A new dwarf chameleon, genus Brookesia, from the Marojejy massif in northern Madagascar

We describe a new species of dwarf chameleon from the Brookesiaminima species group. Brookesiatedisp. n. occurs above 1300 m above sea level on the Marojejy massif in northeastern Madagascar. It is genetically sister to B.peyrierasi, a species occurring in lowlands around the Baie de Antongil, but is genetically strongly divergent from that species in both nuclear (c-mos) and mitochondrial (16S, ND2) genes, and morphologically in its smaller size and distinctly different hemipenis. It is the second species of the B.minima species group from Marojejy National Park, but is not known to occur in syntopy with the other species, B.karchei, due to elevational segregation.


Introduction
Chameleons, family Chamaeleonidae, are a highly diverse group of lizards, occurring across Africa and Madagascar, with a few species also present in southern Europe and Asia (Tilbury 2018). Body size within the family varies more than tenfold, from the giant species Furcifer oustaleti (Mocquard, 1894) at 284 mm maximum snout-vent length (SVL) to the dwarf species Brookesia micra Glaw, Köhler, Townsend & Vences, 2012 whose maximum SVL is ca 20 mm snout-vent length (Glaw et al. 2012a;Villeneuve 2017). According to current phylogenetic understanding of the evolution of these lizards (Tolley et al. 2013), chameleons were ancestrally semi-terrestrial, probably roosting above the ground on small twigs and branches at night, and relatively small-bodied, and sub-tributed across the northern half of Madagascar, with one species also occurring in the Tsingy de Bemaraha Massif in the dry west (Schimmenti and Jesu 1996). Glaw et al. (2012a) recognised three clades within the B. minima group: Clade A (B. desperata Glaw, Köhler, Townsend & Vences, 2012;B. tristis Glaw, Köhler, Townsend & Vences, 2012;B. micra, B. confidens Glaw, Köhler, Townsend & Vences, 2012;B. tuberculata Mocquard, 1894) consisting of species distributed primarily in the far north of Madagascar, Clade B (B. peyrierasi Brygoo &Domergue, 1975 andB. karchei Brygoo, Blanc &Domergue, 1970) consisting of species distributed in northeastern Madagascar, and Clade C (B. ramanantsoai Brygoo & Domergue, 1975;B. exarmata Schimmenti & Jesu, 1996;B. dentata Mocquard, 1900;B. sp. 'Betampona') consisting of species distributed in central eastern and western Madagascar, with B. minima Boettger, 1893 from the Sambirano region in northwestern Madagascar occupying a position sister to Clade B. Some undescribed diversity is still known within this species group, however. Brygoo et al. (1974) reported on four specimens (probably MNHN 1986.876-879, formerly all subsumed under the preliminary number 722/C) belonging to the Brookesia minima group that they collected at 1300 m above sea level (a.s.l.) in Marojejy, from the humid faces of ravines, where they were found in bushy foliage and among mosses. The colouration of the specimens was briefly described, but their overall morphology was not discussed. A further specimen collected from the region of Sambava (presumably MNHN 1974(presumably MNHN .251 or 1974 was also referred to the species. Brygoo and Domergue (1975) elected not to describe the species due to the lack of sufficient data. A specimen was later illustrated by Brygoo (1978), and he also reported therein that the specimens were closer to B. tuberculata than to B. peyrierasi in the morphology of their dorsolateral spines and the crests of the dorsal head. According to Brygoo (1978), G. Ramanantsoa had apparently assembled a collection of ca 40 individuals and indicated to Brygoo his intent to describe the species, which presented also hemipenial differences, but this description was never published.
On a recent expedition to Marojejy, we collected new material of this species, and can finally resolve the question of the identity of this population. In the present study we provide a formal taxonomic description of the new species, and provide genetic evidence for its close relationship to B. peyrierasi.

Materials and methods
Specimens were found via searching by torchlight at night and photographed the following morning. They were anaesthetised and subsequently euthanised by oral of lidocaine. Tissue samples were taken and stored in pure ethanol. They were then fixed with 90% ethanol and deposited in 75% ethanol for long-term storage. ZCMV and FGZC refer to field numbers of M. Vences and F. Glaw, respectively. Specimens were deposited in the Zoologische Staatssammlung München, Germany (ZSM). Reference is also made to the Muséum National d'Histoire Naturelle, Paris, France (MNHN) and University of Michigan Museum of Zoology, Ann Arbor MI, USA (UMMZ).
Morphological analysis was conducted following Glaw et al. (2012a). Measurements were taken by MV to the nearest 0.1 mm using a digital calliper. The following measurements were taken: TL (total length); SVL (snoutvent length); TaL, tail length; HW, maximum head width; HH, maximum head height; ED, eye diameter; FORL, forelimb length.
For molecular analysis, we expanded the DNA sequence data set of Glaw et al. (2012a) for the two mitochondrial genes of NADH dehydrogenase subunit 2 (ND2) and 16S rRNA (16S), and the nuclear gene for oocyte maturation factor mos (c-mos). We added to the existing data sequences for the two individuals of the new species described herein, as well as one additional individual of B. peyrierasi. Genomic DNA was extracted and amplified using standard protocols as described in Glaw et al. (2012a). Gene fragments were amplified using the primers ND2F17 (5′-TGACAAAAAATTGCNCC-3′; Macey et al. 2000) and ALAR2 (5′-AAAATRTCTGRGTTGCATTCAG-3′; Macey et al. 1997) for ND2, 16SA-L (5′-CGCCTGT-TTATCAAAAACAT-3') and 16S-BH (5'-CCGGTCT-GAACTCAGATCACGT-3') for 16S (Palumbi et al. 1991), and CO8 (5′-GCTTGGTGTTCAATAGACT-GG-3′) and CO9 (5′-TTGGGAGCATCCAAAGTCTC-3′) for c-mos (Han et al. 2004). PCR products were purified with ExoSAPIT (Thermo Fisher Scientific, Waltham, MA, USA), sequenced on an automated DNA sequencer (ABI 3130 XL; Applied Biosystems). We used the software Co-donCode Aligner (CodonCode Corporation) to check and correct errors in the DNA sequences based on the chromatograms, and to trim poor-quality stretches and primer sequences from the beginning and end of the sequences. We verified that none of the sequences of the protein-coding genes (ND2 and c-mos) contained stop codons, and retained hypervariable stretches in the 16S gene, as an exploratory analysis removing these yielded a similar tree topology. All new sequences obtained in the present study were submitted to GenBank (accession numbers MK452364-MK452366, MK457367-MK457369, and MK457452-MK457454); for accession numbers of previously published sequences, see Glaw et al. (2012a).
Alignment and analysis of sequences, separately for each marker, was performed in MEGA7 (Kumar et al. 2016). Sequences were aligned using the Clustal algorithm, and the most suitable substitution models determined under the Bayesian Information Criterion, implemented in MEGA7. Phylogenetic analyses were conducted under the Maximum Likelihood optimality criterion, with subtree-pruning-regrafting branch swapping (SPR level 5) and with 500 heuristic bootstrap replicates. Bootstrap values <50% (suggesting the respective node is recovered by less than half of the replicates) were omitted from the figures. We decid-ed to analyse sequences for each marker separately rather than concatenating them because (1) a concordant signal of markers with different mode of inheritance (nuclear and mitochondrial) provides additional support for distinct evolutionary lineages under the genealogical concordance species criterion (Avise and Ball 1990), and (2) different sets of individuals have been sequenced for the three markers which would inflate a concatenated matrix with missing data, even for the two mitochondrial markers (for which the sequences in part refer to different individuals which we preferred not to combine to form chimera sequences).
This work is published in digital form and represents a published work under the amended International Code of Zoological Nomenclature. It is published in a journal with an ISSN (print: 1435ISSN (print: -1935online: 1860-0743), and the article and the new name it contains are registered in ZooBank (www.zoobankorg). The Life Science Identifier (LSID) of the publication is urn:lsid:zoobank.org:pub:09A33D57-6581-4A45-99CC-1E1DE17C996E. This article is deposited in the following digital repositories: CLOCKSS, Zenodo, Library of Congress, Portico.

Results
Phylogenetic analysis of the nuclear (c-mos) and mitochondrial (ND2, 16S) markers yielded largely concordant results regarding the phylogenetic placement and molecular divergence of the specimens from Marojejy (Figs 1, 2). While deep relationships among clades of the Brookesia minima clade could not be reliably resolved in our single-marker analyses, the Marojejy specimens were in each case placed sister to B. peyrierasi, with moderate to strong support from bootstrap analysis (ND2: 99%; 16S: 84%; c-mos: 99%). However, the genetic differentiation between these two lineages were substantial: the Marojejy specimens differed from B. peyrierasi by 8.1-9.0% uncorrected pairwise distance (p-distance) in the 16S gene, 17.5-19.3% in the ND2 gene, and 1.7-2.7% in the usually very conserved c-mos gene. Their differentiation to the other species of the B. minima clade amounted to 10.8-17.5% for 16S, 26.1-29.2% for ND2, and 2.7-5.5% for c-mos.
Our genetic data confirm for the first time the presence of B. peyrierasi in a lowland site on the mainland opposite to its type locality, the islet of Nosy Mangabe (Fig.  3). One specimen from Ambodivoangy clustered with 98-100% bootstrap support with Nosy Mangabe specimens in the analyses of all three gene fragments (Figs 1, 2). Despite a substantial genetic differentiation between the Ambodivoangy and Nosy Mangabe specimens (3.2% in 16S, 7.9% in ND2, 0.7% in c-mos), at present there is no reason to consider them taxonomically distinct. This confirms B. peyrierasi as being a species distributed in lowlands of mainland Madagascar near Nosy Mangabe, and we here also assign a specimen from the Masoala peninsula (Fig. 4G, H) to this species based on morphology and biogeography, pending genetic confirmation. Two female specimens from Sambava MNHN 1974.251-252 that were assigned to B. peyrierasi based on external morphology by Glaw et al. (1999) and a further uncollected juvenile from Sambava, depicted in Glaw and Vences (1994), are here omitted from further discussion because neither hemipenial nor molecular data are available from these specimens.
Morphological analysis revealed differences between the Marojejy specimens and B. peyrierasi in body size (B. peyrierasi being slightly larger; see diagnosis below and Table 1). We did not observe additional morphological differences, e.g. in the arrangements of scale rows dorsally on the head, and the two Marojejy individuals differed distinctly from each other in the appearance of their supraocular scales (Fig. 5). However, hemipenial morphology was drastically different between the Marojejy specimens and B. peyrierasi (Fig. 6), and the diagnostic characters were constant in all specimens of B. peyrierasi studied so far (Glaw et al. 1999) and in the two Marojejy individuals (see below for more details).
In addition, B. peyrierasi has so far been collected in lowland localities only (Fig. 3), whereas the two specimens of the Marojejy lineages came from forests ≥ 1300 m a.s.l. Taken together, the concordant and strong differences in mitochondrial and nuclear genes, body size, and hemipenial morphology provide unambiguous evidence for a species-level differentiation of the Marojejy lineage, and we therefore describe and name it as a new species as follows.  Table 1 Remark. This species has been previously referred to as follows: B. "minima" - Brygoo et al. (1974) B. minima s. l. -Brygoo andDomergue (1975) B. aff. minima -Brygoo (1978) B. cf. minima - Glaw and Vences (1994) Possibly also within the definition of B. minima by Raxworthy and Nussbaum (1995); see Referred specimens below.  formerly housed in the Collection of the Institut Pasteur de Madagascar, were apparently collected ('event date') on 4 December 1972 at 'Marojezy' (=Marojejy), 1300 m a.s.l. according to the VertNet database (http://portal. vertnet.org/search, accessed 25 December 2018) and are probably the specimens referred to by Brygoo et al. (1974); we tentatively refer these specimens to this species without having examined them. The whereabouts of additional potential specimens (no. 722 and 723c) and of the ca 40 specimens collected by G. Ramanantsoa reported by Brygoo (1978) is unknown.
Specimens UMMZ 203615-203621 reported by Raxworthy and Nussbaum (1995) to be identical to B. minima may also be attributable to this species, but apparently come from lower elevation (200-800 m a.s.l.) and therefore cannot be referred to this species until they are genetically and morphologically investigated. In any case, given the morphological variation of B. tedi (see Discussion below) and the presence of one other species of the B. minima group at Marojejy (B. karchei), the identification of all these additional specimens as B. tedi will remain tentative unless molecular data can be obtained or fully diagnostic morphological or osteological characters defined in the future.

Diagnosis.
A diminutive chameleon species assigned to the genus Brookesia on the basis of its small body size, short tail, crests of the head, dorsolateral spines and molecular relationships. Brookesia tedi sp. n. is morphologically characterised by the following unique suite of characters (n = 2 males): (1)   ti and Jesu (1996) by the presence of a supraocular cone (vs absence), slightly smaller relative head height (HH/ SVL 0.16-0.18 vs 0.23), and slightly smaller total length (TL 29.5-31.7 vs 33.3 mm); from males of B. minima by slightly longer relative tail length (TaL/SVL 0.74-0.92 vs 0.65-0.73), slightly wider relative head width (HW/SVL 0.20-0.21 vs 0.16-0.19), presence of supraocular cone (vs absent), and presence of a distinct pelvic spine (vs absent or indistinct); from males of B. ramanantsoai by smaller body size (SVL 15.3-18.2 vs 21.7 mm) and total length (TL 29.5-31.7 vs 39.0 mm), and wider relative head width (HW/SVL 0.20-0.21 vs 0.16); from males of B. dentata by smaller total length (TL 29.5-31.7 vs 43 mm) and presence of supraocular cone (vs absent); from females of B. karchei (as reported by Glaw et al. 2012a; no data from males are available) by longer relative tail length (TaL/SVL 0.74-  The holotype of Brookesia tedi sp. n. from Marojejy in lateral view (above), and dorsal (middle) and lateral (below) views of the heads of the types of B. tedi sp. n., in comparison with a representative male specimen of B. peyrierasi from the type locality Nosy Mangabe. The lateral view of the holotype (marked with an asterisk) has been mirrored to be in the same orientation as those of the other specimens. Scale bar: 1 mm. 0.92 vs 0.66), slightly wider relative head width (HW/SVL 0.20-0.21 vs 0.17), absence of lateral tail spines (vs presence), and fewer dorsolateral spines (8-10 when countable vs 13); and from males of B. tuberculata by smaller supraocular cone and globular hemipenis (vs tubular).
Molecular data clearly identify B. peyrierasi as the sister species of B. tedi sp. n. (Figs 1, 2); from males of B. peyrierasi the new species is distinguished by smaller body size  and total length , more distinct supraocular cone, and rather globular hemipenis lacking apical spines (vs tubular hemipenis with paired sets of apical spines; Fig. 6).
Description of the holotype. Adult male in good state of preservation (Figs 5, 7). Both hemipenes everted. Measurements in Table 1. Lateral crest on head weakly devel-oped, barely recognizable; prominent orbital crests, and a transverse row of enlarged tubercles at the posterior edge of the head that separates the head from the body, no distinct posterior crest; a pair of curved parasagittal crests that start above the eyes and begin to converge before terminating at the transverse row of enlarged tubercles; depression between the eyes with short longitudinal median crest; three pointed tubercles on each side of posterior margin of head; scattered slightly enlarged tubercles on lateral surfaces of head; orbital crest denticulated; distinct supraocular cone present; supranasal cone small, not projecting beyond snout tip; head longer (5.1 mm) than wide (3.2 mm); chin and throat with evenly spaced distinctly enlarged tubercles. Dorsal surface of body without a vertebral ridge or keel; 10/8 (left/right) dorsolateral spines (pointed tubercles) form an almost complete longitudinal line on the body; posteriormost pointed dorsolateral tubercle being largest, above insertion point of hindlimb (the pelvic spine), very slightly projecting backwards; no dorsal pelvic shield in sacral area, but distinct pelvic spine; pointed dorsolateral tubercles almost equally spaced, except for first three tubercles on left side; dorsal surface of tail lacking distinctly enlarged tubercles; scattered enlarged tubercles laterally on anterior third of tail; lateral surface of body with evenly spaced enlarged rounded tubercles; venter with distinctly enlarged rounded tubercles; scattered, distinctly pointed tubercles on limbs; no pointed tubercles around cloaca; longitudinal row of slightly enlarged tubercles lateral on anterior tail; no dorsal, lateral or ventral spines on tail; no enlarged tubercles on ventral surfaces of tail.
The hemipenis is rather globular (though not as globular as that of B. minima), short and broad, with a flattened apical end with a clear lip around its circumference (Fig. 6). The whole of the apex is somewhat tilted sulcally. A pair of structures emerge from the apical surface, each of which consists of three fleshy lobes, of which the middle lobe is the shortest. The truncus is smooth and lacks any trace of calyces.
In life, overall colouration light to dark brown, lighter ventrally, the dorsal head and dorsum down to the tail grey. Tubercles and patches of various other colours, including a number of nearly black spots, dot the flanks. Rectangular patches of grey invade the flank from the dorsum. The eye is rayed in shades of brown, with three especially light rays ventrally, the anterior two of which traverse across the upper and lower lips. A brown stripe bordered with cream traverses the supraocular cones. Figure 6. The preserved hemipenes of Brookesia tedi sp. n. and its sister species, B. peyrierasi in sulcal view. Not to scale. Note that the spines of B. peyrierasi are equally visible in asulcal view, and in all studied individuals of this species (see Glaw et al. 1999).
The limbs are distinctly darker than the body. For further detail, see Figure 7. After two years in ethanol, the body colour is muted and more homogeneous. The rays on the eye and stripe between supraocular cones are still distinct, but the grey invading the flanks from the dorsum is less distinct.
Variation. For morphological measurements and proportions see Table 1. The male paratype ZSM 439/2016 differs from the male holotype by a far more prominent supraocular crest, formed by six large pointed tubercles, including a large supraocular cone; by three parallel dorsal crests between parasagittal crests; less prominent enlarged tubercles on throat, chin and venter; indistinct and smaller dorsolateral tubercles (not reliably countable), and a prominent enlarged pointed tubercle on lateral side of head. Its hemipenes are less well extruded than those of the holotype, and the fleshy apical lobes, although present, are difficult to identify.
Etymology. The species name is a patronym dedicated to Ted Townsend, in recognition of his important contributions to the phylogenetics and systematics of squamates, chameleons, and Brookesia in particular.
Conservation status. This species is currently only known from relatively high elevation on the Marojejy massif. We follow assessments for other chameleon endemics from this area on Marojejy, specifically Calum-ma jejy and C. peyrierasi, which are found somewhat higher but probably have a similar level of microendemism. We consider the species Vulnerable under IUCN Red List criterion Vulnerable D2: as far as is known, B. tedi has a highly restricted area of occupancy (= extent of occurrence) of under 150 km 2 (this is the area of 1200 m a.s.l. and above in Marojejy), and is known from a single threat-defined location at 1300 m a.s.l. and above in Marojejy National Park. Two plausible future threats, namely decrease in efficacy of protection on Marojejy, and fire, could rapidly drive the species to becoming Critically Endangered.