This constitutes the first phylogenetic study using a complete sampling of Stenodactylus taxa and including 207 specimens from across the entire distribution range of North Africa and Arabia (Figure 1). This has enabled a robust phylogenetic reconstruction (see Figure 2 and Additional file 2: Figure S1), the uncovering of intraspecific diversity and, in some cases, the unveiling of interesting distribution patterns (see below). The phylogenetic results show a high level of support in most of the nodes and a striking agreement with the phylogenetic analyses of Stenodactylus by Arnold (1980) , based on morphological data, increasing our confidence that the recovered topology represents the true evolutionary history of the genus.
Monophyly of Stenodactylus
Despite the general concordance between morphological and phylogenetic conclusions, one important discrepancy is observed: while morphology supports the inclusion of P. khobarensis in the genus Stenodactylus, the results of our molecular analyses indicate that Pseudoceramodactylus and Stenodactylus are not even sister genera (Figure 2). Kluge (1967)  transferred P. khobarensis to the genus Stenodactylus based on a “large number of external (meristic and mensural) and internal morphological similarities”, including relevant characters like the phalangeal reduction to a formula of 220.127.116.11.3 on both fore and hind limbs and a very high scleral ossicle number (20–28). Arnold (1980) , despite pointing out some unique scale characters of P. khobarensis, retained it in Stenodactylus and considered the scalation characters as “convincing pointers to holophyly”. However, according to a recent molecular analysis of the group by Fujita and Papenfuss (2011)  based on independent samples and sequences of different mitochondrial and nuclear regions, two representatives of Tropiocolotes branched between P. khobarensis and the six species of Stenodactylus included in the analysis (see Figure 1 of ). In order to deal with the non-monophyly of Stenodactylus, the genus Pseudoceramodactylus was resurrected. This pattern is repeated and further investigated in our study, with a complete taxon sampling of Stenodactylus and the inclusion of a greater number of representatives of Tropiocolotes, resulting in the splitting of the latter genus into two groups, a surprising but not strongly supported, albeit consistent, result.
We performed a series of constraint analyses in which Stenodactylus and Pseudoceramodactylus were forced to form a monophyletic group. Results clearly show that our dataset cannot reject the alternative hypothesis of a monophyletic Stenodactylus + Pseudoceramodactylus group (Table 2). In order to further investigate this, the dataset of Fujita and Papenfuss (2011)  was subjected to the same ML topological tests, but also could not reject the alternative hypothesis of monophyly of Stenodactylus + Pseudoceramodactylus (AU P = 0.074; SH P = 0.092). In view of the confusing molecular evidence and taking into account the morphological data, we think that the resurrection of Pseudoceramodactylus was precipitated, but in the meanwhile, this change accommodates for both the paraphyly reported by Fujita and Papenfuss and confirmed here, and the hypothesis of monophyly of Stenodactylus + Pseudoceramodactylus. We recommend not performing any further changes at the generic level before an in-depth revision clarifies the evolutionary relationships between the genera Stenodactylus, Pseudoceramodactylus and Tropiocolotes.
Systematics and evolution
The well-supported clade A is formed by the morphologically similar S. pulcher, S. arabicus and the lineage S. cf. arabicus and, according to the inferred dates, the split between the former and the two latter species dates back to approximately 17 Ma ago (95% HPD: 11.3-23.6) (Figure 2). On the one hand, variability within S. pulcher is very low, probably as a result of the two specimens analyzed being from very close localities. On the other hand, the S. cf. arabicus lineage from the Sharqiya Sands (formerly Wahiba Sands), as already highlighted by Fujita and Papenfuss (2011) , is genetically very distinct from all other populations of S. arabicus included in our study in both mitochondrial and nuclear markers (Additional file 3: Table S2c and Figure 2), where almost all alleles are lineage-specific (see Results and Figure 3). This supports the idea that the Sharqiya Sands are isolated and surrounded by some areas of unsuitable habitat for sand dune specialists like this species [88–90]. Further morphological and molecular studies including more specimens from putative contact zones and faster nuclear markers are expected to give S. cf. arabicus formal recognition.
Clade B is well-supported (ML 100%, BI 1.0) and was also recovered by the morphological analysis of Arnold (1980) . Stenodactylus doriae and S. leptocosymbotes are reciprocally monophyletic and form the relatively well-supported sub-clade B1 (Figure 2). Our molecular results agree with the results of the morphological analysis by Arnold (1980) , who also recovered the two species as sister taxa based on three synapomorphies. The two species diverged approximately 7.0 Ma ago (95% HPD: 4.2-10.1) (Figure 2) and, like the two North African sister species S. sthenodactylus and S. mauritanicus, they are ecologically distinct. Stenodactylus leptocosymbotes is an arid-adapted species that lives on relatively hard, although usually sandy, substrates being replaced by its sister species, S. doriae, on soft, wind-blown sand [34, 91]. Thanks to its morphological and physiological adaptations, the latter is able to live in hyper-arid sand dune environments like for example the Eastern Rub al Khali , one of the largest and driest sand deserts in the world . Given the clear morphological and ecological differences between these two species and the apparent absence of morphologically intermediate individuals [31, 34], it seems reasonable to deduce that allele sharing in RAG-2 (see Results), which is limited to the ancestral allele, is the result of incomplete lineage sorting rather than ongoing gene flow between the two species. Variability within S. leptocosymbotes is rather low (Additional file 3: Table S2b) and the number of samples included permit to observe only moderate geographical structuring (Figures 1 and 2, Additional file 2: Figure S1). In contrast, S. doriae, shows a higher level of genetic differentiation, with the Sharqiya Sands lineage being quite divergent (Additional file 3: Table S2c and Figure 2), as already mentioned by Fujita and Papenfuss (2011) .
Sister to sub-clade B1 is a group composed by S. slevini, S. grandiceps and S. affinis, for which support is relatively low (ML 62, BI = 0.95). The topology within this sub-clade differs from the morphological hypothesis of Arnold (1980) , which supported the following relationship: (S. grandiceps (S. affinis (S. slevini (S. leptocosymbotes, S. doriae)))). Stenodactylus slevini is the only member of the group with two divergent lineages, one limited to Jordan and the other with representatives from East Arabia. Although the divergence based on mitochondrial data is clear (Additional file 2: Figure S1), there is no supporting nuclear data available (Figure 3), and no obvious morphological differences (pers. obs.). With the only exception of the soft wind-blown sand specialist S. doriae, all remaining representatives of clade B plus two other species, the African S. sthenodactylus and the Arabian S. yemenensis, appear to occupy rather similar spatial niches. These six species are adapted to living on relatively hard ground, coarse sandy planes, large wadis and sandy substrates and, based on their head dimensions, probably feed on similar-sized prey [31, 32, 34]. As a consequence of that, these species rarely coexist and have largely allopatric distribution ranges, while in places where they coincide they are not syntopic [31, 33, 34]. The analysis of the nuclear allele networks (Figure 3) indicate that the morphologically and ecologically similar and phylogenetically closely related S. leptocosymbotes, S. slevini, S. grandiceps and S. affinis do not share a single allele in the c-mos and RAG-2 genes analyzed, even though the results of the calibration analyses suggest that S. grandiceps and S. affinis diverged later (6.7 Ma ago; 95% HPD: 4.1-9.3) than other lineages for which extensive allele sharing in the RAG-2 has been identified (S. doriae and S. leptocosymbotes; see above and Results). These differences of the level of lineage sorting in some of the morphologically well-recognized species may also be the result of differences in effective population sizes, which affect the lineage coalescence time .
In sub-clade C1, S. petrii is grouped together with the North African endemic S. stenurus that branches inside it (Figure 2). As a result, S. petrii is paraphyletic and constitutes the only exception among the otherwise monophyletic Stenodactylus species. The results of the topological tests (Table 2) indicate that our dataset most probably rejects the monophyly of this species (AU:0.036, SH:0.210, BF:2.578). Stenodactylus stenurus was described by Werner (1899)  and synonymized ten years later by the same author . It remained in synonymy until Kratochvil et al. (2001)  recognized it as a valid species, based on a multivariate analysis of several metric and scalation characters. It is noteworthy that the representative of S. stenurus included in our analysis is one of the specimens used by Kratochvil et al. (2001)  in their study.
The highly divergent lineage that includes specimens from Egypt and Israel (see Results) is estimated to have split from specimens further west in Algeria, Morocco, Western Sahara and Mauritania approximately 6.1 Ma ago (95% HPD: 3.9-8.6) (Figure 2). In fact, the northern Sinai populations of S. petrii have been reported to be morphologically distinct and, as a result of that, were considered a different species (S. elimensis) by Barbour (1914) , now under the synonymy of S. petrii[31, 98]. Yet, specimens from this area included in our analyses do not present considerable genetic differences with the rest of the Egyptian and Israeli specimens (Figures 1 and 2, Additional file 2: Figure S1). It should be pointed out that the type locality of S. petrii is Egypt and, thus, this lineage represents the 'true' S. petrii. The pattern in the nuclear genes, with numerous unique alleles for this lineage (Figure 3), contrasts with the situation in S. stenurus that lacks unique alleles. This suggests that further analyses and a thorough taxonomic revision including more samples of S. petrii, especially from not sampled areas of Algeria and Libya, and mainly S. stenurus will be necessary in order to evaluate the status of the populations assigned to the two species. With this evidence it will be possible to differentiate between a single species with high genetic variability (petrii), two species (petrii in the East and stenurus in the West) or three species, if stenurus proves to be distinct from the more western forms.
The two North African species of sub-clade C3, S. sthenodactylus and S. mauritanicus, are shown to be reciprocally monophyletic and highly divergent (Additional file 3: Table S2a), while their separation dates back to approximately 10.0 Ma (95% HPD: 6.6-13.7) (Figure 2). These results help to clarify the status of these two taxonomically controversial taxa that were treated as two different subspecies by Loveridge (1947)  and Sindaco and Jeremcenko (2008) , as the same monotypic species by Arnold (1980)  and that were finally considered as full species by Baha el Din (2006) , who found them in sympatry at particular localities in northern Egypt. As observed by Baha el Din (2006) , although these two sister species can be morphologically similar and share similar habits, they are ecologically different. Stenodactylus mauritanicus is restricted to fairly mesic coastal semi-desert under the influence of the Mediterranean (see Figure 1), where it inhabits flat rock-strewn sand and gravel plains with fairly good vegetation cover. On the contrary, S. sthenodactylus inhabits areas of the Sahara that are far more arid and inhospitable than the ones of its sister species (see Figure 1), being the only vertebrate to be readily found in some parts of the Western Desert of Egypt . It prefers gravelly and coarse sandy plains and large wadis and, although the species is typical of hard coarse substrates, it sometimes penetrates some dune areas .
The distributions of these two species, as introduced by the present study, give insights into the controversial taxonomic status and frequent misidentification of the two forms . Our analysis concludes that S. sthenodactylus extends west from the Middle East and Egypt, previously thought to be its eastern limit, across the Sahara and into Mauritania (Figure 1). Stenodactylus mauritanicus is confirmed to be present in Egypt  and has a wide, almost continuous distribution roughly along the northern margin of the Sahara desert. The two species are found in sympatry or in close proximity in Egypt and coastal Mauritania, yet retain distinct mtDNA lineages and exhibit only limited allele sharing in the nuclear markers, most of which is due to sharing of ancestral alleles and hence is likely to represent incomplete lineage sorting (see Figure 3).
Stenodactylus sthenodactylus presents high variability, both at genetic (see Results) and morphological levels . Its three deep lineages are estimated to have diverged approximately 4.8 Ma ago (95% HPD: 2.8-6.9) (Figure 2). According to Baha el Din (2006) , some morphological characters appear to correlate with environmental factors, with populations from hyper-arid places showing a very slender body, less contrasting pattern and tubular nostrils, while populations from more mesic areas being usually more robust, with thick limbs, big heads and marked pattern [31, 36, 98]. The populations from coastal regions in southeast Egypt are especially distinct and, according to Baha el Din (2006) , they resemble specimens of S. s. zavattarii from Kenya, which Loveridge (1957)  synonymized with S. sthenodactylus. Two specimens of this form were included in our phylogenetic analyses (see Figure 2 and Additional file 2: Figure S1), and indeed they belong to a clade with samples from south and southeast Egypt. These results suggest that some of the morphological variability between populations of S. sthenodactylus may also be supported by molecular data. A nomenclatural revision of North African Stenodactylus (work in progress) is essential for stability before any changes are performed, while further work focused on the contact zones between the three lineages and combining detailed morphological analyses with additional nuclear data is needed in order to determine if they deserve formal recognition.
On the other hand, the high genetic variability within S. mauritanicus (Figure 2 and Additional file 3: Table S2b) does not seem to correlate with differences in morphology. This species is fairly uniform morphologically, with populations from the West being a bit larger than Egyptian ones but generally maintaining the same proportions, pattern and scalation across most of its distribution range . Nevertheless, the intra-specific divergence is estimated to date back to 6.6 Ma ago (95% HPD: 4.0-9.5) and the six mitochondrial lineages present a clear geographical pattern (Figure 1 and Additional file 2: Figure S1). The relationship between these lineages, however, is not clear and neither is any structure observed in the nuclear alleles (Figure 3), both facts being mirrored in the low-supported nodes of the concatenated phylogeny (Figure 2).
Origin, biogeography and diversification of Stenodactylus
Reconstruction of ancestral areas with both parsimony and ML methods (Figure 4) suggests that the genus Stenodactylus originated in Arabia approximately 30 Ma ago (95% HPD: 20.7-39.2) (Figure 2), a time of high geological instability as a result of the onset of major seismic and volcanic events in the general area of Ethiopia, northeast Sudan and southwest Yemen . These major volcanic and tectonic events, centered over the Afar region, marked the onset of the formation of some of the most relevant and complex physiographical features in the contact zone between Africa and Arabia, like the Gulf of Aden, the Red Sea and the elevation of the Afro-Arabian rift-flanks to heights above 3600 m [1, 101, 102].
The tempo and mode of the deep splits in Stenodactylus bear a striking resemblance to the basal splits that occurred in the African-Eurasian snake genus Echis[13, 103], which suggests a common biogeographical pattern for both groups. The distribution of the members of Arabian clade B (S. doriae, S. leptocosymbotes, S. slevini, S. grandiceps, S. affinis) and the mainly African clade C (S. petrii, S. stenurus, S. yemenensis, S. mauritanicus, S. sthenodactylus) (Figures 1 and 2) extend primarily on the opposite sides of the Red Sea, mimicking the situation of the sister taxa E. coloratus (mainly Arabian) and E. pyramidum (mainly African). The split between these two Stenodactylus groups dates back to 21.8 Ma ago (95% HPD: 15.4-29.1) (Figure 2), which roughly coincides with the split between E. coloratus and E. pyramidum calculated at approximately 19.4 Ma ago. The dates of these phylogenetic events follow a well-studied phase of volcanism and strong rifting initiated at approximately 24 Ma ago, that appeared in an almost synchronous way throughout the entire Red Sea . Therefore, it is possible that the formation of the Red Sea acted as a vicariant event separating the aforementioned clades of Stenodactylus, as also suggested by Pook et al. (2009)  for the genus Echis. The agamid lizards of the genus Uromastyx is yet another group that could have been affected by such an event, although in this case the split between the Arabian and African clades seems to have happened later, at 11–15 Ma ago. Amer and Kumazawa (2005)  attributed this split to a dispersal event from Arabia into North Africa, coinciding with climatic changes towards aridity in this latter area, rather than to vicariance. However, since earlier dates had also been calculated for the split between African and Arabian Uromastyx that coincide with the inferred dates for Stenodactylus and Echis (18 Ma ago; ), a reassessment of the calibration dates of Uromastyx using relaxed clock methods like the ones applied by Pook et al. (2009)  and in the present study seems necessary (work in progress).
The split between the Arabian S. yemenensis and the ancestor of the African S. mauritanicus and S. sthenodactylus on either sides of the Red Sea also parallels the splits between Arabian and African sister clades of the E. pyramidum complex  and Uromastyx ocellata and U. ornata. Although the divergence time estimate for the Stenodactylus members (15.4 Ma ago (95% HPD: 10.5-20.8), Figure 2) predates the ones of the other two groups by almost 7 Ma, the split between African and Arabian lineages might be explained by the complex geology of the Red Sea. Several recurrent episodes during the Miocene caused the desiccation and refilling of this tectonically active rifting area [1, 105] and provoked the severing of the land bridges that had existed after the initial formation of the Red Sea in the early Miocene. So, the separation between S. yemenensis and the ancestor of S. mauritanicus and S. sthenodactylus was probably also the result of vicariance, similarly to Echis and Uromastyx. After this event, S. yemenensis would have remained isolated at the coastal side of the southern Arabian highlands (Figures 1 and 2).
In Arabia, an example of a similar biogeographical pattern caused by a different biogeographical process is the case of the ecologically similar sister species of clade A, S. pulcher and S. arabicus (including S. cf. arabicus), which, according to the results (Figure 2) and the geological data available, are hypothesized to result from vicariance caused by the uplift of the Yemen Mountains approximately 18 Ma ago [1, 101, 102]. The splits within clade B, however, seem more difficult to interpret, as little information is available on the geological and climatic history of the interior of Arabia. A general pattern could be proposed with a first North–South split between the ancestors of S. doriae, S. leptocosymbotes and S. slevini, S. grandiceps, S. affinis, respectively, followed by the posterior range expansion of some of these species. Interestingly, in Arabia, even though evidence exists for an increase in aridification , it has been hypothesized that at the same time an important river system, as evidenced by the fluvial sediments, could characterize the interior of the peninsula [93, 107]. Such dynamic scenery could be responsible for the rapid diversification within clade B, having caused fragmentation of the distribution range of the ancestor(s) and the different lineages to split allopatrically.
The onset of diversification in clade B coincides in time with the split between the African S. mauritanicus and S. sthenodactylus in sub-clade C3 (Figure 2). These speciation events match very closely the estimates of the formation, in the late Miocene, of a major east-Antarctic ice sheet with its associated polar cooling, which triggered the aridification of mid-latitude continental regions and a shift in North Africa from forest to dry open woodlands and savannahs [4, 20, 108]. The two North African forms, S. mauritanicus and S. sthenodactylus, seem to have diverged in ecological niche, with one form adapted to mesic environments and the other occupying much dryer areas, respectively. It has been proposed that the gradual increase in aridity that took place in northern Africa during the late Miocene accelerated the diversification process in reptile faunas . The estimated divergence times of the North African Stenodactylus seem to corroborate a common emerging pattern among European biota, according to which the speciation events in many reptile and amphibian groups do not coincide with the accentuated environmental instability during the Pleistocene, but rather date back into the Miocene and proceeding through the Quaternary, when many species and populations originated [109, 110].
It has been suggested that 18 Ma ago, Africa connected with Eurasia through the closure of the Eastern Mediterranean seaway (the Gomphotherium land bridge) . This land bridge later became disconnected temporarily but it has been continuously present since approximately 15 Ma ago. It is interesting to notice that, despite the existence of a continuous passage between Arabia and Eurasia, our phylogeny suggests that colonization of Eurasia by members of the genus Stenodactylus occurred much later and was very restricted geographically. In fact, only two Stenodactylus species extend their ranges into Eurasia (S. affinis and S. doriae). From these two, only samples of S. affinis from Eurasia (Iran) were available, while for the other species a specimen from neighboring Kuwait was included. In both species, however, the low intraspecific genetic variability suggests that the colonization of Eurasia was a very recent event (Figure 2 and Additional file 3: Table S2b). One possible explanation of this biogeographical pattern may be the existence of ecologically and morphologically very similar forms in Iran like Crossobamon (formerly a member of Stenodactylus) and Agamura, which may compete with Stenodactylus and therefore may have not allowed it to expand further outside the narrow coastal strip in southwestern Iran (Arnold, 1980). This situation is completely different than the one in North Africa, where no ecological analogs to Stenodactylus seem to exist and therefore several of its species are found across an area of more than 10 million Km2[31, 33, 98, 111, 112].