The multilocus phylogenetic trees recovered from both optimality criteria (MP and ML) provide robust support for the monophyly of (1) the F. labordi + F. lateralis clade, (2) the F. lateralis complex, and (3) F. labordi clade. However, phylogenetic trees resulting from the analysis of individual loci are not well resolved. The mitochondrial genes 16S and ND2 are inconclusive with regard to the relationship of F. lateralis with its sister species F. labordi. Additionally, the individually analysed nuclear loci do not recover much phylogeographic structuring within the F. lateralis complex, supporting the perspective that a single locus will often be unable to accurately facilitate species identification, even for DNA barcoding (Roe et al., 2010; Niemiller et al., 2011).
Separate analysis of the mitochondrial and nuclear data reveals several interesting patterns. For MXRA, the resulting median-joining network reveals a similar result as the mitochondrial data, with some exceptions. Most surprising is that individual TRA143 is homozygous for a MXRA haplotype also seen in north-west individuals (Fig. 2b), yet this specimen is recovered as part of the eastern mtDNA clade and has only ‘eastern’ haplotypes with the other nuclear gene RAG1. TRA143 is a F. lateralis specimen collected at the very southern range of the eastern clade, lending evidence that this individual could be a hybrid. However, it is difficult to distinguish between hybridization and incomplete lineage sorting in recently evolved groups (Polihronakis, 2009), and either of these are possible causes of discrepancies in mtDNA and nDNA presented in this study. The median-joining network for MXRA revealed several other differences between this locus and the mtDNA. However, these discrepancies did not result from individuals distributed at the extremes of their ranges, suggesting that this might instead be an example of incomplete lineage sorting.
In the RAG1 median-joining network, there is some support for the mtDNA clades, but the pattern is much less clear. Particularly interesting is that the mtDNA eastern clade is broken into two sections, but these do not appear geographically correlated (Fig. 2c). In addition, the identical haplotype group ‘TYPE3’ contains a mixture of individuals from the three different mtDNA clades. While most belong to the north-west mtDNA clade, this group also includes a large number of southern mtDNA individuals and one eastern mtDNA individual. Because these individuals are not found where the clade distributions meet and because these results contradict both the mtDNA tree and MXRA median-joining network, it is unlikely that this represents continued gene flow between clades. Instead we suspect that this represents a case of incomplete lineage sorting. However, as is the case above with the discrepencies found with MXRA and the mtDNA clades, more loci, samples and analyses are needed to test between these possibilities.
The combined molecular data support three distinct north-west, eastern and southern clades within a F. lateralis species complex with the north-west and southern clades recovered as sister groups. This supports an initial divergence in the F. lateralis complex between populations in the more humid east and those in the drier south and west. These clades are also highly differentiated with mitochondrial Dxy values as high as 7.8%. Taken alone, these data suggest that these clades warrant species-level recognition, as this has been shown to be characteristic of sister species in other taxa using a multimethod approach similar to this study (Brown et al., 2002; Rissler & Apodaca, 2007) and also across diverse species criterion (see Bradley & Baker, 2001; Hart & Sunday, 2007). The lower level of divergence between the southern and north-west clades (Dxy = 3.6%) is also highly suggestive that these two genetic clades are different cryptic species.
There is some genetic structure within each clade that corresponds with geography, but faster-evolving loci and more samples will be needed to further elucidate these patterns. The eastern and southern clades have a higher mitochondrial diversity than the north-west clade. On the basis of the isolation-by-distance (Slatkin, 1993), this result is surprising because the north-west clade is the most widely distributed. The low genetic diversity in this clade may be indicative of recent population expansion northwards. We hypothesize expansion northwards based on greater genetic structuring in the samples distributed in the most southern regions of this clade’s range (with Makay populations clustering together – see Fig. 1a).
These results demonstrate that Brygoo’s (1971) morphologically distinct form in the south-west deserves species recognition, as also discussed by Boumans et al. (2007) based on their 16S mitochondrial data. Boumans et al. (2007) also found evidence for several other genetically structured clades within the complex, but overall support values were low. The early east/west divergence found in our study is also congruent with several other taxa. The geckos, Phelsuma madagascariensis and Phelsuma kochi, are closely related species that have disjoint distributions in eastern and western Madagascar, respectively (Raxworthy et al., 2007). In addition, the lemur Eulemur fulvus rufus contains two genetically distinct groups distributed between eastern and western parts of Madagascar (Pastorini et al., 2003). This east–west split has been found in other taxonomic groups as well, including other geckos, tree boas and frogs (eg. Nussbaum & Raxworthy, 1998; Andreone et al., 2002; Orozco-ter Wengel et al., 2008– but see the supplementary material of Vences et al., 2009 for a full review of speciation studies). The south/north-west divergence is less common, but has been highlighted in other studies. For example, there is a genetically distinct southern group and a widely dispersed north-western group within the Malagasy plated lizard Zonosaurus laticaudatus (Raselimanana et al., 2009).
Morphometric support for cryptic species
Canonical variates analyses of both males and females found morphological differences among the three clades, with CV1 accounting for 87.6% of the variation among clades for males and 77.0% among females. CV1 likely relates to changes in the landmarks associated with the head casque (see Table 2 and Fig. S1), with eastern individuals having a much lower head casque that extends posteriorly, almost flattening into the dorsal ridge of the body. The head casque on the southern clade individuals are the most elevated, with north-west clade individuals being slightly lower. Higher casque height has been correlated with stronger bite force in lizards (Herrel et al., 2001) because this development results in the enlargement of the medialis and profundus portions of the external jaw adductor (Rieppel, 1981). Interestingly, in the field we noted differences in aggressive behaviour exhibited between these three groups, with southern individuals showing the greatest tendency to bite when handled or placed with conspecifics in collecting bags. This suggests that the southern F. lateralis group may have more aggressive behaviour that may explain a larger head casque and greater bite force. In contrast, eastern individuals (with flattened head casques) were extremely docile, displaying little aggression (Florio, pers. obs.). Head casque differences between clades could also be attributed to differential dietary preferences. One study found that exaggerated head morphology and absolute bite force are correlated with prey size in chameleon species found in open habitats (Measey et al., 2011). Little is known about dietary preference in Furcifer chameleons, but it is interesting to note that individuals in the southern clade (with the highest head casques) are generally distributed in open spiny desert habitat (Florio, pers. obs.).
We found both similarities and differences between our morphological results and previous studies. While Hillenius (1959) noted a prevalence of white lines under the tail in the F. lateralis complex distributed in southern Madagascar, we found this to be more common in the eastern individuals (data not shown). However, our results support Hillenius’ view that axillary pits are absent in lizards of the southern region of the island. Brygoo (1971) noted larger individuals in the south-west that are less rich in colour, similar to our findings that southern individuals are often pale green and large in size (see Proposed Taxonomic Revision section).
ENMs for species delimitation and phylogeographic studies
Ecological niche models have the potential to help delimit cryptic species when combined with other data, such as phylogenetic and morphological analyses, by providing evidence of (1) improved niche descriptions in split versus lumped taxonomies, (2) niche divergence and (3) geographic isolation between lineages (Raxworthy et al., 2007; Rissler & Apodaca, 2007; Leachéet al., 2009). Separate ENMs for each of the three groups were statistically significant and provided a better fit to the data than the model treating F. lateralis as a single species, and niches in geographic space are unique to each species (although partially overlapping). These results agree with the morphological and genetic divergences found among these three potential cryptic species.
In addition to aiding in species delimitation, our ENMs also provide insights into the processes driving divergence in the F. lateralis complex. On the basis of the largely nonoverlapping distribution of the ENMs for the eastern vs north-west/southern clades, initial divergence may have been driven by adaptation across the western to eastern climate gradient of Madagascar. This provides support for the Ecogeographic Constraint model of diversification (Yoder and Heckman, 2004) that proposes that the distinction between the climate and resulting vegetation of western and eastern Madagascar was the driver of initial divergence between widely distributed animal groups. With regarad to the humid eastern distribution of F. lateralis, this ecological divergence is especially significant in that its sister species F. labordi occupies the more arid southern and western regions of Madagascar, as do the other closely related species (e.g. F. oustaleti, F. verrucosus, and F. antimena) within the complex. If the Ecogeographic Constraint model is correct, this predicts that future speciation within the eastern clade will be constrained to humid eastern habitats. This may also include future range expansion into the north-east, much of which has not yet been occupied by this species. By contrast, the ENMs for the sister southern and north-west clades are overlapping, and the distributions of these clades meet at (or close to) the Mangoky River, suggesting that this river (which is one of the biggest in Madagascar) may have acted as a barrier to gene flow. Martin (1972) first proposed a riverine barrier model of diversification for Madagascar in lemurs, and more recently large rivers have been shown to drive divergence in several lemur groups (see Goodman and Ganzhorn, 2003; Pastorini et al., 2003). Recent fieldwork resulted in the discovery of individuals belonging to the north-west clade distributed just south of the Mangoky River. This could be due to human-mediated dispersal or could indicate that the Mangoky River is no longer an isolating barrier between clades (although it may have caused initial divergence). Further analysis of this potential contact zone is needed to better test this hypothesis. Although preliminary, current evidence suggests that at least two different speciation processes may have driven diversification within the F. lateralis complex.
The ENMs also provide interesting information on the potential distribution of F. lateralis. The southern clade of the F. lateralis complex has a disjunct area of over-prediction in north-eastern Madagascar (Sambava region) where southern F. lateralis are not known to occur. This region may represent an area of potential endemism for other Furcifer species (Raxworthy et al., 2002).
Proposed taxonomic revision
Given the results of this study, we conclude that the F. lateralis complex represents three species corresponding to the three clades recovered from phylogenetic analysis. These clades are not only well supported, but individuals between clades also display differences in head casque morphology and have diverged in environmental niche space. Although we are not the first to propose or suspect that F. lateralis is a species complex (Angel, 1921; Brygoo, 1971), this is the first study to employ a combination of methods, and to find broad support for three species within the group. Because there are three available names for the species within the complex –F. lateralis (Gray, 1845), F. lambertoni (Angel, 1921) and F. lateralis major (Brygoo, 1971) – it is important to assign their types to the appropriate species identified in our study. The adult male syntype of F. lateralis falls within the eastern CVA group of F. lateralis and was not statistically different from the other eastern individuals included in the analysis using permutation tests of the Mahalanobis and Procrustes distances (although the Procrustes distances was also not statistically different to the north-west group). Although this syntype has no precise locality within Madagascar, based on the CVA results, we are confident that Gray’s (1845) description refers to the eastern group recovered in this study, and other morphological characters also diagnose the F. lateralis syntypes as eastern individuals (see below).
The holotype of F. lambertoni MHNP 1921.269 collected from Antananarivo (Central High Pleateau) is a juvenile (SVL = 57 mm, with an open mouth) and therefore could not be included in the geometric morphometric analyses. Upon our examination of the F. lambertoni holotype, we found a weak gular crest present on the anterior part of the chin, which is a typical condition in smaller juvenile F. lateralis. All other diagnostic characters (see below) also show no differences between the F. lambertoni holotype and F. lateralis (from eastern and central Madagascar). We also note that we have collected typical F. lateralis from the F. lambertoni type locality, Antananarivo. We therefore agree with Hillenius (1959) that F. lambertoni is a junior synonym of F. lateralis.
Brygoo’s (1971) description of F. lateralis major from Tanandava, south-western Madagascar, closely corresponds with the morphology of the southern group recovered in our study. These individuals have higher head casques and have a larger SVL. Unfortunately, we were unable to examine the holotype (designated by implication, see Klaver & Böhme, 1997), because this specimen (457/C, Brygoo’s chameleon collection) could not be located during the time of this study and has not been catalogued at the Muséum national d’Histoire naturelle (Paris). However, based on the excellent description and illustration of the holotype, we are confident that the southern group should be assigned to this taxon. We thus formally elevate here F. lateralis major to the rank of full species: Furcifer major.
Because there is no available name for the north-west species, we describe it as a new species: Furcifer viridis new species (Fig. 5).
Figure 5. Live adults of the Furcifer lateralis complex: (a) male F. lateralis (Kianjavato), (b) female F. lateralis (Mandalahy Forest), (c) male Furcifer major (Andoharano), (d) female F. major (Andranohinaly Village), (e) male Furcifer viridis (Anoalakely), (f) female F. viridis (Makay Massif).
Download figure to PowerPoint
Holotype: AMNH 152603 (RAX 5989), a mature male, collected 1 March 2003 at Ambinanitelo (14.22556°S, 48.96297°E), 1250- to 1300-m elevation, Tsaratanana Massif, Mahajanga Province, Madagascar, by N. Rabibisoa, S. Mahaviasy & N. Rakotozafy. Left hind limb was removed and preserved in ethanol for DNA extraction.
Paratopotypes: AMNH 152604 (RAX 5990) and AMNH 152606 (RAX 5995) – both mature females collected in the same locality and on the same date as the holotype.
Other specimens: See Supporting information.
Diagnosis: A Furcifer chameleon from Madagascar with a double row of scales along the dorsal body ridge, which can be distinguished from all other species with this character by the following: 14–18 tubercles on the parietal crest (F. major 10–13, F. campani 7–8); axillary pits always present or at least indicated (F. major absent); a typical solid green adult colour in life, with or without a single pale line on the flank (F. lateralis and F. campani with complex pale and dark spotting, often on a dark brown or reddish brown background, and for F. campani 2–3 pale lines on lateral body); maximum SVL of 120 mm for males and 98 mm for females (F. lateralis with maximum SVL of 98 mm for males, and 92 mm for females; F. campani with maximum SVL of 70 mm for males and females), head casque height/head height > 0.5 (F. campani and F. lateralis < 0.5); and no regular rows of enlarged round tubercles on flanks (F. campani, regular rows of enlarged round tubercles on flanks). Furcifer viridis is also diagnosable from other species based on phylogenetic analysis of the mitochondrial and nuclear loci (see results).
Description of holotype: Male, in good condition, but missing the left hind limb, which was removed and preserved in ethanol for DNA extraction; hemipenes everted; SVL, 81 mm; tail, 89 mm; axilla–groin distance, 56 mm; eye horizontal diameter, 9 mm.
Head lacks a rostral appendage, the orbital crests do not make contact anteriorly at the snout tip; orbital and lateral crests (the latter of which are weakly defined) form a dorsal helmet; helmet posteriorly comes to a blunt point and is elevated above the dorsal ridge of the body; parietal crest weakly developed, formed by a row of 18 tubercles; temporal crest not obvious behind eye; orbital crest rounded in lateral view and formed by a single row of scales; no occipital lobes or folds on each side of the head; gular crest formed by a row of pointed tubercles, which continues to the thorax. Head casque height, 8.5 mm; head height, 15 mm; ratio of head casque height/head height, 0.57.
Dorsal ridge of body with a vertebral double row of rounded tubercles that do not form a crest and flanked below by two other rows of regularly arranged tubercles; body laterally with homogeneous scalation that lack enlarged tubercles; thorax with a ventral crest of short pointed tubercles; body otherwise lacks a ventral crest of pointed tubercles; axillary pits present; limbs with scattered slightly enlarged rounded tubercles, tail with a vertebral double row of rounded tubercles that do not form a crest, feet without tarsal spines. Hemipenes quadriform with calyces on the truncus; apex smooth with a pair of elongate pedunculi bearing 14–16 short papillae; and a small denticulated auricula on the external lateral side and a tuft of papillae on the sulcal side at the base of each pedunculus.
In preservation, the coloration of head, body, limbs and tail is black, with a single pale lateral line present on the flank running from above the front limb to just anterior, and above, the hind limb insertion point. There are a few reddish brown blotches on the posterior head and neck. Ventrally there is a prominent white line that begins in the gular region and fades out under the tail. Just behind the cloaca, there is a pair of short white lines that extend 4 mm onto the ventral tail base.
Variation: See Table 3 for summary data of the examined material (listed in Supporting information). Adult males vary greatly in size (SVL, 65–120 mm) but females never exceed 98 mm (SVL, 64–98 mm). Females and juveniles lack the swollen tail base of adult males. The posterior head casque in females is lower than that in equivalent sized males. The parietal crest tubercle count varies from 14 to 18. Specimens always have axillary pits or they are at least indicated, with specimens from the most southern populations tend to have more poorly developed axillary pits. The pair of white ventral tail base lines may be weakly marked in some specimens.
Table 3. Morphological variation in all Furcifer species with a vertebral double row of tubercles on the body. All measurements in mm. Coloration is based on live animals at rest.
|Maximum male SVL||120||117||98||70|
|Maximum female SVL||98||112||92||70|
|Tubercles on parietal crest||14–18||10–13||14–20||7–8|
|Head casque height/head height||> 0.5||> 0.5||< 0.5||< 0.5|
|Enlarged round tubercules on flanks||–||–||–||+|
|Body mostly solid green or brown||+||+||–||–|
|Body with pale and dark spotting||–||–||+||+|
|Number of pale lines on lateral body||1||1||1||2–3|
In life, the unstressed coloration of the body, head, limbs and tail in adult males is vivid green, with a white lateral line on the body flank and a white labial line. The casque, eye turret and flanks may be marked with scattered small pale blue flecks. Adult females have a similar coloration except that the ground colour may be pale green, pale brown, or greenish brown, with sometimes a pale yellowish brown vertebral line, and without blue flecking. The white lateral body line and labial line may also be weakly displayed in females. Juveniles have a more uniform pale brown or green body coloration. Preserved specimens are often dark, with a weak pale lateral line on the flanks and a bright white mid-ventral line on the throat and body. Some specimens have pale gular skin between the scales that gives a striated appearance on the throat.
Distribution: This species has a large distribution throughout western and northern Madagascar, ranging from Ambodiampana (13.7°S, 49.6°E) in the north-east, to the Makay Massif in the south (21.6°S, 45.1°E) and extending into the interior of the island as far east as Mandoto (19.6°S, 46.3°E – see localities on Fig. 1). At the type locality, it has been found at a maximum elevation of 1250–1300 m, but the species also occupies lower coastal areas. In western Madagascar, F. viridis occupies dry deciduous forest, scrub and grasslands; and on the western high plateau and in northern Madagascar, it occupies more humid and transitional forests, grasslands and scrub. Like F. lateralis and F. major, F. viridis individuals are tolerant of habitat degradation and are often found near rice fields, streams and rivers.
Remarks: Hillenius (1959) reported geographic variation in F. lateralis concerning the development of axillary pits and the ventral white tail line. The western and north-western populations that he examined represent populations of this new species, and he noted a general trend for these populations to have better-developed axillary pits and a more obvious ventral white tail line.
Etymology: This species is named to recognize the predominantly green body.