Phylogenetic systematics of day geckos, genus Phelsuma, based on molecular and morphological data (Squamata: Gekkonidae) Author Rocha, Sara Author Rösler, Herbert Author Gehring, Philip-Sebastian Author Glaw, Frank Author Posada, David Author Harris, James Author Vences, Miguel text Zootaxa 2010 2429 1 28 journal article 10.5281/zenodo.194693 4c0d1aa7-64e9-4c3d-a673-8155067872b7 1175-5326 194693 Phelsuma barbouri group (contains P. barbouri and P. pronki ) These two species were recovered as sister taxa with strong support by both the mitochondrial and nuclear sequences, and the combined analysis (clade Q in Rocha et al. 2009 ; not shown here). They were previously thought to be closely related ( Glaw et al. 1999 ) but later P. pronki was reclassified within the P. klemmeri group based on being a non-egg-gluer ( Van Heygen 2004 ), contrary to P. barbouri . The molecular data suggests instead that P. klemmeri does not belong to any of the previously identified groups; although it may be more related to the P. l i n e a t a and P. l a t i c a u d a species groups (see below). Phelsuma barbouri and P. pronki individuals exhibit a genetic distance of only 5.1% (uncorrected p-distance, cytochrome b ), which is low compared to other interspecific differentiation values within this genus and points to their relatively recent differentiation. Phelsuma barbouri and P. pronki share a rather depressed body and head shape, and a rather similar and distinct colour pattern without clear predominance of green, but differ in the number of midbody scales (higher in P. barbouri ), number of ventral scales and egg-gluing behaviour ( Tables 1–2 ). FIGURE 1. Schematic cladogram summarizing phylogenetic relationships of species groups proposed herein, species contained in each group, and characters useful for their phenotypic characterization. The tree is adapted from Rocha et al. (2009) and represents the BI 50% majority-rule consensus tree of combined mitochondrial and nuclear data. Species groups containing more that one species are colour coded. FIGURE 2. Molecular phylogeny of the Phelsuma dubia group as calculated from the multi-gene dataset of Rocha et al. (2009). Shown are the respective clades from the Bayesian 50% majority-rule consensus trees based on (a) the combined dataset of mitochondrial and nuclear genes, (b) the mitochondrial genes only, and (c) the nuclear genes only. Asterisks mark nodes supported by Bayesian posterior probabilities of 99% or higher. ML bootstraps supports above 50 are given below the corresponding branch. Relevant PP values above 95 are given above respective branches (italics). Clade numbers M and O are as in Rocha et al. (2009) and discussed in the text. P. barbouri is the most ground-dwelling species within the genus, found on and around the exposed rocky outcrops and boulders of unforested habitat areas of Madagascar's central mountain chain. In contrast, P. pronki is an arboreal species in the mid-altitude rainforests bordering Madagascar's central high plateau. The divergence between these two species is likely a recent one, and may have been driven by ecological specialization along the sharp ecotone between these habitat types . The distribution of P. pronki is very poorly known (Glaw & Vences 2007), and the species is probably rare and/or localized. Additional surveys are needed to understand if these two taxa are geographically isolated or have a contact or hybrid zone.

Phelsuma dubia group
(contains P. berghofi , P.	dubia ,	P.	flavigularis ,	P.	hielscheri ,	P.	malamakibo ,	P.	modesta ,	P.	nigristriata	and	P.
ravenala )

Evidence for a close relationship between P. dubia , P. flavigularis , P. malamakibo , P. b e rg h o f i and P. hielscheri and of these with P. m o d e s t a and P. nigristriata is apparent from the molecular phylogenies ( Fig. 2 ; clade M), corroborating morphological data. This group is strongly supported by the combined phylogeny and mtDNA data and also recovered (though not strongly supported – 89 PP, including also P. serraticauda ) by the nuclear phylogeny. Within this clade, P. hielscheri , P. berghofi and P. malamakibo (clade O ) plus P. flavigularis , form a generally strongly supported clade ( Fig. 2 ), but the position of this latter species respective to others differs between the mitochondrial and combined phylogenies (where P. flavigularis is basal to remaining three species) and the nuclear data (which instead recovers P. flavigularis as sister to P. b e rg h o f i and P. malamakibo , with P. hiescheri basal to these three). The placement of this group of species ( P. malamakibo , P. berghofi , P. hielscheri and P. flavigularis ) with the P. dubia / P. ravenala clade is strongly supported by mtDNA data and, although weakly, also by the nuclear data. The placement of P. hielscheri (and P. malamakibo ) in the P. dubia group is congruent with the morphological data ( Table 1–2 ; see also Van Heygen 2004 ) but contrasts with recent results from Raxworthy et al. (2007) in which their individual of P. hielscheri was more closely related to P. lineata and P. laticauda . Specimens of P. h i e l s c h e r i have previously been considered as P. dubia (see Glaw et al. 1999 ), and the differences between the two species only recently noted (Rösler et al. 2001). In our analysis, even when it is not placed in a monophyletic group with P. m a l a m a k i b o and P. berghofi , P. hielscheri is always recovered as included in the P. dubia group, and never close to the P. lineata or P. laticauda species groups as in the analysis of Raxworthy et al. (2007) . A reanalysis of the data of these authors as available from Genbank revealed that they had only a very short partial sequence of cytochrome b available for this species (193 bp; EF434870 ); which, when aligned with our data, clustered deep inside the Phelsuma lineata group. Since our samples come from a well-identified specimen of P. hielscheri and the various markers we sequenced are congruent in placing it into the P. dubia group, we conclude that the deviant placement of the species in the phylogeny of Raxworthy et al. (2007) is likely erroneous. The mitochondrial genes reveal additional deep differentiation within P. dubia (although not apparent in the nuclear phylogeny, possibly due to incomplete lineage sorting). P. dubia has a patchy distribution along the northwestern coast of Madagascar , possibly with disjunct populations ( Glaw & Vences 1994 ). We find a deep divergence between the individual from the northern isolate (Ambanja) and the southern one (Antsalova) (12.2% uncorrected p-distance in the cytochrome b marker). The included specimen from Moheli, Comoros , is very closely related to specimens from northwestern Madagascar (Ambanja), as are specimens from Zanzibar , Tanzania : see Rocha et al. 2007 ). Recently, Raxworthy et al. (2007) described P. ravenala from the eastern coast of Madagascar as a new species from the P. dubia group. Although genetic differentiation relative to P. dubia was minimal (~0.3%, 12S rRNA uncorrected p-distance; 0.4–0.7% cytochrome b distance according to our data), allopatric distribution, morphological differentiation and better fit of distribution models when considering it a different species were invoked to argue for its specific status. Our P. dubia individuals from the Antsalova region in western Madagascar exhibit a much higher degree of differentiation relative to remaining P. dubia (14% uncorrected p-distance at cytochrome b ) and might also be allopatrically distributed. This indicates that the taxonomy in the P. dubia / ravenala complex of species merits further investigation; P. dubia is at the moment paraphyletic with respect to P. r a v e n a l a, and possibly contains yet further undescribed species, with the western Antsalova population being one candidate. Also, the validity of P. r a v e n a l a is in need of confirmation, especially regarding its morphological differentiation (in number of midbody scales, colour and number of preanofemoral pores; see Raxworthy et al. 2007 and Tables 1–2 ). The second major clade in the P. dubia group is composed of P. m o d e s t a and the Comoran P. nigristriata . The sister group relationship of these two species is supported by nuclear as well as mitochondrial data. The placement of this clade with the P. dubia group is strongly supported by the combined and mitochondrial data sets but unresolved with nuclear data alone. Our results are partly in agreement with the opinion of Nussbaum et al. (2000) who considered the subspecies of Phelsuma modesta as invalid forms representing colour variations. The level of differentiation between two of the subspecies ( P. modesta modesta from Tolagnaro and P. m. isakae from near the type locality Isaka) is very low (0.1% uncorrected p-distance in cytochrome b ) and the blue head colouration of males, the only character to distinguish P. m. i s a k a e from P. m. modesta is also seen in specimens from the Tolagnaro region, usually assigned to P. m. modesta . On the other hand, P. modesta leiogaster presents a higher level of differentiation (6%), but the topology recovered differs between mitochondrial and nuclear genes. Species in the Phelsuma dubia group share egg-gluing behaviour (except for P. nigristriata ), absence of nostril-rostral contact, a relatively low number of preanofemoral pores in males (30 or less), and keeled dorsal and lateral scales (except for P. ravenala ). Most species have a weak expression of green colour, and in some species of the group the ventrals and/or subcaudals are keeled ( Tables 1–2 ).