On the taxonomic validity of Boiga whitakeri Ganesh et al., 2021 with new insights on Boiga dightoni (Boulenger, 1894) (Reptilia: Squamata: Colubridae)

Colour polymorphism has been previously reported in several colubrid snakes including Boiga spp. In this paper, we report colour variations within the poorly known southern Indian Boiga dightoni , provide the first molecular data for this species, from two lo ­ calities (including the type locality) and compare them with data from other congeners. Additionally, we provide detailed dentition and hemipenis descriptions for B. dightoni . Molecular data for B. dightoni show very little difference (0.2–0.4% 16S; 0.9–1.2% cyt b ) to the recently described Boiga whitakeri , also from southern India. We have re­examined and present new information on the pholidosis of the type specimens of B. whitakeri and reconsider its taxonomic status. On the basis of molecular data and overlapping morphological characteristics, we argue that Boiga whitakeri and Boiga dightoni are conspecific, and place B. whitakeri under the subjective synonymy of the latter. Furthermore, we show that colour polymorphism in B. dightoni is a gender­independent character and that both colour morphs are found in high as well as low elevations and partly in sympatry. A revised key to the Boiga ceylonensis complex is provided.


Introduction
The colubrid snake genus Boiga Fitzinger, 1826 is rep resented by 37 currently recognised species distributed from the southern Palaearctic and the Oriental region to the northern and eastern coasts of Australasia (Uetz et al. 2022).Of these, eight species viz., Boiga beddomei (Wall, 1909), B. dightoni (Boulenger, 1894), B. thackerayi Giri et al., 2019, B. whitakeri Ganesh et al., 2021, B. forsteni (Duméril, Bibron & Duméril, 1854), B. flavivi ridis Vogel & Ganesh, 2013, B. nuchalis (Günther, 1875) and B. trigonata (Schneider, 1802) are found in the West ern Ghats of peninsular India.Boiga beddomei, B. dightoni, B. thackerayi and B. whitakeri are endemic to the Western Ghats (Ganesh et al. 2021).Among these, Boiga whitakeri is the most recently described species, based on two specimens from the southern Western Ghats (Ganesh et al. 2021).Prior to this, Ganesh et al. (2020) clarified the status of B. ceylonensis and B. beddomei based on morphological data and restricted these taxa to Sri Lanka and India, respectively.Ganesh et al. (2020) speculated that Indian records of B. ceylonensis might actually repre sent B. thackerayi, which was confirmed subsequently by analysis of molecular data from across the species' range (Ganesh et al. 2021).Ganesh et al. (2021) also provided molecular data for the holotype of Boiga whitakeri and all previously unsam pled species of the genus Boiga from across peninsular In dia, except B. dightoni.Boiga dightoni was originally de scribed based on a single female specimen collected from "Pirmed" (now Peermed, Kerala state, India) (Boulenger 1894) and appears to be a rarely encountered snake.Since its description, only a few studies reported the occurrence of this species from different parts of the southern West ern Ghats and none so far from the type locality (Inger et al. 1984;Murthy 1984;Kanagavel and Ganesh 2021).
During our recent fieldwork in the southern Western Ghats, we collected two individuals of Boiga sp., one from Peermed and the other from Arippa, Kerala.The speci men from Arippa superficially resembled the holotype of B. whitakeri, in colour and inconspicuous dorsal mark ings, and the specimen from Peermed resembled the para type of B. whitakeri in having prominent dorsal bands.We generated molecular and further morphological data for these two individuals and compared them with the types and with nontypes of other Boiga spp.from the Western Ghats.In this work, we reassess the taxonomic status of B. whitakeri in light of new data on scale variation, and we report colour polymorphism within Boiga dightoni.
We extracted genomic DNA from liver samples stored in absolute ethanol at -20°C, using the DNeasy (QiagenTM) blood and tissue kit following the manufac turer's protocol.We amplified partial sequences of two mitochondrial genes, 16S rRNA (16S) and cytochrome b (cyt b).Respective primers for these genes are as fol lows: 16SarL and 16SbrH (Palumbi et al. 1991) and CS1L and LTyph2R (Adalsteinsson et al. 2009).PCR conditions were as follows: Fragments of 16S were am plified using an initial denaturation at 95°C for 5 min, followed by 39 cycles of denaturation at 95°C for 45 sec, annealing at 50.4°C for 45 sec and extension at 72°C for 1 min 30 sec.Final extension was at 72°C for 10 min.Fragments of cyt b gene were amplified using an initial denaturation at 95°C for 5 min, followed by 35 cycles of denaturation at 95°C for 45 sec, annealing at 48°C for 45 sec and extension at 72°C for 55 sec.Final exten sion was at 72°C for 10 min.PCR reactions were carried out in 25 µl reactions containing 11µl of Takara emerald RR310B mastermix, 12 µl of PCR grade H 2 O, 0.5 µl of each forward and reverse primers and 1µl (60-80 ng) of template DNA.PCR amplifications were carried out in S1000TM Thermal Cycler (BioRad, USA).Ampli fied PCR products were run on a 2% agarose gel and viewed with an Essential V4 (UVITEC Cambridge, UK) gel documentation system to confirm the PCR amplifica tion.PCR products were purified and Sanger sequenced in both directions at Barcode Biosciences (Bangalore, India) using the same primers that were used for ampli fication.
Bidirectional sequences were checked manually using CHROMAS (http://technelysium.com.au/wp/chromas) and aligned using ClustalW with default prior settings implemented in MEGA 7 (Tamura et al. 2011;Kumar et al. 2016).We checked for unexpected stop codons in the proteincoding gene cyt b by translating nucleotide align ments to amino acids in MEGA7 (Kumar et al. 2016).The new sequences generated in this study were concat enated with data for twentythree other Boiga and three outgroups (Telescopus tripolitanus, T. variegates and Toxicodryas pulverulenta) (Appendix 1).
Maximum Likelihood (ML) analysis was performed using IQTREE (Nguyen et al. 2015), implemented in the web server version (http://iqtree.cibiv.univie.ac.at) (Trifinopoulos et al. 2016).The IQ-TREE server used Modelfinder (Kalyaanamoorthy et al. 2017) to find the best-fit evolutionary model for each of the four sug gested partitions (16S: TIM2+F+I+G4; Cytb position 1: TIM2+F+G4; Cytb position 2: TN+F+I+G4; cyt b posi tion3: TIM3+F+G4).Bayesian (BI) phylogenetic anal ysis was carried out with MrBayes 3.2 (Ronquist et al. 2012), with default prior settings and implementing the best-fit models and partitioning scheme as determined by Partition Finder V2. (Lanfear et al. 2017) with default settings.The best-fit scheme comprised three partitions, by gene and codon position (16S & cyt b position 1: GTR+I+G; cyt b position 2: TrN+I+G; cyt b position3: TVM+G).Four separate MCMC runs were initiated from random trees and allowed to run for ten million genera tions, sampling every 1000 generations.Analyses were terminated when the standard deviation of split frequen cies was less than 0.005, the first 25% of trees were dis carded as "burnin", and trees were constructed under the 50% majority consensus rule.Support for internal branches in ML and BI trees was quantified using Ul trafast Bootstrap (1000 pseudoreplicates) and posterior probability, respectively.
The numbers of dorsal scale rows are reported for one head length behind the head, at midbody (i.e., at the level of the ventral plate corresponding to half of the total ven tral number), and at one head length anterior to the vent respectively.Dorsal scale row reduction formulae were based on Dowling (1951a).Ventral scale counts and hemi penial descriptions follow Dowling (1951b) and Dowling and Savage (1960), respectively.The terminal scute is not included in the number of subcaudals.Values for symmet ric head characters are given in left/right order.
The following measurements were taken: snoutvent length (SVL); tail length (TL); head length (HL: distance between posterior edge of last supralabial and tip of the snout); head width (HW: at angle of jaws); head depth (HD: height at the occipital region); Frontal length (FL: at the longest point); frontal width (FW: at the widest point on the anterior region); eye diameter (ED: horizontal di ameter); eye to nostril distance (E-N: anterior corner of eye to posterior edge of nostril); eye to snout distance (E-S: anterior corner of eye to tip of snout); frontal to snout (FrSN: anterior end of frontal to tip of snout); inter-orbit al distance (IO: measured at the anterior edge of eyes); number of dorsal scale rows (DSR).All linear measure ments, except SVL and TL were taken using Mitutoyo dial vernier callipers (to 0.1 mm).SVL and TL were mea sured using a thread and metal scale (to 1 mm).
Among the specimens checked in this study, the holo type (BNHS 3597) of Boiga whitakeri is in a poor state of preservation and the paratype (BNHS 1863) of this spe cies is also damaged, especially its anterior ventral scales.Thus, the number of ventral scales provided here for the paratype of Boiga whitakeri (BNHS 1863) is not com plete Appendix 3C).For the ventrals that are damaged, we counted the adjacent dorsal scale rows assuming that one first-row dorsal corresponds to one ventral here.
Location records for both Boiga dightoni and B. nuchalis used for the map (Fig. 1) are based on the litera ture and specimens examined during the study (Appendix 4).Additionally, we downloaded researchgrade data for both of these species from the citizen science portals, iNaturalist (https://www.inaturalist.org)and India Biodi versity Portal (https://www.indiabiodiversity.org).All the records for both species were checked individually wher ever we could count the dorsal scales on one side, espe cially for the records from the southern Western Ghats.Doubtful records or records with poor photographs were not considered for plotting on the mapping and distribu tion (see also Discussion).
To obtain counts of teeth by a noninvasive procedure, the head of the holotype of Boiga dightoni was subject ed to microtomographic analysis at the Museum für Naturkunde Berlin, using a Phoenix nanotomXray|s tube.

Molecular phylogenetics
The inferred phylogenies are broadly congruent with those presented by Ganesh et al. (2021).Boiga ceylonensisis sister to B. dightoni + B. whitakeri with strong sup port (ML 98, BI 1.0) and this clade is sister to B. nuchalis.Boiga nuchalis from Yercaud (BNHS 3618) is nested with other B. nuchalis from the Western Ghats (Fig. 2).Boiga cf.ranawanei (sensu Ganesh et al. 2021) is sister to B. flaviviridis (a dry zone species found in thorn forests and scrub jungle).The holotype sequence of B. whitakeri is sister to the sample from Arippa (BNHS 3617) with strong and moderate support in BI and ML, respectively (BI 0.98, ML 83) and these two samples are together sis ter to the B. dightoni from the type locality with strong support (ML 95, BI 1.0) (Fig. 2).
The uncorrected pairwise genetic distance between the two samples of B. dightoni and the holotype of B. whitakeri, is 0.9-1.2% and 0.2-0.4% in cyt b and 16S, respectively (Table 1).These distances are almost all smaller than intraspecific distances for the other conge ners B. beddomei 2.5%, B. flaviviridis 2.2%, B. nuchalis 0.4-2.3% and B. thackerayi 0.6-3.6% in cyt b and B. beddomei 0.8%, B. nuchalis 0% and B. thackerayi 0.2-0.8% in 16S.2).Due to the poor state of preservation, we were unable to acquire the complete scale reduction formula for the paratype of B. whitakeri (BNHS 1863) but it has 21 MDSR until the level of the 145 th ventral and 19 DSR at the level of the 160 th ventral.This matches the dorsal scale reduction range of the sym patric Boiga nuchalis (see scale redcution formula).Fur thermore, except for the dorsal scale rows, there is a close similarity in the arrangements of head scalation, ventrals and subcaudals between specimens identified as belong ing to these "three" species (Table 2).
Based on the morphological data from the two speci mens collected during this study, including the specimen from the type locality (Peermed, Kerala) of Boiga dightoni, we confidently identify these two specimens as B. dightoni.Our morphological examination of the types and nontype materials of Boiga whitakeri, B. dightoni and B. nuchalis provide evidence that led us to conclude that the holotype of B. whitakeri is conspecific with B. dightoni.This is consistent with our molecular analyses, in which the holotype of B. whitakeri is nested within the samples (including the topotype) that we identify as B. dightoni.3C-D, 4, 6 A-C).This also applies to the holotype of B. whitakeri (Fig. 3B), which had markings on the body in life that disappeared in the preservative (Ganesh et al. 2021).A specimen from Aanapara, Kerala reported by Kanagavel and Ganesh (2021) also belongs to this morph, with very faint bands.
Morph 2 (n = 5).Olive greenish dorsum with black bands (76-80) on the body, with distinct marks on the head and a postocular stripe that ends shortly behind the fissure of the mouth, and irregular small dark blotches along the paraventral scales (Figs 3C-D, 5, 6D-F, J-L, P-R).The topotypic specimen (ZSICZRCV7541) of B. dightoni collected during this study is of this morph (Fig. 3C) and we observed several specimens from museum collections of this morph including a specimen (ZSI/SRS/S73) col lected from the Anamalais in Southern India.
Based on the specimens examined here, it is also clear that these two colour morphs are not explained by sex ual dichromatism because both male and female speci mens are known for both morphs.For example, the male specimens BNHS 3597 and ZSICZRCV7541 and the female specimens BMNH 1946.1.1.32 andBMNH 1940.10.13.19 belong to Morph 1 and 2, respectively.Both the morphs are found in sympatry in at least one location (Arippa, Kerala), so they additionally cannot be explained as purely geographic variation.Furthermore, these colour morphs cannot be currently explained as simple ontogenetic variation, because all the specimens examined here are adults.

Description of hemipenis of ZSI-CZRC-V-7541 (Fig. 7).
The right hemipenis is fully everted and removed in situ for further analysis.The hemipenis is subcylindrical and moderately elongate (length: 17.0 mm, maximum width: 5.7 mm), extending to the 7 th subcaudal.The sulcus is un divided, bounded by thick walls on both sides, and termi nates at the centre of the lobe.It can be differentiated into three zones; the proximal zone is covered with 4-6 rows of spines (~40% of the total length), the middle zone with 5 or 6 rows of spinulate flounces arranged transversely (~35% of the total length), and the distal calyculate area (~25%) with 4 or 5 rows of irregular calyces with papilate edges.The sulcus spermaticus is exposed before enter ing the calyculate area.There is not much variation in the arrangements of spines and body calyces on sulcate and asulcate sides.The overall structure of the hemipenis of ZSICZRCV7542 is similar to that described for ZSI CZRCV7541.
Mandibular bone with 20/20 posteriorly curved teeth, shorter than maxillary and palatine teeth, gradually de creasing in size posteriorly.Medial to each mandibular tooth is a single replacement tooth in different growth stages.Teeth 1, 3-7, 9, 11, 13, 15-17, and 19 missing,  tooth 2 loose on left side, teeth 1, 3, 5, 7, 9, 11, 13, 15, 17,  Distribution.Based on currently available data, Boiga dightoni is widely distributed in the southern West ern Ghats (south of the Palghat Gap), at elevations of 9-1258 m (Appendix 4).Murthy (1984) extended the northern range of this species to Topslip in Anamalais.Murthy (1984) reported 23 dorsal scale rows at midbody for the specimen he collected, which is known only for B. dightoni among Western Ghats' Boiga.The identity of this specimen (ZSI/SRS/S-73) is confirmed by photo graphs presented by Murthy (1984).With an additional specimen from the same locality (ZSICZRCV7541), we reconfirm the distribution of B. dightoni in Topslip in the Anamalai hills.The northernmost known distribu tion of B. dightoni is based on a specimen (BNHS 1842) from Palagapandy in the Nelliyampathy Hills, Kerala, a specimen that was previously (Ganesh et al. 2020) mis identified as B. nuchalis.The southernmost known oc currence of this species is Ponmudi in Kerala (Fig. 1).Thus, B. dightoni is found only south of the Palghat Gap in the Western Ghats.Boiga dightoni in parts of its range is probably sympatric with B. nuchalis and B. thackerayi  immediately south of the Palghat Gap, based on distribu tion data (Fig. 1) and the sequences reported by Ganesh et al. (2021).

Discussion
Intraspecific colour polymorphism has been reported in several colubrid genera (Pavón-Vázquez et al. 2011;van Rooijen et al. 2011;Cox and (Mohapatra et al. 2009;Tillack et al. 2021;Weinell et al. 2021).Our results demonstrate that B. dightoni is polymorphic in colouration with no marked sexual dimorphism or obvious ontogenetic or geographic component to this variation.
The holotype of Boiga dightoni (in preservative) is uniform in colour without dorsal markings and this prob ably led to several misidentifications in the past.Beyond the specimens examined here, it is probable that several individuals of B. dightoni are misidentified as B. nuchalis based on colour pattern.For example, at least two records (https://www.inaturalist.org/observations/37460080,86841689) identified as B. nuchalis may actually represent B. dightoni.As mentioned above, only a few records of B. dightoni are available in the literature and this might be mainly because of its overall similarity with B. nuchalis, a much more commonly encountered species that partly overlaps in geographic range with the former.It is likely that, at least in some places, these two species are sym patric.Hence, we hereby caution against identifying these species solely based on the colour pattern, especially from the southern Western Ghats where both B. dightoni and B. nuchalis are present.Our results highlight the importance of careful examination of type specimens when describ ing new, similar and closely related species, especially in the absence of molecular data.Wherever possible, it is also preferable to select wellpreserved and undamaged specimens when designating namebearing types.
During this study, we also examined the type series of Boiga thackerayi.In the original description (Giri et al. 2019), the midbody scales were reported as being disposed in 17 rows for the holotype (BNHS 3569) and paratype (BNHS 3571), and 19 for the other paratype (BNHS 3570).Based on this, Ganesh et al. ( 2021) used 17-19 midbody scale rows as a character for B. thackerayi in their key.However, our examination reveals that these two individuals (BNHS 3569 and BNHS 3571) also have 19 midbody scale rows (Appendix 5), the same as the other specimen (BNHS 2372).Here, we provide the correct scale reduction formulae for these two specimens (Appendix 5) and update an identification key to the B. ceylonensis complex.

Appendix 4
Gazetteer of confirmed locality records for Boiga dightoni and B. nuchalis in India.Localities where we verified only images for confirmation are marked with an asterisk.

Figure 1 .
Figure 1.Updated distribution of Boiga dightoni and B. nuchalis in the Western Ghats and B. nuchalis in peninsular India.

BMNH:
The Natural History Museum, London, UK; FMNH: Field Museum of Natural History, Chicago, USA; BNHS: Bombay Natural History Society; MCZ: Museum of Comparative Zoology, Cambridge, USA; Both specimens of Boiga sp.(ZSICZRCV7541 and BNHS 3617) collected during this study match well with the holotype of B. dightoni based on scalation data, main ly in having 23 MDSR (also see scale reduction formula for more characters).However, these two specimens dif fer significantly in scalation from B. whitakeri (based on data provided by Ganesh et al. 2021), despite the strong molecular similarity.MDSRs in B. dightoni are predom inantly 23 from the 10 th ventral throughout most of the midbody (up to ventrals 123-144) in all of the specimens examined during the study, but the point of reduction from 23 to 21 differs slightly among the specimens examined (see scale reduction formula below).Boiga whitakeri is diagnosed from congeners mainly based on the presence of 19 MDSR as provided in the original description (Ga nesh et al. 2021), but we counted 23 rows in the holotype (scale redcution formula & Table

Figure 2 .
Figure 2. ML phylogeny showing relationships of the newly sampled Boiga (in blue) and sequences of other available congeners.ML bootstrap support and BI posterior probability support = />75 or 0.75 is shown at each internal branch.Holotype of B. whitakeri (BNHS 3597) labelled in red.Inset image: head closeups of the respective samples in life or freshly roadkill specimen (BNHS 3617).Outgroups are pruned from this tree.
On the other hand, the type series and two other specimens of Boiga nuchalis examined here have 21 dorsal scale rows at midbody (Scale reduc tion formula; Appendix 2).This further confirms that the paratype (BNHS 1863; Appendix 3) of B. whitakeri is rather B. nuchalis.Because the holotype and paratype of B. whitakeri clearly represent two already described species, we relegate Boiga whitakeri Ganesh, Mallik, Achyuthan, Shanker and Vogel, 2021 to the junior subjec tive synonymy of Dipsas dightoni Boulenger, 1894.Morphology.A mediumsized Boiga (greatest TL 1000 mm (male), 935 mm (female)); 229-249 ventrals, 99-112 divided subcaudals; 13/14 teeth on maxilla and 7 on palatine; dorsal scales smooth, 23:23:19 in rows; dorsal scale reduction from 23 to 21 rows occurs be tween ventrals 123-144 and the reduction from 21 to 19 occurs between ventrals 148-155.Dorsum reddish dun to olive greenish with dorsal light brown to dark bands.Head with dark marking dorsally (rarely absent) and a dark laterocular stripe present.Colouration in life andpreservative.Based on the (live and museum) specimens examined and information available from the literature, we report two different co lour morphs in B. dightoni.Morph 1 (n = 5).Reddish duncoloured dorsum with faint reddish bands on the body (rarely absent) with or without distinct dark marking on the head, and ven tral scales uniformly creamish white (Figs 3A-B, 4, 6 A-C, G-I, M-O).Holotypes of both B. dightoni and B. whitakeri are of this colour morph with no mark ings on the body in preservation.However, it might be noted that the recently collected specimen from Arippa (BNHS 3617) had faint markings on the body at the time of collection (3 rd February 2022) that disappeared in the preservative (Figs 2,
Dorsal scale row reduction formulae for some of the Boiga specimens examined in this study presented be low.*between ventral 9 and 14 counts are not possible because of the damaged vertebral region.Additionally, there are several reductions and additions of the paraver tebral scale row between the corresponding ventrals of 194 to 209.*** dorsal scales damaged and is not possible to find the area of scale reduction:

Table 1 .
Pairwise genetic distances (%) between the Boiga spp.from the Western Ghats, Eastern Ghats and Sri Lanka for both mitochondrial 16S and cyt b genes.

Table 2 .
Meristic and morphometric data (in mm)for Boiga dightoni and Boiga nuchalis examined in this study.