An integrative taxonomic revision of lesser gymnures (Eulipotyphla: Hylomys) reveals five new species and emerging patterns of local endemism in Tropical East Asia

Abstract

We here present a comprehensive integrative taxonomic review of the genus Hylomys, using molecular (mitochondrial genomes and up to five nuclear loci) and morphological data from museum specimens across its distribution, resulting in the description of two new species and the elevation of three subspecies to specific status. This revision significantly increases the known diversity of Hylomys from two to seven extant species, challenging the traditional view of species-level diversity within gymnures. We discuss the implications of the taxonomic findings for conservation, particularly in relation to the restricted distribution ranges of several species that may be threatened by habitat loss and/or climate change. Our research emphasizes the importance of scientific collections and underscores the potential of museum genomics and additional field sampling to identify new species and improve our understanding of species diversity in poorly studied regions. Speciation events within Hylomys occurred during the Late Miocene and Early Pliocene, possibly driven by shifting climate conditions such as the strengthening of the Indian monsoon and the expansion of seasonally dry conditions. This study supports northern Sumatra and the southern Annamites as centres of localized endemicity and suggests the need for additional small mammal surveys across Sumatra’s Barisan Range.

biodiversity, biogeography, Hylomys, Indochina, new species, mitogenomics, museum genomics, Sundaland, systematics

INTRODUCTION

The family Erinaceidae is composed of two subfamilies each with a highly distinct external morphology and life history: the spiny hedgehogs (Erinaceinae), and the soft- to stiff-furred gymnures and moonrats (Galericinae). The members of these subfamilies also differ in their habitat requirements: hedgehogs are distributed throughout the deserts, steppes, savannas, and temperate forests of Africa and Eurasia, but extant gymnures and moonrats are restricted to Tropical East Asia and the South China-Viet Nam subtropical evergreen forests. However, the current distribution range of Galericinae is relictic, since fossil evidence supports a much wider Miocene distribution, including Europe, south-west and central Asia, and East Africa (Zijlstra and Flynn 2015). Galericinae are composed of six extant genera and three major clades: (i) the early-branching Otohylomys, endemic to Lao P.D.R. limestone karsts; (ii) Echinosorex (Blainville, 1838) and Podogymnura (Mearns, 1905), endemic to Sundaland, and Mindanao and Dinagat, respectively; (iii) Neotetracus (Trouessart, 1909) and Neohylomys (Shaw and Wong 1959), restricted to the South China-Viet Nam subtropical evergreen forests, and Hylomys (Müller 1840), a widely distributed genus that spans Sundaland and Indochina (He et al. 2012, Bannikova et al. 2014, Zeng et al. 2022).

The taxonomy of Galericinae is far from stable. There has been much discussion on whether Neotetracus and Neohylomys should be retained as different genera (Honacki et al. 1982, Corbet 1988, Mein and Ginsburg 1997, Wilson and Reeder 2005, Engesser and Burkart 2011, He et al. 2012, Bannikova et al. 2014) or be synonymized with Hylomys (Frost et al. 1991, Gould 2001, Jenkins and Robinson 2002). Similarly, Best (2018) did not recognize the genus Otohylomys (Bannikova et al. 2014), rendering the genus Hylomys (s.l.) paraphyletic (Bannikova et al. 2014, Zeng et al. 2022). Conversely, the genus has been accepted in the Mammal Diversity Database (MDD 2023). At the intrageneric level, the species limits within Podogymnura have been re-evaluated in the light of additional geographic sampling and molecular evidence supporting four species within this genus (Balete et al. 2023). This rich species diversity within Mindanao sharply contrasts with that of its sister monotypic genus Echinosorex, which shows little genetic structure across a much wider distribution range (Zeng et al. 2022). Finally, the monotypic Neotetracus sinensis (Trouessart, 1909) and Hylomys species diversity are probably also underestimated (Ruedi et al. 1994, Ruedi and Fumagalli 1996, Bannikova et al. 2014, Zeng et al. 2022).

Three species of Hylomys are currently recognized (Ruedi et al. 1994, Bannikova et al. 2014, Best (2018), MDD 2023, Zijlstra 2023): †Hylomys engesseri (Mein and Ginsburg 1997), an extinct Early Miocene species from north-western Thailand described from a M3 molar; Hylomys parvus (Robinson and Kloss, 1916), extant and endemic to the higher slopes of Mt. Kerinci (Sumatra); and Hylomys suillus (Müller, 1840), extant and distributed throughout Indochina (H. s. peguensis, H. s. siamensis, and H. s. microtinus), Malay Peninsula and Sumatra (H. s. maxi), Tioman Island (H. s. tionis), Java (H. s. suillus), and Borneo (H. s. dorsalis). However, these two extant species are not reciprocally monophyletic in several phylogenies with poor support in key nodes. Additionally, intraspecific genetic divergence in H. suillus populations is higher than that among H. parvus and some H. suillus subspecies (Ruedi and Fumagalli 1996, Bannikova et al. 2014, Zeng et al. 2022). According to the genetic species concept (Bradley and Baker 2001), the cytochrome b genetic distance found between several Hylomys suillus populations is indicative of species rank rather than subspecific status (Ruedi and Fumagalli 1996, Bannikova et al. 2014, Zeng et al. 2022). Abd Wahab et al. (2022) interpreted this, together with cranial differences, as sufficient evidence and gave specific status to the subspecies H. s. maxi, H. s. suillus, H. s. dorsalis, H. s. siamensis, and H. s. microtinus, and maintained the subspecies ranks of H. s. tionis and H. s. peguensis due to lack of molecular data. However, Erinaceidae mitochondrial DNA is characterized by an extremely rapid mutation rate and biased base composition, which has led to mitonuclear discordance and conflicts in eutherian phylogenetic studies (Waddell et al. 1999, Nikaido et al. 2001, 2003). In this context, Bannikova et al. (2014) highlighted the importance of including nuclear loci evidence in determining species delimitations in Hylomys.

In this study, we perform a comprehensive re-evaluation of Hylomys using molecular (mitochondrial genomes and up to five nuclear loci) and morphological data from museum specimens across its distribution range. By re-assessing the diversity of Hylomys with these different lines of evidence, and an improved geographic sampling, we elevate three subspecies to species status and describe two new species. This increases the number of extant species in the genus Hylomys from two to seven. We also discuss regional patterns of local endemism and mechanisms that may have been important in driving speciation in Tropical East Asia.

MATERIALS AND METHODS

Materials

Our molecular and morphological analyses included H. parvus and H. suillus, and all subspecies of the latter (Fig. 1; Supporting Information, Table S1). Topotypes of H. parvus, H. s. suillus, H. s. dorsalis, and H. s. tionis, a specimen 58 km away from the type locality of H. s. peguensis (designated as neotype in taxonomic section), and holotypes of the two new species were sampled for DNA sequencing. The holotypes/syntypes of H. parvus and all H. suillus subspecies, except H. s. peguensis and H. s. suillus, were examined and measured (Ruedi et al. 1994; Supporting information, Table S1). The syntypes of H. s. peguensis are lost and the syntype of H. s. suillus (RMNH 39017) has the skull inside the mounted skin and could not be measured (see species accounts in taxonomic section). The skulls of the types of H. s. tionis and H. s. siamensis were badly damaged and/or immature, and thus excluded from morphological analyses. Several paratypes of H. parvus and H. maxi were also included in morphological analyses (Supporting Information, Table S2). GenBank sequences from gymnures of the genera Echinosorex, Neotetracus, and Neohylomys were included as outgroups in phylogenetic analyses (Supporting Information, File S1). Genbank cytochrome b sequences of Hylomys were included in the pairwise uncorrected pairwise genetic distances plot, while available nuclear sequences were included in haplotype network plots (Supporting Information, File S1).

Figure 1.

Map of tropical East Asia indicating sample localities in this study (circles) for each of the Hylomys taxa. Samples with associated DNA data are shown as large outlined circles, samples with associated morphological data are shown as smaller circles lacking an outline, and samples with molecular and morphological data are shown as large outlined circles with a black point inside. Inlaid maps highlight the elevational separation of specimens. Darker shades are associated with higher elevation. K-P: Kangar-Pattani Line vegetation transition.

All genetic samples were obtained from historical or modern specimens housed in the following natural history museums: Academy of Natural Sciences of Drexel University (ANSP), Estación Biológica de Doñana-CSIC (EBD), Lee Kong Chian Natural History Museum (ZRC), Museum of Vertebrate Zoology (MVZ), Museum Zoologicum Bogoriense (MZB), Museum of Zoology of the Universiti Malaya (MZUM), Natural History Museum of Geneva (MHNG), Naturalis Biodiversity Center (RMNH), Smithsonian National Museum of Natural History (USNM), and Thailand Natural History Museum (TNHM); and the mammal tissue collection of Institut des Sciences de l’Evolution de Montpellier (T). Among the 85 tissue samples included in this study, 67 were historic and 18 frozen, nine of which represented museum loans, and nine of which were obtained from specimens collected during three field trips to Borneo (Supporting Information, Table S1; Camacho-Sanchez et al. 2019, Hinckley et al. 2022). Animal care and use committees’ approved protocols, sampling and field research permits for these campaigns are specified in Hinckley et al. (2022).

Molecular methods

DNA was extracted using the QIAamp DNA Mini Kit (Qiagen) following the manufacturer’s protocol or phenol-chloroform and purified with an ethanol precipitation. Amicon Ultra 0.5 mL 10kDa filters (UFC501096, Merck Millipore, Darmstadt, Germany) were used instead of ethanol precipitation to purify extracts from historic samples. Museum samples were processed in isolated historical DNA laboratories at the Smithsonian Institution’s Museum Support Center and Estación Biológica de Doñana-CSIC following strict protocols to control for contamination, such as the use of filter pipette tips and inclusion of negative extractions in every batch of DNA extractions. To sequence the mitogenomes, modern sample dual-indexed libraries were generated using a Kapa Biosystems Library Preparation Kit or a modified Illumina protocol that relies on separate ThermoFisher reagent kits for the end-repair, A-tailing, and ligation enzyme reactions, and KAPA HiFi HotStart ReadyMix for the indexing step. Historic sample libraries were prepared with the previous protocols, the single tube protocol described in Carøe et al. (2018), or a modification of the ‘single strand’ SRSLY NGS Library Prep Kit protocol described in Troll et al. (2019) with half-volume reactions to reduce costs. Libraries were quantified with a Qubit fluorometer or qPCR prior to equimolar pooling and sequencing. Dual-indexed libraries were shotgun sequenced on Illumina HiSeq 2000 (using 2 × 100 PE), MiSeq (using 2 × 250 PE) or NovaSeq (using 2 × 50 PE) platforms at the Genetic Resources Core Facility (GRCF) of Johns Hopkins University, Smithsonian National Museum of Natural History (NMNH), or Oklahoma Medical Research Foundation NGS Core.

We targeted five nuclear loci, three exons: ApoB, BDNF, GHR10; and two exons + introns: PCSK2 and POU2F2. Primers for ApoB (583 bp), PCSK2 (434 bp), and POU2F2 (356 bp) were developed for different mammals in Jiang et al. (1998) and were tested here on Hylomys. In addition, two short fragments of BDNF (101 bp) and GHR10 (165 bp) were designed for this study to amplify historic DNA from museum specimens in a single multiplex. Sequence and polymerase chain reaction (PCR) conditions for these primers are detailed in the Supporting Information, File S1, Tables S3 and S4. Three to four PCR replicates were performed and sequenced per historic sample to identify allelic dropout vs. apparent mutations caused by damage or degradation of the DNA (Yuan et al. 2021). Amplicon size exceeded read length in ApoB (583 bp). This generated missing data in the centre of this alignment (18 bp), which was trimmed. USNM and ANSP samples were Sanger sequenced at NMNH’s Laboratory of Analytical Biology, while the remaining samples were sequenced on a MiSeq at GRCF or a partial NovaSeq lane at the Oklahoma Medical Research Foundation NGS Core. Tailed primers were used to amplify a unique combination of forward and reverse barcodes for specimens sequenced on an Illumina MiSeq (2 × 250bp) (Supporting Information, File S1). PCR products were run on a 2% agarose gel viewed with BIO-RAD Image LabTM software (Bio-Rad Laboratories) for relative quantification of PCR products for equimolar pooling.

Preprocessing and quality scanning of sequencing data

Reads were demultiplexed from the sequencing cores, and adaptor removal and quality trimming were performed with CUTADAPT 4.1 (Martin 2011) using paired-end mode for USNM and ANSP samples and Trimmomatic (Bolger et al. 2014) with the sliding window parameter set to 5:20, read minimum length parameter to 50 bp (HiSeq/MiSeq modern sample libraries) or 30 bp (Novaseq modern and all historic sample libraries), and leading and trailing to 5 for the remaining samples. Quality scans were run on the raw fastq files before and after trimming with FastQC (Andrews 2010), and viewed with MultiQC (Ewels et al. 2016) enabling global trends and biases to be quickly identified.

Assembly, alignment, and phylogenetic analysis of mitochondrial genomes

Lineage-specific references were obtained by mapping iteratively the best-quality samples to the H. suillus mitogenome AM905041, using GENEIOUS PRIME 2023.0.4 mapper with medium-low sensitivity and up to five iterations. Quality-trimmed reads were mapped to their lineage-specific reference with BWA-MEM (Li 2013). The output BAM files were sorted, PCR duplicates were removed with SAMtools (Li et al. 2009), and libraries from the same specimens were merged. BAM files were imported to GENEIOUS where consensus sequences were called with minimum 3x coverage and 65% threshold.

Trailing Ns were removed, and sequences with more than 30% of ambiguities and those below 10 000 bp were not considered for the core multiple sequence alignment. Consensus sequences were annotated in GENEIOUS using as a reference the mitogenome AM905041, and they were aligned with the MAFFT v.7.450 GENEIOUS plugin (Katoh and Standley 2013) under default parameters. The control region was removed from the mitogenome assemblies because it was poorly assembled in many historic samples and has been shown to provide low phylogenetic resolution and overestimation of divergence times (Duchêne et al. 2011). Protein-coding regions were translated and inspected for frameshift mutations and for the presence of unexpected stop codons. This core alignment was used to map five historical samples (MVZ186403, USNM583813, THNHM03942, THNHM03952, and USNM320500) using MAFFT flag -add. These samples had lower sequencing depths and had not passed initial quality filters (>30% missing data), but included important samples such as the holotype of Hylomys macarong sp. nov. and neotype of H. peguensis. The multiple sequence alignment was partitioned into the different genes according to the curated annotations. For protein-coding genes, the three codon positions were written to different partitions. GBlocks (Castresana 2000) was run on each partition using a window of size 1 nt for column filtering and columns with more than half gapped positions were removed.

Raw genetic distances were plotted against Tamura and Nei (1993) corrected distances to evaluate the saturation of each partition, with APE (Paradis and Schliep 2019) in R (R Core Team 2013). The third codon positions of all protein-coding genes were saturated, and, thus, they were excluded from downstream phylogenetic inference (Lemey et al. 2009). The final mitogenome alignment had 11 292 positions, of which 2237 were parsimony informative in the 29 ingroup Hylomys individuals, and five outgroups sequences from Neotetracus, Neohylomys, and Echinosorex, with 9.2% of characters undetermined. All ingroup species were represented by several individuals, except the species H. parvus and H. suillus. These were represented by a single sequence given that four and seven historic specimen samples, respectively, failed to generate molecular data due to DNA damage (Supporting Information, Table S1). Maximum likelihood (ML) phylogenetic inference was carried with IQ-TREE 2.2.5 (Minh et al. 2020). The determination of the models of molecular evolution and the best-partition scheme were carried out with IQ-TREE’s ModelFinder implementation (Kalyaanamoorthy et al. 2017). The input 51 partitions were merged using the ‘greedy’ strategy (Lanfear et al. 2012) until BIC stabilized into seven partitions. FreeRate heterogeneity model was considered and each partition was allowed to have its own evolutionary rate (Chernomor et al. 2016). ML tree reconstruction was run following the best-partition model. Ultrafast bootstrap support (UFBoot) was computed with 1000 bootstrap replicates. An UFBoot of 95% or above indicates good support, and it is equivalent to approximately 80% of standard bootstrap support (Minh et al. 2013).

Bayesian tree inference was performed in BEAST 2.4.6 (Bouckaert et al. 2014). The mitogenome alignment was trimmed with AMAS to remove the outgroup (Echinosorex gymnura), and one sequence per putative species was kept, plus two supported and divergent lineages within H. peguensis, and H. maxi from both sides of the Strait of Malacca. The alignment was split into the same partitions as with the ML inference, the third codon position was removed, and the partitions were trimmed with GBlocks using the same parameters as above. The final DNA matrix had 11 473 positions, 11 sequences, and 0.9 % of missing data. The partition scheme was determined with PartitionFinder 2.1.1 (Lanfear et al. 2017), using unlinked branch lengths, the models of evolution restricted to those allowed in BEAST, model selection based on BIC, and the greedy search algorithm. The 51 partitions were merged into two (7817 nt + 3656 nt). BEAUTi was used to prepare the input for BEAST. We used a unique relaxed clock model sampled from a log-normal distribution. For each partition, a different site model was estimated by model averaging in bModelTEST (Bouckaert and Drummond 2017), selecting mutation rate estimation, transition–transversion split, and empirical frequencies priors. A birth–death process was selected to model speciation along the tree. We calibrated the tree by imposing a 16 Myr time constraint at the root of the tree, incorporated as a log-normal prior (offset = 16, SD = 1.25, mean = 0) (Parham et al. 2012). This constraint was informed by the presence of Hylomys and Neotetracus fossil records (Hylomys engesseri and Neotetracus butleri) from the end of the Early Miocene (inferior MN4 biochronological zone: c. 17–18 Mya; Li Mae Long, Lamphun, Thailand; Mein and Ginsburg 1997), and served as the lower boundary for estimating the time of separation between these lineages. We ran two chains of 25 million generations, sampled every 1000 generations. The convergence of the two chains was checked in TRACER 1.7.2 (Rambaut et al. 2018) for each parameter in the combined log file after 10% burn-in. All parameters had estimated sample sizes above 200. Trees from both runs were combined with LogCombiner after removing the first 10% as burn-in, and a maximum clade credibility tree was generated with TreeAnnotator, with median node heights considered as node age.

Genotyping and alignment of nuclear data

For nuclear sequences, we imported the trimmed reads to Geneious 11.0.5 and mapped them to the closest homologous sequences found in GenBank (Supporting information, File S1) with Geneious Mapper, low-medium sensitivity and up to 5 iterations. We called consensus sequences in Geneious with a minimum 5x coverage and 75% consensus threshold. Each locus was aligned independently with MAFFT v7.450 Geneious plugin automatic algorithm (Katoh and Standley 2013) under default parameters. Alleles were manually genotyped based on the original BAM alignments and comparison of each of the sample’s replicates.

Structure of nuclear genetic diversity and species delimitation

Genetic structure of nuclear loci and species limits were inferred based on mutual allelic exclusivity through haplowebs and conspecificity matrices as implemented in HaplowebMaker and CoMa web tools (Spöri and Flot 2020). Given that the haploweb approximation illustrates a single locus at a time, we employed a ‘conspecificity matrix’ approach to combine the structure produced by the combination of different nuclear markers and turn them into a single graphical output to reveal separately evolving metapopulations present in the dataset. A median joining network algorithm was selected, with indels treated as a fifth state (given that long strings of missing nucleotides were lacking) and columns with missing data masked. Historic samples were included in haplowebs but removed from the CoMa input alignments given that these were not represented in most loci alignments (Supporting Information, File S1). Uncorrected pairwise genetic distances were computed with APE (Paradis and Schliep 2019) in R based on all available cytochrome b sequences in our dataset and GenBank with less than 5% of missing data.

We consider species as separately evolving metapopulation lineages (De Queiroz 2007), follow the species delimitation framework of Esselstyn et al. (2021), and recognize as species, groups of specimens that are geographically, genetically, and morphologically cohesive.

Morphological data collection

The specimens examined here and included in craniodental morphometric analyses are housed in the following natural history repositories: ANSP, ZRC, MHNG, Muséum National d’Histoire Naturelle (MNHN), MZB, Natural History Museum of London (NHMUK), RMNH, and USNM (Supporting Information, Table S1). In addition, selected external measurements were retrieved from aged specimens housed at the Museum of Comparative Zoology at Harvard University (MCZ), EBD-CSIC, and MZUM. Two-hundred and ninety-nine specimens were aged based on dental eruption, following Ruedi et al. (1994). Two-hundred and thirty-two of these had fully erupted permanent dentition (adults) and were included in downstream morphometric analyses. We combined data from males and females in morphometric analyses following Ruedi et al. (1994) and Gould (2001). However, if certain characters were sexually dimorphic, results would be interpreted in the context of phenotypic variation among putative species inclusive of potential patterns of sexual dimorphism, thereby providing a more conservative assessment of divergence (Meik et al. 2018). Craniodental measurements were taken with high-precision electronic digital calipers to the nearest 0.01 mm. Selected external measurements (in millimetres) were taken from field catalogues of the authors or from skin labels of voucher specimens. These included total length (TL), length of tail (T), length of ear from notch (E), and weight in grams (WT). Whenever the length of head and body (HB) was not specified on a label, this was determined by subtracting length of tail from total length. We generated two separate variables for length of hindfoot: ‘hindfoot without nail’ (HF) for European and Asian museum specimens, and ‘hindfoot with nail’ (HFN) for American museum specimens. Several ear and hindfoot outliers were removed after remeasuring study skins/comparing these erroneous measurements with specimens of the same series.

We analysed four morphometric datasets: the first dataset (D1) was published by Ruedi et al. (1994), the second and third datasets (D2 and D3) were generated in this study. D2 represents a subset of D3 that just includes measurements present in D1 and was created to test for observer measurement error. D4 was created by combining the data from D1 and D2 to increase sample size and geographic coverage. D1 included the data of 103 specimens from MHNG, MNHN, MZB, NHMUK, RMNH, and 16 craniodental variables collected in Ruedi et al. (1994) and defined by Heaney and Morgan (1982): greatest length of skull (GLS), condylobasal length (CBL), braincase breadth (BB), interorbital breadth (IOB), rostral length (ROL), rostral breadth (ROB), postpalatal length (PPL), postpalatal depth (PPD), palatal width at third upper molars (M3B), M1 to M1 width (M1M1), length of upper toothrow (IM3Sb), and P4 to M3 length (P4M3); or defined by Ruedi et al. (1994): length of mandible (LMA), length of lower toothrow (IM3I), height of coronoid process (HCO), and length of angular process (LAP). D2 included 96 specimens from ANSP, ZRC, and USNM (Supporting Information, Table S1) and the same craniodental variables as D1, collected by A. Hinckley over a period of 3 weeks after being trained by M. Ruedi (e.g. placing the caliper on upper or lower jaws on the exact landmarks and keeping consistency in caliper and skull orientation for each measurement to minimize observer bias). D3 was composed of the same 96 specimens and 16 variables of D2 but also included seven additional craniodental variables (defined in Supporting Information, File S1): nasal length (NL), nasal breadth (NB), braincase depth (BD), mastoid breadth (MAB), length of upper toothrow modified from Ruedi et al. (1994; IM3Sa), first incisive length (I1L), and alisphenoid canal length (ASC). D4 included D1 and D2 (16 craniodental variables), and selected external measurements of D1 + D2, MCZ, EBD-CSIC, and MZUM specimens (199 with craniodental data and 225 with external measurement data; Supporting Information, Table S1). D2, D3, and D4 included all seven named or putative extant species, while D1 lacked the two new undescribed species. Cranial and dental nomenclature follows Frost et al. (1991), Ruedi et al. (1994), and Jenkins and Robinson (2002). Dental notations are specified in the text with uppercase (premaxillary and maxillary teeth) and lowercase letters (mandibular teeth): incisor (I/i), canine (C/c), premolar (P/p), and molar (M/m); for instance, M1 refers to the first upper molar and p2 refers to the second lower premolar.

Morphometric statistical analyses

Principal components analysis (PCA) of craniodental variables were carried out to assess morphometric variation and to visualize the morphometric distinctiveness of named and putative species. We log-transformed each measurement prior to computing PCA so that the data were analysed on the basis of correlations instead of covariances. PCA was implemented in R with the ‘prcomp’ command (R Core Team, 2013), and results were extracted and visualized with factoextra and ggplot2 (Wickham et al. 2016, Kassambara and Mundt 2017). PCAs were initially run independently for datasets D2–4 (D1 was already run in Ruedi et al. 1994; Supporting Information, File S1). ASC was excluded from D3 PCA since this measurement was just collected for a subset of the taxa. Dataset 4 (without selected external measurements) PCA is presented in the main text and as the reference due to its greater sample size and the consistency found with D2 PCA results (both in terms of distribution of samples and variable loadings), which suggested a low effect of observer bias (Supporting Information, File S1). Uni/bivariate plots of craniodental and selected external measurements were computed to include damaged but informative specimens, increase geographic coverage and the number of types in our analyses. Standard summary statistics were generated from univariate measurements for all 23 craniodental variables. Specimens ANSP 20372 and 20374 are in a subadult–adult transition stage (permanent dentition almost fully erupted) and were kept in the PCA due to its importance but not in the descriptive statistics analyses. Several damaged and unsequenced incertae sedis specimens (MZB 3172–5 and RMNH 5137–8; see taxonomic section) were also excluded from the main manuscript descriptive statistics’ analyses. Specimens from Kon Tum province, Central Highlands region of Viet Nam (MNHN1929320–5) were not sequenced but were assigned to peguensis following preliminary PCA results and kept in downstream descriptive statistics’ analyses.

Ecological data

Descriptions of habitats where specimens were collected, syntopic species that shared these habitats, stomach content, and reproductive data were taken from specimen tags and field notes written by the authors and different historic collectors; these notes are available in the archives of EBD-CSIC, ANSP, and the Division of Mammals at USNM. This approach was complemented with additional data from the literature (e.g. Robinson and Kloss 1918, de Schauensee and Ripley 1939, Miller 1942, Harrison and Traub 1950, Rudd 1965, Medway 1969, Van Sung 1976, Ruedi et al. 1994, Camacho-Sanchez et al. 2019).

RESULTS

Phylogenetic analyses

The mitogenome maximum likelihood phylogenetic tree had two highly supported clades (99% UFBoot support): a core Sunda clade (H. suillus, H. vorax sp. nov., H. maxi, H. dorsalis, and H. parvus) and an Indochinese clade (H. macarongsp. nov. and H. peguensis) (Fig. 2A). All named or putative species stood on long branches. Five of the lineages, represented by more than one sample, were monophyletic clades with 100% UFBoot support. The basal relations of the Sunda taxa were not fully resolved, and they were supported in all cases by short internal branches. There was moderate differentiation between specimens of H. maxi from both sides of the Strait of Malacca and little differentiation between the mainland and Tioman Island (formerly known as H. s. tionis). Two main clades were strongly supported (98% UFBoot support) within Indochina, corresponding to southern Viet Nam (H. macarong sp. nov.), and the remaining Indochinese populations (H. peguensis). Substantial structure was detected within H. peguensis, with nominotypical populations from SC Myanmar and NW Thailand being sister to populations from CE Thailand, Cambodia, Lao P.D.R., and CN Viet Nam (H. p. siamensis and H. p. microtinus). The holotype of H. macarong sp. nov. and the neotype of H. peguensis were clustered with high support within the diversity of H. macarong sp. nov. and H. peguensis, respectively, despite the higher percentage of missing data of these samples (Supporting Information, File S1, Fig. S1). Little differentiation was shown in Neohylomys hainanensis (Shaw and Wong 1959) between the mainland (Vinh Phuc, Viet Nam; MVZ186403) and Hainan Island (NC063830), suggesting these currently isolated populations are possibly conspecific (c. 0.01 substitutions/site; Supporting Information, File S1, Fig. S1).

Figure 2.

A, mitochondrial DNA (11 292 nt) maximum-likelihood consensus phylogeny of Hylomys, reconstructed with IQTREE 2. UFBoot support (%) is indicated for the main nodes. The geographic origin of the ingroup species is indicated by colour on the adjacent vertical bar. Colours match the map (B) and other figures. B, map showing the distribution of Hylomys species. Areas where the taxonomic status of populations is unknown are coloured black. Distribution ranges in this map are approximate (e. g. Borneo is homogeneously colored but this does not mean that H. dorsalis is distributed throughout the entire island). See Figure 1 for a more precise map. C, mitochondrial DNA (11 473 nt) dated Bayesian maximum clade credibility tree inferred with BEAST 2. Node age (i.e. median node height) is indicated next to each node with the posterior probability (PP) in parentheses whenever PP < 1. Displayed bars represent the central 95% posterior density of node age. The density of the posterior distribution of the tree root age has been added.

The Bayesian and the ML phylogenies shared a congruent topology (Fig. 2). However, the Bayesian tree provided a higher support for the basal position of the Javanese H. suillus and H. parvus with respect to the other Sunda species (PP = 0.98, vs. UFBoot support 78%). All other relationships were highly supported (PP = 1), except for the clade H. maxi–H. vorax (PP = 0.92). The time to the most recent common ancestor (TMRCA) of Hylomys was estimated at 7.8 MYA (95% high posterior density, HPD: 6.0–10.0).

Structure of nuclear genetic diversity and molecular species delimitation

The conspecificity matrix supported five to seven separately evolving metapopulation lineages, which showed mutual allelic exclusivity for most loci (Fig. 3). These corresponded with H. maxi, H. dorsalis, H. peguensis (nominotypical populations), H. peguensis (previously considered H. s. siamensis, and H. s. microtinus), and H. suillus + H. parvus, and with the two latter possibly subdivided in two (former H. s. siamensis, former H. s. microtinus, H. suillus, and H. parvus; Fig. 3A). Nevertheless, this dataset was generated based on modern (frozen) samples and did not include the putative species H. macarong sp. nov. and H. vorax sp. nov.. The species H. maxi, H. dorsalis, H. peguensis, H. suillus, and H. parvus did not share alleles at the two longer (356–565 bp) and more informative nuclear loci (ApoB and POU2F2; Supporting Information, Fig. S2).

Figure 3.

A, conspecificity matrix of Hylomys inferred with CoMa based on the five nuclear loci shown in the Supporting Information, Figure S2 obtained from modern (frozen) samples. The conspecificity score of a given pair of individuals is computed as the number of partitions for which there is mutual allelic exclusivity. Samples are coloured according to their taxonomy and match the other figures. Six or seven squares of high conspecificity scores (5–3) are shown along the diagonal. B, haplotype web of a short fragment of GHR10 (165 bp) including sequences from both modern and historical samples and all named or putative species. The size of the pie charts corresponds to the frequency of the haplotype. Haplotypes found co-occuring in heterozygous individuals are connected with Bezier curves.

The short (165 bp) and less informative fragment of GHR10 was the only nuclear locus with data for all seven named or putative extant species of Hylomys, since historical samples of H. macarong sp. nov. and H. vorax sp. nov. could not be amplified with BDNF (Sanger sequencing) primers. Indochinese and closer relatives H. macarong sp. nov. and H. peguensis show mutual allelic exclusivity in GHR10, but the former has the same haplotype as Sundaic H. suillus and H. parvus (Fig. 3B). Sumatran highland H. vorax sp. nov. and H. parvus also show mutual allelic exclusivity, but the former shares its haplotype with Sumatran lowland/Malay Peninsula H. maxi (Fig. 3B).

Cytochrome b pairwise uncorrected genetic distances among the seven major mitochondrial lineages of Hylomys (Fig. 2) and five separately evolving metapopulation lineages according to nuclear evidence (Fig. 3A) were above 11%, a value indicative of specific recognition in other mammals under the genetic species concept framework (Fig. 4; Bradley and Baker 2001). High or moderate per cent sequence divergence levels were also observed within H. peguensis (up to c. 9%) and H. maxi (c. 6% between Malay peninsula and Sumatran populations), suggesting that some of these populations could deserve species-level recognition if confirmed by additional geographical sampling and nuclear and morphological evidence (Fig. 4).

Figure 4.

Cytochrome b pairwise uncorrected genetic distances among and within species of Hylomys, computed with the ‘ape’ package in R. Black vertical lines at 0.02 and 0.11, illustrate the ranges of Bradley and Baker (2001), where ‘genetic distance values <2% were indicative of intraspecific variation; values between 2 and 11% had a high probability of being indicative of conspecific populations or valid species and merit additional study concerning specific status; and values >11% were indicative of specific recognition’.

Morphometrics

Morphometric results (Fig. 5) support the distinctiveness of the seven major mitochondrial clades (Figs 2, 4) and five putative species delimited by CoMa (Fig. 3). There is minor overlap between allopatric H. suillus with H. maxi, H. vorax sp. nov., and H. macarong sp. nov. (Fig. 5A), but the Sumatran species H. maxi, H. parvus, and H. vorax sp. nov. occupy distinct regions of PCA morphospace, particularly the latter with parapatric and closest relative H. maxi (Fig. 5A, B). Similarly, there is little overlap between ‘mainland’ species H. maxi, H. peguensis, and H. macarong sp. nov.. Only ANSP 20372, a H. maxi Sumatran specimen in subadult–adult stage overlaps with H. macarong sp. nov. (Fig. 5A), while the latter and H. peguensis show minor overlap (Fig. 5B) or no overlap in PCAs with fewer samples (Supporting Information, Fig. S3) and additional diagnostic variables, such as first incisor length (Supporting Information, Fig. S4). The first principal component accounted for a large part (70.3%) of the variance, was correlated with size, and discriminated the larger sized H. maxi and H. dorsalis from the smaller sized H. macarong sp. nov., H. vorax sp. nov., and H. parvus. PC2 explained 10.6% of the variance and was mainly correlated with shape robustness/broadness, and rostrum length to a lesser degree. PC2 discriminated H. vorax sp. nov. and H. dorsalis, which have a relatively broad interorbital constriction and narrow rostrum, from H. maxi, H. peguensis, and H. macarong sp. nov., which have a relatively narrow interorbital constriction and broad rostrum. In fact, species present in Sumatra (H. maxi, H. vorax sp. nov., and H. parvus), in the mainland (H. maxi vs. H. peguensis and H. macarong sp. nov.), and others across Sundaland (H. dorsalis vs. H. maxi/H. suillus/H. parvus) can be easily distinguished based on the combination of rostrum length and rostrum breadth (Fig. 6).

Figure 5.

Morphometric variation in Hylomys. A, first and second principal components from an analysis of 16 cranial dimensions, computed with Dataset 4 (with the exclusion of external measurements); B, second and third principal components of the same analysis. Asterisks denote undamaged holotype or neotype specimens that could be included in this analysis.

Figure 6.

Bivariate plot of rostrum length (ROL) and rostrum breadth (ROB) measurements.

Taxonomy

The generic diagnosis of Hylomys by Müller (1840) was based on H. suillus and was brief. Diagnostic features for the genus were subsequently noted by Anderson (1874), Corbet (1988), Frost et al. (1991), Jenkins and Robinson (2002), and He et al. (2012), but these studies did not include specimens of the species H. parvus or the new species here described, the whole distribution extent of H. suillus s.l. (e.g. Sumatra and Myanmar), and/or included the monotypic Neotetracus sinensis, Neohylomys hainanensis, and/or Otohylomys megalotis (Jenkins and Robinson 2002) within the diversity of Hylomys. Based on the comparative cranial and dental characters provided by Jenkins and Robinson (2002), the new molecular evidence of Bannikova et al. (2014), and additional morphological data of this study, we here provide an emended diagnosis to more fully define and integrate the morphological diversity of the different species in this genus.

Hylomys Müller, 1840

Type species:

Hylomys suillus Müller, 1840: 436.

Included species:

The type species, plus H. parvus, H. engesseri, H. macarong sp. nov., H. vorax sp. nov., and H. dorsalis, H. maxi, and H. peguensis (recognized at species rank, below).

Distribution:

Currently known from Sundaland (including Borneo, Java, Sumatra, and the Malay peninsula) and Indochina (including Myanmar, Thailand, Cambodia, Lao P.D.R., Viet Nam, and southern China).

Emended diagnosis:

The genus Hylomys is defined phylogenetically as the most recent common ancestor of extant H. parvus, H. dorsalis, H. maxi, H. peguensis, H. macarong sp. nov., and H. vorax sp. nov., and extinct H. engesseri, and by the following combination of morphological characters. Small-sized gymnures, HB from c. 98–157 mm, W c. 40–80 g, GLS from c. 29–39 mm (Table 1). Hylomys is characterized by brown fur that can range from harsh to moderately soft. The dorsum fur is characterized by long and thick spinous black guard hairs and golden-brown guard hairs that produce a golden-streaked appearance. Black guard hairs are stiff and black tipped, gradually turning to a light silvered grey towards its proximal end. Golden-brown guard hairs are stiff and have a short black tip followed by a narrow yellow to golden band which can be more or less expanded and that sharply transitions to a proximal dark silvered grey. Hylomys parvus is an exception to this general appearance as its soft dorsal fur lacks spinous/stiff guard hairs (Ruedi et al. 1994). Dorsum coloration generally acquires a more ochraceous coloration towards the rump and face, and a yellow hue in the shoulder area. The dorsal coloration has limited variation, with only the Bornean population adults exhibiting a moderately distinct to faint black sagittal stripe (but less marked than that of Neohylomys hainanensis), generally restricted to the nape/shoulder area, but which can extend to the rump. Variation on dorsum fur length, guard hair thickness and coloration is slight. Higher elevation individuals seem darker and to have a softer and more dense fur, while lower elevation individuals can be harsher and more colourful (e.g. become very reddish in H. maxi at 100 m a.s.l.; Ruedi et al. 1994). Golden-brown guard hairs are paler, longer, and thicker in Indochinese populations, conferring the dorsum fur a lighter and more yellow appearance than that of Sundaic populations. Ventral coloration is grey/brown, paler than the dorsum, buff or white tipped with a grey base, and can be subject to seasonal, elevational, sexual, and/or ontogenetic variation within populations. In fact, this variation within populations seems generally greater than that between most species. Only H. peguensis is distinct from the other recognized or putative species, regardless of sex or season, as it exhibits buff coloration throughout most of its range. Adults can have a pale brown or ochraceous throat coloration, generally during spring or summer, possibly part of the breeding season. This darker coloration can extend across all the venter in males of Bornean and Da Lat populations, making these seasonally distinct to the other species. As with dorsal pelage, fore- and hindfeet are lighter in Indochinese than Sundaic populations. The legs are short and the feet plantigrade. Tails are bicoloured in all species but H. vorax sp. nov. and one H. maxi specimen from north Sumatra. The hindfoot sole and tarsal regions have short and applied hairs and have generally pale hair tufts over nails. Tail is very short (7–32 mm), 7–28% of head and body length and essentially naked with some short and applied hairs. Females have four axial and two inguinal mammae, the two upper axial are frequently hidden at the armpit area, perhaps suggesting why Anderson (1874) described just two axial and two inguinal mammae; adult males have a slight swelling in the uro-genital area and a penis with many spines; anal glands on anterior margin of anus.

Table 1.

Craniodental and selected external measurements (mm), weight (g) of adult Hylomys, including new species from Mount Leuser and the Da Lat–Dak Lak Plateaus. Mean, SD (±), ranges, and sample size (N) are specified for each measurement.

	Hylomys maxi	Hylomys vorax sp. nov.	Hylomys parvus	Hylomys dorsalis	Hylomys suillus	Hylomys macarong sp. nov.	Hylomys peguensis
GLS	36.9 (± 0.8)	34.5 (± 1.2)	30.9 (± 1.0)	36.6 (± 0.9)	34.8 (± 1.1)	34.8 (± 1.0)	34.4 (± 0.9)
	35.5–38.4	33.4–36.3	29.2–33.4	34.9–38.4	32.2–37.9	33.6–35.9	32.3–36.0
	N = 28	N = 5	N = 15	N = 25	N = 57	N = 7	N = 17
CBL	36.2 (± 0.9)	33.3 (± 1.3)	29.8 (± 1.0)	35.5 (± 0.8)	33.8 (± 1.3)	33.8 (± 0.9)	33.4 (± 1.1)
	34.0–38.2	31.9–35.1	28.1–32.1	33.9–37.5	30.4–37.0	32.8–34.9	30.4–35.4
	N = 35	N = 5	N = 15	N = 30	N = 62	N = 6	N = 21
IOB	8.7 (± 0.4)	8.8 (± 0.6)	8.3 (± 0.3)	9.5 (± 0.2)	8.9 (± 0.3)	8.3 (± 0.3)	8.2 (± 0.3)
	8.1–9.6	7.8–9.4	7.7–9.1	9.2–10.0	8.1–9.6	7.9–8.8	7.5–9.0
	N = 32	N = 7	N = 14	N = 29	N = 43	N = 9	N = 19
BB	15.0 (± 0.3)	14.8 (± 0.2)	13.6 (± 0.3)	15.5 (± 0.4)	14.6 (± 0.4)	14.4 (± 0.3)	14.3 (± 0.4)
	14.1–15.6	14.5–15.1	13.1–14.0	14.9–16.3	13.8–15.6	14.1–15.0	13.7–14.8
	N = 30	N = 5	N = 14	N = 26	N = 58	N = 7	N = 18
ROL	16.9 (± 0.5)	16.0 (± 0.7)	14.2 (± 0.4)	17.3 (± 0.5)	16.1 (± 0.8)	15.1 (± 0.5)	14.6 (± 0.7)
	16.1–18.2	15.2–17.3	13.6–15.1	16.1–18.1	14.5–18.1	14.6–15.7	13.1–15.8
	N = 37	N = 7	N = 15	N = 33	N = 47	N = 10	N = 25
ROB	5.9 (± 0.3)	4.7 (± 0.2)	4.7 (± 0.2)	5.0 (± 0.2)	5.3 (± 0.3)	5.2 (± 0.2)	5.5 (± 0.3)
	5.2–6.4	4.5–5.0	4.4–5.0	4.6–5.5	4.7–6.1	4.9–5.6	5.0–6.0
	N = 34	N = 7	N = 14	N = 34	N = 46	N = 9	N = 24
PPL	13.2 (± 0.5)	11.8 (± 0.5)	10.7 (± 0.3)	13.0 (± 0.4)	12.4 (± 0.8)	12.5 (± 0.5)	12.6 (± 0.6)
	12.3–14.2	11.1–12.3	10.1–11.2	12.3–13.8	10.8–14.3	11.9–13.1	11.5–14.0
	N = 35	N = 5	N = 13	N = 29	N = 47	N = 5	N = 20
PPD	9.2 (± 0.3)	8.7 (± 0.3)	8.2 (± 0.4)	9.0 (± 0.2)	8.6 (± 0.3)	9.1 (± 0.3)	8.9 (± 0.3)
	8.6–9.7	8.3–9.1	7.4–8.6	8.4–9.4	8.0–9.1	8.7–9.7	8.4–9.4
	N = 29	N = 6	N = 13	N = 28	N = 42	N = 9	N = 19
M3B	6.9 (± 0.3)	6.2 (± 0.1)	5.3 (± 0.4)	6.5 (± 0.4)	6.3 (± 0.3)	6.5 (± 0.3)	6.3 (± 0.3)
	6.2–7.5	6.1–6.4	4.7–6.0	5.5–7.4	5.7–6.9	6.1–6.8	5.5–6.8
	N = 29	N = 7	N = 14	N = 33	N = 48	N = 7	N = 24
M1M1	11.7 (± 0.3)	10.2 (± 0.3)	9.4 (± 0.4)	10.6 (± 0.5)	10.5 (± 0.4)	10.9 (± 0.4)	10.9 (± 0.5)
	10.9–12.5	9.8–10.8	9.0–10.2	8.2–11.3	9.8–11.4	10.2–11.5	9.8–11.6
	N = 28	N = 7	N = 15	N = 29	N = 57	N = 10	N = 19
IM3SB	18.5 (± 0.6)	17.3 (± 0.5)	15.7 (± 0.6)	18.1 (± 0.5)	17.5 (± 0.8)	16.6 (± 0.5)	16.9 (± 0.7)
	17.6–19.6	16.7–18.1	14.6–16.9	17.5–19.0	15.3–18.8	16.1–17.4	15.1–18.2
	N = 31	N = 7	N = 17	N = 32	N = 62	N = 10	N = 25
P4M3	8.8 (± 0.3)	7.7 (± 0.2)	7.2 (± 0.2)	8.1 (± 0.2)	7.9 (± 0.3)	8.0 (± 0.2)	8.0 (± 0.4)
	8.3–9.9	7.5–8.2	6.9–7.5	7.7–8.6	7.3–8.6	7.6–8.3	7.1–8.6
	N = 25	N = 7	N = 14	N = 29	N = 43	N = 10	N = 20
LMA	26.3 (± 0.7)	24.0 (± 0.9)	21.3 (± 0.7)	25.3 (± 0.7)	24.5 (± 1.0)	24.9 (± 0.8)	24.2 (± 0.8)
	24.8–27.9	23.0–25.6	20.1–22.2	23.6–26.6	22.4–26.8	23.9–25.9	22.4–25.4
	N = 32	N = 6	N = 13	N = 27	N = 42	N = 8	N = 19
IM3I	18.2 (± 0.4)	17.1 (± 0.4)	15.3 (± 0.5)	17.7 (± 0.4)	17.0 (± 0.5)	16.4 (± 0.4)	16.5 (± 0.6)
	17.6–18.9	16.6–17.7	14.5–16.0	16.7–18.4	15.7–18.2	15.8–17.1	15.1–17.4
	N = 26	N = 7	N = 16	N = 26	N = 57	N = 8	N = 20
HCO	9.4 (± 0.4)	7.8 (± 0.4)	7.5 (± 0.3)	8.9 (± 0.5)	8.2 (± 0.6)	8.8 (± 0.2)	8.9 (± 0.5)
	8.6–10.7	7.4–8.6	6.9–7.9	8.2–9.9	6.6–9.5	8.5–9.1	8.0–9.8
	N = 32	N = 7	N = 13	N = 28	N = 42	N = 7	N = 17
LAP	9.7 (± 0.6)	8.4 (± 0.5)	7.6 (± 0.5)	9.2 (± 0.6)	8.8 (± 0.6)	10.1 (± 0.4)	9.1 (± 0.5)
	8.8–10.8	7.9–9.1	6.8–8.1	7.8–10.3	7.7–10.3	9.4–10.6	8.4–10.6
	N = 27	N = 6	N = 13	N = 28	N = 41	N = 8	N = 18
NL	12.0 (± 0.9)	12.1 (± 0.9)	9.5	14.0 (± 0.7)	11.2 (± 0.9)	13.0 (± 0.6)	11.8 (± 1.0)
	10.6–14.3	11.0–13.4	9.5	12.6–15.0	9.4–12.9	12.1–13.9	10.3–13.6
	N = 23	N = 7	N = 1	N = 22	N = 20	N = 10	N = 12
NB	2.6 (± 0.2)	2.2 (± 0.2)	2.0	2.5 (± 0.1)	2.4 (± 0.2)	2.4 (± 0.3)	2.2 (± 0.2)
	2.2–3.0	2.0–2.4	2.0	2.3–2.8	2.1–2.9	2.0–3.0	1.9–2.5
	N = 21	N = 7	N = 1	N = 22	N = 20	N = 10	N = 13
BD	10.0 (± 0.3)	9.1 (± 0.2)	8.6	9.8 (± 0.2)	9.2 (± 0.3)	9.7 (± 0.3)	9.3 (± 0.2)
	9.3–10.5	8.9–9.5	8.6	9.6–10.1	8.8–9.8	9.4–10.2	9.0–9.6
	N = 23	N = 5	N = 1	N = 20	N = 20	N = 8	N = 12
MAB	14.5 (± 0.3)	13.6 (± 0.6)	12.2	14.6 (± 0.3)	13.7 (± 0.3)	14.2 (± 0.4)	13.9 (± 0.4)
	13.9–15.0	13.0–14.1	12.2	13.8–15.1	13.2–14.5	13.5–14.9	13.2–14.7
	N = 20	N = 4	N = 1	N = 20	N = 20	N = 7	N = 11
IM3SA	18.9 (± 0.5)	17.8 (± 0.4)	16.1	18.5 (± 0.4)	17.2 (± 0.4)	17.4 (± 0.5)	17.3 (± 0.7)
	18.3–19.7	17.2–18.5	16.1	18.0–19.2	16.0–17.9	16.8–18.2	15.8–18.3
	N = 11	N = 7	N = 1	N = 22	N = 19	N = 10	N = 12
I1L	2.7 (± 0.2)	2.3 (± 0.1)	2.5	2.6 (± 0.2)	2.5 (± 0.3)	3.3 (± 0.7)	2.6 (± 0.3)
	2.2–3.3	2.2–2.6	2.5	2.1–3.1	2.0–3.0	2.6–4.5	2.1–3.1
	N = 18	N = 6	N = 1	N = 19	N = 18	N = 10	N = 13
ASC	0.2 (± 0.1)	0.4 (± 0.2)	-	-	-	0.5 (± 0.3)	1.0 (± 0.2)
	0.1–0.4	0.2–0.5	-	-	-	0.2–1.0	0.7–1.2
	N = 4	N = 4	N = 0	N = 0	N = 0	N = 8	N = 8
TL	146.4 (± 6.4)	144.6 (± 3.6)	-	154.7 (± 11.5)	142.0 (± 9.2)	161.3 (± 12.4)	150.7 (± 12.6)
	140.0–160.0	139.0–150.0	-	139.0–175.0	124.0–152.0	135.0–171.0	140.0–172.0
	N = 8	N = 7	N = 0	N = 20	N = 10	N = 7	N = 7
HB	139.3 (± 7.8)	123.1 (± 3.4)	107.2 (± 4.4)	135.2 (± 8.1)	129.5 (± 9.8)	138.4 (± 5.4)	129.4 (± 12.4)
	123.0–156.0	117.0–128.0	100.0–115.0	119.0–150.0	98.0–152.0	130.0–146.0	103.0–157.0
	N = 54	N = 7	N = 14	N = 54	N = 52	N = 8	N = 29
T	16.2 (± 3.0)	21.4 (± 0.6)	23.1 (± 5.4)	19.0 (± 3.5)	18.8 (± 4.3)	23.2 (± 2.4)	21.3 (± 2.5)
	7.0–24.0	20.5–22.0	15.0–32.0	11.0–26.0	12.0–28.0	20.0–25.0	17.0–25.0
	N = 46	N = 7	N = 14	N = 49	N = 46	N = 8	N = 18
HF	25.2 (± 1.2)	(±)	23.1 (± 0.8)	25.7 (± 0.9)	24.4 (± 1.7)	23.5 (± 0.7)	22.5 (± 1.4)
	23.0–28.0	-	21.9–25.0	24.0–28.0	20.0–28.0	23.0–24.0	20.0–25.0
	N = 31	N = 0	N = 16	N = 13	N = 42	N = 2	N = 14
HFWN	26.3 (± 1.1)	26.3 (± 0.8)	-	27.7 (± 1.8)	24.8 (± 1.2)	25.7 (± 1.5)	24.8 (± 1.2)
	24.0–29.0	25.5–27.5	-	24.0–32.0	23.0–26.0	23.0–27.0	22.0–27.0
	N = 32	N = 7	N = 0	N = 37	N = 9	N = 7	N = 14
E	16.4 (± 1.3)	19.1 (± 0.7)	16.5 (± 1.3)	17.4 (± 1.7)	16.6 (± 1.5)	18.1 (± 1.4)	16.4 (± 2.5)
	13.9–21.0	18.0–20.0	14.0–19.0	12.0–21.0	13.0–21.0	16.0–20.0	11.0–20.0
	N = 53	N = 7	N = 16	N = 53	N = 52	N = 10	N = 26
W	64.2 (± 9.2)	44.4 (± 5.0)	-	60.9 (± 8.4)	49.7 (± 7.4)	-	54.2 (± 12.7)
	43.0–80.0	39.6–53.8	-	46.0–75.0	40.0–64.0	-	45.0–73.0
	N = 29	N = 6	N = 0	N = 14	N = 8	N = 0	N = 4

The skull is relatively small with an elongate rostrum; posterior end of nasals spans the level of the antorbital rim in some species; antorbital fossae are moderately deep; maxilla and parietal are separated by a narrow frontal bone strip; supraorbital processes absent to well developed; frontals not inflated to strongly inflated; interorbital region minimally constricted; anterior parietal process is well developed or at least distinct; braincase inflated; sagittal crest inconspicuous to prominent, generally restricted to interparietal but can extend anteriorly to frontal in H. macarong sp. nov.; nuchal crest of varying height that originates laterally as low projections from dorsal margin of each mastoid that meet mid-dorsally to form a broadly tapered arch with which the sagittal crest converges [as described for Podogymnura in Balete et al. (2023)]; occipital crest absent to prominent, generally does not extend ventrally more than a third of the occipital bone; zygomatic arch complete and moderately developed, posteroventral process on zygoma absent or indistinct, deep nasolabialis fossa, small jugal bone; epipterygoid process moderately developed; basioccipital–petrosal suture closed forming distinct foramen, bullae incomplete, condylar foramen in condyle emargination; posterodorsal region of the premaxilla is rarely in contact with the anterodorsal region of the frontal (0–4.15 mm). Anterior palatine foramina positioned slightly anteriorly/posteriorly or at level of maxilla/palatine suture. Anterior opening of infraorbital canal dorsal to P4. Paraoccipital process almost indistinct to somewhat prominent. Dental formula: 3/3 1/1 4/4 3/3 = 44; I1 large and caniniform, C1 larger than adjacent teeth, P1 present, P3 slightly larger or equal to P2, P2 larger or equal to P1. P3 has one root or two roots that are fused in most species; molariform and tribosphenic P4, quadrate M1 and M2; M3 c. half the crown area of M1 and M2; lower incisors spatulate and procumbent, relatively small i1, i1 longer than i2 and i3, i2 longer than i3; c1 procumbent and larger than adjacent teeth; p1 present, crown generally extended anteriorly, p3 can have one or two roots and is larger or equal in size to p2. Mandible relatively elongated, with coronoid process long and narrow to moderately wide, condyloid process elongated, angular process narrow and average to long.

As Ruedi et al. (1994) pointed out, the morphological distinction among the taxa of H. suillus was historically almost exclusively based on variations in colour, despite the apparently high craniodental geographically structured differentiation exhibited by this genus. In the following accounts, we unite these two lines of morphological evidence and diagnose the two species named by Müller (1840) and Robinson and Kloss (1916), give species-level recognition to three subspecies, provide emended diagnoses, and describe two new species. To facilitate taxonomic comparisons, we group sympatric or parapatric taxa based on their distribution (mainland Asia and Sumatra), and allopatric taxa, based on phylogenetic relatedness and geographical proximity (Sundaland clade).

Hylomys maxi Sody, 1933 stat. nov.

Hylomys suillus maxi Sody, 1933: 438. Original description.

Hylomys suillus tionis Chasen, 1940: 12.

Hylomys suillus suillus Corbet, 1988: 122 (part).

Holotype:

RMNH 23895, an adult male collected by Sody on 4 January 1930.

Type locality:

‘Giesting, Lampoengs, S. Sumatra’ (=Giesting, Mt. Tanggamus, Lampung, Sumatra, Indonesia; c. –5.44, 104.71).

Paratypes (5):

Five according to Sody (1933), although Sody did not examine them: ‘For examining the Atchinese specimens I have to thank Mr. F.J. Naing-Golan’. Naturalis Biodiversity Centre holds three of these (RMNH 23896–98), the other two might be lost.

Emended diagnosis:

A large-sized Hylomys (average HB = 139 mm, W = 64 g, GLS = 36.9 mm) characterized by a brown suffusion in throat (from late April to mid-November but absent from late February to late April), bicoloured (except specimen RMNH 23896) and short tail (average T = 16 mm, T/HB = 11.7%), long hindfoot (average HF = 25 mm, HF + nail = 26 mm), uniformly brown-coloured and relatively short ears (average E = 16 mm, E/GLS = 45%). Its skull can be distinguished from all relatives by its relatively larger molars (Supporting Information, File S1, Fig. S5) and from all relatives but allopatric H. suillus by its long and relatively broad rostrum (Fig. 6). Its skull can also be distinguished from all other relatives by the combination of the following characters: obtuse-angled notch between anterior side of premaxilla; posterior end of nasals does not extend to the level of the antorbital rim; dorsal region of maxillary and frontal have a rugged appearance with capillary grooves and pores; prominent antorbital ridge/flange; anterior side of the antorbital rim is posterior to second cusp of M1; robust supraorbital processes and dentition; short alisphenoid canal (average ASC = 0.2 mm), prominent posterior cuspule in P2 in Sumatran specimens with unworn dentition; P3 is much larger than P1; P4 has a thick labial cingulum; posterior nasal spine present; prominent occipital crest that extends more ventrally than in most other species; single root/two well fused roots in p3; anterior margin of p1 crown extends to root of c1.

Comparisons:

Distinguished from parapatric H. parvus by its larger size (HB: 123–156 mm vs. 100–115 mm, GLS: 35.5–38.4 mm vs. 29.2–33.4 mm), relatively shorter tail (average: 16 vs. 23 mm) and harsher dorsal fur. Its skull has a more angular and robust appearance than H. parvus, its supraorbital process is well developed vs. absent in H. parvus, and its dentition is robust (particularly the canines and adjacent teeth) vs. gracile in H. parvus (Ruedi et al. 1994). The notch between premaxilla tips is obtuse vs. acute-angled in H. parvus. The frontal bone and dorsal region of maxillary is more rugged, with greater capillary grooves and pores, than in H. parvus. It has a prominent antorbital ridge and occipital crest, and thick P4 labial cingulum, while these are absent or vestigial/inconspicuous in H. parvus. It has a single root/two well fused roots in P2 and p3 vs. two non-fused roots in H. parvus (Jenkins and Robinson 2002).

Comparisons with additional species have been included in the following accounts, after such species have been formally defined.

Distribution, habitat, and natural history:

Currently known from forests between c. 100 and 2000 m a.s.l. on Sumatra and 600 and 1700 m a.s.l. on the Malay peninsula, possibly extending north up to the Kangar Pattani Line (K-P) vegetation transition, which cuts across the Thailand–Malaysia border (see comments on the following species account). Hylomys maxi seems more abundant in hill and montane forests and was not recorded in dipterocarp forest in surveys on Northern Sumatra and Peninsular Malaysia (Langham 1983, Boubli et al. 2004). This species has been collected in ‘jungle filled hollow surrounded by grasslands’ (Fred Ulmer field notes, ANSP archives) and in secondary vegetation on the edge of mountain forests, in highly anthropized areas (gardens, golf courses, etc.; authors’ unpublished data). Robinson and Kloss (1918) stated that the species was ‘quite numerous at dusk, running about in a patch of sweet potato (Ipomoea batatas) planted in a flat place among rocks at the edge of a torrent’. In Peninsular Malaysia it shares habitat with Crocidura fuliginosa, Crocidura cf. neglecta, Crocidura malayana, Suncus malayanus, Euroscaptor malayanus, Tupaia glis, Berylmys bowersi, Leopoldamys sabanus, Leopoldamys ciliatus, Maxomys surifer, Maxomys rajah, Maxomys whiteheadi, Maxomys inas, Niviventer cremoriventer, Niviventer cameroni, Sundamys muelleri, Sundamys annandalei, Rattus exulans, Callosciurus caniceps, Dremomys rufigenis, Lariscus insignis, and Sundasciurus tahan (Robinson 1911, Harrison and Traub 1950, Rudd 1965, Langham 1983, Ruedi 1995, Omar et al. 2011, 2013, authors’ unpublished data). In Sumatra, H. maxi shares its habitat with Crocidura beccari, Crocidura lepidura, Crocidura aequicauda, Tupaia javanica, Tupaia ferruginea, Tupaia tana, Maxomys inflatus, Maxomys rajah, Maxomys surifer, Mus caroli, Niviventer fraternus, Lariscus niobe, and Sundasciurus altitudinis (Robinson and Kloss 1918, Miller 1942, Ruedi 1995, authors’ unpublished data).

Hylomys maxi is omnivorous. In Peninsular Malaysia stomach contents contained tapioca bait and several crickets (N = 1; Harrison 1961). In Sumatra, stomachs contained insects of various orders, principally small beetles (N = 2–3). In captivity they feed on fish, shrimp, insects (cockroaches, mealworms, grasshoppers), raw meat, and soft fruit, showing less interest in the latter (Davis 1965, Medway 1969, Genoud and Ruedi 1996). In Selangor, a pregnant individual with two embryos was recorded in March and females with enlarged mammae were recorded from May to November (Medway 1969; USNM specimens). Hylomys maxi is cathemeral and found alone or in small groups (2–3 individuals; Fred Ulmer field notes, Medway 1969). In West Malaysia, average densities in favourable habitats are estimated at 3–4 individuals/ha. Home ranges are c. 30–40 m in diameter (Rudd 1980).

Conservation:

Recorded in Taman Negara, Kerinci Seblat and Leuser national parks, Pulau Tioman Wildlife Reserve, and Ulu Gombak, Ulu Langat, and Temengor forest reserves. Temengor has been selectively logged since 2001 and captures of H. maxi have not changed between sites with and without logging residues (Yamada et al. 2016). It is unclear how ongoing deforestation in Kerinci Seblat and Leuser National Parks (see H. vorax and H. parvus accounts) is affecting H. maxi, nor the degree of deforestation in Malaysian protected areas. Hylomys maxi has been recorded in anthropized habitats such as gardens and sweet potato patches, but not in palm oil plantations. To assess the conservation status of H. maxi, further research is required to determine population trends within protected areas and to evaluate whether this species can adapt to palm oil plantations, considering much of the regional habitat is undergoing land-use change. Likewise, additional field campaigns combined with habitat suitability modelling seem pivotal to evaluate the species area of occupancy. Conservation planning must also acknowledge that H. maxi contains two divergent lineages that, pending additional research, may deserve species-level recognition (see below), and an altered conservation status. In the meantime, we suggest these two lineages are treated independently by regional government agencies and NGOs.

Comments:

Sumatran populations seem to be smaller on average than Peninsular Malaysia populations and are somewhat differentiated in the craniodental PCA but do overlap, although this could be due to ontogenetical sampling biases. Sumatran populations also differ from Peninsular Malaysia populations with cytochrome b p-distances of c. 6% and have a private allele for the BDNF fragment, but not for GHR. Further geographical and nuclear loci sampling, particularly across Sumatra (which is poorly represented in museums), will be required to assess whether Sumatran and Peninsular Malaysia’s populations represent two different species. Abd Wahab et al. (2022) retained Tioman Island populations as a subspecies of H. suillus (H. s. tionis), but our results suggest it is found within the diversity of H. maxi. We do not believe this population deserves subspecies-level recognition given that this taxon was described based on pelage coloration, which we now know, after looking at a much larger series than Chasen (1940), that it is highly prone to ecophenotypic, ontogenetic, and seasonal variation.

Hylomys peguensis Blyth, 1859

Hylomys peguensis Blyth, 1859: 294. Original description.

Hylomys siamensis Kloss, 1916: 10.

Hylomys suillus microtinus Thomas, 1925: 497.

Hylomys suillus suillusCorbet, 1988: 122 (part).

Subspecific taxonomy

Hylomys peguensis peguensis Blyth, 1859: 294.

Hylomys peguensis siamensis (Kloss), 1916: 10.

Hylomys siamensis Kloss, 1916: 10.

Hylomys peguensis microtinus (Thomas), 1925: 497.

Hylomys suillus microtinus Thomas, 1925: 497.

Holotype:

The original description does not specify a reference number. Adult male and female collected by Major Berdmore and housed in the Indian Museum (Zoological Survey of India, National Zoological Collection), at Kolkata (syntypes). Preserved in spirit according to the description but one of the skulls was subsequently extracted (Anderson 1874). Khajuria et al. (1977) revised the mammal types in this museum and did not cite these specimens. The staff at the Zoological Survey of India National Zoological Collection have stated that these syntypes are lost (Uttam Saikia pers. comm.).

Type locality:

‘Schwe Gyen, in the valley of the Sitang river, Tenasserim province’ (Shwegyin, valley of the Sittang river, Bago, Myanmar).

Paratypes:

Non-existent.

Neotype designation:

Given that the type series is lost and that there are no available topotypes, we here designate as a neotype the geographically closest specimen to the type locality that is available: USNM 583813. This specimen is an adult male collected at ‘Kinmun, 5 Km N E Of, By Road, Yetagon Myaung, Kyaikhtiyo Wildlife Sanctuary, Mon, Myanmar’ (386 m a.s.l.; 17.4442, 97.0994) on 15 March 2002 by J.F. Jacobs. This locality is approximately 58 km away from the type locality and there are no geographical barriers in between. The neotype specimen morphology matches the species description and Anderson (1874) redescription of H. peguensis (venter coloration, nasals extend to the level of the antorbital rim, strongly developed cingulum in P4) and it is unlikely to represent any other species, given that the other two mainland Asian Hylomys species allopatric distribution ranges do not span this region. The specimen consists of a museum study skin with accompanying cranium and mandible, remaining body in fluid, and frozen tissue samples; skin has retained the phallus and all four limbs and is in a perfect condition; skull has a small area of the palate, posterior to the right incisive foramen, broken, but is otherwise in a good condition. Total length 172 mm, tail 20 mm (13% HB), hindfoot 26 mm, ear length 19 mm, weight 73 g. Skull measurements in Supporting information, Table S2. 3D surface scans of this neotype skull are available at MorphoSource: https://doi.org/10.17602/M2/M560027 (cranium); https://doi.org/10.17602/M2/M560031 (mandible). Pictures of this holotype skin and skull are shown in Figures 7A and 8A. DNA sequences associated with this neotype have been deposited in GenBank: OR554485, OR554506, OR554522, OR554414, OR554417, OR554461.

Figure 7.

Dorsal and ventral views of the skins of (A) Hylomys peguensis (USNM583813, neotype), and (B) Hylomys macarong (USNM 320500, holotype). Photo credits: Katie Sayers.

Figure 8.

Skulls of (A) Hylomys peguensis (USNM583813, neotype) and (B) Hylomys macarong (USNM 320500, holotype). All images are at the same scale. Photo credits: Katie Sayers.

Emended diagnosis:

A medium-sized Hylomys (average HB = 129 mm, W = 54.2 g, GLS = 34.4 mm) characterized by a brown dorsum fur with a yellow tinge and relatively lighter coloured appearance than Sundaic relatives (except at the southern tip of its range), silvery tinged with buff venter coloration, bicoloured and relatively long tail (average T = 21.3 mm, T/HB = 17%), light brown feet, short hindfoot (average HF = 22.5 mm, HF + nail = 24.8 mm), relatively short ears (average E = 16.4 mm, E/GLS = 50%) with a whitish/light brown rim (except in one Cambodian specimen). Its skull is short (average GLS = 34.4 mm) and has a relatively short rostrum (average ROL = 14.6 mm). It can be distinguished from all other relatives by the combination of the following characters: obtuse-angled notch on anterior side of premaxilla; nasals generally extend to the level of the antorbital rim (9/11 specimens); highly to moderately developed supraorbital processes; antorbital ridge/flange poorly developed; relatively shallow antorbital fossa; anterior side of the antorbital rim is anterior or in line with M1 second cusp; lacks posteriorly extended horizontal crest in maxillary, behind infraorbital canal; long alisphenoid canal (average ASC = 1 mm); ectotympanic anterior process is robust and thickens dorsoventrally towards the tip, interior side of ectotympanic anterior process is not elongated and expanded anteriorly; I2 is oriented posteriorly; flat braincase at the occipitum (average BD = 9.3 mm); relatively gracile mandibular ramus (average LAP = 9.1 mm) with narrow coronoid process; average sized upper first incisor (average I1 = 2.6 mm) not projected towards lingual side, so that space between incisors does not narrow towards the tip; labial cingulum in P4.

Comparisons:

It is currently thought to have an allopatric distribution with Sumatran and mainland Asian H. maxi, from which it can be distinguished based on its generally lighter dorsum (brown with yellowish tinge vs. without), venter (buff vs. grey-brown) and feet coloration (light vs. dark brown), particularly on its forefeet (light brown vs. dark brown/black), presence of whitish/pale brown in ear rim (except in Cambodian population), longer tail and shorter hindfoot (Supporting information, File S1, Fig. S6), shorter skull length (32.3–36.0 vs. 35.5–38.4 mm), shorter rostrum length (RL: 13.1–15.8 vs. 16.1–18.2 mm), shorter upper and lower toothrows (IM3SA: 15.8–18.3 vs. 18.3–19.7 mm, IM3I: 15.1–17.4 vs. 17.6–18.9 mm), flatter braincase at occipitum (BD: 9.0–9.6 vs. 9.3–10.5 mm), and longer alisphenoid canal (0.7–1.2 vs. 0.1–0.4 mm). Posterior end of nasals generally extends to the level of the antorbital rim but it does not in H. maxi. It has similar sized supraorbital processes to H. maxi despite its much smaller skull size. Antorbital flange is thinner in H. peguensis giving a shallower appearance to the antorbital and nasolabilis fossa than H. maxi. Anterior side of the antorbital rim is anterior or in line with M1 second cusp, while it is posterior in H. maxi. Its robust ectotympanic process thickens dorsoventrally towards the tip, while it is generally more gracile and dorsoventrally even in H. maxi. Its P4 labial cingulum is generally thinner/less prominent than H. maxi.

Comparisons with additional species have been included in the following accounts, after such species have been formally defined.

Distribution, habitat, and natural history:

Distributed across most of Indochina’s hills and mountains, except the Da Lat and Dak Lak Plateaus and surrounding lowlands. Its southern distribution limit seems to be in Perlis (Malaysia), possibly at the K-P vegetation transition, mirroring the distribution of other small mammals (Hinckley et al. 2023). In southern Myanmar, specimens were caught at c. 400 m a.s.l. along a fast-running stream with boulders in a heavily cut secondary forest (Jacobs field notes, USNM Mammal Division archives). In northern Myanmar, it was collected dead on a path on a thickly wooded hillside at c. 900 m a.s.l (Anderson 1874). In central and north Thailand confined to hilly and mountain areas, with dense undergrowth (Lekagul and McNeely 1977), and under logs in a grove of wild bananas (Allen and Coolidge 1940). In Huai Kha Khaeng Wildlife Sanctuary, central Thailand, H. peguensis has been recorded in dry evergreen and mixed deciduous forests with a relatively closed canopy (>60%), and a variable, but reduced ground cover (<60%) at an altitude of 550–650 m a.s.l., but not in a mixture of dry mixed deciduous forest and dry dipterocarp forest at an elevation of 400–500 m a.s.l., and a relatively open canopy (<60%) with thick ground cover (>60%; Walker and Rabinowitz 1992). In northern Lao P.D.R. it was recorded in a small, vegetated gully with a small running stream, in a secondary and highly degraded semi-deciduous forest (authors’ unpublished data). In northern Viet Nam the species has been recorded from c. 100–1000 m a.s.l., but it is most abundant at higher elevations (900–1000 m a.s.l.) in habitats with plentiful logs (Abramov et al. 2013, this study), and on stream banks with wild bananas and reeds in open spaces near forest (Van Sung 1976). In Kiri Rom Plateau, Cambodia it has been recorded at 700 m a.s.l., in the grass at the edge of a pine forest (Bernard Feinstein’s field notes, USNM Mammal Division archives).

In the north-west of its distribution, H. peguensis appears to be parapatric with Neotetracus sinensis, which occurs at higher elevations (1500–2700 m) (Van Sung 1976, Corbet 1988).

It seems to have a parapatric or allopatric distribution with Neohylomys hainanensis, which has recently been recorded in Viet Nam, in evergreen mixed forest north of the Red river (Cao Bang, Vinh Phuc, and Bak Kan provinces) at elevations of 300 to c. 850 m a.s.l, but not at higher elevations of 1500–1800 m a.s.l. (Abramov et al. 2018, this study). Although both species might overlap in their elevation ranges, they have not been recorded in sympatry; in fact, H. peguensis has been recorded in the Red River valley (Lao Cai province) but not north of it. In Kyaikhtiyo Wildlife Sanctuary, Myanmar, H. peguensis shares its habitat with Crocidura cf. fuliginosa, Tupaia belangeri, Cannomys badius, Maxomys surifer, Mus cookii, Rattus andamanensis, and Callosciurus phayrei (Jacobs field notes, USNM Mammal Division archives). In Huai Kha Khaeng Wildlife Sanctuary, central Thailand, it shares its habitat with Tupaia belangeri, Cannomys badius, Maxomys surifer, Leopoldamys herberti, Niviventer sp., Menetes berdmorei, and Rattus rattus (Walker and Rabinowitz 1992). In Phu Lom Lo, northern Thailand, it shares habitat with Berylmys berdmorei, Cannomys badius, Leopoldamys herberti, Maxomys surifer, Niviventer mekongis, Rattus nitidus, Rattus rattus, Dremomys rufigenis, and Menetes berdmorei (Robert Elbel’s field notes, USNM Mammal Division archives). In Kiri Rom Plateau, Cambodia, it shares its habitat with Tupaia belangeri, Maxomys surifer, Mus shortridgei, Rattus rattus, and Rattus andamanensis (Feinstein’s field notes, USNM Mammal Division archives).

Earthworms, snails, ants, beetles, butterfly caterpillars, and fruits of Ficus and Melastoma have been recorded in specimen stomachs (Van Sung 1976). Males had larger testicles in March (13 × 6 mm) than in December (7–8 × 3 mm) suggesting this species might breed in spring or summer in northern Viet Nam (Van Sung 1976). Gravid females have been recorded in late-February on the Myanmar–China border, and in mid-May in Ba Vi National Park, Viet Nam. Litter size was 2 (N = 1) and 4 (N = 2), respectively (AMNHM-44272; Abramov et al. 2013). An adult male and female specimen were caught in the same trap during consecutive days suggesting the species might form pair bonds or males may mate guard, and that male and female territories must overlap (perhaps just when breeding; Jacobs field notes, USNM Mammal Division archives).

Conservation:

Recorded in Thung Yai Naresuan, Huai Kha Khaeng and Kyaikhtiyo wildlife sanctuaries, Mae Wong, Ba Vi, Xuan Son, and Preah Monivong Bokor National Parks (Walker and Rabinowitz 1992, Robinson et al. 1995, Abramov et al. 2013, Pavlova et al. 2018; Supporting information, Table S1). Recorded in Khasia pine plantations with native undergrowth vegetation, and in margin between evergreen forest and short-term agriculture but not in Khasia pine plantations with coffee or macadamia plantations (Chaiyarat et al. 2020, THNHM field notes). To evaluate the conservation status of H. peguensis, additional research is necessary to ascertain population trends within protected areas and to understand the species tolerance to environmental degradation in anthropized landscapes. Additional field surveys and habitat suitability modelling is necessary to predict future trends. Some of the described subspecies of H. peguensis may deserve species-level recognition (see below), and as such we suggest each is treated as an independent conservation unit by regional government agencies.

Comments:

This species exhibits higher levels of genetic structure than other relatives but the craniodental morphospace of these lineages (peguensis, siamensis, and microtinus) overlaps, and sample size is low. Thus, further geographic sampling will be required to assess if these populations might deserve specific status. In the meantime, we suggest retaining siamensis and microtinus as subspecies of H. peguensis. The literature regarding the distribution limits of H. peguensis and H. maxi is somewhat confusing. Lekagul and Mcneely (1977) stated that only H. s. siamensis is present in Thailand, indirectly suggesting that H. peguensis distribution should at least extend to the K-P. Chasen (1940) recorded the lighter coloured H. s. siamensis (H. peguensis) as far south as Tapli (Ranong) in the Isthmus of Kra based on specimen ZRC4.5033 and only a few miles north of the Malaysian boundary. However, Hill (1960) describes this supposedly ‘light coloured’ specimen as very brown with a strong reddish tinge and a clear grey underside (features that resemble more H. maxi). Hill (1960) states that H. maxi distribution extends to Tasan (Chumphon) and is present in Pelarit (Perlis). Ruedi et al. (1994) followed Hill (1960). The only molecular evidence we have (two short nuclear fragments of specimen MZUM-M00057, Kaki Bukit, Perlis), along with diagnostic craniodental (e.g. rostrum length) and selected external measurement bivariate plots, seem to support Chasen (1940), suggesting that H. peguensis distribution extends to the Malaysian boundary, possibly at the K-P.

Hylomys macarong Hinckley, Lunde & Hawkins, sp. nov.

Hylomys sp. Bannikova et al., 2014: 501. Informal name.

Hylomys suillus microtinus, Best, 2018: 328.

Hylomys suillus ssp. 2 Zeng et al., 2022: preprint fig. 3. Informal name.

Zoobank registration:

urn:lsid:zoobank.org:act:A1305A1B-AA64-479F-9D88-59EAE5CCE2F0.

Holotype:

USNM 320500, an adult male collected by Bernard R. Feinstein on 21 July 1961. The specimen consists of a museum study skin with accompanying cranium and mandible; skin is in relatively good condition, but is partially open along the suture line, and has a twisted right forelimb with the most external phalanx missing (clipped for DNA extraction), right ankle skin is pierced by tibia or calcaneus, partially chewed ears (possibly by ants, as reported for other specimens of the type series), and broken proximal and distal sections of tail, exposing the internal wire; skull is well prepared but has damaged pterygoid processes, left zygomatic arch, dorsal side of angular process and tips of left i1 and i2; and left P1 and occipital condyle are missing (unerupted and broken, respectively). Total length 165 mm, tail 25 mm (18% HB), hindfoot 27 mm, ear length 18 mm. Skull measurements in Supporting Information, Table S2. 3D surface scans of this holotype skull are available at MorphoSource: https://doi.org/10.17602/M2/M560006 (cranium); https://doi.org/10.17602/M2/M560011 (mandible). Pictures of this holotype skin and skull are shown in Figures 7B and 8B. DNA sequences associated with this holotype have been deposited in GenBank: OR138081 and OR554435.

Type locality:

‘Fyan, Lam Dong, Vietnam’ (now known as Phú Sơn, Lâm Hà District, Viet Nam, 11.88º N, 108.2º E, c. 1200 m a.s.l.; source: US military intelligence).

Paratypes (2):

USNM 320492, 320495: same as type locality; USNM 320485: Thac Datan La, Dalat, Lam Dong, Viet Nam, c. 1450 m a.s.l.; USNM 320490: 1 km E of Poste De Mdrak, Dac Lak, Viet Nam, c. 500 m a.s.l.

Representative DNA sequences:

Deposited (or already available) in GenBank with the following accession numbers: KF783150, KF783151 (cytochrome b, two representatives), (mitochondrial genomes, two representatives). New sequences: OR138079-OR138081 (mitochondrial genomes, three representatives), OR554432-OR554435 (GHR, four representatives). DNA sequences associated with the following voucher specimens: USNM 320485, 320490, 320495, 320500 (this study); USNM 320485, 320501 (Zeng et al. 2022); ZMMU S-190307 (Bannikova et al. 2014).

Etymology:

The specific name ‘macarong’, which means vampire in Vietnamese (Ma cà rồng), acknowledges the long fangs (first upper incisors) that characterize mature males of this species. We suggest the common names Dalat Gymnure, Chuột Voi Đà Lạt, and Gimnuro de Dalat, in English, Vietnamese and Spanish, respectively, given that this species is just currently known from the Da Lat and Dak Lak Plateaus and surrounding lowlands, in southern Viet Nam.

Diagnosis:

A medium-sized Hylomys (average HB = 138.4 mm, GLS = 34.8 mm) characterized by a rusty chest coloration in males (Fig. 7B; possibly just present during spring/summer, since specimens collected in April and possibly winter (inferred due to long, soft, and thick pelage), lacked this coloration), bicoloured and long tail (average T = 23.2 mm, T/HB = 16.8%), light brown feet, short hindfoot (average HF = 23.5 mm, HF + nail = 25.7 mm), ears are uniformly brown-coloured and relatively long (average E = 18.1 mm, E/GLS = 52.6%). Its skull is characterized by: obtuse-angled notch between premaxilla tips; narrow rostrum (average ROB = 5.2 mm); long nasals (average NL = 13 mm; NL/GLS = 37%) that extend to the level of the antorbital rim; highly to moderately developed supraorbital processes; conspicuous and posteriorly expanded horizontal crest behind infraorbital canal; alisphenoid canal intermediate in length among other Hylomys species (average ASC = 0.5 mm); interior side of ectotympanic anterior process elongated and greatly expanded anteriorly; deep braincase at the occipitum (average BD = 9.7 mm); robust mandibular ramus (average LAP = 10.1 mm) with broad coronoid process; very long first upper incisor (average I1 = 3.3 mm), particularly in males (average I1 = 3.6 mm males, 2.9 mm females; Supporting Information, Fig. S7), which is slightly projected towards lingual side, so that space between incisors generally narrows towards the tip; P4 labial cingulum absent or poorly developed.

Comparisons:

In summer, easily distinguished from its closest geographical and genetic relative, H. peguensis, by its rusty venter coloration (in males), which contrasts with H. peguensis’s pale grey coloration (Fig. 7). White rim in ear is absent vs. present in H. peguensis (except in Cambodian specimen). Rostrum is longer (average = 15.1 mm) and narrower (average = 5.2 mm) than H. peguensis (average = 14.6 mm, 5.5 mm). Nasals are longer (average = 13 mm) vs. shorter (average = 11.8 mm) in H. peguensis. Conspicuous and posteriorly expanded horizontal crest behind infraorbital canal vs. less prominent and expanded in H. peguensis. Alisphenoid canal shorter (average = 0.5 mm) vs. longer (average = 1 mm) in H. peguensis. Interior side of ectotympanic anterior process elongated and greatly expanded anteriorly vs. not expanded in H. peguensis. Braincase at the occipitum is deeper (9.4–10.2 mm) vs. flatter (9–9.6 mm) in H. peguensis. Mandibular ramus broader (average LAP = 10.1 mm) vs. narrower (average LAP = 9.1 mm) in H. peguensis, in fact, these species can be diagnosed through the combination of angular process length and rostral breadth (Supporting information, File S1, Fig. S8). Coronoid process is broader vs. narrower in H. peguensis, making the space between the distal region of coronoid and condylar processes frequently more narrow/acute in H. macarong than in H. peguensis (Fig. 8). I1 sexually dimorphic and longer (males: 3.6, 2.6–4.5 mm, females: 2.9, 2.6–3.15 mm) vs. sexually monomorphic and shorter (males: 2.6, 2.1–3 mm, females: 2.6, 2.4–3.1 mm) in H. peguensis. I1 generally slightly projected towards lingual side, so that space between incisors generally narrows towards the tip vs. I1 slightly projected towards labial side, so that space between incisors does not narrow towards the tip in H. peguensis. I2 projected posterolabially vs. posteriorly in H. peguensis. P4 labial cingulum absent/poorly developed vs. generally well developed/conspicuous in H. peguensis.

It is currently thought to have a highly disjunct allopatric distribution with Malayan H. maxi, from which it can be easily distinguished based on its shorter rostrum (RL: 14.6–15.7 vs. 16.1–18.2 mm), upper toothrow (IM3SB: 16.1–17.4 vs. 17.6–19.6 mm) and molars (P4M3: 7.6–8.3 vs. 8.3–9.9 mm). Additionally, it can also be diagnosed from H. maxi due to its generally longer first incisor and through the combination of angular process length and rostral breadth (as with H. peguensis; Supporting information, File S1, Figs S7, S8).

Distribution, habitat, and natural history:

Currently known from the Langbian/Da Lat and Dak Lak Plateaus and surrounding lowlands, in southern Viet Nam. Collected at c. 50–80 m a.s.l. in semi-deciduous dipterocarp forest with extremely dense understory with shrubs, saplings, and lianas; at a site at 1700 m a.s.l. containing both pine savanna (dominated by Pinus kesiya) with a dense grass cover and a tract of primary evergreen hill forest (Adler et al. 2001); and at a Dalat pine forest at c. 1520 m a.s.l. (USNM 269789; J.F. Rock). Specimens were collected under log in grass (N = 6), grass (N = 5), grass area in wood edge (N = 5), under log (N = 1), under broadleaf herbaceous shrub near wet spot (N = 1), and at base of Pandanus in moist area (N = 1). Not recorded in bamboo patches or inside fagaceous forest, as with some sympatric rats, squirrels, and treeshrews (Feinstein’s field notes, USNM Mammal Division archives). Shares its habitat with Crocidura spp., Suncus murinus, Tupaia belangeri, Bandicota savilei, Berylmys berdmorei, Berylmys mackenziei, Chiromyscus langbianis, Leopoldamys milleti, Maxomys moi, Maxomys surifer, Mus carolii, Mus pahari, Niviventer mekongis, Rattus andamanensis, Rattus argentiventer, Rattus rattus, Vandeleuria oleracea, Dremomys rufigenis s.l., and Menetes berdmorei (Feinstein’s field notes, USNM Mammal Division archives; USNM collection; Adler et al. 2001).

A juvenile female and adult male were collected in the same trap during two consecutive days, perhaps suggesting parental care and/or male territory overlap with offspring at least until adulthood. Two juvenile female specimens were also collected 3 m away from each other during two consecutive days. Mature males have a reddish ventral coloration in July, which subadults and females lack, suggesting it might represent a sexual character. This coloration might last until late September, since one specimen (USNM 320485) was moulting it but still had enlarged testes (13 × 6 mm; using H. peguensis specimens as a reference), perhaps suggesting end of breeding season. One female with six mammae was recorded by mid-December (USNM 320490), and juveniles have been recorded from mid-September to mid-October. Specimens were collected with snap traps baited with banana and manioc.

Conservation:

Recorded in Bu Gia Map and Cat Tien National Parks (Bannikova et al. 2014) and in proximity to Kalon Song Mao and Hon Ba Nature Reserves (Supporting information, Table S1). Viet Nam now has the second-highest rate of deforestation in primary forests (Poyarkov et al. 2023). Co-distributed species with a potential similar extent of occurrence than that currently known for H. macarong, such as Pygathrix nigripes (Milne-Edwards 1871), Trachypithecus margarita (Elliot 1909), Nomascus gabriellae (Thomas 1909), and Murina harpioloides (Kruskop and Eger 2008), are listed as endangered or critically endangered (IUCN Red List of Threatened Species 2023). Further research is needed to assess its tolerance to habitat degradation, and to evaluate whether this species is present in other potentially suitable protected areas, such as Ea So, Krong Trai, Nam Nung, and Nam Kar Nature Reserves, and Lo Go Xa Mat and Ta Dung National Parks, in Viet Nam, and Lumphat, Phnum Prech, and Phnom Namlear Wildlife Sanctuaries in Cambodia.

Comments:

Specimens from Kon Tum province, Central Highlands region of Viet Nam (MNHN1929320–5) have not been sequenced but were clustered in the morphospace of H. peguensis and were thus assigned to H. peguensis in univariate descriptive statistics and bivariate plots.

Hylomys vorax Hinckley, Lunde & Hawkins, sp. nov.

Hylomys suillus maxi Miller, 1942: 109–111.

Hylomys suillus maxi Best, 2018: 328.

ZooBank registration: urn:lsid:zoobank.org:act:BA2B0BCE-AF92-4583-8C74-1A3EAA60AE7B.

Holotype:

USNM 271034, an adult male collected by Frederick A. Ulmer, Jr on 1 May 1939. The specimen consists of a museum study skin with accompanying cranium and mandible; skin well prepared, with a small hairless patch on the two thighs and the most external toe of the left hindfoot missing (clipped for DNA extraction), otherwise in good condition; skull well prepared (with some remaining dried tissue attached, allowing for subsequent historical DNA extraction), but has damaged pterygoid processes, left P2 crown, and ectotympanic bone of the right bulla. Total length 148 mm, tail 22 mm (17.5 % HB), hindfoot 27.5 mm, ear length 19 mm, weight 42.5 g. Skull measurements in Supporting Information, Table S2. 3D surface scans of this holotype skull are available at: https://doi.org/10.17602/M2/M560018 (cranium); https://doi.org/10.17602/M2/M560022 (mandible). Pictures of this holotype skull and skin are shown in Figure 9A and the Supporting Information, Figure S9, respectively. DNA sequences associated with this holotype have been deposited in GenBank: OR138102 and OR554454.

Type locality:

‘Bivouac No. 5, Loser Trail, Atjeh’ (now known as Mt. Leuser trail, Aceh, Sumatra, Indonesia). This locality is shown in the map of de Schauensee and Ripley (1939), and we have estimated it at c. 3.85º N, 97.12º E, 2408 m a.s.l. The type locality could be georeferenced after we located two flanking sampling sites of the George Vanderbilt Sumatran Expedition (Ulmer’s field notes, Schauensee and Ripley 1939, Miller 1942): Mt. Leuser peak no. 1 (3.791072º N, 97.211351º E) and Blangbeke (c. 3.891223º N, 97.152307º E). Blangbeke could be easily georeferenced by the authors of this manuscript since it is the only ‘blang’ (open area) with a river on a ‘flat plateau of about 7000 ft’ found to the south of ‘Base camp (3600 ft), Blangnanga’, which is 20 km West of the town of Blankejeren according to Fred Ulmer’s notes. The river is called ‘Waih Blang Bebeke’, which resembles the locality’s name. The type locality ‘Bivouac No. 5’ is shown on the map of de Schauensee and Ripley (1939), just south of the Waih Blang Bebeke headwater. The approximate bivouac location was estimated in this manuscript through the combination of this map and information retrieved with Google Earth topographic layers and street view, and Fred Ulmer’s references and pictures housed at the ANSP Mammal Department archives.

Paratypes (9):

ANSP 20379, 20380, 20381: same as type locality; ANSP 20374, 20375, 20376, 20377, 20378, and USNM 271033: ‘Blangbeke, Atjeh’ (=Blang Beke, Aceh, Sumatra; elevation 2073–2134 m a.s.l.). Paratype ANSP 20377 was illustrated by Frederick A. Ulmer (Supporting Information, File S1, Fig. S11)

Representative DNA sequences:

Deposited in GenBank with the following accession numbers: OR138101 and OR138102 (mitochondrial genomes, two representatives), OR554452-OR554454 (GHR, three representatives). DNA sequences associated with the following voucher specimens: USNM 271034, ANSP 20377, 20380.

Etymology:

The specific name acknowledges the voracious behaviour that Frederick Ulmer, the collector of the type series, described in his field notes: ‘They were voracious beasts often devouring the whole bait before springing the trap. Ham rind, coconut, meat, and walnuts were eaten. One shrew partially devoured the chicken head bait of a steel trap before getting caught in a nearby Schuyler trap baited with ham rind’ (USNM 271033). We recommend the common names Leuser Gymnure, Salak Ba’a Leuser, and Gimnuro de Leuser, in English, Malay, and Spanish, respectively.

Diagnosis:

A medium-sized Hylomys (average HB = 123 mm, W = 44.4 g) characterized by a thick and soft, dark brown fur, a long and monocoloured black tail (average T = 21.4 mm; T/HB = 17.5%), and long ears (average E = 19.1 mm, E/GLS = 56.3%). Ventral coloration is paler than the dorsum, with buff-tipped, dark-grey hairs. Brown-yellowish hue in the throat area is generally inconspicuous and just present in females (but these could be due to phenological sampling biases: specimen series collected in April). All specimens have an ochraceous hue around the cloaca/inguinal area. Fore/hindfeet and tail have a lighter and darker coloration than dorsum, respectively. The skull of this species is distinguished from all other congeneric species by a unique combination of the following characters: acute-angled notch between premaxilla tips; long nasals; narrow rostrum; poorly developed antorbital ridge/flange; broad interorbital constriction; small but prominent supraorbital process in males, which is generally inconspicuous in females; relatively smooth frontal bone and dorsal region of maxillary; anterior palatine foramina (generally) anterior or (rarely) at level of maxilla/palatine suture; generally thick and robust postpalatine torus; lack of posterior nasal spine; elongated pterygoid process; robust and globular ectotympanic process that tappers dorsoventrally towards proximal end; gracile mandible, with relatively short vertical height at coronoid and length of angular process; procumbent incisors; single root in P2; P2 and P3 lack posterior cuspules, even in subadults and young specimens; absent labial cingulum in P4; anterior margin of p1 crown does not extend to root of c1; p3 is much larger than p2 and larger than p1, p1 is generally larger than p2; posterior cuspule of p2 absent, even in subadults and young specimens; single root in p3.

Comparisons:

Compared with its closest geographical and genetic relative, Hylomys maxi, H. vorax is distinguished by its smaller (average HB = 123 mm; W = 44.4 g; GLS = 34.5 mm; LMA = 24 mm; Fig. 9A) vs. larger size (HB = 139 mm; W = 64.2 g; GLS = 36.9 mm; LMA = 26.3 mm; Fig. 9B), uniformly black and relatively longer (average T = 21.4 mm, T/HB = 17.1–18.8%) vs. bicoloured (except specimen RMNH 23896) and relatively shorter tail (average T = 16.2 mm, T/HB = 5.11–16.9%) (Supporting Information, File S1, Fig. S10A), and a relatively longer (E/GLS = 54.8–55.9%) vs. shorter ear (E/GLS = 39.3–48.2%) (Supporting Information, File S1, Fig. S10B). Its fur is softer, denser, and longer (7–8 mm at axial region) than H. maxi (4–5 mm at axial region), although higher elevation (>1100 m a.s.l.) Sumatran H. maxi specimen furs were not measured, and this trait is subject to great ecophenotypic variation in this genus (Ruedi et al. 1994).

Figure 9.

Skulls of (A) Hylomys vorax (USNM 271034, holotype) and (B) Hylomys maxi (USNM 487905, standard specimen since holotype is damaged). All images are at the same scale. Photo credits: Katie Sayers.

The skull of H. vorax has a narrower and more gracile appearance than H. maxi. The notch between the premaxilla tips is acute vs. obtuse in H. maxi (Fig. 10). Its rostrum breadth is narrower (4.5–5.0 mm) vs. broader (5.2–6.4 mm) in H. maxi. Its nasals are relatively longer (average NL/GLS = 36.1%) vs. shorter in H. maxi (average NL/GLS = 32.4%; Supporting Information, File S1, Fig. S10C). Antorbital ridge/flange is undeveloped vs. well developed in H. maxi, particularly between the lacrimal flange and supraorbital process. Its frontal bone and dorsal region of maxillary has a relatively smooth vs. rugged appearance in H. maxi. Its interorbital constriction is proportionately broader than H. maxi (Supporting Information, File S1, Fig. S10D). Zygomata and supraorbital processes are less robust than in H. maxi. Anterior palatine foramina end anteriorly or (rarely) at maxilla/palatine suture vs. posterior to maxilla/palatine suture in H. maxi. Postpalatine torus is generally thicker and more robust than in H. maxi. Posterior nasal spine is absent vs. present in H. maxi. Postpalatal length is shorter (11.1–12.3 mm) vs. longer (12.3–14.2 mm) in H. maxi, while braincase is similar in breadth (average BB = 14.8 vs. 15 mm). Epipterygoid processes are lateroventrally vs. lateral oriented in H. maxi. Its robust ectotympanic process thickens dorsoventrally towards the tip, while it is more gracile and dorsoventrally flattened in H. maxi (Fig. 10B). Occipital crest absent or inconspicuous and shorter vs. prominent and longer in H. maxi. Its mandible is more gracile, with a shorter vertical height at coronoid and shorter length of angular process (HCO = 7.4–8.6 mm; LAP = 7.9–9.1 mm) than H. maxi (HCO = 8.6–10.7 mm; LAP = 8.8–10.8 mm).

Figure 10.

A, dorsal view of the rostrum of two H. vorax specimens (left; ANSP 20374 and 20377) and two H. maxi specimens from Sumatra (right; ANSP 20372 and ZRC4.5051), which illustrates the narrow, notched space between premaxillary tips of H. vorax (vs. broad/obtuse in H. maxi). B, lateral view of the bullae of H. vorax (left; ANSP 20377) and H. maxi (right; USNM487909), which illustrates the robust ectotympanic process that thickens distally in H. vorax (vs. more gracile and dorsoventrally flattened in H. maxi).

Its dentition appearance is less robust than that of H. maxi, with smaller teeth and more space between incisors, canine, and premolars than H. maxi. I1 is more procumbent than H. maxi. P1 is usually similar in size to P3 vs. generally much larger in H. maxi. P4 labial cingulum absent or inconspicuous vs. well developed in H. maxi. Molars are smaller (M1M1 = 9.8–10.8 mm; P4M3 = 7.5–8.2 mm) vs. larger (M1M1 = 10.9–12.5 mm; P4M3 = 8.3–9.9 mm) in H. maxi. Its lower toothrow is shorter (16.6–17.7 mm) vs. longer (17.6–18.9 mm) in H. maxi. Crown height of c1 is generally greater in H. vorax than H. maxi, the latter species c1 is more expanded anteriorly, giving it a less prominent appearance. Anterior margin of p1 crown does not extend anteriorly to the root of c1, but it does in H. maxi. p3 is usually much larger than p2 in H. vorax, while it has a similar size or p3 is just slightly larger than p2 in H. maxi.

Distinguished from H. parvus by its larger size (HB = 117–128 mm; GLS = 33.4–36.3 mm; LMA = 23–25.6 mm vs. HB = 100–115 mm; GLS = 29.2–33.4 mm; LMA = 20.1–22.2 mm) and black monocoloured and relatively shorter (average T/HB = 17.5%) vs. bicoloured and relatively longer tail (average T/HB = 22.4%). Rostrum is longer (ROL = 15.2–17.3 vs. 13.6–15.0 mm) but similar in width (ROB = 4.52–5.05 vs. 4.43–5 mm) to H. parvus, giving H. vorax a more elongated appearance. Toothrow length is longer (IM3Sa = 17.2–18.5 mm; IM3I = 16.6–17.7 mm) than in H. parvus (IM3Sa = 16.1 mm; IM3I = 14.5–16 mm). Braincase and palate breadth are wider in H. vorax (BB = 14.5–15.1 mm; M3B = 6.1–6.4 mm) than H. parvus (BB = 13.1–14 mm; M3B = 4.7–6 mm). Height of the coronoid process is proportionately shorter than H. parvus (Supporting Information, File S1, Fig. S10E). First upper incisor is more procumbent than H. parvus. Males have small but prominent supraorbital processes, which are absent in H. parvus. It has a single root in P2 and p3 vs. two non-fused roots in H. parvus (Jenkins and Robinson 2002).

Distribution, habitat, and natural history:

Currently known from montane forest from 2073 to 2835 m a.s.l. on Mount Leuser, but it seems to be most abundant at elevations ranging from 2073 to 2408 m a.s.l. (Fig. 11). This habitat is characterized by ‘trees averaging only 15 to 40 feet in height, very hard, knotted and twisted. Everywhere the ground and the branches of the trees were covered with a deep carpet of moss and ferns’ (Ulmer’s field notes, Miller 1942). However, one specimen was recorded at 2835 m a.s.l., described as ‘flat heathy plateau mostly bare or covered with grass interspersed with patches of bushes and low trees’ such as ‘rhododendrons, Vaccinium, and Quercus’ (Ulmer’s field notes, de Schauensee and Ripley 1939). Its lower elevational limit is unknown, but possibly around 2000 m since there is a noticeable change from moss forest to foothill jungle according to Miller (1942). Furthermore, the lower elevational limits of H. vorax might be constrained by H. maxi, which has been recorded between less than 100 and 1100 m a.s.l. in Aceh, and up to 2200 m a.s.l. on Mt. Kerinci. We suggest H. maxi and H. vorax have a parapatric distribution in the Gayo Highlands mirroring the elevational segregation of H. maxi and H. parvus on Mt. Kerinci (Ruedi et al. 1994). Additional sampling at Gayo Highlands between 1100 and 2000 m a.s.l. will be required to confirm this hypothesis. Hylomys vorax seems to show preference for moist microhabitats with high cover, since six over nine specimens with associated habitat field notes were caught in a mossy area next to a river. These specimens were caught either under mossy logs (N = 4), under tree roots in thick scrubby jungle (N = 2), in thick growth near blang (=meadow) (N = 1), mossy hole at base of tree (N = 1), or in grass in an open glade near forest (N = 1). Stomachs contained the chewed remains of earthworms (N = 2), and specimens were caught with ham rind or coconut bait (Ulmer’s field notes). A subadult male and two females were caught under the same log. The most likely scenario is that a mature female with teats (USNM 271033) and its offspring (subadults ANSP 20375 and 20378) were collected together, perhaps suggesting parental care and/or territory overlap with offspring at least until adulthood. Three adult specimens had a very strong porcine odour, but this was not mentioned for any of the subadults in Fred Ulmer’s field notes. It shares habitat with Maxomys hylomyoides (Robinson and Kloss 1916), Rattus hoogerwerfi (Chasen 1939), Sundamys infraluteus (Thomas 1888), and Sundasciurus altitudinis (Miller 1942).

Figure 11.

A, trapline set by Frederick A. Ulmer and other collectors of the George Vanderbilt Expedition. B, habitat of H. vorax in Blangbeke (2073 m a.s.l.), Aceh, Sumatra. C, habitat of H. vorax, view from the summit of Gunung Leuser looking eastward. D, paratype of Hylomys vorax (ANSP20377) collected on 12 April 1939 by Fred Ulmer in Blangbeke, Aceh, Sumatra, and illustrated by him (Supporting Information, Fig. S11). Picture credits: ANSP Archives Collection (unaccessioned Vertebrate Zoology 2012-01: George Vanderbilt Sumatra Expedition in 1939: photographs and images by Frederick A. Ulmer, Jr).

Conservation:

This species is found in Gunung Leuser National Park (GLNP). The Leuser ecosystem is designated as a ‘national strategic area’ (NSA) in Indonesia, to safeguard ecosystem services and limit infrastructure and agricultural expansion for ecological preservation. However, recent actions by the Aceh provincial government, which omitted Leuser’s NSA status and proposed conflicting development projects, have created uncertainty about its conservation status and the effectiveness of decentralized forest governance in Indonesia (Sloan et al. 2018). GLNP has lost c. 8000 ha of its primary forest from 2012 to 2017, possibly due to a combination of industrial oil palm plantations, dryland agriculture, and forest fires (Lubis et al. 2020, Condro et al. 2021, Dwiyahreni et al. 2021, Thoha et al. 2022). The Leuser Ecosystem is also threatened by habitat fragmentation due to unofficial road development and conversion of ‘sustainable managed’ buffer forests to palm oil plantations (Sloan et al. 2018). While most deforestation is taking place in lowland areas, it is suggested that the magnitude of climate change effects will be higher on the montane areas of the GLNP (Condro et al. 2021). Finally, it is anticipated that the Trans-Sumatra Highway’s completion, if constructed according to the current plan, would encourage increased anthropogenic pressure on the outskirts of the GLNP, potentially impacting its fragile ecosystem (Sloan et al. 2019). Species co-distributed with H. vorax, like Rattus hoogerwerfi and Presbytis thomasi, have similar or larger occurrence ranges and are listed as vulnerable in the IUCN Red List of Threatened Species. Further research is required to assess H. vorax population trends, explore its distribution beyond Mt. Leuser, and investigate whether H. maxi is displacing H. vorax due to climate-induced elevational changes.

Comments:

Five specimens (MZB 3172–5, RMNH 5137-8) with damaged skulls collected in Pajatoeng Kalang (south of Isaq) at 2000 m were clustered in the morphospace of H. vorax in some preliminary uni/bivariate analyses (MZB 3172–5: HB/E, GLS/IM3Sb, M1M1, IM3I; RMNH 5137-8: HB/E, P4M3) but in between H. maxi and H. vorax (MZB 3172–5: HB/T, GLS/BB; RMNH 5137-8: M1M1, IM3I) or in the range of H. maxi in others (RMNH 5137: T, PPL). Specimens RMNH 5137-8 skins and skulls were examined through pictures and have a bicoloured tail (H. maxi feature), but most cranial or mandible features were consistent with H. vorax. These were excluded from Table 1 and all bivariate plot figures since they were collected in a habitat transition and potentially parapatric (and perhaps even hybridization) area of H. maxi and H. vorax, and were not sequenced nor included in the craniodental PCA due to their damaged state. Lastly, diagnostic measurements of RMNH 5137-8 were not collected by the authors, but by Pepijn Kamminga (senior collection manager at RMNH) and could be subject to observer bias.

Hylomys parvus Robinson and Kloss, 1916

Hylomys parvus Robinson and Kloss, 1916: 269. Original description.

Holotype:

NHMUK 1919.11.5.12, an adult female (skin and skull) collected by H.C. Robinson and C.B. Kloss on 9 May 1914. Head and body length 105 mm, 25 tail mm (23.8% HB), hindfoot 23.5 mm, ear length (na), weight (na). Skull measurements in the Supporting Information, Table S2.

Type locality:

Korinchi Peak, 10 000 ft (=Gunung Kerinci, 3048 m), West Sumatra, Indonesia.

Paratypes (19):

Original description specifies that 20 specimens were examined but does not cite these. These are the specimens we could locate from the type series: NHMUK ZD 1919.11.5.10, RMNH23892, ZRC4.3427, ZRC4.3430, ZRC4.3431, ZRC4.3432, ZRC4.3434, and ZRC4.3935.

Emended diagnosis:

A small-sized Hylomys (average HB = 107 mm, GLS = 30.9 mm) characterized by thick and entirely soft fur, a very long and bicoloured black tail (average T = 23 mm; T/HB = 22.4%), and long ears (average E = 19 mm, E/GLS = 56.3%), supraorbital process absent, dentition is gracile, notch between premaxilla tips is acute-angled, frontal bone and dorsal region of maxillary are smooth, occipital crest minute, P4 cingulum and parastyle absent or inconspicuous, p3 with two non-fused roots.

Comparisons:

Comparisons with Sumatran H. maxi and H. vorax are detailed in these species accounts. It can be distinguished from all relatives by its smaller mandible (LMA < 22.2 mm vs. LMA > 22.4 mm), and by having two non-fused roots in p3 (Jenkins and Robinson 2002).

Distribution, habitat, and natural history:

Sumatran endemic, currently known from montane forest between 2200 and 3330 m a.s.l. on Mount Kerinci (Ruedi et al. 1994). This species has been recorded in sympatry with H. maxi at 2225 m by Robinson and Kloss (1918). It shares habitat with Crocidura beccari, C. lepidura, C. aequicauda, Maxomys hylomyoides, Mus crociduroides, Niviventer fraternus, Rattus korinchi, and Sundasciurus altitudinis (Robinson and Kloss 1918). Hylomys parvus might limit the upper distribution of Crocidura lepidura according to Ruedi (1995).

Conservation:

Recorded in Kerinci Seblat National Park (KSNP). Although the habitat of H. parvus is protected, deforestation is ongoing, since KSNP has lost c. 38 000 ha of its primary forest from 2012 to 2017 (Dwiyahreni et al. 2021). In addition, 12 current road proposals across Kerinci Seblat National Park are being evaluated by the Indonesian Government, which would contribute to extensive habitat fragmentation if approved (Sloan et al. 2019). It is anticipated that the Trans-Sumatra Highway’s completion, if constructed according to the current plan, would also encourage increased anthropogenic pressure on the outskirts of the KSNP, potentially impacting its fragile ecosystem (Sloan et al. 2019). This species was listed as critically endangered in 1996. However, in 2008, it was reclassified as vulnerable, a status that remains current on the IUCN Red List. This conservation status change seems unjustified since it is based on a miscitation of Ruedi et al. (1994) and overestimation of the extent of occurrence of H. parvus. The species was just recorded on Mt. Kerinci in Ruedi et al. (1994), not north and south of it, which reduces considerably its extent of occurrence. Habitat availability and quality has probably decreased from 1996 to 2023 due to climate change. Further research is needed to determine population trends of H. parvus, evaluate if its distribution spans additional mountains besides Mt. Kerinci, and whether H. maxi is outcompeting and displacing H. parvus due to climate change-driven elevational shifts (Ruedi et al. 1994).

Hylomys dorsalis Thomas, 1888 stat. nov.

Hylomys suillus dorsalis Thomas, 1888: 407. Original description.

Hylomys suillus dorsalisThomas, 1889: 229.

Holotype:

Not specified in original description. NHMUK 1895.10.4.2, adult male (skin and damaged skull), collected by John Whitehead in February 1888 was possibly designated as a lectotype by Thomas (1889). However, its field and museum number was not provided, just the date of collection and elevation, but this seemed to be enough to identify it by NHMUK staff. We thus clarify here that NHMUK 1895.10.4.2 should be treated as the lectotype of this species. Head and body length 116 mm, 16 tail mm (13.8% HB), hindfoot 25 mm, ear length (na), weight (na). Skull measurements in the Supporting Information, Table S2.

Type locality:

‘Mount Kina Balu, North Borneo’ (=Gunung Kinabalu, Sabah, Borneo, Malaysia; lectotype was collected at 2438 m a.s.l.).

Paratypes:

Non-existent.

Emended diagnosis:

A large-sized Hylomys (average HB = 135 mm, W = 61 g) characterized by a black dorsal stripe in adults, at least in the nape-shoulder area, which can sometimes be relatively inconspicuous. Dorsum and flank pelage coloration are homogenous. Adults have a ventral brown-grey coloration (at least from late April to late August). Forefoot with long nails. Hindfoot is large (average HF = 25.7 mm, HF + nail = 27.7 mm). Skull is characterized by a unique combination of long nasals (average NL = 14 mm) that generally extend to the level of the antorbital rim, broad interorbital constriction (average IOB = 9.5 mm), and long but narrow rostrum (average ROL = 17.3 mm, ROB = 5 mm). I1 has a prominent posterior cuspule. I2 is generally posterolabially oriented. P1 and P3 are generally similar in size. Vestigial cingulum and small but frequently sharp parastyle in P4.

Comparisons:

Only Hylomys species present in Borneo. Distinguished from all other congeneric species by black dorsal stripe in the shoulder area in adults, and long nasals and broad interorbital constriction (Supporting Information, File S1, Fig. S12). Posterior end of nasals generally extends to the level of the antorbital rim (26/36 specimens examined), while it does not in the other Sundaic Hylomys. Forefoot nails are longer than H. maxi, H. vorax, and, generally, than H. suillus. Supraorbital processes are more developed and prominent than in H. suillus, H. vorax, and H. parvus, but less robust than in H. maxi. I1 is less procumbent than H. vorax and H. suillus. I1 posterior cuspule is more prominent than H. suillus, H. vorax, and H. parvus. Braincase height at the occipitum (9.6–10.1 mm) is greater than H. parvus (8.6 mm), H. vorax (8.9–9.5 mm), and H. suillus (8.8–9.8 mm). Skull is narrower (ROB = 4.6–5.5 mm, M1M1 = 8.2–11.3 mm) and molars are smaller (P4M3 = 7.7–8.6 mm) than H. maxi (ROB = 5.2–6.4 mm, M1M1 = 10.9–12.5 mm, P4M3 = 8.3–9.9 mm).

Distribution, habitat, and natural history:

Bornean endemic distributed from 1280 to 3413 m a.s.l. on Mt. Kinabalu (Lim and Heyneman 1968, Camacho-Sanchez et al. 2019) and 1510–2620 m a.s.l. on Mt. Trus Madi (authors’ unpublished data). It has also been recorded at c. 2050 m a.s.l. on Mt. Tambuyukon (Camacho-Sanchez et al. 2019), 1800–1950 m a.s.l. on Mt. Alab (authors’ unpublished data), c. 2100 m a.s.l. on Mt. Murud (Wiantoro et al. 2009), at an unknown elevation on Mt. Mulu (Cranbrook 1982), and 1000 m a.s.l. in Bario, Kelabit Highlands (MZUM-M 891). The species has been recorded in different habitats such as grass at the edge of forest, between rocks and logs next to a stream, under roots or fallen trees in oak mossy forest, in bamboo patches, grassy banks and flowerbeds around buildings, and inside an abandoned resthouse (Phillipps 2016, authors’ unpublished data). In Kinabalu and Crocker Range Nationals Parks and Trus Madi Forest Reserve this species shares its habitat with Crocidura baluensis/Crocidura foetida s.l., Crocidura cf. neglecta, Chimarrogale phaeura, Palawanosorex ater, Tupaia montana, Leopoldamys sabanus, Maxomys alticola, Maxomys baeodon, Maxomys ochraceiventer, Niviventer rapit, Rattus baluensis, Sundamys infraluteus, Sundasciurus everetti, and Melogale everetti (Camacho-Sanchez et al. 2019, Hinckley et al. 2022, authors’ unpublished data).

Four specimens collected in Kinabalu had remains of caterpillars, centipedes, beetles, and earthworms in their stomachs (Lim Boo Liat and Heyneman 1968). Two specimens collected in Trus Madi contained the remains of soft-bodied insects, caterpillars, and beetle larvae (Harrison 1954). Specimens also ate tapioca and banana bait (authors’ unpublished data). Lactating females have been recorded in late-August (N = 3) and December (N = 1) (Lim Boo Liat and Heyneman 1968, this study). A gravid female with two embryos was recorded in mid-July (USNM 292343). In March, a mature male and female were collected in the same trap (BOR309 and 312). Specimens (N = 11) collected in the same transects (N = 4) in Kinabalu and Crocker Range Parks were on average 155 m from each other (range: 90–220 m). Transects had never more than one adult male but could have more than one female or juvenile, suggesting parental care and/or territory overlap until adulthood or among females. Specimens were recorded in the morning (8:15 and 10 a.m.), afternoon (caught between noon and 7 p.m.) and at dawn, and were possibly also caught at night (authors’ unpublished data). This species makes ‘a sheltered nest out of leaves under rocks or logs’ (Phillipps 2016).

Conservation:

Recorded in Kinabalu, Crocker Range, Pulong Tau, and Mulu National Parks, and in the Trusmadi Forest Reserve (Cranbrook 1982, Wiantoro et al. 2009, this study). The known distribution of H. dorsalis is mostly within protected areas, but further research is needed to determine population trends within these areas. Likewise, the combination of habitat suitability modelling and additional field surveys in Kalimantan are essential to re-assess the conservation status of H. dorsalis. Even though this species is distributed across most of northern Borneo, it is restricted to hill and montane forests over 1000 m a.s.l., which will probably decrease in area because of climate change.

Hylomys suillus Müller, 1841

Hylomys suillus Müller, 1841: 25, 50. Original description.

Holotype:

Not specified in original description.

Syntypes:

RMNH39016 (Jentink 1887: 243a; Jentink 1888: 120a), adult male, relaxed mount and cranium with damaged zygomata and occipital collected by S. Müller in May or July/August 1834 in Batang Singgalang at c. 2000 ft (650 m), Sumatra, Indonesia. RMNH39017, adult female relaxed mount with in situ skull collected by S. Müller on 23 September 1831 in Kaliki, Mt. Gede at c. 1200 ft (400 m), Java, Indonesia (Smeenk et al. unpublished work, P. Kamminga personal communication).

Lectotype:

Sody (1933) assumed that the specimen illustrated by Schlegel and Müller (1843) was from Java, and proposed to restrict the type locality to Java, but did not designate a lectotype. We here designate as lectotype the syntype RMNH 39017.

Type locality:

‘Java en het andere van Sumatra’. Restricted by Sody (1933) to ‘Java’ Indonesia.

Paratypes:

Non-existent.

Emended diagnosis:

A medium sized Hylomys (average HB = 129.5 mm, GLS = 34.8 mm) with an average-sized, bicoloured tail (average T = 19 mm) and a venter coloration that transitions from brown-grey (between May and November) to dark grey (December to April). Skull is characterized by the following features: acute-angled notch between premaxilla tips; posterior end of nasals does not generally extend to the level of the antorbital rim; antorbital ridge absent/vestigial; lacrimal flange poorly developed; supraorbital processes generally absent, rarely developed; I1 procumbent; P4 has a generally straight, sometimes slightly sinuous, cingulum; P4 has a prominent and sharp parastyle; posterior nasal spine present but sometimes poorly developed; occipital crest absent or vestigial in females and absent to poorly developed in males.

Comparisons:

Only Hylomys species present in Java. It has a shorter upper toothrow length than H. maxi (IM3Sa = 16–17.9 vs. 18.3–19.7 mm; Sody 1933), an acute vs. obtuse-angled notch between premaxilla tips in H. maxi, antorbital ridge, lacrimal flange and occipital crest absent/poorly developed vs. well developed in H. maxi, generally straight and thin vs. sinuous and thick P4 cingulum in H. maxi. It can be differentiated from H. vorax based on its bicoloured vs. monocoloured tail, shorter hindfoot (HF + nail = 23–26 vs. 25.5–27.5 mm), and relatively broader rostrum (Fig. 6). It has a larger braincase (BB = 13.8–15.6 vs. 13.1–14 mm, BD = 8.8.–9.8 vs. 8.6 mm, MAB = 13.2–14.5 vs. 12.2 mm) and mandible length (LMA = 22.4–26.8 vs. 20.1–22.2 mm) than H. parvus. See H. dorsalis species account for a comparison with this species.

Distribution, habitat, and natural history:

Javan endemic distributed from sea level (recorded in Jakarta) to 2200 m a.s.l, although it has also been reported to be found exclusively in the montane forests of some isolated volcanoes (Bartels 1937), perhaps because most lowland forests have been logged in Java. Collected in forest, edge of forest, and inside a house in Cibodas (USNM specimens NAMRU). In Cibodas, West Java, it shares its habitat with Crocidura brunnea, C. monticola, C. orientalis, Maxomys bartelsi, Niviventer cremoriventer, Niviventer sp., Rattus argentiventer, Rattus exulans, Rattus tanezumi, and Melogale orientalis (Ruedi 1995, USNM catalogue).

Conservation:

Recorded in Gunung Gede-Pangrango and Mount Halimun Salak National Parks and Kawah Ijen Nature Recreation Park. While H. suillus may not face habitat loss in protected areas of West Java (Higginbottom et al. 2019), it is important to note that most of these areas primarily consist of montane forests, which are susceptible to the impacts of climate change. Thus, further research is needed to determine population trends within protected areas. Likewise, the use of habitat suitability modelling is essential to re-assess the conservation status of H. suillus, given that, even though this species is found across most of Java, its distribution is very fragmented (Ruedi et al. 1994). In essence, the species actual current and future area of occupancy may be much smaller than its overall extent of occurrence. The designation of Gunung Slamet and Gunung Lawu as national parks would greatly benefit this species. These mountains have the second and third largest elevational ranges on Java, and their long-term protection might aid this and other montane species in tracking climate change, but also contribute to east–west population connectivity if forest restoration efforts are promoted.

Comments:

This species exhibits a high degree of morphological variation, spanning about twice the morphospace compared to its relatives in the craniodental PCA (Fig. 5). This variation seems somewhat geographically structured (east and west Java), although there are several outliers that break this pattern. Future studies should examine species limits within H. suillus in the light of molecular evidence and additional geographical sampling.

DISCUSSION

Here we review the taxonomy of the genus Hylomys, with the description of two new species and elevation of three subspecies to specific status. This research has increased the diversity of Hylomys from two to seven extant species (Ruedi et al. 1994, Bannikova et al. 2014). Nuclear and morphological evidence support the specific status of Hylomys macarong, the highly divergent southern Viet Nam matrilineage identified by Bannikova et al. (2014). Mitochondrial and morphological evidence, and a broader geographic sampling, has led to the description as a new species of the higher elevation specimens identified as Hylomys suillus maxi in Miller (1942). Nuclear evidence and a denser taxon sampling support some of the taxonomic changes informally used by Abd Wahab et al. (2022), such as the specific status of H. maxi and H. dorsalis, but not others, such as the species-level recognition of H. siamensis and H. microtinus, which are here considered as subspecies of H. peguensis (discussed below). The names and concepts ‘peguensis’ and ‘tionis’ were considered subspecies of H. suillus in Abd Wahab et al. (2022) but the former is here recognized as a distinct species (H. peguensis) and the latter is considered a synonym of H. maxi.

All five newly recognized species are supported by three independent lines of evidence—morphological, genetic, and biogeographical (contemporary marine barriers between H. suillus, H. dorsalis, and H. vorax/H. parvus; partial contemporary riverine barriers between H. peguensis and H. macarong) or ecological (habitat differentiation in H. vorax/H. parvus–H. maxi, and H. peguensis–H. maxi). Thus, these species merit a ‘green’ category, the greatest degree of taxonomic certainty in the traffic-light system, described by Kitchener et al. (2022). However, the amount and strength of evidence supporting the recognition of each new species varies. The species H. macarong and H. vorax lacked associated fresh tissue samples, so molecular evidence just relied on mitochondrial genomes and a short (165 bp) nuclear gene fragment (GHR10), which was shown to be less informative than other nuclear markers (ApoB, PCSK2, and POU2F2). In fact, while sister-species H. macarong and H. peguensis show mutual allelic exclusivity in GHR10, the former has the same haplotype as the more distantly related H. suillus and H. parvus. Similarly, Sumatran highland H. vorax and H. parvus also show mutual allelic exclusivity, but the former shares its haplotype with Sumatran lowland and Malay Peninsula H. maxi. This study also lacks geographic sampling for intermediate sites (east Cambodia, central Viet Nam, and south Lao P.D.R.) between the known range of sister-species H. peguensis and H. macarong, which slightly overlap in craniodental morphospace but exhibit qualitative differences. With that said, multiple pieces of evidence support these species hypotheses: (i) raw (uncorrected) Cytb sequence divergences among these sister-species are much greater (>12.5%) than that among other gymnures recently described based on mitochondrial and morphological evidence (P. aureospinula and P. intermedia: <5%; Balete et al. 2023); (ii) mitochondrial genome divergence dating supports ancient speciation events (c. 4.3–6.6 Mya; Fig. 2C); (iii) species limits are consistent among mitochondrial and nuclear loci evidence for those species with five-loci data, leading us to believe mitochondrial data represents an informative marker in this genus; (iv) H. maxi–H. vorax show high mitochondrial, phenotypic, and ecological divergence despite lack of physical barriers among them. Therefore, available evidence predominantly supports these species hypotheses, but these should be re-evaluated with additional nuclear data and geographic coverage in the future.

The implications of our taxonomic findings are many. Heaney and Morgan (1982), Jenkins and Robinson (2002), Balete et al. (2023), and this study have increased the number of gymnure species (Galericinae) from six to 15, challenging the traditional view that considered extant gymnures a species-depleted subfamily in comparison to relatively species-rich spiny hedgehogs (Erinaceinae: 18 species). Furthermore, species diversity may still be underestimated in Hylomys and Neotetracus sinensis, and, consequently in Galericinae (Zeng et al. 2022, this study). Thus, gymnures exhibit similar levels of species diversity to spiny hedgehogs, despite the much larger distribution of the latter. This probably remained unnoticed due to the cryptic nature and smaller distribution ranges of gymnures, and disparate museum material. The advent of molecular systematics and new field campaigns have ultimately narrowed the species diversity gap among these subfamilies, shifting macroevolutionary patterns across the Erinaceidae. This scenario mirrors that of colugos (Dermoptera), a morphologically conserved order with a similar relictual distribution and a highly underestimated species diversity (Mason et al. 2016). This taxonomic revision will also have important conservation implications due to the more restricted distributional ranges of most Hylomys species, particularly H. parvus, H. vorax, and H. macarong, which are thought to represent narrow-range endemics that could be threatened by habitat loss and/or climate change. Thus, this study’s findings, conservation and distributional data will be pivotal for future IUCN Red List assessments. These three endemics could represent optimal umbrella species, given that protecting their habitat (highlands of the Leuser Ecosystem and Mount Kerinci and forests of southern Viet Nam) would contribute to conserving the habitat of many other narrow-range endemics (cited in the latter part of the Discussion) and several endangered species, such as the Sumatran tiger, Sumatran rhinoceros, Malayan tapir, black-shanked douc, Annamese langur, and yellow-cheeked gibbon. Protecting Leuser and Kerinci Seblat National Parks from road development and deforestation will not only provide critical habitat for these species, but will also benefit around 6–7 million people through essential ecosystem services, such as water supply, flood prevention, and tourism (Van Beukering et al. 2003, Lubis et al. 2020). Furthermore, the total economic value of the services provided by an intact Leuser Ecosystem is predicted to be greater than that generated from logging and agriculture over a 30-year period (Van Beukering et al. 2003).

The new species, H. vorax and H. macarong, had been accessioned in natural history collections for 84 and 62 years, respectively, prior to identification. This is the case for approximately three-quarters of newly described mammal species, which are based on specimens that had been previously stored in natural history collections, and later identified as new taxa (Kemp 2015). This is not surprising because the whole purpose of a systematic collection is to bring a variety of specimens together in one place for easy comparison. As this study emphasizes, specimens often remain in collections for many decades before receiving careful study, highlighting their under-utilization. Oftentimes, museums prioritize digitization efforts over taxonomic revisions, but providing accurate taxonomy is an important first step for which we will need more properly trained taxonomists (Kemp 2015, Engel et al. 2021). Hence, scientific collections are an important mine of undescribed species, particularly with the advent of the ‘Museum Genomics’ revolution, which now enables cost-effective gene sampling from museum specimens of elusive taxa from poorly sampled and remote regions of the globe (Fontaine et al. 2012, Kemp 2015, Hawkins et al. 2022). Species descriptions are critical to the field of evolutionary biology and, ultimately, improve our understanding of spatial patterns of species richness and endemicity, guiding stakeholders in the design of areas of conservation priority. Nevertheless, as Voss (2022) pointed out ‘the proper goal of taxonomy is not, however, simply to describe new taxa, nor to revise existing ones, but to achieve increasingly accurate and stable classifications that can serve as frameworks for future research on a wide variety of topics’.

The species H. peguensis and H. maxi, belonging to the Indochinese and Sundaic clades, respectively, seem to show a turnover around the K-P (5°N–6.5°N; Fig. 1). This transition is 500 km south of the Isthmus of Kra, at the border of two climate and vegetation zones (seasonal evergreen/monsoon forest and evergreen tropical forest) and represents one of the two major mammal species turnovers between Indochina and Sundaland (Hinckley et al. 2023). In H. maxi, low divergence levels between the mainland and Tioman Island (former H. s. tionis) are consistent with other phylogenetic studies and paleogeographic evidence, which support a recent geological connectivity across these satellite islands and the mainland (Mason et al. 2019, Husson et al. 2020, Hinckley et al. 2022, 2023). Divergence across the Strait of Malacca is moderate and greater than that shown among rat populations (i.e. Sundamys), suggesting that a taxonomic revision of these populations with additional geographical sampling is warranted (Camacho-Sanchez 2017). The species H. peguensis exhibited the greatest levels of genetic structure, with nominotypical populations (west of the Chao Phraya River basin) being sister to H. p. siamensis and H. p. microtinus (east of the Chao Phraya River basin). However, sequence-level divergence is in the ‘grey zone’ of speciation, the craniodental morphospace of these two clades overlaps, and sample size is low, so further geographical coverage and nuclear loci sampling will also be required to re-evaluate the taxonomic status of these subspecies. Finally, Hylomys species-level diversity might also be currently underestimated in Borneo and Sumatra. In Borneo, this research only sampled populations from the Crocker Range and unsuccessfully attempted to sequence a historic specimen from the Kelabit Highlands (MZUM(M)-891), which is relatively isolated from the former. Co-distributed small mammals’ populations from the Crocker Range and Kelabit-Mulu mountains have been supported as heterospecific in some taxa [Crocidura foetida s.l. (Peters, 1870); Hinckley et al. 2022] and conspecific in others [Palawanosorex ater (Medway, 1965) and Sundamys infraluteus (Thomas, 1888); Camacho-Sanchez 2017, Nations et al. 2022]. Furthermore, it also seems likely that this species distribution spans these mountain ranges. The species went unrecorded in the well-sampled Crocker Range NP until 2016, implying that it can be overlooked in understudied areas like Kalimantan’s mountains, such as the highly isolated Meratus Range, which hosts recently described and undescribed endemic birds, reptiles, and fishes in its hill and/or montane forests and along its south- or east-draining rivers (Harvey et al. 2019, Irham et al. 2022, Parenti et al. 2023). Similarly, this research has just sampled the ‘north’ and ‘central’ biogeographic montane zones of Sumatra described in Shaney et al. (2020), each of which is home to a high-elevation endemic (H. vorax and H. parvus). Additional surveys in Sumatra’s ‘north-central’ and ‘south’ biogeographic montane zones might reveal new species of Hylomys, as recently shown in other small vertebrates (Sarker et al. 2019, Shaney et al. 2020).

Divergence dating indicates that speciation events within Hylomys occurred during the Late Miocene and Early Pliocene, from c. 7.8 to 4.3 Mya. Speciation might have been driven by shifting climate conditions, such as the Late Miocene strengthening of the Indian monsoon and the Late Neogene global cooling and drying, when seasonally dry conditions expanded (Morley 2018). These climate conditions might have promoted diversification through several alternative mechanisms: (i) montane forest expansion and connectivity, and dispersal of Hylomys if MRCAs were adapted to montane forests; (ii) vicariance in allopatric habitat refugia in or around mountains during periods of C4 grassland expansion, if MRCAs were adapted to lowland or hill forest (Sheldon et al. 2015, Hinckley et al. 2022); and (iii) adaptation to different habitats following the Indian monsoon strengthening (e.g. MRCA of Sundaic clade to evergreen tropical/hill/montane forest, MRCA of Indochinese clade to seasonal evergreen/monsoon forest; Morley 2018). Sundaic and Indochinese Hylomys diverged anciently (c. 7.8 Mya), and greatly predate the Early Pleistocene divergence of other small mammals distributed across the Isthmus of Kra biogeographic transition (Hinckley et al. 2022, 2023). Diversification in Sundaland occurred rapidly (c. 6.6-4.3 Mya) and when the Sunda shelf was exposed, supporting a habitat-driven diversification scenario (Husson et al. 2020, Hinckley et al. 2022, 2023). Finally, geological events, such as Late Miocene mountain uplift in Sundaland (Mason et al. 2019) and riverine vicariance in Indochina (Chao Phraya River Basin/Mekong River; Klabacka et al. 2020), might have also played an important role in the diversification of Hylomys.

The discovery of Hylomys vorax brings the number of endemic mammals from northern Sumatra (north of lake Toba) to seven or eight, including one flying squirrel [Hylopetes winstoni (Sody, 1949)], one rat [Rattus hoogerwerfi (Chasen, 1939)], three primates [Pongo abelli (Lesson, 1827), Presbytis thomasi (Collett, 1893), and Nycticebus hilleri (Stone and Rehn, 1902)], and one bat [Mormopterus doriae (Andersen 1907)]. The species Rattus blangorum (Miller, 1942), is also endemic to northern Sumatra (described from the middle slopes of Mount Leuser) but was included in the widespread Rattus tiomanicus (Miller 1900) complex in Musser and Califia (1982) and recognized as a separate species by Carleton and Musser (2005) and Denys et al. (2017). Attempts to sequence this species have failed (Camacho-Sanchez and Leonard 2020). These endemics have been recorded at middle and high elevations (1200 m a.s.l.: H. winstoni; 1500–3400 m a.s.l.: P. thomasi; 2134–2835 m a.s.l.: H. vorax and R. hoogerwerfi), but also spanning lowlands (P. abelli, N. hilleri, and M. doriae). Mammal species spanning lowlands exhibit both, limited [Sundamys muelleri, Ratufa affinis (Raffles 1821)] and high [Galeopterus variegatus (Audebert 1799)] genetic structure across Sumatra (Mason et al. 2016, Camacho-Sanchez 2017, authors’ unpublished work). Similarly, highland populations or species can be highly divergent (Hylomys, Rattus, Crocidura, and Sundasciurus altitudinis) or poorly differentiated across this landmass (Sundamys infraluteus; Camacho Sánchez 2017, Camacho-Sanchez and Leonard 2020, Hinckley et al. 2020, 2022). Such discordant patterns among co-distributed taxa are possibly the result of different life-history traits and dispersal capabilities, which ultimately contribute to variable degrees of population connectivity among habitat refugia during climatic oscillations.

This emerging pattern of localized endemicity in northern Sumatra has also been recently recorded in squirrels (Hinckley et al. 2020), teleosts (Lumbantobing 2010, 2014), amphibians (Wostl et al. 2017, Arifin et al. 2018, 2022, O’Connell et al. 2018, Sarker et al. 2019), and reptiles (Iskandar et al. 2017, Shaney et al. 2020). Levels of endemicity still probably remain an underestimation given that Sumatran vertebrates, particularly small mammals, are under-represented in museum collections and, thereby, understudied (Arifin et al. 2022). The combination of museum genomics with additional small mammal sampling across the Barisan Range will surely contribute to new species discoveries, particularly in the understudied northern, north-central, and southern montane zones (described in: Shaney et al. 2020).

Hylomys macarong adds to the many South Vietnamese endemic mammals, including shrews [Chimarrogale varennei (Thomas, 1927), Crocidura phanluongi (Jenkins et al., 2010), and Crocidura phuquocensis (Abramov et al., 2008)], moles [Euroscaptor parvidens (Miller, 1940)], primates [Pygathrix nigripes (Milne-Edwards, 1871) and Trachypithecus margarita (Elliot, 1909)], bats [Myotis phanluongi (Borisenko et al., 2008) and Murina harpioloides (Kruskop and Eger, 2008)], and rodents [Callosciurus honkhoaiensis (Nguyen et al. ,2018) and Leopoldamys milleti (Robinson and Kloss 1922)]. Most of these species are restricted to the Southern Annamites’ montane rain forest (Langbian/Da Lat and Dak Lak plateaus) considered an important centre of local endemism (Rundel 1999, Tordoff 2002, Bain and Hurley 2011, Abramov et al. 2017, Poyarkov et al. 2021). However, some of these species, including Hylomys macarong, are also found in hilly and lowland adjacent areas to these ranges, mirroring the distribution pattern of many reptiles (Poyarkov et al. 2023).

The presence of short, internal branches, potential incomplete lineage sorting, and gene tree discordance, suggests a phylogenomic coalescent-based approach will be required to infer the correct species tree and to reconstruct the historical biogeography of this genus with an improved resolution (Liu et al. 2015). Such research would also benefit from environmental niche modeling to evaluate the role of niche divergence in the speciation of Hylomys, given that some species live at high elevations (H. vorax and H. parvus), others live at high to mid-elevations (H. dorsalis, H. suillus, and H. macarong), and others have wide elevation ranges that span lowlands (H. peguensis and H. maxi). Within Sumatra, morphological evidence and stomach content data could suggest some degree of divergent adaptation among the elevational parapatric sister-pair, H. maxi and H. vorax. The species H. maxi has a more robust dentition and mandible, with relatively larger molars, perhaps adapted to harder food items, and its stomachs have been reported to contain insects of various orders, principally small beetles, and crickets with hard cuticles. On the other hand, H. vorax has a more gracile dentition and mandible, perhaps adapted to softer food items, and its stomachs have been reported to contain earthworms. Nevertheless, this hypothesis is based on just a few observations and should be statistically tested with additional diet data.

CONCLUSION

In this paper we address the systematics of Hylomys through an integrative approach combining Next Generation Museum Genomics and traditional craniodental morphometrics. We present evidence supporting the recognition of two new species, which we describe as H. macarong and H. vorax, and the elevation of H. dorsalis, H. maxi, and H. peguensis from subspecific to specific status. Our findings support northern Sumatra and the Southern Annamites as centres of localized endemicity, and suggest the need for additional small mammal surveys across Sumatra’s Barisan Range.

urn:lsid:zoobank.org:pub:780D1924-E984-4900-88E5-85FE2C5688D8.

ACKNOWLEDGEMENTS

We dedicate this research to the late Frederick A. Ulmer, Lim Boo Liat, and François Catzeflis, who collected important specimens/data that benefited this study, and greatly contributed towards a better understanding of Tropical East Asia’s mammalian diversity. We thank the two anonymous reviewers for their valuable and constructive feedback, which significantly enhanced our manuscript. We thank the following museums and their curators and collection managers for loan of materials, access to specimens, archives, and pictures for this study: EBD (Carlos Urdiales, Manolo López, and Carlos Ruiz), MVZ (Chris Conroy and Jim Patton), MZB, MCZ (Mark Omura), USNM, LKC (Kelvin Lim), MZUM, NHMUK (Roberto Portela), and THNHM (Tadsanai Jeenthong, Wachara Sanguansombat, Cholawit Thongcharoenchaikit, and Chainupon Suwannakunpaisan). We would like to extend additional gratitude for the immense efforts of Ned Gilmore (ANSP), and Pepijn Kamminga (RMNH) for their contributions to this study. We would also like to thank: different collectors listed in the Supporting Information, Table S1 who made available the specimens of study; Mary Faith Flores helped with laboratory work; Katie Sayers took pictures of type specimens; Rose Ragai, Ipe, Manolo López, Paco Carro, Razzak Intang, and Cristopher Baidis participated in field campaigns; Fred Tuh Yit Yu and Linus Gonsilou provided logistic support; Farhan Abd Wahab, Uttam Saikia, and Edson Fiedler Abreu provided information/pictures of MZUM, Zoological Survey of India National Zoological Collection, and AMNH specimens; Lê Tấn Quy helped finding the name ‘ma cà rồng’ for Hylomys macarong and translated to Vietnamese the common name of this species; Daniel Lumbantobing shared insights regarding biogeography and conservation in Sundaland; Megan Viera and Ingrid Rochon helped finding and/or ‘translating’ some collector notes; Smithsonian NMNH (Richard Greene and Sue Zwicker) and Biodiversity Heritage libraries provided access to many important historical papers/books; Sabah Parks, Sabah Forestry Department and Sabah Biodiversity Council provided research permits and logistic support; ANSP granted access to photographs and images of the George Vanderbilt Sumatra Expedition in 1939, which were supplied by Dan Thomas. Logistical support was provided by Laboratorio de Ecología Molecular (LEM-EBD), Doñana ICTS-RBD, and the Laboratories of Analytical Biology of the Smithsonian NMNH.

CREDIT STATEMENT

All authors contributed intellectually; A.H., M.C., M.R., and M.H. collected the material in the field; A.H. and M.H. collected the molecular data; A.H., M.R., and M.H. collected the morphological data; A.H. and M.C. analysed the molecular data; A.H. analysed the morphological data; A.H. drafted the paper with contributions from all authors.

CONFLICT OF INTEREST

The authors declare that they have no conflicts of interest in relation to this work.

FUNDING

The Spanish Ministry of Science and Innovation grants CGL2014-58793- P and CGL2017- 86068- P and PID2020- 120115GB- 100 to J.A.L., and Smithsonian Institution’s discretionary startup funds to M.H. supported this study. A.H. was supported by a Ministerio de Economía y Competitividad contract CGL2014-58793- P and a ‘Margarita Salas’ postdoctoral grant funded by the Ministerio de Universidades de España and the European Union ‘NextGenerationEU’. A.H. was also supported by an Ernst Mayr Travel grant.

DATA AVAILABILITY

DNA sequences obtained in this research are deposited in GenBank under the accession numbers: OR138079–OR138103 (mitochondrial genomes) and OR554392–OR554528 (nuclear loci). Code for mitochondrial assembly and phylogenetic analysis is available at: https://github.com/csmiguel/hylomys_mitogenomes/. 3D surface scans of type specimen craniums and mandibles are available at MorphoSource: https://doi.org/10.17602/M2/M560011https://doi.org/10.17602/M2/M560006, https://doi.org/10.17602/M2/M560031, https://doi.org/10.17602/M2/M560027, https://doi.org/10.17602/M2/M560022, https://doi.org/10.17602/M2/M560018.