Disentangling morphology and genetics in two voles (Microtus pennsylvanicus and M. ochrogaster) in a region of sympatry

Abstract

Species in recent, rapid radiations can be difficult to distinguish from one another due to incomplete sorting of traits, insufficient time for novel morphologies to evolve, and elevated rates of hybridization and gene flow. The vole genus Microtus (58 spp.) is one such system where all three factors are likely at play. In the central United States, the prairie vole, Microtus ochrogaster, and the eastern meadow vole, M. pennsylvanicus, occur in sympatry and can be distinguished on the basis of molar cusp patterns but are known to be exceptionally difficult to distinguish using external morphological characters. Using a combination of morphometrics, pelage color analyses, and phylogenetics, we explored which traits are most effective for species identification and whether these same traits can be used to identify the subspecies M. o. ohionensis. While we were able to identify six traits that differed significantly between M. ochrogaster and M. pennsylvanicus, we also found substantial measurement overlap which limits the utility of these traits for species identification. The subspecies M. o. ohionensis was particularly difficult to distinguish from M. p. pennsylvanicus, and we did not find any evidence that this subspecies forms a distinct genetic clade. Furthermore, the full species M. ochrogaster and M. pennsylvanicus did not form reciprocal clades in phylogenetic analyses. We discuss several possible reasons for these patterns, including unrecognized variation in molar cusp patterns and/or localized hybridization. Overall, our results provide useful information that will aid in the identification of these species and subspecies in the future, and provides a case study of how genetics, morphometrics, and fur color analyses can be used to disentangle signatures of evolutionary history and hybridization.

Classifying species based on distinct, diagnostic traits is a central goal in the field of systematics. In some cases, however, phenotypic and genotypic traits can be ambiguous or misleading, posing challenges for species identification. For example, species can be difficult to tell apart if they are early in the speciation process, that is, if insufficient time has passed for phenotypic changes to accumulate or for genetic haplotypes to be sorted (Coyne and Orr 2004). Hybridization can further complicate this issue as species may produce offspring that appear intermediate between the two parental species or exhibit a mosaic of parental traits (Hubbs 1940; Rieseberg and Ellstrand 1993). Hybridization and short speciation times are both common features of rapid radiations, making them useful for studying these phenomena. Among mammals, the genus Microtus (Rodentia; Cricetidae; Arvicolinae) is particularly well-suited for such studies, as it has been called “probably the most rapid mammalian radiation” (Bastos‐Silveira et al. 2012: 5309) with at least 58 extant species evolving within the last 2 million years (Mammal Diversity Database 2022). Many of these vole species are extremely difficult to differentiate using external characteristics, which could be due to a combination of recent divergence, convergence, and hybridization. In this study, we focus on two vole species with overlapping ranges in the grassland biotic communities of North America: the prairie vole, Microtus ochrogaster; and the eastern meadow vole, M. pennsylvanicus.

The medium-sized prairie vole, M. ochrogaster (37–73 g), has a geographic range that covers a large swath of the central United States and south-central Canada (Laerm and Ford 2007; Fig. 1). Prairie voles prefer grassland habitats where they can construct runways and shallow burrows to hide from predators, and they undergo an annual reproduction cycle which peaks from May to October (Stalling 1990). Generally, prairie voles are described as having dark, grizzled pelage, which becomes more yellow at the ends (Stalling 1990). However, the easternmost subspecies, M. o. ohionensis, has distinctly light ventral pelage and dark dorsal pelage (Laerm and Ford 2007). This subspecies inhabits Ohio, western West Virginia, and eastern Kentucky, and its range is adjacent to the subspecies M. o. ochrogaster to the west. Although the geographic boundary line between these two subspecies has never been well-defined, proposed range maps appear to use the Licking River in Kentucky as the subspecies boundary line (Hall 1981; Adams et al. 2017). Nine subspecies of M. ochrogaster are currently recognized (Mammal Diversity Database 2022) and can be identified based on minute morphological differences and geography (Hall 1981), but M. o. ohionensis was the only subspecies found to be genetically distinct in a recent range-wide microsatellite study (Adams et al. 2017).

Fig. 1.

—Locality map of all samples included in this study. The geographic ranges of M. ochrogaster and M. pennsylvanicus are shown (Ford et al. 2007; Cassola 2016a, 2016b; Jackson and Cook 2020).

Until recently, the meadow vole M. pennsylvanicus (sensu lato) was described as a single species with 28 subspecies ranging across most of North America. However, Jackson and Cook (2020) elevated the western and central clades to full species status bearing the name M. drummondii, and elevated the Florida clade to a full species with the name M. dukecampbelli. Thus, the eastern meadow vole as recognized today (M. pennsylvanicus sensu stricto; 13 subspecies) occurs only in eastern North America, from Georgia and South Carolina in the south to the far northern areas of Quebec. It is not yet clear how far west this species occurs, but the westernmost sample that has been confidently assigned to M. pennsylvanicus sensu strico was from southwestern Ohio (Jackson and Cook 2020); thus, the species is likely sympatric with the prairie vole M. ochrogaster in Ohio, Kentucky, and West Virginia (Fig. 1). Like the prairie vole, the eastern meadow vole lives in grasslands, although it prefers more dense, moist vegetation (Miller 1969). The eastern meadow vole also has an overlapping breeding season and is similar in size to the prairie vole (25–65 g), but typically has lighter pelage, with brownish dorsal hair and silver-gray ventral hair (Ford et al. 2007).

Two other Microtus species live in the eastern United States—the woodland vole M. pinetorum and the rock vole M. chrotorrhinus—which are easily recognized by their unique habitats (deciduous forests and rocky areas, respectively) and by the short tail length of M. pinetorum and the yellow rostrum of M. chrotorrhinus. Microtus ochrogaster and M. pennsylvanicus, on the other hand, often occupy the same habitats and are exceptionally difficult to distinguish based on external characteristics (Adams et al. 2017). Species identification has been found to be particularly challenging in Kentucky (Krupa J.J., University of Kentucky, personal observation), which may be due to the presence of the subspecies M. o. ohionensis in the region; Adams et al. (2017: 184) noted that this subspecies looks “exceedingly like M. pennsylvanicus,” while Bole and Moulthrop (1942: 156) stated that M. o. ohionensis “in coloration so closely resembles Microtus pennsylvanicus that collectors…will likely overlook this animal.” The most reliable distinguishing characteristics for these two species are their mammae counts (six in M. ochrogaster and eight in M. pennsylvanicus) and molar patterns. The middle upper molar of M. ochrogaster (all subspecies) has four enclosed dentin triangles, whereas M. pennsylvanicus has five (Schwartz and Schwartz 2001). Coloration of the ventral fur is also often used in species diagnosis, as M. pennsylvanicus typically has a silver-gray ventrum compared to the buffy-yellow of M. ochrogaster. However, the subspecies M. o. ohionensis was described as having light ventral fur more similar to M. pennsylvanicus (Bole and Moulthrop 1942). Thus, where fur color is ambiguous, biologists might also use tail length to aid in species diagnosis; the tail of M. pennsylvanicus is typically greater than or equal to twice the length of the hindfoot, while the tail of M. ochrogaster is typically less than twice the length of the hindfoot (Schwartz and Schwartz 2001). Several other distinguishing features have been proposed including a coarser pelage in M. ochrogaster and the presence of five plantar tubercles in M. ochrogaster versus six in M. pennsylvanicus; however, these traits have been shown to be unreliable at times (Bole and Moulthrop 1942; DeCoursey 1957; Miller 1969; Henterly et al. 2011).

Here, we explore morphological and genetic patterns of diversity in a region where M. ochrogaster and M. pennsylvanicus are sympatric. In doing so, we aim to identify the external traits that field biologists might use to distinguish these species as well as the subspecies M. o. ohionensis. We address these questions by incorporating multiple lines of evidence, including morphological measurements, fur color, and mitochondrial and nuclear DNA sequence data.

Materials and Methods

Examination and identification of specimens.—

We examined 66 Microtus specimens cataloged in the University of Kentucky Vertebrate Teaching Collection (Supplementary Data SD1). Sex, locality information, and five standard external measurements (head-body length in mm [HB], tail length in mm [TL], hind-foot length in mm [HF], ear length from notch in mm [E], and body mass in grams [M]) were transcribed from specimen tags. If exact coordinates (latitude and longitude) were not available, the specimen was georeferenced using the verbatim locality description and the GEOLocate Web Application (Rios and Bart 2010) and Google Earth Pro (Google 2020).

A dissecting microscope was used to examine the dentition of each specimen. Specimens were classified as adults if they had fully erupted permanent dentition; specimens without fully erupted permanent dentition were considered juvenile and not included in morphological analyses. Specimens were further classified as M. ochrogaster or M. pennsylvanicus based on the presence of four dentin islands on the upper, middle cheek tooth (M2), or five dentin islands, respectively. Following this approach, 31 specimens were classified as M. pennsylvanicus adults and 30 were classified as M. ochrogaster adults (Supplementary Data SD1). Where possible, individuals were further assigned to an M. ochrogaster subspecies on the basis of locality: six specimens of M. ochrogaster from east of the Licking River in Kentucky were classified as M. o. ohionensis; and nine specimens of M. ochrogaster collected outside of central Kentucky were classified as M. o. ochrogaster. All other specimens could not be assigned to a subspecies due to ambiguity in the described subspecific boundaries.

Morphological analyses.—

Morphological measurements were inspected for normality and outliers using QQ plots and Shapiro–Wilk tests, performed in R using the functions ‘qqnorm’ and ‘shapiro.test’ in the stats package (R Core Team 2012). To visualize overall patterns of morphological diversity in this vole system, we performed a principal component analysis (PCA) on all adult samples. The five standard external field measurements (HB, TL, HF, E, and M) were used as input for both analyses, which were conducted using the command ‘prcomp’ from the stats R package. The principal components were visualized using the R package ggbiplot (Vu 2015).

To test whether M. ochrogaster and M. pennsylvanicus differed significantly in any particular external trait, we performed a series of t-tests using the function t.test in the stats R package (R Core Team 2012). For measurements with non-normal distributions, we used a Wilcoxon rank-sum test (function wilcox.test in the stats package) instead of a t-test. We analyzed the five external field measurements as well as four measurement ratios (TV:HB, TV:HF, HF:HB, and E:HB). Only adult specimens were included, and a Bonferroni correction was used to adjust the P-value cutoff for significance to 0.006 (= 0.05 with nine traits). The same approach was also used to determine whether any of the traits were significantly different between males and females of either species.

To determine whether any of our morphometric traits were significantly different between subspecies, we used a reduced data set containing only the individuals that could be assigned to M. o. ochrogaster (n = 9) or M. o. ohionensis (n = 6). Again, only adult specimens were included, and a Bonferroni correction was used to adjust the P-value cutoff for significance to 0.006 (= 0.05 with nine traits). After all analyses, boxplots were generated using the R package ggplot2 (Wickham 2016).

Quantification and analysis of fur color.—

We analyzed fur coloration using reflectance spectra for the ventral fur of each study skin. These measurements were taken using an Ocean Optics USB4000 spectrophotometer and PX-2 pulsed xenon light source (Ocean Optics, Dunedin, Florida) with a wavelength accuracy of ±0.38 nm. Measurements were taken perpendicular to the fur surface in a dark room. For each specimen, three reflectance spectra from different parts of the ventral fur were recorded and averaged. Each spectrum comprises an average of 10 readings collected in 200-millisecond intervals by the operating software SpectraSuite (Ocean Optics). Reflectance data are expressed as the percentage of reflectance relative to a white standard and are represented as total light intensity for a given wavelength.

We analyzed the reflectance spectra for values within the range of colors typically visible to mammals (between 400 and 700 nm) following Krupa and Geluso (2000). Reflectance spectrum data were used to calculate three variables typically used in color analysis: chroma, hue, and brightness. Chroma (sometimes called saturation) is a measure of purity or saturation of a color on a scale of 0–1; hue is a measure of the shape of the color spectrum and is represented as a 360° circle, with pure red, yellow, green, and blue colors located at 0° (or 360°), 90°, 180°, and 270°, respectively; and brightness is the total area under the reflectance curve from 400 to 700 nm. The reflectance spectra were imported into Microsoft Excel and chroma, hue, and brightness values were calculated using the equations provided in Smith (2014). We tested each variable for normality as above and used a t-test was used to test for significant differences in chroma, hue, and brightness between the two species, or an ANOVA to test for significant differences between subspecies. A Bonferroni correction was used to adjust the P-value cutoff for significance to 0.017 (= 0.05 with three measurements).

Ventral fur color was also analyzed using a low-cost photograph-based method. Photographs were taken using an Apple iPhone 7 under the same lighting conditions using a color reference card (ColorChecker Classic Mini; X-Rite, Inc., Grand Rapids, Michigan). We applied a standard white balance and tonal calibration to all photos using the color reference card in the software Adobe Bridge. All photos were then cropped to 2 cm wide and 4 cm tall, centered on the specimen’s ventrum. These cropped photographs were used as input to the R package ImaginR (Stenger et al. 2019), which extracts an average value of chroma, hue, and value from each photograph. Note that ‘brightness’ (calculated from the spectrophotometer data) differs from ‘value’ (measured here); a brightness measurement of 1.0 corresponds to pure white, whereas a value measurement of 1.0 corresponds to a pure, fully saturated color. The measurements of chroma, hue, and value were analyzed as above to identify significant differences between species and subspecies.

Genetic data collection and analysis.—

We collected genetic data from 43 Microtus specimens from the University of Kentucky Vertebrate Teaching Collection: 24 M. ochrogaster, 18 M. pennsylvanicus, and 1 M. drummondii (Supplementary Data SD1). Genomic DNA was extracted using a DNeasy blood and tissue kit (Qiagen, Germantown, Maryland). Fresh tissues were used when available; otherwise, dried tissue was removed from the foot and toe pads of study skins. Tissues originating from study skins underwent an additional 48 h of digestion with Proteinase K (step 1 of the DNeasy extraction protocol). PCR amplification was conducted for the mitochondrial cytochrome-b gene (Cytb) using previously published primers (L14724 and H15915 of Irwin et al. 1991); for exon 28 of the nuclear von Willebrand factor gene (Vwf) using previously published primers (W1 and W3 of Huchon et al. 1999); and for exon 10 of the growth hormone receptor gene (Ghr) using custom primers (forward: 5ʹ-GCG CAA GTA AGC GAC ATT ACA CC-3ʹ, reverse: 5ʹ-GTT CAG TTG GTC TGT GCT CAC-3ʹ). PCR parameters were as follows: an initial stage of 3 min at 94°C; 35 cycles followed the initial cycle in the following order: a denaturation stage of 30 s at 94°C, an annealing stage of 30 s at 54°C, and an extension stage of 1 min at 72°C; after 35 cycles, there was a final extension stage of 4 min at 72°C and the mixture was held at a constant 4°C. Raw PCR products were purified and sequenced on a Sanger instrument by Eurofins (Louisville, Kentucky).

Raw chromatograms were trimmed and edited by eye using 4Peaks (Griekspoor and Groothuis 2005). The software Geneious Prime version 2020.0.5 (Biomatters Ltd. 2020) was used to assemble the forward and reverse strands for each individual. Sites were considered heterozygous if (a) the chromatogram showed two peaks at a single location and the height of the minor peak was ≥80% of the height of the major peak, and (b) the heterozygous position was observed in both the forward and reverse strands.

Additional data and outgroups for phylogenetic analyses were downloaded from the NCBI GenBank sequence database. Specifically, we downloaded DNA sequences for all three loci from one individual of the European species M. agrestis and each of the following North American species: M. californicus, M. chrotorrhinus, M. longicaudus, M. miurus, and M. richardsoni. We also downloaded all available Cytb, Ghr, and Vwf sequences for M. ochrogaster and M. pennsylvanicus (Supplementary Data SD2). These were obtained through GenBank or directly from the authors of one previous study (Jackson and Cook 2020). The program Clustal W (Thompson et al. 1994) was then implemented within Geneious to create an alignment for each locus. Finally, nuclear alleles were phased using the program PHASE v.2.1.1 (Stephens et al. 2001) with default settings: 100 iterations, a thinning interval of 1, a burn-in of 100 iterations, and a probability threshold of 0.9.

Separate nexus files were generated for each of the three loci. We estimated maximum-likelihood phylogenies using the IQTree web server (Trifinopoulos et al. 2016). Within IQTree, a best-fit model of nucleotide substitution was automatically estimated via ModelFinder (Kalyaanamoorthy et al. 2017), and the ultrafast bootstrap algorithm was used with 1,000 replicates to estimate bootstrap support values. We also estimated a Bayesian phylogeny for each nexus file using MrBayes (Huelsenbeck and Ronquist 2001) with an Markov Chain Monte Carlo chain of 10 million generations sampling every 1,000 generations with a burn-in of 20%. All trees were visualized using FigTree (Rambaut 2007) with the European species M. agrestis used as the outgroup following the results of previous phylogenetic studies (e.g., Conroy and Cook 2000).

Results

Morphometric analyses.—

Three of our measurements (TL, HF, and E), and two measurement ratios (TL:HB and HF:HB) failed our tests of normality (Supplementary Data SD3). We report the results of our PCA here for descriptive purposes (Fig. 2), while acknowledging that the assumption of normality was violated in this analysis. The first principal component (PC1, explaining 59.6% of the total variance) was driven by a larger HB, HF, TL, and M in M. pennsylvanicus, while PC2 (explaining 16.7% of the variance) was driven by a large variation in E across all samples; however, this analysis also revealed substantial overlap between M. pennsylvanicus and M. ochrogaster samples, indicating that the two species are morphologically similar. Visual examination of all remaining PCs (PC3–PC5; not shown) did not provide further insight into these patterns.

Fig. 2.

—Bivariate plots of the first two principal components for the morphometric data set. Each point represents an individual. Ellipses denote the 95% confidence intervals around each cluster. Arrows indicate trait loading, that is, how strongly each trait influences the principal component. Two individuals (JJK-4078 and JJK-4355) are discussed in the text and indicated on the plot with black arrows.

We did not identify any significant differences between males and females for any of our measurements in either species. Males and females were, therefore, pooled in subsequent analyses. We did find significant differences between M. pennsylvanicus and M. ochrogaster for HB, TL, HF, and three measurement ratios: TL:HF, TL:HB, and E:HB (Table 1). However, these traits cannot be considered diagnostic, as the measurement ranges for all three traits overlapped substantially (Supplementary Data SD4). We found no significant differences between the subspecies M. o. ochrogaster and M. o. ohionensis (Table 2).

Table 1.

—Averages and standard errors for each measurement of both focal species (Microtus pennsylvanicus and M. ochrogaster). The P-values were calculated using a two-sided t-test or a Wilcoxon rank-sum test (depending on whether or not the trait values were normally distributed; see text) with a Bonferroni-corrected significance value of 0.006. Significant values are denoted by an asterisk, while nonsignificant values are denoted as n.s.

	M. pennsylvanicus (n = 31)	M. ochrogaster (n = 30)	P-value
Head-body length in mm (HB)	157.42 ± 2.25	143.53 ± 2.23	<0.0001 (*)
Tail length in mm (TL)	44.0 ± 1.15	32.9 ± 0.92	<0.0001 (*)
Hind-foot length in mm (HF)	20.58 ± 0.18	18.67 ± 0.24	<0.0001 (*)
Ear length in mm (E)	13.97 ± 0.37	13.87 ±0.23	0.70 (n.s.)
Body mass in grams (M)	38.93 ± 1.32	33.99 ± 1.64	0.031 (n.s.)
TL:HB ratio	0.28 ± 0.005	0.23 ± 0.0047	<0.0001 (*)
TL:HF ratio	2.14 ± 0.053	1.76 ± 0.044	<0.0001 (*)
HF:HB ratio	0.13 ± 0.0017	0.13 ± 0.0022	0.70 (n.s.)
E:HB ratio	0.088 ± 0.0019	0.097 ± 0.0021	0.0015 (*)

Table 2.

—Averages and standard errors for each measurement of the subspecies Microtus ochrogaster ohionensis and M. o. ochrogaster. The P-values were calculated using a two-sided t-test or a Wilcoxon rank-sum test (depending on whether or not the trait values were normally distributed; see text) with a Bonferroni-corrected significance value of 0.006. Significant values are denoted by an asterisk, while nonsignificant values are denoted as n.s.

	M. o. ohionensis (n = 6)	M. o. ochrogaster (n = 9)	P-value
Head-body length in mm (HB)	149.7 ± 5.66	148.4 ± 3.08	0.85 (n.s.)
Tail length in mm (TL)	37.67 ± 2.67	31.1 ± 1.39	0.06 (n.s.)
Hind-foot length in mm (HF)	18.83 ± 0.79	19.1 ± 0.39	0.76 (n.s.)
Ear length in mm (E)	14.33 ± 0.56	13.1 ± 0.35	0.10 (n.s.)
Body mass in grams (M)	37.67 ± 3.79	40.9 ± 2.81	0.51 (n.s.)
TL:HB ratio	0.25 ± 0.01	0.21 ± 0.007	0.055 (n.s.)
TL:HF ratio	2.00 ± 0.09	1.63 ± 0.06	0.009 (n.s.)
HF:HB ratio	0.13 ± 0.004	0.13 ± 0.004	0.56 (n.s.)
E:HB ratio	010 ± 0.004	0.09 ± 0.004	0.19 (n.s.)

Differences in fur color.—

Using spectrophotometer data from the ventral fur of M. ochrogaster and M. pennsylvanicus specimens, we found no significant differences between M. ochrogaster and M. pennsylvanicus for the three color metrics: brightness, chroma, and hue (Supplementary Data SD5). We also did not identify significant differences between species in the chroma, hue, or value of ventral fur color using a photo-based method in the R package imaginR (Fig. 3). Note that chroma (measured from both the spectrophotometer and from photos) and value (measured from photos) both failed tests for normality (Supplementary Data SD3), so Wilcoxon rank-sum tests were used for these analyses rather than t-tests. Among subspecies, we did find significant differences in chroma, hue, and value between M. o. ochrogaster and M. o. ohionensis, and between M. o. ochrogaster and M. p. pennsylvanicus (Fig. 3, Supplementary Data SD5). Visually, we noted that individuals in Kentucky often had intermediate ventral fur compared to the typically buffy-yellow fur typical of M. ochrogaster or the silvery-gray fur typical of M. pennsylvanicus (Fig. 4).

Fig. 3.

—Boxplots showing differences in ventral fur hue, chroma, and value between M. pennsylvanicus and M. ochrogaster using a photo-based method in the R package imaginR (top) and photos of ventral fur color from three example specimens (bottom). The spectrophotometer method yielded similar results shown in Supplementary Data SD4. Measurements that were significantly different are indicated by black brackets above the boxplots. Photos from top to bottom: catalog number 4421 (M. p. pennsylvanicus from Kentucky with classic silvery ventrum), catalog number 4368 (M. o. ohionensis from Kentucky), and catalog number 106 (M. o. ochrogaster from Nebraska with classic buffy/yellow ventrum).

Fig. 4.

—Maximum-clade credibility phylogeny of the mitochondrial gene cytochrome-b. Node values at major clade splits represent bootstrap support values (from IQTree analysis) and Bayesian posterior probabilities (from MrBayes analysis), separated by a slash. Individuals denoted with an asterisk were sequenced in Jackson and Cook (2020); see their Supplementary Data SD2 for specimen identification information. Individuals that were sequenced in this study were collected by JJK and are listed with their catalog numbers in the University of Kentucky Vertebrate Teaching Collection (Supplementary Data SD1). Sequence data for all other individuals were downloaded from GenBank (Supplementary Data SD2). For M. ochrogaster, subspecies names are included if known; individuals without a subspecies name might be M. o. ochrogaster or M. o. ohionensis.

Genetic analyses.—

Our alignment of the mitochondrial Cytb gene was 1,080 base pairs (bp) in length, and 337 of these sites were informative. The phylogenetic analysis of this gene recovered clades primarily associated with M. ochrogaster and M. pennsylvanicus; however, two individuals that were assigned to M. ochrogaster based on dentition (JJK-4355 and JJK-4078) fell within the M. pennsylvanicus clade. All other phylogenetic patterns in this analysis agreed with previous studies (Jaarola et al. 2004; Martínková and Moravec 2012), that is, M. pennsylvanicus formed a clade with M. drummondii, M. dukecampbelli, and M. longicaudus while M. ochrogaster formed a clade with M. chrotorrhinus, M. richardsoni, and M. miurus (Fig. 4). Notably, we did not find any evidence of M. o. ohionensis being genetically distinct (sensu Adams et al. 2017).

The nuclear Ghr gene tree recovered similar phylogenetic relationships, although the overall resolution was poor and many node support values were low (Fig. 5), likely because these gene contained fewer informative sites (45/454 sites in the alignment were informative). Note that the two-phased haplotypes from each diploid individual were treated as separate tips in this analysis. All four haplotypes from the M. ochrogaster samples JJK-4078 and JJK-4355 fell out within the M. pennsylvanicus clade, recapitulating the results of the Cytb analysis. While this phylogeny did recover weakly supported clades within M. ochrogaster, none formed a geographically circumscribed group that might represent M. o. ohionensis.

Fig. 5.

—Maximum-clade credibility phylogeny of the nuclear growth hormone receptor (Ghr) gene. These diploid data were phased; the two haplotypes for each individual are denoted by “a” and “b” at the end of the individual ID label. All other formatting is identical to Fig. 4.

Finally, the nuclear Vwf gene tree differed from both other gene trees (Fig. 6); it recovered M. ochrogaster and M. pennsylvanicus in a single clade to the exclusion of all other species (split node bootstrap support = 92, Bayesian posterior probability = 0.89) and no evidence of a distinct M. o. ohionensis clade. Only 28 of the 523 sites in this alignment were informative. Two poorly supported subclades were recovered within this M. ochrogaster + M. pennsylvanicus grouping: one that was dominated by M. ochrogaster individuals and one that was dominated by M. pennsylvanicus individuals; however, two individuals that were identified as M. ochrogaster on the basis of dentition fell within the predominantly M. pennsylvanicus subclade (JJK-4078 and JJK-4355 [note that these are the same individuals that were phylogenetically mismatched in both other trees]), five individuals identified as M. pennsylvanicus on the basis of dentition fell within the predominantly M. ochrogaster subclade (JJK-4370, JJK-4369, JJK-4363, JJK-4402, JJK-4421), and four individuals identified as M. pennsylvanicus on the basis of dentition had one haplotype from each subclade (JJK-4328, JJK-4321, JJK-4422, JJK-4420).

Fig. 6.

—Maximum-clade credibility phylogeny of the nuclear von Willebrand factor (Vwf) gene. These diploid data were phased; the two haplotypes for each individual are denoted by “a” and “b” at the end of the individual ID label. All other formatting is identical to Fig. 4.

Discussion

Accurate species-level identification——particularly in the field——is critical for most ecological and behavioral research. While mammalogists often rely on external measurements and pelage coloration to distinguish between species, our results suggest that these characteristics do not reliably distinguish the vole species M. ochrogaster and M. pennsylvanicus in our study region where they occur in sympatry. Despite statistical differences between six external traits, too much measurement range overlap occurred for reliable field identification (Table 1). Ventral fur coloration was also not a reliable field trait for distinguishing these two species in our study region. While M. ochrogaster is described as having yellow or buffy ventral fur and M. pennsylvanicus as having gray or silvery ventral fur, we found most individuals in Kentucky to have intermediate buffy-gray coloration. Thus, while ventral fur color might be a useful character in other parts of the their ranges, it cannot effectively be used to identify the two species where they co-occur in Kentucky. Our results align with previous studies that demonstrated limited utility of these external morphological characteristics in distinguishing the two species (DeCoursey 1957; Henterly et al. 2011). However, we acknowledge that this study did not consider mammae count, another external trait often used in species identification; M. ochrogaster is known to have six mammae, whereas M. pennsylvanicus has eight. This character may well be useful for species identification in our study region, but we were not able to include it due to ambiguity in mammae count caused by ventral incisions and stitching of study skins.

Dental characteristics are often used when external morphological characteristics are unreliable, and indeed our results suggest that molar cusp pattern is the most accurate morphological character used to distinguish M. ochrogaster and M. pennsylvanicus. We found that 94% of individuals that were assigned to species based on dentition were also recovered within their respective mitochondrial clade. The exceptions were two individuals from the Bluegrass Region: (1) JJK-4355 from Owen County has a molar cusp pattern consistent with M. ochrogaster (four dentin islands on the second molar) but was recovered in the M. pennsylvanicus clade; and (2) JJK-4078 from adjacent Henry County was also recovered in the M. pennsylvanicus clade and has an M. ochrogaster-like cusp pattern, although the fourth cusp has an abnormal protrusion that makes this species assignment more tentative (Supplementary Data SD6). In terms of other morphological characters, individual JJK-4355 fell with other M. ochrogaster individuals, whereas individual JJK-4078 was morphologically intermediate between M. ochrogaster and M. pennsylvanicus (Fig. 2). Two further individuals from the Cumberland Plateau were identified as M. pennsylvanicus on the basis of molar cusp patterns but as M. ochrogaster in the Vwf gene (JJK-4369 from Rowan County and JJK-4402 from Harlan County; additional genetic data were not available for these two individuals). There are several possible explanations for these findings. First, molar cusp count may simply not be a perfect diagnostic character. While M. ochrogaster is described as having four dentin islands on the second upper molar and M. pennsylvanicus is described as having five (Barbour and Davis 1974; Schwartz and Schwartz 2001), there is a great deal of variation in the shape and size of these islands (DeCoursey 1957) that were not considered in this study. It is possible that subtle ambiguities in cusp size and shape, as well as other sources of variation like molar wear or damage, could lead to misidentification. Second, these two individuals could be recent M. ochrogaster–M. pennsylvanicus hybrids that inherited an M. ochrogaster-like molar cusp pattern while exhibiting greater genetic similarities to M. pennsylvanicus due to backcrossing (see next section for a discussion of the evidence for hybridization). Furthermore, it is also possible that dentition is a good diagnostic trait for these species, but that our genetic analyses placed these individuals in the wrong clade due to incomplete lineage sorting (ILS), the low number of informative sites, or error during the lab work or bioinformatics. Overall, dentition was the most reliable morphological method to differentiate the two species in our study, but because this requires examining a cleaned skull it could present a significant problem for field studies where voles are caught and released. We recommend that future studies should evaluate mammae count as a likely diagnostic trait for field identification of live animals in this region of sympatry.

Our phylogenetic analysis of the mitochondrial cytochrome-b gene was in agreement with previous studies (Jaarola et al. 2004; Martínková and Moravec 2012) indicating that M. pennsylvanicus, M. drummondii, M. dukecampbelli, and M. longicaudus are closely affiliated while M. ochrogaster belongs to a clade with M. chrotorrhinus, M. richardsoni, and M. miurus. However, our nuclear phylogenies contradicted this result, placing M. pennsylvanicus and M. ochrogaster as sister species (Vwf) or else in a polytomy with M. longicaudus, M. chrotorrhinus, and M. richardsoni (Ghr). Dissonance between nuclear and mitochondrial phylogenies (i.e., mito-nuclear discordance) can be caused by a variety of biological factors including hybridization, ILS, and natural selection (Toews and Brelsford 2012). Additional genetic data (particularly from the nuclear genome) will be required to verify the phylogenetic relationships in this group.

It is worth noting that M. pennsylvanicus was, until the year 2020, taxonomically synonymized with the western meadow vole M. drummondii (Jackson and Cook 2020). Thus, previous studies of this species were likely conflating the genetic and morphological variation present in two different lineages. Many of the traits proposed by previous authors are likely sufficient to distinguish M. ochrogaster and the western meadow vole M. drummondii. For example, the westernmost sample of M. ochrogaster that we evaluated in this study had a distinctly yellow ventral fur, which contrasted highly with the fur color of M. drummondii; thus, fur color may be an acceptable diagnostic trait between these species in the western United States and Canada. Additional work is needed to clarify morphological differences between the eastern and western meadow vole and intraspecific variation within the eastern meadow vole.

Possible explanations for genetic discordance.—

In our genetic analyses, we identified several cases of gene tree discordance. First, three voles from Rowan County, Kentucky, clustered with M. pennsylvanicus in our analyses of the Ghr gene (JJK-4363) or Cytb (JJK-4370 and JJK-4421), but all fell out firmly with other M. ochrogaster individuals in the Vwf analysis. Additionally, we identified four individuals collected from Fayette County (catalog numbers JJK-4321 and JJK-4328) and Rowan County (JJK-4420 and JJK-4422) that were heterozygous at several positions in the Vwf gene, and when these heterozygous positions were phased we found that one of their Vwf haplotypes matched other M. ochrogaster individuals and the other matched other M. pennsylvanicus individuals. Finally, we found that the interspecific relationships varied from tree to tree, with M. ochrogaster and M. pennsylvanicus being recovered as sister species in the Vwf analysis, being placed in a clade with M. longicaudus, M. richarsoni, and M. chrotorrhinus in the Ghr analysis, and being placed in completely separate clades in the mitochondrial analysis.

There are several possible explanations for each of these instances of discordance. First, many of these patterns might be explained by ILS (the stochastic retention of ancestral polymorphisms in some but not all lineages). ILS is most often seen in loci that evolve slowly; indeed, both of our nuclear loci (Ghr and Vwf) had fewer informative sites and less resolution relative to the mitochondrial locus Cytb. It is also possible that these patterns could be due to hybridization at either recent or ancient timescales. Hybridization has been observed between several species of Microtus (Bastos-Silveira et al. 2012; Kovalskaya et al. 2014), and is particularly likely among species that are parapatric or sympatric, but the extent to which hybridization occurs across the genus is not well understood. Only one study to date——a microsatellite study of voles in Illinois——has given evidence of admixture between M. ochrogaster and M. pennsylvanicus (Douglas et al. 2021), but their nuclear data suggested that it is quite rare (2 individuals out of 314). Finally, it is possible that these patterns were induced by human error, either in the labwork phase (e.g., caused by interspecific contamination) or in the bioinformatics phase (e.g., failure to identify and remove possible paralogs).

Given that Microtus voles are a recent, rapid radiation, both ILS and hybridization are likely and could offer good explanations for our observed patterns (Knowles and Chan 2008). In our study, perhaps the strongest evidence for hybridization comes from the Vwf gene, where several individuals have mismatched haplotypes (one haplotype matching primarily M. ochrogaster individuals and the other matching primarily M. pennsylvanicus individuals), which is usually indicative of recent admixture. If these species have indeed hybridized, the fact that this pattern was not observed in the nuclear Ghr gene could be explained by the fact that this gene contained very few variable sites overall, or it could be explained by random chance, as hybrids beyond the F1 generation are not expected to have mixed ancestry at every nuclear locus (Mallet 2005). Likewise, the Cytb gene would not be expected to show patterns of admixture, because mitochondrial genes are haploid and maternally inherited; thus, hybrids will exclusively reflect the maternal genotype, rather than a mixture of both parents. Hybridization would also help explain the ambiguous external morphological traits in this system, particularly in central Kentucky where M. ochrogaster and M. pennsylvanicus have overlapping habitat ranges and breeding seasons.

Future researchers should collect additional genetic data to disentangle the effects of ILS and hybridization between M. pennsylvanicus and M. ochrogaster in this system. Modern genetic data sets derived from high-throughput sequencing technologies, which are thousands of times larger than the current data set, will have the power to provide fine-scale resolution on parentage, timing of hybridization, and the direction and magnitude of gene flow. Population-level data on chromosome count are also needed for both species. Chromosome counts of 2n = 46 and 2n = 54 have been documented for M. pennsylvanicus and M. ochrogaster, respectively (Hsu and Benirschke 1977). However, it is not known (a) whether these counts hold true across the entire species ranges, nor (b) whether these differences in chromosome number represent a significant barrier to reproduction. Other voles in the genus Microtus are capable of hybridizing despite different numbers of chromosomes (Bastos-Silveira et al. 2012; Kovalskaya et al. 2014).

Status of Microtus ochrogaster ohionensis.—

We did not find that M. o. ohionensis formed a distinct clade in our genetic analyses, which contrasts with the genetic results of Adams et al. (2017) who used microsatellite data to show that M. o. ohionensis was genetically distinct in population structure analyses. It is possible that the genetic loci we used in this study were simply not informative enough to detect this fine-scale population structure. It is also possible that our geographic sampling of M. ochrogaster was not large enough to capture population-level variation. However, one other important difference between this study and Adams et al. (2017) is that they did not include M. pennsylvanicus samples. If M. o. ohionensis is made up of a high proportion of hybrid M. ochrogaster × M. pennsylvanicus individuals, this would explain why M. o. ohionensis appeared to be genetically distinct in population genetic analyses (Adams et al. 2017) and why M. pennsylvanicus is particularly difficult to distinguish from M. o. ohionensis compared to other subspecies (Bole and Moulthrop 1942). Our morphometric results lend some support to this hypothesis, as they showed that the ventral fur color of M. o. ohionensis is more similar to M. p. pennsylvanicus than expected (Fig. 3, Supplementary Data SD5). Future research should focus on range-wide sampling of M. o. ohionensis to clarify this question.

Significance and directions for future research.—

This study provides a framework for understanding the genetic and morphological patterns that may be encountered by field biologists in the region of sympatry between M. ochrogaster and M. pennsylvanicus and provides several possible explanations for these patterns. We found, as in previous studies, that external morphological traits are not consistently reliable for identification or diagnosis of these species. Further complicating matters is the fact that in some locations these two species coexist in the same microhabitat. This is especially the case in Kentucky where M. ochrogaster and M. pennsylvanicus can both be trapped from the same localities. Molar cusp patterns are the most reliable traits for species identification, yet these too are occasionally equivocal and are nearly impossible to assess in the field. Despite these challenges, this study sheds new light on the evolutionary relationships between these two sympatric vole species, and demonstrates how multiple lines of evidence (morphometrics, fur color analyses, and genetics) can be used in an integrative framework to address questions related to species identification and natural history in other biological systems.

SUPPLEMENTARY DATA

Supplementary data are available at Journal of Mammalogy online.

Supplementary Data SD1.—Information for all specimens included in this study. Catalog numbers are associated with the University of Kentucky Vertebrate Teaching Collection. GenBank accession numbers are provided for specimens from which genetic data were collected from the cytochrome-b (Cytb), growth hormone receptor (Ghr), and von Willebrand factor (Vwf) genes. Specimens included in the fur color analyses are indicated by Xs in the corresponding columns. Measurement data for head-body length (HB), tail length (TL), hind-foot length (HF), and ear length from notch (Ear) are given in millimeters. Body weight data are given in grams. Latitude and longitude are given in decimal degrees. Descriptions for how species, subspecies, sex, and age-classes were assigned can be found in the main text.

Supplementary Data SD2.—GenBank accession numbers for the individuals added to the genetic data set from previously published studies. The genes used in this study were the mitochondrial cytochrome-b (Cytb), the nuclear growth hormone receptor (Ghr), and the nuclear von Willebrand factor (Vwf).

Supplementary Data SD3.—Results of tests of normality for each morphological measurement. Points on each QQ plots are colored according to species (red = M. pennsylvanicus and blue = M. ochrogaster). The P-values resulting from our Shapiro–Wilk tests are also included on each plot, with significant deviations from normality indicated by an asterisk (*).

Supplementary Data SD4.—Boxplots showing differences between M. ochrogaster and M. pennsylvanicus for all external morphological characters collected in this study. Traits that were found to be significantly different are indicated by black brackets above the plots.

Supplementary Data SD5.—Boxplots showing differences in ventral fur color between the species M. ochrogaster and M. pennsylvanicus (top two rows) or between the three subspecies M. o. ochrogaster, M. o. ohionensis, and M. p. pennsylvanicus (bottom row). Traits that were found to be significantly different are indicated by black brackets above the plots. Note that the photo-based method produced values for value, chroma, and hue, whereas the spectrophotometer-based method produced values for brightness, chroma, and hue. Results of the photo-based method subspecies comparisons are shown in Fig. 3 of the main text.

Supplementary Data SD6.—Photos of the molar cusp patterns for two specimens discussed in the text (JJK-4078 and JJK-4355). Typical cusp patterns for the second upper molar (M2) of M. ochrogaster and M. pennsylvanicus are shown on the left (figure based on Schwartz and Schwartz 2001) with the number of dentin islands indicated. Note that individual JJK-4078 has four dentin islands, but the fourth island has an abnormal posterior protrusion that might be considered a fifth cusp by some observers.