A morphological and molecular review of the genus Goniurosaurus, including an identification key

1,2 Vietnam National Museum of Nature, Vietnam Academy of Science and Technology, 18 Hoang Quoc Viet Road, Hanoi, Vietnam. 2,8 Graduate University of Science and Technology, Vietnam Academy of Science and Technology, 18 Hoang Quoc Viet Road, Hanoi, Vietnam. 3,8 Institute of Ecology and Biological Resources, Vietnam Academy of Science and Technology, 18 Hoang Quoc Viet Road, Hanoi, Vietnam. 4 Department of Genetics, Faculty of Biology, University of Science, Vietnam National University Hanoi, 334 Nguyen Trai Road, Hanoi, Vietnam. 4,5 Central Institute for Natural Resources and Environmental Studies, Vietnam National University Hanoi, 19 Le Thanh Tong, Hanoi, Vietnam. 5 Department of Environmental Ecology, Faculty of Environmental Sciences, University of Science, Vietnam National University Hanoi, 334 Nguyen Trai Road, Hanoi, Vietnam. 5 Department of Herpetology, American Museum of Natural History, Central Park West at 79th Street, New York, USA. 1,6,9 Institute of Zoology, University of Cologne, Zülpicher Strasse 47b, 50674 Cologne, Germany. 1,6,9 Cologne, Germany Cologne Zoo, Riehler Straße 173, 50735 Cologne, Germany. 7 Federal Agency for Nature Conservation, CITES Scientific Authority, Konstantinstrasse 110, 53179 Bonn, Germany.


Introduction
The eublepharid genus Goniurosaurus Grismer, Viets & Boyle, 1999 (tiger geckos) currently consists of 23 species associated with granitic or karst formations in Southeast and East Asia. This genus shows a high level of local endemism and most tiger gecko species have been only recorded from a single locality, within unique ecosystems or an isolated archipelago (Orlov et al. 2008;Ziegler et al. 2008;Yang & Chan 2015;Honda & Ota 2017;Zhou et al. 2018;Ngo et al. 2019a;Qi et al. 2020aQi et al. , 2020bZhu et al. 2020aZhu et al. , 2020b. The complex topography and variable climatic conditions throughout the distribution range of the genus may account for the disjunct distribution and occupation of different ecological niches by distinct populations. These factors may have driven natural selection, morphological diversification and phylogenetic evolution within this genus (Vitt et al. 1997;Clements et al. 2006;Sexton et al. 2009;Gomes et al. 2016;Liang et al. 2018;Qi et al. 2020aQi et al. , 2020b. According to recent morphological and molecular analyses, the genus Goniurosaurus is split into four species groups, namely the G. lichtenfelderi group with five species from both mainland and islands in China and Vietnam; the G. kuroiwae group containing six species from the Ryukyu Archipelago, Japan; the G. luii group comprising eight species distributed throughout islands and mainland in Vietnam and China; and the G. yingdeensis group consisting of four species from China (Nguyen et al. 2009;Nguyen 2011;Wang et al. 2013;Honda & Ota 2017;Liang et al. 2018;Qi et al. 2020aQi et al. , 2020bZhu et al. 2020aZhu et al. , 2020b. However, the systematics of the genus Goniurosaurus remains challenging due to ongoing discoveries of further cryptic species and the fact that there is not a single genetic marker that covers all 23 species, precluding a complete generic phylogeny of the group. Five subspecies of the G. kuroiwae group from Japan were reinstated at full species status and a new species, G. sengokui (Honda & Ota, 2017), was recently discovered from the Ryukyu Archipelago (Honda & Ota 2017). In China, two taxa, previously identified as G. luii , were described as distinct species, namely G. kadoorieorum Yang & Chan, 2015and G. kwangsiensis Yang & Chan, 2015(Yang & Chan 2015, and a sister species of G. araneus , namely G. gezhi Zhu, He & Li, 2020, was recently discovered (Zhu et al. 2020a). These three species occur in Southwest Guangxi Province (Zhu et al. 2020a). Four additional species, namely G. kwanghua He, 2020 andG. zhoui Zhou, Wang, Chen &, belonging to the G. lichtenfelderi group, as well as G. gollum Qi, Wang, Grismer, Chen, Lyu &Wang, 2020 andG. varius Qi, Grismer, Lyu, Zhang, Li &Wang, 2020 of the G. yingdeensis group, were recently described from China Qi et al. 2020aQi et al. , 2020bZhu et al. 2020b).
Due to habitat degradation and over-harvesting for the pet trade, wild populations of species of Goniurosaurus from China and Vietnam have been subject to severe declines (Stuart et al. 2006;Yang & Chan 2015;Ngo et al. 2019b). Therefore, all Chinese and Vietnamese tiger gecko species have recently been included in CITES (Convention on International Trade in Endangered Species of Wild Fauna and Flora) Appendix II (CITES 2019; Ngo et al. 2019b). To effectively enforce relevant national and international regulations and to enable sustainable trade in CITES listed species, a detailed guideline for species identification is an essential prerequisite, especially as trade in the morphologically similar species from Japan is currently regulated under a lower protection level, namely CITES Appendix III (CITES Notification No. 2020/068). A proper identification guidance will help to prevent certain taxa from being traded under a wrong name to circumvent trade restrictions.
In this study, phylogenetic analyses were performed across all recorded populations of Goniurosaurus from Vietnam and along with other species from China and Japan using a fragment of the mitochondrial Cytochrome b gene. Based on newly collected data from Vietnam, we expanded morphological descriptions of each species and evaluated intra-and inter-specific morphological variations among the species of Goniurosaurus. We further compared morphological data of Vietnamese tiger geckos with those from China and identified characters that can distinguish the Japanese species. We thereby provide an identification key for all Goniurosaurus species from China, Japan and Vietnam, in order to assist CITES enforcement authorities in species identification.

Field surveys
Field surveys were conducted in June and August 2014, May 2015, July 2016, June and July 2017, April 2018, from April to September 2019, as well as in June and July 2020 in Vietnam. Surveys were conducted for G. catbaensis in Hai Phong City and Quang Ninh Province, for G. huuliensis in Lang Son Province, for G. lichtenfelderi in Hai Duong, Bac Giang and Quang Ninh provinces, and for G. luii in Cao Bang and Lang Son provinces. Animals were captured by hand and subsequently released at the same sites after taking measurements and photos in life. Coordinates of each captured individual were recorded using a GPS device (Garmin GPSmap64, WGS84 datum) and can be shared upon request to the authors.
In addition, several vouchered specimens and small tissue samples of Goniurosaurus deposited in collections of the Institute of Ecology and Biological Resources (IEBR), Vietnam National Museum of Nature (VNMN) and Hanoi University of Science (HUS), Hanoi, Vietnam, were examined for morphological characters and molecular phylogeny.

Molecular data and phylogenetic analyses
All taxa of the genus Goniurosaurus from China, Japan, and Vietnam, for which Cytochrome b data were available in GenBank, were included in the study. Three species, Gekko chinensis (Gray, 1842) (NC027191), Coleonyx mitratus Peters, 1863 (AB853481) and Holodactylus africanus Boettger, 1893 (AB853482), were used as outgroups following Honda et al. (2014).
Small tissue samples were collected from the tail tip of 18 wild specimens from representative localities in Vietnam and preserved separately in 70% ethanol (Merck, Germany). The mitochondrial DNA Cytochrome b (cytb) gene was selected for DNA sequencing. The total genomic DNA was extracted using GeneJet Genomic DNA Purification (ThermoFisher Scientific, Lithuania), following protocols by the manufacturer's instructions. PCR reactions were performed using HotStar Taq Mastermix (Qiagen, Germany) to amplify a fragment of approximately 1000 bps of cytb. The primer pair used for this study was L14731 (5'-TGGTCTGAAAAACCATTGTTG-3'; Honda et al. 2014) and GoniR1 (5'-CTACGGGCTGTCCTCCGATTCAGGTT-3'; this study). The PCR volume consisted of 21 µl: 2 µl of each primer, 5 µl water, 10 µl of Taq mastermix and 1-2 µl DNA template depending on the quality of DNA in the final extraction solution. The PCR was performed at 95 o C for 15 minutes, followed by 35 cycles, for 30 s at 95 o C, 45 s at 48 o C and 60 s at 72 o C with a final elongation step for 6 minutes at 72 o C. Negative and positive controls were used for all DNA extractions and PCR reactions.
PCR products were visualized using electrophoresis through a 1% agarose gel, marker 1kb, 1X TBE, stained with ethidium bromide and photographed under UV light. Successful amplifications were purified using GeneJet PCR Purification Kit (ThermoFisher Scientific, Lithuania). Cleaned PCR products were sent to 1 st Base (Malaysia) for sequencing in both directions.
The obtained sequences were aligned in Sequencher ver. 5.4 (Gene Codes Corp, Ann Arbor, MI, USA) and afterwards aligned using ClustalX ver. 2.1 (Thompson et al. 1997) with default settings. Data were analyzed using maximum parsimony (MP) as implemented in PAUP * 4.0b10 (Swofford 2001) and Bayesian inference (BI) as implemented in MrBayes ver. 3.2 (Ronquist et al. 2012). Settings for these analyses followed Le et al. (2006), except that the number of generations in the Bayesian analysis was increased to 1 × 10 7 . For the maximum likelihood (ML) analysis, we used IQ-TREE ver. 1.6.7.1 (Nguyen et al. 2015) with a single model and 10 000 ultrafast bootstrap replications. The optimal model for nucleotide evolution was set to TPM1uf+I+G as selected by Jmodeltest ver. 2.1.4 (Posada et al. 1998). As the phylogenetic relationships between species and species groups were well resolved and virtually every important node received high statistical support from all analyses, we opted not to partition our data by codon positions.
For the Bayesian analysis, we used the optimal model determined by Jmodeltest with parameters estimated by MrBayes ver. 3.2.7. Four Markov chains, one cold and three heated utilizing default heating values, were sampled every 1000 generations. Log-likelihood scores of sample points were plotted against generation time to detect stationarity of the Markov chains. The burn-in value was set to 50 in the BI analysis, as -lnL scores reached stationarity after 50 000 generations in both runs. Nodal support was evaluated using Bootstrap replication (BP) as estimated in PAUP * 4.0b10 and ultrafast BP in IQ-TREE ver. 1.6.7.1 and posterior probability (PP) in MrBayes ver. 3.2. BP ≥ 70 (Hillis & Bull 1993) and ultrafast BP (UBP) and PP ≥ 95% are regarded as strong support for a clade (Ronquist et al. 2012;Nguyen et al. 2015). The uncorrected pairwise distance (p) were calculated in PAUP * 4.0b10.

Morphological analyses
A total of 486 live individuals and 54 museum specimens of four species from Vietnam were examined for morphological data, comprising 194 individuals of G. catbaensis (21 juveniles, 93 females and 80 males), 80 individuals of G. huuliensis (02 juveniles, 46 females and 32 males), 178 individuals of G. lichtenfelderi (14 juveniles, 72 females and 92 males), and 88 individuals of G. luii (11 juveniles, 43 females and 34 males). Detailed descriptions of each species were based on newly collected data combined with previous literature for the morphological variation, such as for G. catbaensis from Ziegler et al. (2008) and Nguyen (2011), for G. huuliensis from Orlov et al. (2008) and Nguyen (2011), for G. lichtenfelderi from Grismer (2000), Grismer et al. (2002) and Nguyen (2011), and for G. luii from Grismer et al. (1999), Vu et al. (2006) and Nguyen (2011). No specimen of G. araneus was investigated in this study, however, we included morphological data of eight specimens from the type locality in Cao Bang Province, Vietnam and Guangxi Province, China from the literature Chen et al. 2014). The sex of each collected specimen was determined based on the presence (in males) or absence (in females) of large swollen hemipenial bulges. Lizards were categorized into two age classes based on the snout-vent length (G. lichtenfelderi: SVL < 80 mm = juveniles, SVL ≥ 80 mm = adults, while three other species of Goniurosaurus were sorted with SVL < 85 mm = juveniles, SVL ≥ 85 mm = adults).
Measurements were taken with dial calipers to the nearest 0.1 mm at the right side of each individual.
Abbreviations AD = diameter of auditory meatus AG = axilla to groin length, from posterior edge of forelimb insertion to anterior edge of hindlimb insertion BH = maximum body height, from top of dorsal body to belly BW = maximum body width, greatest width of torso, taken at level of midbody CH = cheek height, from posterior edge of labial to top of head at parietal region ED = diameter of eye, greatest diameter of orbit EE = eye to ear distance, from posterior margin of eye to posterior margin of ear FLL = forelimb length, from axilla to the tip of the fourth finger HH = maximum head height HL = head length, from the tip of snout to posterior edge of occiput HLL = hindlimb length, from groin to the tip of the fourth toe HW = maximum head width IO1 = interorbital distance, distance between anteriormost points of eyes IO2 = interorbital distance, distance between posteriormost points of eyes ML = mouth length, from tip of snout to last posterior labial edge MW = mouth width, distance between last posterior labial edges on each side ND = supranasal distance, distance between nares SE = snout to eye distance, measured from tip of snout to anteriormost point of eye SVL = snout-vent length, from tip of snout to vent TaL = tail length, from vent to tip of tail WT = maximum tail width Shapiro-Wilk's test was used to test the assumption of normality. Kruskal-Wallis's test was performed to determine differences concerning the SVL of sex and age classes among species, as well as the extracted PC1 and PC2 values (mentioned below) of a principal component analysis among four tiger gecko species in Vietnam except of G. araneus. For both tests we applied p = 0.05. The SVL variable was excluded from subsequent analyses due to high collinearity with other dimensions. We further performed a Principal Component Analysis (PCA) of log 10 -transformed raw data of 18 remaining morphometric characters (except TaL and WT variables, due to many regenerated tails), using the packages "factoextra" and "FactoMinerR" to detect variances among the four selected Vietnamese species via contribution percentage of the PC1 and other PCs scores (Kassambara 2017). Morphometric variation of each species was illustrated by representatively clustered ellipse spaces with different coded-color within a spatial coordinate of the first two most important dimensions (PC1 and PC2) in the PCA analysis, to visually evaluate the overlap among the four species. In addition, we identified the overall difference in meristic characters among the 17 tiger gecko species native to China and Vietnam assigned to three groups (G. lichtenfelderi, G. luii and G. yingdeensis) by using a Multiple Correspondence Analysis (MCA). The meristic variation among the three groups was also visualized by convex ellipses within a spatial coordinate of the first two most important dimensions (Dim1 and Dim2) in the MCA analysis.

Molecular phylogeny
The final matrix consisted of 949 bp aligned characters with 48 sequences of 14 ingroup and three outgroup taxa. The alignment contained no gaps. In total, 404 characters were found to be parsimony informative. The MP analysis produced 624 most parsimonious trees with 1404 steps (CI = 0.54, RI = 0.85). Our phylogenetic analyses recovered a generally similar topology to those reported by Liang et al. (2018) and Zhu et al. (2020aZhu et al. ( , 2020b. Specifically, the Goniurosaurus lichtenfelderi group and the G. luii group form a monophyletic group with high nodal values from all analyses and the G. kuroiwae group is sister to the other two. In addition, each species group was also strongly corroborated as monophyletic from all analyses ( Fig. 1). Within each species group, the results supported by this study are more similar to those generated by Liang et al. (2018), because Zhu et al. (2020a) used a different mitochondrial marker, the 16S ribosomal RNA gene. Nonetheless, we found several noted discrepancies between the two studies. Specifically, our phylogenetic analyses showed that G. kuroiwae (Namiye, 1912) is polyphyletic with high nodal support for each of the clades. In addition, G. luii was strongly corroborated as monophyletic in our Bayesian and MP analyses (BP = 93%; PP = 98%), but only weakly recovered as monophyletic in Liang et al. (2018) (BP = 67%; PP = 64%). Goniurosaurus lichtenfelderi was recovered as monophyletic with perfect nodal support from all analyses in our study, but G. hainanensis Barbour, 1908 is polyphyletic. On the other hand, both species are not monophyletic in Liang et al. (2018). The positions of G. zhoui and G. bawanglingensis Grismer, Haitao, Orlov & Anajeva, 2002 are interchanged in the trees supported by the two studies with strong nodal support from Liang et al. (2018) and weak corroboration from ours. All surveyed sub-populations of Goniurosaurus from Vietnam were placed in four separate clades corresponding to four distinct species, G. catbaensis, G. huuliensis, G. luii and G. lichtenfelderi with significantly statistical support values, except for the ML analysis of G. luii (UBP = 88%). Intra-specific genetic divergences of these species were relatively low: G. catbaensis (0.7-1.6%), G. huuliensis (0.4-1.2%), G. lichtenfelderi (0-0.7%), and G. luii (0-1.9%) ( Fig. 1; Supp. file 1: Table S1).

Remarks
During our field work in northern Vietnam, including the type locality in Cao Bang Province, no specimen of Goniurosaurus araneus was seen. Therefore, the diagnosis was solely based on the descriptions of Grismer et al. (1999) and Chen et al. (2014). Ziegler, Nguyen, Schmitz, Stenke & Rösler, 2008 Fig. 2

Diagnosis
Body splayed and gracile, SVL 89.3-125.3 mm; external nares bordered by 6-8 nasal scales; supraorbital region with a row of enlarged tubercles; outer surface of upper eyelid composed of granular scales, about the same size of those on top of head and with a row of 6-10 enlarged tubercles; internasals absent; supralabials 8-11; infralabials 7-10; eyelid fringe scales 45-56; postmentals 2-5; gular region below lower jaws without enlarged tubercles; paravertebral tubercles 31-38; scale rows around midbody 112-127, granular scales surrounding tubercles 8-11; axillary pockets deep; subdigital lamellae under fourth toe 22-25; precloacal pores in males 16-23; iris orange-brown; dorsal ground color of head, body and limbs gray-brown to pale brown and mottled with dark brown blotches; nuchal loop thin, posteriorly protracted (in V-shape); dorsal body bands between limb insertions 3-4, thin, yellow, without dark spots; light band on tail base in width of 8-9 scales; ground color of tail black, caudal bands 5, white, completed ventrally; ventral surface of head, body and limbs dull white and immaculate, gular region with brown spots (modified after Ziegler et al. 2008;Nguyen 2011 Table S2); head triangular, wider than neck, covered by uniform granular scales interspersed with tubercles in temporal and occipital regions; scales on rostrum slightly larger and flatter; enlarged supraorbital tubercles in a conspicuous row; middorsal portion of rostral partially sutured dorsomedially, bordered laterally by first supralabial on each side, dorsolaterally by prenasal on each side, and dorsally by two supranasals; internasal (postrostral) scales absent; external nares bordered by 6-8 nasals: anteriorly by prenasal and supranasal, dorsally by supranasal, posteriorly by two slightly enlarged postnasals and 1 or 2 smaller granular scales, and ventrally by prenasal; prenasals with long recurved ventral portion; supranasals triangular, meeting in midline behind rostral suture; preorbital scales 10-13; supralabials 8-11; infralabials 7-10; eyes relatively large, pupils vertical; eyelid fringe scales 45-56, those of upper eyelid slightly enlarged; outer surface of upper eyelid composed of granular scales of about the same size of those on top of head, including a row of 6-10 enlarged tubercles; fold of skin originating from suborbital region extends posteroventrally across angle of jaw; external auditory meatus elliptical with long axis directed dorsoventrally; tympanum deeply recessed; mental triangular, bordered laterally by first infralabial on each side and posteriorly by 2-5 postmentals; postmentals bordered by 6-10 gular scales; gular region below lower jaws without enlarged tubercles; gular scales juxtaposed and granular, abruptly grading posteriorly into flat hexagonal pectoral scales and even larger ventral scales.
Neck narrower than body, covered with uniform granular scales interspersed with several sharply pointed conical tubercles on nape; tubercles on flanks conical, those of vertebral region somewhat lower in profile; dorsal body tubercles surrounded by 9-11 granular scales; dorsal tubercle rows at midbody 19-24; paravertebral tubercles between limb insertions 31-38, distinct vertebral row of tubercles absent; scales around midbody 112-127; larger ventral scales grade abruptly into smaller granular scales immediately anterior to vent at level of preanal pores; males with 18-23 precloacal pores in a transverse continuous series; region posterior to vent covered by flat juxtaposed scales and great hemipenial bulges, containing 2 or 3 enlarged postcloacal tubercles laterally on each side at level of vent; tail long and thin, thick at base, anteriorly with whorls; dorsal tail scales flat, smooth, up to 1.5 times the size of dorsal scales, arranged in more or less regular transverse rows; light band on tail base in width of 8-9 granular scales and with 7-8 tubercles in a transversal series; subcaudals larger than those on dorsal surface of tail.
Limbs relatively long and thin, covered dorsally with granular scales interspersed with several tubercles and ventrally with flat juxtaposed to subimbricate scales; dorsal granular scales grade into slightly flattened subimbricate scales on top of pes and manus; hind limbs slightly larger than forelimbs; larger granular scales on ventral surface of pes and manus; axillary pockets deep; subdigital lamellae wide, 9-11 under first finger, 18-21 under fourth finger, 9-12 under first toe, 22-25 under fourth toe; digits laterally compressed, increasing in length from first to fourth, fifth shorter than fourth (modified after Ziegler et al. 2008;Nguyen 2011).

Coloration in life (Fig. 2)
Dorsal ground color of head, body and limbs grey brown (in animals with SVL 53.4-105 mm) and blotches chestnut brown (in animals with SVL 110-125.3 mm), juveniles without small blotches, adults mottled with few circular blotches on body and limbs and long dark brown blotches on head; few yellow conical tubercles on franks of neck and body, and limbs in adults; iris light orange or red brown; five bands on the dorsal ground, thin, immaculate without dark spots, cream in juveniles and light orange or yellow in adults, all edged anteriorly and posteriorly by thin dark brown bands, including one thin nuchal loop extending from posterior corners of eyes and posteriorly protracted (in V-shape), three body bands between limb insertions, and another one on tail base; ground color of tail dark brown, and grey brown at mottled tail base; 3-5 immaculate white caudal bands, edged anteriorly and posteriorly in black; ventral surface of head, body and limbs dull white, juveniles absolutely immaculate, but adults with few dark spots on limbs, weak brown lateral spotting in gular region, venter and limbs. Orlov, Ryabov, Nguyen, Nguyen & Ho, 2008 Fig. 3 Diagnosis Body splayed and gracile, SVL 97.2-134.6 mm; external nares bordered by 6-8 nasal scales; supraorbital region with a row of enlarged tubercles; outer surface of upper eyelid composed of granular scales, about one half the size of those on top of head and without enlarged tubercles; internasal 1 or 2 (rarely absent); supralabials 9-12; infralabials 9-12; preorbital scales 14-20; eyelid fringe scales 51-59; postmentals 2-4; gular region below lower jaws with enlarged tubercles; paravertebral tubercles 31-37; scale rows around midbody 118-130, granular scales surrounding tubercles 11-13; axillary pockets deep; subdigital lamellae under fourth toe 21-25; precloacal pores in males 25-30; iris reddish brown; dorsal ground color of head, body and limbs dark brown, without small dark brown blotches (dark blotches present only on lower zone of flanks); nuchal loop thin, posteriorly protracted (in V-shape); dorsal body bands between limb insertions 3, thin, immaculate yellow; gular region with brown spots (modified after Orlov et al. 2008;Nguyen 2011 Table S2); head triangular, wider than neck, covered by uniform granular scales interspersed with tubercles in temporal and occipital regions; scales on rostrum slightly larger and flatter; enlarged supraorbital tubercles in a conspicuous row; middorsal portion of rostral partially sutured dorsomedially, bordered laterally by first supralabial on each side, dorsolaterally by prenasal on each side, and dorsally by internasal and two supranasals; internasal 1-3 (rarely absent); external nares bordered by 5-7 nasals (prenasal, anterior and posterior supranasals, 2 slightly enlarged postnasal, and 1 or 2 granular scales); preorbital scales 14-20; supralabials 9-12, grading into granular scales posteriorly; infralabials 9-12; eyes large, pupils vertical; eyelid fringe scales 51-59, those of upper eyelid slightly enlarged; outer surface of upper eyelid composed of granular scales of about one half the size of those on top of head, without enlarged tubercles; fold of skin originating from suborbital region extends posteroventrally across angle of jaw; external auditory meatus elliptical with long axis directed dorsoventrally; tympanum deeply recessed; mental triangular, bordered laterally by first infralabial on each side and posteriorly by 2-4 postmentals; postmentals bordered by 7-10 gular scales; gular region below lower jaws with enlarged tubercles; gular scales juxtaposed and granular, grading posteriorly into flat hexagonal pectoral scales and even larger ventral scales.

Goniurosaurus huuliensis
Neck narrower than body, covered with uniform granular scales interspersed with several sharply pointed conical tubercles on nape; tubercles on flanks conical, those of vertebral region somewhat flatter; dorsal body tubercles surrounded by 11-13 granular scales; dorsal tubercle rows at midbody 19-24; paravertebral tubercles between limb insertions 31-37, distinct vertebral row of tubercles absent; scales around midbody 118-130; ventral scales large; males with 25-30 precloacal pores in continuous series, females without precloacal pores (but pitted scales present); region posterior to vent covered by flat juxtaposed scales and greatly swollen, containing 1 or 2 enlarged tubercles on each side at level of vent; tail thick at base; light band on tail base in width of 9-12 granular scales and with 8-10 tubercles in transversal series.
Limbs relatively long and thin, covered dorsally with granular scales interspersed with several tubercles and ventrally with flat juxtaposed to subimbricate scales; dorsal granular scales grade into slightly flattened subimbricate scales on top of pes and manus; hind limbs larger than forelimbs; axillary pockets deep; subdigital lamellae wide, 10 or 11 under first finger, 18-21 under fourth finger, 11 or 12 under first toe, 21-25 under fourth toe; digits laterally compressed, increasing in length from first to fourth, fifth shorter than fourth (modified after Orlov et al. 2008;Nguyen 2011). (Fig. 3) Dorsal ground color of head, body and limbs signal brown (juveniles), grey-brown (young adults, SVL: 97.2-116 mm) and signal brown (adults), without small dark brown blotches (dark blotches present only on lower zone of flanks); dull white tubercles on dorsal body, limbs in juveniles, few orange tubercles on limbs in adults; iris orange or red brown; five bands on the dorsal ground, thin, immaculate without dark spots, slight bisque in juveniles, orange brown or yellow in adults, all edged anteriorly and posteriorly by thin dark brown bands, including one thin nuchal loop extending from posterior corners of eyes and posteriorly protracted (in V-shape), three body bands between limb insertions, and another one on tail base; ground color of tail dark brown, and signal brown at mottled tail base; 3-6 immaculate white caudal bands; ventral surfaces of head, body and limbs dull white and immaculate except for few dark brown spots on margin regions of gular and limbs. Grismer, Viets & Boyle, 1999 Fig. 4 Diagnosis Body splayed and gracile, SVL 86.5-126.5 mm; external nares bordered by 5-8 nasal scales; supraorbital region with a row of enlarged tubercles; outer surface of upper eyelid composed of granular scales, about one half the size of those on top of head and without enlarged tubercles; internasals 1 or 2; supralabials 8-12; infralabials 8-11; preorbital scales 13-16; eyelid fringe scales 46-61; postmentals 2-6; gular region below lower jaws with enlarged tubercles; paravertebral tubercles 29-38; scale rows around midbody 119-144, granular scales surrounding tubercles 9-14; axillary pockets deep; subdigital lamellae under fourth toe 20-26; precloacal pores in males 23-32; iris brown or bright orange; dorsal ground color of head, body and limbs pale brown to grey brown, mottled with small dark brown blotched; nuchal loop thin, posteriorly protracted (in V-shape); dorsal body bands between limb insertions 3, thin, immaculate yellow; gular region, belly, and ventral surface of limbs with brown spots Vu et al. 2006;Nguyen 2011 Table S2); head triangular, wider than neck, covered by uniform granular scales interspersed with tubercles in temporal and occipital regions; scales on rostrum slightly larger and flatter; enlarged supraorbital tubercles in a conspicuous row; middorsal portion of rostral partially sutured dorsomedially, bordered laterally by first supralabial on each side, dorsolaterally by prenasal on each side, and dorsally by 1 or 2 internasal and two supranasals; internasals 1 or 2; external nares bordered by 5-8 nasal scales (prenasal, anterior and posterior supranasals, 2 slightly enlarged postnasal, and 1-3 granular scales); preorbital scales 13-16; supralabials 8-12, grading into granular scales posteriorly; infralabials 8-11; eyes large, pupils vertical; eyelid fringe scales 46-56, those of upper eyelid slightly enlarged; outer surface of upper eyelid composed of granular scales, about one half the size of those on top of head and without enlarged tubercles; fold of skin originating in suborbital region extends posteroventrally across angle of jaw; external auditory meatus elliptical with long axis directed dorsoventrally; tympanum deeply recessed; mental triangular, bordered laterally by first infralabial on each side and posteriorly by 2-6 postmentals; postmentals bordered by 6-11 gular scales; gular region below lower jaws with enlarged tubercles; gular scales juxtaposed and granular, abruptly grading posteriorly into flat hexagonal pectoral scales and even larger ventral scales.

Goniurosaurus luii
Neck narrower than body, covered with uniform granular scales interspersed with several sharply pointed conical tubercles on nape; tubercles on flanks conical, those of vertebral region somewhat more flat; dorsal body tubercles surrounded by 11-13 granular scales; dorsal tubercle rows at midbody 20-24; paravertebral tubercles between limb insertions 29-38, distinct vertebral row of tubercles absent; scale rows around midbody 119-144; ventral scales large; males with 24-32 precloacal pores in a transverse continuous series, females without precloacal pores (but pitted scales present); region posterior to vent covered by flat juxtaposed scales and greatly swollen, containing 1-3 enlarged tubercles on each side at level of vent; tail thick at base, light band on tail base in width of 9-13 granular scales with 9-10 tubercles in transversal series; ventral caudals of tail base larger than dorsal caudals.
Limbs relatively long and thin, covered dorsally with granular scales interspersed with several tubercles and ventrally with flat juxtaposed to subimbricate scales; dorsal granular scales grade into slightly flattened subimbricate scales on top of pes and manus; hind limbs larger than forelimbs; axillary pockets deep; subdigital lamellae wide, 9-12 under first finger, 17-22 under fourth finger, 10-12 under first toe, 20-26 under fourth toe; digits laterally compressed, increasing in length from first to fourth, fifth shorter than fourth Vu et al. 2006;Nguyen 2011). (Fig. 4) Dorsal ground color of head, body and limbs pale brown to grey brown, mottled with small dark brown circle blotches, juveniles without small dark brown blotches (except juveniles SVL ≥ 72 mm initially present on head); iris red brown or bright orange; five bands on dorsal ground, thin, immaculate without dark spots, slight bisque in juveniles, orange brown or yellow in adults, all edged anteriorly and posteriorly by thin dark brown bands, including one thin nuchal loop extending from posterior corners of eyes and posteriorly protracted (in V-shape), three body bands between limb insertions, and another one on tail base; ground color of tail nearly solid black, and grey brown at mottled tail base; 3-6 immaculate white caudal bands; ventral surfaces of head, body and limbs dull white with a few dark brown spots in gular region, on belly and limbs.   Table S2); head triangular, wider than neck, covered by uniform granular scales interspersed with tubercles in temporal and occipital regions; scales on rostrum slightly larger and flatter; enlarged supraorbital tubercles in a conspicuous row; middorsal portion of rostral partially sutured dorsomedially, bordered laterally by first supralabial on each side, dorsolaterally by prenasal on each side, and dorsally by 1 or 2 internasals and two supranasals; internasals 1-5 (rarely 1:2; 2:2 or 2:3); external nares bordered by 5-10 nasals; preorbital scales 12-18; supralabials 7-10, grading into granular scales posteriorly; infralabials 6-9; eyes large, pupils vertical; eyelid fringe scales 47-58, those of upper eyelid slightly enlarged; outer surface of upper eyelid composed of granular scales of about one-half the size of those on top of head, without enlarged tubercles; fold of skin originating in the suborbital region extends posteroventrally across angle of jaw; external auditory meatus elliptical; tympanum deeply recessed; mental triangular, bordered laterally by first infralabial on each side and posteriorly by 2-5 postmentals; postmentals bordered by 7-10 gular scales; gular region below lower jaws without enlarged tubercles; gular scales juxtaposed and granular, abruptly grading posteriorly into flat hexagonal scales and even larger ventral scales. Neck narrower than body, covered with uniform granular scales interspersed with several sharply conical tubercles on nape; tubercles on flanks conical, those of vertebral region somewhat more flat; dorsal body tubercles surrounded by 10-13 granular scales; dorsal tubercle rows at midbody 15-22; paravertebral tubercles between limb insertions 22-33, distinct vertebral row of tubercles absent; scales around midbody 117-130; ventral scales large; males with 24-33 precloacal pores in a transverse continuous series, females without distinct precloacal pores (but 25-33 pitted scales present); region posterior to vent covered by flat juxtaposed scales and greatly swollen, containing 1 (rarely 2) enlarged tubercles on each side at the level of vent; tail thick at base; light band on tail base in width of 4-7 scales and with 6-10 tubercles in a transversal series.

Goniurosaurus lichtenfelderi
Limbs robust, covered dorsally with granular scales interspersed with several tubercles and ventrally with flat juxtaposed to subimbricate scales; dorsal granular scales grade into slightly flattened subimbricate scales on top of pes and manus; hind limbs larger than forelimbs; axillary pockets shallow; subdigital lamellae wide, 6-9 under first finger, 13-17 under fourth finger, 6-10 under first toe, 16-20 under fourth toe; digits laterally compressed, increasing in length from first to fourth, fifth shorter than fourth (modified after Grismer 2000;Grismer et al. 2002;Nguyen 2011).

Coloration in life (Fig. 5)
Dorsal ground color of head, body and limbs grey and grey brown in juveniles without blotches and chestnut brown in adults with dark brown blotches (in animals with SVL > 80 mm); iris orange or redish brown; four bands on the dorsal ground, thin, yellow in both juveniles and adults, immaculate without dark spots (few dark brown spots present in some animals), all edged anteriorly and posteriorly by thin dark brown bands, including one thin nuchal loop extending from posterior corners of eyes and posteriorly rounded (in U-shape), two body bands between limb insertions, and another one on tail base; ground color of tail dark brown, and grey brown at mottled tail base; 3-5 immaculate white caudal bands except first band slightly yellow, and some animals with regenerated tail present immaculate grey brown without white caudal bands; ventral surfaces of head, body and limbs dull white with a few dark dots on margin regions of belly, limbs, gular region, but immaculate dull white in juveniles (SVL ≤ 65 mm) without dark spots.

Morphological comparisons
Our morphological comparisons of the tiger gecko species (except for G. araneus) from Vietnam revealed an overall significant difference in the snout-vent length (SVL). As the result, G. huuliensis had the largest SVL, while G. lichtenfelderi had the shortest SVL in both adult males and females, compared to the other tiger geckos (P < 0.05). However, the SVL of G. catbaensis was not significantly different from that of G. luii (P > 0.05, Supp. file 2: Fig. S1).
The PCA of 18 selected morphometric characters estimated the first (PC1) and second principal (PC2) components to explain 82.5% (74.5% and 8.0%, respectively) of the variance among the four investigated species of Goniurosaurus in Vietnam (Supp. file 2: Fig. S2). The extracted PC1 and PC2 scores of the PCA differ significantly among the four species (Kruskal-Wallis's test, P < 0.05), of which the values and morphometric spaces of G. lichtenfelderi are not concordant with the three remainders (Fig. S2). Three morphological factors, namely head length (HL), head width (HW), and mouth width (MW) highly account for the overall difference in the PC1 score, while the PC2 score is mainly explained by the body height (BH) and body width (BW) (Supp. file 2: Fig. S2).

Taxonomic review
Our phylogenetic analyses strongly suggest that populations recorded from Vietnam indeed belong to four known species, namely G. catbaensis, G. huuliensis, G. lichtenfelderi and G. luii Vu et al. 2006;Orlov et al. 2008;Ziegler et al. 2008;Nguyen et al. 2009;Nguyen 2011). Despite extensive surveys over the last two decades, G. araneus has not been recorded in Vietnam, leaving its accurate type locality in Cao Bang Province, northern Vietnam ambiguous Ngo et al. 2016). It is possible that the species has been extirpated from Vietnam as a result of overexploitation to supply the international pet trade or that the species has never occurred at the documented type locality Ngo et al. 2016). Similarly, G. luii has not been recorded again at its type locality in China Stuart et al. 2006;Yang & Chan 2015).
Molecular results supported by our study confirm that G. murphyi, described by Orlov & Darevsky (1999), is a junior synonym of G. lichtenfelderi (Grismer 2000). All mainland populations of G. lichtenfelderi found in granitic forests are conspecific with those from the granitic offshore islands in Bai Tu Long National Park, Quang Ninh Province, northern Vietnam (type locality), with an intraspecific genetic distance of less than 0.7% (Supp. file 1: Table S1). However, a few issues need to be resolved in the future. In particular, G. hainanensis was recovered as polyphyletic by Liang et al. (2018) and our phylogenetic analyses. Moreover, Zhu et al. (2020a) suggested that G. luii and G. kadoorieorum are polyphyletic. We highly recommend that they may be synonymized. Samples assigned to G. kuroiwae were not recovered as monophyletic in our phylogenetic analysis. It is possible that more cryptic species from the Japanese group will be discovered. Further studies with additional samples, especially from type localities, should be undertaken to clarify these taxonomic problems.
Regarding morphological analyses, six morphometric characters, namely snout-vent length, head length, head width, mouth width, body height and body width strongly supported the overall difference among the four species in Vietnam and five characteristics, namely fourth toe, body and limb shapes, body bands, precloacal pores and granular scales surrounding dorsal tubercles, mainly accounted for the variation among the 17 tiger geckos from China and Vietnam. Six species of the G. kuroiwae group from Japan were not included in this study. However, the Japanese G. kuroiwae group can be distinguished from Chinese and Vietnamese species by the absence of precloacal pores and unsheathed claws Yang & Chan 2015;Honda & Ota 2017). We highly recommend that all these traits be considered as diagnostic characters for species of Goniurosaurus. Based on our identification key in accordance with the morphological and phylogenetic analyses, all tiger gecko species were grouped into four separate species groups, in agreement with the previous findings Qi et al. 2020aQi et al. , 2020bZhu et al. 2020aZhu et al. , 2020b. Liang et al. (2018), Qi et al. (2020aQi et al. ( , 2020b and our genetic analyses revealed that G. bawanglingensis and G. zhoui are embedded within the G. lichtenfelderi group, while they were previously placed as sister taxa in the G. luii group due to superficial morphological similarities as a result of adaptation to karst habitat Zhou et al. 2018). In accordance with these genetic findings, we found that the two karst-adapted species are morphologically most similar to three remaining granite-stream-adapted species of the G. lichtenfelderi group. Further comparative studies in southern China and Japan may help to resolve the phylogenetic relations of the genus Goniurosaurus.
Due to the high degrees of adaptation to specific microhabitats and local endemism, the genus Goniurosaurus could serve as a model system to study the evolution in lizards. In Vietnam, all species of Goniurosaurus have been found in isolated geographic ranges and none of them occurs in sympatry. For example, the distribution of G. luii stretches to the north of the Lang Son and Cao Bang provinces (Ngo et al. 2016), while G. huuliensis has been recorded in karst forests in the south of Lang Son Province. Goniurosaurus catbaensis is currently known only from offshore islands in Cat Ba National Park and the Ha Long Bay. Goniurosaurus lichtenfelderi can only be found in granitic habitats at several mainland localities and on some offshore islands in the Bai Tu Long Archipelago, which is contiguous with Ha Long Bay in the Gulf of Tonkin (Orlov et al. 2008;Gawor et al. 2016;Ngo et al. 2019a). Goniurosaurus araneus has not yet been re-discovered at its type locality in Cao Bang Province, and we only recorded G. luii inhabiting limestone karst forests of Cao Bang Province Vu et al. 2006;Ngo et al. 2016). The allopatric distribution has also been documented among species of Goniurosaurus from China. To date, neither the four species from the Hainan Archipelago, nor the five species of the G. luii group from Guangxi Province, China, nor the four known species of the G. yingdeensis group from Guangdong Province were recorded to occur in sympatry Zhou et al. 2018;Zhu et al. 2020aZhu et al. , 2020b. A large river (Zuojiang River) is regarded as the potential geographic barrier between the two species G. luii and G. araneus within Nonggang Nature Reserve, China (Chen et al. 2014). Likewise, geographic barriers (e.g., rivers and canyons) are considered to prevent a genetic exchange between members of the G. yingdeensis group (Qi et al. 2020a). The global rising of the sea level during the last melting period of glacial ice has shaped the myriad of archipelagoes (Clements et al. 2006;Ziegler et al. 2008;Liang et al. 2018;Ngo et al. 2019a). Thus, the oceanic barrier known as an important geographic feature has constrained current distributions and limits the genetic exchange of insular tiger gecko species, such as G. catbaensis, species from the Hainan Archipelago and G. kuroiwae group (Ziegler et al. 2008;Honda & Ota 2017;Liang et al. 2018;Zhu et al. 2020b). However, the last glacial maximum period occurring around 10 000-50 000 years ago was not sufficiently long for speciation events within Goniurosaurus (Clements et al. 2006;Sterling et al. 2006;Ziegler et al. 2008). Indeed, by employing a molecular dating method, Liang et al. (2018) estimated that the latest speciation events among tiger geckos took place approximately 2.8-2.9 Ma, in the late Pliocene, between G. yamashinae-G. orientalis, and G. huuliensis-G. luii. It is likely that the global warming during the Pliocene promoted the diversification in the genus, as also reported in the crocodile newts (genus Tylototriton Anderson, 1871) in Asia (Bernardes et al. 2020). Thus, the genus Goniurosaurus might offer a unique opportunity to investigate mechanisms of speciation and evolutionary adaptation, that are expected to highly correlate with a pre-dominantly allopatric mode of diversification.

Implications for conservation
The taxonomy of several existing species complexes requires further investigation and new species are still being discovered. Nevertheless, wild populations of the genus Goniurosaurus are under high risks of extinction, due to habitat destruction and exploitation for national and international trade. Of the 23 described tiger gecko species, 12 have been considered threatened (listed as VU, EN and CR) and one data deficient (DD) in the IUCN Red List (Ngo et al. 2019b). Assessments for the remaining species are still lacking. In order to better regulate the international trade in the species, all species of Goniurosaurus from China and Vietnam were recently listed in CITES Appendix II and the Vietnamese species were also protected under the Governmental Decree 06/2019/ND-CP (Group IIB). Trade data in the US showed that about half of the number of traded tiger geckos were only assigned to the genus level (Ngo et al. 2019b). This fact could be a consequence of the highly similar morphology among tiger gecko species and lacking guidance to distinguish the species (Ngo et al. 2019b). Besides, the conservation status and national protection status are not assessed equally among species. Therefore, the present identification guide will aid to prevent tiger geckos from being traded under a wrong name to circumvent legislations. The detailed descriptions of morphology provided in this study can assist local, national, and international authorities, such as local rangers, CITES authorities, custom officers, in enforcing international regulations. The identification key can also be useful for scientists and breeders to accurately identify tiger gecko species. However, morphological identification of captive lineages should be treated with reservation, as some breeders have produced hybrids (L. Grismer, pers. comm.), and thus molecular identification tools should be consulted as well in these cases.
After the CITES listing, four new species, namely G. gezhi, G. gollum, G. kwanghua and G. varius were recently described from China, which are automatically listed in Appendix II of the Convention (Qi et al. 2020a(Qi et al. , 2020bZhu et al. 2020aZhu et al. , 2020b. As the Japanese species appear to be similarly threatened and to prevent a shift towards international trade, Janssen & Shepherd (2019) and Ngo et al. (2019b) highly recommended that the Japanese species be also included in the CITES Appendices. During the time of writing, the Japanese Ministry for Environment announced to list all six endemic species of Goniurosaurus from Japan in CITES Appendix III to prevent over-harvesting of wild animals for trafficking activities. As a result, the listing of all Japanese species in Appendix III has come into force since 14 February 2021 (CITES Notification No. 2020/068).  Table S1. Pair-wise genetic divergence between species included in this study. Table S2. Morphological (minimum-maximum (mean ± standard deviation)) and meristic characters (minimum-maximum (number of specimens)) of four tiger geckos in Vietnam (except Goniurosaurus araneus). Length given in mm.