Multiple, independent colonizations of the Hawaiian Archipelago by the family Dolichopodidae (Diptera)

Taxonomic sampling

Specimens were collected from 2004 to 2012 from sites across the Hawaiian Islands. The bulk of Hawaiian Dolichopodidae species are endemic to high elevation (900–1,700 m.) rain forest habitats, and thus collecting efforts were concentrated in these areas. Other habitats (e.g., coastal strand, dry and mesic forests, alpine zone) were also sampled, including rocky beaches, the only known habitat of Conchopus, Thinophilus, Asyndetus and Hydrophorus. We succeeded in collecting specimens from 11 of the 14 Hawaiian dolichopodid genera with endemic species known from the islands (Campsicnemus, Conchopus, Thinophilus and eight genera from the Eurynogaster complex, Table S1A in Appendix S1). Data from the Hawaiian Campsicnemus are included here from a previous study from our group, and are described in Appendix A from Goodman et al. (2014). Material was collected by general sweeping of vegetation and leaf litter, pan and Malaise trapping, and hand collecting. To evaluate monophyly of and diversity within the Eurynogaster complex, we included representatives from each of its eight constituent genera (Table 1; Evenhuis, 2005). No Eurynogaster complex lineages were omitted from our sampling. All material was preserved in 95% ethanol.

All material was identified using the most recent key to species in Tenorio (1969) and Evenhuis (2005). Descriptions of new species from within the Eurynogaster complex discovered as a result of this project are in preparation. Unpublished new species included in the study were given letters (e.g., Eurynogaster n. sp. A, B, C, etc.). In addition to the extracted specimens, whenever possible, a series of conspecifics from the same site were also preserved in 95% ethanol. Voucher material has been deposited in the Bernice Pauahi Bishop Museum (Honolulu). In addition, new sequences were generated for outgroup specimens from the non-endemic Dolichopodidae: five specimens of Dolichopus exsul, two specimens of Chrysotus longipalpis, and one specimen each of Condylostylus sp. and Tachytrechus angustipennis. Finally, sequences from Hercostomus indonesianus were also downloaded from GenBank to include in the outgroup (see Table S1A in Appendix S1). Access and collection permits were granted by the State of Hawaii Department of Land and Natural Resources, the National Park Service (Hawaii Volcanoes and Haleakala National Parks), Maui Land and Pineapple, East Maui Irrigation, Parker Ranch, and The Nature Conservancy of Hawaii (Appendix S4).

Phylogenetic analysis

Divergence time estimation in the Eurynogaster complex

To estimate the age of the Eurynogaster complex lineage, divergence time estimation was performed on dataset B using a Bayesian relaxed-clock method implemented in BEAST 1.7.5 (Drummond et al., 2012) on CIPRES (http://www.phylo.org: Miller, Pfeiffer & Schwartz, 2010). The age of the Eurynogaster complex is unknown as representatives of the genus are not known outside of Hawaii and biota in the Hawaiian Islands does not fossilize well. There is a fossil available for one genus that has an endemic species in the Hawaiian Islands (Thinophilus Wahlberg: subfamily Hydrophorinae), but the wide range in ages of the fossils (Baltic amber—Eocene/Oligocene; ca. 35–60 mya) compared with the very young ages of the islands make them unsuitable for use in this analysis. Instead, we used three biogeographic calibrations based on the island ages of Kauai, Maui and Hawaii (see Table S2A and Fig. S2A). We then evaluated the impact of the Kauai calibration by running an analysis with only the Maui and Hawaii calibrations, and also ran two alternate analyses for comparison based on evolutionary rates, described in Appendix S2.

We selected two well-supported nodes for calibration from within a lineage of the genus Eurynogaster that exhibit a clear progression from older to younger islands (Oahu to Maui to Hawaii). We also performed a maximum likelihood ancestral state reconstruction in Mesquite v.2.7.2 (Maddison & Maddison, 2009) to assign ancestral areas to all nodes in the phylogeny. We then selected a third well-supported node for calibration with a clear ancestral range reconstruction to the oldest island of Kauai. All three nodes were calibrated with island dates from Carson & Clague (1995) (Table S2A and Fig. S2A). While island calibrations have been widely used for the estimation of divergence times in Hawaiian lineages (e.g., Rubinoff & Schmitz, 2010; Lerner et al., 2011), it is plausible that divergence among populations occurred prior to island emergence and was thus unrelated, or that it occurred well after the emergence of the younger island (Heads, 2005). Standard deviations were chosen to accommodate some of this uncertainty, including a biologically relevant timeframe during which habitat was likely available on the islands, and the fact that the insects may have colonized the islands well before or after they reached their peak heights (Table S2A).

Divergence time estimation was performed on dataset B described above. The same seven gene concatenated data set (COI, COII, ND2, 12S, 16S, EF1α and CAD) was analysed in each of the analyses described here and in Appendix S2. Partitions and the best fit models of evolution for each partition were selected using BIC in PartitionFinder (Lanfear et al., 2012). Initial analyses indicated that these models overparameterized the data in that the ESS values were extremely low for some parameters, despite being run with very long chains (beast-users Google group discussions). For the final runs, all GTR models were changed to HKY (Table S1B) and ESS increased significantly while divergence times and tree topology did not change. Base frequencies were estimated from the data. The partitioning scheme in the divergence rate analyses differed only slightly from the island calibration analyses in that COI was assigned its own partition (Table S1B). Site and clock models were unlinked and all partitions were analysed using an uncorrelated lognormal relaxed clock except for the partition comprised of CAD (positions 1 & 2) and the EF1α intron, for which a strict clock could not be rejected and was thus applied. The tree-shape prior was linked across partitions and the tree-shape prior was specified as a Yule Process. The xml file was hand edited to include a starting tree, generated using maximum likelihood in RAxML 3.7.2 (Stamatakis, 2006). Two independent MCMC searches were conducted, each running for 50 million generations and sampled every 1000 generations. The number of generations was selected to generate ESS values greater than 200 for each of the parameters (Drummond et al., 2012). Convergence was assessed using Tracer v. 1.7.5 and trees were summarized to one Maximum Clade Credibility (MCC) tree using Tree Annotator v. 1.7.5 after removing a burn-in phase.

Phylogenetic relationships within the endemic Hawaiian Dolichopodidae

The family Dolichopodidae includes more than 6,800 described species (Yang et al., 2006) in 232 genera worldwide (Pape & Thompson, 2013). A total of 29 genera are found in the Hawaiian Islands. Of these, fifteen have been introduced in the past 150 years, most likely through human activity, while the remaining fourteen genera present in the archipelago are known to contain endemic Hawaiian taxa (Table 1). The relationships between Campsicnemus and the Eurynogaster complex and the colonization history of these genera have remained an open question, largely due to the difficulty of placing both in a subfamilial context. While Campsicnemus is clearly placed in the subfamily Sympycninae, the placement of the Eurynogaster complex has been more difficult to ascertain (see Appendix S3). Individual taxa have previously been described as members of the subfamilies Sympycninae, Hydrophorinae, and Thinophilinae. Hardy & Kohn (1964) considered Eurynogaster and associated genera as part of the Sympycninae (see Fig. S3A in Appendix 3). Later, Evenhuis (2005) transferred the entire Eurynogaster complex to the Hydrophorinae. If the current taxonomy placing these lineages in two separate subfamilies is correct, Campsicnemus and the Eurynogaster complex represent independent colonizations to the Hawaiian Islands.

Molecular evidence demonstrates that the endemic Hawaiian dolichopodid fauna is clearly the result of multiple colonizations to the archipelago (Fig. 2, Figs. S1A & S1B in Appendix S1). Several key nodes are well supported and allow us to infer the history of the Hawaiian Dolichopodidae. Conchopus (posterior probability (PP) = 1, bootstrap (BS) = 100: node A, Fig. 2), the Eurynogaster complex (PP = 1, BS = 100: node B, Fig. 2), and Campsicnemus (PP = 1, BS = 98: node C, Fig. 2) are each supported as monophyletic with respect to other dolichopodid genera. Another key node that is supported in both analyses (PP = 1, BS = 99: node D, Fig. 2) is the large clade that includes Campsicnemus and a number of non-Hawaiian genera in the subfamily Sympycninae (e.g., Sympycnus, Teuchophorus) and does not include the Eurynogaster complex. This demonstrates that there were at least three colonizations to Hawaii by the family Dolichopodidae, one each by the three radiations: Campsicnemus, the Eurynogaster complex, and Conchopus.

There is little support for the placement of Thinophilus, so its history of arrival to Hawaii remains enigmatic (Fig. 2, Figs. S1A & S1B in Appendix S1). This genus is known primarily from the Indo-Pacific, with one species each known from the Galapagos Islands and the Hawaiian Islands. Previously it has only been collected from rocky, wet sand on the south shores of Oahu (Carlton & Eldredge, 2009). The specimen included in this study represents the first record from Hawaii Island and suggests that directed collecting on the south shores of Maui and Kauai may turn up additional populations.

Three genera that contain endemic taxa were not included in this study because they are difficult to collect and we did not recover them in our sampling. While their placement must await future work, their omission here does not change the result that the Hawaiian Islands have been colonized multiple times. An additional issue is that support at many nodes in this phylogeny is poor, owing partially to the large divergences between the subfamilies and the incomplete taxon sampling with this enormous family. These issues are also seen in previously published phylogenetic studies of dolichopodid relationships (Lim et al., 2010; Bernasconi, Pollet & Ward, 2007). The lack of support and long branches across most of the rest of this phylogeny preclude identifying the specific sister lineages to the Hawaiian taxa (Fig. 2).

Phylogenetic relationships in the Eurynogaster complex

The Eurynogaster complex, with 66 described species (Yang et al., 2006) and about a dozen awaiting description, comprises the fourth most species-rich radiation of Hawaiian flies, after the Drosophilidae, Campsicnemus, and Lispocephala (Muscidae). Although the Eurynogaster complex is one of the largest radiations of Diptera in Hawaii, phylogenetic relationships in this group have never been studied. This collection of genera are hypothesized to have been derived from a single colonization to the Hawaiian Archipelago (Evenhuis, 2005). Little is known about the biology of these species, but collecting observations suggest that species found on the forest floor and on vegetation tend to be dull coloured, while species found in wet habitat, along seeps, streams and on wet banks tend to have shiny metallic thoraces and/or abdomens.

Molecular phylogenetic results presented here show support for Evenhuis’s (2005) hypothesis of a monophyletic complex of related genera (PP = 1, BS = 100: Fig. 2), as well as support for several of the genera within this radiation. We focused on the smaller dataset (dataset B) to address phylogenetic and biogeographic questions within the Eurynogaster genus complex. Analyses of individual genes are presented in Figs. S1E–S1L in Appendix S1, and final data partitions and evolutionary models are reported in Table S1B in Appendix S1. Tree topologies generated using ML and BI approaches of the concatenated dataset B were very similar; at well-supported nodes, they are identical (Figs. S1C & S1D).

In Fig. 1, the maximum clade credibility tree from the Bayesian analysis performed in BEAST is used to display the patterns within the Eurynogaster complex, and the following PP and BS supports are from the BI performed in MrBayes and ML analysis performed in RAxML (shown in Figs. S1C & S1D). The Eurynogaster complex is split into two clades: Clade A (Adachia + Elmoia + Sigmatineurum + Major, PP = 1, BS = 100) and Clade B (Sweziella + Arciellia + Uropachys + Eurynogaster, PP = 1, BS = 100). Current sampling indicates that the genus Adachia is monophyletic (PP = 1, BS = 100) and sister to a well-supported clade (PP = 1, BS = 100) composed of the genera Elmoia, Sigmatineurum and Major (ESM Clade). Sampling within the ESM clade is not extensive, with only a single representative each of Sigmatineurum and Major. Two representatives of the genus Elmoia were sampled and our results indicate that this genus is paraphyletic with respect to Sigmatineurum and Major. Denser sampling with the ESM clade will be necessary to resolve the placement of the Elmoia taxa.

Clade B includes the large genus Eurynogaster, along with Arciellia, Uropachys and Sweziella. Sweziella, represented by S. tergoprolixa from Maui, is the basal lineage within clade B and sister to the lineage formed by Arciellia, Uropachys and Eurynogaster (PP = 1, BS = 91: Fig. 2). Current sampling indicates that the genus Arciellia and Uropachys are each monophyletic (PP = 1, BS = 100 and PP = 1, BS = 100, respectively) and sister to one another (PP = 1, BS = 100). Eurynogaster is supported as monophyletic (PP = 1, BS = 100). This genus is confusing taxonomically and is in need of revision. There are three undescribed Eurynogaster species that were discovered as part of this work, E. n. spp. A–C. There are also a number of taxa that, while morphologically similar to named taxa, show significant sequence divergence from the described species. This sometimes corresponds to samples having been taken from different islands. For example, E. maculata from Oahu is quite different from the E. sp. nr. maculata samples collected from Maui (E. sp. nr. maculata 141) and Hawaii Island (E. sp. nr. maculata 115 and 126)–they are 3.9% and 3.6% divergent at COI, respectively. Furthermore, one exemplar of E. maculata from Maui is quite similar to E. sp. nr. maculata 141—it is identical at COI–suggesting that cryptic species may exist within the concept of what we currently recognize as E. maculata. This phenomenon is common in large evolutionary radiations in Hawaii (e.g., Bennett & O’Grady, 2011). Another species we sampled, E. cilifemorata, also seems to be a complex of species sampled from Maui and Oahu. Additional sampling within Eurynogaster, as well as thorough taxonomic revisions of the genera within this complex, will be necessary to better delineate species within this rapidly evolving clade.

Finally, four new species within the Eurynogaster complex were discovered as a result of this project, three within Eurynogaster and one within Sigmatineurum. An additional five possible new species (Adachia—1 species; Eurynogaster—4 species) were identified (labeled as “sp. nr”.) and are in the process of examination to confirm their taxonomic status.

Arrival times and biogeography

We estimate that the Eurynogaster complex arrived in the Hawaiian Archipelago 11.83 (9.08–15.04) Ma, approximately within the timeframe that the Northwest Hawaiian Islands of La Perouse, Necker, and Gardner were providing substantial high island habitat (Price & Clague, 2002). This ancient lineage arrived well before the formation of the current high islands about 5 Ma and the arrival of Campsicnemus, which is estimated to have occurred approximately 4.6 Ma (Goodman et al., 2014). Early diversification into five of the eight contemporary genera took place in the older, now eroded, northwest Hawaiian Islands, and five colonizations of these ancestral lineages into the current main (high) islands are needed to explain the contemporary patterns of diversity. All of the diversification within the crown groups has occurred within the past 5 million years (Myr), the timeframe of the current high islands. The most speciose lineage within the Eurynogaster complex, the genus Eurynogaster, began diversifying approximately 2.6 (95% HPD: 1.94–3.26) Ma, about the time Oahu and Maui Nui were forming. We estimate that the small endemic dolichopodid genus Conchopus arrived quite recently—1.77 (95% HPD: 1.09–2.6) Ma (Fig. 1).

Within the Eurynogaster complex, a number of classic biogeographic patterns are evident, some of which are significantly different from what is observed in other large radiations. First, a progression rule pattern (Hennig, 1966) is common in hotspot archipelagos where islands appear along a chronosequence. The typical progression rule pattern seen in Hawaii occurs when the most basally branching taxon is present on Kauai, the oldest island, with more recently branching taxa present on the progressively younger islands of Oahu, Molokai, Maui and Hawaii (Wagner & Funk, 1995). While the progression rule is commonly observed in both the Hawaiian Drosophila (Bonacum et al., 2005) and Campsicnemus (Goodman et al., 2014) lineages, it is less prevalent in Eurynogaster. Only a single lineage of the genus Eurynogaster shows a clear progression from Oahu to Maui to Hawaii (Fig. 1).

Another phenomenon observed in Hawaiian lineages is within-island diversification, where species break up to diversify into new populations and eventually sibling species on the same island. This has been thought to be an uncommon occurrence, in part because it is fairly uncommon across the historically best-studied group in the islands, the Hawaiian Drosophila—for whom diversification primarily occurs following inter-island dispersal. However, even within this iconic group, there are examples and it has been very well studied in the sympatric sibling pair D. silvestris and D. heteroneura (Carson, 1982; DeSalle et al., 1987; Price & Boake, 1995). Newer examples are now accumulating across taxonomic groups (e.g., Goodman, Welter & Roderick, 2012; Eldon et al., 2013; Bennett & O’Grady, 2013; Liebherr, 2015), exposing how variable a process diversification can be, and how dependent it is on the dispersal capabilities of the groups studied (Price & Wagner, 2004). The Eurynogaster complex shows at least five instances of within-island diversification. Uropachys is a genus of six species only known from Kauai. Three Uropachys species were sampled for this study and are supported as a monophyletic clade, indicating they diversified there. This pattern is also observed in Adachia, where A. hispida and A. apicenigra have both formed on Hawaii, and in several clades of the genus Eurynogaster where diversification has occurred on Oahu and Hawaii. While it is possible that additional sampling, and subsequent discovery of new species, may alter this inference, it is clear that diversification within an island is a pattern seen in many other Hawaiian groups, including the genus Campsicnemus (Goodman et al., 2014) and the well-studied Hawaiian Drosophila (O’Grady et al., 2011).

Colonization of and diversification within the Hawaiian Islands

It is clear that the endemic Dolichopodidae of Hawaii arrived to the archipelago in at least three successful colonization to radiation sequences over the last 12 Myr (Eurynogaster complex, 11.8 Ma; Campsicnemus, 4.6 Ma; Conchopus, 1.8 Ma) –demonstrating that dispersal to and establishment within this remote island group is more common than has been documented in other groups. This is fascinating because it means that three separate radiations occurred despite the excellent dispersal capabilities of these animals. In order to multiply into radiations, they must have been able to generate reproductive isolation rapidly enough to overcome gene flow from their highly vagile conspecifics. Members of this family are known to have complicated courtship behavior (Zimmer, Diestelhorst & Lunau, 2003). Though this has never been studied in the Hawaiian fauna, it may be a contributing factor to the development of reproductive isolation as has been shown with the Hawaiian Drosophilidae (Kaneshiro, 1976; Price & Boake, 1995), Laupala (Grace & Shaw, 2011) and Nesosydne (Goodman et al., 2015), and suggests fruitful research directions. There seems to be no correlation between the age of colonization and the diversity of each lineage.

MacArthur & Wilson (1967) stated that “an island is closed to a particular species when the species is excluded... by competitors already in residence...” The Hawaiian Islands were clearly not closed to dolichopodid flies that arrived after the first wave 12 Ma. This suggests that, at the arrival of each new lineage, there was still plenty of ecological opportunity available or these insects are ecologically labile and able to adapt easily when faced with niches already occupied by competitors. Both statements may be true.

Very little is known about the ecology of the Dolichopodidae in Hawaii, but they are known to be predatory from observations elsewhere in the world (Ulrich, 2005). For the Eurynogaster complex (which only occur in Hawaii), there is only a single published account in the literature that includes ecological observations (Williams, 1938). Despite the dearth of ecological data available, we have some evidence to support the idea that the dolichopodids seem to adapt easily. In our 2014 study, we used morphological colouring together with field observations to infer that the Hawaiian Campsicnemus have rapidly diversified into three ecological types: (1) brown, low vegetation and litter dwellers, (2) black water skaters and (3) yellow canopy dwellers. Interestingly, the black water skaters and yellow canopy dwellers are restricted to the Pacific. Furthermore, the yellow canopy dwellers are endemic to Hawaii (Goodman et al., 2014). The Hawaiian Conchopus may have also undergone a shift in ecological type. This lineage can be traced back to East Asia (Takagi, 1965), where they are known primarily from barnacle colonies in the marine tidal zone, living in the interstices or in nearby cracks in the rocks and feeding on tiny invertebrates (Sunose & Sato, 1994). There are no native barnacles in the Hawaiian Islands, and Conchopus there are known from puka (holes) in beach rocks deriving from volcanic flows. Once established in this habitat, they radiated into six known species.

Prevailing dogma among Hawaiian evolutionary biologists in the past 30 years has been that colonization events to the archipelago are rare and colonization within the islands follow a few well-defined patterns, such as the progression rule (Wagner & Funk, 1995). Recent molecular phylogenetic studies are beginning to overturn these overly simplified notions (Heaney, 2007; Bellemain & Ricklefs, 2008), finding that colonization and diversification are based on a combination of factors. These include characters linked to the dispersal and adaptability of the lineage in question and the ecological and environmental context of the islands when that lineage arrives. The current study highlighting the multiple colonizations that Dolicohpodidae have undergone in the past, and the specific patterns of diversification within the Eurynogaster complex, further demonstrate that there are no simple “rules” and each colonization event should be considered an independent event. The Hawaiian Dolichopodidae are an intriguing example of repeated, overlapping evolutionary radiations, ripe for field studies that can begin to untangle their propensity to speciate and ecological lability.