The Malacca Strait separates distinct faunas of poorly-flying Cautires net-winged beetles

Sampling and sequencing

Cautires net-winged beetles from the Malay Peninsula and Sumatra were included in the dataset (Table S1). The available material contained 140 samples from Sumatra and 248 samples from the Malay Peninsula. The samples were collected in 18 localities from lowlands to 2,400 m above sea level (Fig. 1; Table S1). The collecting was approved by the permit No TS/PTD/5/4Jld48(41); some material was collected on public land, outside protected areas, no protected species were collected.

The total DNA was extracted from metathoracic muscles using the DNeasy tissue kit (Qiagen N.V., Venlo, Netherlands). Due to financial constraints that limit the genetic sequencing of hundreds of samples and the problem of identifying suitable genomic markers for the species-level phylogeny, only mitochondrial fragments were amplified: rrnL–tRNA-Leu–nad1 (∼810 bp), the 3′end of cox1–tRNA-Leu–cox2 (∼1,100 bp), and nad5–tRNA-Phe–tRNA-Glu–tRNA-Ser (∼1,310 bp). The primers and PCR conditions followed Sklenarova, Chesters & Bocak (2013). The PCR products were purified using PCRμ96 Plates (EMD Millipore Co., Burlington, MA, USA) and sequenced by an ABI 3130 automated sequencer using the Big Dye Sequencing Kit 1.1 (Thermo Fisher Scientific Inc., Foster City, CA, USA). The chromatograms produced by Sanger sequencing were edited using Sequencher 4.8 (Gene Codes Inc., Ann Arbor, MI, USA) and the new data (GenBank accession codes AB123456–AB123456, Table S1) were aligned with the previously published sequences representing several Metriorrhynchini genera as an outgroup (Table S2; Sklenarova, Chesters & Bocak, 2013; Sklenarova, Kubecek & Bocak, 2014).

Species delimitation, phylogenetic analyses, reconstruction of ancestral areas and dating

The taxonomy of South East Asian Cautires has not been revised and original descriptions are uninformative. Only the Malay fauna was recently studied (Jiruskova & Bocak, 2015; Jiruskova, Motyka & Bocak, 2016). Therefore, we could not formally identify many Sumatran species and we had to delimit them here. We did not have a chance to test intrinsic reproductive isolation. Therefore, using the biological species concept (sensu Mayr 1942, see De Queiroz, 2005 for further discussion), we hypothesized that the sets of individuals which differ morphologically from other individuals are intrinsically reproductively isolated and represent separate biological species. We used external morphology, that is, the relative size of eyes in males, coloration, the shape of the pronotum, elytral cells and elytral costae. The characters were studied using a binocular microscope Olympus SXZ-16 under magnification 6–100×. Further, the genitalia of all species were dissected and cleaned from muscles and fat bodies to observe detailed structures. Genitalia often serve as a reproductive isolating mechanism and closely related net-winged beetle species regularly differ in their morphology (Malohlava & Bocak, 2010; Bocak & Yagi, 2010; Masek et al., 2015; Fig. S1). Further, we used mitochondrial rrnL, cox1 and nad5 markers to identify genetic differentiation between morphology-based species (Table S3). Although maternally inherited, these markers are commonly used to identify species limits (Ahrens et al., 2016, but Baselga et al., 2013). The comparison of color patterns, morphology and molecular differentiation can identify whether some species are color polymorphic. If we identified a set of individuals with highly similar morphology and mtDNA sequence, the difference in coloration was not considered as a proof of intrinsic reproductive isolation and we designated the individuals belonging to various color pattern subsets as a single putative biological species.

All individuals were dry-mounted and the color pattern of each individual was described and each individual was assigned to one of the following groups: (1) the pronotum and elytra completely black; (2) the pronotum and the humeral part of elytra light brown; (3) the pronotum red, humeri or at least the humeral part of elytral costae red; (4) the pronotum black, the basal part of elytra brightly colored, that is, red or brown; (5) the pronotum brightly orange, at least the humeral part of elytral costae brightly orange or whole elytra orange; (6) the pronotum black, the humeral part of elytra black, their apical part red; (7) the whole upper side of the body yellow; (8) the pronotum completely red or with a black patch in the middle, elytra black. The geographic distribution of color patterns was mapped.

Mitochondrial DNA fragments were separately aligned with MAFFT 7.017 plug-in (Katoh & Standley, 2013) in Geneious R7.1.9 (Biomatters Inc., Newark, NJ, USA) and G-Ins-i algorithm. The alignments of the protein-coding genes cox1, cox2, nad1 and nad5 were checked by amino acid reading frames and manually corrected where necessary. The rrnL fragment has a complex loop structure and its alignment may be more complicated if loops are extensive (Zhong & Zhang, 2012). The longest identified indel contained four positions, and therefore, we did not use a structural alignment approach to assess the homology of individual nucleotides and gaps. The concatenated supermatrix was analyzed under the maximum likelihood (ML) criterion using IQ-TREE 1.6.0 (Nguyen et al., 2015) with 5,000 UFboot iterations and partitioned by genes. Optimal models of evolution were identified by ModelFinder (Kalyaanamoorthy et al., 2017) implemented in IQ-TREE (Table S2).

The dataset for subsequent analyses was pruned to a single representative of each putative species (Figs. 2 and 3). The reduced dataset contained 76 terminals and Xylobanus sp. as a single outgroup and was used for both dating and area reconstruction analyses. The splits were dated in Beast 1.8.1 using the fixed topology inferred from the analysis of the pruned dataset (Drummond et al., 2006; Suchard & Rambaut, 2009). The HKY model, Yule Process and Lognormal Uncorrelated Relaxed Clock, as proposed in the Beast 1.8.1 manual, were set in the Beast analysis after the application of the GTR+I+G model did not reach convergence (Drummond et al., 2012; Drummond & Bouckaert, 2015). As no fossils of Metriorrhynchini are available, we used rates of molecular evolution proposed by Papadopoulou, Anastasiou & Vogler (2010): the 0.0168 substitutions/site/my/lineage for cox1 fragment, 0.0054 subs/s/my/l for rrnL fragment and 0.012 subs/s/my/l for nad5 fragment. The Markov chain Monte Carlo (MCMC) parameters were set to 5 × 10⁷ million generations with sampling every 5,000 generations and the effective sample size values. The pre-stationary phase was identified in Tracer 1.6 (Rambaut et al., 2014) and the initial 1.25 × 10⁷ generations were discarded as burn-in.

Additionally, the ancestral areas were inferred using the ML framework in BioGeoBEARS (Matzke, 2013) implemented in RASP 4.0 (Yu et al., 2015), using the dataset containing 76 terminals (a single terminal per species). We compared all alternative models of colonization, all also with +J (Matzke, 2014) which tests founder-event speciation (Table S4). The localities were assigned to respective taxa and coded for two analyses. (i) general geographic origin (A) Sumatra and (B) Malaya; (ii) specific geographic origin: (A) Malay Highlands, (B) Malay lowlands and lower elevation forest <1,000 a. s. l., (C) Sumatra Jambi (Tujuh and Kerinci) (D) Sumatra Barat (Merapi, Maninjau, Talamau) and (E) Sumatra Utara (Sibayak and Sinabung) and (F) Java. The phylogenetic distribution of aposematic patterns was noted for each terminal and their geographic distribution was summarized on the map of the western part of the Sunda Shelf. We do not present the formal reconstruction of the evolution of color patterns because their distribution mostly agrees with specific geographic distribution (see the analysis above).

Sanger sequencing, alignment, phylogenetic analyses

Three mtDNA fragments, rrnL, cox1 and nad5 were assembled in the dataset of 388 ingroup and 18 outgroup specimens. The ingroup was represented by 369 cox1 fragments (95% completeness, 1,104 homologous positions in the MAFFT alignment), 178 rrnL (45%, 817 positions) and 368 nad5 (95%, 1,322 positions) (Table S1). Cautires was retrieved as a monophyletic group albeit with moderate support BS 90%, similar to the relationships among the deepest clades (Figs. 2 and 3; Fig. S1). The shallower splits had high bootstrap (mostly BS ≥ 99%). A well-supported split separates the Cautires pauper species-group and the Cautires s. str. clade (i.e., all Cautires including the C. obsoletus species-group as defined by Dudkova & Bocak (2010), Jiruskova & Bocak (2015)).

Species identification and distribution

Using morphology, we identified 76 species, 28 of them formally named, and the results of morphology-based species limits were compared with their genetic divergence (Fig. S1; Table S3). The delimitation widely agrees and in the cases of ambiguous support, the species are morphology-based using genitalia (e.g., C. rianganus and C. tapahensis) or the shape of the pronotum (e.g., C. corporaali and Cautires sp. AW; see illustrations in Fig. S1). Altogether 39 species were recorded on the Malay Peninsula and 39 species in Sumatra. Only two species (Cautires sp. G and C. rianganus) were simultaneously recorded in both landmasses. The highest local diversity was identified in the lower montane forests of both regions: the Cameron Highlands (24 spp.), the Sinabung and Sibayak volcanoes (10 spp.) and the Kerinci massif (22 spp.; Fig. 3, Inset (A)–(B)). About two-thirds of species were recorded only in a single locality.

Dispersal-extinction cladogenesis including founder-event speciation (DEC+J; Table S4) was identified as the most appropriate model of ancestral area reconstruction. Two deeply split clades, designated as Cautires s. str. and C. pauper group, split in the Paleocene (64.1 mya; Figs. 2 and 3; Fig. S2). The C. pauper group comprised six putative species; five of them from Sumatra and C. pauper from the lowlands of the Malay Peninsula (Fig. S2). Cautires s. str. started its diversification early and 20 splits were identified from 53 to 26 mya. One deeply rooted species occurs in Sumatra (Cautires sp. U), but as it is a single species, we cannot estimate its colonization history. Further, three Sumatran clades, representing 18 spp. in total, split from their Malay relatives at 26.5, 17.5 and 4.7 mya (Fig. 3; Figs. S2 and S3). Three Malay species were identified within these clades as single-species terminal lineages and their splits from the closest Sumatran relatives cannot be reliably dated (Fig. 3). We inferred 11 range shifts from Malaya to Sumatra, 10 range shifts in the opposite direction. Additionally, we inferred five transfers between the Malay lowlands and mountains (Fig. S3).

Distribution of color patterns

Similar coloration was identified in a high number of unrelated taxa (Fig. 2; Fig. S4; Table S5) and individual patterns occurred in restricted ranges: the lowland Malay Cautires were less brightly colored and had a brown to orange-brown pronotum and the humeral part of elytra with a gradual transition between bright and dark colored parts (Fig. 4). Most Malay montane species were uniformly black (13 spp.), some had a dark red colored pronotum and humeri (Fig. S5). The Sumatran low elevation species are uniformly black, have a red colored pronotum and black elytra, a black pronotum and the red basal part of elytra or they have a brown to orange-brown pronotum and the humeral part of the elytra (Fig. 4; Fig. S4). The Sumatran montane species are bright colored and they usually have a high-contrast border between dark and bright parts: 21 species are orange and black, further species are uniformly bright colored or they have the black pronotum and basal part of elytra in contrast with their bright red apical part. Distribution of all patterns is summarized in Fig. 4. The populations of a single species were generally uniform in color pattern and the observed differences were subtle. Seldom two patterns were found within a single population (observed in Cautires spp. AH, AX and T) or in two geographically distant populations (C. rianganus and C. jasarensis).

Origin of Cautires and their diversification

Deep-rooted Oriental Cautires lineages originated out of the studied region, namely in drifting India or in a contact zone between India and continental Asia at the time of their collision (55–35 mya, Sklenarova, Chesters & Bocak, 2013). Additionally, synonymous nucleotide divergence and saturation limit the robustness of deep mtDNA-based topologies. Therefore, no splits beyond ∼30 mya are considered (Fig. 3; Fig. S3). Fossil and tectonic calibrations are unavailable and the secondary calibration would be ambiguous due to sparse sampling and large differences between various analyses (Hunt et al., 2007; McKenna et al., 2015; Toussaint et al., 2017a; Bocak et al., 2016; Zhang et al., 2018; Kusy et al., 2018). Therefore, we used the rates of mitogenome evolution and discussed only Late Paleogene and Neogene splits. Our rate-based dating is supported by the congruence of the evolution of the Sumatran Cautires fauna and tectonics of Sumatra (Fig. 3).

We suppose that Cautires started their diversification in the region in the Oligocene (Fig. 3). Until the Lower Pleistocene, the C. pauper group contained only Sumatran species and only 1.5 mya a single species colonized Malaya (Figs. 2 and 3). Cautires s. str. is a lineage of Malay origins and almost all ancestral lineages only occurred on the Malay Peninsula or Asian continent, respectively, prior to the Upper Oligocene (26.5 mya; Fig. 3). The Sumatran fauna consists mostly of terminal subclades nested in older, more inclusive Malay groups similarly to Platerodrilus Pic, 1921 net-winged beetles (Masek et al., 2015). The oldest Sumatran highly diverse clade split from Malay relatives in the Upper Oligocene (10 Sumatran spp. and two Malay spp. in terminal positions; Fig. 3). Further splits between Malay and Sumatran lineages are dated to the Miocene (17.5 and 6.2 mya) and later in the Pliocene (three events, 4.7, 4.6 and 3.6 mya). Hence, the Sumatran Cautires diversified with an apparent delay compared to the Malay fauna in accord with the tectonic history (Fig. 3; Hall, 2002). The origin of the oldest Sumatran clade supports the existence of an island chain in the region from the Upper Oligocene to the Lower Miocene (Hall, 2002; Malohlava & Bocak, 2010; Masek et al., 2015). The colonization direction was asymmetrical from the Upper Oligocene until the end of the Upper Pliocene. In this period, we identified six colonization events from the Malay Peninsula to Sumatra which gave origin to multi-species clades, and no colonization in the opposite direction (Fig. 4). This pattern agrees with the hypothesis that areas with recent histories of size expansion should obtain higher levels of immigration (De Bruyn et al., 2014) and it was earlier documented in net-winged beetle Scarelus Waterhouse, 1879 and Platerodrilus (Malohlava & Bocak, 2010; Masek et al., 2015). The phylogeny and reconstruction of ancestral areas suggest a founder effect diversification model and subsequent in situ speciation producing different faunas of neighboring landmasses (Figs. 3 and 4; Matzke, 2014; Demos et al., 2016).

We identified colonization events in both directions during the Quaternary and this is in line with repeated low sea levels (Voris, 2000). We propose that colonization dynamics shifted from a tectonic to a climatic dominated regime in the last 5 million years. Aside from two species recorded from both regions, all colonization events resulted in the origin of a separate species or a whole local clade (Fig. 3). These colonization events contributed to the observed species-level diversity (Fig. 2). Although the sampling is apparently incomplete, we can conclude that most Cautires have small ranges and that the geographic speciation mode of speciation is frequent between Sumatra and the Malay Peninsula (Barraclough & Vogler, 2000; Ikeda, Nishikawa & Sota, 2012). We suppose that intensive faunal exchange between these regions would result in the presence of widespread species on both sides of the Malacca Strait. Our data do not support such a prediction.

Further aspects of the colonization and diversification history are the origin and uniqueness of the montane faunas. We identified 19 Cautires with distribution limited to the montane forests in the Malay Central Range and none of them is distributed in a wide range of elevations (Table S1; Jiruskova, Motyka & Bocak, 2016). Two clades of the Malayan montane fauna started their in situ diversification 28.1 and 12.6 mya and they represent about a half of the diversity reported from Malay mountains. Despite the limited extent of mountain regions on the Malay Peninsula and a turbulent climatic history which could have potentially caused complex altitudinal range shifts over such long periods, these two clades are dominantly mountainous and only a single species, Cautires sp. S, was inferred to be a member of the mountain clade yet was distributed in lower elevations (Fig. 4; Fig. S1). Additional nine species were recorded in the Malay mountains and the time of their split from lowland relatives cannot be exactly estimated. The Malay Central Range is a biodiversity hotspot with ancient and diverse fauna similar to other tropical mountains (Merckx et al., 2015). Similarly to the Malay Peninsula, we found a high turnover between Sumatran mountains and lowlands. Only three species were recorded simultaneously in two mountain regions of Sumatra—Cautires spp. B, N and AN. We conclude that the in situ diversification of montane species contributed to high alpha-taxonomic diversity also in Sumatra (Merckx et al., 2015; Demos et al., 2016).

To avoid a potential source of error when closely related species are delimited, we can alternatively consider the phylogeny of Cautires only from the Rupelian Stage of the Oligocene (∼30 mya) to the beginning of the Pliocene (∼5 mya). Cautires s. str. already contained 17 separate lineages at the beginning of the Oligocene, all (with one exception) known only from the Malay Peninsula (Cautires sp. G, see Fig. 3). Before Sumatra was uplifted, some of these lineages dispersed to proto-Sumatran islands and diversified there. Further colonization events are hypothesized from the Malay Peninsula to Sumatra in the Upper Miocene and the Lower Pliocene, about 5–7 mya (Fig. 3). At the beginning of the Pliocene, the Cautires s. str. hypothetically contained 45 lineages, 17 of them Sumatran (Fig. 4). The data suggest that despite geographic proximity, colonization events had been rare in the region for a long time and unique Malayan and Sumatran faunas were established already before the Pliocene/Pleistocene period.

Why does the Malacca Strait separate different faunas?

The Malacca Strait is shallow and, especially in the southern part, very narrow, so it should not represent a major dispersal barrier for flying insects (Fig. 1; Balke et al., 2009; Taänzler et al., 2014; Toussaint et al., 2017b). Additionally, very similar ecosystems are currently present on the Malay Peninsula and Sumatra and we suppose that the narrow Malacca Strait never separated ecosystems whose differences could substantially lower the colonization success (Thomas, 1994; Morley, 2000). We cannot exclude a possibility that some species might colonize other islands on the Sunda Shelf first and only then colonize Sumatra or Malaya, respectively. The distinct fauna of Borneo (Sklenarova, Chesters & Bocak, 2013; Masek et al., 2018) indicates that colonization via Borneo did not dominate and can only marginally affect the species-level structure of both studied faunas. Observed small ranges point to low colonization capacity which makes the shortest distance colonization direction the most probable (Lester et al., 2007; Bray & Bocak, 2016).

Hence, we can discuss further factors which might be responsible for the observed distribution. The reconstruction of the paleoclimate during recent glacial maxima indicates that the Malacca Strait was covered by semi-dry savannah (Cannon, Morley & Bush, 2009; DiNezio & Tierney, 2013). We identified a lower abundance and diversity of Cautires in lowland localities characterized by a more pronounced dry season than in mountain ecosystems. The ecosystems of the exposed Shelf during glacial maxima were unfavorable (Gathorne-Hardy et al., 2002) and although more colonization events were recovered since the Pliocene, most species are endemic to the respective area.

Factors hypothetically decreasing the colonization potential are different aposematic color patterns in the lowlands and individual mountain massifs of the Malay Peninsula and Sumatra. Similarly-colored Cautires are unrelated (Fig. 2) and the geographic distribution of various aposematic patterns is limited (Fig. 4). Based on these facts, we can hypothesize that dispersing Cautires regularly entered the area where their aposematic signal was uncommon or absent. Therefore, we propose that some dispersing populations could be wiped out by local predators unfamiliar with their allochthonous aposematic signal before they could adapt to local mimetic complexes, that is, they could be under antiapostatic selection decreasing colonization rates (Beatty, Beirinckx & Sherratt, 2004; Sherratt, 2008).