Systematics, biogeography and phylogenomics of northern bog lemmings (Cricetidae), cold-temperate rodents of conservation concern under global change

Abstract

Northern bog lemmings, Mictomys (Synaptomys) borealis, are currently being assessed for protections under the U.S. Endangered Species Act. A major impediment to comprehensive evaluation is a deficiency of data towards understanding the biology of these rodents. Inherent rarity and scarce specimen sampling, despite a continent-wide distribution, has precluded our ability to implement modern methods for resolving taxonomy, evolutionary history, and investigating multiple other species traits. Here we use a maternally inherited locus (mitochondrial cytochrome b) and between 5,939 and 11,513 nuclear loci from reduced representation sequencing (ddRADseq) to investigate the evolutionary history of northern bog lemmings. We 1) qualify evidence based on morphological and early molecular studies for the genus assignment of Mictomys, 2) test the validity of multiple sub-species designations, 3) provide spatial and temporal historical biogeographic perspectives, and 4) discuss how incomplete sampling might influence conservation efforts. Both mitochondrial and nuclear datasets exhibit deep divergence and paraphyly between two recognized species, the northern (Mictomys borealis) and southern (Synaptomys cooperi) bog lemmings. Based on mtDNA, the geographically isolated subspecies (M. b. sphagnicola) was found to be divergent from all other specimens. The remainder of the species exhibited shallow intra-specific differentiation in mtDNA, however nuclear data supports genetic distinction consistent with four geographic subspecies. Recent coalescence of all northern bog lemmings (except for M. b. sphagnicola) reflects divergence in multiple refugia through the last glacial cycle, including a well-known coastal center of endemism and multiple regions south of continental ice-sheets. Regional lineages across North America suggest strong latitudinal displacement with global climate change, coupled with isolation-reconnection dynamics. This taxon suffers from a lack of modern samples through most of its distribution, severely limiting interpretation of ongoing evolutionary processes, particularly in southern portions of the species’ range. Limited voucher specimen sampling of vulnerable populations could aid in rigorous conservation decision-making.

Introduction

The northern bog lemming, Mictomys (Synaptomys) borealis, Richardson 1828, is pending a federal listing decision by the U.S. Fish and Wildlife Service under the Endangered Species Act (ESA). However, M. borealis is one of the most data-deficient species among all North American small mammals, from the perspective of its evolutionary history, ecology, and life history. It is seemingly rare in the sense that it is generally detected in very low abundance despite a continent-wide range, but little is known of its population densities, demographics, or connectivity to quantify the presumption of relative rarity. From an evolutionary perspective, genetic diversity, historical demography, geographic variation, or even taxonomy based on phylogenetic relationships remain poorly investigated. There is, therefore, an urgent need to more fully understand fundamental aspects of the biology of this species that is hypothetically declining (Baltensperger et al. 2022). This species succinctly demonstrates a pervasive and continuing problem for conservation of biodiversity, in that rare species (or rarely detected species; pervasive scarcity) are difficult to study, even given advances in detection technology and genomics. Although museum specimens can be an excellent resource for understanding species’ biology (e.g., Roberts et al., 2016), often, insufficient samples exist within biorepositories to fuel modern analytical methods (Remsen, 1995). From what is known, M. borealis is a cold-temperate species, generally associated with low-lying cool, mesic, bog-like habitats, broadly distributed across mid-to-high latitudes of North America from northwestern Alaska, northernmost Quebec, and Newfoundland and Labrador in the north to fragmented pockets of the boreal zone within the northernmost contiguous United States (Fig. 1). Much of the narrative to petition this species for listing under the ESA suggests that suitable cold bog habitats are relictual, increasingly disconnected, and under threat from both anthropogenic climate trends and local disturbances, especially at southern distributional latitudes (Jones & Melton, 2014). These southern regions are also coincident with the distribution of seven of nine putative subspecies, but they collectively have the fewest available specimens for analysis (Fig. 1).

Figure 1.

Maps of North America showing the range-wide distribution of Mictomys borealis (retrieved from the IUCN website https://www.iucnredlist.org/species/42638/22377185). A – The distribution of specimens sampled for this study (locality numbers correspond to specimen data in Supplementary Materials Appendix A). Also shown are estimated subspecies distributions (based on Hall, 1980). B – The distribution of available georeferenced samples of Mictomys borealis within publicly searchable museum archives, including only results from ARCTOS (https://arctos.database.museum/). Samples are divided into thirds by longitude across North America and summarized in each region using histograms by total numbers of specimens collected through time (1901–1930, 1931–1960, 1961–1990, 1991–2020). Specimens represented only by study skin and skeleton are shown by open circles and gray bars, and specimens with tissues available or fluid-preserved are shown by closed circles and black bars.

Current taxonomy (Mammal Diversity Database, 2023) places northern bog lemmings within the Family Cricetidae, Subfamily Arvicolinae, Tribe Lemmini, and Genus Mictomys, True 1894. Generic distinction between Mictomys and Synaptomys is based on morphological and fossil evidence (Abramson & Nadachowski, 2001; Koenigswald & Martin, 1984; Markova et al., 2018; Martin et al., 2003; Repenning, 2001; Meade et al., 1992; Tesakov & Bondarev, 2021). Early genetic evidence corroborates this nomenclature, given differences in karyotype (Hoffmann & Nadler, 1976), and a lack of monophyly of M. (Synaptomys) borealis and S. cooperi, based on mitochondrial DNA (Buzan et al., 2008; Steppan & Shenk, 2017). Despite these findings, there has been resistance to accept the current nomenclature, instead opting to keep both species of bog lemmings within Synaptomys, pending additional evidence from modern data (Cassola, 2017; Knox-Jones et al., 1986; Musser & Carleton, 2005). Here we explicitly consider northern bog lemmings as Mictomys borealis, and use this taxonomy as a basis for testing the validity of relationships among genera.

The evolutionary timeline for bog lemmings also remains unclear. Repenning & Grady (1988) hypothesized that Mictomys is a derived form that evolved rapidly from Synaptomys in the mid-Pleistocene (1.3–1.6 Ma) whereas Martin et al. (2003) suggested independent early Pleistocene evolution of each of these genera. From an ecological perspective, bog lemming species are generally considered closely related based on similar habitat associations (mesic bog habitats) and similar external morphology (Banfield, 1974; Rose & Linzey, 2022). Within M. borealis the nine putative subspecies originally designated based on geographic variation in morphology have been reviewed in detail elsewhere, but their validity has been questioned as tentative (Fig. 1B; Hall, 1981; Howell, 1927; Bradley, 2017).

Since the early taxonomic and ecological studies on M. borealis, modern techniques have been developed that have improved our knowledge of the biology of bog lemmings. Ecological niche modeling and related spatial analyses have revolutionized our ability to assess bog lemming distributions as related to climatic variables, and to generate hypotheses for future (and past) distributions (Baltensperger & Huettman, 2015a, 2015b; Hope et al., 2015). Stable isotope ecology has allowed for comparative analysis of diet and trophic position from the perspectives of both population and community ecology (Baltensperger et al. 2015). Genetic metabarcoding has recently provided even more detailed insight to dietary composition among M. borealis from Alaska (Baltensperger et al. 2022). The availability of museum specimens that made these studies possible is due to rigorous spatial and temporal sampling from one broad region (eastern Beringia and westernmost North America). Specimen materials generated from these field efforts were preserved explicitly to fuel analyses based on future technological advances, such as those cited, that did not exist at the time of most specimen acquisition (Cook et al., 2017; Galbreath et al., 2019).

In this study, we implement reduced representation genomic sequencing coupled with more traditional amplicon sequencing of mitochondrial DNA to further extend the scientific value of existing M. borealis specimen resources. We first assess competing hypotheses of relationships between members of the Tribe Lemmini (including the genera: Myopus, Lemmus, Mictomys, and Synaptomys) to evaluate the status of genera among these groups and to integrate the timeframe of diversification into knowledge of the historical biogeography of North America. Second, we provide a preliminary range-wide phylogeographic assessment of M. borealis, highlighting the strengths of robust genomic data coupled with the limitations of sparse sampling. Finally, given pending decisions towards conservation actions for M. borealis, we provide a discussion of the continued merits of whole-specimen vouchering for species of perceived rarity and/or recent decline.

Materials and methods

Range-wide quantification of sample availability

To assess availability of samples of M. borealis from publicly databased museum archives, we reviewed the geographic distribution of historic specimen collections through time, as well as the availability of potential high quality genomic resources. We searched both ARCTOS (https://arctos.database.museum/) and GBIF (https://www.gbif.org/) specimen data portals and downloaded all available specimen records of M. (Synaptomys) borealis that reflect physical voucher specimen materials (i.e., not including photo evidence or reports of occurrence). We removed specimen records with no georeferenced locality information, or date of collection. Specimens were then binned in 30-year intervals of collection (1901–1930, 1931–1960, 1961–1990, 1991–2020). Given that ARCTOS data provides specimen preservation information, specimens reported through this data portal were also split into those consisting only of traditional study skin and/or skeleton materials, and those that were either preserved as fluid vouchers (ethanol preserved) or with frozen or fluid preserved tissues. Given that ddRADseq methods generally require non-degraded DNA, these divisions of specimens are considered to generally reflect their suitability for genomic analysis (Galbreath et al. 2019).

Genetic sampling and DNA extraction

To generate Cytb sequences, samples of tissue were obtained through field sampling and institutional loans, and analyzed in combination with data from available sequence databases (n_tot=54; Supplemental Materials, Appendix A; Fig. 1). Samples used for analyses that were collected from Manitoba were sampled using pitfall cups (AGH), and those on loan from Alaska were obtained through use of either snap traps or pitfall cups (MacDonald & Cook, 2009). Cytb sequences represent six of the nine recognized subspecies of M. borealis, based on the geographic distribution of specimens, including M. b. chapmani (n=3), M. b. dalli (n=39), M. b. innuitus (n=1), M. b. sphagnicola (n=4), M. b. smithi (n=4), and M. b. wrangeli (=truei; n=3). Additionally, multiple outgroup samples were obtained representing most other genera of lemmings across the Holarctic, including Phenacomys (P. ungava, n=1; P. intermedius, n=1) and taxa within the Tribe Lemmini other than M. borealis, including Lemmus lemmus (n=2), L. sibiricus (n=2), L. trimucronatus (n=3), Myopus schisticolor (n=2), and Synaptomys cooperi (n=9). For nuclear analyses a subset of M. borealis (n=21) representing the same subspecies (except for M. b. chapmani for which no DNA was available, and M. b. sphagnicola, where the DNA was not of sufficient quality for nuclear sequencing methods), L. trimucronatus (n=2) and S. cooperi (n=5) provided DNA of sufficient quality for robust sequencing. DNA was extracted from muscle tissue using standard salt extraction methods (Fleming & Cook, 2002).

mtDNA Cytb PCR and sequencing

For most samples, the Cytb gene was PCR amplified using primers L14729/H15985 and optimized thermocycling conditions (Lebedev et al., 2007). Sanger sequencing was performed on an ABI 3730 by Genewiz LLC (South Plainfield, NJ). Raw reads were cleaned and aligned using Geneious (Kearse et al., 2012). For samples of M. b. sphagnicola, DNA was extracted using a QIAamp DNA micro kit (Qiagen) and the complete Cytb gene was targeted for PCR amplification using primers MTCB-F/-R (1,143 bp; Naidu et al., 2012) or, if DNA quality was insufficient to yield a complete Cytb sequence, novel internal primers were used to target a fragment of the Cytb gene (176 bp; ME-mammals-3F: 5’ ATR GCA ACA GCM TTY ATA GG 3’; ME-mammals-3R: 5’ GTD GCT TTR TCT ACW GAR AA 3’). PCR products were Sanger sequenced on an ABI 3730xl by the Yale DNA Analysis Facility (New Haven, CT). Raw reads were cleaned and aligned using BioEdit v. 7.0 (Hall, 1999)

Nuclear ddRADseq sequencing and variant calling

Genomic DNA was quantified using Quant-iT Picogreen dsDNA Assay (Invitrogen). Samples with good yields (>100ng) of high molecular weight DNA were submitted to the University of Minnesota Genomics Center (Minneapolis) for double digest restriction site associated DNA sequencing (ddRADseq; Peterson et al., 2012), using SbfI and TaqI restriction enzymes (New England Biolabs). Libraries were sequenced with 150 bp single-end reads across a quarter lane of a NextSeq 550 (Illumina, USA), using a High-Output flow cell. Raw reads were trimmed and padding sequences removed with gbstrim.pl (https://bitbucket.org/jgarbe/gbstrim/src/master/gbstrim.pl). Trimmed reads were filtered for quality with process_radtags in Stacks v2.54 (Rochette et al., 2019). Briefly, this removes reads with uncalled bases and low-quality scores using a sliding-window approach. Loci were discovered and SNPs called with the denovo.pl pipeline in Stacks, using M=n=3, which optimized the number of polymorphic loci while minimizing the number of potentially paralogous loci, resulting in 129,231 loci and 92,251 SNPs (Rochette & Catchen, 2017). We reduced SNPs to one per locus (--write-single-snp) using the populations module in Stacks, keeping 41,036 SNPs, and filtered for quality and completeness using the R packages SNPfiltR (DeRaad, 2022) and vcfR (Knaus & Grünwald, 2017). To maximize the number of variable SNPs retained for M. borealis, we first removed S. cooperi and L. trimucronatus individuals. We set a minimum coverage (read depth) of 5 and genotype quality of 30, filtered heterozygotes for an allele balance ratio between 0.25–0.75, set a maximum read depth of 50, a maximum proportion of missing data per sample of 0.58, and a maximum proportion of missing data per SNP of 0.8, retaining 5,939 SNPs with 3.78% missing data. These values were determined in SNPfiltR by iteratively assigning cut-offs for missing data, checking PCAs to make sure missing data was not driving spatial differentiation and assigning final values that resulted in most data coupled with robust inference. This SNP-set was used for all analyses except the RAxML phylogenetic tree. A number of recent studies have demonstrated the utility of ddRADseq data for multi-species phylogenetic analyses, although with precautions for careful assessment of SNP yield versus missing data (e.g., Leaché et al., 2015a, 2015b; Salas-Lizana & Oono, 2018). Given that there was a large amount of missing data for S. cooperi and L. trimucronatus at the 5,939 SNPs used for the M. borealis, we used relaxed filters for RAxML analysis of multiple species. For the outgroup SNP-set, we implemented the same coverage, genotype-quality, and allele balance ratio filters, but kept every individual and set the maximum proportion of missing data per SNP at 0.6. These filters retained 11,513 SNPs with 21.05% missing data overall.

Phylogenetic analyses

For mtDNA Cytb, we first analyzed only the ingroup consisting of all Mictomys samples, through estimation of a Bayesian tree with BEAST v2.6.3 (Bouckaert et al., 2019), with models of evolution inferred during analysis (bModelTest, Bouckaert & Drummond, 2017), using the transition-transversion split option and empirical frequencies. We used a lognormal relaxed molecular clock with a mutation rate of 5% per million years, as estimated for Lemmus (Hope et al., 2014) and the Coalescent Constant Population tree prior. We ran the MCMC for 10,000,000 generations, sampling every 5,000. Based on the log file from this analysis interpreted in Tracer v1.7.1 (Rambaut et al., 2018), the rate of evolution did not significantly deviate from a constant rate. As such, given that the coalescent timeframe for evolution within the genus Mictomys was the primary focus of this work, a subsequent phylogeny including outgroup taxa was performed using the same methods, but applying a strict molecular clock so that rate variation along deep (long) branches did not influence coalescence estimates for intraspecific crown taxa (Mictomys and Synaptomys). Resulting phylograms were midpoint rooted and visualized with posterior probabilities in FigTree v1.4.4 (Rambaut, 2012). Finally, we estimated average corrected pairwise sequence divergence between all genera in DnaSP (Librado & Rozas, 2009).

For nuclear ddRADseq data, using the outgroup SNP-set, we ran RAxML (Stamatakis, 2014) to estimate a maximum likelihood (ML) phylogeny, removing heterozygous sites that would lead to invariant sites if phased using the python script (https://github.com/btmartin721/raxml_ascbias), yielding 6,789 variant sites. We tested branch support with 100 bootstraps and again visualized trees in FigTree.

Population genetics/genomics

Based on the mtDNA Cytb data, genetic diversity analyses used all available sequences of Mictomys combined. We calculated summary statistics in DnaSP including nucleotide diversity (π), and average corrected pairwise sequence divergence for samples representing each putative sub-species, also being coincident with mtDNA clades. For tests of demographic change through time, we used DnaSP to calculate Tajima’s D (Tajima 1989), assessing significance with 10,000 coalescent simulations, and a mismatch distribution for pairwise sequence divergence.

Based on the M. borealis nuclear SNP-set, we conducted a PCA using the R package adegenet (Jombart, 2008), keeping the first 10 principle components. We calculated expected heterozygosity for each M. borealis population in adegenet and calculated observed heterozygosity for each individual by dividing the number of heterozygous SNPs by the total number of SNPs. We calculated F_ST using the R package StAMPP (Pembleton et al., 2013), following the Weir & Cockerheim (1984) method. We also used StAMPP to calculate Nei’s genetic distance between each individual, which was used to generate an unrooted neighbor-joining network in Splitstree (Huson, 1998). We ran STRUCTURE for 500,000 MCMC repetitions, with a 100,000 burn-in, for K=2–7 and 10 replicates for each value of K. We used the admixture model and did not use prior population information. We used the Evanno method to determine the best value of K (Evanno et al., 2005).

Results

Sampling

Sampling was heavily weighted to available specimen resources from eastern Beringia (M. b. dalli from Alaska and the Yukon Territory) reflecting greater sampling effort from this region (Fig. 1; Cook et al., 2017). Sampling from elsewhere across North America reflected a scarcity of available high quality genomic resources in existing biorepositories (Fig. 1; Fig. S1), and roughly half of available Cytb sequences on GenBank were not represented by existing voucher specimens from which additional DNA might be obtained for nuclear analyses (Supplemental Materials, Appendix A). The only samples available for analysis of putative M. b. sphagnicola consisted of old museum study skins that were sampled and yielded only poor-quality DNA and limited, but important sequence information for Cytb.

Phylogenies

Considering Cytb data for the Mictomys ingroup, the best model of substitution inferred from bModelTest was TN93+I+G. There was minimal phylogeographic structure evident based on current sampling (Fig. 2A; Supplemental Materials Fig. S2). Coalescence of the entire species was dated to 0.3Ma, coincident with diversification through the last three glacial cycles. The basal divergence at this time was between M. b. sphagnicola and the remainder of the species. All other subspecies and lineages coalesce at ~0.14Ma, coincident with differentiation since the Sangamon interglacial. Some sub-groups of specimens within Alaska were well-supported, although eastern Beringian specimens collectively did not form a supported monophyletic group. Additional branches with moderate to high support included two specimens from southeast Alaska, sampled from Betton Island and Revillagigedo Island, three specimens from northwest Montana, two specimens from southwest Manitoba, and two specimens from southeast Manitoba. An additional specimen from mainland southeast Alaska, within the geographic range of the subspecies M. b. wrangeli (UAM:69950), was aligned with the matrilineage of specimens from interior Alaska (M. b. dalli).

Figure 2.

A – Bayesian chronogram based on the mitochondrial cytochrome b gene for genealogical relationships within Mictomys borealis, and phylogenetic relationships among related species. Numbers along branches show posterior probability support for clades, and italicized bold numbers on nodes show coalescent times in millions of years before present. B – Maximum Likelihood phylogram showing relationships among taxa based on reduced representation genomic SNPs (11,513 loci). Numbers along branches show bootstrap support for clades. For both trees, support values for clades ≤ 0.75/75 are excluded. Specimen labels reflect sample data in Supplemental Materials Appendix A and numbers represent collection localities in accordance with Fig. 1A.

Inclusion of samples of other species resulted in long branches between species, coupled with shallow intraspecific divergence (Fig. 2A). Myopus formed the basal split rendering all recognized genera within the Tribe Lemmini as a paraphyletic group with respect to Phenacomys, although none of the basal nodes in the Cytb tree was well supported. Given a timeframe of 2–3.5Myr, intra-generic relationships among the Tribe Lemmini and allies warrants further investigation with nuclear data. However, there was unanimous support for all genera as reciprocally monophyletic. Age of divergence among these taxa is coincident with the early Pleistocene with 95% CI of all divergence times overlapping (not shown). However, use of a strict molecular clock likely does not account for homoplasy and variable rates through time, and even earlier (late Pliocene) divergences are possible. Given that Mictomys and Synaptomys collectively do not form a monophyletic group, nor are they reciprocally monophyletic, and divergences among all groups are very deep, assignment to independent genera is supported. Further, average pairwise sequence divergence among genera within the Tribe Lemmini ranged from 9.7–13.8% (Table 1). Mictomys and Synaptomys were 13.4% divergent.

Table 1.

Average pairwise sequence divergence (%) between all species considered, based on available Cytb data. Sample sizes (n) are provided.

	n	2.	3.	4.	5.	6.	7.	8.
1. Mictomys borealis	54	13.4	13.2	10.1	9.7	12.8	14.4	16.2
2. Synaptomys cooperi	9	--	13.2	13.1	13.1	12.5	14.3	13.5
3. Lemmus trimucronatus	3	--	--	11.5	11.8	13.2	14.4	13.3
4. Lemmus lemmus	2	--	--	--	5.2	13.8	14.0	14.0
5. Lemmus sibiricus	2	--	--	--	--	13.7	15.2	14.4
6. Myopus schisticolor	2	--	--	--	--	--	17.3	16.1
7. Phenacomys intermedius	1	--	--	--	--	--	--	6.0
8. Phenacomys ungava	1	--	--	--	--	--	--	--

Based on SNP data from the RAxML phylogeny, the best model of substitution was GTR+G. Mictomys, Lemmus, and Synaptomys all form monophyletic groups, and Synaptomys is well-supported as a sister taxon to Lemmus, as opposed to Mictomys (Fig. 2B). Within Mictomys, all eastern Beringian specimens representing M. b. dalli form a loosely supported monophyletic group. The two insular specimens from southeast Alaska also form a monophyletic group, and along with the single specimen (UAM:69950) collected from the southeast Alaska mainland, there is moderately high support for a southeast Alaska lineage, consistent with the subspecies M. b. wrangeli. This mainland specimen is also consequently an example of mito-nuclear discordance among subspecies assignments, given the Cytb sequence is aligned with M. b. dalli, but the nuclear phylogeny aligns it with M. b. wrangeli. There is also moderate support for a central Canada lineage based on two specimens from Manitoba coincident with the subspecies M. b. smithi. The single specimen from Quebec (eastern Canada) is divergent from all other Mictomys, coincident with the subspecies M. b. innuitus (Fig. 2B).

Population genetics/genomics

Considering all Mictomys sequences combined, nucleotide diversity for the Cytb data was low (π = 0.005) and the value of Tajima’s D (D = −2.07; P = 0.003) suggests significant demographic post-glacial expansion. Intra-specific genetic divergence between samples representing putative subspecies of M. borealis based on Cytb sequences are shallow, ranging from 0.46–1.87% (Table 2). There is no apparent isolation by distance and highest divergence exists between samples of M. b. sphagnicola and all other samples.

Table 2.

Genetic differentiation between designated groups of specimens representing putative subspecies. Numbers above the diagonal are average pairwise sequence divergence (%) based on Cytb sequences. Numbers below the diagonal are F_ST values based on genome-wide SNP data. Mictomys borealis smithi is split into west and east considering separate supported mtDNA lineages from these regions. NA = not applicable based on a lack of sampling for genome-wide SNPs from these regions.

SNPs\Cytb	dalli	wrangeli	chapmani	smithi (w)	smithi (e)	innuitus	sphagnicola
M. b. dalli	--	0.46	0.53	0.93	0.46	0.79	1.37
M. b. wrangeli	0.26	--	0.80	0.76	0.85	0.65	1.07
M. b. chapmani	NA	NA	--	1.27	0.50	1.00	NA
M. b. smithi (west)	NA	NA	NA	--	0.89	0.90	1.87
M. b. smithi (east)	0.41	0.32	NA	NA	--	0.64	1.07
M. b. innuitus	0.63	0.60	NA	NA	0.29	--	1.07
M. b. sphagnicola	NA	NA	NA	NA	NA	NA	--

Based on SNP data for Mictomys, the first two axes of the principal components analysis account for 24.06% and 12.51% of total genomic variance (Fig. 3A; variance for other axes are included in Supplementary Materials Appendix A). Four groups are spatially separated across each of the first two axes, again representing eastern Beringia (M. b. dalli), southeast Alaska (M. b. wrangeli), central Canada (M. b. smithi), and eastern Canada (M. b. innuitus). The mainland sample from southeast Alaska is spatially intermediate between insular individuals and other mainland samples. The unrooted Splitstree topology also supports these groups but also suggests closer relationships between samples that are geographically most proximate (Fig. 3B). Values of F_ST show higher differentiation between groups of samples that are geographically more distant (Table 2), suggesting multiple possibilities that either multiple lineages are distributed longitudinally across North America, or there is a gradation of genomic isolation by distance, although this could not be tested with current sampling. Values for observed heterozygosity were similar except for the two insular specimens (M. b. wrangeli) from southeast Alaska that exhibited very low heterozygosity (Fig. 4). Based on the Structure analysis, the best supported number of clusters was K=4. The Structure plot clearly showed genomic distinction of the Quebec specimen (Fig. 3C). Other groups shared variable proportions of ancestry, but groups again supported both regional distinction and closer genomic relatedness with geographic proximity. There is also a clearly independent genomic signal (assigned as a separate group) from southeast Alaska, although a high proportion of alleles from the mainland southeast Alaska specimen was assigned to both eastern Beringia and central Canada, suggestive of potentially complex admixture in this region. A table of sample quality and genomic read information for all samples included in SNP analyses is provided (Supplemental Materials Appendix A)

Figure 3.

Population genomic relationships based on reduced representation genomic SNPs (5,939 loci). A – PCA plot showing variance among samples based on the first two out of ten total principal components. B – Splitstree distance topology based on a Neighbor-joining network algorithm. C – Structure plot for K=4 groups showing proportional assignment of alleles among samples. Specimen labels reflect sample data in Supplemental Materials Appendix A. For A and B, samples are colored according to geography that reflects putative subspecies assignments. For C, colors reflect group assignment of loci and the majority color for each sample is concordant with geographic subspecies designations apart from a single specimen from mainland southeast Alaska (UAM:69950).

Figure 4.

Bean-plots of observed heterozygosity. Width of plots reflects numbers of samples with similar values. Samples are grouped by geographic subspecies assignment and values are based on reduced representation genomic SNPs (5,939 loci).

Discussion

There are a number of meaningful results from both mtDNA and nuclear data presented here that span multiple scales of analysis and provide insight with respect generally to the management and conservation of rare species. We have 1) provided multiple lines of evidence that define systematic relationships, and support firm recognition of northern bog lemmings as Mictomys borealis, distinct at the generic level from Synaptomys, 2) highlighted that existing intra-specific diversity within M. borealis is congruent with current sub-species designations, and 3) of these sub-species, M. b. sphagnicola is the most distinct. However, additional sampling is required to rigorously assess the status and trends of extant populations, particularly across central and eastern North America.

Higher-level taxonomy among lemmings

Recognition of northern bog lemmings as Mictomys has been noted in the literature, primarily through paleontological studies that have discussed the diversification sequence for different lemming genera based on the fossil record (e.g., Martin et al., 2003). From these debates, the evolution of bog lemmings among other lemming groups is hypothesized to be generally coincident with the late Pliocene or Early Pleistocene, to have occurred either in Europe or Far East Asia, and to potentially consist of a contemporary radiation leading to multiple lemming genera around this timeframe. Our data support published evidence for rapid diversification of lemming genera (e.g., Repenning & Grady, 1988), considering deep branches among all genera and a lack of posterior support at the base of the phylogeny, given the mtDNA data. Due to lack of monophyly for both mtDNA and nuDNA, there is no support for inclusion of both northern and southern bog lemmings within the single genus Synaptomys. These genetic data therefore provide robust support for the validity of Mictomys (as a monotypic genus), and strengthen evidence for uniqueness of this species, with no signal of ongoing geneflow between bog lemming species based on our sampling, and very high divergence suggesting that inter-specific hybridization would be unlikely. From a conservation standpoint, such evolutionary independence reduces the burden of managing potential hybrid taxa, although signals of inter-specific hybridization should continue to be an area of ongoing assessment, especially from areas of sympatry, given the small sample sizes available for this initial genetic study. Although M. borealis and S. cooperi are not easily distinguished based on external morphology, clear cranial and dental character differences exist between the two species coupled with only limited areas of sympatry (Rose & Linzey, 2021).

Intra-specific evolution within Mictomys

In addition to unambiguous taxonomy, conservation decision-making can benefit from knowledge of diagnosable intraspecific groups, demographic trends, and areas of endemism (Hope & Frey, 2022). Historically those key analyses have relied on morphology and ecology to establish putative demographic independence of units of analyses, normally at the level of sub-species. With the advent of phylogeographic studies based on mtDNA and matrilineal histories, recognition of intraspecific lineages has provided a mechanism for assessing the validity of described subspecies, although still with shortcomings, including only single-locus analysis and uniparental inheritance (Burbrink et al. 2000; Zink, 2004). However, the primary advantage of evidence based on mtDNA is rapid differentiation of this gene among mammalian taxa, providing insight into regional histories across shallow (late-Pleistocene) timeframes (e.g., Hewitt, 2004). Of interest, and compared to most other boreal small mammals across North America (e.g., Arbogast & Kenagy, 2001; Hope et al., 2020; Jackson & Cook, 2020), there is a distinct lack of intra-specific mtDNA differentiation among Mictomys across its vast continental range and a very shallow coalescent history with most samples (except M. b. sphagnicola) sharing a common ancestor as recently as the last interglacial period. It is possible that this species has experienced source-sink dynamics repeatedly through the Pleistocene where only a single refugial population was responsible for range-wide post-glacial recolonization following the most recent glacial phase. However, we suggest that the primary explanation for a lack of mtDNA phylogeographic structure is poor sampling. Through our preliminary phylogeny reconstructions, as additional samples were added, additional long branches with strong support were observed that are coincident with geographic regionalization. The most notable of these (with moderate support) is the basal split of samples from south of the St. Lawrence River coincident with the distribution of M. b. sphagnicola. A lack of monophyly of these three samples likely reflects extremely short sequences available for two of the samples that were sourced from dried skin. This taxon is geographically isolated, genetically diagnosable, and occurs in the most peripheral and at-risk part of the species-wide range. Undoubtedly, addition of more samples across the distribution of M. borealis will continue to refine our knowledge of spatial differentiation.

Thousands of putatively unlinked SNPs from genomic sequencing has substantially compensated for a lack of phylogeographic signal based on mtDNA. Although sample sizes are still small, a number of genetically discrete groups were retrieved. A shallow history of differentiation among all groups was again supported by nuclear data, when considering relatively short branch lengths between subspecies (Fig. 2B) and that a small but consistent proportion of nuclear alleles was shared between groups, likely representing incomplete lineage sorting (Fig. 3C). The exception is the individual from Quebec representing M. b. innuitus, which is basal to other lineages considering the nuclear phylogeny. There is also a signal of potential refugial isolation of individuals representing M. b. wrangeli in the Alexander Archipelago of Southeast Alaska, and this lineage is most similar to the bulk of individuals sampled from interior Alaska and northwestern Canada (M. b. dalli). Notably, the one individual sampled from mainland Southeast Alaska exhibited a shared evolutionary history between three groups, shown by non-trivial proportions of nuclear ancestry across Groups 1–3 as well as mito-nuclear discordance based on phylogenies.

Given the evidence, and allowing for low sample sizes, we can suggest a phylogeographic history for Mictomys that reflects long-term isolation (multiple glacial cycles) and divergence of M. b. sphagnicola from the remainder of the species range, and subsequently, more recent isolation of populations in multiple refugia through the Last Glacial phase. Although there is no evidence to support a Beringian refugial population, there is evidence for endemism within the Southeast Alaska refugium with a coalescent timeframe consistent with multiple other mammal species (Sawyer et al. 2019), and multiple refugia south of continental ice-sheets, similarly seen more broadly among mammals (Hope et al. 2020). Following the Last Glacial Maximum, populations expanded to their present range, and have been predicted to continue northward niche expansion, particularly at current northern range limits (Hope et al., 2015; Baltensperger et al., 2015b). It is also possible that distinct phylogeographic transitions (breaks) exist across North America but current sampling precludes identification of where and how many there are. Finally, there is evidence of admixture between subspecies/lineages, suggesting potential genetic isolation by distance between previously discrete populations and ongoing genomic mixing that could progressively lead to loss of regional genomic identity through time. Undoubtedly, current populations experiencing physical isolation from the remainder of the species range have highest likelihood for maintaining local adaptive potential and independent evolutionary trajectories. However, anecdotal data from contemporary field sampling for M. b. sphagnoicola (northern bog lemming working group, personal communication) supports extreme rarity of this subspecies, and genomic heterozygosity among insular samples of M. b. wrangeli shows very low diversity and likely associated small population sizes.

Future sampling of rare species can constructively inform conservation decisions

Much of the existing literature on northern bog lemmings hinges on an underlying assumption of inherent rarity of this species (Jones & Melton, 2014). Further, extreme rarity and ongoing endangered species assessments can influence decision-making towards continued and future development of specimen resources (Malaney & Cook, 2019). It is not clear, based simply on data-deficiency, that Mictomys is extremely rare, in decline, or unable to ecologically adapt to changing conditions. This species is difficult to detect with conventional live-trapping methods, and even museum-style snap traps have seemed more effective through incidental capture along runways as opposed to attracting bog lemmings to bait directly (DuBois, 2016). Pitfall-style traps have also been successful in some instances, detecting and sampling both northern and southern bog lemmings (DeGraff & Yamasaki, 2001; Rose 2006). Some studies have shown northern bog lemmings to be at least locally common (Christian, 1993; Edwards, 1952; Salt, 2000; MacDonald & Cook, 2001). However, most of the ecological knowledge for M. borealis is based on limited field studies, and on surrogate information from S. cooperi (e.g., Banfield, 1974), the latter species which we have shown here to be highly divergent from M. borealis. The majority of modern studies on the biology of M. borealis have relied on museum archives of specimens, especially those requiring genomic samples (current study) or isotopic resources for assessment of diet (Baltensperger et al. 2015; 2022). Given minimal ecological consequences of scientific collecting and potential benefits of robust specimen resources for conservation decisions, additional specimen representation from across the range of M. borealis is warranted (e.g., Patterson, 2002; Hope et al. 2018).

Our meta-analysis of existing specimen resources of M. borealis clearly shows that the vast majority of specimens have been sampled from northwestern North America (eastern Beringia). The major timeframe for collections across central and eastern North America precluded the era of genomic resources (i.e., no tissues preserved) and subsequently virtually no specimens were collected from the most recent three decades. Much of the range of this species remains completely unsampled (Fig. 1; Fig. S1). It is possible that the uneven geographic distribution of bog lemming specimen resources may reflect real differences in abundance across the entire range, but it also reflects disproportionate sampling effort particularly from the northwest region through conservation-based research initiatives that have used targeted techniques for sampling whole small mammal communities (e.g., Cook et al. 2017). Minimally, these efforts demonstrate that northern bog lemmings from eastern Beringia (M. b. dalli) are relatively abundant. Our genomic data support relatively high heterozygosity among this subspecies (Fig. 4) and similarly high or even higher heterozygosity among samples from elsewhere in North America suggesting that less well-sampled regions along the southern periphery of the species range and through central and eastern Canada may also support robust populations. There is still no knowledge of many aspects of the natural history of M. borealis, for which specimen resources constitute the most rigorous way to stimulate investigations (Baltensperger et al., 2022). This includes study of wildlife disease and parasite faunas/microbiomes, co-evolution of hosts and other components of biodiversity, adaptive evolution under global-change scenarios, and documenting the existence of rare species in a given locality and timeframe, considering these small mammals are so often misidentified (Galbreath et al. 2019). Particularly for wide-ranging species that are presumed to be rare, additional specimen sampling remains a critical need.

Conclusions

This is the first genetic appraisal of Mictomys borealis based on wide geographic sampling and specimens that represent multiple named subspecies. Our data provide strong support for taxonomic recognition of Mictomys borealis for all northern bog lemmings. Differentiation within this species is recent and reflects phylogeographic structure that is inherent across other boreal-associated small mammals of North America (e.g., Arbogast & Kenagy, 2001; Hope et al., 2020; Jackson & Cook, 2020), providing further evidence for the generalized impact of climate change when coupled with geography to shape evolutionary legacies across multiple species. Our data cannot reject that some regional populations of northern bog lemmings, especially within peripheral areas (M. b. wrangeli and M. b. sphagnicola), are of particularly high conservation concern. However, more rigorous inference of ongoing population genetic trajectories, genetic diversity, and functional evolutionary responses to ongoing environmental change will necessitate further tissue samples from throughout the species’ range, using sampling methods that may result in whole-body voucher specimens. The best evidence that exists for informing decision making for northern bog lemmings is from rigorous and ongoing sampling efforts in the range of M. b. dalli, a relatively abundant and generalist taxon. Perhaps the best way forward would include similar field efforts from across the boreal zone of North America.

Data availability statement

Mitochondrial cytochrome b sequences can be found in GenBank (Accession numbers: OR350870-OR350919). ddRADseq raw sequence reads are deposited in the SRA (BioProject PRJNA996183). All specimen data are provided online in the Supplemental Materials Appendix A.