Intergeneric relationships
Previous partulid molecular phylogenies [35],[39],[43]-[45] have either ignored or under-sampled Western Pacific taxa, the most comprehensive representation being 4 western species of Partula[44]. We were able to incorporate 14 Western Pacific species (Table 2) in our analyses, including for the first time Samoana fragilis, the sole western member of that genus, and 2 of the 3 endemic Partula species in Palau, the westernmost archipelago of the family’s range (Figure 1). Twelve of the Western Pacific partulids clustered as expected with Central/Eastern Pacific congeners in our molecular phylogenetic trees. However, the two Palauan species, P. thetis and P. calypso, were topologically distinct from all of their congeners, forming discrete and robustly supported clades with Samoana species. For our relatively conserved marker, the large nuclear ribosomal gene fragment, the two Palauan Partula species formed a shallow polytomy with Samoana fragilis, within a polytomous Samoana clade (Figure 2). The faster-evolving mt COI marker yielded a more fully resolved inference of Palauan partulid genealogical relationships – see Figure 3 for the salient segment of this much larger phylogenetic tree (the complete topology is available in Additional file 1: Figure S1). Here, P. thetis and P. calypso were sister to all genotyped Samoana species.
The unexpected phylogenetic placement of the two Palauan Partula taxa with Samoana species raises a question concerning their generic status. Pilsbry and Cooke [42] distinguished the genera Partula and Samoana primarily on male genital characters: the latter being differentiated into distinct epiphallus and penis segments connected by an extension of the penial retractor muscle that is absent in other partulids. Kondo’s Ph.D. thesis [56] contains detailed diagrams of the reproductive tracts of the three endemic Palauan species (P. thetis, P. calypso and P. leucothoe, the latter was unavailable for this study) and all three unambiguously displayed the male genital anatomy characteristic of the genus Partula. These anatomical data, together with the robust, well-resolved mt gene tree topology (Figure 3) demonstrate that the genus Partula is paraphyletic and that the distinct genital anatomy of Samoana species is derived. In contrast to Kondo and Burch’s [48] hypothesis of intergeneric relationships - (Eua (Samoana, Partula)) - our data are consistent with (Eua (Partula1 (Partula2, Samoana))) where Partula1 represents all non-Paluan members of the genus and Partula2 represents the two genotyped Palauan species.
West–east dispersal of Samoana
The most remarkable feature of the distribution of Samoana is the presence of a single isolated species, S. fragilis, bearing the characteristic Samoana genital anatomy [56], in the Western Pacific (Figure 1). Genotyping S. fragilis museum specimens sampled in 1945 allowed us to put this geographic disjunction into a phylogenetic context for the first time. Our two molecular markers yielded very different levels of phylogenetic resolution. The more conserved nuclear large ribosomal gene fragment produced a shallow clade with a basal polytomy such that interrelationships among Western (including the two Palauan Partula species) and Eastern Samoana taxa were unresolved (Figure 2). The mt marker provided much better resolution (Figure 3) and we base our discussion of the evolutionary history of Samoana on this topology. It shows that the Samoana clade is anchored in the West by its exclusive sister relationship with Palauan Partula species and by the placement of the Marianas species S. fragilis at its base - the sole western species is not a derived founder lineage (Figure 3). This is a somewhat surprising result because it places the inferred evolutionary origins of the genus not in the Eastern Pacific, home to 21/22 Samoana species (Figure 1), but instead in the far western portion of the partulid range (Figure 3).
The mt gene tree topology (Figure 3) is consistent with a western origin of the genus Samoana followed by progressive eastward colonization of derived lineages. S. fragilis is sister to the Samoan congener S. thurstoni, although the posterior probability value for this node is low. A derived clade, that approximates a stepping-stone dispersal pattern from West to East, is evident for the easternmost archipelagoes with Samoa as the regional source. Three closely-related Samoan species – S. conica, S. stevensonia and S. abbreviata – are robustly sister to derived congers in the Society, Austral and Marquesan archipelagoes, the latter two forming monophyletic clades. Reconstructing the multi-archipelagic dispersal pathways of these Eastern lineages is complicated by a basal polytomy involving Austral taxa and two Society Island lineages (Figure 3). However, the sister status of the Marquesan taxa with two Society Island congeners, S. diaphana and S. burchi, is consistent with an earlier inference [44] of dispersal and colonization among these two archipelagoes. Our results corroborate a previous allozyme study [57] regarding Marquesan monophyly and adaptive radiation – Marquesan taxa exhibit a diversity of distinct shell phenotypes. Interestingly, so do the two Society island montane forest sister taxa: the thin-shelled S. diaphana (Moorea and Thaiti) and the thick-shelled Tahitian endemic S. burchi[54].
Although our data provide new insights into the evolution and biogeography of the genus Samoana, they raise obvious follow-on questions: if the genus originated in the West, why does almost all the extant diversity occur in Eastern archipelagoes and why are Samoana species missing from a huge swathe of the family’s range from the Marianas to Fiji (Figure 1)? We consider it unlikely that the genus extended its range across this ~5000 km gap in a single dispersal event and hypothesize that at least some intermediate archipelagoes such as the Caroline Islands, Bismarck Archipelago, Solomon Islands, Santa Cruz Islands and Vanuatu may have supported populations at some time in the past. Present day absence of the genus Samoana from at least some of these archipelagoes is probably not due to a lack of dispersal ability. Our data indicate that a Samoan lineage was capable of stepping-stone colonization of Eastern archipelagoes up to 3000 km distant, a range that ostensibly puts the Western archipelagoes of Vanuatu, Santa Cruz and the Solomon Islands within equivalent reach. The absence of the genus Samoana from much of the Western range of the family may stem from (as yet unidentified) ecological factors that have driven original populations to extinction and prevented rare long-distant migrants from reestablishing new populations.
West–east Partula disjunction – product of a single dispersal event
The ~2000 km geographic disjunction – from the Lau Islands of Fiji to Rarotonga in the Cook Islands – separating Western and Eastern members of the genus Partula (Figure 1) has yet to be put into a comprehensive phylogenetic framework. Although our sampling of the genus Partula is heavily weighted toward Eastern populations, we have genotyped 11 species from across the Western range (Marianas, Palau, Carolines, Bismarck Archipelago and the Massim Region of Papua New Guinea, Vanuatu, Wallis and Futuna, Fiji). This broad sampling allows some confidence that we have captured much of the regional phylogenetic framework, if not all of its detail. Similarly, for the Eastern Partula species, our taxon sampling is geographically comprehensive: the 26 species genotyped include the sole non-anthropogenic Cook Island species, P. assimilis, as well as representatives from the all of the salient Society Islands except for Tahaa, which shares the same lagoon as Raiatea.
Our phylogenetic analyses of the large nuclear gene fragment consistently produced a very poorly resolved topology for the non-Palauan Partula species, as previously reported [44], e.g., Figure 2 contains two polytomies, each with a mix of Western and Eastern species. We are therefore basing our inferences of West–east Partula relationships on the mt topology. As shown in Figure 4, the Western taxa place at the base of the non-Palauan Partula clade separated by a robust (PPS = 100) node from all Eastern congeners. This is consistent with a single dispersal event from West to East (Figure 1).
Phylogenetic relationships among the non-Palauan Western Partula taxa were obscured by a basal polytomy (Figure 4). Within-archipelago monophyly was evident only for Marianas taxa and these were weakly sister to the Micronesian P. emersoni. The phylogenetic affinities of Papua New Guinean populations of P. carteriensis and P. similaris, flagged as likely anthropogenic introductions [28], remained uncertain. The two easternmost taxa had distinct topological placements: P. subgonochila (Wallis & Futuna) was a member of the basal polytomy whereas the Fijian P. lirata was sister to one of the Vanuatu species, P. auraniana (Torres Islands). These disparate results imply that a more comprehensive sampling of Western taxa is required to capture a high-resolution understanding of regional evolutionary history. Unfortunately, this will have to be accomplished quickly given the ongoing decline and extirpation of many Western Pacific partulid populations and species [24],[58]-[61].
Eastern Partula diversification
Figure 5 shows, in outline form, the mt tree topology of the Eastern Partula clade labeled by source island – a full-sized topology containing details of individual genotype identification is available in Additional file 1: Figure S1. Apart from major gaps for the islands Tahaa (0/6 species) and Raiatea (7/34 species), sampling of the Society Islands endemic Partula radiation was almost complete: Bora Bora (1/1), Huahine (3/3), Moorea (7/7) and Tahiti (7/8 – the missing species, P. cythera, has not been seen since its discovery on a remote interior mountain slope in the 1920’s [62]).
The only Eastern Partula species not endemic to the Society Islands – P. assimilis of Rarotonga (Cook Islands) – is positioned within the Society Island clade (Figure 5), weakly sister to a minor, divergent Moorean P. suturalis clade and to the Bora Bora endemic P. lutea, with which it shares a large nuclear ribosomal genotype (Figure 2) as well as allozymic affinities [43]. We agree with Johnson et al. [43] that P. assimilis probably represents a founder lineage from a Society Islands source; the only know Partula transplant from that hotspot archipelago apart from anthropogenic populations of the Tahitian endemic P. hyalina[34].
Concerning the Society Islands topology, a number of important stem nodes within the clade were poorly supported and, consistent with parallel studies on this archipelago’s biota [53], there was little agreement with progression rule [4] expectations for the archipelago, i.e., topological congruence with the chronological sequence of island formation. Only one (Huahine) of three older Society Islands represented in the topology (Bora Bora, Raiatea, Huahine; all members of the Leeward Island subgroup) positioned basally. The oldest Society Islands with partulids, Bora Bora and Raiatea, placed in derived positions, weakly sister to a heterogeneous assemblage that included the Rarotonga species, two Moorean clades and one Tahitian clade (Figure 5).
One consistent topological difference between the Leeward and Windward Island lineages concerned their degree of within-island monophyly. Genotyped snails from each Leeward Island formed well-supported (PPS ≥97) single island-specific clades. Snails from each Windward Island formed four discrete lineages, three of which had exclusive phylogenetic relationships (PPS ≥96) with snails in the other Windward island, as described in detail in an earlier study [35]. However, the phylogenetic affinities of the largest and most deeply branched of the four Tahitian mt clades remained enigmatic. This robust (PPS = 99) Tahitian clade positioned on a basal polytomy that encompassed both Leeward and Windward Island lineages and it lacked an identifiable sister lineage, either on Moorea, or on any of the other Society Islands (Figure 5).
The Figure 5 topology is incompatible with the prevailing speciation model for Moorean and Tahitian Partula that views all congeners on each Windward Island as the product of a single colonization event: a Leeward Island source for Moorea and a single Moorean source for the subsequent colonization of Tahiti [49]. The model has a narrow base of empirical support: allozyme phylogenies from 24 species of Partula that have been weighted on the basis of allele frequency [43],[49]. These data overturned previous inferences – based on reproductive relationships and morphological similarities – that Tahiti was colonized by multiple Moorean lineages [63]. However, the allozyme data were less than robust: an unweighted (UPGMA) analysis of the same allozyme dataset yielded a substantially distinct topology in which the Tahitian species were no longer monophyletic [43] and a later reanalysis yielded another topology consistent with back-migration from Tahiti to the southern part of Moorea [64]. Salient DNA phylogenies, whether using nuclear ribosomal [44, this study] or mt [35, 50–52, this study] markers have failed to recover the model’s predicted topology. The mt studies, in particular, have consistently revealed the presence of multiple exclusive sister relationships among subsections of Moorean and Tahitian Partula mt treespace [35, 50–52, this study].
Despite these incongruences, the Windward Island Partula speciation model [49] has remained the consensus conceptual framework for speciation studies of Moorean and Tahitian Partula species: speciation is viewed as having occurred in situ within each island [43],[46],[49]-[53],[65]. Consistent with this view, examples of Moorean/Tahitian mt polyphyly have been interpreted either as being due to convergent molecular evolution (for restriction fragment length polymorphism data [23],[50], or to retained ancestral mt polymorphisms [51]-[53], rather than being products of inter-island gene flow involving multiple discrete lineages. However, both of these non-gene flow interpretations are problematic. The large majority of mt COI nucleotide substitutions observed among Moorean and Tahitian taxa are synonymous (data not shown) and are therefore unlikely to stem from selection-driven convergent evolution. The ancestral polymorphism interpretation requires a vicariant model of inter-island genetic differentiation that is inapplicable to Moorea and Tahiti, two spatially discrete islands that have never been joined [66],[67].
We propose a new speciation model for Tahitian Partula taxa that involves four discrete founding lineages. Three of the four have explicit phylogenetic ties to Moorean congeners, as detailed earlier [35], but the largest, comprising half of Tahitian mt treespace, is of undetermined provenance (Figure 5). This new model is consistent with parallel Society Island studies that have uncovered a role for multiple independent colonizations in the evolution of Tahiti’s endemic biota [53].
Age of the Society Islands Partula radiation
A standout feature of the biogeography of Partulidae is the markedly asymmetric distribution of alpha diversity across the family’s 10,000 km range (Figure 1). Half of the nominal species diversity is endemic to six Society Islands spanning a mere 320 km of Oceania with a quarter being endemic to a single island: the 167 km2 Raiatea [23]. The extraordinary concentration of 59 species of Partula within this single hotspot archipelago raises questions of taxonomic equivalency (species descriptions are typically based on shell phenotype distinctions) and lineage persistence across the range. While we cannot engage meaningfully here with the fraught topic of Partula species designations [52],[64],[68], we can address the issue of lineage persistence by constructing a time-calibrated phylogeny for our molecular dataset.
Figure 6 shows, in outline form, the time-calibrated Partulidae mt tree obtained with BEAST (see details of the topology in Additional file 2: Figure S2). Three calibration points based on the geological ages of Bora Bora, Raiatea and Tahiti, were used to date nodes supporting the entire Society Island clade, the endemic Raiatean clade and the largest endemic Tahitian clade, respectively. The BEAST topology differed little from that of the Bayesian analyses (Figures 3,4 and 5) except that the non-Palauan Western Partula species now formed a discrete clade, sister to their Eastern congeners (Figure 6). It yielded a greater inferred age for the Society Island Partula radiation, both for the entire archipelago (3.27 Mya), and for the single island of Raiatea (2.71 Mya), than for the much more widespread sister Western Partula clade (2.41 Mya), the genus Eua (2.31 Mya) and the family range-spanning genus Samoana (2.12 Mya), although note that the 95% Highest Posterior Density age intervals for many of these nodes overlap (Figure 6).
These time-calibrated data imply that the heightened alpha diversity characteristic of the Society Island Partula radiation, and of the island of Raiatea, resulted at least in part from a longer uninterrupted diversification timeframe than was available elsewhere in the family’s range. For this within-archipelago radiation, cladogenesis has apparently unfolded in a qualitatively different manner to that experienced by partulids elsewhere, being characterized (until very recently) by greater lineage longevity coupled with much less lineage dispersal. Within the archipelago, the hyper-speciose Raiatean radiation is correlated with that island’s age, area and relief. Despite its relatively advanced age (Figure 5), Raiatea is second only to Tahiti in area and has retained the high island profile (1017 m maximum elevation) required for autonomous rainfall generation and rain forest formation [69],[70]. Outside of the Society Island Partula radiation, the small Palauan Partula and Eua clades also show evidence for long-term persistence within modest ranges, but lineage diversification in the Western Partula and Samoana crown clades occurred across much more extensive ranges and necessarily involved multiple episodes of among-archipelago dispersal (Figures 3 and 6).
Divergent histories of Eastern Partula and Samoana radiations
Although the genera Partula and Samoana overlap in the eastern edge of the family’s range, co-occurring in the Society Islands, they appear to have experienced distinct patterns of regional cladogenesis [35]. Partula has many more regional species (59 vs. 15 respectively), but Samoana has a greater eastern range, encompassing the Marquesas and Austral Archipelagoes as well as the Society Islands (Figure 1). They also differ in their regional phylogenetic/population genetic profiles: eastern Samoana species have much lower collective genetic diversity levels and more pronounced phylogenetic cohesiveness [35],[55],[57]. The BEAST topology (Figure 6) yielded new insights into the evolutionary origins of these regional distinctions, showing that the genus Partula established eastern populations much earlier (3.27 Mya) than did Samoana (1.12 Mya). This divergent chronology probably contributed to the regional generic species richness disparity, but not to their distributional disparity. Despite its much longer regional timeframe, the genus Partula has effectively been marooned on the Society Islands, establishing just one inter-archipelago founder species in Rarotonga (Cook Islands), excepting anthropogenic introductions of P. hyalina[34]. In contrast, the later arriving Eastern Samoana lineage has proven to be a much more effective regional disperser, on both intra-archipelago (many species have multi-island distributions: 33, 35, 54, 55, 57] and inter-archipelago scales. Why this should be is unclear, although it may be relevant that many eastern Samoana species produce exceptionally sticky mucus [54],[57] and this trait could plausibly increase the likelihood of rare inter-island/archipelago phoretic dispersal events, e.g., on avian vectors.