphological and genetic analyses suggest that southern African crown crabs, Hymenosoma orbiculare, represent five distinct species. Crustaceana 80: 667–683. Emanuel, B.P., Bustamante, R.H., Branch, G.M., Eekhout, S. & Odendaal, F.J. 1992. A zoogeographic and functional approach to the selection of marine reserves on the west coast of South Africa. S. Afr. J. Mar. Sci. 12: 341–354. Excoffier, L., Smouse, P.E. & Quattro, J.M. 1992. Analysis of molecular variance inferred from metric distances among DNA haplotypes - application to human mitochondrial-DNA restriction data. Genetics 131: 479–491. Excoffier, L., Laval, G. & Schneider, S. 2005. Arlequin ver. 3.0: an integrated software package for population genetics data analysis. Evol. Bioinform. Online 1: 47–50. Felsenstein, J. 1985. Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783–791. Gibson, R.N. 1982. Recent studies on the biology of intertidal fishes. Ocean. Mar. Biol. Ann. Rev. 20: 363–414. Gilchrist, J.D.F. & Thompson, W.W. 1908. The Blenniidae of South Africa. Ann. S. Afr. Mus. 6: 97–142. Griffiths, C.L., Robinson, T.B., Lange, L. & Mead, A. 2010. Marine biodiversity in South Africa: an evaluation of current states of knowledge. PLoS ONE 5: e12008. doi:10.1371/ journal.pone.0012008. Harrison, T.D. 2002. Preliminary assessment of the biogeography of fishes in South African estuaries. Mar. Freshw. Res. 53: 479–490. Hausdorf, B. 2011. Progress toward a general species concept. Evolution 65: 923–931. Heads, M. 2005. Towards a panbiogeography of the seas. Biol. J. Linn. Soc. 84: 675–723. Hedrick, P.W. 2005. A standardized genetic differentiation measure. Evolution 59: 1633–1638. Heller, J. & Dempster, Y. 1991. Detection of 2 coexisting species of Oxystele (Gastropoda, Trochidae) by morphological and electrophoretic analysis Journal of Zoology 223: 395– 418. Heller, R. & Siegismund, H.R. 2009. Relationship between three measures of genetic differentiation G(ST), D-EST and G’(ST): how wrong have we been? Mol. Ecol. 18: 2080–2083. von der Heyden, S. 2009. Why do we need to integrate population genetics into South African marine protected area planning? Afr. J. Mar. Sci. 31: 263–269. von der Heyden, S. 2011. ‘Carry on sampling!’ – assessing marine fish biodiversity and discovery rates in southern Africa. Divers. Distrib. 17: 81–92. von der Heyden, S., Prochazka, K. & Bowie, R.C.K. 2008. Significant population structure and asymmetric gene flow patterns amidst expanding populations of Clinus cottoides (Perciformes, Clinidae): application of molecular data to marine conservation planning in South Africa. Mol. Ecol. 17: 4812–4826.



von der Heyden, S., Lipinski, M.R. & Matthee, C.A. 2010. Remarkably low mtDNA control region diversity in an abundant demersal fish. Mol. Phylogenet. Evol. 55: 1183–1188. Irwin, D.E. 2002. Phylogeographic breaks without geographic barriers to gene flow. Evolution 56: 2383–2394. Jost, L. 2008. G(st) and its relatives do not measure differentiation. Mol. Ecol. 17: 4015–4026. Jost, L. 2009. D vs. G(st): Response to Heller and Siegismund (2009) and Ryman and Leimar (2009). Mol. Ecol. 18: 2088–2091. Katoh, K., Misawa, K., Kuma, K. & Miyata, T. 2002. MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform. Nucleic Acids Res. 30: 3059–3066. Knowlton, N. 1993. Sibling species in the sea. Annu. Rev. Ecol. Syst. 24: 189–216. Larmuseau, M.H.D., Van Houdt, J.K.J., Guelinckx, J., Hellemans, B. & Volckaert, F.A.M. 2009. Distributional and demographic consequences of Pleistocene climate fluctuations for a marine demersal fish in the north-eastern Atlantic. J. Biogeogr. 36: 1138–1151. Lee, W.J., Conroy, J., Howell, W.H. & Kocher, T.D. 1995. Structure and evolution of teleost mitochondrial control regions. J. Mol. Evol. 41: 54–66. Lima, D., Freitas, J.E.P., Araujo, M.E. & Sole-Cava, A.M. 2005. Genetic detection of cryptic species in the frillfin goby Bathygobius soporator. J. Exp. Mar. Biol. Ecol. 320: 211–223. Lukoschek, V., Waycott, M. & Marsh, H. 2007. Phylogeography of the olive sea snake, Aipysurus laevis (Hydrophiinae) indicates Pleistocene range expansion around northern Australia but low contemporary gene flow. Mol. Ecol. 16: 3406–3422. Lutjeharms, J.R.E. 2006. The Agulhas Current, 1st edn. Springer, Berlin. Maggs, C.A., Castilho, R., Foltz, D., Henzler, C., Jolly, M.T., Kelly, J. et al. 2008. Evaluating signatures of glacial refugia for North Atlantic benthic marine taxa. Ecology 89: S108–S122. Meirmans, P.G. 2006. Using the AMOVA framework to estimate a standardized genetic differentiation measure. Evolution 60: 2399–2402. Neethling, M., . Matthee, C.A., Bowie, R.C.K. & von der Heyden, S. 2008. Evidence for panmixia despite barriers to gene flow in the southern African endemic, Caffrogobius caffer (Teleostei: Gobiidae). BMC Evol. Biol. 8: 325. Palumbi, S., Martin, A., Romano, S., McMillian, W.O., Stice, L. & Grabowski, G. 1991. The Simple Fool’s Guide to PCR. University of Hawaii, Honolulu, 46 pp. Penrith, M.-L. 1969. The systematics of the fishes of the family Clinidae in South Africa. Ann. S. Afr. Mus. 55: 1–121. Petit, J.R., Jouzel, J., Raynaud, D., Barkov, N.I., Barnola, J.-M., Basile, I. et al. 1999. Climate and atmospheric history of the past 420,000 years from the Vostok ice core, Antarctica. Nature 399: 429–436. Posada, D. & Crandall, K.A. 1998. MODELTEST: testing the model of DNA substitution. Bioinformatics 14: 817–818. Prochazka, K. 1994. The reproductive biology of intertidal klipfish (Perciformes: Clinidae) in South Africa. S. Afr. J. Zool. 29: 244–251. Prochazka, K. & Griffiths, C.L. 1992. Observations on the distribution patterns, behaviour, diets and reproductive cycles of sand-dwelling clinids (Perciformes: Clinidae) from South Africa. Environ. Biol. Fish. 35: 371–379.

2519

Ramsay, P.J. 1995. 9000 years of sea-level change along the southern African coastline. Quat. Int. 31: 71–75. Ramsay, P.J. & Cooper, J.A.G. 2002. Late Quaternary sea-level change in South Africa. Quat. Res. 57: 82–90. Rice, W.R. 1989. Analyzing tables of statistical tests. Evolution 43: 223–225. Riddle, B.R., Dawson, M.N., Hadly, E.A., Hafner, D.J., Hickerson, M.J., Mantooth, S.J. et al. 2008. The role of molecular genetics in sculpting the future of integrative biogeography. Prog. Phys Geogr. 32: 173–202. Rocha, L.A. & Bowen, B.W. 2008. Speciation in coral-reef fishes. J. Fish Biol. 72: 1101–1121. Ryman, N. & Leimar, O. 2009. G(st) is still a useful measure of genetic differentiation - a comment on Jost’s D. Mol. Ecol. 18: 2084–2087. Sachs, J.P., Anderson, R.F. & Lehman, S.J. 2001. Glacial surface temperatures of the southeast Atlantic Ocean. Science 293: 2077–2079. Sambrook, J., Fritsch, E.F. & Maniatis, T. 1989. Molecular Cloning: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Sevilla, R.G., Diez, A., Noren, M., Mouchel, O., Jerome, M., Verrez-Bagnis, V. et al. 2007. Primers and polymerase chain reaction conditions for DNA barcoding teleost fish based on the mitochondrial cytochrome b and nuclear rhodopsin genes. Mol. Ecol. Notes 7: 730–734. Shanks, A.L. 2009. Pelagic larval duration and dispersal distance revisited. Biol. Bull. 216: 373–385. Shimodaira, H. & Hasegawa, M. 1999. Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Mol. Biol. Ecol. 16: 1114–1116. Smith, M.M. 1995. Clinidae. In: Smiths’ Sea Fishes (M.M. Smith & P.C. Heemstra, eds), pp. 758–769. Struik Publishers, Cape Town. Swofford, D.L. 2003. PAUP*. Phylogenetic Analysis Using Parsimony (*and Other Methods). Version 4.0b10. Sinauer Associates, Sunderland, MA. Taylor, M.S. & Hellberg, M.E. 2005. Marine radiations at small geographic scales: speciation in neotropical reef gobies. Evolution 59: 374–385. Teske, P.R., von der Heyden, S., McQuaid, C.D. & Barker, N.P. 2011. A review of marine phylogeography in southern Africa. S. Afr. J. Sci. 107, Art #514. Van Andel, T.H. 1989. Late Pleistocene sea-levels and the human exploitation of the shore and shelf of southern Africa. J. Field Archaeol. 16: 133–155. Vrijenhoek, R.C. 2009. Cryptic species, phenotypic plasticity, and complex life-histories: assessing deep-sea faunal diversity with molecular markers. Deep-Sea Res. Part II 56: 1713–1723. Wallace, A.R. 1898. The Malay Archipelago, the Land of the Orangutan and the Bird of Paradise; a Narrative of Travel, with Studies of Man and Nature. Macmillan and Co., London. Zwickl, D.J. 2006. Genetic algorithm approaches for the phylogenetic analysis of large biological sequence datasets under the maximum likelihood criterion. PhD dissertation, The University of Texas at Austin. Received 19 November 2010; revised 27 July 2011; accepted 31 July 2011.
