INTRODUCTION

The nominal taxon Marmosa elegans coquimbensis was named by Tate (1931) based on a single specimen in the Field Museum of Natural History (FMNH) collected on 20 June 1923 by Colin C. Sanborn at or near Paiguano (30°02′ S, 70°27′ W, ca. 1600 m elev.) in the Coquimbo region of Chile. Tate's trinomial usage was maintained by most subsequent researchers (e.g., Cabrera, 1958) until elegans Waterhouse, 1839, and other species of fat-tailed mouse opossums were transferred to the genus Thylamys Gray, 1843, by Gardner and Creighton (1989). Most subsequent authors have treated coquimbensis as a subspecies (Palma, 1997) or a junior synonym (Creighton and Gardner, 2008) of T. elegans, but we, together with S. Jansa (Giarla et al., 2010), transferred coquimbensis from the synonymy of T. elegans to that of its sister species T. pallidior (Thomas, 1902) based on morphological traits of the holotype (FMNH 22302). If valid, our reidentification of FMNH 22302 as T. pallidior would have been the first record of that species from the Coquimbo region. Recently, however, Boric-Bargetto et al. (2016) analyzed cytochrome b sequence data from several specimens of Thylamys collected near Paiguano and recovered them as members of a northern haplogroup of T. elegans. Although the authors admitted the possibility that T. elegans and T. pallidior might be sympatric near Paiguano, they concluded that the haplogroup in question should be called T. e. coquimbensis. In order to test their conclusions, we extracted and sequenced DNA from the 96 year old holotype of coquimbensis for phylogenetic analysis.

Materials and Methods

A single claw was clipped from the dried study skin of FMNH 22302. To avoid contamination from exogenous DNA, all pre-PCR laboratory procedures were performed in a biological safety cabinet with UV sterilization in a lab where mammalian DNA is never amplified and in which contaminating mammalian PCR products are unlikely to be present. Moreover, no other didelphid DNA samples had ever been present in the building where the work was conducted prior to the experiment. The sample was soaked in ethanol overnight and then rinsed with water three times. The same wash procedure was performed the next day. On the third day, the sample was allowed to air-dry before DNA extraction was performed with a modified DNeasy Blood and Tissue Kit (Qiagen) protocol. The sample was added to a mixture of 40 µl proteinase K, 160 µl Buffer ATL, and 30 µl 1 M dithiothreitol and incubated at 56° C in a shaking mixer for 24 hours. The digested sample was then lysed by adding 200 µl Buffer AL, mixing, and incubating at 70° C for 10 minutes. DNA was precipitated using 200 µl cold ethanol and incubated at 4° C for one hour. The sample was added to a QIAquick PCR Purification Kit (Qiagen) spin column. This column was chosen over the standard DNeasy spin columns because it is designed to preferentially bind fragments of DNA less than 10 kb. The subsequent wash and elution steps followed the usual DNeasy tissue extraction protocol.

Nested pairs of primers were designed to span 174–359 bp pieces of the mitochondrial cytochrome b (CYTB) gene in the degraded sample. To ensure that primers would amplify both a relative of Thylamys pallidior and T. elegans, primers were designed using a consensus sequence of two T. pallidior CYTB sequences (GenBank accession numbers HM583398 and HM583387) and one T. elegans whole mitochondrial genome sequence (NC005825). Primers were designed in Geneious R9 (Biomatters, Inc.) using the Primer3 algorithm (Untergasser et al. 2012), and only regions that exhibited sequence conservation between the two reference species were used (table 1). Seven sets of PCR reactions were conducted using seven primer pairs. Each reaction contained 13 µl of GoTaq Green Master Mix (Promega), 9 µl of water, 1 µl of each 10 µ M primer solution, and 1 µl of undiluted sample DNA. A negative control reaction, where no DNA was added, was conducted for each separate PCR mixture. The reaction mixture was PCR amplified using an initial 2 min. melting phase at 95° C; then 40 cycles of 30 s at 95°, 30 s at 55°, and 30 s at 72°; then 5 min at 72°. PCR product size was verified on a 1% agarose gel. PCR products were purified using ExoSAP-It (Thermo Fisher) and sent to GENEWIZ (South Plainfield, NJ) for Sanger sequencing in both directions.

For each PCR amplicon, chromatograms were examined, trimmed, and assembled in Geneious. To determine whether individual amplicons might be derived from contaminant DNA, each was subjected to a standard nucleotide BLAST search (Altschul et al., 1990) against GenBank's nonredundant nucleotide database. After ruling out contamination, FMNH 22302 CYTB amplicons were assembled to a reference Thylamys pallidior sequence and trimmed to include only the CYTB coding region. A representative set of 72 Thylamys CYTB sequences and one sequence from the outgroup Lestodelphys halli were downloaded from GenBank (appendix 1). This set included sequences from all the Thylamys species recognized by Giarla et al. (2010), broad geographic sampling within the Elegans Group, and all the sequences that Boric-Bargetto et al. (2016) identified as T. elegans coquimbensis. These sequence data were aligned in Geneious using MUSCLE (Edgar 2004). The alignment was partitioned by codon position, and different partitioning schemes and nucleotide substitution models were tested in PartitionFinder2 (Lanfear et al. 2016). A Bayesian phylogeny was inferred in MrBayes 3.2.6 (Ronquist et al., 2012) using the best-fitting partitioning scheme and substitution models. Two simultaneous Markov chain Monte Carlo (MCMC) runs were initiated, each for 10 million generations and with four chains. Convergence of the MCMC runs was assessed in Tracer 1.7.1 (Rambaut et al., 2018) to ensure that each parameter estimate had an effective sample size greater than 200.

TABLE 1.

Pairs of primers used to sequence overlapping pieces of cytochrome b and associated information about PCR amplicons.

RESULTS

A complete CYTB sequence from FMNH 22302 (GenBank accession MK907779) was assembled from seven individually amplified and overlapping PCR amplicons derived from high-quality Sanger reads. When subjected to individual BLAST searches, each amplicon matched a Thylamys elegans sequence in the nonredundant nucleotide database (table 1), ruling out contamination from nondidelphids. Once assembled, the CYTB amplicons overlapped each other by 84.5 bp on average (range: 20–218 bp), with no conflicting bases. After alignment, PartitionFinder2 identified a best-fitting partitioning scheme and associated nucleotide substation models in which each codon position was allowed its own substitution model (position 1: GTR+G; position 2: GTR+I; position 3: GTR+I+G). The resulting phylogeny unambiguously recovered the holotype of coquimbensis as nested within T. elegans as that species is currently recognized (fig. 1).

FIG. 1.

Bayesian phylogenetic tree of Thylamys cytochrome b sequences. Numbers at nodes indicate posterior probabilities (PP). Filled circles at nodes denote PPs equal to 1.0. Unmarked nodes received PPs less than 0.5. Within T. elegans, tips are labeled with country, region, specimen identifier, and, in parentheses, a GenBank accession number. The holotype of Marmosa elegans coquimbensis Tate, 1931, is in boldface type. For other species, tips of the phylogeny are collapsed and the outgroup is not shown. See appendix 1 for a full list of sequences included in the phylogeny. A full tree file corresponding to this topology is available on TreeBase (doi:  http://purl.org/phylo/treebase/phylows/study/TB2:S25505).

DISCUSSION

Our results unequivocally support Boric-Bargetto et al.'s (2016) conclusion that coquimbensis is more closely related to Thylamys elegans than to T. pallidior, although the gene tree we recovered does not address the issue as to whether coquimbensis should be treated as a subspecies or as a synonym of the former species. Indeed, the trinomial classification of what is currently considered T. elegans is complicated by several divergent but unnamed mtDNA lineages, one consisting of two sequences from Tarapacá (SSUCMa519, SSUCMa520; the “Loa lineage” of Boric-Bargetto et al., 2016), another of a single sequence from Valparaíso (NK96763), and a third of seven sequences from Maule and O'Higgins (NK105928, NK106178, NK160466, NK160518, NK160526, NK160945, NK160972). Of these, only the “Loa” haplogroup from Tarapacá was analyzed by Boric-Bargetto et al. (2016), although the Maule/O'Higgins and Tarapacá sequences had previously been analyzed by Palma et al. (2014).

Our results also raise questions about phenotypic character variation within Thylamys elegans. The type of coquimbensis (FMNH 22302) is—as previously reported by us (Giarla et al., 2010: 45)—phenotypically very similar to T. pallidior, notably in pelage coloration (much paler dorsally than typical T. elegans, and with almost entirely self-white ventral fur), hind-foot length, and bullar inflation. Apparently, the Paiguano topotypes examined by Boric-Bargetto et al. (2016) are also pale-furred, but other specimens that they refer to T. e. coquimbensis (e.g., NK 27583, one of a series of specimens that we examined at the Museum of Southwestern Biology in Albuquerque, NM) are much darker, so this trait is evidently not diagnostic of the haplogroup that they associate with this trinomen.

The other traits of FMNH 22302 that led us to synonymize coquimbensis with Thylamys pallidior (smaller feet, more inflated bullae) were not mentioned by Boric-Bargetto et al. (2016), who emphasized the diagnostic value of nasal morphology, in which their Paiguano specimens were said to resemble T. elegans rather than T. pallidior. We had previously evaluated this character (which was also mentioned by Solari, 2003) as a diagnostic criterion for our revision (Giarla et al., 2010), but found too much intraspecific variation for it to be useful, and we invite readers to inspect Boric-Bargetto et al.'s (2016: fig. 2) illustration to decide this issue for themselves.

In summary, the mtDNA sequence data at hand clearly support Boric-Bargetto et al.'s (2016) decision that the nominal taxon coquimbensis should be associated with Thylamys elegans rather than with T. pallidior, but there remain unresolved issues of trinomial nomenclature and geographic variation that should be addressed in any future revision of these taxa.