Genomic analysis reveals a cryptic pangolin species

Significance

Our results provide consistent and compelling genomic evidence for a previously unrecognized genetic lineage distinct from all the eight currently known pangolin species, justifying its designation as a new pangolin species, which becomes the ninth pangolin species or the fifth Asian pangolin (Manis mysteria). Besides genomic assessment, we also present characterization of the new species’ scales and provide morphological evidence that the new species belongs to the Asian rather than the African pangolins. Our finding greatly expands current knowledge of pangolin diversity and evolution. Our analyses also show genomic signatures of a declining population including lower heterozygosity, higher inbreeding level, and genetic load, which makes it merit urgent conservation concerns and assessment.

Abstract

Eight extant species of pangolins are currently recognized. Recent studies found that two mitochondrial haplotypes identified in confiscations in Hong Kong could not be assigned to any known pangolin species, implying the existence of a species. Here, we report that two additional mitochondrial haplotypes identified in independent confiscations from Yunnan align with the putative species haplotypes supporting the existence of this mysterious species/population. To verify the new species scenario we performed a comprehensive analysis of scale characteristics and 138 whole genomes representing all recognized pangolin species and the cryptic new species, 98 of which were generated here. Our morphometric results clearly attributed this cryptic species to Asian pangolins (Manis sp.) and the genomic data provide robust and compelling evidence that it is a pangolin species distinct from those recognized previously, which separated from the Philippine pangolin and Malayan pangolin over 5 Mya. Our study provides a solid genomic basis for its formal recognition as the ninth pangolin species or the fifth Asian one, supporting a new taxonomic classification of pangolins. The effects of glacial climate changes and recent anthropogenic activities driven by illegal trade are inferred to have caused its population decline with the genomic signatures showing low genetic diversity, a high level of inbreeding, and high genetic load. Our finding greatly expands current knowledge of pangolin diversity and evolution and has vital implications for conservation efforts to prevent the extinction of this enigmatic and endangered species from the wild.

Pangolins (monotypic order Pholidota and family Manidae) are believed to be the world’s most heavily poached and trafficked wild mammals (1–5). Overexploitation driven by escalating demand for their meat as luxury food and scales for traditional medicines has driven pangolins to the edge of extinction (6–11). More than one million pangolins were poached in the decade prior to 2014 (https://www.iucnredlist.org/). Eight extant species of pangolins, all of which are listed in “Appendix I” of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (12), are currently recognized (13–15): the four Asian pangolins belonging to the genus Manis, namely the Malayan pangolin M. javanica, the Chinese pangolin M. pentadactyla, the Indian pangolin M. crassicaudata, and the Philippine pangolin M. culionensis; and the four African pangolins belonging to the genera Phataginus and Smutsia: namely, the White-bellied pangolin P. tricuspis, the Black-bellied pangolin P. tetradactyla, Temminck’s pangolin S. temminckii, and the Giant pangolin S. gigantea.

Notably, Zhang et al. (10) recently found that two mitochondrial (mt) haplotypes, H04 and H08, from 27 pangolin scales confiscated in Hong Kong during 2012 and 2013 constituted a distinct lineage that could not be assigned to three of the four Asian pangolin species, namely the Malayan, Chinese, and Indian pangolins, or any African pangolin. They proposed that these two haplotypes derive either from the Asian Philippine pangolin, which was not included in their study, or an unknown lineage of Malayan pangolin (because of the sister-grouping between them and Malayan pangolin), or a putative pangolin species (10). To test these hypotheses, Hu et al. (16) performed an analysis including the corresponding mtDNA gene fragments from the Philippine pangolin, revealing that neither of the new haplotypes could be assigned to the Philippine pangolin and that the Philippine pangolin formed a sister group with the Malayan pangolin but the two mt haplotypes did not (16). These results refuted the first two hypotheses and thus pointed to the existence of a previously unrecognized pangolin species. However, this new species scenario has only been inferred from analyses of short mt gene fragments, which are limited and prone to bias due to restricted sequence variation, exclusive maternal inheritance, and pseudogene coamplification (14, 17–19). Moreover, there have been no subsequent rediscoveries of haplotypes from the putative new species. Consequently, in the absence of more samples and genomic assessments, it was impossible to conclude confidently whether this cryptic species exists and constitutes a unique pangolin species. If the existence of a new species was to be confirmed, there would be an urgent need to determine its evolutionary history, current genetic diversity status, and survival potential, which are important to understand the diversity and evolution of pangolins, and to develop effective conservation strategies for this mysterious species/population (20).

Here, we report that two additional mt haplotypes (designated YN01 and YN02) from six scale samples originating from two confiscations in Yunnan in 2015 and 2019 are more closely related to the two previously identified mt haplotypes (H04 and H08) than to all other known pangolin species, hence implying the rediscovery of the new species of pangolin suggested by Zhang et al. (10) and Hu et al. (16). The finding of scale samples bearing these haplotypes in multiple independent confiscations taking place at different times in different regions strengthened confidence in the existence of this mysterious species/population. Thus, based on a broader sampling of scales from the cryptic new species obtained from confiscations in Hong Kong and Yunnan, we conducted a comprehensive survey of pangolin scale characteristics and whole-genome data of 138 individuals representing all previously recognized pangolin species and the cryptic new species. Ninety-eight of these whole-genome sequences were newly generated. Our analyses provide robust and compelling evidence that this cryptic species represents a separately evolving pangolin lineage whose demographic history is distinct from all other recognized pangolin species. These findings provide a solid and rigorous basis for its formal recognition as the ninth pangolin species or the fifth Asian pangolin species, supporting a new taxonomic classification of pangolins. The effects of glacial climate changes and recent human activities driven by illegal trade are inferred to have caused its population decline with the genomic signatures showing low genetic diversity, a high level of inbreeding, and a high genetic load. Our study has vital implications for conservation efforts to prevent the extinction of this enigmatic and putatively endangered species from the wild.

Results

Morphological Analysis.

A morphological analysis was performed using 33 scale samples from multiple confiscations in Hong Kong and Yunnan that represented four mt haplotypes (the H04 and H08 haplotypes and the YN01 and YN02 haplotypes) of the cryptic new species (Fig. 1 and SI Appendix, Tables S1 and S2). Five of the 33 scale samples were attached to skin (Fig. 1A), three were attached to claws (one from the forefoot and two from the hindfoot) (Fig. 1B), and the rest were individual disarticulated scales (Fig. 1 C–H), 20 of which were determined to originate from a pangolin’s tail, dorsum, ventral parts, or head (Fig. 1 C–G).

Fig. 1.

Thirty-three scales from multiple confiscations in Hong Kong and Yunnan representing the published H04 and H08 haplotypes and the novel YN01 and YN02 haplotypes of the cryptic new species. (A) Five scales attached to skin fragments. (B) Three scales attached to claws, including two from the hindfoot (H04) and one from the forefoot (H08). (C) A scale from the dorsum of the tail whose origin was deduced from the fact that it is smaller than scales from other body parts, has a V-shaped posterior margin, and lacks a ridge bulge. (D) Thirteen scales from the dorsal side, whose origins were deduced from their large size, their fan-like, rhombic, or inverted trapezoidal shape, and their lack of ridge bulges. (E) Three scales from the ventral side, whose origins were deduced from their small size and the presence of a ridge in the middle of the scale. (F) Two scales from the head, whose origin were deduced from their small size and fan-like or almost round shape. (G) One scale from the lateral rows of the tail, whose origin was deduced from its small size and reduced triangular shape. (H) Five scales from unknown body parts. All body part assignments (C–G) are based on previously reported characteristics (21–23).

The scale samples were found to possess morphological characteristics of Asian pangolins by comparing to those known to distinguish Asian and African pangolins (14, 21). These characteristics are 1) hairs projecting between the scales (Fig. 1 A, D, and E), which are absent in African pangolins; 2) the third claw of the hindfoot being approximately 1 cm longer than the fourth claw (Fig. 1B), whereas these claws are similar in length in African pangolins; and 3) a smooth or V-shaped posterior margin in the scales of the median row on the dorsal side of the tail (Fig. 1C), which have a three-cusped shape in African pangolins. The cryptic new species thus seems to be more closely related to the Asian pangolin species than to African pangolins, in accordance with the earlier haplotype analysis (10, 16), the mt haplotype analysis in the present study (SI Appendix, Fig. S1), and the genomic analyses results below. However, the overlapping scale shape between the new species and the other Asian pangolins made it impossible to distinguish the new species from the known ones on morphological grounds (Fig. 1) (22, 23).

Whole-Genome Dataset Generation.

We obtained whole-genome resequencing datasets representing 138 different individuals from all eight known pangolin species, including 20 Indian pangolins (M. crassicaudata), 19 Chinese pangolins (M. pentadactyla) (24), 20 Malayan pangolins (M. javanica) (24), one Philippine pangolin (M. culionensis) (25), 20 White-bellied pangolins (P. tricuspis), 20 Black-bellied pangolins (P. tetradactyla), 29 Giant pangolins (S. gigantea), and two Temminck’s pangolins (S. temminckii), as well as seven individuals belonging to the putative cryptic new pangolin species (SI Appendix, Table S1). Ninety-eight of these datasets were newly generated in the present study. The average clean data depth and genome mapping rate were 41.59-fold and 91.07%, respectively. A total of 203.19 Mb high-quality autosomal single-nucleotide polymorphisms (SNPs) and 0.23 Mb X chromosome SNPs were obtained. We also obtained 1,773 single-copy orthologous coding sequences comprising 1,872,840 nucleotides from the genomes. In addition, 169 mitogenomic sequences with an average length of 14,919 bp were obtained from the 138 genomic sequences and 31 published sequences (SI Appendix, Table S3) (14, 26–31).

Phylogenomic Analyses.

Seven phylogenies were inferred from autosomal SNPs (ML gene-tree; SNAPP species-tree), X chromosome SNPs (ML gene-tree), concatenated single-copy coding sequences (ML gene-tree; MP-EST and ASTRAL species-trees), and mitogenome sequences (ML gene-tree). In all phylogenies, the individuals from the cryptic new species formed a well-supported and separate monophyletic lineage (ML, MP-EST, and ASTRAL BS = 100%; SNAPP BP = 1.0) that was distinct from the other pangolin species. Moreover, all of the pangolin species including the eight known species and the cryptic new species were separated into two clades (Fig. 2 A–D). One clade comprised four Asian pangolin species and the cryptic new species. The Chinese pangolin diverged first, followed by the Indian pangolin and then the cryptic new pangolin species, and finally the Philippine pangolin and the Malayan pangolin split. The other clade consisted of four African pangolin species. Sister-species relationships between the White-bellied pangolin and the Black-bellied pangolin and also between the Giant pangolin and Temminck’s pangolin were consistently observed within this clade. These interspecific relationships were all supported by full bootstrapping and posterior probabilities (ML, MP-EST, and ASTRAL BS = 100%; SNAPP BP = 1.0). The relationships among the eight well-accepted pangolin species are consistent with the previously reported topology (14, 16). The placement of the cryptic new species as an independent phylogenetic lineage within the Asian pangolins agrees with the earlier analyses of short mt gene fragments (10, 16) and corroborates our morphological analyses.

Fig. 2.

Phylogenomic analyses, genetic difference evaluations, species delimitation, and identification of genomic island of differentiation between the cryptic new species and other Asian pangolin species. (A) Gene tree (ML) inferred from autosomal and X chromosome SNPs. They both obtained a well-supported and consistent phylogenetic tree topology. (B) Gene tree (ML) inferred from mitogenome sequences. (A and B) The number of individuals for each species is indicated (SNP/mitogenome). (C) Species tree (SNAPP) inferred from the autosomal SNPs. (D) Gene tree (ML) and species tree (MP-EST and ASTRAL) inferred from concatenated single-copy coding sequences. Interspecific divergence times (nodes from a to j) are shown on the branches, including median values and their 95% HPDs (purple bars). The red boxes on the nodes (A–D) indicate full bootstrap values (ML, MP-EST, and ASTRAL BS = 100%; SNAPP BP = 1.0) at the nodes representing relationships among pangolin species. (E) Pairwise comparisons of nucleotide site differences between the cryptic new species and the four recognized Asian pangolin species across 50-kb genomic windows. (F) Species delimitation analyses based on autosomal SNPs using the BFD* method. (G–L) Genomic regions of elevated differentiation identified by performing Dxy and Fst analyses between the cryptic new species and the Chinese pangolin (G), Malayan pangolin (H), Indian pangolin (I); and between the Indian pangolin and the Chinese pangolin (J), Malayan pangolin (K); and between the Chinese pangolin and the Malayan pangolin (L). Data points above the horizontal dotted line (corresponding to the 5% right tail of the empirical Fst distribution) and on the right side of the vertical dotted line (the 5% right tail of the empirical Dxy distribution) were identified as genomic islands of differentiation for each species pair. Reproduction-related genes with significant signals are indicated by arrows and red circles.

Divergence Time Estimation.

Divergence times estimated using concatenated single-copy coding sequences (Fig. 2D and SI Appendix, Table S4) show that Asian and African pangolin species diverged at 33.19 Mya [95% highest posterior density (HPD) = 30.34 to 39.80 Mya]. Within Asian pangolins, the Chinese pangolin diverged first at 16.85 Mya (95% HPD = 12.92 to 22.36 Mya), followed by the Indian pangolin at 10.15 Mya (95% HPD = 7.55 to 13.91 Mya). The cryptic new pangolin species diversified from Philippine pangolin and Malayan pangolin about 5.05 Mya (95% HPD = 3.65 to 7.08 Mya). The Philippine pangolin and the Malayan pangolin split most recently at 3.23 Mya (95% HPD: 2.17 to 4.75 Mya).

Genomic Differences, Species Delimitation, and Genomic Islands of Differentiation.

Pairwise comparisons of genetic differences across the genomes suggested that the nucleotide site differences between the cryptic new species and each of the four recognized Asian species (ranging from 180 to 999 bp in 50,000-bp windows) are comparable in magnitude to the species-level divergence among the four recognized Asian species (ranging from 79 to 971 bp in 50,000 bp windows) (Fig. 2E). This supports the genetic distinction of the cryptic new species as a fifth Asian pangolin species. The genetic differences between the mitogenome sequences also indicated divergence at the species level (SI Appendix, Fig. S2).

The species delimitation analyses of autosomal SNPs using the Bayes factor delimitation method (BFD*) identified five candidate species of Asian pangolins based on both the highest marginal likelihood and decisive Bayes factors (BFs) (MLE value = −221,159.61, all BF values > 10) (Fig. 2F). This result supported a division of Asian pangolins into five species, including the four currently recognized Asian species and the cryptic new species. The correspondence between this division and the five separate monophyletic lineages of the Asian pangolin clade observed in the phylogenies (Fig. 2 A–D) provides further support for the existence of a fifth Asian pangolin species. The same pattern of five Asian pangolin species was also suggested by the species delimitation analysis of the mitogenomes, again indicating that the cryptic new species is a distinct Asian pangolin species (SI Appendix, Fig. S3).

The “genomic islands of differentiation” containing genes related to reproductive isolation are often thought to play important roles in speciation (32, 33). We therefore searched for genomic regions of elevated differentiation between the cryptic new species and other Asian pangolin species. This analysis identified the genes related to gamete generation, meiotic cell cycle, fertilization, sterility, intracellular estrogen, and androgen receptors in genomic regions exhibiting elevated differentiation between the cryptic new species and each of the other Asian pangolin species (Fig. 2 G–I). Moreover, some of these reproduction-related genes are also present in such islands of differentiation between other Asian pangolin species (Fig. 2 J–L). For example, a key meiosis-related gene TEX11 (34) was identified in all pairwise comparisons between the cryptic new species and the Indian and Malayan pangolins (Fig. 2 H, I, and K). Two key genes involved in spermatogenesis, SPATA5 and KLHL10 (35, 36), were also identified: The former was identified between the cryptic new species and the Indian pangolin (Fig. 2I), and between the Malayan and the Indian/Chinese pangolins (Fig. 2 K and L), while the latter was identified between the cryptic new species and the Chinese pangolin (Fig. 2G), and between the Malayan and the Chinese pangolins (Fig. 2L); The DCAF1 gene, which is responsible for activating estrogen receptor binding activity (37), was identified between the cryptic new species and the Chinese pangolin (Fig. 2G), and between the Chinese and the Indian pangolins (Fig. 2J). The identification of reproduction-related genes in the genomic islands of differentiation between the cryptic new species and the other Asian pangolin species supports the conclusion that the cryptic new species has undergone species-level differentiation from other Asian pangolins.

Genetic Diversity, Inbreeding Level, and Genetic Load Evaluation.

Whole-genome heterozygosity (He) is a metric used to assess genetic diversity and evolutionary potential. Its value in the eight recognized pangolin species ranges from 0.040% in the Philippine pangolin to 0.295% in the White-bellied pangolin with an average of 0.158%. The He of the new pangolin species was estimated to be 0.094% (Fig. 3A), suggesting a relatively low level of genetic diversity compared to other pangolin species. Indeed, its genetic diversity is comparable to that for some of the world’s endangered mammals (38) (SI Appendix, Fig. S4), including the Aye Aye (0.070%), Malayan tapir (0.090%), Amazon river dolphin (0.090%), Mexican mantled howler monkey (0.100%), and giant otter (0.110%).

Fig. 3.

Genetic diversity, inbreeding levels, and genetic load evaluation. (A) Whole-genome genetic diversity based on the average heterozygosity (He) among all individuals for each species. (B) Inbreeding levels calculated based on the average inbreeding coefficient (F_ROH) among all individuals for each species. (C) Long-ROHs (>1 Mb; ROH_1Mb) detected in the new pangolin species and three other Asian pangolin species. Evaluations of genetic load based on LOF mutations (D), missense mutations (E), deleterious GS mutations (F), and synonymous mutations (G) among all individuals for each species. Error bars indicate SD values based on 100 resamples.

Low genetic diversity is often accompanied by an elevated inbreeding level (39, 40). The genomic inbreeding coefficients (F_ROH) of the eight recognized pangolin species range from 2.32% for the White-bellied pangolin to 34.03% for the Malayan pangolin, with an average of 13.96%. The F_ROH of the new pangolin species is estimated to be 16.10% (Fig. 3B), suggesting a relatively high inbreeding level compared to other pangolin species. In addition, the long blocks of ROH (>1 Mb; ROH_1Mb) indicating inbreeding within the last 50 y were only detected in the new pangolin species (0.18%) and three other Asian pangolin species: the Chinese pangolin (0.06%), Indian pangolin (0.13%), and Malayan pangolin (0.82%) (Fig. 3C).

High inbreeding levels can lead to increased homozygosity of recessive deleterious mutations (24, 41) that disrupt gene function or reduce individual fitness. To estimate genetic load across the pangolin species, including the new pangolin species, we defined genotypes of major homozygous alleles (i.e., alleles homozygous in over 50% of individual Asian and African pangolins) as the ancestral state, while those of minor homozygous and heterozygous mutations represent the derived state. Strikingly, the ratio of the derived homozygous sites to the derived homozygous and heterozygous sites in the new species [0.966 for LOF (loss of function); 0.979 for missense mutations; 0.950 for Grantham score (GS); 0.979 for synonymous mutations] (Fig. 3 D–G) was higher or significantly higher than in six of the eight currently recognized pangolin species (0.759 to 0.951 for LOF; 0.782 to 0.978 for missense mutations; 0.753 to 0.938 for GS; 0.802 to 0.977 for synonymous mutations), indicating that the new species has a relatively high genetic load. Furthermore, enrichment analyses of the 701 genes exhibiting homozygous deleterious LOF mutations in all individuals of the new species revealed that 73 of these genes are involved in 40 pathways that are significantly enriched in functions related to cancer, disease, and immunity (SI Appendix, Fig. S5). This may increase the new species’ susceptibility to disease and reduce its ability to adapt to environmental changes.

Demographic History and Gene Flow.

The PSMC (pairwise sequentially Markovian coalescence) analysis revealed distinct demographic histories for all of the pangolin species, including the new species (Fig. 4A). The new species suffered from an obvious population decline since 300 thousand years ago (kya), mostly likely caused by the Penultimate Glaciation (PG, 300 ~ 130 kya) (42). This led to a continuous decline in effective population size (Ne) from the PG to the Last Glacial Period (LGP, 120 to 12 kya), culminating in a bottleneck around 30 kya. Chinese pangolin and Indian pangolin also experienced a similar demographic trajectory to that of the new species during this period, but at the end of the LGP, the new species experienced slight population expansion, Indian pangolin underwent obvious expansion, and Chinese pangolin tended to stabilize. Conversely, the Malayan pangolin underwent an apparent population expansion in the early LGP, peaked in the middle of LGP and then declined again. Based on its island distribution, the expansion of the Malayan pangolin during the glacial period has been attributed to episodes of falling sea level (24).

Fig. 4.

Demographic history reconstruction and gene flow analyses. (A) Results of a PSMC analysis of Asian pangolin species, including the new species (shown in the upper half of the figure), and African pangolin species (lower half) based on a mutation rate per generation of 1.47 × 10⁻⁸ (43) and an average generation time of 1 y (44). Note that 100 bootstrap replicates were performed to assess variance in the demographic trajectories. Glacial periods, including the PG, LGP, and last interglacial period are indicated by background shading. Surface temperatures (temp.) over the last 1 Mya based on δ ¹⁸O data (45) are indicated by dotted lines. (B–D) The gene flow analysis results for Asian pangolins, including the new species, using the D statistic (B), f4-ratio test (C), and the f-branch method (D). The depth of the color for each block in the matrix represents the degree of migration ratio. The darker color means the higher migration ratio. The gray shading corresponds to tests that are not applicable to the provided phylogeny.

Analyses of the genome-wide SNP data based on the D statistic (Fig. 4B), the f4-ratio test (Fig. 4C), and the f-branch method (Fig. 4D) provided evidence of historical gene flow between the new species, the Chinese pangolin, and the Indian pangolin (D = 0.14 ~ 0.39, |Zscore| > 3, P < 0.001) (Fig. 4B and SI Appendix, Table S5). These three species also had the most similar demographic histories from the PG to LGP, as described above (Fig. 4A). However, compared with the strong gene flow between Chinese and Indian pangolins, the new species showed limited gene flow with these two species (Fig. 4 B–D and SI Appendix, Table S5). No gene flow signal was detected among the African pangolin species (SI Appendix, Table S5).

Discussion

Advances in genomics have been thought to significantly enhance the understanding of biodiversity and evolution, which has a wide application for species delineation and conservation (19, 46–48). The existence of an unrecognized fifth Asian pangolin species was previously suggested by genetic studies based on short mtDNA fragment sequencing from scales confiscated in Hong Kong (10, 16). Our results based on analyses of whole genomes and mitogenomes from scales confiscated in Hong Kong and Yunnan provide consistent and compelling genomic evidence for the existence of such a lineage. Moreover, we have shown that this lineage is distinct from all eight currently known pangolin species, hence justifying its designation as a new pangolin species, which becomes the ninth pangolin species or the fifth Asian pangolin, i.e., Manis sp. This new pangolin species was distinguished from known species by all genomic measures considered here, showing a fully supported independent evolutionary lineage in all phylogenetic trees (Fig. 2 A–D), a similar level of genomic differences to those among other recognized Asian pangolin species (Fig. 2E and SI Appendix, Fig. S2), a species-level recognition in probabilistic species delimitation analyses (Fig. 2F and SI Appendix, Fig. S3), as well as the reproduction-related genes identified in the genomic islands of differentiation between the new species and other Asian pangolin species (Fig. 2 G–L). Our results thus extend and substantiate the results of earlier genetic analyses by not only identifying additional new species haplotypes from independent confiscations but also providing a rigorous and conclusive genome-level assessment of the differences between the new species and all other recognized pangolin species, confirming the existence of a novel species. We also present characterization of the new species’ scales (Fig. 1) and provide clear morphological evidence that the new species belongs to the Asian rather than the African pangolins. Although there is currently no formal description of the type specimen, our comparative analyses of scale characters and DNA across all the pangolin species indicated that this mysterious new species belongs to the Asian pangolin (genus Manis). Considering that the trafficking pressures faced by pangolins merit taxonomic haste, we tentatively name it as Manis mysteria (the Asian mysterious pangolin). In the future, a formal taxonomic description and nomenclature will require the discovery of whole animals and a type specimen. A more precise and in-depth analysis on morphology such as morphometrics and computed tomography scans will be needed.

Although this new species currently appears to be cryptic, our genome analyses indicate that it diverged 5.05 Mya during the Pliocene (Fig. 2D), a slightly younger age than previously reported based on mtDNA (6.95 Mya) (16). Our divergence time estimate indicates that the differentiation and speciation of extant pangolin species, including the new species, occurred from the mid-Miocene through the Pliocene and extended into the Pleistocene. This is consistent with fossil evidence that suggested a global shift to cooler and more arid climates following the mid-Miocene climatic optimum, which may have restricted pangolins’ geographic range to the more tropical environments in Africa and Southeast Asia, including their present-day distribution (14, 49, 50). The resulting radiation formed many reported extinct and extant pangolin species that have been discovered in Asia and Africa throughout the Pliocene and Pleistocene (14), indicating that pangolins once possessed considerably greater diversity than they do at present. It seems likely that the new Asian pangolin species, which was dated back to the Pliocene, emerged during this radiation.

We also provided a continuous reconstruction of the new species’ demographic trajectory, which is clearly distinct from those of other pangolin species (Fig. 4A). Our results indicate that the new species went through a serious population bottleneck caused by the cold climate during the Pleistocene glaciation (Fig. 4A). Alarmingly, like other pangolin species, the new species may be at severe risk due to hunting and trafficking since it was discovered by analyzing scale samples confiscated from smugglers. Such activities would accelerate its population decline.

It is generally believed that smaller populations have lower levels of genetic diversity as well as higher levels of inbreeding and genetic load (51–53). All of these factors can reduce fitness and survival in the face of changing environmental conditions, creating an increased risk of extinction (54). Analyses of the genetic impact of population declines are therefore vital when designing conservation measures for endangered species (55). Our results showed genomic signatures of a declining population including the relatively low genetic diversity when compared to other pangolins as well as high levels of inbreeding and genetic load and that inbreeding still occurred in recent times (Fig. 3). The adverse consequences of population decline are thus apparent in its genome. This implies an urgent need for research to establish a formal nomenclature of this new species based on morphological descriptions of whole specimens as well as geographic range and ecology that would ultimately support conservation measures and assessments independent of those for other pangolin species.

While the geographic distribution of the new species is currently unknown, our results provide some insight into its likely range. One possibility is that the distribution of the new species may overlap with or be adjacent to the ranges of the Chinese and Indian pangolins. These areas include several global biodiversity hotspots, such as Indo-Burma, Himalayas, and the Mountains of Southwest China. This speculation is based on our gene flow results detecting the historical gene flow signal between the new species and both Chinese and Indian pangolins (Fig. 4 B–D), which may have been facilitated by geographic proximity. The other possibility is the isolated islands in Southeast Asia because those islands have been thought to be evolutionary refugia with many relict and endemic species (56, 57), and indeed, the Asian pangolins trafficked into Hong Kong and Yunnan have been reported to be mainly originated from Southeast Asia (10, 24). Unfortunately, this new pangolin species has not been discovered in the wild to date, so its morphological description is based entirely on scattered confiscated scales. Given that the present morphological analyses of the scales can only distinguish Asian and African pangolins, the new species may be morphologically indistinguishable from other Asian pangolin species and was therefore overlooked in the wild. It could also have been overlooked because its range occurs in a region that has not been well studied or because of its elusive behavior. Further ecological and phenotypic data are expected to uncover the veil of this mysterious species.

Material and Methods

Sample Collection and Species Identification.

We collected a total of 127 confiscated pangolin scale samples from Yunnan (samples donated by Yunnan Provincial Forest Public Security Bureau, and the Animal Branch of the Germplasm Bank of Wild Species of Chinese Academy of Sciences) and Hong Kong (samples donated by the Agriculture, Fisheries and Conservation Department of the Hong Kong SAR Government). All necessary research permits and ethics approvals were granted. The 44 confiscated scale samples from Hong Kong were reported and identified by using a 600-bp region of COI gene: 10 scales were assigned to the giant pangolin, four to the Black-bellied pangolin, three to Temminck’s pangolin, and 27 to the cryptic new species (10, 58). The 83 confiscated scale samples from Yunnan were identified in the present study. The pieces of tissue attached to the scales were used for DNA extraction. The outer surface (0.5 mm) of tissues was first removed using a surgical blade to eliminate possible surface contaminants. Then, a small section of tissue was cut out, transferred to a 2-mL eppendorf tube, and washed with ethanol and 10X Phosphate buffered saline two times. DNA from these samples was subsequently extracted using the Magen Hipure DNA Micro Kit (Magen, China). The COI gene was amplified by PCR with the primers F1 (5′-TGR CTM TTT TCA ACA AAT CAC AA-3′) and R1 (5′-GGG RTG YCC GAA GAA TCA GAA T-3′). PCR amplification was carried out using an initial denaturation step (98 °C, 2 min), followed by 35 cycles of denaturation (98 °C, 10 s), annealing (52 °C, 15 s), and extension (72 °C, 1 min) followed by a final extension (72 °C, 5 min). We did not detect any stop codons or indels in the reading frame of these sequences that would indicate putative nuclear mt DNA segments. BLAST searches of GenBank were performed using these COI sequences (GSA: SAMC1131749-SAMC1131786; GenBank: OQ925816–OQ925853) as queries to check for similarities with published pangolin sequences (59), resulting in the identification of 20 Indian pangolins, 20 White-bellied pangolins, 16 Black-bellied pangolins, and 21 Giant pangolins sequences, along with two haplotype sequences from six samples (referred as haplotypes YN01 and YN02 hereafter) that showed the highest similarities (98.17 to 99.83%) with the two haplotype sequences reported as the cryptic new species from 27 samples confiscated in Hong Kong (referred to as haplotypes H04 and H08) (10).

We further verified the species identity of the two newly found haplotypes (haplotypes YN01 and YN02) of the new species using a tree-based method by assessing their phylogenetic clustering with 104 COI haplotypes including 66 previously published haplotypes representing the eight recognized pangolin species (10, 14, 24, 28, 29, 50, 60), 36 generated in this study representing four previously known pangolin species, and two haplotypes previously assigned to the cryptic new species (haplotypes H04 and H08) (10) (SI Appendix, Table S2). The phylogenetic tree confirmed that the two new haplotype sequences reported in the present study (YN01 and YN02) were not grouped with those of the eight currently recognized pangolin species but did cluster with the two previously reported new species haplotype sequences (H04 and H08) to form a monophyletic lineage (SI Appendix, Fig. S1). Based on the species identification analysis using the COI sequences, a total of 127 confiscated pangolin scale samples were used in the analysis, including 20 Indian pangolins, 20 White-bellied pangolins, 20 Black-bellied pangolins, 31 Giant pangolins, and three Temminck’s pangolins scales as well as 33 scales from the cryptic new species.

Morphological Diagnosis.

Thirty-three scale samples from multiple confiscations in Hong Kong (2012 and 2013) and Yunnan (2015 and 2019) representing four mt haplotypes of the cryptic new species (Fig. 1) were used for the morphological analyses. Among the 33 scale samples, five were attached to skins, three were attached to claws (one from the forefoot and two from the hindfoot), and the rest were individual disarticulated scales. In accordance with published descriptions of pangolin scale morphology (22, 61), we first determined which part of the body each scale originated from and then determined whether they belonged to Asian pangolins based on reported differences in the morphological characteristics of different body parts between Asian and African pangolins (14, 21). These morphological characteristics include the length of the claw of the hindfoot, whether there are projecting hairs between the scales, and the shape of posterior margin in the scales of the median row on the dorsal side of the tail. The length of the claw from the hindfoot was determined by measurement using tape, while the other characteristics were determined by visual inspection. The determination of the morphological characteristics has been confirmed by multiple persons.

Whole-Genome Resequencing Data.

Genome resequencing datasets with acceptable DNA quality/quantity were obtained for 111 of the 127 confiscated pangolin scale samples, including 20 Indian pangolins, 20 White-bellied pangolins, 20 Black-bellied pangolins, 31 Giant pangolins, three Temminck’s pangolins, and 17 cryptic new species samples (GSA: CRA010100 and NCBI: PRJNA962256). Illumina sequencing libraries with 500-bp inserts were generated for the samples from the Yunnan confiscations and sequenced on the Illumina NovaSeq platform at Berry Genomics Co. (Beijing, China) to generate 150-bp paired-end reads. The samples from the Hong Kong confiscations were used to construct proprietary DNBSEQ sequencing libraries with 300- to 500-bp inserts and sequenced on the BGISEQ-500 sequencing technology platform at BGI Hong Kong Co. (Hong Kong, China) to generate 150-bp paired-end reads. All the paired-end reads (raw data) were trimmed to remove adapter sequences and low-quality sequences to obtain clean data. For Illumina data, all reads containing >20% low-quality nucleotides (Q <= 5) or >10% ambiguous nucleotides were removed. For BGISEQ data, all reads containing more than 50% low-quality bases (Q <= 20) or more than 3% ambiguous nucleotides were removed. The final mean content of high-quality nucleotide sites across all sequences was 96.84%.

Besides the genomic sequences from five recognized pangolin species and the cryptic new species newly generated here, the genomic sequences from the other three recognized pangolin species, including one Philippine pangolin (25), 20 Chinese pangolins (24), and 20 Malayan pangolins (24), were downloaded from the NCBI GenBank database (SI Appendix, Table S1).

In total, 152 samples of whole-genome resequencing datasets representing all eight currently recognized pangolin species and the cryptic new species were obtained for SNP calling.

SNP Calling.

We used Burrows-Wheeler Alignment-mem (62) to align each high-quality resequencing dataset with the reference high-quality de novo genome of the Malayan pangolin (NCBI accession number: PRJNA529512; genome size: 2.45 Gb, scaffold N50: 13.85 Mb) (24). BAM alignment files were generated using SAMtools v.1.3 (63). PCR duplicates were removed using PICARD (http://picard.sourceforge.net). Scaffolds with lengths below 100 kb were excluded. The Genome Analysis Toolkit (GATK) v.4.1.8 (64) was used for SNP calling. The Indel (Insert and deletions) realignment was performed using the IndelRealigner algorithm. A gvcf file for each sample was obtained using the "HaplotypeCaller" module, and all gvcf files were merged using “CombineGVCFs” in GATK. SNP calling was then conducted with “GenotypeGVCFs” and “SelectVariants” was used to obtain candidate SNPs. To generate high-quality SNPs, the candidate SNPs were filtered using GATK with the following criteria: QUAL < 30.0 || QD < 2.0 || MQ < 40.0 || FS > 60.0 || SOR > 3.0 || MQRankSum < −12.5 || ReadPosRankSum < −8.0 || SB> = −1.0. Then, SNPs with allele frequencies below 20% and a depth distribution of all sites below 2.5% or above 97.5% were filtered using VCFtools v.0.1.13 (65).

For the autosome SNP dataset, the candidate sex chromosome scaffolds from the assembled genomes that showed >60% sequence similarity with sex chromosomes of the domestic dog and those of the human were excluded by LASTZ (24). For autosome analyses, 203.19-Mb SNPs were retained.

Besides the autosome SNP dataset, the X chromosome SNPs dataset was also generated. By bidirectional BLASTing (66) with a cutoff e-value of 10⁻¹⁰ and the optimal matching between the reference Malayan pangolin genome and the most closely related carnivoran species genomes with chromosome-level assemblies and high-quality annotations, including domestic cat (GCA_000181335.4), domestic dog (GCA_014441545.1), Canada lynx (GCA_007474595.1), meerkat (GCA_ 006229205.1), and giant panda (GCA_002007445.2) from the Ensemble database (https://asia.ensembl.org/index.html), a total of 48 unique X-linked genes on 15 candidate sex chromosome scaffolds were identified, and 0.23-Mb SNPs were extracted for X chromosome analyses.

Kinship Analyses to Remove Duplicate Individuals from Samples.

Because the inclusion of samples representing duplicate individuals could bias subsequent analyses, the SNP data for the 152 samples were subjected to a kinship analysis using Kinship-based Inference for Genome-wide association studies (KING) (67) to remove samples representing duplicate individuals. Kinship coefficients were estimated with the “--make-king-table” command in PLINK v.2.0 (68), whose output reflects the proportion of SNPs with identical states (IBS0, identity by state zero) between individuals. Negative coefficients indicate no relationship between individuals, while positive coefficients indicate genealogical links between individuals. After removing 14 potential duplicates and identical twins identified in this way (kinship coefficients > 0.354), 138 unique individuals were retained for further analysis (SI Appendix, Table S1).

Single-Copy Orthologous Coding Sequence Dataset Generation.

In addition to the SNPs dataset, a genomic-level single-copy coding sequence dataset based on orthologous genes identified from the eight recognized pangolin species and the cryptic new species was generated for phylogenetic tree construction. We first downloaded protein sequences from the de novo genomes of the Malayan pangolin (PRJNA529512) (24), Chinese pangolin (PRJNA529513) (24), and White-bellied pangolin (69), as well as from two outgroups from the Order Carnivora, the sister lineage of the Pholidota: domestic dog (PRJNA68156) (70) and domestic cat (PRJNA773801) (71). Single-copy orthologous sequences from these five species were then extracted using OrthoFinder v.2.5.4 (72). For the remaining pangolin species without de novo genome assemblies, we selected one individual with a high resequencing depth (~30-fold), with the exception of the Philippine pangolin, which was represented by a single individual with <30-fold resequencing depth, to extract the corresponding gvcf files, and remerged them using “GenotypeGVCFs” in GATK. The Asian and African pangolin genomes were mapped to Malayan pangolin (24) and White-bellied pangolin (69) genomes, respectively. The single-copy orthologous coding sequences were obtained using CDSs and gene IDs from the reference genomes. Finally, we obtained the 1,773 single-copy orthologous sequences from all pangolin species. Sequences were aligned using PRANK v.170427 (73) and ambiguous sites were removed using Trimal v.1.4.1 (gap = all) (74). We also removed alignments with effective sequence lengths (i.e., those for which the inclusive nucleotide sites do not include Ns for any taxa) below 500 bp.

Mitogenome Assembly and Dataset Generation.

Mitogenome assembly was performed using the clean genomic resequencing data representing 138 pangolin individuals. The published mitogenomes of the eight recognized pangolin species from ref. 14 were used as references for the mitogenome assembly. The mitogenome of the Malayan pangolin species (MG196309) was also used as a reference for the assembly of the cryptic new species due to the close relationship between them (Fig. 2B and SI Appendix, Table S3). The paired-end reads from the genomic resequencing data of each individual were first combined into a single fastq file and then assembled with default parameters in MITObim v.1.9.1 (75), which was run 3 to 5 times independently. Finally, 138 mitogenomes were assembled and checked with manual correction. Additionally, 31 published mitogenomes representing eight pangolin species were downloaded from NCBI and added to the mitogenome analyses (SI Appendix, Table S3). Domestic dog (GenBank accession AB499817) and domestic cat (KP202278) were used as outgroups (SI Appendix, Table S3). A total of 169 mitogenome sequences were used in the analyses after excluding the highly repetitive and poorly assembled D-loop region. Sequences were aligned using PRANK v.170427 (73), and ambiguous sites were removed using Trimal v.1.4.1 (gap = all) (74).

Phylogenomic Analyses.

Phylogenetic analyses were performed based on four datasets: autosomal SNPs, X chromosome SNPs, concatenated single-copy orthologous coding sequences, and mitogenome sequences. The autosomal SNPs were thinned with a 10-kb window size to minimize the influence of linkage imbalance between loci and then 0.21 Mb SNPs after thinning were obtained. Autosomal SNPs and X chromosome SNPs were used to reconstruct Maximum Likelihood (ML) tree using IQtree v.1.6.12 (76). 1,000 bootstraps were performed to evaluate node support. A species-tree was also reconstructed using the SNAPP (77) method by analyzing individuals of each pangolin species with high resequencing depths (~30-fold, although as noted previously, the only data available for the Philippine pangolin derived from an individual with a <30-fold sequencing depth). Only single individuals were used in this case because of the method’s high computational cost. The SNAPP (v.1.5.2) analyses were implemented in BEAST2 v.2.6.6 (77) and were run in parallel for one million generations, sampling every 1,000 steps with two independent replicates. Log files were examined in Tracer v.1.6 (http://beast.bio.ed.ac.uk/Tracer), and ESS (effective sample sizes) values were checked for stationarity across all metrics. After combining all trees sampled from the two independent runs in LogCombiner v.2.6.3 (discarding a burn-in of 10% from each one), a maximum clade credibility tree was constructed using TreeAnnotator and a cloudogram of the posterior distribution of the SNAPP species trees was generated using DensiTree (78).

The concatenated single-copy orthologous coding sequences (1,773 genes and 1,872,840 nucleotides) and the mitogenome sequence (14,919 bp) were also used to reconstruct the ML gene-tree as described above, respectively. In addition, two species-trees were reconstructed using the MP-EST (79) and ASTRAL (v.5.7.8) (80) methods based on the concatenated single-copy orthologous coding sequences. First, the GTRGAMMA model and 100 bootstrap replicates were used to build an ML tree for single genes using RaxML v.8.2.12 (81); then, the combined trees for all single-copy orthologous gene sequences were used in both species-tree analyses. For the MP-EST analysis, five independent runs and five searches per run were performed to check the consistency of the results.

Divergence Time.

The divergence times of the pangolin species were estimated based on the single-copy orthologous coding sequences using the Bayesian MCMC program MCMCTREE, which was implemented in PAML v.4.9 (82) using a GTR model. Birth (λ), death (μ), and sampling (ρ) priors of λ = 1, μ = 1, and ρ = 0 were used. The transition/transversion rate ratio (kappa gamma), the shape parameter for rate heterogeneity between sites (alpha gamma), and the prior on rates (rgene gamma) were specified as (6, 2), (1, 1), and (1, 12.91), respectively. The first 200,000 iterations were discarded as burn-in, and sampling was performed every 50 iterations until 2,000,000 samples were gathered. The key indicator of ESS >200 was evaluated by Tracer 1.6 (http://beast.bio.ed.ac.uk/Tracer) to ensure the convergence, stability, and effective sample size of parameters. The resulting divergence time confidence was displayed using FigTree v.1.4.3 (http://tree.bio.ed.ac.uk/software/figtree).

Three calibration points were applied using soft-bound calibrations: 1) the fossil records (Min: UALVP 50993 and 50994; Max: Molecular estimate) for the split at 37.3 to 66 Mya between the two outgroups, dog and cat (83, 84); 2) the fossil records (Min: Daphoenus and Hesperocyon; Max: UALVP 50993 and 50994) for the divergence at 66 to 87 Mya between the outgroups and pangolins (14, 83); and 3) the molecular divergence time of crown pangolins at 31 to 45 Mya (13, 85).

Genetic Difference Calculation, Species Delimitation Analysis, and Identification of Genomic Islands of Differentiation.

The genetic differences across the genomes of the cryptic new species and the four recognized Asian pangolins were estimated by pairwise comparison of nucleotide sites differences (N_VARIANTS) based on 50-kb windows using VCFtools v.0.1.13 (64) with “vcftools --window-pi”. Because the genomic data were available for seven individuals of the new species (after the removal of the duplicate individuals by kinship analyses), we included the same number of individuals from each of the other species considered in the analysis. The sole exception was the Philippine pangolin, which was represented by only a single individual. The genetic distances between the mitogenomes of the cryptic new species and the four recognized Asian pangolins were calculated using MEGA11 (86) based on the Kimura 2-parameter model; 1,000 bootstraps were conducted.

Species delimitation analyses based on autosomal SNPs from 60 Asian pangolins and seven cryptic new species individuals were conducted using the BFD* method to verify that the cryptic new species is a distinct Asian pangolin species. The BFD analyses were performed using the SNAPP package v.1.5.2 in BEAST2 v.2.6.6 (77), and the autosomal SNPs were thinned with a 200-kb window size. The species delimitation models assuming the existence of two to five species were tested by path sampling with 24 steps in which each step was set to 50,000 generations with a preburnin of 10,000 generations (48). The best model was obtained by comparing the marginal likelihood estimate (MLE) values for each species delimitation model. All BFs were calculated against the best model (runA). Positive BF values indicate support for the best model, with BF values above 10 indicating decisive support (87). Species delimitation analysis based on the mitogenomes was also performed with the Bayesian coalescent-based method using the Bayesian Phylogenetics and Phylogeography v3.4 (88, 89) software. We applied two models, i.e., the A10 model (88) with the gene tree (Fig. 2B) used as the guide tree and the A11 model (90) without the guide tree. We implemented three schemes of priors for ancestral population size (θ) and root age (τ0), following Gaubert et al. (14) and Leaché et al. (88): 1) Gθs (1, 10) and Gτ0 (1, 10); 2) Gθs (2, 2000) and Gτ0 (2, 2000); and 3) Gθs (1, 10) and Gτ0 (2, 2000). Two rjMCMC algorithms (algorithms 0 and 1) and two independent analyses were run using different starting seeds for each set of parametrization for 100,000 generations (with a sampling frequency of five), with 20,000 samples discarded as burn-in. A posterior probability ≥0.95 indicates species-level delineation (88).

Genomic islands of differentiation between the cryptic new species and the four recognized Asian pangolins were identified by pairwise comparisons using Dxy (91) and Fst (92, 93) analyses based on 20-kb windows along the whole genome. These analyses were performed with popgenWindows.py (https://github.com/simonhmartin/genomics_general/blob/master/popgenWindows.py) (94). Both analyses require population data, so the Philippine pangolin could not be included because only a single individual of this species was represented in the available data. Windows where both Dxy and Fst were >=95th quantile were considered to represent genome islands of differentiation (95). The genes within these windows were identified based on human genome annotations by using BLASTP (E-value) ≤ 1.0 × 10⁻¹⁰ (96).

Evaluation of Genetic Diversity, Inbreeding Level, and Genetic Load.

We assessed genetic diversity by calculating the average heterozygosity (He) among all individuals for each species. The heterozygosity for each individual was calculated as the ratio of the number of heterozygous sites to the total number of callable sites across the genome.

The inbreeding level was calculated according to the average inbreeding coefficient (F_ROH) among all individuals for each species. Long runs of homozygosity (ROH) were identified by PLINK v.2.0 (68) using the following parameter settings: --homozyg -homozyg-window-snp 20 -homozyg-kb 100. The inbreeding coefficient (F_ROH) for each individual was calculated as the total length of ROH divided by the total length of the autosomes covered by SNPs (L_Auto) as described by McQuillan et al. (97). Longer ROHs indicate more recent inbreeding; ROHs longer than 1 Mb indicate inbreeding within the last 50 y.

Deleterious mutations (i.e., genetic load) are predicted to disrupt gene function and are thus expected to substantially reduce the mean fitness of individuals (54). We used SnpEff v.4.3t (98) to evaluate the genetic load level by categorizing the derived allele mutations in the coding regions of each individual into LOF, missense, and synonymous mutations. The genotypes of major homozygous alleles (i.e., alleles homozygous in over 50% of individual Asian and African pangolins) were used to represent the ancestral state, while those of minor homozygous and heterozygous mutations represent the derived state. We built databases from the annotations and reference genome sequences of the Malayan pangolin (24). Input files in VCF format were used to annotate SNPs and assigned mutation categories to the input SNPs for each individual. LOF mutations included premature stop codons (nonsense) and splice site–disrupting single-nucleotide variants. The deleterious load was estimated by the ratio of the number of the derived homozygous sites (two per each site) to both the derived homozygous and heterozygous sites (two per homozygous site and one per heterozygous site) for each category (99, 100). The deleteriousness of missense mutations was also evaluated using the GS (101), a measure of the physical/chemical consequences of amino acid changes. GSs equal to or greater than 150 were designated as deleterious (102).

In addition, the GO terms and KEGG functional enrichment of all genes corresponding to homozygous and putatively deleterious LOF mutations were retrieved using the Metascape website (103). GO terms and KEGG pathways with an enrichment factor >1.5 and P-value <0.01 were considered significantly enriched.

To reduce the likelihood that the evaluation of genetic diversity, inbreeding, and genetic load was affected by sample size bias, we resampled 7 individuals from each pangolin species (except the Philippine pangolin and Temminck’s pangolin, which were represented by one and two individuals, respectively) by 100-fold bootstrapping. We obtained the same patterns as those from the dataset without resampling, indicating that the results are not affected by the sample size bias.

Demographic History Reconstruction.

To trace the demographic histories of pangolin species, we used a haploid PSMC model (104) with the following set of parameters: -N25 -t15 -r5 -p “4+25*2+4+6”. To accurately trace their demographic history, the Asian and African pangolins were mapped to the genomes of the Malayan pangolin (24) and the White-bellied pangolin (69), respectively. The density of heterozygous sites was used to estimate changes in effective population size across the diploid genome of the individual with the highest genome coverage (>20×) from each species. The Philippine pangolin was excluded from this analysis due to its low genome coverage. The distribution of the time to the most recent common ancestor between the two alleles across all scaffolds was estimated by assuming a mutation rate per generation of 1.47 × 10⁻⁸ (43) and an average generation time of 1 y (44, 105). To assess variance in the demographic trajectories, 100 bootstrap replicates were performed.

Gene Flow Analysis.

We used the Dinvestigate program in Dsuite (106) to evaluate the gene flow between Asian pangolins (including the new species) and between African pangolins separately based on the species relationships revealed in the phylogenies. Three approaches were used for this purpose: D statistics (107), the f4-ratio (108), and the f-branch method (109). When evaluating the gene flow between Asian pangolins, an African pangolin species was selected as outgroup, and vice versa. For the D statistics and f4-ratio analyses, we ran “Dsuite Dtrios” by inputting the merged vcf file, the tree file, and the sample sets file. For f-branch analysis, we performed “Dsuite Fbranch” with the phylogenetic tree and used the output of the Dsuite Dtrios analysis to map the gene flow intensity to the phylogenetic tree topology.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The newly generated COI haplotypes (GSA: SAMC1131749-SAMC1131786; GenBank: OQ925816–OQ925853) (110) and raw genome resequencing data have been deposited in the GSA (http://gsa.big.ac.cn/) (111) and NCBI (https://www.ncbi.nlm.nih.gov/) (112) archive under project CRA010100 (113) and PRJNA962256 (114), respectively.