The highest rates of extinction have been recorded for endemic species and populations of widespread species inhabiting islands [1,2,3]. The high risk of extinction in this group has been attributed to a range of demographic, environmental and genetic factors [1,2,4]. The genetic characteristics that increase extinction risk in this group are low genetic diversity and inbreeding depression [1,2]. Low genetic diversity increases extinction risk because it reduces the potential for species to adapt to rapid environmental change. In endemics, and isolated populations of widespread species, inbreeding depression is thought to occur because of the low number of individuals forming the founder population and the small size of the extant population [5,6]. The importance of maintaining genetic diversity to reduce extinction risk is recognised by the IUCN and is considered to be a conservation priority [6,7].

Genetic diversity can be influenced by a range of factors including population size, natural selection, mutation rates, gene flow between populations, introgression from hybridisation and historical effects on these factors (e.g., population bottlenecks) [6]. In the terrestrial environment, island endemic species usually have small populations [5,8] and consequently, genetic diversity is predicted to be low as genetic variation is increasingly lost through genetic drift [9]. Isolated island populations of widespread species are also expected to have low genetic diversity because of limited gene flow with other populations and lower localised effective population sizes [6,10]. Therefore, low genetic diversity is expected to be a feature of terrestrial communities on remote islands because they contain a high proportion of endemic species and isolated populations of widespread species.

Coral reef fishes are the most diverse vertebrate communities in the world and are found throughout the tropics, including numerous remote islands. These isolated islands are hotspots of coral reef fish endemicity [11,12,13,14] and also support isolated populations of species with broad geographic ranges. Determining whether these endemics and isolated populations conform to terrestrial patterns of low genetic diversity is crucial to assessing their risk of extinction. Of the known neo-extinctions of reef fishes, all have been from isolated islands or island groups [15,16]. An appreciation of the genetic diversity of endemic reef fishes is also of importance for ascertaining extinction risk of this group to changes in their environment caused by local and global anthropogenic disturbances [17,18].

Previous studies have revealed a range of genetic diversity values for populations of marine fishes, including those inhabiting coral reefs. Haplotype diversity (h) in populations of marine fishes has been found to vary vastly from 0 to 1, while nucleotide diversity (π%) is known to vary from 0.07 to 31.8 [19,20,21,22,23]. For marine fishes, haplotype and nucleotide diversities are generally considered to be low where h and π% are less than 0.5 [19]. Therefore, if reef fishes conformed to predictions based on terrestrial fauna, we would expect endemic species, and isolated populations of widespread fishes, to have low haplotype (h < 0.5) and nucleotide (π < 0.5%) diversities.

This study determines whether endemic and isolated populations of widespread angelfishes, at Christmas and Cocos Islands have low genetic diversity (h < 0.5 and π < 0.5%). These islands are an ideal location to examine genetic diversities of isolated marine species because they are among the most-isolated islands in the tropical Indian Ocean [24]. Three congeneric pygmy angelfish (genus: Centropyge) species were examined. Centropyge joculator is endemic to these two locations and is not found anywhere else in the world [25]. C. bispinosa and C. flavicauda are among the most widely distributed angelfishes, ranging from east Africa to the central Pacific Ocean [25], with Christmas Island supporting geographically-isolated populations of these species. Christmas and Cocos Islands are separated by approximately 1000 km, therefore, we also predicted that the two populations of the endemic species (C. joculator) might be genetically subdivided between these locations due to low genetic exchange [26].

In the terrestrial environment, genetic diversity of endemics and isolated populations of widespread species tends to be low and this may be an important factor contributing to their extinction risk [1,9]. Based on this, and the known range of genetic diversity for marine fishes [19], we predicted that island endemic reef fishes and isolated populations of species with broad geographic ranges would have low haplotype (h < 0.5) and nucleotide (π < 0.5%) diversity. Instead, we found that genetic diversity values of the endemic and isolated angelfish populations (h > 0.98, π > 2.8%) were among the highest recorded for marine fishes [19,21,22,23,36,37,38,39]. Hence, contrary to expectations based on terrestrial species, it appears that there are processes operating that are promoting, rather than reducing, genetic diversity in these endemic and isolated populations.

There are a number of factors that influence genetic diversity. Endemic species usually have low abundance [5,8] and therefore are predicted to have low genetic diversity because of the increasingly deleterious effects of inbreeding and genetic drift on small populations [9,40]. However, marine endemics often have high abundance [41] and thus the effect of drift on genetic diversity may be minimal. Indeed, the abundance of the endemic study species C. joculator is more than 30 times greater than that of the 16 widespread species of angelfishes that are present at Christmas and Cocos Islands [42]. High genetic and nucleotide diversity is expected in fishes that maintain large populations through their evolutionary history [19]. If the current high abundance of C. joculator has been a feature of its evolutionary history then this will have contributed to its high haplotype and nucleotide diversity.

Irrespective of population size, genetic diversity can also be high if endemics occur in refugia (e.g., Pleistocene refugia: [43]) where environmental conditions have been relatively stable over evolutionary time scales allowing for the accumulation of genetic variation (see [44] and references therein). Christmas and Cocos Islands are oceanic islands that have the Indonesian Archipelago as their nearest neighbour. Indonesia is an area where Pleistocene sea level fluctuations caused significant disruption to marine habitats and fauna [45]. If Christmas and Cocos Islands were environmentally stable then they may have served as Pleistocene refugia. Also, C. joculator may have had high genetic diversity if it was once more widespread and as its range contracted this diversity may have been condensed and maintained in the relict populations at Christmas and Cocos Islands. These islands also support endemic relict populations of other reef fishes (e.g., C. flavissima and Stegastes insularis, [25,46]) whose disjunct geographic ranges are separated by the historically tumultuous Indonesian Archipelago [45]. Therefore, the high genetic diversity of C. joculator at Christmas and Cocos may also be because it is endemic to refugia and/or because it is a relict of a once widespread species.

Endemics can also have high genetic diversity if they represent a fusion of clades or have interbred with other species (i.e., introgression from hybridisation) [47,48,49,50]. Christmas and Cocos Islands represents a suture zone where different clades or species come into secondary contact and interbreed [51,52]. C. joculator has both high haplotype and nucleotide diversity, which is a signature of secondary contact between different clades or lineages [19]. Indeed, two genetically distinct groups were detected in C. joculator and while the historical cause(s) for the origin of these groups is not known, it has clearly increased the genetic diversity of this species.

The restricted gene flow between Christmas and Cocos populations of C. joculator supports the notion that endemic reef fishes have limited dispersal [26]. Restricted gene flow may have also influenced genetic diversity in C. joculator. If local adaptation occurs at each island location then this will lead to differences in the gene pools of the two populations. Rare dispersal events may allow for enough gene flow to increase genetic diversity in each population, but gene flow may be insufficient to homogenise the genetic composition or to counteract adaptive changes of both populations.

Genetic diversity in isolated populations of widespread species is predicted to be low because population size is usually small and gene flow to other populations is low [1]. However, genetic diversity was high despite the very small population sizes observed for the two widespread study species at Christmas Island (0.05 to 0.18 individuals per 300 m²: [42]). It is possible these isolated populations still receive sufficient gene flow from other populations because these species are among the most widely distributed angelfishes in the world [25] and gene flow would be required to establish and maintain this broad distributional range. Isolated populations of other widespread Centropyge species also have high haplotype diversity and high genetic connectivity [37,53]. Furthermore, the recent arrival of C. acanthops at the Cocos Islands (Hobbs personal observation) demonstrates that larvae of widespread Centropyge species are capable of dispersing over 2500 km. Broad dispersal in widespread Centropyge may be facilitated by broadcast spawning, a pelagic larval duration of 30 to 35 days and competent swimming abilities of late stage larvae [54,55], which may be further aided by favorable currents at the time of spawning. Moderate to high genetic diversity (haplotype and/or nucleotide) has also been reported in several other widespread reef fishes present at Christmas and Cocos Islands [21,22,23,39]. The gene flow received by populations of these widespread species at Christmas and Cocos Islands may be sufficient to counteract the loss of genetic diversity associated with genetic drift. For the widespread species examined in this study, additional sampling from other populations would be required to confirm that gene flow was the reason for their high genetic diversity at Christmas Island.

Genetic diversity can also be related to taxonomy with some groups having higher diversity due to faster mutation rates [9]. High haplotype diversity in an Atlantic Centropyge angelfish was partly attributed to the high mutation rate of the control region of mtDNA [37]. Similarly, a very high rate of mutation rate has been detected in the mtDNA control region of butterflyfishes [56], which are the sister group to angelfishes [25,57]. Therefore, the high genetic diversity observed in all three angelfishes in this study is probably influenced (at least partly) by the fast mutation rate of the mtDNA control region in this taxonomic group.

In this study we used mtDNA to examine genetic diversity. A more detailed assessment of genetic diversity would require sequencing regions of the nuclear genome and developing microsatellites to estimate genotypic diversity from multiple independent loci. However, mtDNA diversity provides a reasonable proxy for nuclear diversity in marine fishes [58] and therefore we expect that the angelfishes in this study do not have low nuclear DNA diversity. Research on other coral reef fishes has used mtDNA control region and nuclear DNA and both markers produced comparable results: endemics have higher genetic diversity than widespread congeners [50]. The hypotheses of this study did not test if endemics and isolated populations of angelfishes had high genetic diversity, but rather tested if they had low genetic diversity (as per expectations based on terrestrial patterns). Despite the limitations of this study, we are confident that these hypotheses can be rejected (that is, endemics and isolated populations of angelfishes do not have low genetic diversity). Even though we did not sample central populations of the widespread species, it is evident just from sampling the isolated populations that isolated populations do not have low genetic diversity. Both nucleotide and haplotype diversity in the isolated populations are among the highest seen for reef fishes and approach the maximum possible.

This study found that geographically remote populations of endemic and widespread angelfishes had high genetic diversity and thus did not conform to terrestrial patterns of low genetic diversity. Endemic terrestrial species usually have low genetic diversity because they often have low abundance [1,9]. However, endemic reef fishes frequently have high abundance (reviewed in [41]) and therefore they may not have low genetic diversity. In the terrestrial environment, isolated populations of widespread species usually have low genetic diversity because they have small populations and low gene flow. However, the dispersive larval phase of reef fishes increases the potential for genetic exchange over large areas and this gene flow may be sufficient to maintain high genetic diversity in isolated populations [21,22,23]. Further research on the genetic diversity of endemic marine species and populations at remote locations is required to determine if there are differences in geographic patterns of genetic diversity between marine and terrestrial species. Determining patterns of genetic diversity is important because if geographically remote populations of endemic and widespread marine species have high genetic diversity than this should increase their evolutionary potential to adapt to escalating impacts [17,18].