Biogeography of the White-Bellied Carpet Viper Echis leucogaster Roman, 1972 in Morocco, a study combining mitochondrial DNA data and ecological niche modeling

In northwest Africa some species from Sahelian origin appear with relict populations and apparently isolated by the extreme aridity of the Sahara desert. However very tolerant to aridity species could maintain continuous populations as might be the case for Echis leucogaster as indicated by results from genetic analysis and bioclimatic models. Resum: Al nord-oest d'àfrica apareixen un grup d'espècies d'origen sahelià en poblacions relictes i aparentment aïllades pel desert del Sàhara. No obstant això espècies molt tolerants a l‟aridesa podrien mantenir poblacions contínues, com podria ser el cas de Echis leucogaster segons indiquen els resultats de l'anàlisi genètica i els models bioclimàtics. The carpet or saw-scaled vipers of the genus Echis Merrem, 1820 are nocturnal, small (less than 90 cm), fairly stout snakes with a pear-shaped head covered with small scales, prominent eyes with vertical pupils set near the front of the head, and a thin neck. These snakes are found across the semiarid regions of the old world, from Sri Lanka through India, Pakistan, Afghanistan north to Uzbekistan, parts of Iran and the Middle East and in west, northern and east Africa (Gasperetti, 1988; Spawls and Branch, 1995; Wüster, 1997; Arnold et al. 2009). All species have a distinctive threat display, forming c-shaped coils with the body, rubbing their lateral scales with serrated keels to make a hissing sound (stridulation) and striking vigorously. They have a very potent venom and, in many areas, are a common cause of fatal snake bite in people Butll. Soc. Cat. Herp., 18 (2009) 56 (Gasperetti, 1988; Spawls and Branch, 1995; Gutierrez et al. 2006). It has been shown that venom chemistry sometimes varies greatly between different species or populations of Echis, which makes it even more relevant to have a good knowledge of the systematics of the genus. Unfortunately, the saw-scaled or carpet vipers are one of the most taxonomically problematic groups of venomous snakes and, up to date, it is still not certain exactly how many species and subspecies exist within the group (Wüster et al., 1997). For a long time (until the 1980‟s) only two species of Echis were recognized: the Burton‟s carpet viper, E. coloratus Günther, 1878, from Arabia, Jordan, Israel and eastern Egypt; and E. carinatus (Schneider, 1801), believed to occur over most of the range of the genus, from west Africa to India and Sri Lanka (described on the basis of specimens collected in Madras, India). The taxonomy of Burton‟s carpet viper has been unproblematic and the only taxonomic changes include the description of a new species endemic to northern Oman, E. omanensis Babocsay, 2004, and a subspecies from Israel and Jordan named E. coloratus terraesanctae Babocsay, 2003. Molecular analyses (Arnold et al. 2009) have supported the specific status of E. omanensis, while E. c. terraesanctae remains recognized on the basis of morphology only. On the other hand, the systematics of E. carinatus in its broad sense has been confused and unstable in recent decades, especially as a result of the revision by Cherlin (1990), who described several species and subspecies, although his analyses have not been widely accepted by subsequent herpetologists. Morpho-logical taxonomy of the Echis carinatus complex is problematic due to the existence of climatic clines affecting the number of ventral scales (Cherlin, 1981). Recent molecular analyses by Arnold et al. (2009) have shown that E. carinatus sensu stricto is confined to the eastern Arabian Peninsula and Asia, with E. multisquamatus Cherlin, 1990, being genetically very similar to E. c. sochureki Stemmler, 1969, something that had been already shown by Auffeberg and Rehman, (1991) using external morphology. Other subspecies as for instance E. c. astolae Mertens, 1970, from Astolae Island in Pakistan and E. c. sinhaleyus Deriyalaga, 1951 from Sri Lanka have not been tested using molecular data. The study by Arnold et al. (2009) also showed that two further species could be recognized from the Arabian Peninsula: the name E. khosatzkii Cherlin, 1990 being available Butll. Soc. Cat. Herp., 18 (2009) 57 for the morphologically distinct populations from Dhofar, southern Oman and the Hadhramaut, eastern Yemen, and a still undescribed species from Yemen, for which the name E. borkini Cherlin, 1990 may be available. These results suggest that E. pyramidum may not be present in the Arabian Peninsula. The situation in Africa is much more complex and still far from being solved (see for instance Cherlin 1990, Largen and Rasmusen 1993, Spawls and Branch 1995, Wüster et al. 1997, Mazuch 2005; Arnold et al. 2009). Therefore, as suggested by Spawls and Branch (1995), it seems safest to regard the African carpet vipers of the “E. carinatus” group as belonging to three species. These are: the west African carpet viper, E. ocellatus Stemmler, 1970, found in the savannah of west Africa, from southern Mauritania and Senegal east to Nigeria, southwestern Chad and northern Cameroon; the northeast African carpet viper, E. pyramidum (E. & I. Geoffroy St. Hilaire, 1827), found in oases, semi-desert, dry savannah and rocky areas including lava fields across northeast Africa, with apparently isolated populations in northern Egypt, Lybia and Algeria; and the white-bellied carpet viper, E. leucogaster Roman, 1972, found in arid savannah, Sahel, semi-desert and well vegetated wadis across the whole western Sahelian region from the southern half of Mauritania, Senegal and northern Guinea in the west, through central Mali, into western Niger and into the the Hoggar in Algeria, with presumably isolated populations in northern Mauritania, western Sahara and southern Morocco (see Figure 1). This latter species has a similar cephalic scalation than E. carinatus, 27 33 rows of dorsal scales at midbody and 165 180 ventrals (Hughes, 1977). According to Spawls and Branch (1995) and Wüster et al. (1997), it is not clear if the northwestern populations of E. leucogaster from Morocco are really isolated from the Sahelian populations or are the result of lack of collecting in the always-difficult intermediate areas. If these had been truly isolated for a long time, the large geographical area between the northwestern and southern populations might have acted as a barrier preventing contact and promoting genetic divergence. Apart from the obvious biogeographical and evolutionary interest that this may have, this would also be very relevant from a toxicological point of view. It has been shown that antivenoms raised against venom from a population of Echis in one area may be ineffective in treating bites elsewhere, for Butll. Soc. Cat. Herp., 18 (2009) 58 venom chemistry sometimes varies greatly between them (Warrell and Arnett, 1976; Gillisen et al. 1994). Although the number of reported loca-lities for E. leucogaster from the northern populations is still very low in the northwestern area (northern Mauritania, Western Sahara and Morocco), it has increased from two specimens from the same locality in Morocco (Auinet-Torkoz) reported by Bons and Geniez (1996), to a total of nine reported by Aymerich et al. (2004). Of these, two specimens were from two different localities in northern Mauritania and the remaining seven specimens were all from Morocco (see Fig. 1). Most of the Moroccan specimens have been found near Aouinet Torkoz, and the northern-most specimen found so far comes from Amazer, 50 Km south of Ouarzazate; 350 Km northeast in a straight-line from Aouinet Torkoz (Fig. 1). During a recent fieldtrip carried out by the authors in May 2009, a specimen of Echis leucogaster was found at 20:45h dead on the road, close to the Oued Tensif, less than 5 Km up road (northeast) from Tassawant and 10 Km southwest of Agz, Morocco (geographical coordinates in decimal degrees: 30.64212 / -6.58195). For its state of preservation, we could deduce that the animal had been lying there for several days (see Fig. 9A and C in apendix of pag. 127). The area where the specimen was found was a relatively barren, semidesertic and stony plateau at 1059 m of altitude (see Fig. 9B in apendix of pag.127). Bioclimatic data from the nearest meterological station (Ouarzazate) indicate that the region is within the upper thermomediterranean low arid bioclimate belt, with oceanic-high semicontinental conditions (RivasMartinez, 2009). The climate of the locality is characterized by an annual mean temperature of 19.9oC, with annual precipitation of 119 mm and minimum temperature of coldest month of 1.4oC (Worldclim dataset, Hijmans et al., 2005). The specimen shares the typical pattern of coloration and body size of the species but as a result of it poor conservation state it was not possible to carry out any scale counts (see Fig. 9A in apendix of pag. 127). Other species found 5 Km further down the road from the new locality of E. leucogaster (at Tassawant) include several adults and larvae of Bufo brongersmai, Epidalea boulengeri and Bufo mauritanicus, near irrigation canals in cultivated land, and Agama impalearis and Tropiocolotes tripolitanus in the stony semidesert. Butll. Soc. Cat. Herp., 18 (2009) 59 This specimen constitutes the northern-most record of Echis leucogaster ever reported, being 40 Km further north than Amazer (Aymerich et al. 2004; see Fig. 1). Although the specimen was in a poor state of conservation, some details, as for instance the two big inoculating fangs, were very evident (see Fig. 9C in apendix of pag. 127), as well as its stomach content, which included the remains of a Scorpion maurus (see Fig. 9D in apendix of pag. 127). This specimen of E. leucogaster, not only was very relevant from a biogeographical point of view but also constituted a unique opportunity to try to test the degree of genetic isolation of the northwestern populations of E. leucogaster from the Sahelian populations using information from its DNA. In order to do so, a small tissue sample including dry skin, muscle and bone was taken in the field, with special care to avoid the stomach area, and it was kept in absolute ethanol. The specimen was preserved dry at 4oC as a voucher for further morphological and genetic Studies. Once in the laboratory of the Institute of Evolutionary Biology (CSIC-UPF) in Barcelona, the tissue sample was processed using methods described elsewhere (Carranza et al. 2004; 2006) and with special care to avoid contamination. The extracted DNA was used to amplify and sequence the cytochrome b and 16S rDNA mitochondrial fragments, using exactly the same primers and conditions as described by Arnold et al. (2009). The mitochondrial DNA sequences of the new population of E. leucogaster from Morocco were aligned with other Echis samples from GenBank (Table 1) using ClustalX with default parameters (Thompson et al. 1997). The resulting alignment included 1117 base pairs (bp) (731 bp of cytochrome b and 386 bp of 16S rDNA) of which 343 were variable positions. JModeltest v.0.1.1 (Posada, 2008) was used to select the most appropriate model of sequence evolution using the Akaike Information Criterion (AIC). The model selected was the GTR+G, for the data set containing the cytochrome b sequences and the GTR+I+G for the 16S rDNA dataset. The computer program RAxML v.7.0.3 (Stamatakis, 2006) was used for the ML analyses using the “Hard & slow” option, with a heuristic search of 100 trees. The reliability of the ML trees was assessed by bootstrap analysis (Felsestein, 1985) involving 1000 bootstrap replications. Bayesian analyses were performed using MrBayes v.3.1.2 (Huelsenbeck and Ronquist, 2001), with independent models and model parameters applied to each mitochondrial gene partition. All analyses started with randomly generated trees and ran for Butll. Soc. Cat. Herp., 18 (2009) 60 2x10 generations. After checking that stationary had been reached, the first 4000 trees were discarded, and a majority rule consensus tree with branch lengths was generated from the remaining 16,000 trees. In order to interpret the results of the phylogenetic analyses in the light of both the known and the potential distribution ranges of Echis leucogaster, an ecological niche modeling projection was performed. Twenty-five presence locali-ties were digitally georeferenced based on information from Gasperetti (1988), Aymerich et al. (2004) and the new specimen found, using the software Ozyexplorer. Nineteen localities outside the described area for the species were considered as absences. Nineteen climate variables (Worldclim dataset, Hijmans et al., 2005) were extracted using DIVA GIS (Hijmanns et al., 2004) with a very low resolution (0,166o), in order to minimize the effect of possible errors in the accuracy of the georeferenced localities. As the correlation matrix showed that the environmental variables were highly correlated, a principal components regression (PCR) was applied to the variables in order to reduce the dimensions and to identify to the most explanatory variables. All statistical analysis were performed using Statistica 6.0. The results of the phylogenetic analyses are shown in Fig. 3 and clearly indicate that the E. leucogaster from Agz, Morocco is very closely related to the two E. leucogaster samples from Ayoun el‟Atrous (Mauritania). Despite being separated by more than 1600 Km, these samples present an uncorrected genetic divergence (p-distance) of only 1% in the 731 bp of cytochrome b analyzed for this study and are identical in the 386 bp of the 16S rDNA. Since the specimen from Agz is the northernmost sample of E. leucogaster known to date, we can confidently conclude that the genetic divergence between the northwestern and Sahelian populations of E. leucogaster is very low, indicating that these populations are either connected or, if in isolation, the separation occurred very recently. The results of the PCR suggest that the presence of Echis leucogaster in the region might be positively correlated with the increase of temperature and inversely correlated with precipitation (Table 2). This is expected as the species has a mainly Sahelian and perisaharan distri-bution (see Gasperetti, 1988). The most significant variables identified by PCR were successively included in the model excluding variables with highly correlated coefficients. Finally, the selected variables (Table 2) were used in a niche modeling simulation performed with Maxent, a generalButll. Soc. Cat. Herp., 18 (2009) 61 purpose algorithm which makes accurate predictions using data sets that only contain information on the known presence (Philips et al., 2006) and even from small sample sizes (Pearson et al., 2007). In order to determine the accuracy of the model several agreement statistics for the three different thresholds were computed: lowest presence threshold (minimum predicted area with omission error equal to 0, Pearson et al., 2007); OC, optimal cut-off with minimum misclassification rate; and 50, the highest of the 50% predicted values (Table 3). The area under the curve for the model was 0.98, suggesting that climatic variables correctly explained the distribution of Echis leucogaster in the region. The projection map is shown in Fig. 4 and indicates that, under the present climatic conditions, E. leucogaster is likely to have a continuous distribution along the western boundary of the Sahara desert. This is in agreement with the results of the phylogenetic analyses, which suggest that there is almost no genetic divergence between the populations of E. leucogaster analyzed (see Fig. 3). However, the high inaccuracy of the available data jeopardizes the results of the modeling. Future discovery of new localities of this rare species will allow us to either ratify or refute this hypothesis. This apparent continuous distribution in Echis leucogaster predicted by the niche modeling analysis and supported by the genetic data might also apply to other subtropical species with supposedly isolated populations in northwest Africa as for instance Crocodylus suchus, Dasypeltis scabra, Bitis arietans, Lamprophis fuliginosus (Bons and Geniez, 1996; Duplessy et al., 1989). Most probably, these species had a continuous range during the Pleistocene, when much of this northwestern area that is now covered by the Sahara Desert was much more vegetated and would have provided suitable habitat for them. What is not so clear is if, like in the case of E. leucogaster, the actual climatic conditions still allow some contact. The lack of data in the intermediate areas between the Sahel and northwest Africa might be the result of poor sampling caused by the difficult political situation in the region. The present work demonstrates that in cases like this niche modeling can be a very useful tool. However, the degree of geographical isolation, genetic divergence and taxonomic status of these apparently relict populations will only be elucidated when molecular and niche modeling data are available for them, as it is available now for E. leucogaster. Butll. Soc. Cat. Herp., 18 (2009) 62 Acknowledgements This work was carried out under a permit from the Moroccan Government (Department of Water and Forestry resources) to David Donaire.