California's 8 Channel Islands host a large diversity of insects, the vast majority of which are shared with mainland southern California. The existence of a small number of recognized endemic species, however, suggest that, for some lineages, the islands are isolated enough to have permitted significant differentiation. Here we investigate the phylogeographic relationships of 4 beetle species (Thinopinus pictus, Hadrotes crassus, Hypocaccus lucidulus, and Nyctoporis carinata): all occurring on the mainland and on multiple (up to 6) Channel Islands. Sequences of the cytochrome oxidase I mitochondrial gene (and, for one species, an intron in the nuclear guftagu gene) are analyzed by Bayesian, haplotype network, and population genetic methods to examine relationships and gene flow among island and mainland populations. In no instances were all island populations resolved to be monophyletic, and northern (Santa Cruz, Santa Rosa, San Miguel) and southern (San Nicolas, San Clemente, Santa Catalina) island groups generally showed separate relationships to the mainland. Northern island populations of Hy. lucidulus were also found to be closely related to those on the southern island of San Nicolas. Populations on San Clemente and Santa Catalina islands did not show close relationships to each other or to San Nicolas Island populations in any species. San Clemente and especially San Nicolas islands hosted disproportionately high levels of diversity in all species examined. This study suggests that the Channel Islands do not function as a biogeographical unit and that several of the islands exhibit levels of diversity comparable to, or even exceeding, similarly sampled populations on the mainland. Thus, as an insular refuge from southern Californian development, the Channel Islands constitute a center of high conservation importance.
As with island systems around the globe, California's Channel Islands have attracted a great deal of scientific attention. Because islands serve as discrete natural laboratories, their evolutionary and biogeographic histories have been the subject of considerable study and have provided the foundation for key advan ces in modern biology—most obviously as Darwin's fundamental inspiration for the theory of natural selection (Darwin 1859, Losos and Ricklefs 2009), as well as a supporting source for Ernst Mayr's transformative ideas on the evolutionary process (Emerson 2008) and for general theories of diversity-area relationships (MacArthur and Wilson 1967).
Islands are often a natural laboratory in which to study the effects of serious anthropogenic disturbance (Gillespie and Roderick 2002). The California Channel Islands unfortunately share this attribute as well. These islands have a long history of human use, including some of the oldest occupied human settlements in the New World (>13,000 YBP; Reeder et al. 2008), intensive historic ranching, and ongoing recreational and military use. Invasive species—including plants, vertebrates, and invertebrates—have taken a substantial toll on the native species of all the islands (Powell 1994, Junak et al. 1995, Wetterer et al. 2000, Knowlton et al. 2007). Though recent efforts to remove invasive species and restore native habitats are alleviating some threats, the ability of the native systems to recover is uncertain. In addition to threats from invasive species, island populations may be at substantial risk from the effects of climate change because their isolation restricts their abilities to respond to shifts in habitat zones.
California's Channel Islands comprise 8 islands off the coast of southern California, ranging from 2.9 to 249 km2 in size and from 20 to 98 km (Fig. 1) in distance from the mainland. Northern (Anacapa, Santa Cruz, Santa Rosa, San Miguel) and southern (San Nicolas, Santa Barbara, San Clemente, Santa Catalina) island groups are generally recognized. Although much of the geological history of the islands remains unclear, certain details are well established. Most significant is that none of the islands has had any direct mainland connection since their most recent complete emergence from the sea (Wenner and Johnson 1980). San Clemente, Santa Catalina, Santa Cruz, and Santa Rosa have had emergent land area since sometime in the Pliocene (2–5 MYBP), whereas San Nicolas, Santa Barbara, Anacapa, and San Miguel were most likely completely submerged at some point during glacial fluctuations in the earlier half of the past 500,000 years (Sorlien 1994, Dibblee and Ehrenspeck 2002). However, during the most recent glacial maximum (about 17,000–18,000 YBP; Vedder and Howell 1980), when sea levels were lowered by as much as 120 m, all 4 of the northern islands were joined into a single super-island (Santarosae), which was separated by as little as 6 km from the mainland to the immediate east (Wenner and Johnson 1980). Their connection might have persisted until as recent as 9000 YBP (Porcasi et al. 1999, Kennett et al. 2008). At this time, all other islands were larger and closer to each other as well as to the mainland. Thus, there has been increased opportunity for movement to and among islands in recent times.
California's Channel Islands are sufficiently isolated to host substantial numbers of endemic taxa. These include plants (Philbrick 1980, Junak et al. 1995), vertebrate animals (Knowlton et al. 2007), and invertebrates (Miller 1985, Rubinoff and Powell 2004). The origins of these endemic species have been varied, with some species representing ancient relicts (e.g., the island ironwood, Lyonothamnus floribundus, whose fossil record places it formerly at numerous locations in the mainland southwest) and others representing recent offshoots of extant mainland species (such as the island fox, Urocyon littoralis, a close relative of the gray fox, Urocyon cinereoargenteus).
Among insects, the level of endemicity on the Channel Islands is an open question. None of the 8 islands' insect faunas are well surveyed (nor, for that matter, is much of the adjacent mainland; Caterino 2006), and the poor state of taxonomy of many groups precludes confident assessment of true endemism. Nonetheless, a wide range of origins has been noted among insects, with some endemic Lyonothamnus-feeding Lepidoptera representing presumably old relicts (Powell 1994). Endemic Orthoptera mostly represent close relatives to mainland species in genera that tend to show local endemism (e.g., Neduba and Cnemotettix; Rentz and Weissman 1982, Weissman 1985), and this is also true for a recently described endemic beetle (Actium vestigialis; Caterino and Chandler 2010). Others are endemic only at the subspecies level, indicating even closer relationships to mainland forms (Miller 1985).
The biogeographic sources of few native island taxa have been extensively explored. In an analysis of the Lepidoptera faunas of the islands, Powell (1994) found that relationships to the mainland closely reflected overall habitat distribution, with the faunas of drier habitats on the islands closely resembling those of the mainland deserts to the southeast and taxa in more mesic parts of the islands resembling mainland faunas to the north. These similarities suggest no dominant source but rather that Lepidoptera migrants have found their way to suitable habitats on the islands from multiple directions. In contrast, the vast majority of Orthoptera island populations and species find their conspecifics or near relatives in the immediately adjacent Santa Monica Mountains (Weissman 1985).
Besides the question of taxonomic endemicity, Channel Island insects “often vary slightly in color, sculpture, or size from conspecific mainland specimens” (Miller 1985). This variability suggests that some level of cryp tic endemicity may be hidden by a conservative taxonomy. In a comparable case, the Island Scrub-Jay (Aphelocoma insularis), long considered a morphologically distinguishable subspecies of the widespread Western Scrub-Jay (Aphelocoma californica), has recently been shown to be highly divergent from any other west ern Aphelocoma (Delaney and Wayne 2005, Delaney et al. 2008). Although such work has been limited, studies on other taxa have generally shown significant divergence of island populations from mainland conspecifics and limited contact among populations on the different islands (Ramirez and Beck witt 1995, Landry et al. 1999, Eggert et al. 2004, Rubinoff and Powell 2004, Wilson et al. 2009). In addition, many species present on the Chan nel Islands have reached there in very recent times, whether through natural means or through human trans port (Powell 1994, Calderwood et al. 2002, Mahoney et al. 2003, Chatzimanolis and Cate rino 2007a). It is likely that various insect species lie along this entire continuum, including deeply divergent en demics, occasional natural colonists, and recent anthropogenic introductions.
The beetles (Coleoptera) of the Channel Islands have received some previous attention (Miller 1985, Nagano 1985), and this insect order has been the focus of more intensive recent work through our California Beetle Project (Santa Barbara Museum of Natural History 2013). Of 225 beetle species previously recorded from the Channel Islands (Miller personal communication), 38 (17%) are considered to be endemic to one or more islands (Miller 1985). Our recent surveys, however, call these numbers into question: over 640 named species of beetles have now been recorded from the Channel Islands (Caterino et al. unpublished), in addition to dozens of additional morphospecies. Many of these morphospecies are in genera inclined toward local endemism on the mainland, and a sizeable proportion of these are likely to be undescribed endemics. More clearly understanding the origins of the insect fauna of the Channel Islands is critical to managing these potentially rare species and restoring island ecosystems.
In this paper, we analyze phylogeographic patterns among 4 beetle species, all of which occur on the Channel Islands and on the adjacent mainland: Thinopinus pictus LeConte, Hadrotes crassus (Mannerheim) (Staphylinidae), Hypocaccus lucidulus (LeConte) (Histeridae), and Nyctoporis carinata LeConte (Tenebrionidae). We previously studied the distribution of genetic diversity and phylogeographic relationships for Hy. lucidulus and N. carinata on the mainland (Chatzimanolis and Caterino 2008, Caterino and Chatzimanolis 2009, Polihronakis and Caterino 2010a). The other 2 are examined here for the first time. Thinopinus pictus and Ha. crassus are associated with beach wrack in intertidal zones; Hy. lucidulus is associated with coastal dunes; and N. carinata is found in inland, fully terrestrial habitats, where it is associated with fungus on dead wood. Of these 4, only Hy. lucidulus is capable of flight, permitting possible dispersal to and among islands by air. The other species would presumably be limited to dispersal by rafting, whether via beach wrack (e.g., Peck 1994) or extreme coastal flooding events (Wenner and Johnson 1980). There is also the possibility that any of the species might be anthropogenic introductions.
We use data from these species to address several questions of Channel Island biogeography concerning the origins, ages, and uniqueness of populations inhabiting the islands. Specifically, we examine phylogeographic relationships among haplotypes on the various islands and adjacent mainland and use phylogenetic trees to infer colonization frequency and sources for the islands. We use diversity statistics to compare the levels of intraspecific diversity found on the islands to that found in mainland populations. Finally, we use population genetic analyses to assess degree of isolation of populations among islands and between various subsets of the islands and the mainland. Together these analyses will help us address synthetic questions, such as whether the island populations represent endemic radiations and if not, how many colonizations of the islands are necessary to explain the current distributions of each species; whether island populations have similar mainland sources, and what can be inferred from these about routes and means of colonization; whether there is evidence for ongoing contact between island and mainland populations; and whether island relationships reflect purported geological relationships. These results will provide a novel perspective on the assembly and conservation value of the insect fauna of this interesting group of islands.
Methods
Sampling: Taxa and Areas
Focal species were selected based on broad distributions that included the mainland and enough islands to allow multiple comparisons, as well as abundance sufficient to obtain meaningful sample sizes. They also represent a diversity of habitats and life histories, permitting some exploration of relations between these factors and degree of phylogeographic structure. For most of these species our samples represent only a portion of their total distribution but most of the areas that might conceivably be related to island populations. The overall ranges of Thinopinus pictus and Hadrotes crassus extend from British Columbia, Canada, to northern Baja California, although the exact limits are poorly documented. The distribution of Hypocaccus lucidulus is somewhat more restricted, extending north only into Oregon and south into Baja Cali fornia. Most of the range of Nyctoporis carinata, a species confined to California, is represented.
The samples used were gathered over several field trips. Most specimens were collected directly into 100% ethanol and stored at -70 °C. Where possible, we collected 10 individuals of each species from each island. Where speci mens were available from multiple localities within an island, we selected a total of 10 samples with roughly equivalent numbers of sam ples from each sublocality and treated each island as a single population throughout. For studies on the California mainland, 10 individuals per species has been adequate to assess patterns of relationship and interregional diversity in several previous studies (Chatzimanolis and Caterino 2007b, 2008, Polihronakis and Caterino 2010a). The total number of samples of each species from each island is given in Table 1. DNA was extracted from each specimen using Qiagen's DNeasy tissue kit (Valencia, CA), with an identifiable voucher specimen mounted, labeled, and assigned a unique “California Beetle Project” catalog number. Full locality and voucher information are available online ( http://www.sbcollections.org/cbp/cbpdatabase1.aspx). We used previously published data for several taxa and for most mainland samples. These sequences correspond to GenBank entries: EU179681–EU179712 for Hypocaccus lucidulus COI, EU037099–EU037189 and GU049332–GU049339 for N. carinata COI, and GU049270–GU049331 for N. carinata GFT. GenBank accession codes for all newly generated sequences are given in the appendix.
Sampling: Genes
For all 4 species, we generated sequences of the cytochrome oxidase subunit I (COI) gene. This well-studied mitochondrial protein coding gene has been used in a large number of phylogeographic and phylogenetic analyses, especially among insect and arthropod groups (Caterino et al. 2000). We analyzed a fragment approximately 826 bp long; most were amplified using primers C1-J-2183 and TL2-N-3014 (aka Jerry and Pat; Simon et al. 1994). In a few cases, other primers were necessary for successful amplification, but the resulting fragments were trimmed to this length. Though mitochondrial DNA is known to have some limitations for phylogeographic inference (e.g., rapid coalescence times, maternal lineage bias, and inability to detect hybridization; Irwin 2002, Zhang and Hewitt 2003), it has proven useful in establishing preliminary hypotheses of intraspecific phylogenetic histories. A previous study of Nyctoporis carinata (Polihronakis and Caterino 2010a) established the nuclear intron GFT (in the guftagu gene) as informative for population relationships, and we have added sequences of GFT for the island populations of this species as well. GFT haplotypes (alleles) were phased manually for individuals with one polymorphic site. Specimens with 2 or more polymorphic sites were phased using the program Phase v2.1 (Stephens et al. 2001, Stephens and Scheet 2005). Haplotype designations follow Polihronakis and Caterino (2010a). GFT primer sequences are available in Polihronakis and Caterino (2010a) or may be obtained from the authors.
Analyses
Our most basic questions focused on the number of colonizations of the islands from the mainland, relying on a robust phylogeny of island and mainland haplotypes. All sequences for each species were filtered for unique haplotypes using Collapse (ver. 1.2; Posada 2006). Phylogenetic trees of unique haplotypes of all species were generated in MrBayes (ver. 3.1.2; Huelsenbeck and Ronquist 2001). Models were estimated using MrModelTest; the best model was selected as indicated by the Akaike information criterion (AIC). For all MrBayes runs, we ran 4,000,000 generations, with 4 chains, 3 of them heated with the temperature at 0.2, sampling every 1000 generations. The first 25% of the resulting trees were discarded as burn-in, and a majority-rule consensus was generated from the remainder, with consensus indices as posterior probabilities.
Trees were rooted with outgroups (with GenBank accession numbers) as follows: T. pictus rooted with the old world Hadropinus fossor Sharp (GU380341) and Hadrotes crassus (GU226635), both close relatives within the subtribe Staphylinina; Ha. crassus rooted with Hadropinus fossor and T. pictus (GU226619); Hy. lucidulus rooted with congener Hy. bigemmeus (LeConte) (GU380342 and GU380343); N. carinata rooted internally following the results of Polihronakis and Caterino (2010a) and lacking close relatives. Island colonizations were reconstructed on the complete rooted Bayesian topologies by using a binary character in parsimony (using MacClade v 4.06; Maddison and Maddison 2003), scoring each haplotype as present on the mainland or the islands or both (polymorphic). Given the low divergences among some haplotypes and the possibility of direct ancestor-descendant relationships among some, we also used TCS to generate parsimony network topologies (Clement et al. 2000) with all individuals (except outgroups) included.
The relative genetic diversities of island versus mainland populations were assessed using several measures of gene diversity and haplotype richness. Within each species we examined the contribution of each island population to total genetic diversity by using C statistics (Petit et al. 1998) calculated with the program Contrib (ver. 1.02; Petit 2006). These statistics use haplotype frequency data, corrected via rarefaction, to assess relative diversity and differentiation of population samples. We report corrected allelic richness r(N), where N is the minimum number of samples in any population, and CT is the relative contribution of each population to total allelic richness. Where multiple mainland populations were available, they were grouped into the following categories: “northern” for north of Point Conception; “central” for Point Conception to Palos Verdes Peninsula; and “southern” for south of Palos Verdes. For inland populations (N. carinata only), these groups included “northern” for Sierra Nevada Mountains, Tehachapi Mountains, Santa Lucia Mountains, Northwest and Central Transverse Ranges, and Santa Ynez Mountains; “central” for Sierra Pelona, San Gabriel Mountains, and San Bernardino Mountains; and “southern” for San Jacinto Mountains. Finally, to assess level of sequence divergence (or phylogenetic distinctness), we calculated nucleotide diversity (p) for each population using Arlequin (ver. 3.1.1; Excoffier et al. 2005).
To examine phylogeographic structure among islands and between islands and the mainland, we conducted AMOVA analyses in Arlequin under 4 alternative models: a 2-group model (all island populations vs. all mainland populations); a 3-group model (northern islands/southern islands/mainland); a 4-group model (northern islands/mainland north of Pt. Conception/southern islands/mainland south of Pt. Conception); and an n + 1 model, separating all n islands occupied by each species plus the mainland. To detect and localize possible interisland and island-mainland connections, we calculated PhiST values (also in Arlequin) with all populations considered independent (including however many mainland populations were available). Interpopulation divergences were based on Tamura-Nei corrected distances.
Results
Phylogenetic Patterns
Nyctoporis carinata.—In Nyctoporis carinata, the mitochondrial tree (Fig. 2) indicated a single colonization event for the northern islands (Fig. 3) and 2 separate colonizations for Santa Catalina Island. The 16 haplotypes occurring on the northern islands were interspersed, and individual populations did not appear to have been isolated long enough to achieve monophyly. Santa Cruz Island exhibited the greatest diversity, with 9 individuals each possessing a unique haplotype. Haplotypes on Santa Rosa and San Miguel showed much lower diversity. Those on San Miguel showed only direct relationships to ones present on Santa Rosa, reflecting their closer geographic proximity. The mainland origin for this lineage could not be specified very precisely. The closest mitochondrial haplotype occurs in the central portion of the mainland (specifically at Pine Mountain in the central Transverse Ranges), but these were separated by >10 changes and were not connected at the 95% confidence level in the haplotype network. The nuclear haplotype G04 is found on all the northern islands, as well as in northern and central mainland areas. Santa Cruz and Santa Rosa islands shared a unique GFT allele, with a closely related one restricted to San Miguel. Despite its much slower rate of evolution, the nuclear gene also indicates that the northern island lineage has been present long enough to have evolved multiple unique variants.
The Santa Catalina population of N. carinata represented 2 distinct colonizations from separate mainland lineages (Fig. 3), although both appear to have originated from the central region (specifically the San Gabriel Mountains), an area hosting a high diversity of both nuclear and mitochondrial lineages (see Polihronakis and Caterino 2010a). In the COI haplotype network, these 2 lineages were not connected at the 95% confidence level, although individuals exhibiting these haplotypes were intermingled at sampling sites on Santa Catalina. Three distinct nuclear haplotypes were present on Santa Catalina: one shared with northern and central mainland regions, one shared among all mainland regions, and one unique to the island.
Hadrotes crassus.—Hadrotes crassus exhibited 2 quite divergent clusters of mitochondrial haplotypes (Fig. 2), requiring 3 island colonizations (Fig. 3) total to explain their distribution on the Channel Islands. The largest cluster was dominated by a widespread haplotype (haplotype 1) present in the northern mainland region and all islands sampled (Santa Cruz and Santa Rosa, Santa Catalina, San Nicolas and San Clemente). Several similar haplotypes are found on Santa Cruz and Santa Rosa islands, one of which is also present in the central mainland region (Santa Barbara County). A single haplotype from far north (haplotype 8, Sonoma County) on the mainland was found to be directly related to a highly diverse assemblage (5 haplotypes for 7 individuals) of haplotypes from San Clemente Island. A single Santa Rosa Island haplotype was also closely related to this northern one. Our sampling to the north was relatively sparse, however, so it remains to be seen whether this distant relationship might be bridged by intervening samples.
Fig. 2.
Haplotype networks for all species, including networks for both COI and GFT for Nyctoporis carinata. Where multiple subnetworks for a gene/species could not be connected at the 95% confidence level, the relationships among them is indicated by an adjacent Bayesian phylogeny (rooted, although root not shown). In Bayesian trees, branch lengths are to the same scale among species but not between trees and networks (where changes are specifically indicated by small filled circles). Asterisks are shown for Bayesian branches relevant to island lineages that are supported at ≥⃒95% posterior probability. Colors indicate localities of origin for individuals, with white and grays indicating mainland localities. Larger circles are proportional to numbers of individuals with each haplotype. Numbers within or next to circles correspond to haplotype numbers given in Table 1. Haplotype “Hb” in the Hypocaccus lucidulus network represents putative outgroup Hypocaccus bigemmeus.
A second smaller cluster in Ha. crassus comprised 2 haplotypes: one (haplotype 2) was very widespread, again covering northern and southern islands as well as central and northern mainland areas. These 2 divergent clusters were weakly linked by a single haplotype from Santa Cruz Island (haplotype 17), which was not closely related to anything else.
Fig. 3.
Summary diagram showing hypothesized connections among islands and mainland for Nyctoporis carinata, Thinopinus pictus, Hadrotes crassus, and Hypocaccus lucidulus. Arrows represent reconstructed colonization events where mainland origin was unambiguous, as revealed by parsimony mapping on phylogenetic relationships among mtDNA haplotypes. Colored arrows are coded by species (see inset).
Thinopinus pictus.—Most haplotypes of Thinopinus pictus were fairly closely related (Fig. 2), although the 2 haplotypes found on Santa Catalina Island were very divergent from anything else. The distribution of these haplotypes required at least 3 island colonizations (Fig. 3), though the source of the Santa Catalina population cannot be reconstructed. The 3 northern islands sampled mostly formed a tight cluster, with Santa Cruz and Santa Rosa islands sharing haplotypes 5 and 7, the former also found on San Miguel Island. Each northern island also hosted unique haplotypes. The closest mainland haplotypes to these northern islands were found in the central region, specifically along the Santa Barbara County coast. San Nicolas Island exhibited high diversity, with the 10 individuals exhibiting 6 closely interrelated haplotypes. The closest relatives to these are found exclusively in the northern mainland region, specifically in San Luis Obispo County.
Hypocaccus lucidulus.—This species exhibited limited phylogeographic structure (Fig. 2), with at least 6 colonization events required to explain their distribution on the Channel Islands (Fig. 3). Haplotype diversity was very high (52 haplotypes for 126 individuals), and these were resolved into 3 wellsupported lineages (resolved as 3 disconnected networks by TCS). However, despite well-supported relationships among haplotypes within these lineages, haplotype relationships showed little concordance with geography. One large lineage, dominated by haplotypes from north of Point Conception, was resolved as sister to the re maining two. This lineage included haplo types from both northern (Santa Rosa and San Miguel) and southern (San Nicolas) islands, with the central mainland sharing a haplotype with the northern islands. A small haplotype group was found predominantly on San Clemente Island (haplotypes 45–47), with a distant relative found in the northern parts of the mainland. The largest haplotype lineage in cluded the majority of haplotypes from the central part of the mainland (Santa Barbara, Ventura, and Los Angeles counties), a smaller proportion from farther north on the mainland, the single southern mainland haplotype (from San Diego County), the majority of haplotypes found on the northern islands, and an additional common haplotype from San Clemente Island. Haplotype 10 in this cluster was found on Santa Cruz and Santa Rosa, with the closely related haplotype 42 on San Nicolas. Haplotype 9 was shared by Santa Cruz and San Nicolas islands, with the closely related haplotype 36 on Santa Rosa. Haplotype 44 from San Clemente Island was closest to a mainland haplotype found in the central region (the Los Angeles and Ventura county coasts, haplotype 22). The central mainland region is also directly related to single haplotypes from Santa Rosa (haplo type 37) and San Miguel (haplotype 38) islands. Finally, 2 unique haplotypes from Santa Cruz Island (haplotypes 49–50) were most closely related to isolated haplotypes occurring north of Point Conception on the mainland.
Table 2.
Results of Contrib analyses, with corrected allelic richness (r[N]), relative contribution of each population to total allelic richness (CT), and nucleotide diversity (p). Multiple values for nucleotide diversity represent values for all populations lumped into northern and central mainland groups.
Diversity Patterns
Channel Islands populations of most species showed little indication of reduced genetic diversity compared to mainland populations that were sampled. Rather, diversity was high in nearly all species studied, by all measures (Table 2). San Nicolas Island especially emerged as hosting unusually high genetic diversity, exhibiting higher allelic richness (corrected by rarefaction for sampling uneven ness) and nucleotide diversity than any other population, island or mainland, for both Hypo caccus lucidulus and Thino pinus pictus. In Hadrotes crassus, though the northern mainland populations led in these diversity measures, the San Clemente Island population ranked a close second. The San Clemente Island population also ranked second in nucleo tide diversity for Hy. lucidulus, although both central and northern mainland populations showed higher allelic richness. In nucleo tide diversity, Santa Rosa and Santa Catalina island populations of T. pictus ranked second and third, respectively, although the northern mainland population exceeded Santa Rosa in allelic richness. Of all these species, only in N. carinata did island populations show lower genetic diversity than most mainland populations.
Table 3.
PhiST values for island—island population comparisons and island—mainland population comparisons. Asterisks indicate significant isolation between areas. Samples were not available for comparisons marked N/A. Mainland areas marked as “Montane” are noncoastal.
Population Structure Among Island/ Mainland Populations
The phylogeographic structuring among islands and species was predominantly high. Based on pairwise PhiST analyses (Table 3), we found significantly restricted gene flow among most islands in most species. The exceptions mostly involved pairs of northern islands. In N. carinata, though Santa Cruz and Santa Rosa Islands were significantly separated, Santa Rosa and San Miguel were not. In Ha. crassus, most significant comparisons involved Santa Rosa Island (especially in comparison with the southern islands), whereas Santa Cruz Island was not significantly isolated from several other islands, including most of the southern islands. Finally, in Hy. lucidulus, Santa Cruz and Santa Rosa Islands were not significantly separated, though all other interisland comparisons showed significant isolation.
Table 4.
Results of AMOVA analyses, with groups as defined in text. Asterisks indicate significance. Results are shown separately for the 2 markers examined for Nyctoporis carinata. Some tests were inapplicable to some species depending on sampling.
Given the significant restriction in gene flow among most individual islands, it was interesting that some island-mainland population pairs were not significantly isolated according to PhiST. Some ongoing connectivity is indicated among mainland and insular populations in 2 of the 4 species. The lowest level of island isolation was found in Ha. crassus, with only extreme northern populations showing significant isolation from any Channel Island population. In Hy. lucidulus, PhiST indicated that central mainland (Santa Barbara County) populations were not significantly separated from those on the northern islands. Island populations of N. carinata, on the other hand, were all significantly isolated from any mainland population, as were those of T. pictus.
Very few of the 4 alternative groupings of islands or island-mainland populations tested indicated significant similarity according to AMOVA (Table 4). There was no significant among-group variation in any of the 2-group comparisons (mainland vs. island), and these values were <10% for all species. When the southern islands were separated in the 3-group comparisons, the among-group variation increased in all cases, although it remained small (<12%) relative to variation within and among populations in all species except N. carinata. The only significant among-group variation seen in N. carinata was in the 3-group scenario, which was likely driven by deep divergence among mainland populations, as reported in Polihronakis and Caterino (2010a). In general, among-group variation increased with greater partitioning among mainland populations, and few groupings of the island populations captured patterns of variation well.
Discussion
Substantial work has been done on relationships of organisms on California's Channel Islands, much of it with the goal of determining the conservation status of particular islands' populations. In general, previous work has found many island populations to be highly divergent from each other and from mainland populations. In the current study, patterns of relationships are sufficiently varied to preclude inference of general patterns. None of the 4 widespread species examined here resolve island populations to be collectively monophyletic. All species require multiple colonization events to explain the distribution of haplotypes, from at least 2 or 3 coloni zations in Ha. crassus, T. pictus, and N. carinata to at least 6 in Hy. lucidulus. In the latter case, there appear to have been back-colonizations to the mainland as well (or, less parsimoniously, several more island colonizations than this).
The only predictable and consistent pattern from previous work is the generally close relationship among populations on the northern islands of Anacapa, Santa Cruz, Santa Rosa, and San Miguel—expected on the basis of their Pleistocene unity as Santarosae. These islands share some endemic taxa (the tortricid moth Argyrotaenia franciscana insulana: Landry et al. 1999; the slender salamander Batrachoseps pacificus: Jockusch and Wake 2002) and exhibit close population relationships in more widespread species in a variety of taxa, including dune spiders (Lutica: Ramirez and Beckwitt 1995), deer mice (Peromyscus maniculatus: Ashley and Wills 1987, 1989), side-blotched lizard (Uta stansburiana: Mahoney et al. 2003), Loggerhead Shrike (Lanius ludovicianus: Eggert et al. 2004, Caballero and Ashley 2011), and marine eelgrass (Zostera pacifica: Coyer et al. 2008).
The species examined here largely support this pattern, most distinctly in N. carinata where the 3 northern islands sampled (Santa Cruz, Santa Rosa, San Miguel) form a strongly supported, divergent clade unrelated to populations on the one southern island sampled, Santa Catalina. These 3 northern islands also form a distinct cluster in T. pictus. In Ha. crassus and Hy. lucidulus, however, such relationships, though evident, are complicated by low levels of divergence or high migration rates or both. Furthermore, all northern islands have unique mitochondrial haplotypes in all species sampled. So if shared haplotypes are indicative of recent ancestry, there is also evidence that significant evolution has occurred since isolation. On the other hand, it is also possible that the super-island Santarosae exhibited within-island phylogeographic structure prior to separation into the modern islands. Similar isolation-bydistance patterns have been detected within islands, notably within the plant Lithophragma maximum, which is endemic to and highly variable within San Clemente Island (Furches et al. 2009). Clearly the presence of many taxa on the northern Channel Islands predates the most recent separation of these islands, and the sharing of haplotypes/alleles among them cannot be definitively attributed to either modern or historical connections.
The southern islands (Santa Catalina, San Clemente, and San Nicolas) show much lower phylogenetic coherence in the species studied here. Previous studies have suggested relationships between Santa Catalina and San Clemente islands, in particular eelgrass (Coyer et al. 2008), side-blotched lizards (Mahoney et al. 2003), Loggerhead Shrikes (Eggert et al. 2004), and one species of dune beetle (Chatzimanolis et al. 2010). The species examined here show little indication of southern island relationships.
San Nicolas Island shows a close relationship to one or more northern sources in a substantial number of taxa. Several previous studies have supported relationships to the northern islands for a variety of taxa, including tortricid Lepidoptera (Landry et al. 1999, Rubinoff and Powell 2004), deer mice (Ashley and Wills 1987, 1989), eelgrass (Coyer et al. 2008), dune spiders (Ramirez and Beckwitt 1995), and dune beetles (Chatzimanolis et al. 2010). Some Hy. lucidulus haplotypes show a similar San Nicholas-northern island relationship. In other cases, the San Nicolas Island populations appear most closely related to mainland populations well to the north of the Channel Islands. Even more disjunct relationships are shown in T. pictus, where San Nicolas haplotypes are most closely related to some from north of Point Conception. This pattern is also evident in Hy. lucidulus, where a haplotype is shared between San Nicolas Island and northern mainland (San Luis Obispo) populations. The most interesting aspect of these northern relationships for San Nicolas Island is that in most cases they appear to be very recent, involving very closely related or even identical haplotypes. This finding seems to point to a direct colonization route in line with the California Current, which follows the coast southward to Point Conception, continuing directly southward past the western edge of San Miguel Island, then southeastward toward San Nicolas Island rather than continuing along the coastline. Where relationships to the northern islands are observed, they may owe something to a closer proximity of San Nicolas to the northern islands during depressed sea levels (-120 m) in the Late Pleistocene (be tween 10–20,000 YPB; Vedder and Howell 1980).
In general, relationships of the island populations to the mainland cannot be attributed to any consistent source. Shared haplotypes and close phylogenetic relationships between the mainland north of Point Conception and one or more of the islands predominate in the species examined here. However, it is difficult to separate this result from sampling bias in coastal species in more northerly mainland areas—coastal habitats that have unfortunately been severely degraded by human activities. Still, our sampling from the Santa Barbara and Ventura areas is relatively strong, and these mainland areas closest to the northern islands have not contributed to island populations as strongly as areas farther north. This difference is surprising not only because of modern proximity, but because of the very small Pleistocene gap (<6 km) between eastern Santarosae and the mainland directly to the east (Powell 1985, 1994). Furthermore, continued colonization avenues from these near-shore sources have been documented, particularly by direct rafting following seasonal coastal floods (Wenner and Johnson 1980). Where northern relationships have been noted, they have been considered a relictual link to a wetter past (Raven 1967, Powell 1994). Our results, especially where haplotypes are shared, show a much more re cent, probably continuing, route for gene flow from northern populations (especially Hy. lucidulus and Ha. crassus).
Results of our own previous work involving Channel Islands beetles span much of the range reported here, even though until now these results have included only Santa Cruz Island samples. Several species have shown the high levels of island-mainland migration seen here in Hy. lucidulus, revealing multiple unrelated colonizations and source areas. These include Cercyon fimbriatus, an inhabitant of coastal wrack (Chatzimanolis and Caterino 2008); Stictotarsus striatellus, an inhabitant of freshwater streams and ponds (Short and Caterino 2009); and Calathus ruficollis, a widespread inhabitant of drier terrestrial environments (Chatzimanolis and Caterino 2007a). In the latter (a flightless species), we suspected anthropogenic introductions to account for at least some of the apparent island diversity. Other species previously examined have shown more-restricted patterns of genetic variation on and among islands. Island popu lations of the freshwater aquatic Anacaena signaticollis (Santa Cruz) and the terrestrial fungus-feeding Phloeodes plicatus (Santa Cruz, Santa Catalina) represent shallowly independent lineages (the latter separate on the 2 islands; Short and Caterino 2009, Polihronakis and Caterino 2010b). Santa Cruz and Santa Rosa Islands host a deeply divergent lineage of the litterinhabiting weevil Geodercodes latipennis (Polihronakis et al. 2010). Thus the varied results of the present study are probably representative of a wide variety of species histories on the Channel Islands.
Many authors have sought explanations of phylogeographic structure in life history attributes, generally habitat association (Ribera and Vogler 2000, Marten et al. 2006, Abellán et al. 2009) or behavior (especially, in insects, the ability to fly; Smith and Farrell 2006). The underlying commonality in such explanations is dispersal propensity (Avise 1994). Our results do show some similarities with other studies of island and coastal taxa (especially Papadopoulou et al. 2009) in that the species most closely associated with coastal environments— Ha. crassus, T. pictus (both associated with intertidal wrack), and Hy. lucidulus (associated with coastal sand dune)—show lower levels of phylogeographic structure than the lone representative of more interior habitats, N. carinata. This pattern is presumably due to the highly dynamic nature of coastal habitats, which are subject to frequent short- and long-term disturbance (tidal and sea level fluctuations, respectively). Though this study was not designed to test the association of microhabitat and phylogeographic structure, previous work on the California mainland found generally similar patterns in these habitats (Chatzimanolis and Caterino 2008 vs. Caterino and Chatzimanolis 2009). Our results with respect to flight capability also agree with these observations: the lone flying species represented here, Hy. lucidulus, does indeed show the largest number of island colonizations as well as the lowest interisland and island-mainland structuring. However, species within each of these categories often show extremely wide variance when multiple similar species are examined (results herein; Chatzimanolis and Caterino 2008, Caterino and Chatzimanolis 2009, Short and Caterino 2009), and it is apparent that any single ecological predictor provides only a rough expectation of population structure. Species-specific attributes and population size certainly play important roles as well.
Finally, high diversity was apparent on some islands, both in absolute terms given their areas and relative to mainland populations of similar extent (contra Frankham 1997). Our genetic contributions analyses reveal that several islands host exceptionally high levels of haplotype diversity, most notably San Clemente and San Nicolas. The latter is particularly remarkable if we accept its complete submergence in fairly recent time (<500,000 years). The levels of diversity found on San Nicolas in T. pictus, Ha. crassus, and Hy. lucidulus (as well as Coelus pacificus; Chatzimanolis et al. 2010) pose a challenge for this hypothesis. If its biota truly is that recent, the island has supported exceptional diversification over a short span of geological time. Some previous work on this subject has expressed concern with the potential loss of variability due to population depression during times of excessive grazing on the Channel Islands (Wallace and Helenurm 2009), but neither those results nor ours reveal unexpectedly low diversity in the islands. Rather, the opposite may be the case. (Furches et al. 2009, Wilson et al. 2009).
The finding of high levels of unique genetic diversity on California's Channel Islands suggests 2 important, nonexclusive explanations. First, it is quite possible that contemporary mainland populations are in fact more significantly depressed due to anthropogenic impacts. Mainland populations, especially in coastal species, have been extirpated in many areas due to a combination of beach grooming, recreational use, and invasive dune plants (Oppewall 1976, Powell 1981, Dugan et al. 2000, Connor et al. 2002). Indeed, it was impossible to find many of our focal species in historically suitable areas in Los Angeles and Orange counties. Obviously, there is no way to know what sort of diversity these areas once harbored. A previous study focused on mainland populations of one of these species did not reveal significant depression in highly impacted areas (Chatzimanolis and Caterino 2008), but island samples were not available to provide the contrast now observed. Second, it is possible that barriers to gene flow among islands helped prevent a significant amount of homogenization that might have occurred among better-connected mainland populations. Also, avoiding selective sweeps that continuous populations might experience and having genetic drift functioning on separate isolates might help maintain higher overall diversity, despite the seemingly inevitable tendency toward lower diversity within each island. Additional studies with greater density of withinisland sampling will be necessary to begin to address these questions.
Despite many lingering and newfound mysteries, it is clear that, as a whole, California's Channel Islands host important levels of intraspecific diversity. Among the species examined here, there are several endemic lineages: most showing strong among-island diversification patterns and several exhibiting comparable or even greater diversity than conspecific (and similarly sampled) mainland populations. Given the patterns observed, it is clear that much of this diversity arose on the islands themselves. However, some of the islands' apparent diversity may also stand out as a result of declines in mainland populations in the southern part of the California Floristic Province. These and other potential explanations clearly justify further exploration, ideally with a multilocus approach to allow more detailed examination of biogeographic scenarios.
Acknowledgments
We thank Scott Miller for background information and inspiration to focus on the Channel Islands. For assistance with fieldwork and permission to conduct this research, we thank Channel Islands National Park (Ian Williams, Kate Faulkner, Mark Senning, Sarah Chaney, Carolyn Greene), the University of California Santa Cruz Island Reserve (Lyndal Laughrin), The Nature Conservancy, Catalina Island Conservancy, and the U.S. Navy (Grace Smith, Melissa Booker), specifically the Department of the Navy on behalf of the Naval Base Coronado, San Clemente Island Naval Auxiliary Landing Field. We thank Mitchell McGlaughlin for comments and insights on unpublished data, and we thank 2 anonymous reviewers for valuable comments on the preliminary manuscript. Finally, we thank Iliana Ouzounov and Leigh Ann Norris for laboratory assistance. This research was supported by a National Science Foundation grant (NSFDEB0447694) to M. Caterino and a grant from National Geographic Society's Committee for Research and Exploration to S. Chatzimanolis and M. Caterino.
