Phylogeny of the Perognathus longimembris Species Group Based on Mitochondrial Cytochrome-b: How Many Species?

Abstract

The Perognathus longimembris species group currently consists of 3 species and approximately 23 subspecies. Members of this species group have been of considerable interest to biologists in both physiological and ecological studies, and several members are threatened with extinction. No one has yet attempted a phylogenetic study including several populations from each of these 3 species to test the monophyly of these taxa and to elucidate the relationships among the species and subspecies. Mitochondrial DNA sequences from the cytochrome-b gene were used to assess the genetic variation among these species and to make a 1st approximation of the number of species that should be recognized taxonomically in this species group. Results indicate that the Arizona pocket mouse, P. amplus, and the San Joaquin pocket mouse, P. inornatus, are monophyletic entities, and should remain as single polytypic species, whereas the little pocket mouse, P. longimembris, is paraphyletic in the maximum-likelihood and maximum-parsimony phylogenies, suggesting that it might best be split into 2 polytypic species.

biogeography, cytochrome-b gene, molecular systematics, Perognathus amplus, Perognathus inornatus, Perognathus longimembris, phylogeny

The 3 currently recognized members of the Perognathus longimembris species group (Fig. 1) are at times and at some places the most abundant rodent species collected in lowland desert or shrub–steppe habitats in the southwestern United States (Hall 1946; Hoffmeister 1986), although they are often prone to great fluctuations in population numbers (Hall 1946). As the most abundant rodents, they exert a tremendous influence on the distribution and diversity of plant species in their communities through the mechanism of seed caching and seed predation, and in turn are preyed upon by a host of reptilian, avian, and mammalian predators (Brown and Harney 1993). This species group also is of interest because it includes a federally endangered subspecies (P. longimembris pacificus—Swei et al. 2003; references therein), and several federal or state subspecies of special concern (P. amplus ammodytes, P. a. amplus, and P. a. cineris—Arizona Game and Fish Department; United States Department of the Interior, Fish and Wildlife Service 1985, 1996; P. inornatus inornatus, P. i. psammophilus, P. l. bangsi, P. l. brevinasus, and P. l. internationalis—California Fish and Game Department; http://www.dfg.ca.gov/endangered/special_animals.html, last accessed 12 April 2005). The group contains the smallest heteromyid (P. l. pacificus—Hall and Kelson 1959), an example of strong selection for concealing coloration (across range of P. a. ammodytes and P. a. cineris—Benson 1933; Hoffmeister 1986), 1 species with wide variation in chromosomal arm number (FN, P. amplus—McKnight and Lee 1992; Patton and Rogers 1993; Williams 1978), and 1 species whose evolution is characterized by changes in the diploid number of chromosomes (P. inornatus—Patton and Rogers 1993; Williams 1978).

Fig. 1

Distribution of Perognathus amplus, P. inornatus, and P. longimembris, the 3 members of the Perognathus longimembris species group in the western United States, and their respective subspecies boundaries (based on opinions of distribution of Hall and Kelson [1959], Hoffmeister [1986], and Williams et al. [1993]). Two very restricted subspecies not sampled in this study, P. l. salinensis and P. l. tularensis, are not pictured. Numbers on map refer to collecting localities of specimens used in this study, and agree with numbers in  Appendix I and on the phylogenies (Figs. 2 and 3).

Given the potential importance of the longimembris species group to population and community ecology, comparative physiological ecology, evolutionary studies, and conservation, it is desirable that the phylogenetic relationships within and among the 3 current species be examined. McKnight and Lee (1992) examined karyotypes of P. amplus and P. longimembris from Arizona, southern Utah, southern Nevada, and southeastern California, whereas McKnight (1995) produced a phylogeny for these same individual mice based on restriction fragments of the entire mtDNA genome and direct sequences of a portion of the mtDNA cytochrome-b gene. Both of these studies were tests of Hoffmeister's (1986) hypothesis that P. amplus and P. longimembris in Arizona represent 2 ends of a ring species demonstrating circular overlap; both studies concluded that P. amplus and P. longimembris are well-differentiated species. Riddle (1995) included 2 specimens of P. longimembris, and single specimens each of P. amplus and P. inornatus in his study of molecular biogeography of pocket mice and grasshopper mice. He found the longimembris species group to be monophyletic, and his phylogeny reconstructed P. amplus as sister to the other 2 species. Finally, Swei et al. (2003) investigated the genetic structure of numerous populations of endangered or threatened subspecies of P. longimembris from southern California, using a single representative of P. amplus and 2 subspecies of P. inornatus as outgroups to root their tree. They found that the southern California populations of P. longimembris, despite much reduction in population size because of loss of habitat to development, retain substantial population subdivision and uniqueness, although this structure is poorly reflected in the subspecific taxonomy.

In this paper, a phylogeny is constructed for the longimembris species group based on 800 base pairs (bp) of the mitochondrial cytochrome-b gene. Specifically, I examine 4 hypotheses of relationship, including all possible combinations of sister relationships that have been proposed for the 3 taxa. First, the morphological (phenetic) similarity between P. inornatus and P. amplus (Best 1993), and the possession in both species of 2 pairs of medium-sized submetacentric chromosomes with satellites or secondary constrictions (Williams 1978) suggest a sister-group relationship between these 2 species. Second, the extensive similarity of karyotypes in P. amplus and P. longimembris in diploid number (McKnight and Lee 1992) unites these 2 species to the exclusion of P. inornatus, which possesses several populations with karyotypes differing in diploid number (Patton and Rogers 1993). Third, the geographic distribution of the 3 species, timing of the development of potential geological isolating barriers (Colorado River and Sierra Nevada Mountains), and the empirical result of Riddle (1995) all suggest a sister-group relationship between P. inornatus and P. longimembris to the exclusion of P. amplus. The 4th category includes all the possible phylogenetic trees where the 3 species are not reciprocally monophyletic.

The species concept I use in this paper is a modification of the phylogenetic species concept of Mishler and Theriot (2000). The modification is the retention of the subspecies designation for taxa included within monophyletic groups, but the main criterion for species status is monophyly in gene trees. The identification of more closely related taxa as subspecies is preferred because it provides an easy mechanism in the taxonomy to show some of the history of the group. For instance, application of this modified phylogenetic species concept retains the names Perognathus amplus amplus, P. a. cineris, and P. a. taylori, which are more revealing about the relationships among these taxa. An application of the unmodified phylogenetic species concept would produce the names P. amplus, P. cineris, and P. taylori, which reveal nothing about relationship, and do not in themselves show that they are closer to each other than either is to P. flavus, for example.

Materials and Methods

DNA sequences in this paper either were obtained from sequencing efforts herein, were received as electronic sequence data (from J. Patton, Museum of Vertebrate Zoology, University of Claifornia–Berkeley), or were downloaded from the National Center for Biotechnology Information public database (for source of data and National Center for Biotechnology Information accession numbers see  Appendix I). Tissue samples were removed from humanely killed specimens according to animal handling guidelines approved by the American Society of Mammalogists (Animal Care and Use Committee 1998), and in accordance with animal use protocols on file at the University of Illinois, University of California–Davis, and Southwest Missouri State University. For all but 1 sequence, the 5′ end of mitochondrial cytochrome b was amplified using classical polymerase chain reactions between the primer pair gludgL and cytb2H (Palumbi et al. 1991) and was sequenced by using Sequenase (United States Biochemical, Cleveland, Ohio; for details see McKnight [1995]). The remainder of the molecule was amplified by using the primer combination gludgL and cytb3H (Palumbi et al. 1991) and sequenced by using dye-terminator chemistry on an ABI 377 automated sequencer (Applied Biosystems, Foster City, California). The sequence of P. l. gulosus was obtained through polymerase chain reaction by using the same primers (gludgL and cyb3H), then sequenced with dye-terminator chemistry on an ABI 310 sequencer (Applied Biosystems). Sequences produced from different directions were manipulated and joined by using GeneJockey (Biosoft, Cambridge, United Kingdom), and aligned by using Clustal-X (Thompson et al. 1997).

Phylogenies were constructed by using maximum-likelihood (ML) and maximum-parsimony (MP) methods in PAUP* 4.0b 10 (Swofford 1998). The computer program Modeltest 3.06 (Posada and Crandall 1998) was used to parameterize the ML search. The suggested model was TrN+I+G, with base frequencies of A = 0.2857, C = 0.2780, G = 0.1410, and T = 0.2953; transition rates of A/G = 4.1404 and C/T = 6.9454; transversion rates for all transversions of 1.0; proportion of invariant sites = 0.3557; and Γ = 1.1035. Bayesian posterior probabilities were found by using MrBayes 2.01 (Huelsenbeck and Ronquist 2001), performing 1 million generations sampling 1 tree from the hottest of 4 chains every 100th generation; the chain temperature differential was 0.2, and the other Bayesian parameters were 6 character states, Γ distributed rate variation, and other starting parameters set as suggested by Modeltest. To evaluate the likelihood of the 3 possible hypotheses of sister relationship, I generated constraint trees using MacClade 4.03 (Maddison and Maddison 2001), found ML trees conforming to these constraints, and tested them against the ML tree with the Shimodaira–Hasegawa test in PAUP*. Unweighted MP searches were heuristic; starting trees were found by 10 replicates of random taxon addition; tree-bisection–reconnection branch-swapping was performed on 4 trees held at each step. Decay indices were calculated from a consensus tree of the 9 MP trees by using the program TreeRot version 2c (Sorenson 1999). Kimura 2-parameter genetic distances (Kimura 1980) were calculated between taxa by using PAUP*; this statistic was chosen to facilitate comparison to previous studies and opinions regarding genetic distance and species status (Bradley and Baker 2001). All trees were rooted by using P.flavus and P. apache as outgroups.

Results

All 24 specimens in this study have 800 bp of usable sequence from cytochrome b except P. l. virginis Beaver Dam, which is missing 54 bp from the middle of its sequence, and P. l. gulosus, which is missing 38 bp from the 5′ end. The ML phylogeny produced from these data (Fig. 2) and the topologically identical MP bootstrap (not shown) are inconsistent with all 3 possible hypotheses of sister relationship; on this tree P. amplus is monophyletic and sister to the remaining 2 species, and P. inornatus also is monophyletic but arises from within P. longimembris, making the latter species paraphyletic.

Fig. 2

Maximum-likelihood analysis based on 800 base pairs of mitochondrial cytochrome b from specimens of Perognathus amplus, P. inornatus, P. longimembris, and the outgroup taxa P. flavus and P. (flavescens) apache from the western United States, −ln likelihood = 4715.86881; model = GTR+Γ +I; alpha = 1.1035. Numbers preceding taxon names correspond to collecting localities in Fig. 1 and  Appendix I. Numbers above or near branches are Bayesian posterior probabilities (9,001 trees out of 1,000,000 generations). A majority rule consensus tree of 1,000 bootstrap iterations under maximum parsimony is topologically identical to this tree. Edwards AFB is Edwards Air Force Base in southern California.

On the ML (Fig. 2; −ln likelihood = 4715.86881) and MP (Fig. 3) phylogenies produced from these data, the longimembris–inornatus sequences consistently form 3 groups: a group containing P. l. virginis Mesquite, P. l. arizonensis, and P. l. virginis Beaver Dam (monophyletic in MP; not monophyletic in ML and MP bootstrap); a monophyletic P. inornatus; and a sister group to P. inornatus containing the remaining 7 subspecies of P. longimembris. The basal group contains specimens of P. longimembris from northern Arizona, southern Utah, and southeastern Nevada; these 3 samples are arranged on both the ML tree and MP bootstrap tree as 2 sequential branches from the main trunk (Fig. 2), but in the strict consensus tree of the 9 equally most-parsimonious trees (Fig. 3; 819 steps, consistency index = 0.6129, homoplasy index = 0.3871), they formed a single monophyletic clade. Similarly, in the ML searches constrained to make the traditional species monophyletic, this basal group was always sister to the remaining P. longimembris. The middle group contains all known variant populations of P. inornatus, including 2 unnamed karyotypic forms. One form with 52 chromosomes was discovered by D. Laabs (pers. comm.) near Edwards Air Force Base, California. The 2nd form is from Jepson Prairie Reserve, Solano County, California, and is represented here by a single specimen that was found freshly dead, so it could not be karyotyped, but based on geographical grounds it probably represents the same population as a single individual from nearby Lake County that had 60 chromosomes instead of the 56 that is typical of P. inornatus neglectus from the western side of the Central Valley (Slayden 1985; Williams et al. 1993). Both of these forms currently are assigned to P. i. neglectus (Williams et al. 1993). The final group among the longimembris–inornatus includes representatives of 7 different named subspecies, the most divergent of which is P. l. bombycinus.

Fig. 3

Strict consensus of 9 equally parsimonious trees, 819 steps, consistency index = 0.6129, homoplasy index = 0.3871. Numbers above branches are maximum-parsimony bootstrap percentages (1,000 iterations); numbers below branches are decay indices from the program TreeRot (Sorenson 1999). Data for this analysis are from specimens of Perognathus amplus, P. inornatus, P. longimembris, and the outgroup taxa P. flavus and P. (flavescens) apache. Note that maximum parsimony reconstructs the 3 taxa P. l. virginis Beaver Dam, P. l. arizonensis, and P. l. virginis Mesquite as a monophyletic group sister to P. inornatus and the rest of P. longimembris. Numbers preceding taxon names correspond to collecting localities in Fig. 1 and  Appendix I. Edwards AFB is Edwards Air Force Base in southern California.

Some aspects of the phylogeny presented here are well supported, as shown by Bayesian posterior probabilities (Fig. 2) and MP bootstrap proportions and decay indices (Fig. 3). The entire longimembris species group is monophyletic; this result is found in all analyses performed with numerous other species and genera included. The populations assignable to P. amplus form a solidly supported monophyletic group. The combined longimembris–inornatus clade is well supported, as is the monophyletic P. inornatus. Although the ML and MP phylogenies are consistent in reconstructing a paraphyletic P. longimembris, this result is not statistically well supported. The low Bayesian posterior probabilities, MP bootstrap proportions, and decay indices at the crucial nodes (Bayesian: 58%, 88%; MP bootstrap: <50%, 51%; decay indices: 2, 3; Figs. 2 and 3) are indicative of the lack of support for this hypothesis, whereas the Shimodaira–Hasegawa test (Table 1) indicates that the ML tree is a statistically better hypothesis than 2 of the hypotheses of sister relationship, but not different from the hypothesis ((inornatus, longimembris) amplus).

Table 1

Results of Shimodaira–Hasegawa test of 3 possible hypotheses of sister relationship, and the maximum-likelihood phylogeny in Fig. 2. All are species of the genus Perognathus. Asterisk indicates significant at P < 0.05.

Tree	–In likelihood	Difference from best	P
Maximum likelihood	4715.86881	(Best)	NAa
(((amplus, inornatus) longimembris) outgroup)	4735.68232	19.81351	0.011*
(((amplus, longimembris) inornatus) outgroup)	4734.56364	18.69484	0.018*
(((longimembris, inornatus) amplus) outgroup)	4718.54449	2.67569	0.547

a

NA, not available.

Discussion

Number of species in this species group.—What do the data and the phylogeny presented here indicate about the number of species in the longimembris species group? Is it best to continue recognizing the 3 existing species, or should the number be increased or decreased to best represent the inferred evolutionary history of this group? A full answer to this question should await more data, perhaps from nuclear gene regions, but the data at hand, and other aspects of the biology and distribution of these mice, can support some conclusions.

First, taxa currently designated P. amplus certainly are distinct from all other members of the species group; they form a well-supported monophyletic group separate from other taxa (Figs. 2 and 3), and as a group they differ from the other ingroup taxa in this study by an average Kimura 2-parameter distance of 16.60% (range: 13.74–19.80%). It might be worth considering elevating the 3 taxa within P. amplus to species status, because they differ from each other by very high genetic distances, suggesting a long history of independent evolution (amplus versus cineris, 11.80%; amplus versus taylori, 11.60%; cineris versus taylori, 12.44%—Bradley and Baker 2001). Plus, P. a. cineris geographically is isolated from the other 2 taxa. However, there is evidence of extensive introgression between P. a. amplus and P. a. taylori (Hoffmeister 1986; McKnight 1995) so, despite the fact that reproductive isolation or its absence is not a criterion in the species concept used here, it is best to await more data. P. amplus also is allopatrically distributed relative to the other 2 species except at 1 place (Wenden, La Paz County, Arizona), where no evidence of hybridization with P. longimembris has ever been found. Hoffmeister (1986) discusses 3 specimens he suspects are hybrids between P. amplus and P. longimembris collected in places other than Wenden. Examination of 2 of these 3 specimens indicated they were juveniles with domed crania and unworn 3rd upper molars. The fact that they were intermediate between true P. longimembris and true P. amplus in Hoffmeister's principal components analysis could have more to do with their juvenile status than any possible hybrid origin. Karyotypic differences among the subspecies of P. amplus are restricted to differences in number of autosomal arms (FN); all P. amplus have 2n = 56 chromosomes (McKnight and Lee 1992; Patton and Rogers 1993; Williams 1978). The populations in the northern part of Arizona (P. a. cineris and northern P. a. amplus (pergracilis)) have FN = 86. On the basis of its presence in all 3 members of the species group (Patton and Rogers 1993), I consider the 2n = 56, FN = 86 karyotype ancestral for the entire species group. More southerly populations of P. a. amplus have progressively higher FN (88, 90) until one moves into the range of P. a. taylori where FN continues to increase (90, 92) until it culminates at a value of 94 in the vicinity of Tucson, Arizona. McKnight (1995) proposed that FN variants intermediate between FN = 86 and FN = 94 are stabilized intergrades between the northern P. a. amplus (pergracilis) and the southern P. a. taylori. An additional karyotype with FN = 84 is known from P. amplus north of the Gila River and east of the Salt River in Arizona (formerly P. a. jacksoni); this form was not included in this study, but because of their different karyotype it is possible that they represent a 4th taxon in P. amplus.

All taxa currently identified as P. inornatus form a well-supported monophyletic group (Figs. 2 and 3), and thus are consistent with remaining a single species. These taxa differ from P. longimembris by an average Kimura 2-parameter distance of 8.07% (range: 6.89–10.13%) and among themselves by an average distance of 4.43% (range: 1.91–7.61%). As noted earlier, the named and unnamed taxa of P. inornatus all differ in diploid number of chromosomes: P. i. inornatus, 2n = 50; P. i. neglectus, 2n = 56; P. i. neglectus Edwards AFB (Air Force Base), 2n = 52; P. i. neglectus Jepson Prairie, 2n = 60? (D. Laabs, pers. comm.; Patton and Rogers 1993; Slayden 1985; Williams 1978). As Williams et al. (1993) point out, the taxonomic problems within P. inornatus and between it and P. longimembris are great, and will require substantial work and data from several aspects of their biology to clarify.

The number of species within P. longimembris is difficult to decide because of the parphyly of the taxon in the ML and MP trees, and the lack of significant superiority of these trees to those where P. longimembris is a monophyletic sister group of P. inornatus. A further complication to this decision is the disagreement between the ML tree (Fig. 2) on the one hand, and the strict consensus of 9 MP trees (Fig. 3) on the other. The different topologies of these trees result because of the disagreement between the conceptually different analyses of ML and MP, yet the data that address the discrepancy are few in number, and thus the MP bootstrap tree is similar in topology to the ML tree. In fact, when the ML tree topology is analyzed under MP criteria it is only 2 steps longer than an unconstrained MP tree. Nevertheless, the consistent split between the more ancestral P. l. arizonensis, P. l. virginis Beaver Dam, and P. l. virginis Mesquite and the remaining 7 subspecies may warrant taxonomic revision when sequence data from nuclear genes are applied to this question. The Kimura 2-parameter distances between P. l. arizonensis and the remaining taxa are high, and may be inflated because of some questionable automated sequencing base calls in the last one-half of its sequence (see Fig. 2), so these distances have been omitted in all comparisons. Yet, it is not surprising for P. l. arizonensis to show a greater genetic divergence in this larger data set, because it was among the most divergent populations of P. longimembris found by McKnight (1995). The 2 remaining northern taxa (P. l. virginis) differ from the remaining P. longimembris by an average Kimura 2-parameter distance of 8.17% (range: 7.47–8.99%). This value is slightly higher than the average divergence between P. inornatus and P. longimembris, and, although less than Bradley and Baker's (2001) suggested cutoff, supports splitting P. longimembris into 2 species. Among the remaining subspecies of P. longimembris, the most divergent is P. l. bombycinus with an average divergence of 5.21% (range: 4.56–5.77%), whereas P. l. gulosus is the only remaining subspecies with Kimura 2-parameter distances greater than 2% when compared to the remaining subspecies (average: 2.92%, range: 2.39–3.36%). The remaining 5 subspecies (bangsi, brevinasus, internationalis, longimembris, and pacificus) differ by an average Kimura 2-parameter distance of 1.48% (range: 0.25–2.42%). It is worth noting here that in addition to its greater sequence divergence from the other members of its clade, P. l. bombycinus is the only subspecies in P. longimembris to differ in karyotype from the rest of the species; it has a diploid number of 56 with 88 autosomal arms (FN), whereas all other P. longimembris that have been documented have 2n = 56, FN = 86 (McKnight and Lee 1992; Patton and Rogers 1993).

To summarize, despite large genetic differences on the one hand, and differences in diploid number of chromosomes on the other, it is the best reflection of the evolutionary history of these pocket mice to retain 1 species of P. amplus (with 3 subspecies: amplus, cineris, and taylori), and 1 species of P. inornatus (with 2 or 3 named and possibly 2 unnamed subspecies: inornatus, neglectus, psammophilusl, neglectus Edwards AFB, and neglectus Jepson Prairie). If further data support the ML and MP result obtained in this study, then P. longimembris would best be split into 2 species: 1 distributed in northern Arizona, southern Utah, and southeastern Nevada called P. arizonensis (with minimally 3 subspecies: arizonensis, virginis Beaver Dam, and virginis Mesquite), and P. longimembris containing the remaining subspecies currently named in this species.

Completeness of coverage.—The taxa used for this study are a nearly complete sampling of the named variation in the species group (Fig. 1). Although P. amplus historically was split into 7 subspecies, Hoffmeister (1986) performed an extensive morphological analysis of P. amplus and reduced that number to 4, and McKnight (1995) found that mtDNA restriction fragment data indicated the presence of only 3 major haplotype groups, with P. a. amplus and P. a. pergracilis forming 1 group, and P. a. cineris and P. a. taylori the other 2 groups. The variation in P. inornatus also is well represented here: all but 1 of the known karyotypic forms have been sampled, including 2 that have not yet been formally named. The missing form is the 2n = 56, FN = 88 karyotype of P. i. neglectus from the Carrizo Plain (Williams 1978). Although the variation in P. inornatus still needs substantial work before it is fully understood (Williams et al. 1993), the samples in this paper represent what currently is known. Coverage of variation in P. longimembris appears at first not to be nearly as complete, but this is not true for the following reasons. Of the 16 names currently listed by Williams et al. (1993), 8 are sampled in this study, and I use a name (P. l. virginis) that was synonymized with P. l. arizonensis by Hoffmeister (1986) in error in my opinion. Therefore, this study has 9 of 17 named taxa for P. longimembris. Of the 8 missing names, 5 are restricted in distribution only to their type localities (aestivus, kinoensis, salinensis, tularensis, and venustas), and most of these appear from their original descriptions to have been named on the basis of slight color differences. Two names probably are synonyms of P. l. bombycinus; in fact, it was Hoffmeister's (1986) opinion that P. l. pimensis was a synonym of P. l. bombycinus (although this synonymy was ignored by Williams et al. [1993]), and he also pointed out that these mice were nearly indistinguishable from P. l. kinoensis. The 2 remaining subspecies, P. l. nevadensis and P. l. panamintinus, are both broadly distributed subspecies that are distributed generally north of P. l. longimembris and west of P. l. gulosus, and I doubt that they are significantly different in cytochrome-b sequence from P. l. longimembris. Reasoning for this conclusion is that P. l. longimembris and P. l. gulosus are only slightly divergent in cytochrome-b sequence, despite the nearly 600-km distance between the collecting localities. This suggests that P. l. gulosus represents a set of populations that have expanded their distribution northward during the recent interglacial, and it is possible that when sequence data from P. l. nevadensis and P. l. panamintinus are included with this data set, they also will be found to have had a similar post-Pleistocene history and to be quite similar to P. l. longimembris.

Acknowledgments

Field assistance during various aspects of this work was provided by B. Seward, T. Jesse, D. K. McKnight, H. B. Shaffer, and T. Tomasi and the students of Southwest Missouri State University Desert Ecology Field Course. Financial assistance has come from American Museum of Natural History T. Roosevelt Memorial Fund, United States Army-Construction Engineering Research Lab, University of Illinois Research Board and Dissertation Research Fund, National Science Foundation grant BSR90-18686 to H. B. Shaffer, and Southwest Missouri State University Faculty Research Grant to MLM. Permission to trap on their lands was provided by The Navajo Nation, United States Department of the Interior (National Parks Service and Fish and Wildlife Service), and the Departments of Fish and Game of Arizona, California, Nevada, and Utah. Thanks to J. Patton for the pocket mouse sequences. This manuscript was improved by the comments of A. Mathis, J. Patton, H. B. Shaffer, and 2 anonymous reviewers.

Appendix I

Specimens examined.—The 24 specimens used in this study are listed below by the locality numbers used in Fig. 1, followed by the taxon, source of the specimen, and the Genbank accession number. Abbreviations: JLP, J. Patton, gifts of electronic sequence information; MVZ, Museum of Vertebrate Zoology, Berkeley, California, catalog numbers; CSULB, California State University Long Beach catalog number; PC, personal collection of MLM; Pea (and other Pe codes), codes linked to karyological specimens in the laboratory of the late M. R. Lee; UIMNH, University of Illinois Museum of Natural History, Urbana; AFB, Air Force Base; UC, University of California; WFB, Wildlife and Fisheries Biology.

1 P. a. amplus (pergracilis), JLP—MVZ 149945, AY697851; 2 P. a. cineris, PC Pea 66—UIMNH 60489, AY697854; 3 P. a. taylori, PC Pea 67—UIMNH 60541, AY697855; 4 P. i. inornatus, JLP—MVZ 182714, AY697849; 5 P. i. neglectus, JLP—CSULB 11039, AY697847; 6, P. i. neglectus Edwards AFB, JLP—MVZ 182709, AY697848; 7 P. i. neglectus Jepson Prairie, PC—UC Davis WFB-3950, AY697859; 8 P. l. arizonensis, PC Pel 32—UIMNH 60454, AY697861; 9 P. l. bangsi 1, Swei et al. 2003, AY152411; 10 P. l. bangsi 2, Swei et al. 2003, AY152409; 11 P.l. bangsi 3, Swei et al. 2003, AY152410; 12 P. l. bombycinus, PC Pel 49—UIMNH 60462, AY697858; 13 P. l. brevinasus, Swei et al. 2003; AY152412; 14 P. l. gulosus, PC field number MLM 148, AY697860; 15 P. l. internationalismSwei et al. 2003, AY152413; 16 P. l. longimembris 1, JLP—MVZ 145702, AY697850; 17 P. l. longimembris 2, Swei et al. 2003, AY152414; 18 P. l. longimembris 3, Swei et al. 2003, AY152415; 19 P. l. pacificus 1, Swei et al. 2003, AY152417; 20 P. l. pacificus 2, Swei et al. 2003, AY152416; 21, P. l. virginis Beaver Dam, PC Pel 31—UIMNH 60428, AY697856; 22 P. l. virginis Mesquite, PC Pel 33—UIMNH 60441, AY697857; 23 P. apache, PC Peap 11—UIMNH 60543, AY697852; 24 P. flavus, PC Pefl 7—UIMNH 60545, AY697853.

Author notes

Associate Editor was Robert D. Bradley.