Geographic genetic variation in the Coral Hawkfish, Cirrhitichthys oxycephalus (Cirrhitidae), in relation to biogeographic barriers across the Tropical Indo-Pacific

Sample collection

We collected 166 individuals of Coral Hawkfish from 45 sites at 20 locations found in all biogeographic regions and provinces from the TIP (Fig. 1 and Table S1). Specimens were collected on SCUBA using, in different locations and at different times, various combinations of rotenone ichthyocide (Robertson & Smith-Vaniz, 2008), clove oil anesthetic (Piñeros et al., 2019) and either aquarium hand nets, a suction tool or pole-spears. We preserved tissue samples in either DMSO buffer or 95–96% ethanol and stored them at either −20 °C or −76 °C. Specimens were deposited in the fish collections at the Universidad Michoacana de San Nicolás de Hidalgo, Mexico, University of Central Florida, USA, The National Museum of Natural History of the Smithsonian Institution, USA, the King Abdullah University of Science and Technology, Saudi Arabia, University of California Santa Barbara, USA, and Ho Chi Minh City University of Science, Vietnam (Table S1). Organism collection was supported and allowed by the following institutions and permits by Mexican Ministry of Environment and Natural Resources under collection permits numbers (PPF/DGOPA-035/15; and F00.DRPBCPN.DIR.RBAR-100/2015-CONANP), and for El Salvador (MARN-AIMA-004-2013), Panama (SC/A-17-19), Costa Rica (007-2013-SINAC and R-056-2105-OT-CONAGEBIO), and Ecuador (013/2012 PNG; N°21-2017-EXP-CM-2016-DNB/MA and MAAE-DBI-CM-2021-0152).

DNA extraction and sequencing

We extracted total genomic DNA from tissue samples using the phenol-chloroform protocol of Sambrook, Fritsch & Maniatis (1989). We PCR amplified a 652 base pair (bp) fragment of the mitochondrial cytochrome b (cytb) gene, using the primers GluDG (forward) and H16460 (reverse) (Perdices & Doadrio, 2001). Because the samples from Clipperton failed to amplify using these universal primers, we designed specific primers 1F31 ACGGCTGACTAATCCGTA (forward) and 1R61 AATTAGGGATGCGACTTGTCC (reverse). Taking into account the degree of variation found in the cytb, we also amplified a 132 bp fragment of the recombination-activating nuclear gene 1 (RAG1) in a subsample of 45 samples using the primers RAG 1F (forward) and RAG 9R (reverse) (Quenouille, Bermingham & Planes, 2004). We performed all PCRs in a total volume of 12.5–16.5 μL with 1 μL (50–100 ng) DNA template, 0.25 μl of 10 μM for each primer, 0.2–0.5 μl of 50 mM MgCl2, 0.25–0.9 μl of 10 mM dNTPs, 0.0625–0.088 μL of 5 U/μL Taq polymerase (Invitrogen, Waltham, MA, USA), 1.25–1.5 μL of 10X buffer and deionized sterile water to reach the final volume. Those PCRs utilized the following thermocycling conditions: initial denaturation at 94 °C (3 min) and a final extension at 72 °C (10 min), with an intervening 35 cycles of 30 s at 94 °C, 45 s at the annealing temperature (51 °C for cytb; 59 °C for RAG1), and 1 min at 72 °C. We visualized the PCR products on 1.5% agarose gels and sent amplicons to MACROGEN Inc. (Seoul, South Korea) for sequencing.

We edited and aligned all the sequences using chromatograms with the software MEGA v7.0.2 (Kumar, Stecher & Tamura, 2016). We conducted the recombination test for the nuclear genes with the software Split Tree v4 (Huson & Bryant, 2006) for the RAG1 gene. We deposited cytb unique sequences in GenBank (Table S1) and sequences of RAG1 are available at https://doi.org/10.5281/zenodo.11323581.

Data analyses

Phylogenetic analyses and haplotype networks

The phylogenetic analysis for cytb shows a basal polytomy that includes all 59 samples from the various provinces within the TICP and another group composed of all 107 TEP samples. In the TEP group there are two sub-clades, one represented by all but one of 19 samples from Clipperton, plus four of 21 samples from Revillagigedo Archipelago and two of 25 from Mainland Mexico, the CL-TEP sub-clade. The second sub-clade (TEP), which had low support, includes the remaining 82 samples from the (23 samples from Mexico mainland, 17 from the Revillagigedo Archipelago, five from El Salvador, one from Costa Rica mainland, nine from Cocos Island, six from Panama, two from Colombia mainland, seven from Ecuador mainland, 12 from the Galapagos Archipelago) and one from Clipperton Atoll (Fig. 2). The phylogenetic analyses for the RAG1 gene show a basal polytomy without resolution (see Fig. S2).

The cytb haplotype network resolved three haplogroups. One haplogroup (TICP, Fig. 3) includes all the samples from the TICP and includes a mixture of haplotypes from the different regions and provinces, with 1–10 mutation steps separating different haplotypes from their closest neighbors. The TICP haplogroup is separated by 10 mutational steps from haplotypes in the main TEP haplogroup (Fig. 3), which included 44 of 46 samples from the mainland TEP, one of 19 samples from Clipperton Atoll, 17 of 21 samples from the Revillagigedo Archipelago, and all nine samples from Cocos Island and 12 from the Galapagos Archipelago. The TEP haplogroup in turn was separated by six mutation steps from the Clipperton haplogroup (CL-TEP). In contrast to cytb haplotype network, RAG1 resolved no geographic structuring and the two most common alleles in this dataset were represented in all regions and provinces (Fig. 4).

Genetic structure, distances, and genetic diversity

In general, AMOVA results for cytb found the highest genetic variation among groups, with all comparisons being statistically significant (Table 1). The arrangement of three groups (TICP/TEP/CL-TEP) showed the highest and significant value of ΦCT (64.39%) (Table 1). Results for RAG1 should be taken with caution due to the lack of samples in the Red Sea Province (representing the only one of four total regions without RAG1 samples), which we were unable to amplify. The highest percentage and statistically significant values for RAG1 were for the within populations (89.19%) comparison in the three-group arrangement (Western Indian Ocean Province/Indo-Polynesian and Marquesas Island Province/TEP Region; Table 1). For cytb, the K = 3 arrangement maximized the differences between groups in the SAMOVA analysis with the highest and significant value of ΦCT (59.36%), segregating the populations into three groups: TICP, TEP and CL-TEP (Fig. 1 and Table 2). We obtained high and significant levels of genetic differentiation among those three groups. The highest and most significant ΦST value for cytb was obtained between TICP and CL-TEP groups (ΦST = 0.637), and the lowest between TEP and CL-TEP (ΦST = 0.493). The largest genetic distance (pd) detected among groups also was between the TICP and CL-TEP (2.96%), and the lowest between the TEP and CL-TEP (1.58%). For RAG1 the highest pd genetic distance was between TEP and TIP (1.3%) and the lowest between CL-TEP and TIP (0.9%), while the highest and significant ΦST value was between TEP and CL-TEP (0.114) and the lowest between TIP and CL-TEP (0.012) (Table 3). For location/regions comparisons involving cytb the highest and significant ΦST was obtained between Galapagos and Vietnam (0.818), and the lowest between RSP and WIOP (−0.02), while the highest pd distances was obtained between Mexican and Arabian populations (3.228%) and the lowest between the Cortez and Panamic Provinces of the TEP (0.001). For RAG1 the highest pd was between Mexican and Christmas populations (1.799%) and lowest between the Galapagos and Clipperton populations (0.379), whereas the highest and significant ΦST (0.298) was between Mexican and Clipperton populations and the lowest between Dongsha and Kiribati populations (0.001) (for more details see Table S2). For the modified Briggs & Bowen (2012) 4-area scheme the highest pd was between the RSP and TEP (2.94%), and the lowest between RSP and IP-M (1.47%), with the highest and significant ΦST obtained between RSP and TEP (0.617) and the lowest between RSP and WIOP (−0.029) (Table S3).

The Mantel test showed a significant correlation between genetic- and geographic distances across the entire TIP dataset (r = 0.49, p < 0.05) but not within either the TICP (r = 0.11, p > 0.05) or the TEP (r = 0.05, p > 0.05).

Overall, nucleotide diversity ( $\pi$) for cytb was 0.017 and haplotype diversity ( $h$) was 0.976, with values of 0.011 and 0.803, respectively, for RAG1. For the three groups resolved in the haplotype networks CL-TEP shown the highest values for haplotype diversity (cytb $h$ = 0.989 +/− 0.006 and RAG1 $h$ = 0.827 +/− 0.046) and TEP the lowest values for nucleotide diversity (cytb $\pi$ = 0.754 +/− 0.071 and RAG1 $\pi$ = 0.703 +/− 0.101) (Table 4). For genetic diversity according to the Briggs & Bowen (2012) plus Red Sea regionalization, the highest haplotype diversity values were from the Red Sea Province (cytb $h$ = 1 +/− 0.177) and the Western Indian Ocean Province (RAG1 $h$ = 1 +/− 0.500). The lowest value was for Australia (cytb $h$ = 0) and Arabia, Vietnam, Australia and the Galapagos (RAG1 $h$ = 0) (Table 5). For genetic diversity according with location/regions, the highest values were from the Mexican Province (cytb $h$ = 0.989 +/− 0.031) and from South Africa and the Marquesas (RAG1 $h$ = 1 +/− 0.500). The lowest values were from Australia (cytb $h$ = 0) and Arabia, Vietnam, Australia, and the Galapagos (RAG1 $h$ = 0) (Table S4).

Two evolutionary lineages: an effect of the Eastern Pacific Barrier

The results presented here, based upon mtDNA, indicate that C. oxycephalus populations from two separate evolutionary units (lineages) (Fig. 3 and Tables 1–3), one distributed across the Tropical Indo-Central Pacific, and the other restricted to the Tropical East Pacific, separated by ten mutation steps, between 2.7 to 2.9 pd and with significant ΦST results. These results support a previous study by Lessios & Robertson (2006) using the mitochondrial ATPase six and eight genes, which indicated that the two regional clades are relics of long past isolation, with significant genetic structure and no gene flow for at least the past ca. 700,000 years. Their results led them to suggest that C. oxycephalus originated in the TEP and spread westward across the EPB to the TICP. However, they only compared populations in the TEP with those on the western edge of the EPB, not across the entire TICP. Although we also found that the mainland TEP mtDNA group is centrally positioned, the mitochondrial phylogenetic tree and nuclear results do not support the idea of a TEP origin. Further, the fact that the TEP group is not the most diverse genetic one in C. oxycephalus and that the center of diversity of the genus Cirrhitichthys, with seven other species, is in the Western Pacific, seems to indicate a TICP origin for C. oxycephalus. The EPB is a 4,000–7,000 km extension of deep water that has been present for about 65 Myr (Grigg & Hey, 1992; Lessios & Robertson, 2006). Darwin (1872) thought of it as an impassable barrier due to the large deep-sea distances between islands. As with the results presented here, some recent studies found that the EPB can be crossed it remains an effective barrier that isolates populations of some organisms, including a lobster (Panulirus penicillatus), a benthic mollusk (Conus ebraeus), and a widespread Indo-Pacific pipefish (Doryhamphus excisus) on both sides of the EPB (Lessios & Robertson, 2006; Duda & Lessios, 2009; Chow et al., 2011; Iacchei et al., 2016). However, some transpacific reef fishes do have low levels of trans-EPB genetic differentiation (Rosenblatt & Waples, 1986, Lessios & Robertson, 2006). Scheltema (1988) found that species of benthic invertebrates that can cross the EPB are those likely to have an exceptionally long larval life i.e., those with telepathic larvae, while other forms lacking that larval characteristic are unable to do so. The PLD has been frequently linked to the dispersal potential of marine species and hence to the degree of genetic connectivity of populations separated by long distances (e.g., Baums, Paris & Chérubin, 2006; Bowen et al., 2006; Reece et al., 2011). Glynn, Veron & Wellington (1996) indicated that due to the flow rate of ocean currents across the EPB that barrier could be bridged in 60 to 120 days. Romero-Torres et al. (2018) using a spatially-explicit biophysical model of larval dispersal and, assuming extreme El Niño events, found the time of larval transport of 110–130 days from the Line Islands to Clipperton Atoll. The PLD of the Coral Hawkfish ranges from 36–51 days (Brothers & Thresher, 1985; Robertson, Grove & McCosker, 2004). Among members of other genera of Cirrhitids PLDs (based on very few specimens) range from ~45 to 73 days, based on very few specimens (Brothers & Thresher, 1985; DeMartini, Wren & Kobayashi, 2013). However, PLDs do vary within species, often considerably so and that of C. oxycephalus evidently is sufficient to have allowed it to cross the EPB in the distant past.

Most species of hawkfish (five out of seven in the IP-H-M) prefer live coral habitats, although in the Chagos Archipelago, C. oxycephalus apparently prefers rock-pavement habitats (Coker et al., 2015). However, C. oxycephalus is considered to be dependent on live Pocillopora corals on TEP reefs (Robertson & Allen, 1996; Alvarez-Filip & Reyes-Bonilla, 2006). Cirrhitichthys oxycephalus is the only species of this genus in the TEP and one of the few fishes strongly associated with live Pocillopora corals (Robertson & Allen, 1996; Alvarez-Filip & Reyes-Bonilla, 2006), which are not only common element of structural coral reefs in the TEP but also are commonly present as small, attached growths on the rocky shorelines that represent the predominant shallow reef type throughout the TEP. Reduced interspecific competition due to the absence of congeners and other coral-associated species, could have aided the establishment and spread throughout that region of C. oxycephalus.

Tropical Indo-Central Pacific group

Even though many reef fishes show high levels of endemism in different TICP provinces (Bowen et al., 2016; Cowman et al., 2017), the ΦST value and the mixture of cytb haplotypes found in C. oxycephalus between provinces and populations distributed across the breadth of that vast area that extends over approximately half of the circumference of the globe, from the Red Sea, through Western Indian Ocean and the Indo-Malayan archipelago and eastern Australia, to the Line Islands and the Marquesas (Briggs & Bowen, 2012; Toonen et al., 2016), do not provide evidence of phylogeographic pattern across the TICP for this mtDNA. Although we must approach these results with caution, given the low sample sizes in some regions (e.g., only four in Red Sea Province), they still offer the opportunity to formulate phylogenetic and evolutionary hypotheses.

Biogeographic studies in the distribution range of C. oxycephalus across TICP region have generally taken into account four semipermeable barriers, the Indo-Pacific Barrier of the Indo-Malayan archipelago, the Red Sea Barrier, the Marquesas Barrier and the mid-Indian Ocean Barrier (Briggs, 1974; Gaither & Rocha, 2013; Borsa et al., 2016). The first of those is considered to be a harder barrier to dispersal formed by a group of islands with large land areas that can prevent dispersal (Craig et al., 2007), the second is considered to be a ‘softer barrier’ due the unusual environmental conditions and narrowness of the Red Sea entrance (DiBattista et al., 2016), the third barrier is attributable to a combination of geographical isolation (a biogeographical barrier to dispersal), unusually variable sea temperatures for an equatorial archipelago (ecological distinctiveness), hydrographical isolation (cold upwelling, prevailing currents) and young geological age (few marine habitats) (Randall, 2001; Gaither et al., 2010), while the last and largest barrier to shallow-water reef fishes in the Indian Ocean is thought to be the wide expanse of deep sea west of the Lakshadweep-Maldives-Chagos archipelagoes. Many marine species that are found spread across large parts of the TICP have genetic population partitions that geographically coincide with one or more barriers between the Red Sea, the Western and mid-Indian Ocean, and the Indo-Polynesian, the Hawaiian, the Marquesas and the Easter Island Provinces (Williams & Benzie, 1998; Lessios, Kessing & Pearse, 2001; Borsa et al., 2016). However, the six internal barriers within the TICP seem to have had little effect on the dispersal of C. oxycephalus throughout that region except for the stretches of deep water isolating the Hawaiian and Easter Island Provinces from the central Pacific. Since reefs in both of those provinces harbor abundant Pocillopora corals (Grigg, 1983; Glynn et al., 2003; Irving & Dawson, 2013), the oceanic distances isolating the Hawaiian and Easter Island Provinces are less (>2,000 km) than the EPB crossed by Coral Hawkfish and given the versatility in habitat use of the Coral Hawkfish (Coker, Pratchett & Munday, 2009; Coker et al., 2015), the absence of the species in both areas is likely due to local extinction after colonization or to its failure to disperse, mediated by oceanographic conditions, across their isolating barriers, rather than ecological limitations or competition. The Coral Hawkfish joins the small group of reef fishes from a diversity of taxa known to show mismatches between genetic differentiation and other internal TICP geographic barriers, with high inter-oceanic gene flow occurring on a relatively recent evolutionary timescale across TICP ranges as large as that of C. oxycephalus. Those include four species of moray eels (Reece et al., 2011); the snapper Lutjanus kasmira (Gaither et al., 2010), four species of surgeonfishes in the genus Naso (Horne et al., 2008; Horne & Van Herwerden, 2013), the butterflyfish Chaetodon meyeri and the cornetfish Fistularia commersonii (Jackson et al., 2015; Borsa et al., 2016). In addition, while the snapper Pristipomoides filamentosus exhibits a lack of genetic structure across the sampled parts of the TICP that are also occupied by the Coral Hawkfish, it does show differentiation of the Hawaiian Island population (Gaither et al., 2011a). Fistularia commersonii is the only reef-fish to date that has been shown to lack significant mtDNA structure across the entire longitudinal extent of the TIP, from the TEP to the Red Sea. However, unlike the situation in C. oxycephalus, in none of those 12 species has sampling included populations on both sides of all major barriers that separate all their known populations and only two of them (Naso hexacanthus and F. commersonii) have included samples from the Red Sea.

While some reef fishes with relatively short PLDs do show much more marked geographic genetic differentiation than species with longer PLDs (Borsa et al., 2016), PLDs among the 12 species referred to above range from relatively short (~31 days in L. kasmira) to very long (up to 180 days in P. filamentosus). Although it has been hypothesized that the longer the PLD, the farther larvae can disperse, recent studies have concluded that this relationship does not always hold, that PLD alone is not a determining factor in successful colonization, which is likely affected by navigation, swimming ability, oceanic conditions, habitat availability in the colonized region, number of migrants and certain life history traits (Selkoe et al., 2010; Leis, Siebeck & Dixson, 2011; Selkoe & Toonen, 2011; Luiz et al., 2012; Szabó et al., 2014). While the lack of geographic structure across the TICP in C. oxycephalus could reflect ongoing gene flow facilitated by moderate-length PLD or versatility in habitat preferences that enhance dispersal within the region, it could also be due to a recent geographic expansion by a species with a large population size.

Tropical Eastern Pacific group

Results within the TEP reveal two divergent groups with cytb, one composed of haplotypes largely restricted to Clipperton and the other that includes samples collected throughout the rest of the TEP islands and continent, results that are also supported by the RAG1 ΦST test. The two TEP groups are well segregated by six mutation steps (Fig. 3) and show significant values in the ΦST for both genes, significant differentiation in cytb AMOVA and SAMOVA analyses and pd = 1.58% (Tables 1–3). Clipperton Atoll is the most isolated shoaling reef in the TEP, 950 km from the nearest emergent reef at Socorro Island in the Revillagigedos. However, the dispersal capacity of the species seems to be not strongly related to Clipperton’s isolation, since a Mantel test not found correlation between genetic distances and geographic distances among the TEP samples. There are a few Clipperton haplotypes also present at the Revillagigedo Islands and continental Mexico and one from the mainland TEP mtDNA group at Clipperton and such apparent vagrants of other species are seen at various oceanic islands in the TEP (Muss et al., 2001; Robertson, Grove & McCosker, 2004; Palmerín-Serrano et al., 2021). In Hawaii, mesoscale eddy-currents transport fish eggs into entrapment centers where they are retained long enough to mature and then return to the reef of origin (Lobel & Robinson, 1986). Similar processes may tend to limit long-distance dispersal of C. oxycephalus larvae and promote the larval retention and the return of larvae to Clipperton. This process was proposed to account for the isolation of Epinephelus clippertonensis by Craig et al. (2006), which is now known to also occur in the Revillagigedos and mainland Baja (Robertson & Allen, 2015), and a biophysical model by Romero-Torres et al. (2018) indicates potential isolation of two reef-building coral species of Clipperton based on a biophysical model of larval dispersal. Clipperton’s fish fauna includes various local endemic species and oceanographic processes that must maintain the long-term persistence of populations of those species at this island likely also produce genetically differentiated populations that may eventually become endemic species. However, other oceanographic processes in the general area of Clipperton likely enhance larval dispersal further afield, including northwards towards the Revillagigedos (Adams & Flierl, 2010).

Taxonomic considerations

Cirrhitichthys oxycephalus was described from type specimens collected in Indonesia. Randall (1963) included this species in a taxonomic review of the family and synonymized three other species with it, including C. corallicola (Tee-Van, 1940), type locality Gorgona Island, on the Pacific coast of Colombia. The original description of C. corallicola, which incorporated specimens from mainland sites in Mexico to Colombia but none from the TEP oceanic islands, makes no mention of C. oxycephalus. Randall (1963) synonymized C. corallicola with C. oxycephalus after examining “all available specimens of Cirrhitidae” from various museums, including the USNM. That museum houses 31 specimens collected from Gorgona that, based on the 1935 collection date, he likely assessed. Unfortunately, it is unclear from Randall’s (1963) brief reference to C. corallicola whether the TEP population differs morphologically from the TICP populations. However, our cytb results suggest that C. corallicola could be a valid species that encompasses all the TEP populations, as suggested by Lessios & Robertson (2006), while C. oxycephalus extends throughout the remainder of its range across almost the entire TICP. Populations of C. oxycephalus in the TEP and TICP clearly need a detailed integrative systematic analysis that includes morphological and genomic assessments at both intraregional and interregional spatial scales. Such an analysis would expand the range of information about any geographic variation related to barriers within both those major biogeographic regions.