We sequenced fragments of two mitochondrial genes (12S and 16S ribosomal RNA) and one nuclear gene (c-mos). These genes were primarily selected because they are the most widely used in phylogenetic studies of Dipsadinae yielding successful results (e.g., Zaher et al. 2009). Consequently, they are the genes with the largest number of available sequences in GenBank for Philodryas and related genera. We generated sequences of these genes from three species of Philodryas: two individuals of P. aestiva, nine of P. trilineata, and 37 of P. patagoniensis. To enrich our dataset, we included sequences of eight additional Philodryas species available on GenBank: P. aestiva (n = 1), P. agassizii (n = 1), P. baroni (n = 1), P. mattogrossensis (n = 1), P. nattereri (n = 1), P. olfersii (n = 1), P. patagoniensis from Brazil (n = 1), and P. psammophidea (n = 1). We also included the same gene fragments of Chlorosoma viridissima and Xenoxybelis argenteus as outgroups (see Table SS1 for voucher information, GenBank accession numbers, and collection localities). Unfortunately, we were unable to include new sequences of species currently unavailable in GenBank (P. arnaldoi, P. erlandi, P. livida, and P. varia), despite their inclusion in two recent phylogenies (Zaher et al. 2019; Arredondo et al. 2020), including new sequences of P. agassizii.

Total genomic DNA was extracted from tissue samples using a saline DNA extraction method based on lithium chloride, following the protocol of Gemmell and Akiyama (1996). Fragments of the 12S rRNA, 16S rRNA, and c-mos genes were amplified independently via polymerase chain reactions (PCR) using the following primers (see Zaher et al. 2009 for details): (1) for 12S rRNA, L1091mod (5’ CAA ACT AGG ATT AGATAC CCT ACT AT 3’) and H1557 (5’ GTA CRC TTA CCWTGT TAC GAC TT 3’); (2) for 16S rRNA, L2510 (5’ CCG ACT GTT TAM CAAAAA CA 3’) and H3056 (5’ CTC CGG TCT GAA CTC AGA TCA CGTRGG 3’); and (3) for c-mos, S77 (5’ CAT GGA CTG GGA TCAGTT ATG 3’) and S78 (5’ CCT TGG GTG TGATTT TCT CAC CT 3’).

The PCR was performed using a Bio-Rad T-1000 thermal cycler. The annealing temperature was 54 °C for the 12S and 16S gene fragments and 56 °C for the c-mos fragment. Amplified products were sequenced in Macrogen Inc. (Seoul, South Korea).

Most specimens used in the morphological analysis are currently housed in the herpetological collections of the Museo Argentino de Ciencias Naturales “Bernardino Rivadavia” (MACN, Buenos Aires, Argentina) and Museo de La Plata (MLP.JW and MLP.R, Buenos Aires, Argentina). In addition, we examined photographs of key specimens from Natural History Museum (BMNH, London, United Kingdom), Museu de Ciências e Tecnologia da Pontifícia Universidade Católica do Rio Grande do Sul (MCP, Rio Grande do Sul, Brazil) and United States National Museum, Smithsonian Institution (USNM, Washington, United States: lectotype of P. patagoniensis) (File S1). The sampling used for study of morphological characters (such as scale counts, body measurements, and colouration) was substantially larger than that used for other analyses (genetic, hemipenial, cranial osteology, and other morphology traits), resulting in various degrees of freedom among character types (File S1, Table SS2). Morphological comparisons and analyses within Philodryas patagoniensis were based on morphotypes that were highly consistent with different genetic types (haplogroups). Juvenile specimens were identified following Fowler and Salomão (1995), who established the critical snout-vent length thresholds marking the transition to adulthood in males and females of six Philodryas species; in the case of P. patagoniensis, these thresholds are 420 mm in males and 470 mm in females. Sex was determined by making a small incision to check for the presence of hemipenes. Juveniles were sexed when possible, included in scale count and body colouration analyses, and excluded from body measurements. External morphology was assessed through the following body measurement characters: head length (HL) measured to the nearest 0.1 mm, using a digital calliper; snout-vent length (SVL), tail length (TL), and total length (TTL), all measured to the nearest 1 mm by carefully extending the specimens along a fixed ruler. Cephalic and body scale counts followed Dowling (1951) and Giraudo (2001), including formulae for dorsal, infralabial, supralabial, and temporal scales, as well as the number of postocular, preocular, subcaudal, and ventral scales. The raw data collected from scale counts, scale formulae, and external measurements are presented in Table SS2. The colour pattern was described using the 39-colour chart proposed by Yu et al. (2018). The statistical comparison of body measurements was conducted using proportions of each measurement relative to total length, rather than comparing the raw data alone, in order to account for effect of size caused by age of specimens that may exist among individuals in our sample (for details on these widely used morphometric procedures: Baur and Leuenberger 2011). Thus, this procedure was applied to the proportional variables HL/TTL, SVL/TTL, and TL/TTL. Ventral and subcaudal scale numbers, unaffected by total length, were statistically compared as raw data. All comparisons were performed for each combination of morphotype and sex using the Mann-Whitney U-test, implemented in the software PAST v3.04 (Hammer et al. 2001).

Cranial osteology was analysed based on the limited descriptions available for skulls of the genus (Bonino et al. 1987; Lobo and Scrocchi 1994; Di Pietro et al. 2014). Skulls of five specimens of P. patagoniensis and two of P. agassizii used for comparison were prepared following the protocol of Taylor and Van Dyke (1985) (File S1). Bone terminology follows Cundall and Irish (2008). Hemipenes of 10 specimens were prepared using the technique described by Zaher and Prudente (2003), and descriptions follow the terminology proposed by Zaher (1999) (File S1). For genera, species, and higher clade taxonomy, we follow Uetz et al. (2025).

The geographic distribution map of the morphotypes of P. patagoniensis was constructed using QGIS v3.28 software (QGIS Development Team 2024), incorporating the aforementioned museum specimens and supplemented with locality data from citizen science platforms such as iNaturalist (https://www.inaturalist.org; last accessed 26 November 2024) and Ecoregistros (https://www.ecoregistros.org; last accessed 26 November 2024). In the case of citizen science platforms, we only considered records with photographs that allowed for unequivocal classification of the species and morphotype. File S1 provided the complete list of specimens employed for the distribution analysis.

Sequences were aligned using the MUSCLE algorithm in MEGA7 v7.0 (Kumar et al. 2016) and concatenated with SequenceMatrix v1.7.8 (Vaidya et al. 2011). The final matrix comprised 58 terminals and 1161 base pairs (12S with 289 bp, 16S with 379 bp, and c-mos with 493 bp). A haplotype network based on parsimony analysis was constructed using the TCS algorithm (Clement et al. 2000) implemented in PopArt, with each gene alignment of Philodryas patagoniensis analysed separately. The network illustrated geographic relationships among haplotypes, considering their frequencies and geographic origins. Ambiguous connections (loops) in the network were resolved following the method of Crandall and Templeton (1993). Uncorrected genetic p distances between species/linages were calculated using MEGA7 v7.0 (Kumar et al. 2016), excluding positions with gaps and missing data.

We conducted a phylogenetic analysis of Philodryas using a concatenated matrix of the three genes, comprising our sequences and those retrieved from GenBank (1161 bp, 58 terminals). We coded seven external morphological characters related to body measurements and scutellation (four binary and three multistate), five characters concerning the shape and pattern of cephalic scales (three binary and two multistate), and four colour pattern characters (one binary and three multistate), totalling 16 external morphological characters. The states of continuous scutellation characters (i.e., ventrals, subcaudals) were delimited and coded on the basis of the observed discontinuities within the overall variation of a given character. Following this procedure, we selected midpoints within these gaps and defined them as boundaries separating each character state. These were coded following Thomas (1976), Di Pietro et al. (2013), Cacciali et al. (2016a), Rivas et al. (2024), and our observations. Additionally, we included 10 cranial osteological characters involving the nasal, palatine, parabasisphenoid, parietal, prefrontal, premaxilla, and quadrate bones (one multistate and nine binary), coded according to Bonino (1987), Lobo and Scrocchi (1994), and our observations. One binary character concerning hemipenial size was coded following Thomas (1976), Zaher (1999), Cacciali et al. (2016a), and our observations. Furthermore, one biochemical character (toxin type in the venom composition) was included as a multistate character, entirely coded from Tioyama et al. (2023). Detailed descriptions of the morphological characters are provided in File S2, while File S1 contains a list of specimens coded based on personal examination. The 28 morphological characters were combined with the molecular data, and this matrix was used in a total evidence phylogenetic analysis based on the principle of maximum parsimony, as implemented in the software TNT v1.6 (Goloboff et al. 2008; Goloboff and Morales 2023).

The total evidence analysis (1161 base pairs plus 28 morphological characters, and 58 terminals; see matrix in File S3) was performed using the traditional search option, beginning with 100 rounds of Wagner trees. Tree Bisection and Reconnection (TBR) was used as the swapping algorithm, with 10 trees saved per round and trees replaced. Chlorosoma viridissima was designated as outgroup. A strict consensus tree, the absolute Bootstrap support (100 replicates) and Jackknife support (removal probability of 40%, 100 replicates), and the list of synapomorphies for each clade, were obtained using the specific tools of the TNT software. The resulting consensus tree was exported as a tree in which the number of synapomorphies on each branch is represented by varying branch lengths, similar to, but distinct from a phylogram. For this purpose, we used the following script: “ttag = ; blength* n; export > filename;” (where n refers to the number of the tree to be exported, using a filename with the .nex extension). For further details on this and other TNT commands, see Goloboff et al. (2008) and the general software documentation referenced therein.

The parsimony analysis implemented in TCS for each gene alignment separately identified two haplogroups within Philodryas patagoniensis, referred to here as A- and B-haplogroups, each comprising several haplotypes (Fig. 1). Thus, the haplogroups stand out in the networks due to the genetic distances involved (i.e., haplotypes within each group differ by only one or two variable sites, while the haplogroups themselves are separated by more variable sites). Interestingly, the c-mos gene also reveals two haplogroups, although the differences are much less marked than those observed in the mitochondrial genes. Each haplogroup corresponded to a distinctive morphological pattern; thus, we refer to them as morphotypes A and B, respectively. The following sections enumerate and describe the distinguishing characteristics between the two morphotypes to avoid redundancy. Specimens assigned to A-haplogroup (n = 16) are from northern samples (Córdoba, Corrientes, Entre Ríos, and Santa Fe provinces, Argentina; and Rio Grande do Sul, Brazil: northern populations), whereas those in B-haplogroup (n = 22) represent the southernmost samples (Chubut, Buenos Aires, and Rio Negro provinces, Argentina: southern populations). As shown in Figure 1, mitochondrial gene sequences revealed substantial divergence between haplogroups (10 mutational steps in both 12S and 16S), while the nuclear c-mos gene exhibited lower divergence (two mutational steps). The number of haplotypes within A-haplogroup ranged from three (mitochondrial genes) to six (c-mos), while in B-haplogroup it varied from four (12S) to six (16S and c-mos). The uncorrected p distance between both haplogroups was 4% ± 1% for 12S, 3% ± 1% for 16S, and 1% ± 0% for c-mos. Although these values were lower than the mean divergence observed among all species pairs used for comparison (12S: 7.7% ± 1.9%; 16S: 5.2% ± 1.3%; c-mos: 1.1% ± 0.4%), they were equivalent to or exceeded those observed in some species pairs (Table 1).

We analysed the morphological characters based on the assignment of specimens to the two haplogroups/morphotypes. For external morphology, we measured 142 specimens representing both morphotypes (A-morphotype = 42; B-morphotype = 100; see Table 2 for variation in body measurements and scale counts). We found significant differences between morphotypes and between sexes within each morphotype for all body proportions and the number of ventral and subcaudal scales, except HL/TTL, which did not differ between males and females of the A-morphotype (Fig. 2). These results indicate that A-morphotype specimens possess proportionally longer heads, bodies, and tails, as well as higher counts of ventral and subcaudal scales compared to B-morphotype specimens. Statistical significance for these comparisons is detailed in Figure 2.

Although cephalic scales and dorsal scale counts showed considerable variation, we identified a dominant pattern for each scale type within each morphotype (e.g., dorsal scale formula 19–19–15 vs. other less represented combinations, see Table 3), with no differences between sexes. Thus, the dominant patterns, when potentially informative, were used in the phylogenetic analysis. The colouration pattern indicates a general tendency towards increased dark pigmentation in specimens of B-morphotype (e.g., maculated pattern on dorsal scales, presence of black dots on the lateral margins of each ventral scale, and irregular dark blotches on the dorsal head scales; see Figs 3, 4, 5, 6). The configuration and shape of certain cephalic scales also differed between morphotypes, with a general tendency for specimens of A-morphotype to exhibit more enlarged scales (e.g., rectangular loreal scale, larger first temporal scale; Fig. 3). In summary, we found nine external morphology characters that varied between morphotypes (characters 1–3 and 10–15 in File S2). This number increased to 16 external characters by incorporating additional external traits potentially informative for Philodryas systematics (see character descriptions in File S2), which were subsequently used for phylogenetic analysis (see below).

The hemipenes of both morphotypes were found to be extremely similar (Fig. 7). We were only able to detect one relevant character for use in the cladistic analysis, primarily related to the variation observed in the outgroup (see character 27 in File S2: relative size of the hemipenes). Additional hemipenial features will be described in the context of the forthcoming taxonomic revision of P. patagoniensis (see below).

The cranial osteological analysis revealed variations in four bones between the two morphotypes: shape of the vomerine process of the premaxilla (rectangular in A-morphotype, square in B-morphotype, character 17), orientation of the vomerine processes of the premaxilla (divergent in A-morphotype, sub-parallel in B-morphotype, character 18), general appearance of the horizontal dorsal nasal lamina (slender in A-morphotype, robust in B-morphotype, character 20), posterior edge of the horizontal dorsal nasal lamina (notched in A-morphoptype, smooth in B-morphotype, character 21), configuration of the parabasisphenoid rostrum (bidentate in A-morphotype, tridentate in B-morphotype, character 22), and shape of the maxillary process of the palatine (pointed in A-morphotype, rounded in B-morphotype, character 24); see Figure 8 and File S2. These traits were included in the phylogenetic analysis alongside eight of the ten cranial characters previously used by Lobo and Scrocchi (1994); see File S2 for character description.

We obtained 70 equally parsimonious trees (374 steps), in which both clades of Philodryas patagoniensis (A- and B-morphotypes) were consistently recovered with high support (each with Bootstrap 100 and Jackknife 100). As consequence, the consensus tree retained the clades described in the next paragraphs (Fig. 9). The species P. agassizii invariably formed a sister group to the clade comprising all terminals of the B-morphotype (Bootstrap 83; Jackknife 80). In contrast, the clade containing all terminals of the A-morphotype was recovered as sister to the P. agassizii + B-morphotype clade (Bootstrap 85; Jackknife 86). Similarly, the consensus recovered a clade comprising P. baroni and P. trilineata (Bootstrap 100; Jackknife 100), with P. mattogrossensis as their sister taxon, although with low support (Bootstrap and Jackknife < 60). Another clade was identified in which P. psammophidea was nested within P. aestiva (Bootstrap 95; Jackknife 97). These clades exhibited a comparable number of synapomorphies. The nine terminals of P. trilineata formed a clade supported by five morphological (23 anterior dorsal scales, dorsal head scales with black spots, ventral scales with dark spots in a random pattern, marbled dorsal and lateral colouration at midbody, and pointed maxillary process of the palatine) and eight molecular synapomorphies. This clade and a single terminal of P. baroni were grouped based on three morphological (23 dorsal scales at midbody, 17 dorsal scales at the posterior body, and simple, undivided cloacal scale) and 11 molecular synapomorphies. Meanwhile, the three terminals of P. aestiva and the single terminal of P. psammophidea were grouped together based on one morphological (immaculate dorsal pattern at midbody, reversed to stripped pattern in P. psammophidea) and nine molecular synapomorphies. The clade comprising P. aestiva + P. psammophidea was further recovered as sister to the clade here referred to as the patagoniensis group (including both morphotypes of P. patagoniensis and P. agassizii), supported based on one morphological (cloacal scale divided) and four molecular synapomorphies.

The patagoniensis group was defined by three morphological synapomorphies (supralabial formula 7(3,4)*; infralabial formula 9(5); anterior border of the optic foramen positioned anterior to the midline that bisects the orbit*) and six molecular synapomorphies (asterisks indicate unambiguous synapomorphies). Within this group, we found two distinct clades: one corresponding to the A-morphotype and the other to the B-morphotype + P. agassizii. The A-morphotype clade was strongly supported by five morphological (head proportionally long with HL/TTL<3%; lateral edge of the supraocular scale straight in a dorsal view; long first temporal scale with height <45% of the length; ventral scales at midbody with consistently transverse lines and lacking lateral spots*; and dorsal scales at midbody in adults exhibiting a dotted pattern) and 11 molecular synapomorphies. The clade comprising B-morphotype + P. agassizii is supported by six morphological synapomorphies (mean number of ventral scales<178*; mean number of subcaudal scales<87*; dorsal head scales bearing noticeable black spots centrally and occasionally along the margins; sub-parallel vomerine processes of the premaxilla; a robust horizontal dorsal nasal lamina; and a rounded maxillary process of the palatine), but we did not recover molecular synapomorphies. The species P. agassizii exhibited four morphological autapomorphies (13 dorsal scale rows at the anterior body*; 13 dorsal scale rows at midbody*; 13 dorsal scale rows at the posterior body; and an infralabial formula 8(5)*), and 12 molecular synapomorphies. All terminals of the B-morphotype formed a clade supported by two morphological synapomorphies (midbody ventral scales in adults exhibiting a transverse line that may be complete, incomplete, or absent with a lateral black spot always present; and a maculated dorsal and lateral colour pattern at midbody in adults*) in addition to 11 molecular synapomorphies. See File S2 for the details of each character state.

The recovered topology and levels of support enabled the recognition of two distinct taxa within the nominal P. patagoniensis. Notably, one of these was grouped in a clade that included all specimens of P. patagoniensis (Girard, 1858) sensu stricto. In contrast, the other clade comprises all A-haplogroup specimens (A-morphotype), representing a new species that has remained cryptic due to the extensive morphological variability within P. patagoniensis sensu lato. The only specimen of P. agassizii genetically analysed in this study corresponds to the sole available sequence in the GenBank, and (except Zaher et al. 2019 and Arredondo et al. 2020) used in all phylogenetic treatments of this taxon since Zaher et al. (2009). As P. agassizii is represented by a single terminal, it must necessarily group with one clade or another; in our study, it was recovered as sister to the clade formed by all terminals of the B-morphotype (see Discussion). In any case, our phylogenetic analysis further reveals that P. agassizii is more closely related to the terminal taxa of the B-morphotype than to the other Philodryas species analysed, supporting the conclusion that P. agassizii, together with the A- and B-morphotype clades, constitutes one of the three primary forms within the patagoniensis species group. Additionally, accumulating autapomorphies within P. agassizii reinforces its classification as a distinct and well-defined species.

To formally re-describe P. patagoniensis (Girard, 1858) and encompass the full spectrum of morphological variation, we will focus on key aspects and distinctive features that differentiate it (B-haplogroup, B-morphotype, southern populations: see above for the list of characters) from the A-morphotype (A-haplogroup, northern populations), which will be designated as a new species.

In this study we identified two distinct taxa, P. patagoniensis sensu stricto (hereafter referred to as such unless otherwise specified) and P. pseudomambasp. nov., based on variation in two mitochondrial genes (12S and 16S ribosomal RNA) and one nuclear gene (c-mos). The p distance values observed between P. patagoniensis and P. pseudomambasp. nov. exceeded those found between other well-differentiated species pairs within the genus (e.g., P. aestiva–P. psammophidea for 12S, P. aestiva–P. agassizii for c-mos, and P. trilineata–P. baroni for all three genes; see Table 1). Moreover, the genetic distances between P. patagoniensis and P. pseudomambasp. nov. were greater than those reported between closely related species within Dipsadinae. For instance, Lehr et al. (2023) reported a 2.1% genetic divergence in the 12S between Tachymenoides affinis and T. harrisonfordi, while Carvalho et al. (2020) found a 0.43% divergence in the 16S gene between Hydrodynastes gigas and H. bicinctus. More strikingly, several closely related species of Atractus, such as A. iridescens–A. dunni (0%), A. typhon–A. gigas (1%), and A. resplendens–A. duboisi (2%), exhibit very low genetic distances for the 16S gene (Araújo De Oliveira and Hernández Ruz 2016). A similar pattern is observed in Apostolepis albicollaris and A. dimidiata, which show only 1% divergence for the 12S gene and 0% for c-mos (Entiauspe-Neto et al. 2022). Concerning the c-mos sequences, Entiauspe-Neto et al. (2022) have argued that this gene is not particularly informative for phylogenetic reconstruction due to its low number of variable sites at the species level, a limitation also noted in squamate phylogenies (Saint et al. 1998). However, despite its limited variability, the c-mos gene remains valuable for delimitation and corroborating candidate species initially identified through mitochondrial markers (Wüster et al. 2024; present study).

Our objective was to establish a phylogenetic framework to assess the position of the new species relative to P. patagoniensis and other species of Philodryas, rather than to perform a comprehensive phylogenetic analysis of the genus’ internal and external relationships. Accordingly, we do not provide a detailed discussion of the phylogenetic relationships recovered. Nevertheless, we note that, in broad terms, our results are consistent with clades and relationships previously reported in the literature. For instance, our results support the clade formed by P. trilineata + P. baroni (Lobo and Scrocchi 1994; Arredondo et al. 2020; Melo-Sampaio et al. 2021) and the close relationship between P. aestiva and P. psammophidea (Grazziotin et al. 2012; Melo-Sampaio et al. 2021), with the latter consistently exhibiting some degree of association with the patagoniensis clade. The sequences we used for P. psammophidea correspond to two mitochondrial fragments (12S and 16S) originally uploaded to GenBank by Vidal et al. (2010). In this study, P. psammophidea forms a clade with P. aestiva, a pattern also observed in other studies that employed the same sequences (Grazziotin et al. 2012; Figueroa et al. 2016; Melo-Sampaio et al. 2021). This finding did not attract attention, as those studies used a single terminal for each species. Zaher et al. (2019) and Arredondo et al. (2020) used own P. psammophidea sequences (never uploaded to GenBank) and excluded, without explanation, the sequence from Vidal et al. (2010). Consequently, in their analyses, P. psammophidea was placed within the patagoniensis group, whereas P. aestiva appeared as sister to that group, showing no direct relationship with P. psammophidea. These findings are consistent with the clear morphological distinctiveness of both species and their largely disjunct geographic distributions (Giraudo 2001; Nogueira et al. 2019). We therefore believe that the sequence attributed to P. psammophidea by Vidal et al. (2010) actually corresponds to P. aestiva, as supported by our phylogenetic results, in which this sequence nests within the P. aestiva terminals.

Except for Zaher et al. (2019) and Arredondo et al. (2020), who used different terminal taxa, other molecular phylogenetic analyses have placed P. agassizii within a clade with P. patagoniensis sensu lato (Zaher et al. 2009; Grazziotin et al. 2012; Melo-Sampaio et al. 2021). However, these analyses included specimens of P. patagoniensis sensu lato that correspond to P. pseudomambasp. nov., specifically MCP 5753 (which, based on requested photographs, undoubtedly exhibits the A-morphotype; see File S1), and a second voucher specimen from Santa Fe, Argentina, represented only by a tail fragment (MACN 47276), for which sequence data are no longer available in GenBank (the latter generated by Arredondo et al. 2020). Philodryas agassizii, P. patagoniensis, and P. pseudomambasp. nov. formed a clade with moderate support (the patagoniensis group), with P. pseudomambasp. nov. recovered as sister to P. patagoniensis + P. agassizii, and strongly supported by morphological and molecular synapomorphies (see Results section). Zaher et al. (2019) and Arredondo et al. (2020) found subtle differences in the positioning of certain species within the patagoniensis group. In the study by Arredondo et al. (2020), P. agassizii was recovered as sister to P. livida, while the individuals they referred to as P. patagoniensis (now identified as P. pseudomambasp. nov.) form a clade with P. varia. Zaher et al. (2019) found P. agassizii to be sister to the patagoniensis group clade, whereas P. patagoniensis (now identified as P. pseudomambasp. nov.) forms a clade with P. varia. This discrepancy between their results and ours may be attributed to two factors, or even an interaction between both. One possible factor is that Zaher et al. (2019) and Arredondo et al. (2020) used two new sequences of P. agassizii that differ from the one we employed in our study, which has also been used in all phylogenetic analyses involving this species (from Zaher et al. 2009 to Melo-Sampaio et al. 2021). In this regard, we assessed the sequence of P. agassizii used in our study across the three genes (c-mos, 12S, and 16S) by running it through the BLAST+ suite (Camacho et al. 2009), and found no evidence of anomalies. Furthermore, this sequence is the only one available for the species (see Materials and Methods) since the P. agassizii sequences used by Zaher et al. (2019) and Arredondo et al. (2020) were never uploaded to GenBank. Alternatively, the discordance between these studies and ours likely arises from their use of a different (and uncheckable) taxonomic sampling. None of the relevant sequences they used were uploaded to GenBank (their sampling included P. arnaldoi, two new sequences of P. agassizii, P. erlandi, P. livida, and P. varia). For the reasons already explained, none of these species were included in our analysis. Additionally, we used sequences from P. patagoniensis sensu stricto, a species for which no sequences were available to Zaher et al. (2019) and Arredondo et al. (2020). In fact, all the sequences of the species employed in various phylogenies actually correspond to P. pseudomambasp. nov. (e.g., Zaher et al. 2009; Grazziotin et al. 2012). In summary, it is important to highlight that Zaher et al. (2019) and Arredondo et al. (2020) did not use the P. agassizii GenBank sequence we used (the only available), nor did they address its exclusion. The supplementary material table S6 in Zaher et al. (2019) explained the exclusion of several GenBank sequences from their analysis, listing the species whose sequences were chimeras or contained various types of errors, with no mention or justification for the exclusion of the P. agassizii we used in our work. We believe that the most appropriate way to evaluate the status of this P. agassizii sequence would be to include it alongside the new sequences analysed by Zaher et al. (2019) and Arredondo et al. (2020) in a joint phylogenetic assessment to verify its true position within the genus.

Many of the synapomorphies we identified were first discussed by earlier authors, who highlighted several of them in their studies on the variation of external morphology within P. patagoniensis sensu lato. For example, Thomas (1976) noted geographical variation in the number of ventral and subcaudal scales, with counts decreasing from north to south, a pattern that strongly supports the recognition of a northern species (P. pseudomambasp. nov.) and a more southern species (P. patagoniensis) (characters 2 and 3 in the present study, see File S2). Similarly, the same author noted that P. patagoniensis sensu lato from Brazil and adjacent areas of Paraguay exhibit dark pigmentation on the posterior part of each ventral and subcaudal scale, a feature absent in specimens from Argentina and Uruguay (character 14 of our work, File S2). Additional synapomorphies supporting clades in our analysis, including within the internal group, are osteological. Although cranial osteology in snakes is seldom used for species diagnoses due to presumed low variability between closely related taxa (Di Pietro et al. 2014), we employed several osteological characters from Lobo and Scrocchi (1994), along with other characters verified explicitly for this study (characters 17–26 in File S2). The three most parsimonious cladograms recovered by Lobo and Scrocchi’s (1994) displayed numerous polytomies and exhibited limited resolution of clades. In contrast, our total evidence analysis revealed that many osteological characters functioned as synapomorphies for clades or autapomorphies for individual species, underscoring their value when evaluated with molecular data, enhancing the overall resolution of phylogenetic trees. Our analysis recovered that the patagoniensis species group clade was supported by one osteological character (anterior border of the optic foramen positioned anterior to the vertical midline that bisects the orbit), while the clade P. agassizii + P. patagoniensis is supported by three (a sub-parallel vomerine processes of the premaxilla; a robust horizontal dorsal nasal lamina; and a rounded maxillary process of the palatine). Although we counted the teeth, we chose not to use this character source because our counts for both species showed high variability and completely overlapped with the ranges reported in the literature for P. patagoniensis sensu lato (Thomas 1976) and P. pseudomambasp. nov. (Lobo and Scrocchi 1994). Therefore, these characters, and other minor hemipenial traits, did not prove helpful in distinguishing between the two species.

The geographical distribution of P. pseudomambasp. nov. and P. patagoniensis reveals a clear correlation with forested and open habitats, respectively (FAO 2020), with an area of overlap in central Argentina and central Uruguay (Fig. 1).

Philodryas pseudomambasp. nov. exhibits greater snout-vent length (SVL) and tail length (TL), and higher counts of ventral and subcaudal scales than P. patagoniensis. These traits are strongly associated with arboreal habits in snakes (e.g., Harrington et al. 2018). Hartmann and Marques (2005) documented habitat use in specimens corresponding to P. pseudomambasp. nov. from southern Brazil, finding that one-third was located in forested areas and two-thirds in open areas. Similarly, Harrington et al. (2018) reviewed the ecology and traits of arboreal snakes and classified P. patagoniensis sensu lato as semi-arboreal. However, habitat use studies in P. patagoniensis sensu stricto suggest a clear preference for open environments (e.g., Vega and Bellagamba 1990; Di Pietro et al. 2020b). In contrast, the regions inhabited by P. pseudomambasp. nov. offer more potential arboreal niches compared to those occupied by P. patagoniensis, where trees are scarce or absent (Morello et al. 2012). Moreover, differential habitat use in snakes is often associated with variation in food resource utilisation (Luiselli 2006; Goodyear and Pianka 2008).

In the context outlined above, the previously observed geographical variation in the diet of P. patagoniensis sensu lato may offer valuable ecological insight into the morphological differences between P. patagoniensis and P. pseudomambasp. nov. Both species exhibit a broad diet that includes anurans, lizards, snakes (including instances of cannibalism), small birds and mammals (primarily rodents), and even fish and a wide range of arthropods (e.g., ants, coleopterans) (Carreira 2002; Lopez and Giraudo 2008; Quintela and Loebmann 2019; Di Pietro et al. 2020b). The latter prey should be considered secondary, i.e., organisms included in the diet as prey of the actual target prey (e.g., Di Pietro et al. 2020b). However, a notable difference exists in the diet of the species analysed: the consumption of Lycosa spiders as primary prey. This is evidenced by the high frequency of spiders in the stomachs of P. patagoniensis from southern Uruguay and central Argentina (Carreira 2002; Di Pietro et al. 2020b; Chuliver and Scanferla 2024), in contrast to their absence in P. pseudomambasp. nov. from Brazil, northern Uruguay and northern Argentina (Carreira 2002; Hartmann and Marques 2005; López and Giraudo 2008; Quintela and Loebmann 2019). The preference of P. patagoniensis for open areas seems to be linked to its consumption of Lycosa spiders, a genus widely recognised as typical grassland dwellers (Jocqué and Alderweireldt 2005). Accordingly, a diet rich in Lycosa spiders and a predominantly terrestrial lifestyle, are distinguishing traits of P. patagoniensis + P. agassizii compared to P. pseudomambasp. nov. The terrestrial habits of P. patagoniensis and P. agassizii may be partially a result of the scarcity of large forested areas within their range, a factor especially relevant for P. patagoniensis. However, the consumption of Lycosa spiders may also reflect additional ecological or behavioural drivers. Notably, the first reports of spider consumption in the diet of P. patagoniensis were provided by Carreira (2002) and Di Pietro et al. (2020b). The most interesting aspect of the latter study is the significant correlation between prey volume and snake size and the fact that larger individuals continue consuming small prey, particularly Lycosa spiders. In other words, larger specimens of P. patagoniensis tend to consume larger prey while still including Lycosa spiders in their diet. This pattern may reflect the retention of juvenile dietary preferences into adulthood, an idea recently revisited by Chulliver and Scanferla (2024) through a comparative analysis of the morphology and diet of P. patagoniensis sensu lato and P. agassizii. They conclude that P. agassizii is a paedomorphic species whose adults differ from P. patagoniensis by retaining juvenile cranial traits and maintaining a diet rich in spiders, in clear contrast with evidence showing that adult P. patagoniensis never cease to consume Lycosa spiders (Carreira 2002; Di Pietro et al. 2020b).

Finally, this study re-describes P. patagoniensis and formally describes P. pseudomambasp. nov., reflecting the molecular and morphological variation previously observed within P. patagoniensis sensu lato. Distinct patterns of colouration, cranial osteology, scale morphology (shape and counts), body proportions, and aspects of natural history, such as diet and habitat use, allow clear differentiation between the two species.