We surveyed small mammals in two locations in Sarawak, Malaysia (Fig. 1): Mount Mulu (in 2017) and Mount Murud (in 2019). Both surveys used pitfall lines, which typically consisted of 5–10 large, 20–30 L buckets with a tarp drift fence, though occasionally we added smaller 1–3 L buckets. On Mount Mulu we set two pitfall lines of large buckets at 1650 m and one pitfall line of large buckets at 1800 m for a total of ca. 300 pitfall-nights. On Mount Murud we set pitfall lines (mixed large and small buckets) at 1480 m, 1660 m, 1770 m, 2000 m, 2250 m, and 2400 m for a total of ca. 440 pitfall-nights. Specimens were deposited at the Louisiana State University Museum of Natural Science, Baton Rouge, USA (LSUMZ). Specimens were measured, weighed, and then preserved in formalin (often with the skull removed and cleaned) or prepared as dried museum skins with dried and cleaned skeletons. Tissues were preserved in 95% ethanol. All collecting methods followed the recommended guidelines of the American Society of Mammalogists (Sikes et al. 2016).

Using a phenol-chloroform extraction protocol (Tsai et al. 2020), we extracted DNA from skin clips of the holotype of S. ater and a species from an outgroup genus (Solisorex pearsoni Thomas, 1924), as well as fresh tissues from three recently collected specimens from Sarawak that are morphologically aligned with S. ater (LSUMZ 40511, 40514, 40522). We modified the phenol-chloroform extraction of Tsai et al. (2020) by using a refrigerated centrifuge set at 3 °C during the ethanol precipitation steps. In addition to the phenol-chloroform extraction, we extracted genomic DNA from tissue samples (liver) from six Sarawak shrew specimens morphologically similar to S. ater (LSUMZ 40511, 40514, 40522, 40695, 40696, 40697) using Qiagen DNEasy Blood & Tissue kits (Qiagen, Germantown, Maryland) following the manufacturer’s instructions.

Phenol-chloroform extractions from skin clip samples (S. ater holotype and S. pearsoni) were treated with a New England BioLabs (Ipswich, MA) PreCR Repair Kit following the manufacturer’s instructions to repair preservation-related damage. For the phenol-chloroform extractions from the three fresh samples, we mechanically sheared the DNA to a 400–600 bp size range using an Epigentek Episonic sonicator. We prepared genomic libraries for all five phenol-chloroform extractions with a KAPA Hyper Prep kit and dual indexed iTru adapters (Glenn et al. 2019) following Esselstyn et al. (2017). Because the S. ater holotype extraction had a very small amount of DNA, we used diluted index primers at 1.25 μM (a quarter of the standard molarity) for this specimen. We then pooled the libraries and enriched them for the standard Tetrapods 5k loci (Faircloth et al. 2012) and 27 exons using the probe set introduced by Esselstyn et al. (2021) and manufactured by Arbor Biosciences (Ann Arbor, MI). We removed short fragments less than 150 base pairs from the enriched pools with a QIAGEN GeneRead Size Selection kit and confirmed the absence of adapter dimers with an Agilent Bioanalyzer using a DNA-High Sensitivity Kit. We then combined the enriched libraries into an equimolar pool with an unenriched library from the S. ater holotype specimen in order to enhance the likelihood of sequencing mitochondrial fragments from this specimen. Novogene (Beijing, China) sequenced these libraries on an Illumina HiSeq 4000 PE 150 lane (Illumina Inc., San Diego, CA, USA). For the six fresh tissue Qiagen DNEasy extractions, we amplified the mitochondrial protein coding gene cytochrome b [CYTB] using Polymerase Chain Reaction (PCR) following the protocol described in Esselstyn et al. (2009).

We processed the UCEs in PHYLUCE v. 1.7.1 (Faircloth 2016), following Tutorial III guidelines. We processed raw Illumina reads with illumiprocessor (Faircloth 2013) and assembled trimmed reads into contigs using both Trinity v. r2013.08.14 (Grabherr et al. 2011) and SPAdes v. 3.14.1 (Prjibelski et al. 2020); we chose the assembler run that resulted in the largest number of UCE loci for final analyses. Contigs matching the UCE probes were aligned in MAFFT v. 7.475 (Katoh and Standley 2013) and edge-trimmed in PHYLUCE using trimmomatic (Bolger et al. 2014). Unfortunately, we did not obtain UCEs from the S. ater holotype. To obtain mitochondrial bycatch from the UCE data for three samples (the S. ater holotype, LSUMZ 40514, 40522) we also assembled reads into contigs using both metaSPAdes v. 3.14.1 (Nurk et al. 2017) and MEGAHIT (Li et al. 2016) as part of the MitoFinder pipeline (Allio et al. 2020). We then searched through the assembled contigs using several different mitochondrial genomes as references to improve our recovery of the mitochondrial data: Suncus murinus Linnaeus, 1766, Crocidura attenuate Milne-Edwards, 1872, C. dongyangjiangensis Liu Yang et al., 2020, C. lasiura Dobson, 1890, C. russula Hermann, 1780, C. grayi (Dobson, 1890), C. fuliginosa Blyth, 1855, and C. beata Miller, 1910 (GenBank: NC_024604.1, KP120863.2, NC_056167.1, KR007669.1, NC_056768.1, KR537885.1, C_042762.1, KR537889.1). The resulting contigs were combined with the original Trinity or SPAdes contigs from the UCE assemblies and then aligned and annotated for 15 mitochondrial genes using the Crocidura shantungensis Miller, 1901 mitochondrial genome (GenBank: OM038325) as a reference in Geneious. Finally, we downloaded previously published mitochondrial and UCE sequences from other outgroups, mostly matching the taxon sampling in Hutterer et al. (2018).

We assembled an alignment of ten mitochondrial genes using a combination of newly generated sequences and sequences from GenBank (Suppl. material 1). For several outgroup species (Crocidura palawanensis Taylor, 1934, C. russula, C. sibirica Dukelski, 1930, and Suncus murinus), gene sequences were pulled from whole mitochondrial genomes that were available via GenBank. We included CYTB sequences from the recently collected Bornean material, as well as CYTB and 16s rRNA sequences from Suncus dayi to test the relationship proposed by Medway (1965). We partitioned the dataset by gene and, for all but 16s rRNA, by codon position (28 data subsets). We tested for the best partitioning scheme and best fitting model using ModelFinder (Kalyaanamoorthy et al. 2017) in IQ-TREE (Minh et al. 2020). We then conducted a maximum-likelihood (ML) phylogenetic analysis in IQ-TREE, assessing nodal support with 1000 ultrafast bootstrap replicates. In order to verify our results with a lower percentage of missing data, we constructed a second alignment of nine mitochondrial genes by removing eight samples that were only represented with CYTB. We repeated the same IQ-TREE analysis on this second mitochondrial DNA alignment of 20 samples. We generated CYTB p-distances for each sample in the first 28 specimen alignment, and the p-distances of nine mitochondrial genes from the second 20 specimen alignment.

We analyzed the 19-specimen UCE dataset using two different approaches: (1) a concatenated ML analysis, and (2) a two-step species tree analysis. For the concatenated analysis, we first identified a set of alignments that met two requirements: each had to comprise at least 14 sequences (i.e., 75% complete) and be at least 300 bp long. We concatenated the alignments in PHYLUCE and analyzed the dataset in IQ-TREE following the same protocol as for the mitochondrial dataset (except we did not partition the dataset). For the two-step species-tree analysis, we first inferred a gene tree in IQ-TREE for each UCE alignment with at least four sequences and at least 300 bp. For each alignment, we tested substitution models but did not estimate nodal support. We then subjected the resulting gene trees to TreeShrink (Mai and Mirarab 2018) to remove any outlier long branches. This final set of gene trees was then used as input for a species tree analysis in ASTRAL v. 5.7.7 (Zhang et al. 2018). Nodal support was measured using local posterior probabilities, a quartet-based support metric.

The 2017 and 2019 fieldwork in Sarawak recovered three species of shrews: Crocidura foetida, C. neglecta, and several specimens of a medium-sized, dark-colored shrew with relatively short tails. Nearly all shrews were captured in pitfall traps. Crocidura foetida and C. neglecta were captured in the same traplines as the dark-colored shrews, suggesting syntopy among all three species.

The 10-gene mitochondrial dataset comprised 10,399 bp of sequence data; 48.0% of the data matrix was missing. Newly generated mitochondrial sequence data are available on GenBank (Suppl. material 1). ModelFinder partitioned the dataset into seven subsets and identified the best-fitting nucleotide substitution model for each (Suppl. material 2). The mitochondrial phylogeny shows that the holotype of Suncus ater is sister to the six recently collected, dark-colored shrews from northern Sarawak (Fig. 2). Furthermore, the S. ater clade is sister to a clade of the Palawan endemic, Palawanosorex muscorum Hutterer et al., 2018, a recently described crocidurine genus (Hutterer et al. 2018). The second nine-gene mitochondrial dataset had 29% missing data. The resulting estimated phylogeny is similar to the 10-gene topology, with the only changes occurring in the branching pattern of distantly related Crocidura taxa (Suppl. material 3). We found that the average CYTB p-distance between the five P. muscorum samples and the six newly collected, dark colored shrews is 11.77% (SD = 0.50%). These two species are distant relatives of species currently placed in the genus Suncus. The average mtDNA p-distance of P. muscorum and the two newly collected, dark-colored shrews in the nine-gene alignment, without CYTB, is 17.75%. The average nine-gene mtDNA p-distance of the S. ater holotype and the two newly collected, dark-colored shrews is 3.6%.

Figure 2. 

Maximum-likelihood crocidurine mitochondrial gene tree inferred in IQTree. Bootstrap values are given at the nodes. The holotype of ater (MCZ36574) forms a clade with the specimens recently collected in Sarawak. The outgroup branch to Myosorex kihaulei has been truncated.

The complete UCE dataset (which included only those alignments with more than four sequences and that were at least 300 bp long) included 3,757 loci and 2,175,243 bp of sequence data; 12.9% of the alignments overall were represented by missing data. The mean locus length was 579 bp (range: 300–1,864). All Illumina reads and UCE sequences are available as NCBI BioProject PRJNA901984 (Suppl. material 1). UCE data were not recovered from the S. ater holotype. However, the concatenated ML phylogeny places the Sarawak specimens as sister to P. muscorum. The S. ater + P. muscorum clade is not aligned with any species in the genus Suncus (Fig. 3A). The two-step species tree recovered nearly the same topology as the concatenated ML phylogeny with no changes to the relationship between S. ater, P. muscorum, and other members of the genus Suncus (Fig. 3B).

Figure 3. 

Phylogenetic hypotheses from UCE data. The tree topologies are very similar between the two methods. A phylogenetic tree inferred using 3,757 concatenated UCE loci (2,175,243 base pairs) in IQTree. Bootstrap supports are provided at the nodes. The branch leading to outgroup Myosorex kihaulei has been truncated for display B species tree inferred using ASTRAL. Nodal supports are given in local posterior probabilities. ASTRAL tree is presented as ultrametric with uninformative branch lengths.

Mitochondrial DNA from the holotype of Suncus ater revealed that it is closely aligned with the six newly collected, dark-colored shrews from Borneo, and that this clade is sister to the species Palawanosorex muscorum, a recently described genus and species known only from the Philippine island of Palawan, north of Borneo (Hutterer et al. 2018). Phylogenetic estimates from UCE data support the mitochondrial results, showing that the newly collected Bornean shrews, which are mitochondrially aligned to the S. ater holotype, are sister to P. muscorum. However, P. muscorum is much larger, has a much longer tail that lacks any bristles, and has several cranial characters that clearly distinguish it from S. ater (Fig. 4; see detailed comparison by Hutterer et al. 2018: 526), and the average CYTB distance between these two species is 11.77%, indicating a long history of reproductive isolation. Therefore, molecular evidence strongly demonstrates that S. ater has a sister relationship with P. muscorum. For this reason, we transfer S. ater to Palawanosorex. Furthermore, based on mitochondrial DNA and morphological data, we also assign the specimens collected in two localities in Sarawak to P. ater, substantially increasing the known geographic range of this species and demonstrating that it is not confined to the slopes of Mount Kinabalu in Sabah, Malaysia. We note that many of the proposed synapomorphies of the genus Palawanosorex no longer apply. The body size and appendage lengths of P. ater are all much smaller than P. muscorum. Additionally, as noted by Hutterer et al. (2018), P. ater does not have the long claws, bare interdigital surfaces, bristle-free tail, wide antorbital bridge, nor the reduced-size P4 and molars of P. muscorum. We do note that, although lacking in the P. ater holotype, the Sarawak specimens do have dorsal foramina in various stages of fusion (Fig. 5), similar to P. muscorum (Hutterer et al. 2018).

Figure 4. 

The holotype of Palawanosorex ater largely matches the external and cranial measurements of P. ater specimens captured in Sarawak, not P. muscorum. Each panel represents a measurement, and the y-axis represents the measurement length in mm. Each species is represented by a different shape. The upper and lower bounds of the point intervals represent the maximum and minimum values for each measurement for each species. Measurements are limited to those reported in the description of Suncus ater (Medway 1965). Palawanosorex ater measurements are taken from six specimens collected in Sarawak, Malaysia. Palawanosorex muscorum measurements are taken from Hutterer et al. (2018: tables 1, 2). LSUMZ 40695 has a cropped tail and was removed from the total length and tail length measurements.

Figure 5. 

A dorsal and ventral views of the cranium of LSUMZ 40697, Palawanosorex ater, collected on Mount Murud, Sarawak, Malaysia. Photo by Heru Handika B dorsal and ventral views of the holotype of P. ater, MCZ 36574, collected on Mt. Kinabalu, Sabah, Malaysia. Photo by Museum of Comparative Zoology, Harvard University, President and Fellows of Harvard College.