Mascarene stick insect phylogeny

Recent studies have revealed that convergently evolved features have often been the basis of classification [24, 34]. Molecular studies provide a breakthrough in the investigation of these radiations since they avoid the use of circularly interpreted characters. Our molecular phylogeny did not support the traditional classification of Mascarene stick insects, which predicted that the respective fauna consists of five unrelated lineages, and, consequently, implied the occurrence of multiple colonisation events leading to the present-day phasmatodean community in the archipelago [22]. Instead we recovered a single origin of all Mascarene stick insects, forming a strongly supported lineage deeply nested within the highly diverse Australasian Lanceocercata (Fig. 2). All Mascarene genera we sampled here (6) and the majority of species (7) are found on Mauritius, which is generally considered to be the oldest island in the archipelago, its volcanic emergence dated 7.8 mya [5]. Three genera, Apterograeffea, Monandroptera, and Rhaphiderus, also have species on Réunion, which is a much younger island dated at 2.1 mya [5]. The age of Rodrigues has been debated with estimates ranging between 11 mya for the old volcanic ridge and 1.5 mya for the recent emergent part [5], some sources give 15 mya as the oldest date of its emergence [7, 8, 14]. The only reported, endemic stick insect from Rodrigues, Xenomaches incommodus (Butler, 1876), which is an extinct member of the Platycraninae [35], was not available for our analysis.

The Lanceocercata clade has undergone an impressive evolutionary diversification during the past ~40 million years that led to several adaptive radiations in geographical isolation, e.g., New Caledonia and New Zealand, resulting in several morphologically and behaviourally convergent ecotypes [20, 24, 33]. Therefore, it did not come as a surprise that the discovery of Lanceocercata rendered numerous traditional systematic entities artificial. Lanceocercata comprises species from six subfamilies, Pachymorphinae, Tropidoderinae, Xeroderinae, Phasmatinae, Eurycanthinae, and Platycraninae, none of which is monophyletic. This knowledge is significantly enhanced by our finding of another unexpected adaptive radiation in the Mascarene archipelago, which involves further strikingly convergent ecotypes. For instance, the Mascarene Apterograeffea (Fig. 1a), traditionally viewed as a member of the Platycraninae (coconut stick insects and allies), exhibits the same enlarged genae (‘cheeks’) necessary to accommodate the massive mandibular muscles necessary for feeding on hardy palm or screw pine (Pandanus) leaves as do the unrelated Australian and Pacific ecotypes, e.g., Graeffea and Megacrania [22]. On Mauritius, Epicharmus (Fig. 1b), traditionally assigned to Xeroderinae, is a stout winged stick insect with lobes on legs and abdomen, which enhance crypsis of this bark-dwelling species, is most reminiscent to the unrelated Australian Xeroderus and New Caledonian Leosthenes. The female of Rhaphiderus (Fig. 1c), hitherto assigned to Tropidoderinae, is a wingless leaf-imitating form very similar to the Australian Malandania and Tropidoderus. This remarkable diversity in appearance and ecology prompted taxonomists to assume the Mascarene species to be unrelated to each other, especially in regard of the young age of Mauritius, and to incorrectly place them in different subfamilies.

Origins of the Mascarene fauna

Usually, islands are colonised from the nearest land source, and not surprisingly the Mascarene biota shows close affinities to Madagascar, which is approximately 700 km away. Terrestrial vertebrates like giant tortoises [36], Phelsuma day geckos [10] and slit-eared skinks [11] and also recently studied invertebrates such as carnivorous land-snails [12] and the golden orb spider Nephila [13] are few examples among many others [7, 8, 14]. We, in contrast, found no close relationships between any Malagasy lineage of stick insects and those of the Mascarene archipelago and, based on a more restricted sampling, already demonstrated the Mascarene Tropidoderinae, Monandroptera and Rhaphiderus, are deeply nested within the Australasian Lanceocercata [20, 24]. Indo-Pacific ancestors of Mascarene lineages have been revealed before. The most famous example might be the extinct dodo (Raphus) on Mauritius and the solitaire (Pezophaps) from Rodrigues, which form sister taxa with close relatives in Southeast Asia [15, 16]. Nactus geckos and skinks of the genus Leiolopisma are further examples showing Australasian origins [19]. The arrival of Leilopisma skinks in the Mascarenes requires a spectacular transmarine dispersal between 5600 km from West Australia or more than 7000 km from a more tropical region such as New Guinea. A similar route must have been taken by the ancestral Mascarene stick insect. An east–west directed ocean stream towards Madagascar, the South Equatorial current, which was already existent during the Oligocene and Miocene, is most likely responsible for these kinds of long-distance transoceanic passages [37, 38]. It has recently been demonstrated that the hard-shelled, seed-like eggs of certain phasmatodeans can survive a significant amount of time (several months) under marine conditions, e.g., floating on seawater [39]. Eggs of Mascarene stick insects are often glued to twigs and other plant substrate [23], which would further facilitate above-water dispersal. It is noteworthy that the remarkable transoceanic journey via east–west dispersal across the Indian Ocean occurred twice within Lanceocercata and is not restricted to the Mascarene clade alone. Graeffea seychellensis, a member of the true coconut stick insects (Platycraninae) (cf. Fig. 2) that are distributed in Australia and throughout the Southwest Pacific, reached and colonised the Granitic Seychelles. Ocean currents are also likely responsible for repeated dispersals from Mauritius towards Réunion. Both islands have never been connected by land and are separated by 164 km from each other [8], nevertheless members of three Mascarene genera, Apterograeffea, Monandroptera, and Rhaphiderus, colonised Réunion since its emergence. Island colonisations are contingent events. Whereas colonisations of New Caledonia and New Zealand led to flourishing stick insect radiations [20, 24], the Pacific Galápagos Islands and the Hawaiian archipelago are devoid of any phasmatodeans. Peculiarly, Madagascar was never successfully colonised by Lanceocercata [20, 40] either from Australia or the Mascarenes, despite being separated from the latter, viz. Réunion, by only 665 km of ocean [8], far less than the 5600–7000 km that separate Australia from the Mascarenes, and it forms a significantly larger target area than the Mascarene Islands or Seychelles. However, the size and structure of the ancient island community of the Mascarene plateau, which might had much larger landmasses than today, especially during sea-level lowstands [8, 17], remains speculative. We cannot preclude that Lanceocercata might have actually arrived on Madagascar, but was not able to persist and diversify because the Malagasy lineage of stick insects, which colonised the island earlier (between 38 and 51 mya), already underwent a flourishing radiation and largely reduced ecological opportunities for any newcomer.

Timing the Mascarene colonisation

Most surprisingly, we obtained a time estimate for the origin of the Mascarene clade that significantly exceeds the geological ages of the extant emerged islands. At 22.03 mya (15.99–29.39 mya) the Mascarene phasmid lineage began to radiate long before Mauritius is supposed to have emerged (8–10 mya) [5, 8]. This result is largely contrasted by the recent accumulation of molecular phylogenetic studies favouring recent arrival and diversification of terrestrial colonisers, e.g., 3.5–5.1 mya for Phelsuma day geckos [10], ~3 mya for slit-eared skinks [11], 1.9–7.4 mya for hermit spiders and 0.8 mya for Nephila spiders [13]. Austin and Arnold [19] recovered a divergence time of 17 mya for Mascarene Leiolopisma skinks, however this refers to the stem divergence time separating the Mascarene skinks from their most closely related Australian relatives. This might represent an over-estimate, since the age of the branching between the island lineage and the outside relative (the continental sister group) marks the upper time limit of colonisation, because the closest outside relative might be unsampled or extinct [3]. Consequently, the colonisation event could have occurred significantly later. On the other hand, the age of the most recent common ancestor of a multi-species lineage of island organisms marks the lower limit, because it is possible that the ancestral species colonised the island long before the historically known taxa began to diversify (= crown divergence time) [3]. The crown divergence time of Leiolopisma skinks, when speciation on Mauritius began, is dated to 3.4 mya [19], thus being concordant with the island’s geological age. A similar pattern is observed in certain flowering plants (Laurales: Monimiaceae) with a stem divergence time of 31.55 mya between Monimia (Réunion) and its sister Palmeria (Australia) and a crown divergence time of Monimia of 1.34 mya [18]. In contrast, the crown divergence time of the Mascarene palm genus Hyophorbe is observed to lie between 12 and 14 mya, albeit with a lower 95 % confidence interval boundary falling within the assumed age of the Mascarene Islands [41], unlike our results.

We used a twofold conservative approach to date the Mascarene phasmid lineage, since our analysis is based on fossil calibrations, and any given fossil indicates the minimum time of the taxon in question, and we refer to the crown divergence time, a potential under-estimate. Taking this into consideration, it is possible that the Mascarene clade (and the whole phasmatodean tree) is even older than suggested by our analysis. We calibrated the split between Phasmatodea and Embioptera at ~122 mya, whereas recent studies based on transcriptomic data inferred much older estimates of ~164 mya [29] and even at ~250 mya [42] and (based on a combination of nuclear and mitochondrial markers) at ~260 mya [43] for the same node. In a previous less conservative approach based on published mitochondrial DNA (mtDNA) rates only, we estimated the stem divergence time of 39.5 mya for the Mascarene members of Tropidoderinae alone [20]. However, molecular clock analyses based on mitochondrial genes can overestimate divergence times [44]. For instance the proposed ancient split (~10 mya) between the two Galápagos iguanas based on mtDNA suggested speciation on now-sunken islands of the archipelago [2], which was recently challenged by a study using protein-coding nuclear genes and arriving at a much younger divergence (~4.5 mya) on currently subaerial islands [45].

Confidence in our time estimates as proposed here gains further support by the dates we obtained for the inter-island splits of the three phasmid genera that have members on both Mauritius and Réunion. The split between the two species of Apterograeffea is 3.45 mya (1.41–6.02 mya), and 0.05 mya (0.01–0.11 mya) lie between Rhaphiderus scabrosus and R. spiniger. The split between the two Monandroptera acanthomera specimens, which are considered to represent one single species despite their greater genetic distance, is dated 1.4 mya (0.5–2.51 mya). These dates are consistent with dispersal from Mauritius to Réunion following formation of the latter at least 2.1 mya.

A potential explanation for the discrepancy between the young age of Mauritius and the greater age of the respective lineage of stick insects is provided by historical tectonic and volcanic events in the western Indian Ocean. The Mascarene archipelago is considered to be the product of an age-progressive trend of north-to-south volcanic activity, the Réunion hot spot chain, with a northward increase in age of volcanic activity [5]. It is assumed that many parts of the Réunion hot spot chain were above sea level over the last 65 million years at the time of their formation (Fig. 3) and that these islands would have been temporarily available as stepping-stones for lineages dispersing between India and Madagascar [12, 17]. The next major islands that emerged prior to the rise of Mauritius were Saint Brandon (Cargados-Carajos), 385 km north-east of Mauritius, dated 31 mya, and Nazareth Bank, which is nearly contiguous with St. Brandon bank at its northern margin, dated at 35 mya [8]. Based on our conservative divergence time estimates we reject the possibility of an ancestral Mascarene phasmid arriving in Mauritius and instead propose the scenario of a colonisation of either Nazareth Bank or St. Brandon with subsequent radiation and dispersal towards the much younger Mauritius. Smaller uplifts between St. Brandon and Mauritius, e.g., Baissac Bank and Soudan Bank (Fig. 3), might have further facilitated spread of terrestrial faunal elements. This stepping-stone route was already assumed for the extinct dodo on Mauritius and its sister, the solitaire, on Rodrigues, which exhibit an equally deep divergence time between 23–25 mya [15, 16], and also for two sister taxa of land-snails [12]. With only two taxa involved, the alternative, albeit less parsimonious hypothesis that they diverged on a distant landmass and colonised the islands independently, has to be considered [2]. However, the more diverse radiation of stick insects with a distant Australasian origin makes this scenario very unlikely. It is likewise implausible to assume multiple colonisations from still existent landmasses in the Indian Ocean, such as Madagascar, that requires the assumption that each coloniser’s sister taxon went extinct on this landmass after the colonisation event. Another potential source of colonisation could be Rodrigues whose upper age estimate (15 mya) nearly overlaps with our lower estimate of the Mascarene phasmid clade (16 mya). Given that our estimate, if anything, is an under-estimate, the time-span difference between the origin of the Mascarene clade and the emergence of Rodrigues is probably much larger. Furthermore, past studies have shown that Rodrigues is rather a sink than a source for colonisers with not a single organismic lineage of Rodrigues having colonised any other island in the region [14]. Thus, the possibility of Rodrigues as a source for the Mascarene stick insects appears highly unlikely. A further alternative hypothesis in accordance with our data would be that Mauritius is much older than previously thought. A recent study suggests the presence of a Precambrian microcontinent in the Indian Ocean, now covered by younger magmatic deposits that separated from Madagascar 61–83.5 mya and fragmented into a ribbon-like configuration that also comprises Mauritius [6]. However, this conclusion does not necessarily mean that all of these continental fragments, including Mauritius, were above sea level before the more recent volcanic uplift. Nevertheless, exposed lavas in the Mascarenes are often the result of recent volcanic reactivation and might impede the detection of the true age of the underlying geology [7].

Consequently, we favour the assumption that submerged islands such as St. Brandon have existed continuously above sea level since its formation until the emergence of Mauritius, viz. over a period of more than 20 million years.

Speciation in the Mascarene archipelago

We found limited evidence for speciation of stick insects on Mauritius. The dates for the inter-generic splits of Mascarene phasmatodeans predate the island’s age again. It is generally assumed that a minimum island size exists below which speciation is rather unlikely due to the lack of opportunities for geographical isolation and niche partitioning [46]. Small islands, like those forming the current Mascarene archipelago, lack speciation in some organismic lineages, as recently observed for Tetragnatha spiders [47], because these islands do not allow for sufficient geographical isolation (allopatric speciation) and provide fewer ecological opportunities, e.g., less niche space than large islands, that would more readily allow diversification due to ecological specialisation (sympatric speciation). Yet, the weevil genus Cratopus underwent a quite impressive radiation in the archipelago leading to more than 80 species [7]. Before severe habitat destruction since the 17^th century, Mauritius was homogenously covered by semi-dry palm forest and moist tropical rain forest [8], but the much younger Mauritius might have been ecologically more diverse. The current landscape of distantly spaced small islands in the Mascarene archipelago was probably preceded by a period of larger islands that more readily allowed for speciation. Therefore, it appears plausible that at least some speciation took place on (intra-island) or between (inter-island) these former, now submerged northern landmasses, which apparently were much larger than Mauritius, e.g., St. Brandon, Nazareth Bank (Fig. 3), followed by repeated colonisations, 3–6 times, of Mauritius. In fact, multiple dispersals towards Mauritius are not unlikely since stick insects already colonised Réunion at least three times during the past 2 million years alone. Allopatry might have played an important role in regard of the deeper nodes of the Mascarene stick insects and was most likely crucial for more recent speciation events. With the exception of Monoiognosis, which has only very recently diverged into two species on Mauritius, further intra-generic splits occurred only in geographical isolation after colonisation of Réunion. Similar patterns are observed in the Mascarene radiations of giant tortoises [36] and geckos [10], in which allopatric inter-island speciation as well as sympatric intra-island speciation events are reported.