Zoologica Scripta

Range expansion and ancestral niche reconstruction in the Mediterranean diving genus Meladema (Coleoptera, ) VIT SYKORA,DAVID GARCIA-VAZQUEZ,DAVID SANCHEZ-FERNANDEZ &IGNACIO RIBERA

Submitted: 16 July 2016 Sykora, V., Garcıa-Vazquez, D., Sanchez-Fernandez D. & Ribera, I. (2017). Range expan- Accepted: 28 October 2016 sion and ancestral niche reconstruction in the Mediterranean diving beetle genus Meladema doi:10.1111/zsc.12229 (Coleoptera, Dytiscidae). —Zoologica Scripta, 00, 000–000. Species of the genus Meladema (Dytiscidae, ) are some of the largest macroin- vertebrates in the western Palearctic region, being top predators in fishless streams. Two of the three described species, Meladema imbricata (Wollaston, 1871) and Meladema lanio (Fabricius, 1775) are Macaronesian endemics from the Canary Islands and Madeira, respec- tively, while the third, Meladema coriacea Laporte, 1835, is widely distributed from Morocco and the Iberian Peninsula to Turkey, including the Canary Islands. Previous phylogenetic analysis using only mitochondrial markers revealed the existence of two cryptic lineages within M. coriacea, one restricted to and the other including the rest of sampled populations. We reconstruct here the evolutionary history of the species of Meladema using a more comprehensive sampling covering its whole geographical range, adding nuclear markers and Bayesian molecular dating. Using environmental niche modelling, we test for possible differences in climatic preferences among lineages and reconstruct their ancestral climatic niche. Our results strongly supported the existence of four monophyletic lineages represented by the three recognized species plus a fourth cryptic lineage with populations of M. coriacea from the Tyrrhenian islands (Corsica, and Montecristo). This pattern is not likely to be the result of mitochondrial artefacts due to Wolbachia infection, as all 11 tested individuals were negative for this parasite. Dating analysis placed the origin of Mela- dema in the Middle Miocene although diversification among extant Meladema lineages started in the early Pleistocene and took place in a relatively short time period. Phylogeo- graphic analysis inferred a continental origin of Meladema, with an independent colonization of the Macaronesian and Mediterranean islands. From the south-western Mediterranean region, the continental M. coriacea expanded its range up to and Turkey in the northern basin and to Tunisia in the southern. Results of niche modelling showed that sea- sonality is the critical factor in shaping the current distribution of Meladema. Island lineages (M. imbricata, M. lanio and the Tyrrhenian lineage of M. coriacea) occur in sites with low seasonality, within the range of the reconstructed ancestral climatic niche of the genus. On the contrary, continental M. coriacea expanded its range to localities outside the ancestral cli- matic range of the genus, with a higher seasonality and aridity. Corresponding author: Ignacio Ribera, Institute of Evolutionary Biology (CSIC-Universitat Pompeu Fabra), Passeig Maritim de la Barceloneta 37, 08003 Barcelona, . E-mail: ignacio. [email protected] Vit Sykora, Faculty of Science, Charles University in Prague, Albertov 6, 128 43 Praha 2, Prague, . E-mail: [email protected] David Garcıa-Vazquez, Institute of Evolutionary Biology (CSIC-Universitat Pompeu Fabra), Pas- seig Maritim de la Barceloneta 37, 08003 Barcelona, Spain. E-mail: [email protected] David Sanchez-Fernandez, Instituto de Ciencias Ambientales, Universidad de Castilla-La Mancha, Campus Tecnologico de la Fabrica de Armas, Toledo, 45071 Spain. E-mail: david.sfernandez@ uclm.es Ignacio Ribera, Institute of Evolutionary Biology (CSIC-Universitat Pompeu Fabra), Passeig Mar- itim de la Barceloneta 37, 08003 Barcelona, Spain. E-mail: [email protected]

ª 2017 Royal Swedish Academy of Sciences 1 Evolutionary history of Meladema V. Sykora et al.

Introduction molecular clock dating. We also include environmental The diving beetle genus Meladema (Dytiscidae, Colymbeti- niche modelling to test for differences in climatic prefer- nae) is a common inhabitant of Mediterranean streams, in ences among individual lineages, trying to understand the which it is often one of the largest macroinvertebrates. factors determining the current and past distribution of With a body size ranging from 20 to 23 mm, in small, the species of this emblematic genus of freshwater fishless streams it may also be a top predator, both in the macroinvertebrate. adult and larval stages. The genus currently includes three species: Meladema coriacea Laporte, 1835, Meladema imbri- Material and methods cata (Wollaston, 1871) and Meladema lanio (Fabricius, 1775) Sampling, DNA extraction and sequencing (Ribera et al. 2003; Alarie & Hughes 2006). Meladema is We sampled all three known species of the genus Meladema confined to the western Palaearctic region, with one spe- (M. coriacea, M. imbricata and M. lanio) from 38 localities cies (M. coriacea) distributed from Morocco and the Iberian covering the whole range of all species (Fig. 1; see Peninsula to Turkey and the other two endemic to Mac- Table S1 for a list of specimens included in the study). aronesia. The widespread M. coriacea is common in the Specimens were collected and preserved in absolute ethanol western Mediterranean, including the Iberian Peninsula, directly in the field. We extracted the DNA non-destruc- Southern , and all the western tively with commercial kits (mostly ‘DNeasy Tissue Kit’; Mediterranean islands (Nilsson & Hajek 2015). It seems to Qiagen GmbH, Hilden, ) following the manufac- become scarcer towards the east, with few records from turer’s instructions. Specimens and DNA extractions are the Balkan Peninsula (Zaitsev 1972; Nilsson & Hajek kept in the collections of the Institut de Biologia Evolutiva, 2015) and a single specimen reported from Turkey in the Barcelona (IBE) and the Museo Nacional de Ciencias Nat- literature (Gueorguiev 1981; Darilmaz & Kiyak 2009). It is urales, Madrid (MNCN). We used the data of Ribera et al. also present in the Canary Islands, likely resulting from a (2003), to which we added 17 newly sequenced specimens, recent colonization (Ribera et al. 2003). There are refer- especially from areas not previously studied in the Central ences of the presence of M. coriacea in northern France and Eastern Mediterranean, and nuclear data. Six gene and in (Gschwendtner 1936), but even when cor- fragments from five different genes (three mitochondrial rect, these were undoubtedly occasional records. In con- and two nuclear) were obtained in four different amplifica- trast, M. imbricata and M. lanio are endemics to the tion reactions: (i) 30end of Cytochrome Oxidase Subunit 1 0 0 Canary Islands and Madeira respectively, in where they are (COI-3 ); (ii) 5 end of 16S rRNA plus tRNA transfer of 0 of high conservation concern due to their confinement to Leucine plus 3 end of NADH subunit 1 (16S + tRNA- well preserved streams (Machado 1987; Balke et al. 1990; Leu + nad1); (iii) an internal fragment of the nuclear gene Ribera et al. 2003). All three species occupy running Histone 3 (H3); (iv) a fragment of the nuclear gene Wing- waters, mainly deeper pools in streams at intermediate alti- less (Wg) (see Table S2 for the primers used and general tudes (Ribera et al. 2003). cycling conditions). The new 125 new sequences have been In a previous study using only mitochondrial DNA deposited in the EMBL database with accession numbers sequences, Ribera et al. (2003) found that Meladema was in LT602717-LT602841 (Table S1). Sequences of outgroup fact constituted by four genetically distinct lineages with taxa were downloaded from GenBank: a representation of unresolved relationships between them. The two Macarone- other genera of subfamily Colymbetinae, including the sian endemics M. imbricata and M. lanio were found to be closest known relatives of Meladema (Moriniere et al. 2015, respectively monophyletic, but the analyses suggested that 2016), and one genus of Agabinae (Agabus) to root the tree, M. coriacea was not monophyletic, as it was split into a lin- which is likely the sister subfamily of Colymbetinae (Ribera eage restricted to Corsica and another including the rest of et al. 2008; Miller & Bergsten 2014). DNA sequences were sampled populations (Morocco, the Iberian peninsula and assembled and edited using GENEIOUS v1.6 (Biomatters, southern France). Despite uncertainties in the calibration, http://www.geneious.com). Alignments of variable length diversification within the genus was estimated to be of early sequences were obtained with the MAFFT plugin v7.017 Pleistocene origin. None of the lineages seem therefore to (Katoh et al. 2002) in GENEIOUS v1.6. be a Tertiary relict, as hypothesized for other Macaronesian The likely presence of cryptic mitochondrial lineages endemics (Ribera et al. 2003). within Meladema (Ribera et al. 2003) raised the possibility In this study, we present an update of the phylogeny that specimens could be infected with Wolbachia, a mater- and biogeography of the species of Meladema, with a nally transmitted parasite that can alter the patterns of more comprehensive sampling encompassing their whole mtDNA variability (Jiggins 2003). We thus tested for the geographical range, nuclear markers and a Bayesian presence of Wolbachia in 11 selected specimens of the two

2 ª 2017 Royal Swedish Academy of Sciences V. Sykora et al. Evolutionary history of Meladema

Fig. 1 Location of the studied specimens. Black, continental Meladema coriacea; blue, M. coriacea CSM; red, Meladema imbricata; green, Meladema lanio. See Table S1 for details of the localities. lineages of M. coriacea (Table S1). We used specific primers respectively. Molecular clock models were linked in three for the wsp gene, trying to amplify a 632-bp fragment. A partitions corresponding to the mitochondrial protein cod- 0 3 ll sample of the PCR reaction mixture was elec- ing genes (COI-3 + nad1), 16S and nuclear genes trophoresed with a 100-bp DNA ladder on 1% agarose gel (H3 + Wg), and we ran two analyses to test the clock to determine the presence and size of the amplified DNA model (strict or lognormal relaxed) that best fitted the data. bands that were visualized by Sybr-Safe staining. The same We applied an a priori rate of 0.0145 substitutions/site/MY methods have yielded positive results, thus detecting the (standard deviation 0.002) for the protein coding genes and presence of Wolbachia infested individuals, in another genus 0.0016 substitutions/site/MY (standard deviation 0.0002) of the same family (Deronectes, Garcıa-Vazquez & Ribera for 16S. These are substitution rates estimated for a related 2016). group, family Carabidae, based on a combination of fossils and biogeographic events (Andujar et al. 2012). Clock rates Data sets for molecular analyses of H3 and Wg were left with uniform priors due to the We used three different data sets in our analyses: (i) phylo- absence of any suitable estimations of the evolutionary rate genetic and divergence time analysis with the complete, for these nuclear genes. We used a Yule speciation prior combined sequence and with only nuclear data; (ii) coales- for all of the partitions and ran the analysis for 100 million cence analysis with only mitochondrial data; (iii) phylo- generations, logging parameters every 5000 generations. geography and climatic niche reconstruction with Unless otherwise stated, we used a conservative burn-in of mitochondrial and nuclear data. 10% for combined data sets of both mitochondrial and nuclear DNA, after checking convergence of all parameters Phylogenetic and divergence time analysis in TRACER v1.6 (Rambaut et al. 2014). For the phylogenetic and divergence time analysis, we used We performed four runs with different topological con- only specimens with different haplotypes of the COI-30 straints: (i) M. imbricata + M. lanio as monophyletic; (ii) gene, plus the corresponding sequences of the rest of the M. coriacea as monophyletic; (iii) M. lanio, M. imbricata and genes. The final length of the data matrix was 2439 nucleo- M. coriacea (only specimens from Corsica, Sardinia and tides for 52 specimens. Two specimens of M. coriacea Montecristo) as monophyletic; (iv) without any constraints. known to have mtDNA haplotypes of M. imbricata and We then compared tree likelihoods in TRACER v1.6. A con- three specimens of M. imbricata with mtDNA of M. cori- sensus tree was obtained via TREE ANNOTATOR v1.8 (Drum- acea from Tenerife were also excluded from the analyses. mond et al. 2012) and visualized using FIGTREE v1.4.2 These specimens were identified in preliminary phyloge- (http://tree.bio.ed.ac.uk/software/figtree/). netic analyses performed with sequences of all available We also analysed the nuclear genes only to test for specimens of the genus Meladema (see Results below and potential incongruence between the nuclear and mitochon- Ribera et al. 2003). drial genomes. The data set for the nuclear sequence con- The data set was divided into five partitions correspond- tained 46 specimens and the aligned sequence was 805 ing to each gene, and we used Partition FINDER v1.1.1 nucleotides long, and was analysed with a fast maximum- (Lanfear et al. 2012) to estimate the evolutionary model likelihood algorithm implemented in RAxML (Stamatakis that best fitted the data for each partition separately, using et al. 2008) with a single partition and a GTR model. Bayesian Information Criterion (BIC) as selection criteria. Node support was assessed with 500 fast bootstrap replicas. We analysed the data with BEAST 1.8 (Drummond et al. 2012) linking substitution models in two partitions corre- Coalescence analysis sponding to the mitochondrial (COI-30, 16S + tRNA-Leu We used all available mitochondrial sequences of the conti- (16S onwards) and nad1) and nuclear (H3 + Wg) markers, nental M. coriacea for analyses of coalescence and Bayesian

ª 2017 Royal Swedish Academy of Sciences 3 Evolutionary history of Meladema V. Sykora et al.

Skyline Plots (Drummond et al. 2005) to estimate the popu- Meladema lineages to characterize their environmental lation history of this species. A similar methodology has been niche and to test for niche divergence among them. As a successfully applied to relate demographic models and geo- source of climatic data, we used WORLDCLIM v1.4 and graphical expansion of other water species complexes obtained 19 climatic variables (http://www.worldclim.org; (Hidalgo-Galiana et al. 2014). The aligned nucleotide matrix Hijmans et al. 2005) (Table S3). We included localities of of 35 specimens was 1632 nucleotides long. Both coalescence all sequenced specimens, plus additional localities for main- and Bayesian Skyline Plot analyses were carried out in BEAST land from CKMAP v5.4.1 (Latella et al. 2007), which 1.8. The data set was divided into three partitions corre- were considered separately in some specific analyses. To 0 sponding to each mitochondrial gene (COI-3 , 16S and characterize the preferred climatic conditions for each spe- nad1), with a GTR substitution model. We estimated the cies, we represented the location of the occurrence cells of best clock model and set an a priori rate of 0.0134 substitu- each of them across the climatic space of the study area. tions/site/MY with a standard deviation of 0.001 for the Values of all 19 bioclimatic variables were summarized combined sequence, estimated for the family Carabidae using a principal component analysis (PCA) to obtain (Andujar et al. 2012). Preliminary analyses had shown that if uncorrelated environmental factors (Varimax rotation). the sequence was split into different partitions, there were Values of the first two PCA factors were used to plot a severe convergence problems, so we used a single concate- bidimensional climatic space of the study area (e.g. nated partition. To identify the best demographic model, we Sanchez-Fernandez et al. 2013) and to locate sampling ran four analyses, including constant size, exponential localities of each lineage of Meladema, including the Italian growth, logistic growth and expansion growth. Models were mainland localities obtained from CKMAP (Table S3). To then compared using modified Akaike information criterion test for significant differences in climatic conditions among (AICM) with moments estimator (Baele et al. 2012) with 100 Meladema species, we also conducted one-way ANOVA bootstrap replicates, as implemented in TRACER v1.6. The analyses of the first three PCA factors with post hoc tests Markov chain Monte Carlo (MCMC) chain was set to using the Bonferroni correction. All statistical analyses 100 million generations, logging parameters every 5000 were conducted using SPSS 15.0.1 and STATISTICA v8.0 generations. The consensus tree was obtained with TREE (www.statsoft.com, 2007). ANNOTATOR v1.8 and visualized using FIGTREE v1.4.2. The To compare the climatic niche between lineages, we Bayesian Skyline Plot was calculated in TRACER v1.6. generated ecological niche models using MAXENT v3.3.3 (Phillips et al. 2006). To avoid autocorrelation in the vari- Phylogeographic analysis and environmental niche ables, we also used the three-first factors of the climatic modelling PCA as predictors. We included all known localities for all For an estimation of the dynamics of the geographic lineages except for continental M. coriacea, for which we expansion of the genus Meladema, we performed a continu- included only localities with sequenced specimens to avoid ous phylogeography analysis in BEAST v1.8. The data sets uncertainties. The localities with molecular data are wide- for both phylogeographic and climatic niche analyses spread in the climatic space, providing a good geographic included sequences of all available specimens, excluding the and climatic subsample of all known records (Fig. 1; Tables hybrid specimens mentioned above. The aligned nucleotide S1, S3), minimizing the risk of bias due to unbalanced cli- matrix of 63 specimens was 2436 nucleotide long. The data matic and spatial sampling effort. MAXENT is a machine- set was divided into three partitions, COI-30, 16S + nad1 learning method that estimates organisms’ distribution by and H3 + Wg, with estimated optimal substitution models finding the probability distribution of maximum entropy for each partition. Another partition containing specimen (i.e. the most uniform), given the constraint that the coordinates as a continuous trait was created, with a expected value of each environmental predictor under this homogenous Brownian evolutionary model. We set the estimated distribution matches the empirical average of best clock model for mitochondrial and nuclear partitions, sample locations (Phillips et al. 2006). MAXENT simulations with a priori clock rates as for analyses before. We ran the of realized distributions produce continuous suitability estimated optimal coalescent model for 100 million genera- scores for each cell (from 0 to 1), and we calculated Scho- tions with parameters logging every 5000 generations. The ener’s D metric (Schoener 1968) using ENMtools (Warren consensus tree was obtained in the same way as with previ- et al. 2010) to assess niche similarity among populations, = ½ ous analyses. The geographic range dynamics of Meladema where D (pX, pY) 1 S (pXi pYi). This metric was obtained using SPREAD v1.0.6 (Bielejec et al. 2011) and assumes probability distributions defined over geographic visualized in GOOGLE EARTH v7. space, in which pXi (or pYi) denotes the probability We used environmental niche modelling based on large- assigned by the modelling method to species X (or Y)in scale climatic data and known presence localities of all the i cell ranging from 0 (niches do not overlap) to 1

4 ª 2017 Royal Swedish Academy of Sciences V. Sykora et al. Evolutionary history of Meladema

(niches completely overlap). The null hypothesis of the homogenous Brownian model of evolution. We ran the background similarity test states that observed niche over- ancestral character state reconstruction analysis for lap between taxa is explained by regional similarities in 100 million generations with parameters logging every available background environments. This hypothesis 5000 generations. For each of the main nodes, we identi- involves a two-tailed test, so it is rejected if the observed D fied the geographic area with climatic conditions encom- between two taxa falls outside the 95% confidence limits of passed by the 95% interval of the reconstructed value for the null distribution. Niche conservatism is supported all three PCA factors. when the observed value of D is larger than the upper 95% confidence limit of the null distribution, suggesting Results that niches are more similar than expected based on their Data sets for molecular analyses background environments (i.e. species are occupying niches There were no length differences in protein coding genes. that are as similar as possible given what is available). The only length differences in alignment were three inser- Niche divergence is supported when the observed value of tions in ribosomal genes in outgroup taxa and one single D is smaller than the lower 95% confidence limit of the nucleotide deletion in several specimens of M. coriacea. null distribution, suggesting that niches are more divergent The GTR + I + G (Tavare 1986) was identified as the 0 than expected based on background divergence. Because best substitution model for the COI-3 , 16S and nad1 parti- niche differences may be simply result of the spatial auto- tions, and the TRN + I + G (Tamura & Nei 1993) for the correlation of the used explanatory environmental variables H3 partition. For the Wg partition, the K80 + I (Kimura (background environmental divergence, Warren et al. 1980) was identified as the most suitable, but as it is not 2008), strong evidence for niche divergence requires two implemented in BEAST 1.8, we used the HKY model instead conditions: (i) that niche characteristics differ between the (Hasegawa et al. 1985), with equal base frequencies (Drum- two considered populations and (ii) that these differences mond et al. 2012). The same substitution models were used are greater than the background environmental divergence for each gene in all data sets. (McCormack et al. 2010). Niche conservatism, on the other hand, would be supported if niche differences were smaller Phylogenetic and divergence time analysis than the obtained background environmental divergence. The strict clock was significantly better than the lognormal Thus, comparison of environmental characteristics from clock when using the combined data (Table 1), and in con- these two classes of data should allow discrimination sequence we used the strict clock in all analyses. The com- between differences as a result of simple spatial autocorre- parison of likelihood scores among trees with topological lation caused by geographic distance and strong niche constraints showed no significant differences between like- divergence that occurs because two species occupy different lihood scores of the tree without any constraints and trees environmental conditions. To test the null hypothesis that with constrained clades (with <5 units of AICM between niches are similarly divergent in comparison with back- the most divergent values, Table 2), so we used the topol- ground environments, we used the ‘background similarity ogy without any prior constraints. test’ procedure implemented in ENMtools, estimating 100 The analysis of phylogenetic relationships including only niche overlap values generated by comparing model suit- specimens with unique haplotypes supported a sister rela- ability values of one population to those generated from tionship of Corsican, Sardinian and Montecristo specimens random cells drawn from the geographic range of the other of M. coriacea (M. coriacea CSM onwards) and the rest of population (Warren et al. 2010). The background area of the genus (Bayesian posterior probability, pp = 1; Fig. 2). each species should be adjusted to the habitat available and The rest of specimens of M. coriacea (continental M. cori- should be biologically realistic (Warren et al. 2010). As acea onwards, although including specimens from the three of the lineages within the genus are insular (see Balearic and Canary Islands and ) were placed as Results), the background area was restricted to the islands in which the species appear. For continental M. coriacea, we Table 1 Likelihood and AICM scores for the BEAST analyses of the defined the background as the whole study area. combined data with different clock models, calculated in TRACER We also performed an ancestral character state recon- Analysis Clock model Likelihood score AICM struction analysis in BEAST v1.8 to estimate ancestral cli- matic preferences of individual Meladema lineages. All Phylogeny and dating Lognormal clock 7875.6 15994.9 substitution models and settings were the same as in coales- Phylogeny and dating Strict clock 7889.8 15962.3 cence analyses above. We created another partition repre- Phylogeny (nuclear genes) Lognormal clock 2218.6 4494.6 Phylogeny (nuclear genes) Strict clock 2240.5 4560.6 senting the first three best scoring PCA factors used in previous PCA analysis as a continuous trait, with a AICM, Akaike information criterion. Lower AICM values indicate a better model fit.

ª 2017 Royal Swedish Academy of Sciences 5 Evolutionary history of Meladema V. Sykora et al.

Table 2 Likelihood and AICM scores for the BEAST analyses of the respectively, M. lanio (pp = 1), M. imbricata (pp = 1), combined data with different topological constraints under a strict M. coriacea CSM (pp = 1) and the continental M. coriacea clock, calculated in TRACER (pp = 0.98) (Fig. 2). The dating analysis using an a priori rate obtained for Topological constraint Likelihood score AICM the family Carabidae showed a very long branch between Meladema lanio + Meladema imbricata 7890.2 15964.0 the ingroup and outgroup taxa, with an estimation for Meladema coriacea 7889.3 15959.5 the origin of the genus Meladema at ~14.4 Ma (million M. lanio + M. imbricata + M. coriacea CSM 7889.9 15959.4 years ago) (95% CI of 9.99–20.02 Ma), in the Miocene No constraints 7889.8 15962.3 (Fig. 2). The crown age of the genus Meladema was esti- fi AICM, Akaike information criterion. Lower AICM values indicate a better model t. mated at ~1.49 Ma (95% CI of 1.01–2.05 Ma), with the sister of a clade including M. imbricata and M. lanio youngest speciation between M. imbricata and M. lanio (Fig. 2), but with a low support (pp < 0.50). On the con- dated to ~1.17 Ma (95% CI of 0.77–1.69 Ma), in a rela- trary, the monophyly of M. imbricata plus M. lanio was tively short time period in the Pleistocene strongly supported (pp = 0.96), as well as monophyly of, (Fig. 2).

Agabus bipustulatus

Neoscutopterus hornii Outgroup 16.17 Hoperius planatus 21.45 fuscus Bunites distigma 12.7 grapii 11.5 18.14 Rhantus cheesmanae 4.11 Rhantus suturalis 15.91 10.79 Rhantus gutticollis 2.51 Rhantus souzannae

cori12c (Corsica) coriacea 0.25 cori12e (Corsica) 14.42 cori12b (Corsica) 0.38 0.09 cori12d (Corsica)

0.23 cori5 (Sardinia) CSM 0.16 cori12f (Corsica) 0.11 cori290 (Montecristo)

lanio9a (Madeira) lanio

0.18 lanio9b (Madeira) lanio298 (Madeira) 1.49 1.17 0.08 0.03 lanio8a (Madeira) imbri17b (Tenerife)

mrct coriacea imbri15a (Gomera) imbricata 0.25 0.04 imbri15b (Gomera) 0.09 imbri17a (Tenerife) 0.19 imbri5d (Tenerife) 0.17 imbri7a (La Palma) 1.38 0.08 imbri6b (La Palma) Meladema 0.05 cori6 (Algeria) 0.11 cori11c (France) cori1a (Morocco) 0.04 cori11b (France) 0.18 0.05 cori104 (Spain) cori293 (Spain) 0.1 cori13a (Spain) 0.05 cori292 (Spain) 0.39 cori861 (Spain) 0.15 cori10a (Spain) cori2a (Morocco) 0.12 0.29 cori19a (Tenerife) 0.2 cori294 (Turkey) 0.09 cori291 (Spain) 0.28 0.03 cori84 (Tunisia) cori14c (Spain) Posterior probability: cori860 (Spain) 0.2 0.07 cori23a (Mallorca) ≥ 0.95 0.17 cori4 (Algeria) 0.03 cori21a (Morocco) ≥ 0.85 0.16 cori1064 (Malta) cori14a (Spain) Meladema imbricata ≥ 0.5 0.13 0.11 cori20a (Gran Canaria) 0.05 cori2b (Morocco) MiocenePlio. Pleistocene 1916 13 10 4 3 1.5 1.4 1.3 1.2 1.1 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Million years ago

Fig. 2 Reconstructed phylogeny of Meladema obtained with BEAST with the combined nuclear and mitochondrial sequence. Numbers at nodes, estimated age (Ma); coloured points, node support. See Table S1 for details of the specimens.

6 ª 2017 Royal Swedish Academy of Sciences V. Sykora et al. Evolutionary history of Meladema

For nuclear partitions, the lognormal relaxed clock was Sardinia and Montecristo. The basal splitting events within considered as a more suitable clock model (Table 1). In the continental M. coriacea took place in north-western Africa, H3 nuclear protein coding gene sequence, there were only with subsequent expansions through the western Mediter- two variable positions, one synonym change synapomor- ranean and possibly several secondary colonizations of the phic for M. imbricata plus M. lanio and another with Canary Islands. The eastern most sampled specimens ambiguous callings in most individuals, without a clear pat- (Malta and Turkey) were not derived from the northern tern. In the Wingless nuclear protein coding gene sequence, lineages (Iberian peninsula and southern France), but origi- there were three variable nucleotide positions with changes nated directly from within the north African stock (see shared by: (i) M. imbricata + M. lanio; (ii) M. coriacea CSM Data S1 for a phylogeographic reconstruction visualized in and (iii) M. coriacea CSM + M. imbricata + M. lanio.As Google Earth). happened in H3, there was more variability among out- group taxa. The monophyly of M. coriacea CSM and the The presence of Wolbachia sister relationship of M. imbricata and M. lanio were thus None of the 11 tested specimens was positive for the pres- corroborated by nuclear DNA (Fig. S1). ence of Wolbachia.

Coalescence analysis Environmental niche modelling The best demographic model for the continental M. cori- The main two factors of the climatic PCA jointly acea was an expansion growth, although with an AICM accounted for a high percentage of the total variance almost identical to the constant size model (Table 3). The (78.7%; the first three factors 85.8%). The first factor was analysis using the expansion growth model dated the age of positively correlated with annual mean temperature and the most recent common ancestor (MRCA) of the conti- negatively with annual precipitation and was interpreted as nental M. coriacea lineage to ~145 thousand years before representing an ‘aridity’ gradient with higher values for hot present (Ka) (95% confidence interval 71–246 Ka), in the and dry sites. The second axis was negatively correlated Late Pleistocene (Table 3). All analyses using different with temperature seasonality and was interpreted as repre- demographic models estimated a Late Pleistocene age of senting a ‘seasonality’ gradient. The third factor (with the MRCA of continental M. coriacea (Table 3). <10% of variance explained) showed a positive correlation The Bayesian Skyline Plot showed a continuous increase with precipitation of the coldest quarter and negative with in the population size of continental M. coriacea during the mean temperature of the wettest quarter, and its interpreta- last 56 Ka, with a slight decrease towards the present tion was less clear. The environmental space of continental (Fig. S2). M. coriacea almost covered that of all other lineages, with the only exception of M. lanio (Fig. 3). Among the island Phylogeographic analysis lineages, M. lanio and M. imbricata were closer to each The phylogeographic analysis inferred a continental origin other, although M. lanio occupied sites with lower seasonal- of Meladema, although the lack of well-defined outgroups ity, while M. coriacea CSM occupied relatively more sea- did not allow a precise reconstruction. Macaronesian and sonal sites (Fig. 3). The environmental space of the Mediterranean islands were colonized independently from peninsular Italian M. coriacea fitted into the space occupied the continent. There were two main clades within the by continental M. coriacea and M. coriacea CSM, with a M. coriacea CSM lineage, one with exclusively specimens higher seasonality than in M. imbricata and M. lanio from Corsica and a second with specimens from Corsica, (Fig. 3). Results of the ANOVA analyses showed differences among the four main lineages for the first three main PCA factors Table 3 Ages (in thousands of years before present) of the MRCA (Fig. 4). Post hoc analyses with Bonferroni corrections of continental Meladema coriacea estimated by the coalescence anal- fi ysis, with the 95% confidence intervals of the MRCA ages and showed that continental M. coriacea occurs in signi cantly likelihood and AICM scores of individual demographic models more arid localities than M. coriacea CSM (first factor), and M. lanio in significantly less seasonal localities than conti- Demographic model MRCA age (Ka) 95% CI (Ka) Likelihood score AICM nental M. coriacea and M. coriacea CSM (second factor;

Constant size 0.170 (0.084–0.281) 2327.1 4692.2 Fig. 4). ’ Logistic growth 0.117 (0.061–0.194) 2328.7 4698.7 Results of the MAXENT analysis estimated by Schoener s Expansion growth 0.145 (0.071–0.246) 2327.2 4692.1 D showed a low niche overlap among the four main lin- Exponential growth 0.123 (0.061–0.209) 2327.7 4697.5 eages (Fig. S3), with the highest between continental M. co- AICM, Akaike information criterion; MRCA most recent common ancestor. Lower riacea and M. coriacea CSM (0.48) and M. imbricata and AICM values indicate a better model fit. M. lanio (0.41). The lowest overlaps were between M. lanio

ª 2017 Royal Swedish Academy of Sciences 7 Evolutionary history of Meladema V. Sykora et al.

Fig. 3 Pooled climatic space of the lineages of Meladema as represented with the two-first principal component analysis factors of the climatic variables (Table S3). Grey background, climatic space covered by the whole studied area; in colours, climatic space occupied by the different lineages plus continental Italy. White dots indicate the localities of continental Meladema coriacea with molecular data. and both lineages of M. coriacea, continental (0.08) and three recognized species (Ribera et al. 2003). This lineage CSM (0.09) (Fig. 5; Table S4). In the background similar- was apparently morphologically indistinguishable from the ity tests, the null hypothesis of no niche difference was widespread M. coriacea, but found only in Corsica. This rejected in both directions in four of the six pairwise com- study had, however, two important shortcomings: it was parisons. In the other two, the comparison of M. imbricata based on mitochondrial data only, and the geographical against M. coriacea CSM and of M. lanio against M. coriacea sampling was very incomplete. Our results using both CSM only one direction showed niche divergence (Fig. 5). mitochondrial and nuclear DNA and a more comprehen- sive sampling corroborated the existence of this cryptic lin- Reconstruction of ancestral climatic preferences eage within the genus. Although we found very small We obtained the 95% Highest Posterior Density (HPD) amount of variability in the nuclear genes used in our anal- intervals in the BEAST analysis for the first two PCA factors yses, the variable nucleotide positions within them sup- (Table S5). The analysis showed that all studied specimens ported the existence of a distinct lineage of M. coriacea of M. coriacea CSM, M. imbricata and M. lanio occurred in from Corsica, Sardinia and the island of Montecristo, in climatic conditions reconstructed as being ancestral for the the Tuscan archipelago (M. coriacea CSM). The genus and whole genus Meladema (Fig. 6). On the contrary, some cur- each of four previously suggested lineages were strongly rent localities of continental M. coriacea (mostly in Mor- supported as monophyletic, although the internal relation- occo, Spain and Turkey) did not fit into the 95% HPD ships within the genus were not resolved reliably except for interval for at least one reconstructed PCA factor, suggest- the sister relationship between the Macaronesian species ing occurrence in non-ancestral conditions for the genus M. lanio and M. imbricata. (Fig. 6; Table S5). When considering the ancestral condi- The origin of Meladema was dated back to the Middle tions of each lineage individually, for all insular lineages, Miocene. However, the diversification among extant Mela- the climatic conditions of the sampled localities fall within dema lineages started much later, in the Early Pleistocene. the 95% HPD interval of the ancestral conditions of the According to our estimate, these events took place in a rel- same lineage only. This is in contrast to some specimens of atively short time period, suggesting the possibility of a continental M. coriacea, which fall within the 95% HPD nearly simultaneous speciation. The poor support for the interval of other lineages (Fig. 6; Table S5). relationships among the lineages with an exception of the sister relationship of M. lanio and M. imbricata would also Discussion imply that the speciation took place in a short sequence of Phylogeny and age of Meladema events. These age estimates are very close to those sug- Previous work on the genus Meladema had revealed the gested by the analysis of Ribera et al. (2003), which used a existence of a fourth, cryptic lineage in addition to the standard rate of variation for the mitochondrial genes of

8 ª 2017 Royal Swedish Academy of Sciences V. Sykora et al. Evolutionary history of Meladema

present a large population of M. coriacea before the isola- tion from the mainland, which may have dampen the mor- phological variation in the CSM lineage. In contrast, if the colonization of Macaronesian islands was only by a small number of ancestral Meladema individuals, this might have resulted in a founder effect-driven fast morphological evo- lution. Another explanation could be an occasional gene flow between the CSM lineage and continental M. coriacea preventing morphological differentiation, while not being enough to leave a clear signal in mtDNA. However, we do not have any clear evidence for either of these two scenar- ios (which are not mutually exclusive). A third possible explanation for the absence of morphological differences between M. coriacea CSM and continental M. coriacea could have been the presence of Wolbachia, which can cause cyto- plasmic incompatibility and thus reproductive isolation (Werren et al. 2008). Recent works gave an estimate of 40% of Wolbachia prevalence among (Zug & Hammerstein 2012) and 31% in a group of families of aquatic beetles, among them Dytiscidae (Sontowski et al. 2015), a frequency agreeing with recent reports of the same family (Garcıa-Vazquez & Ribera 2016). However, none of the specimens of Meladema tested was positive for the pres- ence of these endosymbiotic bacteria.

Phylogeography and climatic preferences of Meladema Fig. 4 ANOVA comparison of the climatic space represented by the Our phylogeographic analysis suggests the origin of Mela- three main axes of the climatic principal component analysis of the four lineages of Meladema. Letters indicate significant differences dema in the western Mediterranean, most probably in between lineages (Bonferroni’s post hoc tests, P < 0.05). southern Iberia or in north-western Africa, during the Middle Miocene. This implies that for most of their evolu- tionary history, the ancestors of extant Meladema lineages 2% per million years (Brower 1994). The age estimated for evolved under the tropical climate present in southern Eur- the crown diversification of Meladema was also in good ope and North Africa during that time (Ruddiman et al. agreement with that in the analysis of Colymbetinae phy- 1989; Griffin 2002; Jimenez-Moreno et al. 2010; Kohler€ logeny of Moriniere et al. (2016), including a wider repre- et al. 2010). The climatic change from wet tropical to sentation of the subfamily and different calibration Mediterranean, with a stronger seasonality and dry sum- schemes. Nevertheless, the lack of internal calibration mers, occurred at ~3.2 Ma, in the Pliocene (Suc 1984). points for Meladema or closely related outgroups still leaves This could have confined the ancestral Meladema to regions some level of uncertainty in molecular clock estimates in with a climate more similar to that in Miocene. Our recon- our analyses. This is especially the case for the long stem struction of ancestral climatic preferences implies that branch of Meladema, estimated here to date from the mid- M. lanio, M. imbricata and M. coriacea CSM are probably dle Miocene, but from the late Eocene in Moriniere et al. relicts inhabiting sites with conditions closer to the ances- (2016). tral for the whole genus. In contrast, continental M. cori- As was already stated by Ribera et al. (2003), M. coriacea acea was the only lineage able to colonize also sites with CSM is morphologically extremely similar to the continen- climatic conditions non-ancestral for Meladema. Environ- tal M. coriacea, and no diagnostic differences have been mental niche modelling analyses showed that seasonality is identified so far, not even in male genitalia. In contrast, the most important factor in current distribution of Mela- there are clear differences both in genitalia and in external dema. All island lineages (M. lanio, M. imbricata and M. co- morphology between Macaronesian species and M. coriacea riacea CSM) inhabit on average less seasonal sites than (Machado 1987). This could have several possible reasons. continental M. coriacea, which is the lineage occupying the One explanation could be a difference in population sizes. highest diversity of climatic conditions. Among the island It is likely that in the case of M. coriacea CSM, there was lineages, endemic Macaronesian species inhabit the least

ª 2017 Royal Swedish Academy of Sciences 9 Evolutionary history of Meladema V. Sykora et al.

Fig. 5 Background similarity test comparisons between the four lineages of Meladema. Observed values of niche overlap (arrows) were tested against null distributions (100 replicates) generated by comparing model suitability values of one lineage with those generated from random cells drawn from the background area of the other lineage. Only in (C) and (D), one of the comparisons was not significant (see text for details). —A. Continental Meladema coriacea 9 M. coriacea CSM; (B) continental M. coriacea 9 Meladema lanio; (C) Meladema imbricata 9 continental M. coriacea CSM; (D) M. coriacea CSM 9 M. lanio; (E) M. imbricata 9 continental M. coriacea; (F) M. imbricata 9 M. lanio. seasonal sites, while Meladema CSM generally occupies being tertiary relicts of laurisilva forests on Madeira or the moderately seasonal localities. The estimated time of the Canary Islands, as already suggested by Ribera et al. (2003). colonization of Macaronesia by the ancestors of M. lanio There are several possible scenarios for the colonization and M. imbricata clearly excludes the possibility of them of these archipelagos. One is that both were independently

10 ª 2017 Royal Swedish Academy of Sciences V. Sykora et al. Evolutionary history of Meladema

Fig. 6 Map of the areas that currently have climatic conditions similar to the reconstructed ancestral values of each of the Meladema lineages, as measured with the 95% confidence interval of the first two axes of the climatic principal component analysis. —A. Genus Meladema; (B) continental Meladema coriacea; (C) M. coriacea CSM; (D) Meladema imbricata; (E) Meladema lanio. colonized from the continent; the Canary Islands (M. im- France or from Italy. Although we sampled only Corsica, bricata) likely from north-western Africa, and Madeira Sardinia and Montecristo, it is likely that also other islands (M. lanio) from south-western Iberia. During the Pleis- such as Pianosa, Capraia or Elba are inhabited by M. cori- tocene, there were several islands between Madeira and acea CSM (Rocchi et al. 2014). On the contrary, Malta continental , even during the Last Glacial Maximum (and likely , from where no specimen could be (Fernandez-Palacios et al. 2011). The two colonizations obtained) were likely colonized through other routes. could have originated from a common continental stock or After the speciation, continental M. coriacea started to they may have previously speciated, reaching their recov- spread to north-western Africa, Iberia, southern France, ered sister relationship through a process of coalescence Turkey and several Mediterranean islands (Malta, Balearic with the extinction of some of the continental lineages. A Islands), including one or more secondary recent coloniza- second possibility is that there was a single colonization tions of the Canary Islands. The phylogenetic assignment process from the continent, either to Madeira or the Can- of continental Italian, Sicilian and Balkan M. coriacea ary Islands, with subsequent dispersal between islands. remains uncertain, as no specimens from these regions However, as Meladema is absent on Salvages and no other could be sequenced. In the case of , M. coriacea was islands were probably present between Madeira and Canary reported for the south, in the Attica region and the Islands in the past (Fernandez-Palacios et al. 2011), the dis- Cyclades (Oertzen 1886). As the Turkish specimen from persion between both archipelagos seems rather unlikely. the Izmir province included in our analysis is nested deep The estimated age of the M. coriacea CSM lineage (Pleis- within the continental M. coriacea clade, it seems that also tocene) indicates that the isolation occurred much later Bulgaria, Greece and Cyclades are likely inhabited by the than the end of the Messinian Salinity Crisis, which lasted same lineage. The phylogenetic assignment of M. coriacea from 5.93 to 5.33 Ma and was the last time when islands in from mainland Italy is more problematic. According to the Tyrrhenian Sea were connected to the continental Europe ‘Checklist of Italian Fauna’ database (CKMAP2000; http:// (Krijgsman et al. 1999; Manzi et al. 2013; Perez-Asensio www.faunaitalia.it/ckmap/), M. coriacea is widely distributed et al. 2013). This suggests that ancestral Meladema were in the Italian peninsula and Sicily, with the northernmost able to colonize islands across the sea, probably either from records from Liguria. The fact that M. coriacea CSM

ª 2017 Royal Swedish Academy of Sciences 11 Evolutionary history of Meladema V. Sykora et al. inhabits islands in the Tuscany archipelago very close to monophyletic and sisters. The speciation among extant lin- the Italian mainland (Montecristo) cannot be taken as evi- eages likely took place in a very short time period in the dence that Italian Meladema also belong to the CSM lin- Early Pleistocene. This could have been the result of the eage. These islands are known to be inhabited by some change from a tropical climate in the Miocene to the typi- Corso-Sardinian endemics not found in mainland Italy (e.g. cal Mediterranean seasonality in the Pliocene. Climate Agabus aubei Perris, 1869, Agabus rufulus Fairmaire, 1859 or niche modelling showed that seasonality is the most impor- Rhithrodytes sexguttatus Aube, 1838 among the Dytiscidae, tant factor in the distribution of Meladema. All three island Rocchi et al. 2014). Moreover, Meladema has never been lineages inhabit less seasonal sites, likely relicts of climatic found in North Africa east of Tunisia (Touaylia et al. conditions ancestral to Meladema. On the contrary, conti- 2010), or in the but Turkey. This suggests that nental M. coriacea was able to colonize more seasonal sites M. coriacea reached western Turkey and Greece probably and expand its range across the Mediterranean. from southern Italy or the Balkans, which would thus be inhabited by the same lineage (continental M. coriacea). Acknowledgements Results of niche similarity test do not provide clear evi- We especially thank S. Bouzid, A. Castro, A. Cieslak, M.G. dences from a climatic niche perspective. The single speci- Parıs, A. Rudoy and R. Vila for providing specimens for men from Malta belongs to the continental M. coriacea study; A. Izquierdo and H. Lehmann (MNCN) and R. lineage, suggesting that the Sicilian populations may also Alonso (IBE) for laboratory work; D.T. Bilton for discus- belong to the same lineage. The aquatic Coleoptera fauna sions on our work with Meladema and two anonymous ref- of Malta is known to be strongly related to that of Sicily erees for comments. DG-V had a FPI PhD grant from the and mainland Italy (e.g. among the Hydraenidae, with a Spanish Government. DSF was supported by a ‘Juan de la high degree of local endemism, all species known from Cierva’ postdoctoral contract from the Spanish Ministry of Malta occur also in Sicily, Mifsud et al. 2004), which can Economy and Competitiveness and another postdoctoral also be taken as evidence of the absence of M. coriacea contract funded by Universidad de Castilla-La Mancha and CSM outside the northern Tyrrhenian islands. the European Social Fund (ESF). This work has been We can only speculate what is the reason of the wide partly funded by projects CGL2010-15755 and CGL2013- distribution of continental M. coriacea through the Mediter- 48950-C2-1-P to IR. ranean, while all other lineages are exclusively restricted to different archipelagos. Similarly to all other Meladema lin- References eages, continental M. coriacea could have initially been Alarie, Y. & Hughes, S. (2006). Re-descriptions of larvae of Hope- restricted to sites with climate resembling the wet tropical rius and Meladema and phylogenetic implications for the tribe climate of the Miocene, but developed capabilities neces- Colymbetini (Coleoptera Dytiscidae). Memorie della Societa Ento- sary for survival during dry periods after the climatic mologica Italiana, 85, 307–334. change from a wet tropical to Mediterranean climate Andujar, C., Serrano, J. & Gomez-Zurita, J. (2012). Winding up 3.2 Ma (Suc 1984). To test this hypothesis, it would be the molecular clock in the genus Carabus (Coleoptera: Cara- bidae): assessment of methodological decisions on rate and node necessary to obtain experimental data on the physiological age estimation. BMC Evolutionary Biology, 12, 40. tolerance of the different species. However, it could also be Baele, G., Lemey, P., Bedford, T., Rambaut, A., Suchard, M. A. & that continental Meladema expanded its range, at least ini- Alekseyenko, A. V. (2012). Improving the accuracy of demo- tially, just because it was in the continent, that is, without graphic and molecular clock model comparison while accommo- geographical barriers to impede its displacement, similarly dating phylogenetic uncertainty. Molecular Biology and Evolution, – to what is hypothesized to have happened in other groups 29, 2157 2167. € of aquatic beetles during the interglacials (Garcıa-Vazquez Balke, M., Hendrich, L. & Cuppen, J. G. M. (1990). Wasserkafer von den Islas Canarias (Coleoptera: Haliplidae, Dystiscidae, & Ribera 2016). Only at a later stage, it may have acquired Gyrinidae, Hydrochidae, Hydrophilidae, Hydraenidae, Dryopi- the dispersal or physiological capabilities necessary to cross dae). Entomofauna, 11, 349–373. the Straits of Gibraltar or to re-colonize the Canary Bielejec, F., Rambaut, A., Suchard, M. A. & Lemey, P. (2011). Islands. SPREAD: spatial phylogenetic reconstruction of evolutionary dynamics. Bioinformatics, 27, 2910–2912. Conclusions Brower, A. V. (1994). Rapid morphological radiation and conver- fl Our results strongly support the existence of a cryptic lin- gence among races of the butter y Heliconius erato inferred from patterns of mitochondrial DNA evolution. Proceedings of the eage of Meladema on the northern Tyrrhenian islands. National Academy of Sciences, 91, 6491–6495. Thus, M. coriacea as currently understood is not a mono- Darilmaz, M. C. & Kiyak, S. (2009). Checklist of Gyrinidae, Hali- phyletic species. In contrast, the Macaronesian species plidae, Noteridae and Dytiscidae of Turkey (Coleoptera: Ade- M. lanio and M. imbricata were found, respectively, phaga). Journal of Natural History, 43, 1585–1636.

12 ª 2017 Royal Swedish Academy of Sciences V. Sykora et al. Evolutionary history of Meladema

Drummond, A. J., Rambaut, A., Shapiro, B. & Pybus, O. G. Lanfear, R., Calcott, B., Ho, S. Y. W. & Guindon, S. (2012). Par- (2005). Bayesian coalescent inference of past population dynam- titionFinder: combined selection of partitioning schemes and ics from molecular sequences. Molecular Biology and Evolution, 22, substitution models for phylogenetic analyses. Molecular Biology 1185–1192. and Evolution, 29, 1695–1701. Drummond, A. J., Suchard, M. A., Xie, D. & Rambaut, A. (2012). Latella, L., Ruffo, S. & Stoch, F. (2007). The project CKmap Bayesian phylogenetics with BEAUti and the BEAST 1.7. Molec- (Checklist and distribution of the Italian fauna) methods and ular Biology and Evolution, 29, 1969–1973. informatical techniques. Memorie del Museo Civico di Storia Natu- Fernandez-Palacios, J. M., de Nascimento, L., Otto, R., Delgado, rale di Verona, 17,15–19. J. D., Garcıa-del-Rey, E., Arevalo, J. R. & Whittaker, R. J. Machado, A. (1987). Los Ditıscidos de las Islas Canarias (Coleoptera: (2011). A reconstruction of Palaeo-Macaronesia, with particular Dytiscidae). La Laguna, Spain: Instituto de Estudios Canarios, reference to the long-term biogeography of the Atlantic island CSIC. laurel forests. Journal of Biogeography, 38, 226–246. Manzi, V., Gennari, R., Hilgen, F., Krijgsman, W., Lugli, S., Garcıa-Vazquez, D. & Ribera, I. (2016). The origin of widespread Roveri, M. & Sierro, F. J. (2013). Age refinement of the Messi- species in a poor dispersing lineage (diving beetle genus Dero- nian salinity crisis onset in the Mediterranean. Terra Nova, 25, nectes). PeerJ, 4, e2514. 315–322. Griffin, D. L. (2002). Aridity and humidity: two aspects of the late McCormack, J. E., Zellmer, A. J. & Knowles, L. L. (2010). Does Miocene climate of North Africa and the Mediterranean. Palaeo- niche divergence accompany allopatric divergence in Aphelo- geography, Palaeoclimatology, Palaeoecology, 182,65–91. coma jays as predicted under ecological speciation?: insights Gschwendtner, L. (1936). Monographie der pal€aarktischen Dytisci- from tests with niche models. Evolution, 64, 1231–1244. den. (Begonnen on Alois Zimmermann, fortgesetzt von L. Mifsud, D., J€ach, M. A. & Schuh, R. (2004). The Hydraenidae Gschwendtner.). VII. Colymbetinae. (Colymbetini: Rhantus, (Insecta: Coleoptera) of the Maltese Archipelago (Central Nartus, Melanodytes, Colymbetes, Meladema). Koleopterologische Mediterranean). Annalen des Naturhistorischen Museums in Wien. Rundschau, 22,61–102. Serie B, Fur€ Botanik und Zoologie, 105, 429–440. Gueorguiev, V. B. (1981). Resultat de L’expedition zoologique du Miller, K. B. & Bergsten, J. (2014). The phylogeny and classifica- musee national de prague en Turquie Coleoptera: Haliplidae, tion of diving beetles (Coleoptera: Dytiscidae). In D. A. Yee Dytiscidae, Gyrinidae. Acta Entomologica Musei Nationalis Pragae, (Ed) Ecology, Systematics, and Natural History of Predaceous Diving 40, 399–424. Beetles (Coleoptera: Dytiscidae) (pp. 49–172). New York, USA: Hasegawa, M., Kishino, H. & Yano, T. (1985). Dating of the Springer. human-ape splitting by a molecular clock of mitochondrial Moriniere, J., Michat, M. C., J€ach, M. A., Bergsten, J., Hendrich, DNA. Journal of Molecular Evolution, 22, 160–174. L. & Balke, M. (2015). Anisomeriini diving beetles - an Hidalgo-Galiana, A., Sanchez-Fernandez, D., Bilton, D. T., Cies- Atlantic–Pacific Island disjunction on Tristan da Cunha and lak, A. & Ribera, I. (2014). Thermal niche evolution and geo- Robinson Crusoe Island, Juan Fernandez? Cladistics, 31, 166– graphic range expansion in a species complex of western 176. Mediterranean diving beetles. BMC Evolutionary Biology, 14, 187. Moriniere, J., Van Dam, M. H., Hawlitschek, O., Bergsten, J., Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G. & Jarvis, Michat, M. C., Hendrich, L., Ribera, I., Toussaint, E. F. A. & A. (2005). Very high resolution interpolated climate surfaces for Balke, M. (2016). Phylogenetic niche conservatism explains an global land areas. International Journal of Climatology, 25, 1965– inverse latitudinal diversity gradient in freshwater arthropods. 1978. Scientific Reports, 6, 26340. Jiggins, F. M. (2003). Male-killing Wolbachia and mitochondrial Nilsson, A. N. & Hajek, J. (2015). Catalogue of Palearctic Dytisci- DNA: selective sweeps, hybrid introgression and parasite popula- dae (Coleoptera). Online version 1.1.2015. Available via http:// tion dynamics. Genetics, 164,5–12. www2.emg.umu.se/projects/biginst/andersn/PAL_CAT_2015.pdf Jimenez-Moreno, G., Fauquette, S. & Suc, J. P. (2010). Miocene (last accessed 14 July 2016) to Pliocene vegetation reconstruction and climate estimates in Oertzen, E. V. (1886). Verzeichnis der Coleopteren Griechenlands the Iberian Peninsula from pollen data. Review of Palaeobotany und Creta. Berliner Entomologische Zeitschrift, 30, 189–293. and Palynology, 162, 403–415. Perez-Asensio, J. N., Aguirre, J., Jimenez-Moreno, G., Schmiedl, Katoh, K., Misawa, K., Kuma, K. & Miyata, T. (2002). MAFFT: a G. & Civis, J. (2013). Glacioeustatic control on the origin and novel method for rapid multiple sequence alignment based on cessation of the Messinian salinity crisis. Global and Planetary fast Fourier transform. Nucleic Acids Research, 30, 3059–3066. Change, 111,1–8. Kimura, M. (1980). A simple method for estimating evolutionary Phillips, S. J., Anderson, R. P. & Schapire, R. E. (2006). Maximum rates of base substitutions through comparative studies of nucleo- entropy modeling of species geographic distributions. Ecological tide sequences. Journal of Molecular Evolution, 16, 111–120. Modelling, 190, 231–259. Kohler,€ C. M., Heslop, D., Krijgsman, W. & Dekkers, M. J. Rambaut, A., Suchard, M. A., Xie, D. & Drummond, A. J. (2014). (2010). Late Miocene paleoenvironmental changes in North Tracer v1.6, Available vai http://beast.bio.ed.ac.uk/Tracer (last Africa and the Mediterranean recorded by geochemical proxies accessed 14 Jul. 2016). (Monte Gibliscemi section, Sicily). Palaeogeography, Palaeoclima- Ribera, I., Bilton, D. T. & Vogler, A. P. (2003). Mitochondrial tology, Palaeoecology, 285,66–73. DNA phylogeography and population history of Meladema div- Krijgsman, W., Hilgen, F. J., Raffi, I., Sierro, F. J. & Wilson, D. ing beetles on the Atlantic Islands and in the Mediterranean S. (1999). Chronology, causes and progression of the Messinian basin (Coleoptera, Dytiscidae). Molecular Ecology, 12, salinity crisis. Nature, 400, 652–655. 153–167.

ª 2017 Royal Swedish Academy of Sciences 13 Evolutionary history of Meladema V. Sykora et al.

Ribera, I., Vogler, A. P. & Balke, M. (2008). Phylogeny and diver- Supporting Information fi si cation of diving beetles (Coleoptera: Dytiscidae). Cladistics, 24, Additional Supporting Information may be found in the – 563 590. online version of this article: Rocchi, S., Terzani, F., Cianferoni, F. & Forbicioni, L. (2014). Table S1. List of specimens included in the different Materiali per una fauna dell’Arcipelago Toscano. XXXI. Con- analyses with specimen code, voucher, locality, geographi- tributo alla conoscenza della coleotterofauna acquatica dell’Arci- pelago Toscano (Coleoptera). Annali del Museo Civico di Storia cal coordinates and Genbank accession numbers. In bold, Naturale “G. Doria”, 106,1–74. sequences newly obtained for this study. Specimens with * Ruddiman, W. F., Sarnthein, M., Baldauf, G., Curry, W. B., were tested for the presence of Wolbachia infection Dupont,L.M.,Janecek,T.,Pokras,E.M.,Raymo,M.E., Table S2. DNA extraction and sequencing. (A) Primers Stabell,B.,Stein,R.&Tiedemann, R. (1989). Late Miocene used for amplification and sequencing. In brackets, length to Pleistocene evolution of climate in Africa and the low-lati- of the amplified fragment (ingroup only). (B) Standard PCR tude Atlantic: overview of Leg 108 results. Proceedings of the conditions for the amplification of the studied fragments Ocean Drilling Program. ScientificResults, 108, Table S3. (A) Bioclimatic variables of each locality 463–484. Sanchez-Fernandez, D., Abellan, P., Picazo, F., Millan, A., Ribera, obtained in WORLDCLIM (http://www.worldclim.org) used in I. & Lobo, J. M. (2013). Do protected areas represent species’ the PCA (Principal Component Analysis). (B) Geographical optimal climatic conditions? A test using Iberian water beetles coordinates and values of the first three axes of the climatic Diversity and Distributions, 19, 1407–1417. PCA of the four lineages of Meladema Schoener, T. W. (1968). The Anolis lizards of Bimini: resource Table S4. Pairwise values of observed niche overlap – partitioning in a complex fauna. Ecology, 49, 704 726. (Schoener’s D-value) between the four lineages of Meladema Sontowski, R., Bernhard, D., Bleidorn, C., Schlegel, M. & Gerth, Table S5. Specimens used in the reconstruction of the M. (2015). Wolbachia distribution in selected beetle taxa charac- fi terized by PCR screens and MLST data. Ecology and Evolution, ancestral niche from the scores of the rst two principal 5, 4345–4353. component analysis (PCA) axes. Overlap “YES” indicates Stamatakis, A., Hoover, P. & Rougemont, J. (2008). A rapid boot- that the two PCA scores of the climatic conditions of the strap algorithm for the RAxML Web servers. Systematic Biology, locality in which the specimen was found fall within the – 57, 758 771. 95% Highest Posterior Density (HPD) intervals of the Suc, J. P. (1984). Origin and evolution of the Mediterranean vege- reconstructed conditions of the most recent common ances- tation and climate in Europe. Nature, 307, 429–432. tor of the respective lineage Tamura, K. & Nei, M. (1993). Estimation of the number of nucleotide substitutions in the control region of mitochondrial Figure S1. Ultrametric phylogenetic tree of Meladema DNA in humans and chimpanzees. Molecular Biology and Evolu- obtained with RAxML using only the nuclear sequence tion, 10, 512–526. (H3 + Wg) (see Methods for details). Numbers in nodes, Tavare, S. (1986). Some probabilistic and statistical problems in bootstrap support values the analysis of DNA sequences. Lectures on Mathematics in the Figure S2. Bayesian Skyline Plot of continental Mela- – Life Sciences, 17,57 86. dema coriacea showing reconstructed changes in its effective Touaylia, S., Garrido, J., Bejaoui, M. & Boumaiza, M. (2010). A population size during the last 60 Ka. Thin blue lines, 95% contribution to the study of the aquatic fi (Coleoptera: Dytiscidae, Gyrinidae, Haliplidae, Noteridae, con dence intervals; horizontal axis, time (1000 years ago); Paelobiidae) from northern Tunisia. The Coleopterists Bulletin, vertical axis, effective population size (NeT) 64,53–72. Figure S3. Environmental niche models (ENMs) using Warren, D. L., Glor, R. E. & Turelli, M. (2008). Climate niche MAXENT for: (A) continental Meladema coriacea; (B) M. cori- identity versus conservatism: quantitative approaches to niche acea CSM; (C) Meladema imbricata and (D) Meladema lanio. evolution. Evolution, 62, 2868–2883. Higher MAXENT probabilities (red colours) represent areas Warren, D. L., Glor, R. E. & Turelli, M. (2010). ENMTools: a more suitable for the species according to the MAXENT mod- toolbox for comparative studies of environmental niche models. Ecography, 33, 607–611. els, lower values (blue) represent less suitable areas. These Werren, J. H., Baldo, L. & Clark, M. E. (2008). Wolbachia: master continuous measures of habitat suitability were used in manipulators of invertebrate biology. Nature Reviews Microbiology, ENM-based tests of niche similarity (see Fig. 5 and 6, 741–751. Table S4) Zaitsev, F. A. (1972). Fauna of the USSR: Amphizoidae, Hygrobiidae, Data S1. kml file obtained with BEAST and SPREAD to visu- Haliplidae, Dytiscidae, Gyrinidae, Coleoptera, Vol. IV. Jerusalem, alize in Google Earth the phylogeographic reconstruction Israel: Israel Program for Scientific Translations. through time of the genus Meladema (see main text for Zug, R. & Hammerstein, P. (2012). Still a host of hosts for Wol- bachia: analysis of recent data suggests that 40% of terrestrial details). species are infected. PLoS One, 7, e38544.

14 ª 2017 Royal Swedish Academy of Sciences