OUP user menu

Taxonomic Relationship between Sorex disparand S. gaspensis: Inferences from Mitochondrial DNA Sequences

Judith M. Rhymer, Jessica M. Barbay, Heather L. Givens
DOI: http://dx.doi.org/10.1644/BER-003 331-337 First published online: 12 April 2004


The long-tailed shrew (Sorex dispar) occurs at upper elevations on wooded slopes in the Appalachian Mountains from North Carolina to Quebec. Size in long-tailed shrews is clinal, decreasing from south to north. A closely related species is the slightly smaller Gaspé shrew (Sorex gaspensis), known primarily from the Appalachians farther north in the Gaspé Peninsula of Quebec. Long-tailed shrews captured in northern Maine appear to fit within the size range of the Gaspé shrew which brings up the question: are these specimens small bodied long-tailed shrews at the northern end of their range, Gaspé shrews at the southern edge of their range, or are the 2 species part of the same continuous distribution and, in fact, conspecific? A morphological comparison with other studies indicates that a continuous cline cannot be ruled out. Phylogenetic analysis of mitochondrial DNA d-loop sequences showed that S. gaspensis and S. dispar cluster with no taxonomic or geographic structure, suggesting that they are conspecific.

Key words
  • latitudinal size cline
  • geographic variation
  • d-loop mitochondrial DNA sequence
  • Otisorex
  • Species at Risk
  • Appalachian Mountains

The long-tailed or rock shrew (Sorex dispar) occurs at upper elevations on wooded talus slopes in the Appalachian Mountains from North Carolina to Quebec (Kirkland and Van Deusen 1979). There is also a population in southeastern New Brunswick (Kirkland et al. 1979). The long-tailed shrew was considered monotypic until Schwartz (1956) described a larger southern Appalachian subspecies (S.d. blitchii). Based on variation in 18 skull and external morphological characteristics, Kirkland and Van Deusen (1979) placed the boundary between subspecies in the vicinity of the Pennsylvania, Maryland, and West Virginia borders (Fig. 1A). Despite this discrete taxonomic split, size in long-tailed shrews is actually clinal, with decreasing size from south to north (Kirkland and Van Deusen 1979). A closely related species is the slightly smaller Gaspé shrew (S. gaspensis), known only from the Appalachians in the Gaspé Peninsula of Quebec and New Brunswick, and from a disjunct population on Cape Breton Island in Nova Scotia (Fig. 1A). The Gaspé shrew was first described as paler dorsally than the long-tailed shrew, but more recent descriptions indicate that the only significant difference between the 2 species is size (French and Kirkland 1983; Kirkland and Van Deusen 1979). The range of Gaspé shrews is only about 1–3° latitude north of the known range of long-tailed shrews (Kirkland 1981).

Fig. 1

A) Geographic distribution of S. dispar subspecies S. d. dispar (diagonal lines) and S. d. blitchi (horizontal lines) and S. gaspensis (stippled areas—from Kirkland 1981). Area in rectangle is enlarged in Fig. 1B; B) Collection sites for long-tailed shrews from Maine and Quebec and Gaspé shrews from Quebec: 1 = Parc Mégantic, 2 = Mont Gosford, 3 = Tricky Bluffs, 4 = Ironbound Mountain, 5 = Green Mountain, 6 = Telephone Hill, 7 = Allagash Mountain, 8 = Clear Lake Mountain, 9 = Priestly Mountain, 10 = Rocky Brook Mountain, 11 = Deboullie Mountain, 12 = Gardner Mountain, 13 = Black Mountain, 14 = Deschene, Mont Albert, 15,16 = Charles Vallée, Indian Falls, Morency, Petite Cascapédia (localities in the Gaspé region are grouped because they are very close together. Triangle symbol represents Mt Carleton, New Brunswick (see Fig. 2).

Extensive sampling throughout a species' range is required to discriminate between discrete and clinal patterns of geographic variation (Barrowclough and Flesness 1996). It is possible that if the size cline in long-tailed shrews is projected to the latitude of the Gaspé Peninsula, it could account for differences in size between the 2 species. Kirkland (1981) suggests otherwise, but the northern portion of the range of S. dispar has been poorly sampled, so the actual slope of the cline is unknown. Kirkland and Van Deusen (1979) measured only 3 individuals from Maine, with the majority of their New England sample from New Hampshire. They also included specimens from the disjunct Cape Breton, Nova Scotia population in their morphological comparison of S. gaspensis with S. dispar.

We sampled long-tailed shrews in the Appalachian Mountains of northwestern Maine, north of the region previously sampled, and posed the question, are these specimens small long-tailed shrews at the northern edge of their range, Gaspé shrews at the southern edge of their range, or are the 2 species essentially part of the same continuous distribution and, in fact, conspecific? We compared external morphology and mitochondrial DNA (mtDNA) sequences of shrews collected in northwestern Maine to long-tailed shrews and Gaspé shrews collected in Quebec.

Materials and Methods

Sample Collection.—Shrews were collected from populations in the Appalachian Mountains in Quebec and Maine (Fig. 1B). Those from Québec were collected between 1995 and 2000 by J. Jutras. They were trapped in August and September using both pitfall and funnel traps (Kirkland and Sheppard 1994). Gaspé shrews (n = 14) were collected from 8 sites in the Gaspé Peninsula, Quebec (48°35′–9°04′N latitude), and long-tailed shrews (n = 4) from 2 sites in southern Quebec (45°16′–45°25′N), near the Maine-New Hampshire border (Fig. 1B).

Seventeen shrews were collected from populations at 11 sites in northwestern Maine (45°46′–46°58′N; Fig. 1B), as part of a statewide survey of rare species. Shrews were trapped in August and September of 2001, and all were captured using snap traps. Traps were set in rocky, wet areas, typically on talus slopes, at sites with elevations 294–577 m. Collection protocols conformed to the American Society of Mammalogists guidelines (Animal Care and Use Committee 1998).

Morphological analysis— Measurements of shrews from Quebec were done by S. St. Onge, and we measured the Maine shrews. Total length (mm), tail length (mm), and greatest skull length (mm) were measured with a ruler and calipers using standard procedures (Martin et al. 2001). We initially divided specimens into 3 groups: Maine long-tailed shrews, Quebec long-tailed shrews, and Quebec Gaspé shrews. For each group, we calculated the mean and standard deviation of total length, tail length, and greatest skull length. Measurements were compared using t-tests with the Bonferroni adjustment for multiple comparisons, to determine whether differences between groups were statistically significant. Multivariate discriminant function analysis was also used to compare morphological variation between long-tailed shrews and Gaspé shrews. Statistical analyses were done using SYSTAT v.10 software (SPSS Inc. 2000).

Genetic analysis.— DNA was extracted from liver tissue for the Maine specimens and from tongue muscle tissue for the Gaspé and long-tailed shrews from Quebec, using the QIAamp DNA Mini Kit (Qiagen Inc., Valencia, California) with the optional additional cleaning step. The 5′ region of the d-loop (displacement loop or control region) was amplified using primers Sorex tRNApro and Sorex CSB-F (Stewart and Baker, 1994a). The PCR protocol included an initial denaturation step at 94°C for 5 min; followed by 35 cycles of denaturation at 94°C, 45 s; annealing at 50°C, 1 min; and extension at 72°C, 1 min; with a final 5 min extension at 72°C. Each 50 µl PCR reaction included 10–50 ng DNA template, 1 × buffer, 1.5 mM MgCl2, 0.2 mM dNTPs, 0.5 µM each primer and 1.25 U Invitrogen (Carlsbad, California) Taq DNA polymerase. PCR products were purified using Nanosep (Gelman Laboratory, Ann Arbor, Michigan) microconcentration tubes (30K) and were sequenced on an ABI 373 Stretch automated sequencer (Applied Biosystems Inc., Foster City, California). Both strands were sequenced to clarify any ambiguities and were aligned using the Clustal algorithm in Sequence Navigator software (Applied Biosystems Inc.) and by eye.

The 5′ end of the Sorex d-loop has a tandem repeat region and we used the terminology coined in previous studies of Otisorex shrews (Stewart and Baker 1994a, 1994b). All tandem repeat copies, which ranged from 5 to 8 per individual, were compared to determine if there were mutational changes among repeats within individuals. For phylogenetic analyses, we compared only the unique flanking sequence on both the 5′ and 3′ ends of the repeat region, the last tandem repeat, and the imperfect repeat as these regions are considered homologous among individuals (Stewart and Baker 1994b). The high rate of concerted evolution among tandem repeats within an individual (homogeneity among repeats within an individual and heterogeneity among repeats from different individuals) makes it possible to compare only 1 copy of the repeat for each individual (Stewart and Baker 1994b; see also Fumagalli et al. 1996). Maximum likelihood analysis was used to generate a phylogenetic tree, with S. hoyi thompsoni and S. cinereus acadicus as the outgroup taxa. These species occur in Maritime Canada and are representative of 2 phylogenetic lineages of closely related Otisorex shrews (Stewart and Baker 1994b). Levels of resolution on nodes were estimated by 1,000 random bootstrap replications of the data. Maximum likelihood was done using the HKY85 + Γ + I substitution model (Hasegawa et al. 1985). The transition-to-transversion ratio, proportion of invariable sites, and shape parameter α of the gamma distribution were estimated from the data. Average distances were calculated among individuals within and among groups. Phylogenetic analyses were done using PAUP 4.0 b10 (Swofford, 2002).


Morphological analysis.— Results of t-tests indicated that there were no significant differences in measurements between long-tailed shrews from Quebec and Maine, so these specimens were combined for a comparison with Gaspé shrews. Total length and greatest skull length, but not tail length, of Gaspé shrews were significantly smaller than those of long-tailed shrews (Bonferroni adjusted P < 0.0001). Fewer cases were available for comparison using multivariate discriminant function analysis because more than half of the skulls of specimens collected in Maine were crushed during collection. Results were similar to those of the univariate analysis, however, with Gaspé shrews significantly smaller than long-tailed shrews (Wilks' lambda = 0.281, P < 0.0001) and most of the difference being attributable to total length and greatest skull length. Long-tailed shrews were correctly classified 90% of the time and Gaspé shrews, 82% of the time. However, a graphical comparison of S. dispar and S. gaspensis in this study with those from previous studies (French and Kirkland 1983; Kirkland and Van Deusen 1979) suggests that a continuous cline in morphology cannot be ruled out (Fig. 2). There is potential for systematic measurement bias when samples are measured by different observers, but the range of variability in each study shows substantial overlap in measurements from locality to locality.

Fig. 2

Morphological comparison of S. dispar from Maine and Quebec and S. gaspensis from the Gaspé Peninsula with results from previous studies in the mid-Atlantic (NY, NJ, PA ∼ 40°–44°N), NH (New England, primarily New Hampshire 44°16′N), Maine and Quebec (45°16′– 46°58′N), Nova Scotia (Cape Breton 46°75′N), NB (New Brunswick, Mount Carleton Park 47°23′N), Gaspé Peninsula 48°35′–49°04′N). Box represents X̄ ± 1 SD; vertical line is range, included where available (1 = Kirkland and Van Deusen 1979; 2 = this study; 3 = French and Kirkland 1983).

Genetic analysis.— Total length of the d-loop region amplified varied from 600 to 837 base pairs (bp), depending on the number of tandem repeats within individuals. In the case of 1 long-tailed shrew from Maine (Rocky Brook Mt sample number 84) and 1 Gaspé shrew (Charles Vallée 4), there were even differences in the number of repeats between DNA strands (Table 1). Similar results have been observed in several shrew species (Fumagalli et al. 1996; Stewart and Baker 1994a, 1994b). This phenomenon is speculated to result from the formation of secondary structure within the tandem repeats and/or from strand slippage during replication (Fumagalli et al. 1996; Stewart and Baker 1994a, 1994b).

View this table:
Table 1

Shared mitochondrial DNA d-loop haplotypes and numbers of tandem repeats for S. dispar and S. gaspensis (Mt. = Mountain).

HaplotypeLocalityPopulation and sample no.SpeciesNo. of tandem repeats
1MaineDeboullie Mt. 17S. dispar7
Rocky Brook Mt. 107S. dispar6
Priestly Mt. 14S. dispar6
QuebecCharles Vallée 4S. gaspensis5, 6
Charles Vallée 5S. gaspensis6
2QuebecMont Albert 1S. gaspensis6
Charles Vallée 3S. gaspensis5
Parc Mégantic 15S. dispar6
Parc Mégantic 16S. dispar6
MaineDeboullie Mt. 113S. dispar6
3MaineDeboullie Mt. 4S. dispar7
Deboullie Mt. 120S. dispar6
4MaineRocky Brook Mt. 84S. dispar7, 8
Telephone Hill 63S. dispar5
5MaineIronbound Mt. 73S. dispar7
QuebecMont Gosford 17S. dispar6
6MaineGreen Mt. 14S. dispar7
QuebecCharles Vallée 7S. gaspensis8
7QuebecCharles Vallée 13S. gaspensis7
Mont Mégantic 18S. dispar7

Analysis of the tandem repeats from each individual indicates that most copies were identical. We analyzed reduced datasets following Stewart and Baker (1994a, 1994b), using 284 bp of sequence for each individual. Sequences included the last tandem repeat (79 bp), the imperfect repeat (76 bp) and unique flanking sequence on both the 5′ (122 bp) and 3′ (7 bp) ends of the amplified region. Of the 284 characters, 226 were constant and 58 were variable. Only 16 of the variable characters were phylogenetically informative. The transition-to-transversion ratio was 5.78:1 for all taxa including outgroups and 31:0 for the ingroup alone.

Maximum likelihood analysis using the HKY85 + Γ + I mutation model indicated that long-tailed shrews from Maine and Quebec and Gaspé shrews clustered together with a bootstrap value of 87%; however, no species-level or geographic structure was apparent within this cluster (—Ln likelihood = 769.066) (Fig. 3). Under this model, the shape parameter (α) of the gamma distribution was estimated as 0.42 and the proportion of invariant sites (I) as 0.14. There were 7 shared and 15 unique haplotypes among 35 individuals, with one-third of long-tailed shrews from Quebec and Maine (7 of 21) having the same haplotype as 5 of 14 Gaspé shrews, although they did not necessarily have the same number of tandem repeats (Table 1). This close similarity between long-tailed and Gaspé shrews was supported by close genetic distances—the average distance between putative species was 1.6%, similar to those among individuals within the same species (∼1.7–1.8%—Table 2). These are relatively minor differences compared to genetic distances between S. dispar— S. gaspensis and the 2 outgroup Otisorex taxa (∼14–16%).

Fig. 3

Mitochondrial DNA d-loop phylogeny of long-tailed shrews and Gaspé shrews based on maximum-likelihood analysis using the HKY85 + Γ + I model (bootstrap values >50% above nodes). Shared haplotypes between ≥2 individuals are numbered (see Table 1 for details). QU = Quebec; ME = Maine; Mt. = Mountain.

View this table:
Table 2

Matrix of X̄ (±SD) genetic distances among long-tailed and Gaspé shrews and outgroup taxa (using HKY85 distances with rates assumed to follow gamma distribution with α = 0.42).

S. disparS. gaspensisS. c. acadicus
S. dispar0.0170.012
S. gaspensis0.0160.0070.0180.008
S. cinereus acadicus0.1410.0160.1520.018
S. hoyi thompsoni0.1640.0160.1610.0241.169


Univariate and multivariate analyses of morphological characters indicated that there is a significant difference in size between long-tailed shrews from northwestern Maine and southern Quebec and Gaspé shrews from the Gaspé Peninsula, with Gaspé shrews being generally smaller. The size cline of long-tailed shrews predicts that individuals from farther north will be smaller (Kirkland and Van Deusen 1979). Because long-tailed shrews and Gaspé shrews were collected from sites that were not continuous, the apparent size difference we found might merely be a reflection of clinal size differences that exist throughout the range of S. dispar. Geographic ranges of long-tailed and Gaspé shrews were once thought to be disjunct (Kirkland, 1981; Fig. 1A), but sampling for this study provided evidence that long-tailed shrews occur farther north than previously observed. Many of the Maine long-tailed shrew specimens were collected from sites beyond the proposed northern extent of the species' range (Kirkland and Van Deusen 1979), especially those from Deboullie, Black, and Gardner Mountains (Fig. 1B). This suggests that more thorough sampling in the Appalachians northeast of Maine through New Brunswick to the Gaspé Peninsula could indicate that the range of long-tailed and Gaspé shrews is actually continuous. A continuous range would suggest that differences between the 2 closely related taxa are likely the result of clinal variation in morphology and probably not to the evolution of separate species (Barrowclough and Flesness 1996). Similar size clines have been observed for Eurasian Sorex (Mezhzherin 1964). It has also been shown that size is not necessarily a good indicator of evolutionary divergence in small mammals, as it tends to be phenotypically plastic and, therefore, influenced by an individual's rearing environment (Patton and Brylski 1987; Ralls and Harvey 1985; Smith and Patton 1988).

Phylogenetic analysis showed that all 3 samples of shrews (Quebec and Maine long-tailed shrews plus Gaspé shrews) cluster together in 1 highly resolved group (bootstrap value of 87%) with no apparent structure within the group. In addition, average genetic distances between species are no greater than those among individuals of the same species, and some long-tailed shrews shared haplotypes with Gaspé shrews (Tables 1 and 2). These data support the hypothesis that S. d. dispar at the northern end of its range and S. gaspensis are conspecific. Ecologically, both species prefer cool, moist rocks along wooded mountain slopes and have similar diets (Kirkland 1981; Whitaker and French 1984).

A recent study of ornate shrews (S. ornatus) showed that there were broad-scale phylogeographic differences, with separate lineages from the southern, central and northern parts of the range (Maldonado et al. 2000). In that case, differences are concordant with past Plio-Pleistocene topographic barriers, and other vertebrate groups show similar geographic partitioning. The northern Appalachian Mountains were still covered by an ice cap in northern Maine, New Brunswick, and Quebec as recently as 12,000 years ago and emerged from the last period of glaciation only about 9,000–10,000 years ago (Davis and Jacobson 1985), so it is not suprising that no genetic geographic structuring has occurred among S. d. dispar and S. gaspensis populations. A study of mountain populations of white-toothed shrews (Crocidura russula) in Europe showed that those at higher altitudes on different mountains were more similar to one another than those lower on the slopes, possibly because they colonized high altitude sites at essentially the same time, resulting in spatial clustering of similar genotypes (Ehinger et al. 2002).

We obtained about 600 to 837 bp of sequence for each individual, but there were many identical mutations throughout the amplified region because tandem repeats tended to be the same within an individual. Although there were effectively only 284 bp of sequence used for phylogenetic analyses, these data were sufficient to show very close relationships between long-tailed and Gaspé shrews, yet very strong differences between these individuals and the 2 outgroup species, which represent 2 divergent lineages of Otisorex shrews (Stewart and Baker 1994b). Mean interspecific distances between other Otisorex taxa are similar to those we observed between S. dispar-S. gaspensis and S. hoyi thompsoni and S. cinereus arcticus (Table 2). We conclude that S. dispar from the northern part of its range and S. gaspensis are likely con specific. A more complete genetic analysis of S. dispar populations throughout their range, including S. d. dispar from New Brunswick, S. d. blitchi from the south, as well as S. gaspensis from the disjunct population in Nova Scotia, is underway (D. Stewart, pers. comm.). It is more likely that genetic divergence has occurred between populations or subspecies in the southern part of the range than in the northeastern United States and eastern Canada, because the southern Appalachians served as a refugium during the Wisconsin glaciation (Guilday et al. 1977).

Taxonomic considerations.— Helbig et al. (2002) maintain that populations at opposite ends of a cline should not be considered separate species. In accordance with the taxonomic convention of retaining the name of the species first described, all populations in this study should be recognized as S. dispar (Batchelder 1911) rather than S. gaspensis (Anthony and Goodwin 1924). It is generally accepted that even subspecies should be diagnosable, with a high degree of probability of being able to predictably assign individuals to a given taxonomic group (Barrowclough and Flesness 1996; Mayr 1969). Despite this, many subspecies have been recognized based only on mean differences, a practice that Patten and Unitt (2002) contend is especially problematic when considering populations along a cline, where any number of subspecies could be named if adequate sampling were done at various localities along the cline. If future analyses indicate a disjunction between current populations of S. gaspensis and those of S. dispar, the former could be considered for subspecific status as S. d. gaspensis. Regardless, only a few, small populations exist in Canada and either taxonomic designation should result in the retention of the current listing as a Vulnerable Species at Risk.


La musaraigne longicaude (Sorex dispar, Batchelder 1911) se retrouve à des altitudes élevées des Appalaches entre la Caroline du Nord et le Québec. La taille typique de la musaraigne longicaude diminue en fonction de la latitude, celles du nord étant plus petites que celles du sud. Une espèce très proche de la musaraigne longicaude, mais un peu moins grande, est la musaraigne de Gaspésie (Sorex gaspensis, Anthony and Goodwin 1924), connue seulement plus loin au nord, dans les Appalaches de Gaspésie et dans la région du Mont Carleton au Nouveau-Brunswick. Il en existe aussi une population à part au Cap-Breton, en Nouvelle-Écosse. La taille des musaraignes longicaudes prélevées dans le nord du Maine semble correspondre à celle de la musaraigne de Gaspésie, ce qui entraîne les questions suivantes: est-ce que ces spécimens seraient des musaraignes longicaudes de petite taille, vivant à la limite septentrionale de leur aire de distribution? S'agirait-il plutôt de musaraignes de Gaspésie, retrouvées à la limite méridionale du leur? Ou bien, est-ce que les deux types feraient en effet partie d'une même répartition continue et ainsi d'une même espèce? L'analyse morphologique de la longueur totale, de la longueur du crâne et de la longueur de la queue des spécimens en question montre que ceux de Gaspésie étaient significativement plus petits que ceux retrouvés dans le nord du Maine; néanmoins, une comparaison avec d'autres études indique qu'une continuité clinale ne peut pas être exclue. L'analyse phylogénétique des séquences d′ADN mitochondrial (dans la région d-loop) a démontré que nos prélèvements de S. gaspensis et de S. dispar se regroupent autour d'une valeur de bootstrap de 87%, sans structure taxinomique ni géographique, ce qui implique que les individus étudiés appartiennent à la même espèce.


We thank J. Jutras and S. St. Onge of Société de la Faune et des Parcs du Québec for providing tissue samples and measurements for the Quebec Gaspé and long-tailed shrew specimens. We thank P. Singer at the University of Maine DNA Sequencing Center for processing samples and R. Worvill of Acadia University for the summary in French. The University of Maine provided funding for the research. W. Glanz, C. Campbell, F. Servello and N. Jacobs provided helpful comments on an earlier version. This is article 2682 of the Maine Agricultural and Forestry Experiment Station.


  • Associate Editor was Eric A. Rickart.

Literature Cited

View Abstract