A phylogeny of the genus Chamaeleo with

investigation of cryptic speciation

By

Devon Campbell Main

Submitted in fulfilment of the requirements for the degree

MAGISTER SCIENTIAE in Zoology

In the Faculty of Science

University of Johannesburg

Supervisor: Prof Bettine van Vuuren

Centre for Ecological Genomics and Wildlife Conservation, Department of Zoology,

University of Johannesburg

Co-supervisor: Prof Krystal Tolley

South African National Biodiversity Institute (SANBI)

Centre for Ecological Genomics and Wildlife Conservation, Department of Zoology,

University of Johannesburg

March 2019

i

“Real knowledge is to know the extent of one’s ignorance.”- Confucius

ii Table of Contents:

Declaration

Acknowledgments

Chapter 1 General Introduction

Chapter 2 Genus-wide phylogeny and species delimitation within Chamaeleo

Appendix I

Appendix II

Chapter 3 Biogeography of Chamaeleo and ecological partitioning of Chamaeleo dilepis

Appendix III

Appendix IV

Chapter 4 General Conclusion

iii Declaration:

AFFIDAVIT: MASTER AND DOCTORAL STUDENTS TO WHOM IT MAY CONCERN

This serves to confirm that I, Devon Campbell Main, ID Number 9401245028080, Student number 201327040 enrolled for the Qualification Masters in Zoology in the Faculty of Science, herewith declare that my academic work is in line with the Plagiarism Policy of the University of Johannesburg with which I am familiar.

I further declare that the work presented in the dissertation is authentic and original unless clearly indicated otherwise and in such instances full reference to the source is acknowledged and I do not pretend to receive any credit for such acknowledged quotations, and that there is no copyright infringement in my work. I declare that no unethical research practices were used or material gained through dishonesty. I understand that plagiarism is a serious offence and that should I contravene the Plagiarism Policy notwithstanding signing this affidavit, I may be found guilty of a serious criminal offence (perjury) that would amongst other consequences compel the University of Johannesburg to inform all other tertiary institutions of the offence and to issue a corresponding certificate of reprehensible academic conduct to whomever requests such a certificate from the institution.

Signed at Johannesburg on this 7th day of March 2019

Signature Print name Devon Campbell Main

STAMP: COMMISSIONER OF OATHS

Affidavit certified by a Commissioner of Oaths

This affidavit conforms with the requirements of the JUSTICES OF THE PEACE AND COMMISSIONERS OF OATHS ACT 16 OF 1963 and the applicable Regulations published in the GG GNR 1258 of 21 July 1972; GN 903 of 10 July 1998; GN 109 of 2 February 2001 as amended.

iv Acknowledgments:

This dissertation represents the product of many hours of research, and indeed frustration, trying to disentangle the evolutionary intricacies of this unique clade of reptiles. I consider it a great privilege to have been given the opportunity to work on such fascinating and enigmatic creatures, and even more so to have been afforded the chance to work with such experts in the field.

Accordingly, it is to my supervisors that I owe my most endearing gratitude. Prof Bettine van

Vuuren provided me with unwavering support and level-headed guidance throughout the past two years, always providing me with encouragement, even during the most frustrating times, for that support I am incredibly thankful. Prof Krystal Tolley was a real beacon of knowledge and she was and still is a role-model for a conscientious and ever-curious work ethos. Her criticisms were instrumental in improving my understanding of difficult concepts, she pushed me to be the best scientist I could be and taught me to think critically and always ask questions.

I would also like to thank Tyron Clark for the many meaningful conversations we shared relating to my research, many of my most important inferences were borne during those conversations.

My friend Gillian Fourie provided me with invaluable assistance with all of my graphics, for that I feel very much indebted to her.

I would also like to acknowledge those working at the South African National Biodiversity

Institute (SANBI) in Kirstenbosch, in particular, Nic Telford who was very helpful and accommodating towards me on my visits to the SANBI lab, and Jody Taft who was incredibly

v patient and helpful to me during my qGIS woes. I also need to make mention of Hanlie

Engelbrecht who made herself available to me during many of my analyses for guidance and advice.

Thanks also go out to those who share a lab with me at the Centre for Ecological Genomics and Wildlife Conservation, in particular I would like to thank Heather Webster, John Chau and

Daniela Monsanto for reading through many drafts of my dissertation and Shilpa Parbhu for helping me with my final formatting. I am also grateful to Sam van Zwieten, Claudia Schnelle, and Erin Crowhurst, who were third years at the time, and took time out of their holidays to help me with my lab work.

Finally I would like to thank my family for supporting me and allowing me to pursue a career in science.

vi 1 Chapter 1

2

3 General Introduction:

4

5 The stochastic nature of the evolution of sexually reproducing organisms has long presented a

6 challenge to the categorization and subsequent documentation of the planet’s biodiversity (De

7 Queiroz 2007; Avise and Robinson 2008; Carstens et al. 2013). Consequently, a plethora of

8 approaches to this conundrum have been proposed, many of which are borne on the notion that

9 ‘species’ form axiomatic entities of life (De Queiroz 2005). This dogma originates from the

10 work of the Swedish botanist Carolus Linnaeus who envisaged a classificatory system of

11 binomial nomenclature which posited that all organisms could be classified into distinct

12 ‘species’ which form basal units within a greater taxonomic hierarchy (Linnaeus 1735).

13 Linnaeus’ taxonomic framework was based for the most part on morphological similarities

14 between conspecifics. Common understanding at the time was that ‘species’ were

15 unchangeable entities of life, this simplified their categorization. Darwin (1859), in his

16 pioneering “On the origin of the species by means of natural selection”, was one of the first

17 people to challenge the idea that species were unchangeable. He hinted at all species having a

18 transient nature on an evolutionary timescale and originating from a single common ancestor

19 (Brooks and McLennan 1999).

20

21 In suggesting an evolutionary ancestry for species, Darwin essentially sparked the beginning

22 of a new biological field, that of evolutionary biology (De Queiroz 2005). Ironically, the field

23 of evolutionary biology further complicated the species problem. Instead of recognizing

24 species as mere units of systematic classification, evolutionary biologists see species as units

25 of evolutionary change (Brooks and McLennan 1999). Since Darwin, taxonomists and

26 evolutionary biologists alike have grappled with the task of proposing a universally true

1 27 definition for species. The mid 20th century saw a boom in biological research punctuated by

28 the assimilation of Darwin’s theory of evolution by natural selection, and Gregor Mendel’s

29 Mendelian genetics, in a movement known as the Modern (Evolutionary) Synthesis (Mallet

30 1995; De Queiroz 2005; Rose and Oakley 2007). Ernst Mayr was a prominent figure of the

31 Modern Synthesis; he proposed what is popularly known as the biological species concept

32 (Mayr 1942; De Queiroz 2005). Mayr defined species as populations of interbreeding

33 organisms (or those that could potentially interbreed) that were reproductively isolated from

34 one another (Mayr 1942). This definition attracted much praise (and criticism), but it also laid

35 the foundation for other species definitions, many of which were based on criteria applicable

36 only to the fields of expertise of their proposers. For instance, ecologists made use of an

37 ecological species concept and geneticists advanced the phylogenetic species concept (De

38 Queiroz 2005). The stochasticity of natural selection and the mechanisms of inheritance have

39 made it essentially impossible to provide a single species concept that is explicitly true for all

40 life forms and it might even be argued that the notion of species is a fictitious human construct

41 (Carstens et al. 2013; Mishler and Wilkins 2018; Zachos 2018). That said, the concept of

42 species is unlikely to be disregarded anytime in the near future and it may indeed be very

43 valuable to conservation, be it a fictitious construct or not (De Queiroz 2007; Zachos et al.

44 2013; Garnett and Christidis 2017).

45

46 Just as species themselves have evolved over time, the concept of biological species has

47 evolved significantly since its inception in Linnaean taxonomy. Recent advances in molecular

48 biology have initiated a general change in opinion among taxonomists from the biological

49 species concept initially proposed by Mayr to an array of phylogenetic species concepts (Isaac

50 et al. 2004; Zachos et al. 2013). However, the general consensus among biologists in the 21st

51 century is that it is best to make use of integrative taxonomy which assimilates multiple species

2 52 concepts and views congruency between them as a requirement for the characterization of new

53 species (Fujita et al. 2012; Puillandre et al. 2012).

54

55 A positive consequence of the inclusion of molecular techniques in the integrative taxonomy

56 approach is the recognition of cryptic diversity – multiple species that were previously

57 described under the heading of a single species using morphological criteria; this was the norm

58 prior to the Modern Synthesis (Bickford et al. 2007). Morphological criteria alone often either

59 fail to recognize recent adaptive variation, or, morphological variation is present but this

60 variation is not reflected in the genetics of the radiation. Such discordance could result from a

61 variety of phenomena that may include phenotypic plasticity, hybridization, and incomplete

62 lineage sorting (Seehausen 2004; Lefébure et al. 2006; Mossel and Roch 2007; Degnan and

63 Rosenberg 2009; Pfennig et al. 2010). These phenomena are evident in the recent adaptive

64 radiations of African cichlids, Darwin’s finches, and Anolis lizards to mention but a few (see

65 e.g. Losos et al. 1998; Seehausen 2006; Losos and Ricklefs 2009).

66

67 The incorporation of molecular techniques in taxonomy can sometimes result in taxonomic

68 inflation (Isaac et al. 2004). Taxonomic inflation is the artefactual increase in biodiversity not

69 due to the discovery of new species but rather due to seemingly arbitrary changes in the way

70 organisms are classified. Such changes are sometimes prompted by subspecific differences that

71 exceed the controversial yet popularly accepted ‘barcode gap’. The barcode gap is a widely

72 accepted notion which postulates that interspecific genetic differences must exceed those found

73 within lineages by an order of magnitude (the so-called 10x rule) (Lefébure et al. 2006;

74 Monaghan et al. 2006; Bickford et al. 2007). Due to the heterogeneity of the rate of molecular

75 evolution across the kingdoms of life, certain groups do not have a valid barcoding gap. In

76 some cases the barcoding gap may be present for a few taxa in a group and this may give the

77 artificial impression that it exists for all taxa, in such cases, an unrealistically low interspecific

3 78 genetic distance threshold might be proposed and therefore subspecies might be artificially

79 elevated to species.

80

81 Both taxonomic inflation and the identification of cryptic species can potentially have far-

82 reaching outcomes as many conservation initiatives quantify biodiversity in terms of ‘species

83 diversity’ and hence funding is generally directed at the conservation of listed species (Isaac et

84 al. 2004). It may be argued then that taxonomic inflation and the discovery of cryptic species

85 should benefit the conservation of the planet’s biodiversity. Unfortunately, conservation

86 resources are limited to the point that a concept of conservation triage, akin to the triage used

87 by paramedics, has been proposed to help prioritize conservation efforts (Bottrill et al. 2008).

88

89 So, attempting to conserve all species on earth is probably too audacious a goal. With that said,

90 identifying what truly warrants conserving is a task that deserves much attention. While the

91 notion of species may be a controversial one, evolutionary biologists and ecologists recognize

92 that discrete evolutionary lineages hold adaptive evolutionary potential (Moritz 1994; Crandall

93 et al. 2000). In acknowledging this it is evident that while attempting to conserve all listed

94 species might be impractical, conserving lineages that have evolutionary significance, typically

95 mediated by historical geographic barriers or corridors might hold promise for the preservation

96 of distinct adaptive evolutionary potential. The recognition of lineages with independent

97 evolutionary trajectories has hence become a central theme in many conservation initiatives –

98 these lineages have been referred to as evolutionarily significant units (hereafter ESUs) (Moritz

99 1994; Crandall et al. 2000; Fujita et al. 2012). Akin to ESUs is De Queiroz (2007)’s since

100 widely accepted recognition of species forming independently evolving metapopulation

101 lineages (i.e. the General Lineage Concept of Species). De Queiroz (2007) has made a

102 distinction between the problem of ‘species concepts’ and ‘species delimitation’ strategies.

103 While conflicting theories exist pertaining to the mechanisms behind speciation, De Queiroz

4 104 (2007) argues that the resulting outcome of these mechanisms is ultimately something that can

105 be unanimously agreed upon: molecular partitioning on an evolutionary timescale. Quantifying

106 the degree of isolation between metapopulation lineages, and when this isolation may be

107 considered taxonomically significant, is something that has attracted much recent attention

108 among taxonomists and molecular systematists (Reid and Carstens 2012; Carstens et al. 2013;

109 Yang and Rannala 2014; Kapli et al. 2017).

110

111 Moritz (1995) advocates the use of molecular phylogenies in unraveling taxonomic

112 relationships and identifying ESUs. In the 21st century, molecular phylogenies have become

113 the norm in inferring taxonomic relationships and species delimitation, however, the

114 parameters by which these phylogenies are carried out and interpreted are a source of much

115 debate (Monaghan et al. 2006; Edwards 2009; Fujita et al. 2012; Yang and Rannala 2012). A

116 requirement for a lineage to be considered as evolutionarily significant according to Moritz

117 (1994) is reciprocal monophyly, at least for mitochondrial markers – this is a consensus

118 accepted by many (Fujita et al. 2012; Puillandre et al. 2012). Crandall et al. (2000), however,

119 argue that while reciprocally monophyletic lineages may be of evolutionary significance, this

120 criterion fails to recognize recent adaptive divergence which could hold much evolutionary

121 potential as in the case of African cichlid and Darwin’s finch radiations (Grant et al. 2004;

122 Seehausen 2006; Wagner et al. 2012).

123

124 New coalescent phylogenetic approaches circumvent the problems introduced by Moritz’s

125 (1994) requirement of reciprocal monophyly by using multilocus data and in so doing,

126 recognize the evolutionary potential of recent adaptive variation (Degnan and Rosenberg 2009;

127 Yang and Rannala 2010; Fujita et al. 2012). Different regions of the genome are under different

128 selective pressures (although debated, some appear to be under no selection at all) and therefore

129 different genes evolve at different rates. This can result in different gene fragments producing

5 130 conflicting trees which may in turn be in conflict with the actual species tree. The Multispecies

131 Coalescent Model (MSC) uses multilocus data to model different evolutionary scenarios and

132 thereby account for discordance between gene trees and species trees as well as ancestral

133 polymorphism (Rannala and Yang 2003; Xu and Yang 2016). Since the derivation of the MSC,

134 software packages that operate on the basis of the MSC, such as Bayesian Phylogenetics and

135 Phylogeography (BPP), have been developed to help delimit species with conflicting gene trees

136 and ancestral polymorphism (Yang 2015; Leaché et al. 2018). In some cases, multilocus data

137 are a luxury that cannot, for financial or practical reasons, be acquired. In such cases, other

138 species delimitation techniques such as the Bayesian Generalized-Mixed Yule Coalescent

139 (hereafter bGMYC) model (Pons et al. 2006; Reid and Carstens 2012) and Multi-rate Poisson

140 tree processes (hereafter mPTP) (Kapli et al. 2017) can be used in coalescent approaches that

141 do not require multilocus data.

142

143 Underlying the diversification and spatio-temporal distribution of biological organisms is a

144 complex array of historical geological and climatic events. The study of biogeography lies at

145 this interface between biology and geography (Cracraft 1975; Nelson 1985; Smith 1989; Yu et

146 al. 2015). In addition to providing frameworks from which species can be delimited from one

147 another, molecular phylogenies have become fundamental in disentangling biogeographic

148 relationships between taxa (Moritz 1995; Ree et al. 2005). The use of fossil data and molecular

149 clocks allow for phylogenetic approximations of divergence times which can then be correlated

150 with the timing of carbon-dated geo-climatic events. Indeed, many phylogenetic species

151 delimitation studies have made use of time-calibrated phylogenies in a geological context to

152 better understand the factors driving interspecific diversification (Kuriyama et al. 2011; Giribet

153 et al. 2012; Ceccarelli et al. 2014; Jongsma et al. 2018). Not only does phylogenetic

154 biogeographic inference allow for correlations to be made between evolutionary diversification

155 of species and historical geography, it also allows for the reconstruction of ancestral ecosystem

6 156 compositions through the identification of patterns of shared biogeographic affinities among

157 diverse biological taxa (Morley 2007; Linder et al. 2012). In addition to seeking to understand

158 the geological and climatic factors that underlie the evolutionary history of species in a

159 phylogeny, many studies make use of ecological niche modelling in an attempt to identify

160 unique bioclimatic preferences among closely related taxa which may have resulted in

161 ecological partitioning and subsequently speciation between taxa (Raxworthy et al. 2007; Blair

162 et al. 2013; da Silva and Tolley 2017). This approach has grown in popularity with the

163 implementation of the programme MaxEnt, which is presumably able to reliably model suitable

164 climate while making use of presence-only data (Phillips and Dudík 2008; Elith et al. 2011).

165

166 Cryptic diversity appears ubiquitous across the kingdoms of life with studies revealing its

167 persistence in crustaceans, spiders, bats, moths, and beetles among others (Mayer and von

168 Helversen 2001; Hendrixson and Bond 2005; Monaghan et al. 2005; Lefébure et al. 2006;

169 Dumas et al. 2015). The herpetofauna (reptiles and amphibians) are no exception, with much

170 documented cryptic diversity (see e.g. Shaffer et al. 2015; Vasconcelos et al. 2016). Within the

171 reptiles, the Chamaeleonidae are a relatively young lineage (ca. 90 my old) compared to other

172 squamate families. They are widespread across Africa and Madagascar, with a few species in

173 the Middle East, southern Europe, and southern Asia (Vidal and Hedges 2005; Wiens et al.

174 2012; Tolley et al. 2013). Previous studies have shown members of this family to exhibit

175 cryptic diversity, with documented examples in Rhampholeon, Bradypodion and Calumma

176 (Tolley et al. 2004; Gehring et al. 2012; Branch et al. 2014; Prötzel et al. 2015).

177

178 The genus Chamaeleo is a pan-African genus of chameleons that also includes all European,

179 Middle Eastern, and Asian species and was first described by Linneaus in 1758 (Tilbury 2018).

180 While the taxonomy within the Bradypodion, Rhampholeon and Calumma genera has been

181 somewhat resolved, for the most part through molecular means (Tolley et al. 2004; Branch et

7 182 al. 2014; Prötzel et al. 2015, 2017), there remains a paucity of phylogenetic work on the genus

183 Chamaeleo. As far as African chameleon genera are concerned, Chamaeleo tend to occupy

184 more xeric environments; most species prefer savannah and savannah-woodland, with a few

185 species tolerating semi-desert and one species (C. namaquensis) occupying true desert (Tilbury

186 2018). While a few Chamaeleo species will venture into dense forest, the genus does not appear

187 to have a clear affinity for forest and the early Eocene presence of forest (Morley 2007; Kissling

188 et al. 2012) more likely fragmented, rather than connected, their populations. Where other

189 chameleon genera may have experienced episodes of habitat fragmentation during arid periods

190 of the Miocene (i.e. Trioceros – see Ceccarelli et al. 2014), these episodes of aridity would

191 have likely reduced forest to savannah and thereby facilitated connectivity for Chamaeleo. The

192 recent taxonomic elevation of the sub-genus Trioceros (previously classified under

193 Chamaeleo) to genus level hints at the lack of evolutionary clarity within these chameleon taxa

194 (Tilbury and Tolley 2009).

195

196 A case in hand may be the common flap-necked chameleon (Chamaeleo dilepis Leach 1819).

197 This species is a large, typical chameleon widespread throughout sub-Saharan Africa (Largen

198 and Spawls 2006; Reaney et al. 2012; Tilbury 2018). The large range of this species (arguably

199 the largest of any chameleon species), taken with the confounding phenotypic variation that

200 exists across its range, renders the current taxonomy questionable (Largen and Spawls 2006;

201 Tilbury 2018). To further add to the taxonomic confusion surrounding this species, a recent

202 preliminary study found potential species level genetic distances between populations within

203 this species and suggested that the species may in fact be a complex of at least two distinct

204 species (Main et al. 2018). However, Main and co-workers only sampled populations within

205 South Africa so it is possible that there may be even more distinct lineages within this species

206 across its entire range. Clearly, an in-depth phylogeny of this group, placed within the context

207 of the entire genus, is needed if its taxonomy is to be resolved. Such would allow for the

8 208 implementation of conservation management actions that preserve all lineages and thereby

209 prevent a loss of unrecognized genetic diversity.

210

211 Here, in chapter two I use two mitochondrial (ND4 and 16S) and two nuclear (PRLR and

212 RAG1) markers to construct a comprehensive phylogeny of the genus Chamaeleo to address

213 the taxonomic ambiguity that currently plagues the group. Furthermore, I make use of various

214 species delimitation models to identify potential cryptic speciation within the genus. In chapter

215 three, I estimate a time-calibrated phylogeny using a secondary calibration point inferred from

216 Tolley et al. (2013)’s large-scale fossil-calibrated phylogeny of the Chamaeleonidae and using

217 this time-calibrated phylogeny I carry out an ancestral area reconstruction of the group

218 according to the dispersal-extinction cladogenesis (DEC) model implemented in RASP in order

219 to test whether Chamaeleo has Southern African origins. In addition, I perform ecological

220 niche modelling on different cryptic species within the C. dilepis complex determined in

221 Chapter 2 to examine whether there exists ecological support for these species.

222

223 REFERENCES:

224 Avise JC, Robinson TJ. 2008. Hemiplasy: a new term in the lexicon of phylogenetics.

225 Systematic Biology 57: 503–507.

226 Bickford D, Lohman DJ, Sodhi NS, Ng PKL, Meier R, Winker K, Ingram KK, Das I. 2007.

227 Cryptic species as a window on diversity and conservation. Trends in Ecology and

228 Evolution 22: 148–155.

229 Blair ME, Sterling EJ, Dusch M, Raxworthy CJ, Pearson RG. 2013. Ecological divergences

230 and speciation between lemur (Eulemur) sister species in Madagascar. Journal of

231 Evolutionary Biology 26: 1790-1801.

232 Bottrill MC, Joseph LN, Carwardine J, Bode M, Cook C, Game ET, Grantham H, Kark S,

233 Linke S, McDonald-Madden E, Pressey RL, Walker S, Wilson KA, Possingham HP.

9 234 2008. Is conservation triage just smart decision making? Trends in Ecology and

235 Evolution 23: 649–654.

236 Branch WR, Bayliss J, Tolley KA. 2014. Pygmy chameleons of the Rhampholeon platyceps

237 complex (Squamata: Chamaeleonidae): Description of four new species from isolated

238 ‘sky islands’ of Northern Mozambique. Zootaxa 3814: 1–36.

239 Brooks DR, McLennan D. a. 1999. Species: turning a conundrum into a research program.

240 Journal of nematology 31: 117–33.

241 Carstens BC, Pelletier TA, Reid NM, Satler JD. 2013. How to fail at species delimitation.

242 Molecular Ecology 22: 4369–4383.

243 Ceccarelli FS, Menegon M, Tolley KA, Tilbury CR, Gower DJ, Laserna MH, Kasahun R,

244 Rodriguez-Prieto A, Hagmann R, Loader SP. 2014. Evolutionary relationships, species

245 delimitation and biogeography of Eastern Afromontane horned chameleons

246 (Chamaeleonidae: Trioceros). Molecular Phylogenetics and Evolution 80: 125–136.

247 Cracraft J. 1975. Historical Biogeography and Earth History: Perspectives for a Future

248 Synthesis. Annals of the Missouri botanical Garden 62: 227–250.

249 Crandall KA, Bininda-Emonds ORR, Mace GM, Wayne RK. 2000. Considering evolutionary

250 processes in conservation biology. Trends in Ecology and Evolution 15: 290–295.

251 Darwin C. 1859. On the origin of the species by means of natural selection. London: John

252 Murray.

253 Degnan JH, Rosenberg NA. 2009. Gene tree discordance, phylogenetic inference and the

254 multispecies coalescent. Trends in Ecology and Evolution 24: 332–340.

255 Dumas P, Barbut J, Ru B Le, Silvain JF, Clamens AL, D’Alençon E, Kergoat GJ. 2015.

256 Phylogenetic molecular species delimitations unravel potential new species in the pest

257 genus spodoptera guenée, 1852 (Lepidoptera, Noctuidae). PLoS ONE 10: 1–20.

258 Elith J, Phillips SJ, Hastie T, Dudík M, Chee YE, Yates CJ. 2011. A statistical explanation of

259 MaxEnt for ecologists. Diversity and Distributions 17: 43–57.

10 260 Edwards SV. 2009. Is a new and general theory of molecular systematics emerging?

261 Evolution 63: 1–19.

262 Fujita MK, Leaché AD, Burbrink FT, McGuire JA, Moritz C. 2012. Coalescent-based species

263 delimitation in an integrative taxonomy. Trends in Ecology and Evolution 27: 480–488.

264 Garnett ST, Christidis L. 2017. Taxonomy anarchy hampers conservation. Nature 546: 25–

265 27.

266 Gehring PS, Tolley KA, Eckhardt FS, Townsend TM, Ziegler T, Ratsoavina F, Glaw F,

267 Vences M. 2012. Hiding deep in the trees: Discovery of divergent mitochondrial

268 lineages in malagasy chameleons of the Calumma nasutum group. Ecology and

269 Evolution 2: 1468–1479.

270 Giribet G, Sharma PP, Benavides LR, Boyer SL, Clouse RM, Bivort BLDE, Dimitrov D,

271 Kawauchi GY, Murienne J, Schwendinger PJ. 2012. Evolutionary and biogeographical

272 history of an ancient and global group of arachnids (Arachnida: Opiliones:

273 Cyphophthalmi) with a new taxonomic arrangement. Biological Journal of the Linnean

274 Society 105: 92–130.

275 Grant PR, Grant BR, Makert JA, Keller LF, Petren K. 2004. Convergent evolution of

276 Darwin’s finches caused by introgressive hybridization and selection. Evolution 58:

277 1588–1599.

278 Hendrixson BE, Bond JE. 2005. Testing species boundaries in the Antrodiaetus unicolor

279 complex (Araneae: Mygalomorphae: Antrodiaetidae): ‘Paraphyly’ and cryptic diversity.

280 Molecular Phylogenetics and Evolution 36: 405–416.

281 Isaac NJB, Mallet J, Mace GM. 2004. Taxonomic inflation: Its influence on macroecology

282 and conservation. Trends in Ecology and Evolution 19: 464–469.

283 Jongsma GFM, Barej MF, Barratt CD, Burger M, Conradie W, Ernst R, Greenbaum E,

284 Hirschfeld M, Leaché AD, Penner J, Portik DM, Zassi-Boulou AG, Rödel MO,

285 Blackburn DC. 2018. Diversity and biogeography of frogs in the genus Amnirana

11 286 (Anura: Ranidae) across sub-Saharan Africa. Molecular Phylogenetics and Evolution

287 120: 274–285.

288 Kapli P, Lutteropp S, Zhang J, Kobert K, Pavlidis P, Stamatakis A, Flouri T. 2017. Multi-rate

289 Poisson tree processes for single-locus species delimitation under maximum likelihood

290 and Markov chain Monte Carlo. Bioinformatics 33: 1630–1638.

291 Kissling WD, Eiserhardt WL, Baker WJ, Borchsenius F, Couvreur TLP, Balslev H, Svenning

292 J-C. 2012. Cenozoic imprints on the phylogenetic structure of palm species assemblages

293 worldwide. Proceedings of the National Academy of Sciences 109: 7379–7384.

294 Kuriyama T, Brandley MC, Katayama A, Mori A, Honda M, Hasegawa M. 2011. A time-

295 calibrated phylogenetic approach to assessing the phylogeography, colonization history

296 and phenotypic evolution of snakes in the Japanese Izu Islands. Journal of

297 Biogeography 38: 259–271.

298 Largen MJ, Spawls S. 2006. Lizards of Ethiopia (Reptilia Sauria): An annotated checklist,

299 bibliography, gazetteer and identification key. Tropical Zoology 19: 21–109.

300 Leaché AD, Zhu T, Rannala B, Yang Z. 2018. The Spectre of Too Many Species. Systematic

301 Biology.

302 Lefébure T, Douady CJ, Gouy M, Gibert J. 2006. Relationship between morphological

303 taxonomy and molecular divergence within Crustacea: Proposal of a molecular

304 threshold to help species delimitation. Molecular Phylogenetics and Evolution 40: 435–

305 447.

306 Linder HP, de Klerk HM, Born J, Burgess ND, Fjeldså J, Rahbek C. 2012. The partitioning

307 of Africa: Statistically defined biogeographical regions in sub-Saharan Africa. Journal

308 of Biogeography 39: 1189–1205.

309 Linnaeus C. 1735. Systema naturæ (10th edn). Stockholm: Laurentii Salvii.

310 Losos JB, Jackman TR, Larson A, de Queiroz K, Rodrıguez-Schettino L. 1998. Contingency

311 and determinism in replicated adaptive radiations of island lizards. Science 279: 2115–

12 312 2118.

313 Losos JB, Ricklefs RE. 2009. Adaptation and diversification on islands. Nature 457: 830–6.

314 Main DC, Jansen van Vuuren B, Tolley KA. 2018. Cryptic diversity in the common flap-

315 necked chameleon Chamaeleo dilepis in South Africa. African Zoology 53: 11–16.

316 Mallet J. 1995. A species definition for the modern synthesis. Trends in Ecology and

317 Evolution Evolution 10: 294–299.

318 Mayer F, von Helversen O. 2001. Cryptic diversity in European bats. Proceedings of the

319 Royal Society B: Biological Sciences 268: 1825–32.

320 Mayr E. 1942. Systematics and the origin of species, from the viewpoint of a zoologist.

321 Cambridge: Harvard University Press.

322 Mishler BD, Wilkins JS. 2018. The Hunting of the SNaRC: A Snarky Solution to the Species

323 Problem. Philosophy, Theory, and Practice in Biology 10.

324 Monaghan MT, Balke M, Gregory TR, Vogler AP. 2005. DNA-based species delineation in

325 tropical beetles using mitochondrial and nuclear markers. Philosophical transactions of

326 the Royal Society of London. Series B, Biological sciences 360: 1925–1933.

327 Monaghan MT, Balke M, Pons J, Vogler AP. 2006. Beyond barcodes: complex DNA

328 taxonomy of a South Pacific Island radiation. Proceedings of the Royal Society B:

329 Biological Sciences 273: 887–893.

330 Moritz C. 1994. Defining ‘evolutionarily significant units’ for conservation. Trends in

331 Ecology and Evolution 9: 373–375.

332 Moritz C. 1995. Uses of molecular phylogenies for conservation. Philosophical Transactions

333 of the Royal Society of London Series B: Biological Sciences 349: 113–118.

334 Morley RJ. 2007. Cretaceous and Tertiary climate change and the past distribution of

335 megathermal forests. In: Tropical Rainforest responses to Climate Change. Springer

336 Praxis Books. Berlin, Heidelberg: Springer. pp. 1–34.

337 Mossel E, Roch S. 2007. Incomplete lineage sorting: consistent phylogeny estimation from

13 338 multiple loci. IEEE/ACM transactions on computational biology and bioinformatics

339 7(1):166-171.

340 Nelson G. 1985. A decade of challenge the future of biogeography. Plate Tectonics and

341 Biogeography 4: 187–196.

342 Pfennig DW, Wund MA, Snell-Rood EC, Cruickshank T, Schlichting CD, Moczek AP. 2010.

343 Phenotypic plasticity’s impacts on diversification and speciation. Trends in Ecology and

344 Evolution 25: 459–467.

345 Phillips SJ, Dudík M. 2008. Modeling of species distributions with Maxent: new extensions

346 and a comprehensive evaluation. Ecography 31: 161–175.

347 Pons J, Barraclough TG, Gomez-Zurita J, Cardoso A, Duran DP, Hazell S, Kamoun S,

348 Sumlin WD, Vogler AP. 2006. Sequence-Based Species Delimitation for the DNA

349 Taxonomy of Undescribed Insects. Systematic Biology 55: 595–609.

350 Prötzel D, Ruthensteiner B, Scherz MD, Glaw F. 2015. Systematic revision of the Malagasy

351 chameleons Calumma boettgeri and C. linotum (Squamata: Chamaeleonidae). Zootaxa

352 4048: 211–231.

353 Prötzel D, Vences M, Scherz MD, Vieites DR, Glaw F. 2017. Splitting and lumping: An

354 integrative taxonomic assessment of Malagasy chameleons in the Calumma guibei

355 complex results in the new species C.gehringi sp. nov. Vertebrate Zoology 67: 231–249.

356 Puillandre N, Modica M V., Zhang Y, Sirovich L, Boisselier MC, Cruaud C, Holford M,

357 Samadi S. 2012. Large-scale species delimitation method for hyperdiverse groups.

358 Molecular Ecology 21: 2671–2691.

359 De Queiroz K. 2005. Ernst Mayr and the modern concept of species. Proceedings of the

360 National Academy of Sciences 102: 6600–6607.

361 De Queiroz K. 2007. Species concepts and species delimitation. Systematic Biology 56: 879–

362 886.

363 Raxworthy CJ, Ingram CM, Rabibisoa N, Pearson RG. 2007. Applications of ecological

14 364 niche modeling for species delimitation: A review and empirical evaluation using day

365 geckos (Phelsuma) from Madagascar. Systematic Biology 56(6): 907-923

366 Rannala B, Yang Z. 2003. Bayes estimation of species divergence times and ancestral

367 population sizes using DNA sequences from multiple loci. Genetics 164: 1645–1656.

368 Reaney LT, Yee S, Losos JB, Whiting MJ. 2012. Ecology of the flap-necked chameleon

369 Chameleo dilepis in Southern Africa. Brevoria: 1–18.

370 Ree RH, Moore BR, Webb CO, Donoghue MJ. 2005. A likelihood framework for inferring

371 the evolution of geographic range on phylogenetic trees. Evolution 59: 2299–2311.

372 Reid NM, Carstens BC. 2012. Phylogenetic estimation error can decrease the accuracy of

373 species delimitation: a Bayesian implementation of the general mixed Yule-coalescent

374 model. BMC evolutionary biology 12: 196.

375 Rose MR, Oakley TH. 2007. The new biology: beyond the Modern Synthesis. Biology Direct

376 2: 30.

377 Seehausen O. 2004. Hybridization and adaptive radiation. Trends in Ecology and Evolution

378 19: 198–207.

379 Seehausen O. 2006. African cichlid fish: a model system in adaptive radiation research.

380 Proceedings of the Royal Society B: Biological Sciences 273: 1987–1998.

381 Shaffer HB, Gidiş M, McCartney-Melstad E, Neal KM, Oyamaguchi HM, Tellez M,

382 Toffelmier EM. 2015. Conservation genetics and genomics of amphibians and reptiles.

383 Annual Review of Animal Biosciences 3: 113–38.

384 da Silva J, Tolley KA. 2017. Diversification through ecological opportunity in dwarf

385 chameleons. Journal of Biogeography 44(4): 834-847.

386 Smith CH. 1989. Historical biogeography: Geography as evolution, evolution as geography.

387 New Zealand Journal of Zoology 16: 773–785.

388 Tilbury C. 2018. Chameleons of Africa: an atlas: including the chameleons of Europe, the

389 Middle East and Asia (Illustrate). Frankfurt: Edition Chimaira.

15 390 Tilbury CR, Tolley KA. 2009. A re-appraisal of the systematics of the African genus

391 Chamaeleo (Reptilia: Chamaeleonidae). Zootaxa 2079: 57–68.

392 Tolley KA, Tilbury CR, Branch WR, Matthee CA. 2004. Phylogenetics of the southern

393 African dwarf chameleons, Bradypodion (Squamata: Chamaeleonidae). Molecular

394 Phylogenetics and Evolution 30: 354–365.

395 Tolley KA, Townsend TM, Vences M. 2013. Large-scale phylogeny of chameleons suggests

396 African origins and Eocene diversification. Proceedings of the Royal Society B:

397 Biological Sciences 280: https://doi.org/10.1098/rspb.2013.0184.

398 Vasconcelos R, Montero-Mendieta S, Simó-Riudalbas M, Sindaco R, Santos X, Fasola M,

399 Llorente G, Razzetti E, Carranza S. 2016. Unexpectedly high levels of cryptic diversity

400 uncovered by a complete DNA barcoding of reptiles of the socotra archipelago. PLoS

401 ONE 11: 1–19.

402 Vidal N, Hedges SB. 2005. The phylogeny of squamate reptiles (lizards, snakes, and

403 amphisbaenians) inferred from nine nuclear protein-coding genes. Comptes Rendus -

404 Biologies 328: 1000–1008.

405 Wagner CE, Keller I, Wittwer S, Selz OM, Mwaiko S, Greuter L, Sivasundar A, Seehausen

406 O. 2012. Genome-wide RAD sequence data provide unprecedented resolution of species

407 boundaries and relationships in the Lake Victoria cichlid adaptive radiation. Molecular

408 Ecology 22: 787–798.

409 Wiens JJ, Hutter CR, Mulcahy DG, Noonan BP, Townsend TM, Sites JW, Reeder TW. 2012.

410 Resolving the phylogeny of lizards and snakes (squamata) with extensive sampling of

411 genes and species. Biology Letters 8: 1043–1046.

412 Xu B, Yang Z. 2016. Challenges in species tree estimation under the multispecies coalescent

413 model. Genetics 204: 1353–1368.

414 Yang Z. 2015. A tutorial of BPP for species tree estimation and species delimitation. Current

415 Zoology 61: 854–865.

16 416 Yang Z, Rannala B. 2010. Bayesian species delimitation using multilocus sequence data.

417 Genetics 107: 9264–9269.

418 Yang Z, Rannala B. 2012. Molecular phylogenetics: Principles and practice. Nature Reviews

419 Genetics 13: 303–314.

420 Yang Z, Rannala B. 2014. Unguided species delimitation using DNA sequence data from

421 multiple loci. Journal of Molecular Biology and Evolution 31: 3125–3135.

422 Yu Y, Harris AJ, Blair C, He X. 2015. RASP (Reconstruct Ancestral State in Phylogenies): A

423 tool for historical biogeography. Molecular Phylogenetics and Evolution 87: 46–49.

424 Zachos FE. 2018. (New) Species concepts, species delimitation and the inherent limitations

425 of taxonomy. Journal of Genetics 97(4): 811-815.

426 Zachos FE, Apollonio M, Bärmann E V., Festa-Bianchet M, Göhlich U, Habel JC, Haring E,

427 Kruckenhauser L, Lovari S, McDevitt AD, Pertoldi C, Rössner GE, Sánchez-Villagra

428 MR, Scandura M, Suchentrunk F. 2013. Species inflation and taxonomic artefacts-A

429 critical comment on recent trends in mammalian classification. Mammalian Biology 78:

430 1–6.

17 431 Chapter 2

432

433 Genus-wide phylogeny and species delimitation within Chamaeleo

434

435

436 ABSTRACT:

437 The availability of new methods with which to delimit species is fast uncovering cryptic

438 diversity throughout the kingdoms of life, and repeatedly highlighting the gaps in traditional

439 morphological taxonomy. An unfortunate consequence of the limited reliability of

440 morphological taxonomy is that many cryptic species go unrecognised. Chamaeleo is a genus

441 of chameleons comprising 14 species with large, overlapping distributions and characterized

442 by considerable phenotypic variation. It is thought to have substantial cryptic diversity and has

443 long been in need of a taxonomic revision informed by molecular data. Accordingly, the aims

444 of the present study were to carry out a comprehensive phylogeny of the genus, accompanied

445 by species delimitation methods. Two mitochondrial (ND4 and 16S) and two nuclear (RAG1

446 and PRLR) markers were sequenced for representative individuals of every known species in

447 the genus, spanning their natural ranges where possible. Bayesian and maximum likelihood

448 analyses were carried out using the mitochondrial dataset, as well as the combined data. Species

449 delimitation was investigated using the Bayesian General Mixed Yule-Coalescent model,

450 distance-based DNA barcoding, and the Bayesian Phylogenetics and Phylogeography

451 programme. The analyses suggested the presence of 18 species within Chamaeleo, splitting C.

452 dilepis and C. gracilis into three candidate species each, and C. anchietae into two candidate

453 species. Additionally, the results suggest that C. necasi should be synonymised with one of the

454 candidate species of C. gracilis. These findings lay a phylogenetic foundation for a taxonomic

455 re-evaluation of the genus and hold implications for the conservation of the group at large.

18 456 INTRODUCTION:

457 Both the concept of species, and how they are delineated, have evolved considerably over the

458 past century with a general change in opinion from species being immutable units of

459 classification to species being transient, spatio-temporal units of evolutionary change (Brooks

460 and McLennan 1999; De Queiroz 2005, 2007). Notwithstanding, the species conundrum still

461 attracts much debate, with some even suggesting that species concepts should be done away

462 with altogether (Mishler and Wilkins 2018; Zachos 2018). While this debate pertains to the

463 processes underlying speciation events and the exact definition of species, in line with the

464 general lineage concept of species, there is a prevailing consensus that isolated metapopulation

465 lineages accumulate genetic differences that are evolutionarily tangible (De Queiroz 2007).

466 Given the evolutionary timescale of speciation events, most modern species delimitation

467 strategies make inferences on the basis of phylogenetic relationships between taxa and attempt

468 to account for lineage-specific divergence rates and molecular clock-informed divergence time

469 estimates (Moritz 1994; Crandall et al. 2000; Yang and Rannala 2010; Fujita et al. 2012).

470

471 Due to its ease of availability, single locus data such as mitochondrial DNA (mtDNA) is often

472 used to delimit species, this initially being used as the benchmark marker through distance-

473 based DNA barcoding (Lefébure et al. 2006; Puillandre et al. 2012). DNA barcoding has since

474 uncovered many cryptic lineages within single ‘morphospecies’ (see e.g. BinaPerl et al. 2014;

475 Vasconcelos et al. 2016; Chambers and Hebert 2016). However, there are shortfalls associated

476 with delimiting species purely on the basis of the genetic distances that separate taxa. In

477 particular, barcoding (or the distance-based approach underpinning barcoding) does not

478 operate within an evolutionary framework, and the rate of mtDNA evolution is heterogeneous

479 across the kingdoms of life – so in some taxa barcoding gaps are not effective for delimiting

480 species (Galtier et al. 2009; Puillandre et al. 2012). Many studies have also uncovered cryptic

481 diversity using evolutionary relationships inferred from actual phylogenies and coupled these

19 482 findings with distance-based thresholds (Hamilton et al. 2011; Main et al. 2018; Engelbrecht

483 et al. 2019). Given these criticisms, species delimitation methods for single-locus markers have

484 been developed that operate in a coalescent framework, and model changes in branching rates

485 that may indicate changes from a coalescent process to a Yule process (essentially identifying

486 changes between intra-specific and inter-specific processes). These include the General Mixed

487 Yule-Coalescent (GMYC) model (Pons et al. 2006; Fujisawa and Barraclough 2013) and its

488 Bayesian implementation bGMYC (Reid and Carstens 2012) as well as Multi-Rate Poisson

489 Tree Processes mPTP (Kapli et al. 2017). When only single-locus data are available, these

490 methods may be seen as the gold standard in species delimitation, however, when using only a

491 single locus to infer species, ancestral polymorphism, incomplete lineage sorting, and species-

492 tree-gene-tree discordance can all potentially confound results. To account for these factors,

493 methods that make use of the Multispecies Coalescent (MSC) on multilocus data, such as the

494 programme Bayesian Phylogenetics and Phylogeography (BPP), are able to account for gene

495 tree uncertainty by inferring a species tree from multiple gene trees and exploring different tree

496 topologies in a Bayesian reversible jump Markov Chain Monte Carlo (rjMCMC) framework

497 (Yang and Rannala 2014; Yang 2015).

498

499 Chamaeleo is a pan-African genus of chameleons with some members also occurring in

500 Southern Europe, Southern Asia and the Middle East (Tilbury 2018). This genus is comprised

501 of 14 species (Uetz et al. 2019), however, many of these species have large, overlapping

502 distributions with confusing phenotypic variation within and between them rendering their

503 current taxonomy somewhat questionable (Largen and Spawls 2006; Main et al. 2018; Tilbury

504 2018). Indeed, until 2009 the genus included the sub-genus Trioceros which was subsequently

505 elevated to its own genus, highlighting the need for a taxonomic revision of the group with

506 possible conservation implications (Tilbury and Tolley 2009). Certain species of Chamaeleo

507 in particular, which all broadly resemble one another, present a real quandary to taxonomy: C.

20 508 anchietae, C. dilepis, C. gracilis, C. laevigatus, C. necasi and C. senegalensis. Chamaeleo

509 anchietae encompasses three subspecies, each with small, isolated distributions (Table 1 and

510 Fig 1; also see Tilbury 2018 and Uetz et al. 2019).

511 Figure 1: Type localities (inferred from vague locality names) of the species and subspecies of Chamaeleo that closely 512 resemble one another morphologically and have overlapping distributions (from Table 1).

513 The taxonomy of C. dilepis has repeatedly confused taxonomists since the species was first

514 described by Leach in 1819 (see Tilbury 2018), probably because its range is larger than that

515 of any other chameleon species and it exhibits perplexing phenotypic variation across this

516 range. It currently comprises eight subspecies, each of which have vaguely understood

517 distributions and in some cases imprecise type localities with substantial distributional overlap

518 between them (Table 1 and Fig 1). Additionally, phenotypic differences between subspecies

519 are not always obvious (see Tilbury 2018). Chamaeleo gracilis is a tropical savannah inhabitant

520 with one subspecies described in addition to the type species; these two variants of C. gracilis

21 521 have overlapping distributions with subspecies of C. dilepis, C. senegalensis and C. necasi.

522 Chamaeleo laevigatus is a species whose taxonomy has also presented a challenge, with debate

523 as to whether it is in fact its own species or whether it is just a variant of Chamaeleo

524 senegalensis (Joger 1990; Chirio and LeBreton 2007; Trape et al 2012; but see Tilbury 2018

525 for a review). Chamaeleo necasi is a recently described species (Ullenbruch, Krause and

526 Bohme, 2007) with a vague type locality and limited information pertaining to the extent of its

527 geographical range (Tilbury 2018). Additionally, C. necasi resembles C. gracilis

528 morphologically and while the type locality is vague for C. necasi (Table 1), it appears likely

529 that its range overlaps both that of C. gracilis and C. senegalensis. Chamaeleo senegalensis

530 has been suggested to include C. laevigatus, its range also overlaps with both that of C. gracilis

531 and C. necasi and it morphologically resembles that of C. gracilis (Tilbury 2018). It is clear

532 that the taxonomy of the abovementioned species is in need of a phylogenetically informed

533 revision, given that morphology appears confounding within this genus.

534 Table 1: Type localities and authorities of species and subspecies of Chamaeleo that closely resemble one another 535 morphologically and have overlapping distributions. Localities are literal from the original descriptions. DRC = Democratic 536 Republic of Congo.

Species Species Subspecies Subspecies Type locality authority authority Chamaeleo anchietae Bocage 1872 anchietae Bocage 1872 Huila, near Mossamedes, Angola marunguensis Laurent 1952 West of Lake Tanganyika, DRC mertensi Laurent 1952 Muneshele, Kivu, DRC vinckei Laurent 1952 Katanga, DRC Chamaeleo dilepis Leach 1819 dilepis Leach 1819 “Congo” idjwiensis Loveridge 1942 Idjwi Island, Lake Kivu, DRC isabellinus Gunther 1893 Shire Islands South of Lake Nyassa, Malawi martensi Mertens 1964 Pemba Island, Tanzania petersii Gray 1865 Mozambique quilensis Bocage 1866 Rio Quilo, north of Cabinda, Angola roperi Boulenger 1890 Kilifi, north of Mombasa, Kenya ruspolii Boettger 1893 Ogaden, Somaliland Chamaeleo gracilis Hallowell 1844 gracilis Hallowell 1844 Liberia etiennei Schmidt 1919 Banana, DRC Chamaeleo necasi Ullenbruch, Krause and NA Togo Bohme 2007 Chamaeleo laevigatus Gray 1863 NA 500 miles south of Khartoum, Sudan Chamaeleo senegalensis Daudin 1802 NA Senegal

22 537 While 30% of all chameleon species are currently listed as threatened by the International

538 Union for Conservation of Nature (IUCN), no Chamaeleo species fall within threatened

539 categories (IUCN, 2017). The majority of threatened chameleon taxa are narrow endemics and

540 depend heavily on intact forest, making deforestation a very real risk to their status in the wild.

541 In contrast, the affinity of most Chamaeleo species for savannah-type habitats, coupled with

542 their moderate to large distributions and tolerance of modified habitats, might mean that the

543 threat of habitat loss is less pervasive to them than other chameleon species. In contrast,

544 Chamaeleo species make up some of the most heavily exported species for the pet trade

545 (Jenkins et al. 2014). The Convention for International Trade in Endangered Species of Fauna

546 and Flora (CITES) provides lists of species for which international trade is permitted at

547 regulated levels, and a list of species whose trade is prohibited; the assignment of species to

548 these lists is in part dependent on their threat status. Because CITES export quotas are set per

549 species, the future of potential unrecognised cryptic species of Chamaeleo hinges on an

550 accurate species inventory of the genus. If species that are currently understood to have

551 widespread distributions are found to comprise multiple species with narrow distributions then

552 the current export quota for these species might be unsustainable.

553

554 Given that a preliminary study with limited sampling pointed to cryptic diversity and

555 potentially multiple species within Chamaeleo dilepis (Main et al. 2018), cryptic lineages are

556 expected in the widespread species of Chamaeleo. To examine this, a comprehensive

557 phylogeny of the genus Chamaeleo was constructed, including multiple samples from all

558 known species. Species delimitation methods were used to examine the potential for

559 unrecognised cryptic diversity which point to candidate species, and to examine the validity of

560 existing species.

23 561 MATERIALS AND METHODS:

562 Laboratory Protocols:

563 Multiple samples representing all currently known species of Chamaeleo were acquired from

564 the South African National Biodiversity Institute (SANBI) tissue bank. In some cases, multiple

565 samples were available from across the range of species. Sequence data for Chamaeleo

566 specimens of known provenance were also downloaded from GenBank, including multiple

567 sequences of C. dilepis from Main et al. (2018). In addition, sequence data for a representative

568 of an outgroup genus (Trioceros) were downloaded, with the final dataset comprising 56

569 individuals (Appendix I).

570 Figure 2: Sampling localities for all Chamaeleo taxa in this study.

24 571 Total genomic DNA was extracted from tail clips or liver samples following standard salt

572 extraction (Aljanabi and Martinez 1997) or Cetyl Trimethyl Ammonium Bromide (CTAB)

573 (Allen et al. 2006) extraction protocols. The mitochondrial NADH subunit 4 gene (ND4), 16S

574 ribosomal subunit (16S), nuclear recombination activating gene 1 (RAG1) and nuclear

575 prolactin receptor gene (PRLR) were PCR amplified using the primers F3 and R4 for ND4

576 (Raxworthy et al. 2002), L2510 and H3080 for 16S (Romano and Palumbi 1997), F118 and

577 R1067 for RAG1 (Matthee et al. 2004), and ChamF1 and ChamR1 for PRLR (Tolley et al.

578 2013). Fragment amplification was carried out in 30 µL reaction volumes with 25-60 ng µL-1

579 genomic DNA, 10x thermophilic buffer (20 mM Tris-HCl [pH 8.0], 100 mM NaCl, 0.1mM

580 EDTA, 1mM DTT), 2.5 mM MgCl2, 0.25 µM of each primer, 0.2 mM dNTPs, and 1 Unit

581 Super-Therm Taq DNA polymerase. The thermocycling profile followed 4 minutes of initial

582 denaturation at 95 ºC followed by 38 cycles for ND4 and 35 cycles for 16S, RAG1 and PRLR

583 of 45 seconds at 94 ºC, 45 seconds for primer specific annealing, and 1 minute extension at 72

584 ºC, with a final extension of 10 minutes at 72 ºC. Annealing temperatures were 52-55 ºC for

585 ND4, 48-52 ºC for 16S and 57 ºC for RAG1 and PRLR. Amplifications were performed on a

586 GeneAmp PCR 2720 system (Applied Biosystems, Foster City, CA, USA). After verifying

587 successful amplifications on a 1% agarose gel, amplicons were cycle sequenced using Big Dye

588 chemistry and analysed on an ABI3730xl automated sequencer (Applied Biosystems).

589 Sequences were manually verified in Geneious 8.19 (Kearse et al. 2012) and aligned using the

590 Muscle plug-in in Geneious.

591

592 Phylogenetic Analyses:

593 Phylogenetic analyses were run for the mitochondrial only dataset and for the combined mt

594 and nuclear datasets (mt–ND4: 856 bp, 16S: 494 bp, nuclear–PRLR: 595 bp, RAG1: 943 bp).

595 The analyses were also run on separate gene trees (total mitochondrial, PRLR and RAG-1).

596 The genus Chamaeleo forms a polytomy with six other extant arboreal chameleon genera

25 597 (Tolley et al. 2013). Given this, a representative from one genus (Trioceros) was used as an

598 outgroup taxon (Appendix I). Bayesian analyses were implemented in MrBayes 3.2.2

599 (Huelsenbeck and Ronquist 2001) on the mitochondrial dataset and the combined dataset. For

600 the latter, a partition homogeneity test was carried out in PAUP 4.0a164 (Swofford 2003) that

601 confirmed topological congruence between gene trees (p=0.41) and suggested that the gene

602 fragments could be combined without conflict. The combined dataset was partitioned by

603 marker and the optimal model of evolution determined for each marker using jModelTest

604 (Darriba et al. 2012), and incorporated into the analyses (16S: GTR; ND4 GTR + I + G; PRLR

605 and RAG1: HKY + G). The mitochondrial only dataset was also partitioned according to the

606 same scheme as above.

607

608 Both analyses were run with a random starting tree and four independent chains of 50 million

609 generations, sampling every 1000 generations. Convergence and adequate mixing of chains

610 was assessed in Tracer 1.7 (Rambaut et al. 2018) and convergence was considered adequate

611 when ESS > 200 for all the parameters. The first 10% of trees were discarded as burn-in. Nodes

612 with posterior probabilities (pp) ≥ 0.90 were considered moderately-supported, while those

613 with > 0.95 pp were considered well-supported. In addition to the Bayesian analysis, a

614 maximum likelihood (ML) analysis was carried out using RaxML (Stamatakis 2006) on the

615 same two datasets, implementing the same partitioning scheme as in the Bayesian analysis with

616 the GTRGAMMA model of evolution and 1000 bootstrap replicates. Here, nodes that received

617 ≥ 75% were considered supported. All phylogenetic analyses were run on the CIPRES Science

618 Gateway (Miller et al. 2010; http://www.phylo. org/sub_sections/portal/).

619

620 Species Delimitation:

621 Due to the uncertainty of species delimitation protocols in accurately accounting for

622 evolutionary process, several analyses were carried out. Firstly, species were inferred using a

26 623 Bayesian implementation of the General Mixed Yule-Coalescent (bGMYC) model (Reid and

624 Carstens 2012), which allows for objective delimitation as it does not require a priori species

625 assignment. The GMYC, first implemented by Pons et al. (2006), lies at the interface between

626 population genetics and phylogenetics and attempts to determine the point of departure at

627 which coalescent processes (intraspecific) become Yule processes (interspecific). Pons et al.

628 (2006)’s implementation of the GMYC makes use of a single ultrametric tree, with the

629 shortcoming that it is unable to account for uncertainty in tree topology or phylogenetic error.

630 Reid and Carstens (2012)’s Bayesian implementation of the GMYC allows for integration over

631 multiple trees in an MCMC analysis, thereby accounting for both phylogenetic error and

632 uncertainty in tree topology. Bayesian GMYC (bGMYC) analyses were carried out in R using

633 the package bGMYC 3.0.1 (Reid and Carstens 2012) on the mtDNA and the combined datasets,

634 separately. For each dataset, BEAST 2 (Bouckaert et al. 2014) was first used to generate

635 ultrametric trees. For the combined dataset, the same partitioning scheme and substitution

636 models were used in BEAST as those used in the Bayesian and maximum likelihood analyses.

637 For each dataset, two independent runs of 100 million generations were sampled every 1000

638 generations and assessed in Tracer for adequate convergence and mixing (ESS ≥ 200). After

639 removing 50% burn-in, the trees from each of the runs were combined in LogCombiner

640 (Bouckaert et al. 2014) and the last 500 trees were saved for the bGMYC analysis. In the

641 bGMYC package, a random sample of 100 trees from the 500 trees generated with BEAST

642 was obtained and pruned to remove outgroup taxa. On this random sample of 100 trees, a

643 bGMYC Markov Chain Monte Carlo (MCMC) of 50000 generations was run with a burn-in

644 of 40000 generations and thinning set to every 100 to produce 1000 samples. Default priors

645 were used for the Yule and coalescent rate change parameters, the upper threshold was set to

646 30 and the starting value of the threshold parameter was set to 23.

27 647 The bGMYC method of species delimitation was developed for the use of single-locus data

648 and although it performs well for some multi-locus studies (Ceccarelli et al. 2014; Engelbrecht

649 et al. 2019), it has also been found to over-split lineages (Carstens et al. 2013). Two factors

650 that may confound the inference of species from single-locus data are ancestral polymorphism

651 and incomplete lineage sorting which may result in discordance between species trees and gene

652 trees. Therefore, candidate species were also inferred using the programme Bayesian

653 Phylogenetics and Phylogeography (BPP 3.4 - Yang 2015). BPP implements a Bayesian

654 multispecies coalescent approach on multilocus data to infer species. BPP was run on the

655 combined dataset using the joint species-delimitation and species-tree inference option (A11 –

656 Yang, 2015). This option uses the nearest neighbour interchange (NNI) and subtree pruning

657 and regrafting (SPR) algorithms to adjust tree topology and to explore species tree scenarios

658 given multiple gene trees, in so doing it does not rely on a fixed, user-specified guide tree and

659 is able to account for uncertainty among gene trees (Yang and Rannala 2014). Option A11 of

660 BPP may lump two species together, but it will not split pre-defined species. Species

661 delimitation with BPP is heavily dependent on the selection of priors. Two priors of interest

662 when running BPP are the ancestral population size (q) and the root age (t0). To ensure the

663 robustness of species inferences, multiple different combinations were tested for these priors,

664 modelling large and small ancestral population sizes as q ~ I(3, 0.2) and q ~ I(3, 0.002)

665 respectively and deep and shallow root ages as t0 ~I(3,0.2) and t0 ~I(3,0.002) respectively

666 (Yang and Rannala 2010, 2014; but also see Wüster et al. 2018). All BPP runs were carried out

667 for 100000 generations with a burn-in of 20000 generations, sampling every 2 generations.

668

669 Finally, a DNA barcoding method was used to test species inferences. DNA barcoding is a

670 distance-based approach to species identification that relies on the premise that between-

671 species genetic distances exceed within-species genetic distances by an order of magnitude.

672 This genetic distance threshold is known as a barcoding gap (Lefébure et al. 2006; Puillandre

28 673 et al. 2012). SpeciesIdentifier 1.8 (Meier et al. 2006) was used to infer a sequence divergence

674 threshold for species from frequency distributions of inter and intraspecific distances within

675 the Chamaeleo genus. Given that the 16S dataset was the most complete, and the only dataset

676 encompassing the full species complement for the genus, this dataset was used to infer species

677 with SpeciesIdentifier. The results from the two model based species delimitation analyses

678 were used to guide input for SpeciesIdentifier, in order to estimate genetic distances between

679 the candidate species inferred by those methods.

680

681 RESULTS:

682 Phylogenetic analyses:

683 Topologies of the mitochondrial and the combined dataset were not in conflict for either the

684 Bayesian or maximum likelihood analyses. The combined dataset resulted in higher nodal

685 support than the mitochondrial dataset for both analyses (Fig. 3; Fig. 1, Appendix II).

686 Chamaeleo forms a well-supported, monophyletic clade with C. namaquensis as the sister

687 taxon to all other Chamaeleo species. In general, deep nodes were well supported, as were the

688 nodes defining most currently described species. The split between C. dilepis/gracilis and C.

689 laevigatus/senegalensis clades has low bootstrap support (60%) but is supported by Bayesian

690 posterior probability. A number of the shallow nodes do not show full support by both analyses,

691 but these are all within the C. dilepis clade. Although most of the currently described species

692 are supported (i.e. 12 species), C. gracilis is paraphyletic and C. necasi is embedded within a

693 clade of C. gracilis.

29 694 Figure 3: A Bayesian consensus tree for Chamaeleo. Node support is indicated for both Bayesian and likelihood analyses. 30 695 Species delimitation:

696 The bGMYC and BPP analyses yielded identical results, both of which retrieved 18 species

697 within Chamaeleo, splitting the C. dilepis into three separate species, C. anchietae into two

698 species, and C. gracilis into three species (Fig. 4). Furthermore, C. necasi and C. gracilis clade

699 3 were retrieved as a single species. The results were the same for the mtDNA (not shown) and

700 the combined datasets, and the BPP yielded the same delimitations and posterior probability

701 support regardless of ancestral population size and root age prior.

702 Figure 4: Species delimitation results for three different methods (bGMYC, BPP, and genetic distance-based DNA barcoding) 703 for the Bayesian phylogeny produced from the combined dataset. Breaks in coloured bars indicate interspecific splits inferred 704 by each method. Chamaeleo necasi is indicated in bold and with a dark yellow bar nested within the light yellow Chamaeleo 705 gracilis 3; all methods lumped these two species together.

31 706 The DNA barcoding approach was largely congruent with the other methods, with the

707 exception of splitting the C. dilepis complex into five (rather than three) species (Fig. 5; those

708 to the right of the barcode code gap shown in grey), with agreement between all methods for

709 the presence of C. dilepis 1 and 2 as species, but barcoding split C. dilepis 3 into three species.

250

Intraspecific distances that exceed the barcode gap 200

150

100 Frequency pairwise differences

50

0 0 > 20.0% 0.0% to 0.5% to 0.0% 1.0% to 0.5% 1.5% to 1.0% 2.0% to 1.5% 2.5% to 2.0% 3.0% to 2.5% 3.5% to 3.0% 4.0% to 3.5% 4.5% to 4.0% 5.0% to 4.5% 5.5% to 5.0% 6.0% to 5.5% 6.5% to 6.0% 7.0% to 6.5% 7.5% to 7.0% 8.0% to 7.5% 8.5% to 8.0% 9.0% to 8.5% 9.5% to 9.0% 9.5% to 10.0% to 9.5% 10.0% to 10.5% to 10.0% 11.0% to 10.5% 11.5% to 11.0% 12.0% to 11.5% 12.5% to 12.0% 13.0% to 12.5% 13.5% to 13.0% 14.0% to 13.5% 15.5% to 15.0% 16.0% to 15.5% 16.5% to 16.0% 17.0% to 16.5% 17.5% to 17.0% 18.0% to 17.5% 18.5% to 18.0% 19.0% to 18.5% 19.5% to 19.0% 20.0% to 19.5% 14.0% to 14.49% to 14.0% 15.0% to 14.49%

Intraspecific Interspecific Sequence divergence for 16S (%)

710 Figure 5: Frequency histogram of pairwise differences between and within candidate species of Chamaeleo for 16S. Blue bars 711 denote intraspecific distances, and yellow indicate interspecific distances. The barcode gap is shown in grey. Intraspecific 712 distances that exceeded the barcode gap were for individuals within the Chamaeleo dilepis 3 clade (Fig. 4).

713 Chamaeleo dilepis 1 with a central African sampling distribution best matches the type locality

714 of C. dilepis dilepis (“Congo” which is most likely either the Democratic Republic of Congo,

715 or the Republic of Congo) and the clade is therefore considered to represent the nominate

716 species. The sampling area for the Chamaeleo dilepis 2 clade covers a large area in Southern

717 Africa whereas, the samples making up the C. dilepis 3 clade were all from East Africa (Fig.

718 6).

32 719 Figure 6: The presumed range of each Chamaeleo dilepis candidate species (coloured polygons) revealed by species 720 delimitation as well as all records of C. dilepis sensu lato (black dots) and sample sites included in the phylogeny (colour 721 symbols). The dark grey polygon represents the presumed distribution of C. dilepis (adapted from Tolley 2014).

722 For C. anchietae, the sampling locality of C. anchietae 1 (Huila Province, Angola) best

723 matches the type locality of the nominate species of C. anchietae (also Huila Province, Angola)

724 and the sampling locality of C. anchietae 2 (Kundelungu National Park, DRC) is closest the

725 type locality of C. anchietae vinckei (Katanga, DRC ; Table 1, Fig. 7). Chamaeleo anchietae

726 has also been recorded from Malawi, and although samples from that area were not included

727 in the phylogeny, the Malawian records are presumably from clade 2. In addition, there are a

728 number of C. anchietae records from Rwanda, however, given that the one Rwandan “C.

729 anchietae” included in the phylogeny was in the C. dilepis 3 clade, this puts the Rwandan

730 records in doubt (Fig. 3).

33 731 Figure 7: The presumed range of each Chamaeleo anchietae candidate species revealed by species delimitation (coloured 732 polygons) as well as all records from across the known range of C. anchietae (black dots) and sample sites included in the 733 phylogeny (coloured symbols). The dark grey polygons represent the recorded distribution of C. anchietae (Tolley et al. 2015).

734 The sampling localities of the three C. gracilis clades are separated by large geographic

735 distances, particularly as compared to its currently known distribution (Fig. 8). However, the

736 area covered by C. gracilis 3 is the closest match to the type locality of nominate C. gracilis

737 (“Liberia”; Fig. 8) with one sample originating in that country (Appendix I). The single sample

738 of C. necasi is embedded within C. gracilis 3 clade.

34 739 Figure 8: The presumed range of each Chamaeleo gracilis candidate species (coloured polygons) revealed by species 740 delimitation as well as all records from across the known range of C. gracilis (black dots) and sample sites included in the 741 phylogeny (colour symbols). The dark grey polygon represents the currently known distribution of C. gracilis (adapted from 742 Tolley et al. 2014).

743 Chamaeleo laevigatus and C. senegalensis are supported as separate species by each of the

744 delimitation methods (Fig. 3). Each of the sampling sites fall within the currently known

745 distribution of these two species (Fig. 9).

35 746 Figure 9: Sampling localities (coloured symbols) of Chamaeleo senegalensis and C. laevigatus used in this study as well as 747 all known records of C. senegalensis (white dots) and C. laevigatus (black dots). The currently known distribution for C. 748 senegalensis (dark grey polygon) and C. laevigatus (light grey polygon) are also shown (Tolley and Trape 2014; Wilms et al. 749 2013).

750 DISCUSSION:

751 The phylogenetic analysis complemented by model based species delimitation shows that

752 Chamaeleo likely comprises 18 species in contrast to the currently recognised 14 species. There

753 is cryptic diversity present in three of the currently recognised species, C. dilepis, C. gracilis,

754 and C. anchietae, while C. necasi does not appear to be a valid species (Fig. 3). In general,

755 other described species are well supported and monophyletic (Fig. 2).

756

757 Large genetic differences within C. dilepis were found across a limited area of its range in

758 South Africa, suggestive of species level differentiation (Main et al. 2018). However, that study

36 759 was not comprehensive, lacking most species and covering only a limited geographic range,

760 which prevented conclusions regarding species delimitation. In the present study, this

761 constraint was overcome by covering a greater range of the potential diversity within

762 Chamaeleo by including all species, and multiple samples from across the range of the more

763 widespread species. Indeed, cryptic diversity is certainly present within C. dilepis pointing to

764 three candidate species delimited here by several model based delimitation methods. These

765 candidate species are logical in terms of geography, with C. dilepis 1 found in Central Africa,

766 C. dilepis 2 for the most part found in Southern Africa and C. dilepis 3 found in East Africa

767 (Fig. 6). However, C. dilepis 1 and C. dilepis 2 appear to be parapatric near southern Angola,

768 southern DRC and northern Zambia. In the present phylogeny sampling is not comprehensive

769 enough in this region to discern where the transitions between these clades are on the landscape.

770

771 Chamaeleo dilepis 1 was inferred from multiple individuals from three African countries

772 (DRC, Republic of Congo, and Angola). Given that these individuals were sampled relatively

773 far apart from one another (approximately 1800 km between each sampling locality), more

774 comprehensive sampling along intermediate points between the sampling localities used in the

775 present study might improve our understanding of relationships within this candidate species.

776

777 The geographic range covered by the sampling of C. dilepis 2 is comparatively large,

778 encompassing much of South Africa and also extending into Mozambique, Malawi, Namibia,

779 Zambia, Zimbabwe, Botswana, southern Angola, and southern Tanzania. Main et al. (2018)

780 found relatively deep divergences within this same clade, and suggested that the differences

781 could be at the species level, although cautioned that firm conclusions could not be made due

782 to limited sampling. Here, the comprehensive species sampling and delimitation methods

783 instead indicate that differences within C. dilepis 2 are not at the species level. The lack of

784 shared haplotypes and large sequence divergences observed by Main et al. (2018) therefore

37 785 point to strong population level structure within C. dilepis 2. This new finding highlights the

786 importance of comprehensive sampling in inferring species from phylogenies as limited

787 sampling can confound such inferences.

788

789 Chamaeleo dilepis 3 is an East African clade, which includes individuals from Tanzania

790 (including Zanzibar), Kenya, and Rwanda. This is the only C. dilepis clade for which

791 congruency in species delimitation was not observed between all three methods. Distance-

792 based barcoding splits this clade into three separate species, however, both BPP and bGMYC

793 support this clade as a single species. Taking a conservative approach, guided by the

794 congruency between the more advanced delimitation methods, point to this clade comprising

795 a single species.

796

797 In Chamaeleo, cryptic diversity is not unique to C. dilepis with two candidate species

798 consistently delineated for C. anchietae across all methods. The distance separating C.

799 anchietae 1 from the most westerly sampled C. anchietae 2 is relatively large, in excess of

800 1500 km, and there are no additional records of C. anchietae in this region suggesting that the

801 gap is genuine. That is, C. anchietae 1 is found in Angola, whereas samples of C. anchietae 2

802 are from South-Eastern DRC and Tanzania. There is also a record of this species in northern

803 Malawi (Tilbury 2018), which is essentially midway between the two samples included in clade

804 2, suggesting that populations from Malawi should be assigned to clade 2. Furthermore, the

805 two areas have different climates and habitats, separated by an arid corridor (Beck et al. 2018),

806 and this could have promoted their diversification. However, these inferences are preliminary

807 as sample sizes are small and geographic coverage is incomplete. Furthermore, confirmation

808 that the intermediate area of the arid corridor does not contain populations of C. anchietae

809 sensu lato would be useful to understand these distributions.

38 810 Interestingly, an individual identified as C. anchietae (ZMFK 54962) from Rwanda was found

811 to group within the C. dilepis 3 clade (Fig. 6 and 7). This clarifies some taxonomic confusion

812 over the potential Rwandan distribution of this species that stemmed from the misidentification

813 of this specimen (Tilbury 2018). Given that this Rwandan individual is within the C. dilepis 3

814 clade, it seems likely that other C. anchietae records from Rwanda (Fig. 7) are actually C.

815 dilepis. However, more comprehensive sampling of chameleons identified as C. anchietae

816 from Rwanda would help clarify this situation.

817

818 Chamaeleo anchietae is morphologically similar to C. laevigatus and they are presumed to be

819 closely related (Tilbury 2018). However, this was not reflected in the phylogeny. Indeed, C.

820 anchietae is deeply divergent from most other Chamaeleo species. This is not a unique case of

821 discordance between morphology and genetics, in fact, many other studies have illustrated that

822 this is a common phenomenon (e.g. Graham et al. 1998; Hamidy et al. 2011; DeBiasse and

823 Hellberg 2015; Sharma et al. 2015). While morphology was once the sole basis for species

824 identification, it has become clear that when used in concert with other data (e.g. molecular,

825 ecological, physiological, behavioural etc.), there is greater confidence for inferences

826 pertaining to species delimitation.

827

828 Challenges with Chamaeleo classification are also evident within C. gracilis where three

829 candidate species have been identified. These results appear to make geographic sense,

830 although samples included were limited in geographic coverage, with C. gracilis 1 occurring

831 in a high elevation region of Tanzania, C. gracilis 2 in western Chad, and C. gracilis 3 across

832 Guinea, Sierra Leone and Liberia in West Africa. The sampling regions are separated by more

833 than 3000 km and have vastly different climates (Beck et al. 2018) suggesting that

834 diversification between the three clades could have been related to environmental drivers.

835 However, there are many records of C. gracilis from the intervening areas so it would be

39 836 important to examine this further through the inclusion of additional representative specimens

837 covering a broader geographic range. In addition to unveiling cryptic diversity in three species

838 of Chamaeleo, all delimitation methods support C. laevigatus and C. senegalensis as sister

839 species, but clearly separate. This is an important finding as Chirio and LeBreton (2007) (cited

840 by Tilbury, 2018) do not recognise C. laevigatus but rather consider it as variant of C.

841 senegalensis. Furthermore, the divergence between C. senegalensis and C. laevigatus is deeper

842 than that between many other sister species and this lends even more to the support of C.

843 laevigatus and C. senegalensis as separate species. A clear mismatch exists between

844 distribution records for C. senegalensis and C. laevigatus and their respective IUCN

845 distribution polygons (Fig. 9). A C. laevigatus sample from the present study (taken from

846 southern Chad) supports the IUCN distribution polygon for C. laevigatus, however, if the

847 records for C. senegalensis are valid, then there is extensive geographic overlap between C.

848 laevigatus and C. senegalensis which has probably perpetuated the confusion surrounding the

849 validity of these species. The IUCN presumed distribution polygon (Wilms et al. 2013) for C.

850 senegalensis is in need of a revision if the eastern-most records for C. senegalensis are accurate.

851

852 Taxonomy

853 The C. dilepis complex consists of three species-level clades which, pending a morphological

854 reassessment of specimens from these areas, should each be described as their own species.

855 Given that Chamaeleo dilepis 1 was sampled from across central Africa, and the type locality

856 is “Congo” (Table 1; Fig. 1), this clade probably represents the nominate species and would

857 retain the name Chamaeleo dilepis.

858

859 For both C. anchietae and C. gracilis there is cryptic diversity that is at odds with current

860 subspecific assignments. Chamaeleo anchietae 1 was sampled from Huila Province in Angola,

861 matching the type locality of the nominate species, making it possible to assign C. anchietae 1

40 862 to C. anchietae anchietae (Table 1; Fig 1). Chamaeleo anchietae 2 was sampled from both the

863 DRC and Tanzania and the individual from the DRC was sampled in Kundelungu National

864 Park which is near the type locality of C. anchietae vinckei (Katanga, DRC) (Table 1; Figs 1

865 & 7). Assuming that this pattern holds given an updated phylogeny with additional geographic

866 coverage, both subspecies should be elevated to specific status, and the subspecies rankings

867 should be abandoned. In this case, C. anchietae anchietae would retain the nominate name.

868 Given that Malawian chameleons identified as C. anchietae (see Tilbury 2018; Fig. 7) from

869 the Nyika Plateau fall geographically between both C. anchietae 2 sample sites from the

870 present study (Fig. 7), it seems plausible that populations from Malawi are also part of the C.

871 anchietae 2 clade. However, inclusion of Malawian C. anchietae samples in a phylogeny would

872 confirm this.

873

874 In the case of C. gracilis, the present study adds two potential candidate species. Chamaeleo

875 gracilis 3 from Liberia, Sierra Leone and Guinea represents C. gracilis given type locality is

876 “Liberia” (Figs 1 & 8). Additionally, given the nesting of C. necasi within this clade, C. necasi

877 should be synonymised with C. gracilis. Chamaeleo gracilis 1 and 2, while inferred from very

878 limited sampling, strongly suggest the presence of at least two additional species within C.

879 gracilis. Indeed, additional sampling with greater geographic coverage, might unveil additional

880 species, or at least lend stronger support to the presence of these two candidate species. As well

881 as being at the species level, the three clades do not match the subspecies distributions. This

882 suggests that the subspecies concept should be abandoned for this group.

883

884 CONCLUSION:

885 The results presented here provide strong evidence for the presence of cryptic lineages within

886 Chamaeleo, some of which can be considered candidate species. In particular, C. dilepis

887 comprises three species, C. gracilis three species, and C. anchietae two species. Additionally,

41 888 the results suggest that C. necasi should be synonymised with the nominate species of C.

889 gracilis (C. gracilis 3). Chamaeleo laevigatus and C. senegalensis are separate species. These

890 findings suggest a revision of the taxonomy of Chamaeleo is needed incorporating

891 morphological assessment. Some species were poorly sampled (C. gracilis 1 and 2; C.

892 anchietae 2) and it is possible that more extensive sampling would reveal additional cryptic

893 taxa, or at a minimum would lend support toward their description as separate species and a

894 better understanding of their distributions. Given that CITES regulations and the issuing of

895 export permits serve as a measure to ensure that exports are not detrimental to wild populations,

896 an accurate understanding of the taxonomy of Chamaeleo may be fundamental in ensuring that

897 species are not overexploited.

898

899 REFERENCES:

900 Aljanabi SM, Martinez I. 1997. Universal and rapid salt-extraction of high quality genomic

901 DNA for PCR-based techniques. Nucleic acids research 25: 4692–4693.

902 Allen GC, Flores-Vergara MA, Krasynanski S, Kumar S, Thompson WF. 2006. A modified

903 protocol for rapid DNA isolation from plant tissues using cetyltrimethylammonium

904 bromide. Nature Protocols 1: 2320–2325.

905 Beck, HE, Zimmermann NE, McVicar TR, Vergopolan N, Berg A, Wood EF. 2018. Data

906 descriptor: Present and future Koppen-Geiger climate classification maps at 1-km

907 resolution. Scientific Data 5:180214.

908 Bina Perl RG, Nagy ZT, Sonet G, Glaw F, Wollenberg KC, Vences M. 2014. DNA

909 barcoding Madagascar’s amphibian fauna. Amphibia-Reptilia 35(2).

910 Bouckaert R, Heled J, Kühnert D, Vaughan T, Wu CH, Xie D, Suchard MA, Rambaut A,

911 Drummond AJ. 2014. BEAST 2: A Software Platform for Bayesian Evolutionary

912 Analysis. PLoS Computational Biology 10: 1–6.

913 Brooks DR, McLennan D. a. 1999. Species: turning a conundrum into a research program.

42 914 Journal of nematology 31: 117–33.

915 Carstens BC, Pelletier TA, Reid NM, Satler JD. 2013. How to fail at species delimitation.

916 Molecular Ecology 22: 4369–4383.

917 Ceccarelli FS, Menegon M, Tolley KA, Tilbury CR, Gower DJ, Laserna MH, Kasahun R,

918 Rodriguez-Prieto A, Hagmann R, Loader SP. 2014. Evolutionary relationships, species

919 delimitation and biogeography of Eastern Afromontane horned chameleons

920 (Chamaeleonidae: Trioceros). Molecular Phylogenetics and Evolution 80: 125–136.

921 Chambers EA, Hebert PDN. 2016. Assessing DNA barcodes for species identification in

922 North American reptiles and amphibians in natural history collections. PloS ONE 11(4):

923 e0154363.

924 Chirio L, LeBreton W. 2007. Atlas des reptiles du Cameroun. MNHN, IRD, Paris.

925 Crandall KA, Bininda-Emonds ORR, Mace GM, Wayne RK. 2000. Considering evolutionary

926 processes in conservation biology. Trends in Ecology and Evolution 15: 290–295.

927 Darriba D, Taboada G, Doallo R, Posada D. 2012. jModelTest 2: more models, new

928 heuristics and parallel computing. Nature methods 9: 772.

929 DeBiasse MB, Hellberg ME. 2015. Discordance between morphological and molecular

930 species boundaries among Caribbean species of the reef sponge Callyspongia. Ecology

931 and Evolution 5(3): 663-675.

932 Engelbrecht HM, Branch WR, Greenbaum E, Alexander GJ, Jackson K, Burger M, Conradie

933 W, Kusamba C, Zassi-Boulou A-G, Tolley KA. 2019. Diversifying into the branches:

934 species boundaries in African green and bush snakes, Philothamnus (Serpentes:

935 Colubridae). Molecular Phylogenetics and Evolution 130: 357–365.

936 Fujisawa T, Barraclough TG. 2013. Delimiting species using single-locus data and the

937 generalized mixed yule coalescent approach: A revised method and evaluation on

938 simulated data sets. Systematic Biology 62: 707–724.

939 Fujita MK, Leaché AD, Burbrink FT, McGuire JA, Moritz C. 2012. Coalescent-based species

43 940 delimitation in an integrative taxonomy. Trends in Ecology and Evolution 27: 480–488.

941 Galtier N, Nabholz B, Glemin S, Hurst GDD. 2009. Mitochondrial DNA as a marker of

942 molecular diversity: a reappraisal. Molecular Ecology 18: 4541–4550.

943 Graham SW, Kohn JR, Morton BR, Eckenwalder JE, Barrett SCH. 1998. Phylogenetic

944 congruence and discordance among one morphological and three molecular data sets

945 from Pontederiaceae. Systematic Biology 47(4): 545-567.

946 Hamidy A, Matsui M, Shimada T, Nishikawa K, Yambun P, Sudin A, Kusrini MD, Kurniati

947 H. 2011. Morphological and genetic discordance in two species of Bornean

948 Leptobrachium (Amphibia, Anura, Megophryidae). Molecular Phylogenetics and

949 Evolution. 61: 904-913.

950 Hamilton CA, Formanowicz DR, Bond JE. 2011. Species delimitation and phylogeography

951 of Aphonopelma hentzi (araneae, mygalomorphae, theraphosidae): Cryptic diversity in

952 north american tarantulas. PLoS ONE 6: 12–16.

953 Huelsenbeck JP, Ronquist F. 2001. MrBayes: Bayesian inference of phylogenetic trees.

954 Bioinformatics 17: 754–755.

955 Jenkins R, Measey GJ, Anderson CV, Tolley KA. 2014. Chapter 10: Chameleon

956 conservation. In: Tolley KA, Herrel A (Eds.) The Biology of Chameleons. University of

957 California Press.

958 Joger U. 1990. The herpetofauna of the Central African Republic, with the description of a

959 new species of Rhinotyphlops (Serpentes: Typhlopidae), pp. 85-102. In: Peters G,

960 Hutterer R (Eds.) Vertebrates in the tropics. Museum Alexander Koenig, Bonn.

961 Kapli P, Lutteropp S, Zhang J, Kobert K, Pavlidis P, Stamatakis A, Flouri T. 2017. Multi-rate

962 Poisson tree processes for single-locus species delimitation under maximum likelihood

963 and Markov chain Monte Carlo. Bioinformatics 33: 1630–1638.

964 Kearse M, Moir R, Wilson A, Stones-Havas S, Cheung M, Sturrock S, Buxton S, Cooper A,

965 Markowitz S, Duran C, Thierer T, Ashton B, Meintjes P, Drummond A. 2012. Geneious

44 966 Basic: An integrated and extendable desktop software platform for the organization and

967 analysis of sequence data. Bioinformatics 28: 1647–1649.

968 Largen MJ, Spawls S. 2006. Lizards of Ethiopia (Reptilia Sauria): An annotated checklist,

969 bibliography, gazetteer and identification key. Tropical Zoology 19: 21–109.

970 Lefébure T, Douady CJ, Gouy M, Gibert J. 2006. Relationship between morphological

971 taxonomy and molecular divergence within Crustacea: Proposal of a molecular

972 threshold to help species delimitation. Molecular Phylogenetics and Evolution 40: 435–

973 447.

974 Main DC, Jansen van Vuuren B, Tolley KA. 2018. Cryptic diversity in the common flap-

975 necked chameleon Chamaeleo dilepis in South Africa. African Zoology 53: 11–16.

976 Matthee CA, Tilbury CR, Townsend T. 2004. A phylogenetic review of the African leaf

977 chameleons: Genus Rhampholeon (Chamaeleonidae): The role of vicariance and climate

978 change in speciation. Proceedings of the Royal Society B: Biological Sciences 271:

979 1967–1975.

980 Meier R, Shiyang K, Vaidya G, Ng PKL. 2006. DNA barcoding and taxonomy in diptera: a

981 tale of high intraspecific variability and low identification success. Systematic Biology

982 55: 715–728.

983 Mishler BD, Wilkins JS. 2018. The Hunting of the SNaRC: A Snarky Solution to the Species

984 Problem. Philosophy, Theory, and Practice in Biology 10.

985 Moritz C. 1994. Defining ‘evolutionarily significant units’ for conservation. Trends in

986 Ecology and Evolution 9: 373–375.

987 Pons J, Barraclough TG, Gomez-Zurita J, Cardoso A, Duran DP, Hazell S, Kamoun S,

988 Sumlin WD, Vogler AP. 2006. Sequence-Based Species Delimitation for the DNA

989 Taxonomy of Undescribed Insects. Systematic Biology 55: 595–609.

990 Puillandre N, Modica M V., Zhang Y, Sirovich L, Boisselier MC, Cruaud C, Holford M,

991 Samadi S. 2012. Large-scale species delimitation method for hyperdiverse groups.

45 992 Molecular Ecology 21: 2671–2691.

993 De Queiroz K. 2005. Ernst Mayr and the modern concept of species. Precedings of the

994 National Academy of Sciences 102: 6600–6607.

995 De Queiroz K. 2007. Species concepts and species delimitation. Systematic Biology 56: 879–

996 886.

997 Rambaut A, Drummond AJ, Xie D, Baele G, Suchard MA. 2018. Posterior Summarization in

998 Bayesian Phylogenetics Using Tracer 1.7. Systematic Biology 67: 901–904.

999 Raxworthy CJ, Forstner MRJ, Nussbaum RA. 2002. Chameleon radiation by oceanic

1000 dispersal. Nature 415: 784–787.

1001 Reid NM, Carstens BC. 2012. Phylogenetic estimation error can decrease the accuracy of

1002 species delimitation: a Bayesian implementation of the general mixed Yule-coalescent

1003 model. BMC evolutionary biology 12: 196.

1004 Romano SL, Palumbi SR. 1997. Molecular evolution of a portion of the mitochondrial 16S

1005 ribosomal gene region in scleractinian corals. Journal of molecular evolution 45: 397–

1006 411.

1007 Sharma PP, Fernandez R, Eposito LA, Gonzalez-Santillan E, Monod L. 2015. Phylogenomic

1008 resolution of scorpions reveals multilevel discordance with morphological phylogenetic

1009 signal. Proceedings of the Royal Society B:Biological Sciences 282: 20142953.

1010 Stamatakis A. 2006. RAxML-VI-HPC: Maximum likelihood-based phylogenetic analyses

1011 with thousands of taxa and mixed models. Bioinformatics 22: 2688–2690.

1012 Swofford DL. 2003. PAUP*. Phylogenetic analysis using parsimony (*and other methods).

1013 Version 4. Sinauer associates, Sunderland.

1014 Tilbury C. 2018. Chameleons of Africa: an atlas: including the chameleons of Europe, the

1015 Middle East and Asia (Illustrate). Frankfurt: Edition Chimaira.

1016 Tilbury CR, Tolley KA. 2009. A re-appraisal of the systematics of the African genus

1017 Chamaeleo (Reptilia: Chamaeleonidae). Zootaxa 2079: 57–68.

46 1018 Tolley, K. 2014. Chamaeleo dilepis. The IUCN Red List of Threatened Species 2014:

1019 e.T176308A1438077. http://dx.doi.org/10.2305/IUCN.UK.2014-

1020 3.RLTS.T176308A1438077.en. Downloaded on 19 February 2019.

1021 Tolley, K., Anderson, C.V. & Tilbury, C. 2015. Chamaeleo anchietae. The IUCN Red List of

1022 Threatened Species 2015: e.T172526A1344263.

1023 http://dx.doi.org/10.2305/IUCN.UK.2015-4.RLTS.T172526A1344263.en. Downloaded

1024 on 19 February 2019.

1025 Tolley, K., Luiselli, L., Niagate, B., Penner, J., Trape, J.-F., Rödel , M.-O., Segniagbeto, G. &

1026 Wagner, P. 2014. Chamaeleo gracilis. The IUCN Red List of Threatened Species 2014:

1027 e.T172536A1344573. http://dx.doi.org/10.2305/IUCN.UK.2014-

1028 3.RLTS.T172536A1344573.en. Downloaded on 19 February 2019.

1029 Tolley, K. & Trape, J.-F. 2014. Chamaeleo laevigatus. The IUCN Red List of Threatened

1030 Species 2014: e.T172542A1344815. http://dx.doi.org/10.2305/IUCN.UK.2014-

1031 3.RLTS.T172542A1344815.en. Downloaded on 19 February 2019.

1032 Tolley KA, Townsend TM, Vences M. 2013. Large-scale phylogeny of chameleons suggests

1033 African origins and Eocene diversification. Proceedings of the Royal Society B:

1034 Biological Sciences 280: https://doi.org/10.1098/rspb.2013.0184.

1035 Trape J-F, Trape S, Chirio L. 2012. Lezards, crocodiles et tortues d’Afrique occidentale et du

1036 Sahara. IRD editions, Marseilles.

1037 Ullenbruch K, Krause P, Bohme W. 2007. A new species of the Chamaeleo dilepis group

1038 (Sauria Chamaeleonidae) from West Africa. Tropical Zoology. 20:1-17

1039 Uetz, P., Freed, P. & Hošek, J. (eds.), The Reptile Database, http://www.reptile-database.org,

1040 accessed [14 January 2019]

1041 Vasconcelos R, Montero-Mendieta S, Simó-Riudalbas M, Sindaco R, Santos X, Fasola M,

1042 Llorente G, Razzetti E, Carranza S. 2016. Unexpectedly high levels of cryptic diversity

1043 uncovered by a complete DNA barcoding of reptiles of the socotra archipelago. PLoS

47 1044 ONE 11: 1–19.

1045 Wilms, T., Wagner, P., Penner, J., Rödel , M.-O., Luiselli, L., Segniagbeto, G., Niagate, B.,

1046 Carpenter, A. & Trape, J. 2013. Chamaeleo senegalensis. The IUCN Red List of

1047 Threatened Species 2013: e.T176312A15898112.

1048 http://dx.doi.org/10.2305/IUCN.UK.2013-1.RLTS.T176312A15898112.en.

1049 Downloaded on 19 February 2019.

1050 Wüster W, Chirio L, Trape JF, Ineich I, Jackson K, Greenbaum E, Barron C, Kusamba C,

1051 Nagy ZT, Storey R, Hall C, Wüster CE, Barlow A, Broadley DG. 2018. Integration of

1052 nuclear and mitochondrial gene sequences and morphology reveals unexpected diversity

1053 in the forest cobra (Naja melanoleuca) species complex in Central and West Africa

1054 (Serpentes: Elapidae). Zootaxa 4455: 68–98.

1055 Yang Z. 2015. A tutorial of BPP for species tree estimation and species delimitation. Current

1056 Zoology 61: 854–865.

1057 Yang Z, Rannala B. 2010. Bayesian species delimitation using multilocus sequence data.

1058 Genetics 107: 9264–9269.

1059 Yang Z, Rannala B. 2014. Unguided species delimitation using DNA sequence data from

1060 multiple loci. Journal of Molecular Biology and Evolution 31: 3125–3135.

1061 Zachos FE. 2018. (New) Species concepts, species delimitation and the inherent limitations

1062 of taxonomy. Journal of Genetics 97(4): 811-815.

48 1063 Appendix I

1064 Table 1: Species, field numbers, GenBank accession numbers for four genes and sampling localities of Chamaeleo used in the study, as well as outgroup taxa. TBA indicates sequence data that 1065 will be deposited in GenBank upon submission of a manuscript to a peer-reviewed journal. NA indicates not available

Species Catalogue or 16S ND4 RAG1 PRLR Locality

field number

Chamaeleo africanus TR 4462 TBA TBA NA NA Chad

Chamaeleo africanus MVZ 238898 HF570439 HF570562 HF570763 HF570852 NA

Chamaeleo calcaricarens MVZ 241362 HF570442 HF570565 NA HF570855 Ethiopia

Chamaeleo chamaeleon TR 4152 NA TBA NA NA Algeria

Chamaeleo chamaeleon SRM Chach1 NA TBA NA NA NA

Chamaeleo arabicus MVZ 246129 HF570441 HF570564 HF570765 HF570854 NA

Chamaeleo calyptratus SRM Chaca4 TBA TBA NA NA NA

Chamaeleo calyptratus ZSM 801 2000 HF570443 HF570566 HF570766 HF570856 NA

Chamaeleo zeylanicus MVZ 248410 HF570447 HF570572 HF570770 HF570861 Sri Lanka

Chamaeleo monachus MVZ 236483 HF570445 HF570569 HF570768 HF570859 Socotra

Chamaeleo dilepis KTH09 111 TBA TBA TBA NA Angola

Chamaeleo dilepis KTH09 87 TBA TBA TBA TBA Angola

Chamaeleo dilepis KTH09 285 HF570444 HF570567 HF570767 HF570857 Angola

Chamaeleo dilepis MBUR 08007 TBA TBA TBA TBA Republic of Congo

Chamaeleo dilepis MBUR 0826 TBA TBA TBA TBA Republic of Congo

Chamaeleo dilepis CT 401 TBA TBA TBA TBA Democratic Republic of the Congo

Chamaeleo dilepis CT 403 TBA TBA TBA TBA Democratic Republic of the Congo

Chamaeleo dilepis KTH09 286 TBA TBA TBA TBA Angola

Chamaeleo dilepis DM29 TBA TBA TBA TBA Johannesburg, Republic of South Africa

49 Chamaeleo dilepis RSP 328 TBA TBA TBA TBA Tswalu, Republic of South Africa

Chamaeleo dilepis AMB 7969 TBA TBA TBA TBA Namibia

Chamaeleo dilepis CT 053 TBA TBA TBA TBA Malawi

Chamaeleo dilepis CT 382 TBA TBA TBA TBA Tanzania

Chamaeleo dilepis PEM R18937 TBA TBA TBA TBA Zambia

Chamaeleo dilepis CT 059 TBA TBA TBA TBA Botswana

Chamaeleo dilepis EI 0488 TBA TBA TBA TBA Mozambique

Chamaeleo dilepis EI 0472 TBA TBA TBA TBA KZN, Republic of South Africa

Chamaeleo dilepis MBUR 01460 TBA TBA TBA TBA Mpumalanga, Republic of South Africa

Chamaeleo dilepis CT 057 TBA TBA TBA TBA Zambia

Chamaeleo dilepis MOZ17 174 TBA TBA TBA TBA Mozambique

Chamaeleo dilepis CT 684 TBA TBA TBA TBA Tanzania

Chamaeleo dilepis CT 213 TBA TBA TBA TBA Kenya

Chamaeleo dilepis CT 216 TBA TBA TBA TBA Tanzania

Chamaeleo dilepis CT 679 TBA TBA TBA TBA Kenya

Chamaeleo dilepis* ZMFK 54964 TBA TBA NA NA Rwanda

Chamaeleo dilepis CT 383 TBA TBA TBA TBA Zanzibar

Chamaeleo gracilis CT 200 TBA TBA TBA TBA Tanzania

Chamaeleo gracilis TR 4429 TBA TBA TBA TBA Chad

Chamaeleo gracilis CT 088 FJ717748 HF570568 FJ746587 HF570858 Guinea

Chamaeleo gracilis MW 06086 TBA TBA NA NA Sierra Leone

Chamaeleo gracilis HB 372 TBA TBA TBA TBA Liberia

Chamaeleo necasi ZFMK 77059 HF570446 HF570570 HF570769 NA Togo

Chamaeleo laevigatus TR 4423 TBA TBA TBA NA Chad

50 Chamaeleo laevigatus HB 375 NA TBA NA NA Ethiopia

Chamaeleo laevigatus MBUR 08434 TBA TBA NA NA Ethiopia

Chamaeleo senegalensis PEM R16627 FJ717752 HF570571 FJ746590 HF570860 NA

Chamaeleo senegalensis SRM Chase1 TBA TBA NA NA NA

Chamaeleo anchietae KTH09 113 HF570440 HF570563 HF570764 HF570853 Angola

Chamaeleo anchietae KTH09 284 TBA TBA TBA TBA Angola

Chamaeleo anchietae KTH09 329 TBA TBA TBA NA Angola

Chamaeleo anchietae CT 409 TBA TBA TBA NA Democratic Republic of the Congo

Chamaeleo anchietae Mahler 0639 TBA NA NA NA Tanzania

Chamaeleo namaquensis WC09 012 HQ130516 FJ981754 FJ984185 HQ130604 Angola

Chamaeleo namaquensis HB 118 TBA TBA NA TBA Namibia

Chamaeleo namaquensis KTH11 22 TBA TBA NA NA Republic of South Africa

Trioceros jacksonii CAS199070 FJ717774 AF4423229 FJ746608 HQ130610 NA

1066 *Corresponding museum specimen catalogued as C. anchietae

51 1067 Appendix II

1068 Figure 1: A Bayesian consensus tree for the mitochondrial dataset for Chamaeleo. Node support is indicated for Bayesian 1069 and likelihood analyses.

52 1070 Chapter 3

1071

1072 Biogeography of Chamaeleo and ecological partitioning of Chamaeleo dilepis

1073

1074

1075 ABSTRACT:

1076 Chamaeleo is a pan-African genus of chameleons that also extends into southern Europe,

1077 southern Asia, and the Middle East, with much geographic overlap and confusing phenotypic

1078 variation between its taxa. The large, overlapping ranges of members of Chamaeleo present an

1079 interesting but complex case for biogeographic ancestral area reconstruction. A previous study

1080 (Chapter 2) found strong evidence for the presence of cryptic species within the Chamaeleo

1081 dilepis complex. The aims of the present study were twofold: firstly, to reconstruct the ancestral

1082 biogeographic history of the genus Chamaeleo and secondly, to investigate whether there is

1083 evidence for niche partitioning between the three cryptic species within the C. dilepis complex.

1084 A time-calibrated phylogenetic analysis was run on a dataset encompassing all known species

1085 of Chamaeleo. Dispersal-Extinction-Cladogenesis (DEC) was used on this time-calibrated

1086 phylogeny to reconstruct the ancestral biogeography of Chamaeleo. Additionally, Maxent was

1087 used to carry out ecological niche modelling on the three clades within C. dilepis. The common

1088 ancestor of Chamaeleo emerged during the Eocene, most likely in the Zambezian region.

1089 Chamaeleo probably exploited mesic corridors that opened up during forest contractions of the

1090 Oligocene and Miocene to move into North Africa and Eurasia. Niche partitioning is evident

1091 between C. dilepis clades and only the suitable niche of C. dilepis 2 appears relatively stable

1092 since the late Pleistocene. These results provide a glimpse at the evolutionary biogeographic

1093 history of Chamaeleo and could aid future revisions of distribution maps pertaining to the C.

1094 dilepis complex.

53 1095 INTRODUCTION:

1096 Climate change is at the helm of evolutionary diversification on earth. As climate shifts, certain

1097 phenotypes will not be able to adapt to the new conditions, particularly if they have narrow

1098 thermal or moisture tolerances. Climate is also the basis for the prevailing vegetation and

1099 consequently the prevailing biome in a region (Pörtner 2001, 2002; Morley 2007; Somero

1100 2010; Kissling et al. 2012). Together with geological change, climate change has played a

1101 pivotal role in the present day distribution of the planet’s fauna and flora. The study of

1102 biogeography is tasked with correlating the evolutionary history of different organisms with

1103 the timing of known geological and climatic changes (Cracraft 1975; Nelson 1985). Recent

1104 advances in molecular techniques have prompted a number of studies correlating a molecularly

1105 informed evolutionary history with known geological and climatic events (Tolley et al. 2006,

1106 2011, 2013; Kuriyama et al. 2011; Measey and Tolley 2011; Barlow et al. 2013; da Silva and

1107 Tolley 2017; Jongsma et al. 2018).

1108

1109 Chamaeleo is a pan-African genus of chameleons with uncharacteristically (for chameleons)

1110 large ranges with overlap between species. The genus currently comprises 14 species, however,

1111 with large intraspecific variation, this number is likely an underestimate of the actual species

1112 diversity within the genus (Chapter 2). In Chapter 2, three species-level clades were revealed

1113 within the Chamaeleo dilepis complex. One of these clades was sampled from near the type

1114 locality (“Congo”) of typical C. dilepis and this clade therefore represents C. dilepis while the

1115 other two clades are additional candidate species within the complex. In contrast to the forest

1116 affinity of most other chameleon genera, Chamaeleo tends to occupy more mesic or xeric

1117 regions, with an affinity for savanna, grassland (in the case of C. anchietae) and even desert

1118 (Tilbury 2018). It is therefore possible that the genus originated in Southern Africa, coinciding

1119 with cooler and drier climate that prevailed in that region during the Oligocene and Miocene

1120 (Morley 2007; Feakins and deMenocal 2010). Given that multiple species delimitation methods

54 1121 suggested that the three C. dilepis clades each represent full species (Chapter 2), it is expected

1122 that the bioclimatic preferences of these clades will be distinct from one another and that they

1123 will show minimal overlap in suitable climatic extent through time. Such a finding would lend

1124 support to the clades representing different species.

1125

1126 While species of the Chamaeleo genus tend to occupy large ranges in the present day, and they

1127 appear to be mesic habitat generalists, it is still possible that their suitable climate space

1128 contracted during paleoclimatic cycling of the Pleistocene. The Last Interglacial (LIG) covers

1129 a period from 129 000-116 000 years ago during which some of the hottest African climates of

1130 the Pleistocene were experienced (Partridge et al. 1997; Govin et al. 2014; Maslin et al 2014;

1131 Tierney et al. 2017). Precipitation during this period varied considerably with latitude;

1132 Southern Africa experienced a largely arid climate during this period while the central

1133 equatorial African climate was wet and monsoonal. In contrast, the Last Glacial Maximum

1134 (LGM), which occurred approximately 20 000 years ago, represents the last period of extensive

1135 glaciation at northern hemispheric latitudes and consequently resulted in cooler temperatures

1136 in equatorial Africa (though no glaciation there). However, Southern Africa actually saw a

1137 warming climate during the LGM (Wu et al. 2007). Given this, it is expected that, in line with

1138 the apparently mesic bioclimatic preference of the Chamaeleo genus, the LIG and LGM would

1139 have, in general, provided smaller suitable climatic envelopes than the present day for C.

1140 dilepis. In particular, C. dilepis 1 and C. dilepis 3 are both centred more toward the lower,

1141 tropical latitudes (Fig. 1) and so the monsoonal climate during the LIG would have presumably

1142 contracted their suitable climate space.

55 1143 Figure 1: Sampling points used for the Maxent ecological niche modelling analysis. Chamaeleo dilepis clades are colour 1144 coded: C. dilepis 1 = green; C. dilepis 2 = blue; C. dilepis 3 = purple.

1145 The widespread distribution of C. dilepis 2 is centred more toward the higher latitudes reaching

1146 into the subtropics (Fig. 1) and this would be expected to favour the stability of this species’

1147 suitable climate space through time, especially because the effects of the LIG and LGM were

1148 presumably more moderate toward the south.

1149

1150 To examine whether Chamaeleo did indeed originate in Southern Africa, the biogeographic

1151 history of Chamaeleo was evaluated using an ancestral area reconstruction carried out on a

1152 time-calibrated phylogeny of the genus. Additionally, to test whether there exists ecological

1153 support for C. dilepis species delimited in Chapter 2, suitable climatic space for each C. dilepis

1154 clade was modelled using the programme Maxent for the current day, the LIG and the LGM.

56 1155 MATERIALS AND METHODS:

1156 Biogeographic history

1157 Using the combined four-gene dataset from Chapter 2, a time-calibrated phylogeny was

1158 estimated in BEAST 2 (Bouckaert et al. 2014). With a paucity of fossil data for Chamaeleo

1159 and much uncertainty as to the taxonomy of the few fossils available (Bolet & Evans 2014), it

1160 was decided that a secondary calibration point would be incorporated in lieu of fossil

1161 calibrations. The large-scale fossil-calibrated phylogeny of the Chamaeleonidae family was

1162 used as a source for the secondary calibration point for the most recent common ancestor of

1163 Chamaeleo (Tolley et al. 2013). A lognormal distribution with an offset mean age of 43 million

1164 years ago (mya) and a standard deviation of 5 mya was applied and this produced a distribution

1165 that closely reflected Tolley et al. (2013)’s 95% HPD confidence intervals for the same node.

1166 The origin of the family has been shown to be African (Tolley et al. 2013), but that analysis

1167 included all chameleon genera as well as a sister family (Agamidae). The current analysis was

1168 restricted to understanding the history within Chamaeleo, not the entire family. Therefore, the

1169 inclusion of other genera and sister families would produce a time-tree with deep nodes, that

1170 are not necessary for the present analysis, and could potentially bias the dates for the target

1171 clade backwards. Therefore, one outgroup genus was included and the root of the tree was

1172 coded as having an African origin, as in Tolley et al. (2013).

1173

1174 The dataset was partitioned by marker, applying optimal substitution models inferred using

1175 jModelTest 2 (Darriba et al. 2012 - 16S: GTR; ND4 GTR + I + G; PRLR and RAG1: HKY +

1176 G). PRLR and RAG1 were unlinked for substitution models and clock models. Given that they

1177 form part of the same locus, the mitochondrial markers (16S and ND4) were linked across

1178 substitution models and clock models. Trees were linked across all markers. A lognormal

1179 relaxed molecular clock (Drummond et al. 2006) was set for all markers and a Birth-Death tree

1180 prior was used. Two independent runs were carried out with the analysis set to 200 million

57 1181 generations, sampling every 5 000 trees. Convergence was assessed for each run in Tracer 1.7

1182 (Rambaut et al. 2018) and determined adequate when the ESS for each parameter ³ 200. The

1183 last 10 000 trees of both runs were then combined in LogCombiner 2.5.1 (Bouckaert et al.

1184 2014) and a maximum clade credibility tree with mean tree heights was determined in

1185 TreeAnnotator 2.5.1 (Bouckaert et al. 2014). All BEAST analyses were run on the CIPRES

1186 Science Gateway (http://www.phylo. org/sub_sections/portal/).

1187

1188 An ancestral area reconstruction for the genus Chamaeleo was made using the time-calibrated

1189 phylogeny, with terminal tips coded to the following biogeographic regions: A = Eurasia

1190 (Europe, Asia and the Middle-East), B = North Africa, C = Congolian, D = Sudanian, E =

1191 Ethiopian-Somalian (includes Ethiopian and Somalian sub-regions), F = Zambezian, G =

1192 Southern African (see Fig. 3 inset). These areas were chosen on the basis of biogeographic

1193 partitioning of the African continent (Linder et al. 2012), and were broad enough to prevent

1194 overparameterization whilst still allowing for the hypothesis that Chamaeleo had Southern

1195 African origins to be tested. The Dispersal-Extinction-Cladogenesis (DEC) reconstruction

1196 approach implemented in the programme Reconstruct Ancestral State in Phylogenies (RASP:

1197 Yu et al. 2015) was used to carry out an ancestral area reconstruction on the time-calibrated

1198 phylogeny. Dispersal probabilities between regions were set according to geographic proximity

1199 of regions. That is, regions not bordering each other were given low probability of dispersal

1200 between them. Time-dependent dispersal constraints between regions were also set, based on

1201 paleo-vegetation maps of Morley (2007) and Kissling et al. (2012). Given the relatively xeric

1202 tolerability and savannah affinity of most Chamaeleo species (Tilbury 2018), the late Eocene

1203 climatic optimum of approximately 50 mya that promoted forest expansion was presumed

1204 unfavourable to Chamaeleo diversification and dispersal. Conversely, arid episodes of the

1205 Miocene (ca. 20-13 mya) were assumed to have favoured Chamaeleo dispersal and

1206 diversification and dispersal probabilities were set to account for these assumptions (Morley

58 1207 2007; Kissling et al. 2012; Tolley et al. 2013). Geographically adjacent regions were given a

1208 dispersal probability of 1.0 between them during favourable climatic dispersal conditions and

1209 this was reduced to 0.5 during unfavourable dispersal conditions. Regions not geographically

1210 adjacent to one another were given dispersal probabilities corresponding to their proximity to

1211 one another, regions with another biogeographic region in between them were given a dispersal

1212 probability of 0.25 during favourable conditions which was reduced to 0.1 during unfavourable

1213 conditions (Table 1, Appendix IV). For regions with two or more biogeographic regions

1214 between them, dispersal was constrained to a probability of 0.

1215

1216 Niche modelling

1217 Current climatic niche

1218 Two new candidate species were revealed for Chamaeleo dilepis in Chapter 2 (Clades 2 and

1219 3). To investigate potential climatic partitioning between these candidate species and the

1220 nominate species (Clade 1), niche modelling was carried out in the programme Maxent 3.1.4

1221 (Phillips and Dudík 2008; Elith et al. 2011). There are many records for C. dilepis that were

1222 not used in Chapter 2’s phylogeny, therefore, the general location of each C. dilepis clade from

1223 the phylogeny was used to guide the assignment of additional records to clades. These records

1224 were taken from Tolley et al. (2016)’s species occurrence dataset assimilated from various

1225 sources (GBIF; Herpnet; Natural History Museum London; Bates et al. 2014). This increased

1226 the sampling complement to 80 individuals for C. dilepis 1, 1062 individuals for C. dilepis 2,

1227 and 80 individuals for C. dilepis 3 (Fig. 1). A total of 19 bioclimatic variables were downloaded

1228 from WorldClim (Hijmans et al. 2005; Warren et al. 2008). These variables are averaged over

1229 50 years and include temperature and precipitation data. Each clade was analysed separately in

1230 Maxent and duplicate points were removed from the analysis. In each case, the most suitable

1231 of the 19 bioclimatic variables were determined by running preliminary models. These models

1232 were run under the default programme settings except the maximum number of iterations was

59 1233 set to 5000 to allow for the models to converge. For each model, ten replicates were run, each

1234 with a random subsample of 25% used for testing so that the dependence of the models on

1235 presence data could be determined. For each clade, the averaged results from the 10 runs were

1236 examined and in each case the area under the curve (AUC) and the receiver operating

1237 characteristic (ROC) plots were noted. This information, together with the three jackknife tests

1238 (AUC, test gain, and regularized training gain) was used to refine the bioclimatic variable

1239 dataset, omitting variables that did not contribute more than 1% to the model or permutation

1240 importance. Additionally, any variable whose presence hampered the model’s predictive

1241 performance was also excluded. The model was then re-run on this refined dataset and the

1242 above process was repeated again until all variables used contributed favourably to the model’s

1243 predictive performance.

1244

1245 Past climatic niche

1246 While presence data is only available for the current day, paleoclimatic reconstructions of the

1247 19 bioclimatic variables used above are made available by WorldClim for both the Last

1248 Interglacial (LIG) and the Last Glacial Maximum (LGM) at 30-arc second and 2.5-arc minute

1249 resolution respectively. Although the LIG and LGM both occurred during the Pleistocene,

1250 which is far more recent than the diversification within the C. dilepis clade (see Results; Fig.

1251 2), these two time slices represent the most extreme climates experienced during the

1252 Pleistocene and allow for the reconstruction of more recent climate-driven range expansions

1253 and contractions. For each clade, the current climatic niche modelled above was projected

1254 back onto the LIG and LGM bioclimatic variables after the climatic dataset had been refined

1255 only to include the most informative variables.

60 1256 RESULTS:

1257 Within Chamaeleo, C. namaquensis diverged from other species approximately 40 mya (Fig.

1258 2), followed by the divergence of Chamaeleo anchietae during the late Oligocene (ca. 29 ± 5

1259 mya). Eurasian taxa split from other Chamaeleo during the early Miocene (ca. 20 mya) and

1260 most species level diversification within the Eurasian clade occurred in the mid-Miocene.

1261 Diversification within the C. dilepis clade, and for the other African species C. senegalensis,

1262 C. gracilis, and C. laevigatus, occurred from the late Miocene to the start of the Pliocene (Fig.

1263 2).

61 1264 Figure 2: A time-calibrated phylogeny of Chamaeleo. 95% High density probability confidence intervals are indicated by 1265 blue bars. Black circles represent nodes with ³ 0.95 posterior probability support. Geologic epochs are indicated.

62 1266 Biogeographic history

1267 The ancestral area reconstruction suggests that the most recent common ancestor (MRCA) of

1268 Chamaeleo most likely originated within the Southern African/Zambezian or Southern

1269 biogeographic regions (Fig. 3). Chamaeleo namaquensis’ sister taxon relationship with the rest

1270 of Chamaeleo, together with its Southern African distribution suggests that the MRCA of the

1271 genus was initially to the south with additional species diversifying to the north. Further

1272 support for this northward dispersal and diversification hypothesis is given by the more recent

1273 dispersal of Chamaeleo out of Africa and into Eurasia in the early Miocene (ca. 23 mya ±

1274 5mya) which suggests their presence in north Africa was recent (Fig. 2 and 3).

63 1275 Figure 3: A time-calibrated phylogeny of Chamaeleo and a representative of its closest related genera showing the results of 1276 a DEC ancestral area reconstruction. Coloured doughnut charts represent the proportional likelihood values of each ancestral 1277 area. Letters at each terminal tip indicate the input area coding for that taxon. Ancestral areas of negligible proportional 1278 likelihood are combined and shown in black.

64 1279 Niche modelling

1280 The variables that best explained the suitable climatic niches of C. dilepis clades were in most

1281 cases related to precipitation (Table 2). In particular, precipitation of the driest month (Bio 14)

1282 and annual precipitation (Bio 12) were found to be of the three best explaining variables for

1283 the C. dilepis 1 and C. dilepis 3. Temperature appears important in the niche suitability of C.

1284 dilepis 2 with two of its three best explaining variables pertaining to temperature: temperature

1285 seasonality (Bio 4) and mean temperature of the driest quarter (Bio 9).

1286 Table 2: The three most important bioclimatic variables for the prediction of suitable niches for three clades of Chamaeleo 1287 dilepis and their respective contributions to the model.

Species Bioclimatic variable Detail % contribution*

C. dilepis 1 Bio 14 Precipitation of driest month 17.0 (7.2)

Bio 7 Temperature annual range 15.4 (2.1)

Bio 12 Annual precipitation 13.7 (2.1)

C. dilepis 2 Bio 4 Temperature Seasonality 51.1 (61)

Bio 18 Precipitation of warmest quarter 14.5 (1.7)

Bio 9 Mean temperature of driest quarter 7.4 (17.1)

C. dilepis 3 Bio 14 Precipitation of driest month 20.2 (2.7)

Bio 12 Annual precipitation 18.7 (17.4)

Bio 15 Precipitation seasonality 15 (1.1)

1288 *The permutation importance of each variable is indicated in brackets next to the percentage contribution

1289 The current suitable niches for each of the clades (Fig. 4) are reasonable approximations of the

1290 distribution of points that were input, suggesting that the model is not under- or over-predicting

1291 (Appendix III, Fig. 1). There may be greater connection between C. dilepis 1 and C. dilepis 3

1292 in the present day and LGM than during late Pleistocene (LIG). While the inland suitable

1293 climate of C. dilepis 2 did not extend as far south in the LGM as it does currently, LGM

1294 predictions for all C. dilepis clades are largely similar to the current day. A slightly larger area

1295 of suitable climate space is evident during the LGM than current day for C. dilepis 3 and for

1296 C. dilepis 1 the opposite is true however in both cases these differences are negligible. During

1297 the LIG, C. dilepis 2 and C. dilepis 3 may have experienced a contraction in their suitable

65 1298 climate space with the suitable climate space of C. dilepis 1 shifting northwards, C. dilepis 2

1299 contracting slightly shifting westwards, and C. dilepis 3 contracting into small pockets.

66 Last Interglacial Last Glacial Maximum Current

1300 Figure 4: Predicted suitable climate space of Chamaeleo dilepis clades based on a minimum training presence logistic threshold for three time slices (Last Interglacial, Last Glacial Maximum, 1301 and Current). Probability of climatic suitability is indicated to the left of the figures for each species. Areas of overlap in suitable climate space between clades are indicated in black.

67 1302 DISCUSSION:

1303 Biogeographic history:

1304 Initial diversification within the Chamaeleo genus coincides with the onset of global cooling

1305 during the late Eocene which ended approximately 34 mya. Chamaeleo namaquensis and the

1306 C. anchietae clade showed an early divergence from the rest of the genus, around 40 and 30

1307 mya, respectively. Divergence among the other species occurred primarily in the Miocene,

1308 starting around 25 mya. Global retraction in forests (Morley 2007, Kissling et al. 2012) that

1309 became prevalent during the Miocene could have made the Congolian and Zambezian regions

1310 more penetrable for Chamaeleo allowing them to expand northward and diversify as the forests

1311 retracted. Diversification within the Chamaeleo dilepis clade took place from the late Miocene

1312 to the Pliocene (ca. 10-5 mya; Figs 2 & 3).

1313

1314 The early Eocene presence of forest in the tropics would have likely made dispersal through

1315 the Congolian and northern Zambezian regions challenging for the apparently more mesic/arid

1316 adapted Chamaeleo genus. Chamaeleo namaquensis is arguably the only chameleon occupant

1317 of true desert in the world and currently confined to the Southern African sub-region. This

1318 species diverged from the rest of Chamaeleo during the mid-Eocene (ca. 42 mya) and the

1319 ancestral area reconstruction suggests that it was originally in Southern African /Zambezian

1320 region (Fig. 3). This finding supports what was hypothesized at the onset of the study, that

1321 Chamaeleo originated in Southern Africa. Chamaeleo namaquensis, due to its affinity for

1322 desert, is in fact more terrestrial than it is arboreal, with shorter limbs and a shorter tail than

1323 relative to body size than other arboreal chameleons (Herrel et al. 2013; Tilbury 2018). If the

1324 ancestral body type of C. namaquensis was similar to that of today, the tropical forests of the

1325 Eocene would have probably been impenetrable to a terrestrial chameleon with a phenotype

1326 adapted to arid conditions. In the 15 million years following the initial divergence of C.

1327 namaquensis from the rest of Chamaeleo, aridity in southwest Africa (Feakins and deMenocal

68 1328 2010) intensified and this probably favoured diversification for other species in the genus.

1329 Chamaeleo anchietae diverged, approximately 29 mya during the Oligocene (Figs 2 & 3). This

1330 species currently occupies mesic, high altitude montane shrubland and grassland plateaus and

1331 its phenotype appears less arboreally adapted than other species in the genus (Tilbury 2018).

1332 The current known distribution of C. anchietae is fragmented with pockets in high altitude

1333 plateaus of south-western Angola, south-eastern DRC, central Tanzania and the Nyika Plateau

1334 in northern Malawi. These pockets seem to track the distribution of an arid corridor connecting

1335 southwest Africa with the Horn of Africa which is thought to have formed during the Miocene

1336 (Bobe 2006). It makes sense for a species with a propensity for a relatively temperate climate

1337 and open habitat to have made use of such a corridor to disperse, especially given that its

1338 phenotype might have hindered its dispersal through dense forest. Prior studies have shown

1339 this to have been the case for xeric amphibians, large mammals and certain arid flora (Poynton

1340 1995; Bobe 2006; Bellstedt et al. 2012).

1341

1342 During the early Miocene, Eurasian species dispersed out of Africa (ca. 23 mya). Globally, the

1343 onset of the Miocene brought warmer temperatures which resulted in the expansion of forests

1344 in most of the tropics, however, in Africa, this period is thought to have undergone a reduction

1345 in forest, with an expansion in grasslands (Morley 2007). This would have opened up corridors

1346 (like the arid corridor mentioned above) for Chamaeleo species to disperse north of the

1347 Congolian region into North Africa and the Ethiopian region and subsequently into Eurasia.

1348 Species level diversification appears to have reached a peak from the mid to late Miocene (ca.

1349 15-5 mya) during which global cooling and drying favoured the replacement of much African

1350 forest with more temperate savannah grassland, opening up more niches for chameleons with

1351 less of a propensity for forest (Morley 2007; Kissling et al 2012; but see also Ceccarelli et al.

1352 2014).

69 1353 The C. dilepis clade likely emerged in the Zambezian and Congolian regions approximately 10

1354 million years ago, towards the end of the Miocene and the beginning of the Pliocene. This

1355 follows the gradual replacement of forested habitats with grassland in central East Africa,

1356 however, the early-to-mid-Pliocene superseding this global cooling and aridification actually

1357 resulted in hotter and wetter climates in sub-Saharan Africa (deMenocal 2004; Feakins and

1358 deMenocal 2010). This would have increased forest cover, especially in equatorial regions,

1359 which might have fragmented areas of suitable habitat for C. dilepis and this may have favoured

1360 allopatric speciation within the C. dilepis complex.

1361

1362 Ecological speciation typically follows populations diverging into distinct ecological niches;

1363 this ensures that the two populations evolve independently without any gene flow between

1364 them eventually to become separate species (Losos et al. 1998; Losos and Ricklefs 2009).

1365 Accordingly then, closely-related species that are parapatric should be expected to inhabit

1366 different niches. Therefore, ecological niche modelling is a valuable tool to assist in the

1367 validation of species that have been inferred on a phylogenetic basis (Raxworthy et al. 2007;

1368 Blair et al. 2013; da Silva and Tolley 2017). While the models used here are based entirely on

1369 climate and therefore carry the limitation of not accounting for other influences such as

1370 vegetation, elevation, terrain or interspecific competition, they still allow for some general

1371 inferences to be made. Firstly, they show support for the presence of distinct, non-overlapping

1372 climatic niches for each clade (Fig. 4), with unique bioclimate preferences among clades (Table

1373 1). This may have promoted the clades being spatially separate. Secondly, the models suggest

1374 that there was essentially no overlap between niches during the LIG and LGM and so the spatial

1375 separation of the niches appears stable through time, which may maintain their genetic

1376 distinctiveness. If anything, the clades are more connected spatially now than they were during

1377 the late Pleistocene. In the phylogeny (Fig. 2), C. dilepis 3 diverges from the MRCA of the C.

1378 dilepis complex first and this suggests that C. dilepis 1 and 2 were possibly more connected

70 1379 spatially in the past. This is however not reflected by the models, which indicate greater

1380 connectivity between C. dilepis 1 and 3 than between C. dilepis 1 and 2. This suggests that the

1381 more recent (late Pleistocene) paleoclimatic models cannot be used as a proxy to understand

1382 diversification that took place in the late Miocene. Instead, something happened in the late

1383 Miocene that promoted the divergence of C. dilepis 3 from the rest of C. dilepis. One possibility

1384 is geological influences. The Albertine Rift is the western branch of the East African Rift which

1385 runs southward from Lake Albert in Uganda to Lake Malawi in Malawi. This rift is thought to

1386 have started forming as a consequence of volcanic activity in the mid-Miocene (13-12 mya)

1387 (Harðarson 2014). Given that this rift lies at the interface of the distributions of C. dilepis 1

1388 and C. dilepis 3 and that the formation of the rift and the divergence of C. dilepis 3 from the

1389 greater C. dilepis clade are almost coeval (C. dilepis 3 diverged after the rift started forming,

1390 but the rift formed very slowly), it seems plausible that the formation of this rift might have

1391 led to the isolation and subsequent divergence of C. dilepis 3.

1392

1393 As hypothesized, suitable climate space for all three clades showed contractions during the

1394 LIG. This contraction in suitable climate space was least severe for C. dilepis 2 and most

1395 notable for C. dilepis 3. This was expected given the higher latitude of C. dilepis 3 which saw

1396 more arid climates during the LIG than the equatorial C. dilepis 1 and C. dilepis 3 which would

1397 have experienced, hot, wet monsoonal climates during this period (Govin et al. 2014). The

1398 suitable niche of C. dilepis 2 appears the most stable through time, at least during the

1399 paleoclimatic cycling of the Pleistocene.

1400

1401 CONCLUSION:

1402 The results presented here suggest that Chamaeleo originated during the Eocene in the

1403 Southern African and Zambezian bioregions and made use of mesic corridors produced from

1404 Oligocene climatic cooling and central African forest contractions during the Miocene to

71 1405 infiltrate central, East and North Africa, Ethiopia and Eurasia. Additionally, there is evidence

1406 for niche partitioning between clades within the C. dilepis complex. The climatic niche of C.

1407 dilepis 2 appears relatively stable through the late Pleistocene while that of C. dilepis 1 and C.

1408 dilepis 3 appear sensitive to climatic warming as the models predicted shifts or contractions,

1409 respectively. These results lay a foundation for a reassessment of the current distribution of

1410 taxa within the C. dilepis complex. IUCN distribution maps currently exist for the species

1411 complex on the whole, however, given that there appear to be at least three species within this

1412 complex, updated distribution maps will need to reflect this if the taxonomy is revised.

1413

1414 REFERENCES:

1415 Araújo MB, Thuiller W, Pearson RG. 2006. Climate warming and the decline of amphibians

1416 and reptiles in Europe. Journal of Biogeography 33: 1712–1728.

1417 Barlow A, Baker K, Hendry CR, Peppin L, Phelps T, Tolley KA, Wüster CE, Wüster W.

1418 2013. Phylogeography of the widespread African puff adder (Bitis arietans) reveals

1419 multiple Pleistocene refugia in southern Africa. Molecular Ecology 22: 1134–1157.

1420 Bellstedt DU, Galley C, Pirie MD, Linder HP .2012. The migration of the palaeotropical arid

1421 flora: Zygophylloideae as an example. Systematic Botany 37(4):951-959.

1422 Blair ME, Sterling EJ, Dusch M, Raxworthy CJ, Pearson RG. 2013. Ecological divergences

1423 and speciation between lemur (Eulemur) sister species in Madagascar. Journal of

1424 Evolutionary Biology 26: 1790-1801.

1425 Bobe R. 2006. The evolution of arid ecosystems in eastern Africa. Journal of Arid

1426 Environments 66: 564-584.

1427 Bolet A, Evans SE. 2014. Chapter 9: Fossil history of chameleons. In: Tolley KA, Herrel A

1428 (Eds.) The Biology of Chameleons. University of California Press.

1429 Bouckaert R, Heled J, Kühnert D, Vaughan T, Wu CH, Xie D, Suchard MA, Rambaut A,

1430 Drummond AJ. 2014. BEAST 2: A Software Platform for Bayesian Evolutionary

72 1431 Analysis. PLoS Computational Biology 10: 1–6.

1432 Ceccarelli FS, Menegon M, Tolley KA, Tilbury CR, Gower DJ, Laserna MH, Kasahun R,

1433 Rodriguez-Prieto A, Hagmann R, Loader SP. 2014. Evolutionary relationships, species

1434 delimitation and biogeography of Eastern Afromontane horned chameleons

1435 (Chamaeleonidae: Trioceros). Molecular Phylogenetics and Evolution 80: 125–136.

1436 Cracraft J. 1975. Historical biogeography and earth history: perspectives for a future

1437 synthesis. Annals of the Missouri botanical Garden 62: 227–250.

1438 Darriba D, Taboada G, Doallo R, Posada D. 2012. jModelTest 2: more models, new

1439 heuristics and parallel computing. Nature methods 9: 772.

1440 Drummond AJ, Ho SYW, Phillips MJ, Rambaut A. 2006. Relaxed phylogenetics and dating

1441 with confidence. PLoS Biology 4: 699–710.

1442 Elith J, Phillips SJ, Hastie T, Dudík M, Chee YE, Yates CJ. 2011. A statistical explanation of

1443 MaxEnt for ecologists. Diversity and Distributions 17: 43–57.

1444 Feakins SJ, deMenocal PB. 2010. Chapter 4: Global and African regional climate during the

1445 Cenozoic. In: Werdelin L (Ed.) Cenozoic Mammals of Africa. California Scholarship

1446 Online.

1447 Govin A, Varma V, Prange M. 2014. Astronomically forced variations in western African

1448 rainfall (21°N-20°S) during the Last Interglacial period. Geophysical Research Letters

1449 41: 2117-2125.

1450 Harðarson, BS. 2014. Structural geology of the western branch of the East African Rift:

1451 tectonics, volcanology and geothermal activity. Presented at Short Course IX on

1452 Exploration for Geothermal Resources, organized by UNU-GTP, GDC and KenGen, at

1453 Lake Bogoria and Lake Naivasha, Kenya, Nov. 2-23, 2014.

1454 Herrel A, Tolley KA, Measey GJ, da Silva, JM, Potgieter DF, Boller E, Boistel R,

1455 Vanhooydonck B. 2013. Slow but tenacious: an analysis of running and gripping

1456 performance in chameleons. Journal of Experimental Biology 216: 1025-1030.

73 1457 Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A. 2005. Very high resolution

1458 interpolated climate surfaces for global land areas. International Journal of Climatology

1459 25: 1965–1978.

1460 Jongsma GFM, Barej MF, Barratt CD, Burger M, Conradie W, Ernst R, Greenbaum E,

1461 Hirschfeld M, Leaché AD, Penner J, Portik DM, Zassi-Boulou AG, Rödel MO,

1462 Blackburn DC. 2018. Diversity and biogeography of frogs in the genus Amnirana

1463 (Anura: Ranidae) across sub-Saharan Africa. Molecular Phylogenetics and Evolution

1464 120: 274–285.

1465 Kissling WD, Eiserhardt WL, Baker WJ, Borchsenius F, Couvreur TLP, Balslev H, Svenning

1466 J-C. 2012. Cenozoic imprints on the phylogenetic structure of palm species assemblages

1467 worldwide. Proceedings of the National Academy of Sciences 109: 7379–7384.

1468 Kuriyama T, Brandley MC, Katayama A, Mori A, Honda M, Hasegawa M. 2011. A time-

1469 calibrated phylogenetic approach to assessing the phylogeography, colonization history

1470 and phenotypic evolution of snakes in the Japanese Izu Islands. Journal of

1471 Biogeography 38: 259–271.

1472 Linder HP, de Klerk HM, Born J, Burgess ND, Fjeldså J, Rahbek C. 2012. The partitioning

1473 of Africa: Statistically defined biogeographical regions in sub-Saharan Africa. Journal

1474 of Biogeography 39: 1189–1205.

1475 Losos JB, Jackman TR, Larson A, de Queiroz K, Rodrıguez-Schettino L. 1998. Contingency

1476 and determinism in replicated adaptive radiations of island lizards. Science 279: 2115–

1477 2118.

1478 Losos JB, Ricklefs RE. 2009. Adaptation and diversification on islands. Nature 457: 830–6.

1479 Maslin MA, Brierley CM, Milner AM, Shultz S, Trauth MH, Wilson KE. 2014. East African

1480 climate pulses and early human evolution. Quarternary Science Reviews 101: 1-17

1481 Matthee CA, Tilbury CR, Townsend T. 2004. A phylogenetic review of the African leaf

1482 chameleons: Genus Rhampholeon (Chamaeleonidae): The role of vicariance and climate

74 1483 change in speciation. Proceedings of the Royal Society B: Biological Sciences 271:

1484 1967–1975.

1485 Measey GJ, Tolley KA. 2011. Sequential fragmentation of Pleistocene forests in an East

1486 Africa biodiversity hotspot: Chameleons as a model to track forest history. PLoS ONE 6.

1487 deMenocal PB. 2004. African climate change and faunal evolution during the Pliocene-

1488 Pleistocene. Earth and Planetary Science Letters 220: 3-24.

1489 Morley RJ. 2007. Cretaceous and Tertiary climate change and the past distribution of

1490 megathermal forests. In: Tropical Rainforest responses to Climate Change. Springer

1491 Praxis Books. Berlin, Heidelberg: Springer. pp. 1–34.

1492 Nelson G. 1985. A decade of challenge the future of biogeography. Plate Tectonics and

1493 Biogeography 4: 187–196.

1494 Partridge TC, deMenocal PB, Lorentz SA, Paiker MJ, Vogel JC. 1997. Orbital forcing of

1495 climate over South africa: A 200 000-year rainfall record from the Pretoria saltpan.

1496 Quarternary Science Reviews 16: 1125-1133.

1497 Phillips SJ, Dudík M. 2008. Modeling of species distributions with Maxent: new extensions

1498 and a comprehensive evaluation. Ecography 31: 161–175.

1499 Pörtner H. 2001. Climate change and temperature-dependent biogeography: Oxygen

1500 limitation of thermal tolerance in animals. Naturwissenschaften 88: 137–146.

1501 Pörtner HO. 2002. Climate variations and the physiological basis of temperature dependent

1502 biogeography: Systemic to molecular hierarchy of thermal tolerance in animals.

1503 Comparative Biochemistry and Physiology - A Molecular and Integrative Physiology

1504 132: 739–761.

1505 Poynton JC. 1995. The “arid corridor” distribution in Africa: a seacrh for instances among

1506 amphibians. Madoqua 19(1): 45-48

1507 Rambaut A, Drummond AJ, Xie D, Baele G, Suchard MA. 2018. Posterior Summarization in

1508 Bayesian Phylogenetics Using Tracer 1.7. Systematic Biology 67: 901–904.

75 1509 Raxworthy CJ, Ingram CM, Rabibisoa N, Pearson RG. 2007. Applications of ecological

1510 niche modeling for species delimitation: A review and empirical evaluation using day

1511 geckos (Phelsuma) from Madagascar. Systematic Biology 56(6): 907-923

1512 Romano SL, Palumbi SR. 1997. Molecular evolution of a portion of the mitochondrial 16S

1513 ribosomal gene region in scleractinian corals. Journal of molecular evolution 45: 397–

1514 411.

1515 da Silva J, Tolley KA. 2017. Diversification through ecological opportunity in dwarf

1516 chameleons. Journal of Biogeography doi:10.1111/jbi.12966.

1517 Somero GN. 2010. The physiology of climate change: how potentials for acclimatization and

1518 genetic adaptation will determine ‘winners’ and ‘losers’. Journal of Experimental

1519 Biology 213: 912–920.

1520 Tilbury C. 2018. Chameleons of Africa: an atlas: including the chameleons of Europe, the

1521 Middle East and Asia (Illustrate). Frankfurt: Edition Chimaira.

1522 Tierney JE, deMenocal PB, Zander PD. 2017. A climatic context for the out-of-Africa

1523 migration. The Geological Society of America doi:10.1130/G39457.

1524 Tolley KA, Alexander GJ, Branch WR, Bowles P, Maritz B. 2016. Conservation status and

1525 threats for African reptiles. Biological Conservation 204: 63-71.

1526 Tolley KA, Burger M, Turner AA, Matthee CA. 2006. Biogeographic patterns and

1527 phylogeography of dwarf chameleons (Bradypodion) in an African biodiversity hotspot.

1528 Molecular Ecology 15: 781–793.

1529 Tolley KA, Makokha JS, Houniet DT, Swart BL, Matthee CA. 2009. The potential for

1530 predicted climate shifts to impact genetic landscapes of lizards in the South African

1531 Cape Floristic Region. Molecular Phylogenetics and Evolution 51: 120–130.

1532 Tolley KA, Tilbury CR, Measey GJ, Menegon M, Branch WR, Matthee CA. 2011. Ancient

1533 forest fragmentation or recent radiation? Testing refugial speciation models in

1534 chameleons within an African biodiversity hotspot. Journal of Biogeography 38: 1748–

76 1535 1760.

1536 Tolley KA, Townsend TM, Vences M. 2013. Large-scale phylogeny of chameleons suggests

1537 African origins and Eocene diversification. Proceedings of the Royal Society B:

1538 Biological Sciences 280: https://doi.org/10.1098/rspb.2013.0184.

1539 Warren DL, Glor RE, Turelli M. 2008. Environmental niche equivalency versus

1540 conservatism: quantitative approaches to niche evolution. Evolution 62: 2868–2883.

1541 Wu H, Guiot J, Brewer S, Guo Z. 2007. Climatic changes in Eurasia and Africa at the last

1542 glacial maximum and mid-Holocene: reconstruction from pollen data using inverse

1543 vegetation modelling. Climate Dynamics 29: 211-229.

1544 Yu Y, Harris AJ, Blair C, He X. 2015. RASP (Reconstruct Ancestral State in Phylogenies): A

1545 tool for historical biogeography. Molecular Phylogenetics and Evolution 87: 46–49.

77 1546 Appendix III

1547 Figure 1: Current day predictions of suitable climate for three clades of C. dilepis with presence records shown.

78 1548 Appendix IV

1549 Table 1: Dispersal probabilities for different time periods used in the DEC analysis

0-2 mya Eurasia North Africa Congolian Ethiopian/Somalian Sudanian Zambezian Southern

Eurasia 1 0.25 0 0.25 0 0 0 North Africa 0.25 1 0 0.25 1 0.25 0 Congolian 0 0 1 0.25 1 0.25 1 Ethiopian/Somalian

0.25 0.25 0.25 1 1 1 0 Sudanian 0 1 0.25 1 1 0.25 1 Zambezian 0 0.25 0.25 1 0.25 1 1 Southern 0 0 0 0 0 1 1 1550 1551

2-11 mya Eurasia North Africa Congolian Ethiopian/Somalian Sudanian Zambezian Southern

Eurasia 1 0.25 0 0.25 0 0 0 North Africa 0.25 1 0.25 1 1 0.25 0 Congolian 0 0.25 1 0.5 1 1 0 Ethiopian/Somalian

0.25 1 0.25 1 1 0.5 0 Sudanian 0 1 0.5 1 1 0.5 0 Zambezian 0 0.25 1 1 0.5 1 1 Southern 0 0 0 0 0 1 1 1552

79

11-30 mya Eurasia North Africa Congolian Ethiopian/Somalian Sudanian Zambezian Southern

Eurasia 1 0.1 0 0.25 0 0 0 North Africa 0.1 1 0.5 0.5 1 0.25 0 Congolian 0 0.5 1 1 1 1 0 Ethiopian/Somalian

0.25 0.5 1 1 1 1 0 Sudanian 0 1 1 1 1 0.5 0 Zambezian 0 0.25 1 1 0.5 1 1 Southern 0 0 0 0 0 1 1 1553 1554

30-50 mya Eurasia North Africa Congolian Ethiopian/Somalian Sudanian Zambezian Southern

Eurasia 1 0.1 0 0.1 0 0 0 North Africa 0.1 1 1 1 1 0.25 0 Congolian 0 0 1 1 1 1 0 Ethiopian/Somalian

0.1 1 1 1 1 1 0 Sudanian 0 1 1 1 1 0.75 0 Zambezian 0 0.25 1 1 0.75 1 1 Southern 0 0 0 0 0 1 1 1555

80 1556 Chapter 4

1557

1558 General Conclusion:

1559

1560 The underlying premise of this dissertation was the driving need for a molecularly informed

1561 revision of the taxonomy of the Chamaeleo genus. Large, overlapping distributions and

1562 puzzling phenotypic variation between certain species (C. dilepis, C. gracilis, C. anchietae, C.

1563 necasi, C. senegalensis, and C. laevigatus in particular) have confounded the group’s

1564 morphological taxonomy. Tilbury and Tolley’s (2009) reappraisal of Chamaeleo, which

1565 resulted in the elevation of the sub-genus Trioceros to its own genus, highlighted a paucity in

1566 our understanding of the evolutionary history of these taxa. Prior to this dissertation, few

1567 studies had investigated the systematics of members of Chamaeleo molecularly, however,

1568 many authors had commented on the taxonomic quandary of the C. dilepis complex (Klaver

1569 and Bohme 1986; Klaver and Bohme 1997; Largen and Spawls 2006), and as Tilbury (2018)

1570 put it, “Chamaeleo dilepis single-handedly demonstrates the limitations of a taxonomic system

1571 which relies heavily on external morphology and the influences on this of ontogenic change,

1572 sexual dimorphism, altitudinal effects on phenotype and simple intraspecific variation on

1573 morphology.”. Main et al. (2018) found large intraspecific differences between C. dilepis

1574 clades in South Africa and suggested that there might be cryptic species within the C. dilepis

1575 complex. Main et al. (2018)’s findings, however, did not allow for conclusions to be made

1576 regarding species as the study was preliminary in nature and did not include sampling from

1577 across the range of C. dilepis and did not include all species in the genus.

1578

1579 Accordingly, the first aim of this dissertation was to carry out a genus-wide molecular

1580 phylogeny of Chamaeleo, including all species representatives and sampling across the natural

1581 ranges of most species in order to disentangle the evolutionary relationships within the group.

81 1582 Furthermore, a second aim was to investigate cryptic speciation in the genus using two model-

1583 based species delimitation methods (bGMYC and BPP) and distance based species delimitation

1584 as the larger sampling complement of this genus-wide phylogeny would help overcome the

1585 limitations of Main et al. (2018). These two aims underpinned Chapter 2 of this dissertation.

1586 Not surprisingly, pervasive cryptic diversity was uncovered, with the phylogenetic analysis

1587 revealing 18 species in contrast to the currently recognised 14 species within Chamaeleo. In

1588 particular, cryptic diversity was evident in three species with C. anchietae comprising the

1589 nominate species as well as C. anchietae vinckei which, pending a morphological re-

1590 evaluation, should be elevated to specific status, C. dilepis comprising the nominate species as

1591 well as two additional candidate species and C. gracilis also comprising the nominate species

1592 and two additional candidate species. Twelve of the 14 recognised species were shown to be

1593 monophyletic, however, C. necasi shares a terminal lineage with the nominate form of C.

1594 gracilis, putting the current taxonomy of C. necasi into question and suggesting that it should

1595 be synonymised with C. gracilis. Furthermore, C. senegalensis and C. laevigatus were

1596 recovered as independent species, putting to rest the debate pertaining to their taxonomic status

1597 (Joger 1990; Chirio and LeBreton 2007; Trape et al 2012; Tilbury 2018). Chamaeleo

1598 namaquensis was recovered as the sister taxon to the rest of Chamaeleo and a Eurasian clade

1599 of Chamaeleo was found to be monophyletic.

1600

1601 The large, imbricating ranges of Chamaeleo, together with an apparent tolerance of more arid

1602 climates (compared to other chameleon genera) and an affinity for mesic savannah (Tilbury

1603 2018) probably contributed to the widespread cryptic diversity within the group, with

1604 prevailing morphological conservatism making it difficult to discern all species

1605 morphologically. Understanding phylogenetic relationships and uncovering cryptic diversity

1606 on the genus scale were just two pieces of a complex evolutionary puzzle. Accordingly,

1607 unravelling the biogeographic history and investigating the geo-climatic drivers behind the

82 1608 evolution of Chamaeleo became a key focus of the third chapter of this dissertation. An

1609 ancestral area reconstruction of the genus was made in order to test a hypothesis that the genus

1610 originated in Southern Africa. Additionally, ecological niche modelling was implemented for

1611 the C. dilepis clades delimited in Chapter 2 in order to determine whether there exists

1612 ecological support for the clades as distinct species.

1613

1614 Chamaeleo was found to have likely diverged from other genera in Southern Africa or

1615 Zambezia bioigeographic regions approximately 50 mya, a period that corresponds loosely

1616 with the early Eocene climatic optimum which was characterised by hotter, wetter climates

1617 globally. This supported the hypothesis that Chamaeleo originated in Southern Africa. Internal

1618 diversification within the genus was shown to have initiated approximately 39 mya, towards

1619 the end of the Eocene which coincides broadly with a period of global cooling (Morley 2007;

1620 Kissling et al. 2012). Chamaeleo namaquensis, the only chameleon occupant of true desert,

1621 was found to branch off from the rest of Chamaeleo first, with intensified southwest African

1622 aridity during the Oligocene and Miocene presumably promoting its independent

1623 diversification from other species in the genus. The next species to diverge from the rest of

1624 Chamaeleo was C. anchietae, approximately 29 mya in the Oligocene, and this species likely

1625 made use of an arid corridor that opened up during the Miocene (Bobe 2006) to disperse north-

1626 eastwards from southwest Africa. Eurasian Chamaeleo species were shown to have dispersed

1627 out of Africa during the early Miocene, 23 mya. This period coincides with globally warmer

1628 temperatures but a replacement of forest with grasslands in East Africa which, due to their

1629 mesic habitat preference, would have opened corridors for the dispersal of Chamaeleo. Most

1630 species-level diversification in Chamaeleo was found to have occurred during the mid-to-late

1631 Miocene (15-10 mya). Chamaeleo dilepis was found to diversify in the late Miocene and early

1632 Pliocene, as with the diversification of other species in the genus, this diversification appears

1633 to have followed periods of cooling and aridity resulting in the replacement of forest by

83 1634 grasslands. For C. dilepis, the period following its diversification (the mid-Pliocene; 5-3 mya)

1635 was hot and wet (deMenocal 2004; Feakins and deMenocal 2010), and would have favoured

1636 the expansion of forests. These forests might have fragmented C. dilepis and a lack of gene

1637 flow between these fragmented populations might have promoted allopatric speciation with the

1638 complex. It is also possible that the formation of the Albertine Rift was responsible for the

1639 isolation of C. dilepis 3 from the rest of the species complex.

1640

1641 Ecological niche modelling revealed that each clade of C. dilepis had its own bioclimatic

1642 preference and occupied climatically unique areas. Minimal overlap was evident for suitable

1643 climate space between clades, even during the LGM and LIG. This supported the status of

1644 these clades as independent species. Given that C. dilepis, diversified at the end of the Miocene,

1645 several million years prior to the LIG and LGM, climate models for these time periods could

1646 not be used to unravel finer patterns pertaining to the diversification of C. dilepis, however, the

1647 suitable climate for these clades appears to have been even less connected during the LIG and

1648 LGM than it is now. Because the LIG and LGM represent the most extreme climates

1649 experienced during paleoclimatic cycling of the Pleistocene, it seems plausible to suggest that

1650 barriers to gene flow have persisted since the LIG. The suitable climatic niche of C. dilepis 2

1651 was found to be the most stable through time, while that of C. dilepis 3 was found to be the

1652 least stable through time, fragmenting into small pockets during the LIG.

1653

1654 The findings of this dissertation have shed light on the evolutionary history and taxonomy of

1655 Chamaeleo, uncovering pervasive cryptic diversity and suggesting that the group’s

1656 diversification tracked historical climatic cooling and aridification in Africa which promoted a

1657 contraction of forests and an expansion of savannah and grasslands. Perhaps most importantly,

1658 this dissertation provides a foundation from which to re-evaluate both the taxonomy and the

1659 known distribution of various members of the genus. One of which, C. dilepis, has caused

84 1660 notable confusion among taxonomists in the past, and these findings will help ameliorate much

1661 of this confusion. This dissertation does however carry some limitations, for certain species (C.

1662 gracilis 1 and 2; C. anchietae 2) there is limited sampling coverage. Additionally, niche

1663 modelling was constrained by the use of only climatic data and thus the inclusion of vegetation,

1664 elevation or terrain data may allow for more confident inferences to be made. These limitations

1665 provide ample scope for further research on this enigmatic genus of reptiles. Greater sampling

1666 coverage in the taxa poorly sampled here might even reveal more cryptic diversity within the

1667 genus.

1668

1669 REFERENCES:

1670 Bobe R. 2006. The evolution of arid ecosystems in eastern Africa. Journal of Arid

1671 Environments 66: 564-584.

1672 Chirio L, LeBreton W. 2007. Atlas des reptiles du Cameroun. MNHN, IRD, Paris.

1673 Feakins SJ, deMenocal PB. 2010. Chapter 4: Global and African regional climate during the

1674 Cenozoic. In: Werdelin L (Ed.) Cenozoic Mammals of Africa. California Scholarship

1675 Online.

1676 Joger U. 1990. The herpetofauna of the Central African Republic, with the description of a

1677 new species of Rhinotyphlops (Serpentes: Typhlopidae), pp. 85-102. In: Peters G,

1678 Hutterer R (Eds.) Vertebrates in the tropics. Museum Alexander Koenig, Bonn.

1679 Kissling WD, Eiserhardt WL, Baker WJ, Borchsenius F, Couvreur TLP, Balslev H, Svenning

1680 J-C. 2012. Cenozoic imprints on the phylogenetic structure of palm species assemblages

1681 worldwide. Proceedings of the National Academy of Sciences 109: 7379–7384.

1682 Klaver CJJ, Bohme W. 1986. Phylogeny and classification of the Chamaeleonidae (Sauria)

1683 with special reference to hemipenis morphology. Bonner zoologische Monographien

1684 22:1-64.

1685 Klaver CJJ, Bohme W. 1997. Chamaeleonidae. In: Klaver CJJ, Bohme W (Eds.) Das

85 1686 Tierreich. Berlin: Walter de Gruyter.

1687 Largen MJ, Spawls S. 2006. Lizards of Ethiopia (Reptilia Sauria): An annotated checklist,

1688 bibliography, gazetteer and identification key. Tropical Zoology 19: 21–109.

1689 Main DC, Jansen van Vuuren B, Tolley KA. 2018. Cryptic diversity in the common flap-

1690 necked chameleon Chamaeleo dilepis in South Africa. African Zoology 53: 11–16.

1691 deMenocal PB. 2004. African climate change and faunal evolution during the Pliocene-

1692 Pleistocene. Earth and Planetary Science Letters 220: 3-24.

1693 Morley RJ. 2007. Cretaceous and Tertiary climate change and the past distribution of

1694 megathermal forests. In: Tropical Rainforest responses to Climate Change. Springer

1695 Praxis Books. Berlin, Heidelberg: Springer. pp. 1–34.

1696 Tilbury C. 2018. Chameleons of Africa: an atlas: including the chameleons of Europe, the

1697 Middle East and Asia (Illustrate). Frankfurt: Edition Chimaira.

1698 Tilbury CR, Tolley KA. 2009. A re-appraisal of the systematics of the African genus

1699 Chamaeleo (Reptilia: Chamaeleonidae). Zootaxa 2079: 57–68.

1700 Trape J-F, Trape S, Chirio L. 2012. Lezards, crocodiles et tortues d’Afrique occidentale et du

1701 Sahara. IRD editions, Marseilles.

86