Comparative skeletal anatomy of salt marsh and western harvest mice in relation to locomotor ecology

Abstract

The salt marsh harvest mouse (Reithrodontomys raviventris) is an endangered species, endemic to the San Francisco Bay Estuary, that co‐occurs with the more broadly distributed species, the western harvest mouse (Reithrodontomys megalotis). Despite their considerable external morphological similarities, the northern subspecies of salt marsh harvest mice have relatively longer and thicker tails than do western harvest mice, which may be related to their abilities to climb emergent marsh vegetation to avoid tidal inundation. We used micro‐CT to compare post‐cranial skeletal anatomy between the salt marsh and western harvest mouse, to examine whether the salt marsh harvest mouse's restriction to brackish marshes is associated with skeletal adaptations for scansorial locomotion. We found that salt marsh harvest mice exhibited a deeper 3rd caudal vertebra, a more caudally located longest tail vertebra, craniocaudally longer tail vertebrae, and a longer digit III proximal phalanx than western harvest mice. These phalangeal and vertebral characteristics are known to decrease body rotations during climbing, increase contact with substrates, and decrease fall susceptibility in arboreal mammals, suggesting that the salt marsh harvest mouse may be morphologically specialized for scansorial locomotion, adaptive for its dynamic wetland environment.

We used micro‐CT to compare post‐cranial skeletal anatomy between the salt marsh and western harvest mouse, to examine whether the salt marsh harvest mouse's restriction to brackish marshes is associated with skeletal adaptations for scansorial locomotion. We found that salt marsh harvest mice exhibited a deeper 3rd caudal vertebra, a more caudally located longest tail vertebra, cranio‐caudally longer tail vertebrae, and a longer digit III proximal phalanx than western harvest mice. These phalangeal and vertebral characteristics are known to decrease body rotations during climbing, increase contact with substrates, and decrease fall susceptibility in arboreal mammals, suggesting that the salt marsh harvest mouse may be morphologically specialized for scansorial locomotion, adaptive for its dynamic wetland environment.

1. INTRODUCTION

Ecomorphological approaches contribute to our understanding of how organisms interact with their environment and help identify their preferred types of locomotion and habitat (Kingsley et al., 2017; Nations et al., 2019; Samuels & Van Valkenburgh, 2008), particularly for taxa in which behavioral data are sparse. Such is the case with the salt marsh harvest mouse (Reithrodontomys raviventris; hereafter SMHM), endemic to the wetlands of the San Francisco Bay Estuary, including the Suisun Marsh and San Pablo Bay estuaries (Figure S1; Fisler, 1965; U.S. Fish and Wildlife Service, 2010). These marshes experience a semidiurnal tide sequence (two pairs of high and low tides; Fisler, 1965) of elevated water levels, during which the SMHM rarely travel upland. Instead, during flooding, the SMHM climbs dense emergent vegetation in order to avoid inundation (Shellhammer, 1982; Smith et al., 2014). In the late 1960's, SMHM populations in the Corte Madera marshes experienced stark population declines and possible extirpation when mice were unable to climb to safety during winter super tides which completely submerged the area, thereby emphasizing the importance of emergent vegetation within and around marshes (Figure S1; Fisler, 1965). Additionally, moving through the highly structured vegetation and accessing elevated nesting locations provide refuge and protection from predators (Fisler, 1965; Johnston, 1956; Smith et al., 2014).

The SMHM is clearly adapted to its dynamic, wetland habitat considering its distribution, abundances, and persistence in the brackish marsh environment, but behavioral data on its specific locomotor preferences and performance capabilities are mostly limited to opportunistic, anecdotal observations. Nevertheless, if it is adapted to a scansorial existence, then its skeletal anatomy might show adaptations for climbing reflective of other scansorial and/or climbing taxa. Conversely, the congeneric western harvest mouse (Reithrodontomys megalotis; hereafter WHM), with which the SMHM co‐occurs throughout much of its range, is both more broadly distributed across the western United States (Webster & Jones, 1982) and more prevalent in upland grass habitats (Fisler, 1965; Smith et al., 2014; Sustaita et al., 2011). However, the species overlap where both upland grasses and marsh flora (e.g., Pickleweed, Salicornia pacifica; saltgrass, Distichlis spicata; tricorner bulrush, Schoenoplectus americanus) are present (Fisler, 1965; Smith & Kelt, 2019). The microhabitat differences (Bias & Morrison, 2006; Sustaita et al., 2011) and shared ancestry (Arellano et al., 2005; Statham et al., 2016) provide a “natural experiment” for comparing the degree of morphological specialization of the SMHM to the brackish marsh environment, while minimizing the confounding effects of geography and phylogeny.

To this end, we measured skeletal attributes of the SMHM and WHM that were most likely to reflect functional differences in their climbing or ambulatory locomotion. We note that we have focused our comparisons on the northern subspecies of the SMHM (R. raviventris halicoetes); previous studies have shown that the southern subspecies (R. raviventris raviventris) has a significantly shorter tail to body length ratio when compared to WHM (Statham et al., 2021). Although the two harvest mouse species considered in this study are very similar morphologically (Shellhammer, 1984; Statham et al., 2021; Sustaita et al., 2018), subtle differences in their skeletal structures, particularly in tail and limb characteristics typically associated with climbing ability (Cartmill, 1985; Hildebrand & Goslow, 2001), could have important locomotor consequences (Lovejoy et al., 1999). Climbing adaptations regarding the size of the feet as well as the length, and muscularization of the tail are often accompanied by modifications in skeletal anatomy.

Relatively long digits, prominent manus/pes friction ridges, a specialized forelimb shape, a long prehensile tail, and a small body size with respect to the substrate, serve to increase stability and decrease slips while climbing (Cartmill, 1985; Clemente et al., 2019; Fabre et al., 2013; Hildebrand & Goslow, 2001; Kingsley et al., 2017; Nations & Olson, 2015; Nations et al., 2019). The ability to wrap hands and feet around the substrate increases correctional torque abilities, thereby combating yaw, roll, and pitch forces exerted on the climbing animal (Cartmill, 1985). In many small mammal species, increased digit and relative phalanx lengths are correlated with increased arboreality and climbing ability in mammals (Ji et al., 2002; Kirk et al., 2008; Nations et al., 2019).

Long tails can also be used as balancing mechanisms and/or prehensile appendages to increases grip force (Cartmill, 1985; Hildebrand & Goslow, 2001; Horner, 1954). In two geographically isolated populations of deer mice, the forest cohort had significantly longer tails (both relatively and absolutely) than the ground dwelling population (Kingsley et al., 2017). Tail prehensility is also related to the number and robustness of the caudal vertebrae (Luger et al., 2020; Organ, 2010; Schmitt et al., 2005). Arboreal chameleons and primates with long, prehensile tails have a greater number of caudal vertebrae (either total or within the proximal region of the tail) than terrestrial, congeneric species (Luger et al., 2020; Organ, 2010; Schmitt et al., 2005). An increase in the number of craniocaudally shorter vertebrae, allows for increased flexibility in terms of narrow branch coiling (Luger et al., 2020; Schmitt et al., 2005). In laboratory mice induced to climb during adolescence, developmental plasticity led to 30% larger transverse processes on the transitional caudal vertebra compared to the control, non‐climbing group (Byron et al., 2011). New World monkeys that use their tails to wrap around branches have laterally longer transverse processes and larger hemal projections on distal caudal vertebrae than their non‐prehensile tailed relatives (Organ, 2010). The prominence of vertebral processes on all regions of the prehensile tail may function to increase the surface area available for muscle attachment (Luger et al., 2020). Additionally, due to their structure and arrangement, prehensile caudal vertebrae can tolerate a larger amount of torsion and bending than non‐prehensile vertebrae, especially in the distal region (Organ, 2010).

We used high resolution computed tomography (micro‐CT) to measure the following skeletal features in northern SMHM and WHM specimens, which are associated with arboreality in other species (Byron et al., 2011; Ji et al., 2002; Luger et al., 2020; Organ, 2010; Russo, 2015): the number of caudal vertebrae, proximal vertebral process prominence, tail vertebrae shape (craniocaudal length, width, and depth) and phalangeal indices. We hypothesized the relatively longer tail (Sustaita et al., 2018) and climbing behavior of the SMHM to be associated with longer, and more numerous, caudal vertebrae; more prominent vertebral processes; and longer relative phalanx lengths than the WHM. This research also has practical implications for wildlife conservation and management, because the most distinctive quantitative characteristics for distinguishing between the endangered northern SMHM and WHM in the field are tail length (relative to body length) and tail diameter (measured 20 mm distal to the rump) (Fisler, 1965; Shellhammer, 1982; Sustaita et al., 2018). However, a comprehensive analysis of underlying skeletal attributes has not been conducted.

2. MATERIALS AND METHODS

2.1. Specimen collection and micro‐CT scanning

We restricted our study to frozen carcasses obtained from the field within the past 18 years to avoid potential distortions to the tail and digits of historical museum skins during the specimen preparation and preservation process, and to minimize the potential effects of time. However, because the SMHM is a California state, federal, and IUCN listed endangered species, they are fully protected from lethal take. Thus, harvest mouse carcasses (n = 9 Reithrodontomys raviventris [one of which is of the southern subspecies, R. r. raviventris] and n = 8 R. megalotis) were obtained from the California Department of Fish & Wildlife and other state, federal, and private entities in possession of federal permits that work with SMHM. The carcasses (resulting from incidental mortalities) were opportunistically collected from July 2005 to August 2019 in trapping locations throughout the Suisun Marsh and North San Francisco and San Pablo Bay areas, with the exception of one specimen of the southern subspecies (R. raviventris raviventris) that was obtained from the South San Francisco Bay area. Species identities were confirmed genetically by comparison of cytochrome b gene sequences to a reference database of sequences from SMHM and WHM (Statham et al., 2016, 2021).

The specimens were sent to the Pratt Lab at California State University, Bakersfield for high resolution computed tomography (HRCT) scanning (SkyScan 2211, Bruker Corporation, Billerica, MA, USA). Samples were rotated 180° and scanned every 0.30–0.40 degrees with the x‐ray source set at 60 kV with a current of 500 μA. Images produced were 1920² pixels/image with resolution set between 25 and 35 μm depending on the size of the specimen. Specimens were too long for a standard scan, so to obtain a data set that included the whole animal, different portions of the animals were scanned and stitched together during the reconstruction step (InstaRecon Software, Champaign, IL, USA). An exception to these settings, one specimen was scanned at the University of Washington Friday Harbor Marine Laboratory micro‐CT scanning facility (Bruker SkyScan 1173 micro‐CT system; voltage: 65 kV, current: 123 μA, 2240² pixels/image, 24.87 μm² pixel size).

2.2. Volume rendering and qualitative analysis

The volume reconstructions from the micro‐CT scans were visualized using CTvox (version 3.30, Bruker micro‐CT; Figure 1), and the total number of caudal vertebral bodies composing the tail were enumerated for each individual. Identifying where the caudal vertebrae begin (cranially) is challenging because sacral vertebral number varies among species, and the boundaries between sacral and caudal vertebrae are not clear (Kingsley et al., 2017). Kingsley et al. (2017) numbered caudal vertebrae posterior to the sixth sacral vertebra in the Peromyscus spp. they studied. However, studies of other murid (Bab et al., 2007) and cricetid (Brito et al., 2021; Quintela et al., 2017) taxa have indicated that there are four sacral vertebrae. We determined the cranial start of the caudal vertebral sequence by locating the distinct gap between successive vertebral bodies, which occurred just beyond the caudal end of the 4th fused sacral vertebra (Figure S2). Because the distal‐most vertebrae are faint and difficult to see, two of us (KMW and DS) independently assessed the number of caudal vertebrae for each specimen. The values obtained for the numbers of caudal vertebrae were then compared. Where there was disagreement, we jointly reviewed the micro‐CT scans to reach a consensus. To compare the caudal vertebral and phalangeal morphology of the SMHM and WHM, the micro‐CT volumes were then uploaded into Seg3D (Seg3D, version 2.4.0, NIH NCRR). The “threshold” tool in Seg3D was used to create three‐dimensional, scaled mesh models of each specimen's skeleton from the micro‐CT scans (henceforth termed “isosurfaces”). Because of the variability in the quality of isosurfaces (even for the same specimen), we performed additional analyses to explore the consequences of this variability on downstream measurements, by building and repeatedly measuring multiple isosurfaces for selected specimens; this procedure, and the results, are described in the supporting information (Text S1).

FIGURE 1.

A three‐dimensional model from a micro‐CT scan of Reithrodontomys raviventris specimen (viewed in CTvox); (a) left‐lateral view; (b) right lateral view; (c) dorsal view; (d) ventral view.

2.3. Model‐based linear skeletal measurements

We exported isosurfaces as .obi files and used Geomagic Wrap (version 2012.0.2, 3D Systems) to refine the models before measurement. We scaled the isosurfaces to the appropriate image pixel size of each specimen's micro‐CT scan. We then smoothed the models using the “relax” and “smooth” functions to remove artefacts and noise generated by the scanning process (Geomagic Design X User Guide 2013). We successfully built models of sufficient quality for measurement for 17 specimens, on which we identified the following skeletal features (based on Byron et al., 2011; Ji et al., 2002; Kingsley et al., 2017; Kirk et al., 2008; Luger et al., 2020; Organ 2010) and performed the following skeletal measurements described below three times each (non‐consecutively) to mitigate errors in measurement.

2.3.1. Body size

Body size was incorporated into the analyses as a covariate (below) to account for the potential difference in overall body size between species. We used body length (excluding the tail) as a proxy for overall body size because body mass was unreliable, given the desiccated conditions of some of the carcasses. Due to variation in the postmortem positions of the specimens in the micro‐CT scans, we measured body length in a series of repeatable increments (Figure S3).

2.3.2. Vertebral metrics

Arboreal primates, for which ample published data is available, tend to have an increased number of caudal vertebrae in the proximal region of the tail (Organ, 2010). Considering that proximal caudal vertebrae articulate via zygapophyses, rotation around junction sites is limited, and multiple, sequential vertebrae are needed to permit flexibility in this region (Organ, 2010). Regarding the dimensions of each vertebra, arboreal species tend to have proximal vertebrae with higher aspect ratios (i.e., vertebrae that are shorter in craniocaudal length with relatively longer zygapophyses, transverse, and neural processes) than their terrestrial comparators (Luger et al., 2020). To capture variation in these aspects in the mice, we measured:

1. Craniocaudal length of caudal vertebrae: measured at one‐half the depth of the vertebra, from the cranial to caudal ends of the vertebral body (lateral view of organism; Figure 2a).

2. Caudal vertebra depth: measured at one‐half the craniocaudal length of the vertebra, from the ventral to dorsal side of the vertebral body, including the neural process (lateral view of organism; Figure 2b).

3. Caudal vertebra width: measured mediolaterally at one‐half the craniocaudal length of the vertebra, from the right to left side of the vertebral body (dorsal view of organism; Figure 2c).

4. Transverse process length: measured from the middle of the transverse process and vertebral body junction to the most distal point on the transverse process (from the ventral view of organism; Figure 2d).

FIGURE 2.

Western harvest mouse (Reithrodontomys megalotis) skeletal isosurfaces: Lateral view of 9th caudal vertebra showing (a) craniocaudal length and (b) dorsoventral depth measurements; (c) dorsal view of 9th caudal vertebra showing mediolateral width measurement; (d) ventral view of 2nd caudal vertebra showing transverse process length measurement; dorsal view of right manus showing (e) combined carpus and 3rd metacarpal length; (f) digit III proximal phalanx length; and (g) digit III intermediate phalanx length measurements.

From these measurements we computed the following indices for all specimens (n = 17): the widths and depths of the 2nd, 3rd, 9th, and 15th (chosen for consistency with previous studies) caudal vertebrae were divided by their craniocaudal lengths to obtain frontal and sagittal plane (respectively) “aspect ratios.” Furthermore, we also divided the depth by the width of these vertebrae to obtain transverse plane aspect ratios. Collectively, the above aspect ratios impart information on the relative robusticity of the vertebral elements in different planes that might be reflective of differences in loading regimes (e.g., the caudal vertebral structure of New World monkeys; Organ, 2010). Finally, for a subset of specimens (n = 12) we were able to obtain craniocaudal length measurements for most (if not all) of the caudal vertebrae along the length of the tail, which allowed us to test for differences across successive vertebral elements (after Kingsley et al., 2017).

2.3.3. Phalangeal metrics

High phalangeal indices functionally support grasping (especially in the narrow branch niche) and are associated with increasing arboreality in mammals (Ji et al., 2002; Kirk et al., 2008). To obtain these indices for the mice, we measured the following:

• vCumulative carpus and third metacarpal length: measured from the distal end of the radioulnar articulation to the most distal, central end of the third metacarpal. The distal end of the metacarpal (i.e., the metacarpophalangeal joint) was determined by either the flexion of the proximal phalanx or a visible indentation in the isosurface (view of organism varied based on manus positioning relative to the body; Figure 2e)
• viDigit III proximal phalanx length: measured from the center of the distal end of the third metacarpal to the proximal interphalangeal joint. The proximal end of the proximal phalanx was denoted by either the point flexion of the intermediate phalanx or indentations visible in the model separating the phalanges (if the intermediate phalanx was extended; view of organism varied based on manus positioning relative to the body; Figure 2f).
• viiDigit III intermediate phalanx length: measured from the center of the distal end of the proximal phalanx to the center of the distal end of the intermediate phalanx. Both endpoints were determined following the same methodology described above under Digit III proximal phalanx length (view of organism varied based on manus positioning relative to the body; Figure 2g).

The “phalangeal index” of digit III with respect to the cumulative carpus and metacarpal length was calculated after Kirk et al. (2008) as the sum of the proximal and intermediate phalanx lengths divided by the cumulative carpus and 3rd metacarpal length. Ratios of the proximal phalanx and intermediate phalanx each relative to the cumulative carpal and metacarpal length were also obtained for each individual (after Ji et al., 2002; Kirk et al., 2008).

2.4. Statistical analyses

For comparisons between species in body, vertebral, and phalangeal variables, we used average values across repeated measurements for each specimen. We then performed a principal component analysis (using function prcomp of R Package “stats;” R Core Team, 2021) to reduce the original set of 17 vertebral and phalangeal variables to 9 principal components (PC's) that explained ~98% of the variance for subsequent tests of differences between species, to retain the maximum amount of information possible, while maintaining a favorable cases: independent variables ratio. We recognize that our small sample sizes violate assumptions of parametric tests and restricted statistical power, and hence our ability to detect significant differences between species. As a result, we employed non‐parametric approaches using the lm.rrpp function of R Package “RRPP” (Collyer & Adams, 2021; see supporting Text S1 for details) to perform an ANOVA (assuming Type III sums‐of‐squares, set to 9999 permutations) on the first 9 PC's. RRPP analyses are robust to high‐dimensional data sets (Collyer & Adams, 2018); however, we felt that reducing the dimensionality via PCA (to essentially half the number of original variables) was warranted given the limited number of specimens available for analysis. The RRPP approach generalizes univariate ANOVA statistics to multivariate data by using the SS calculated from the trace of the sum of squares and cross‐products matrix (Collyer, 2021). Nevertheless, we also performed a traditional MANOVA in the permutation‐based framework of RRPP (based on 10,000 permutations) using function manova.update (Collyer, 2021; Collyer & Adams, 2021). We then complemented this analysis with a series of univariate ANOVA models including species, the species × body length interaction, and body length as predictors. If the interaction species × body length interaction was not significant, it was omitted from the formula and the analysis was repeated. Subsequently, if the effect of body length was also not significant, we removed from the model to avail residual degrees of freedom for testing the main effect of primary interest, “species.”

We performed a separate test for differences across vertebral elements and between species for the subset of specimens for which we obtained length measurements for most (if not all) of the caudal vertebrae along the length of the tail. This analysis was restricted to the first 25 (of ~30) vertebrae, due to missing values resulting from difficulties in accurately reconstructing the tiniest bony elements of the distal tail. Here again, we averaged the three trials for each specimen, and individual caudal vertebra lengths were treated as a succession of “repeated” measurements within each specimen, to account for the lack of independence among successive vertebrae. Furthermore, to account for potential differences in body size we divided the craniocaudal vertebra lengths by body length. We used the same lm.rrpp R function as described above (assuming Type III sums‐of‐squares, set to 9999 permutations), but in a mixed within‐ and between‐subjects ANOVA design, with “vertebral identity” included as the within‐subjects factor and “species” as the between‐subjects factor. The means squares of these terms were tested over those of the following to obtain the appropriate tests of effects: specimen nested within species, vertebral identity nested within specimen, and species × vertebral identity. A significant species × vertebral identity interaction would imply that the degree of difference between species in vertebral length differs depending on location along the length of the tail.

Because our SMHM sample includes one southern subspecies specimen (R. r. raviventris), and differences in external tail and body dimensions have been detected between the two SMHM subspecies (Statham et al., 2021), we repeated the analyses excluding this case to determine if and how overall differences between species are affected, for which we refer readers to the supporting information, noting that the results reported below did not differ substantively upon exclusion of the southern subspecies (Figures S4–S6 and S8; Tables S3–S5).

3. RESULTS

3.1. Overall differences in skeletal characteristics

There was no significant difference in body length between species, although the SMHM specimens tended to be larger (F _1,15 = 3.51, p = 0.0812; Table 1; Figure 3). Salt marsh harvest mice had significantly longer tails (F _1,15 = 1.87, p = 0.0251), but the difference was not significant (p = 0.143) when the marginally non‐significant effect of body length (F _1,14 = 1.55, p = 0.054) was included in the model (Table 1; Figure 3). The number of caudal vertebrae ranged from 26 to 31 among specimens (Figure 3) and did not differ significantly with body size (F _1,14 = 1.15, p = 0.860; Table 1) or between species (F _1,14 = 0.0232, p = 0.503), even upon removal of the non‐significant effect of body length (F _1,15 = 0.0582, p = 0.427).

TABLE 1.

Results of RRPP ANOVAs testing for differences between salt marsh (Reithrodontomys raviventris) and western (R. megalotis) harvest mice in body length, tail length, and number of caudal vertebrae.

	df	SS	MS	Rsq	F	Z	Pr(>F)
Body length
Species	1	116.31	116.31	0.189	3.51	1.37	0.0812
Residuals	15	497.55	33.17	0.811
Total	16	613.86
Tail length
Species	1	120.12	120.42	0.0924	2.38	1.09	0.143
Body length	1	221.08	221.08	0.170	4.38	1.55	0.054
Residuals	14	707.17	50.51	0.542
Total	16	1303.76
Reduced model
Species	1	375.51	375.51	0.288	6.07	1.87	0.0251
Residuals	15	928.25	61.88	0.712
Total	16	1303.76
Number of caudal vertebrae
Species	1	1.028	1.028	0.0324	0.469	0.0232	0.503
Body length	1	0.068	0.0676	0.00213	0.0308	−1.15	0.860
Residuals	14	30.70	2.19	0.966
Total	16	31.77
Reduced model
Species	1	1.00	1.00	0.0315	0.488	0.0582	0.427
Residuals	15	30.76	2.05	0.968
Total	16	31.77

Note: Non‐significant species × body length interaction terms for the analyses of tail length and number of caudal vertebrae were removed from models to strengthen tests of main effects.

FIGURE 3.

Box and whisker plots of interspecific variation in (a) body length, (b) tail length, and (c) number of postsacral caudal vertebrae among salt marsh (R. raviventris, n = 9) and western (R. megalotis, n = 8) harvest mouse specimens from micro‐CT‐based skeletal measurements.

To parse out the skeletal morphometrics driving the majority of variance between species, a principal component analysis (PCA) was performed. The first four principal components explained 82.5% of the variance in the data (PC1 = 53.0%, PC2 = 13.5%, PC3 = 9.0%, PC4 = 7.0%; Figure 4), beyond which additional components marginally contributed to the variance explained. Based on the rotated variable loadings (Table S1), increasing in the positive direction on PC1 indicates a wider 3rd caudal vertebra (0.36), longer (craniocaudal) 9th (0.51) and 15th (0.48) caudal vertebrae. Increasing positively on PC2 signifies deeper 2nd (0.41) and 3rd (0.33) caudal vertebrae, and a longer cumulative carpus and 3rd metacarpal (0.45). Decreasing PC2 reflects a longer 15th caudal vertebra (−0.41). Increasing along PC3 indicates a relatively wider 15th caudal vertebra (0.33) and a longer digit III proximal phalanx (0.44), whereas decreasing along this axis reflects a longer 2nd caudal vertebra transverse process (0.51). Increasing PC4 reflects a longer 3rd caudal vertebra transverse process (0.35) and deeper 15th caudal vertebra (0.31), whereas negative PC4 values indicate a deeper 3rd caudal vertebra (−0.43) and a longer digit III proximal phalanx (−0.51). The two species were differentiated the most along PC1 and PC4, such that the SMHM averaged higher along PC1 and lower along PC4 (Figure 4). The ANOVA revealed a significant difference between species (F _1,15 = 3.56, p = 0.011; Table 2). This result was corroborated by the MANOVA test (Pillai's trace = 0.905, Z = 2.245, df = 1,15, p = 0.0061).

FIGURE 4.

Principal component analysis of 17 micro‐CT‐based skeletal measurements of the caudal vertebrae and manus of salt marsh (R. raviventris; black triangles) and western (R. megalotis; gray circles) harvest mouse specimens. (a) Positive along PC1 indicates a wider 3rd caudal vertebra, and longer 9th and 15th caudal vertebrae. Positive along PC2 indicates deeper 2nd and 3rd caudal vertebrae, and a longer cumulative carpus and 3rd metacarpal length. (b) Positive along PC3 indicates a relatively wider 15th caudal vertebra and a longer digit 3 proximal phalanx. Positive along PC4 reflects a longer 3rd caudal vertebra transverse process and deeper 15th caudal vertebra, whereas negative PC4 values indicate a deeper 3rd caudal vertebra and a longer digit 3 proximal phalanx.

TABLE 2.

The results of RRPP ANOVAs testing for multivariate differences along the first 9 components from a principal components analysis (PCA) of 17 vertebral and digit III phalangeal micro‐CT‐based skeletal measurements, and along the first 7 components of a PCA of 14 vertebral shape indices, between salt marsh (Reithrodontomys raviventris) and western (R. megalotis) harvest mice.

	df	SS	MS	Rsq	F	Z	Pr(>F)
Principal components of vertebral and digit III phalanx dimensions
Species	1	2.79	2.79	0.192	3.56	2.19	0.011
Residuals	15	11.76	0.784	0.808
Total	16	14.55
Principal components of vertebral shape indices
Species	1	0.173	0.173	0.0921	1.52	0.955	0.173
Residuals	15	1.703	0.114	0.908
Total	16	1.88

3.2. Vertebral characteristics

We tested for univariate differences along each dimension, and only found significant differences in the 3rd caudal vertebra after accounting for the effects of body length. The SMHM had a deeper 3rd caudal vertebra than did WHM (F _1,14 = 13.23, p = 0.0021; Table 3; Figure 5), after accounting for the non‐significant effect of body length (F _1,14 = 0.224, p = 0.648). Other notable, but non‐significant, differences include: a longer 3rd caudal vertebra transverse process in WHM (F _1,14 = 3.78, p = 0.0762; Table 3; Figure 5), accounting for the significant effect of body length (F _1,14 = 9.65, p = 0.0092); a longer 2nd caudal vertebra in SMHM (F _1,14 = 4.35, p = 0.0553; Table S2; Figure S7), accounting for the significant effect of body length (F _1,14 = 15.14, p = 0.0010); and a deeper 2nd caudal vertebra in SMHM (F _1,15 = 4.32, p = 0.0556; Table S2) after removing the non‐significant (F _1,14 = 1.20, p = 0.2925) effect of body length. The SMHM had significantly greater caudal vertebra 3 depth: width ratios (F _1,14 = 9.74, p = 0.0072; Table 3; Figure 5), accounting for the significant effect of body length (F _1,14 = 12.31, p = 0.0025), and greater caudal vertebra 3 depth: length ratios (F _1,14 = 8.70, p = 0.0071; Table 3; Figure 5), accounting for a marginally non‐significant effect of body length (F _1,14 = 4.61, p = 0.0503). We performed a PCA on all vertebral shape variables, collectively. However, there were no significant differences between species along the first 7 PC's that explained ~97% of the variance (F _1,15 = 1.52, p = 0.173; Table 2).

TABLE 3.

Results of RRPP ANOVAs on the dimensions and shapes of the 3rd caudal vertebra.

	df	SS	MS	Rsq	F	Z	Pr(>F)
Transverse process length
Species	1	0.0717	0.0717	0.157	3.78	1.43	0.0762
Body length	1	0.183	0.183	0.401	9.65	2.28	0.0092
Residuals	14	0.266	0.0187	0.581
Total	16	0.457
Vertebral body depth
Species	1	0.425	0.425	0.460	13.23	2.65	0.0021
Body length	1	0.0072	0.0072	0.0078	0.224	−0.366	0.648
Residuals	14	0.449	0.0321	0.487
Total	16	0.923
Depth:length ratio
Species	1	0.0795	0.0795	0.368	8.70	2.23	0.0071
Body length	1	0.0421	0.0421	0.195	4.61	1.60	0.0503
Residuals	14	0.128	0.0091	0.593
Total	16	0.216
Depth:width ratio
Species	1	0.0777	0.0777	0.331	9.74	2.30	0.0072
Body length	1	0.0981	0.0981	0.418	12.31	2.56	0.0025
Residuals	14	0.112	0.0080	0.475
Total	16	0.235

Note: The full models included species × body length interactions, none of which were significant and hence were subsequently dropped from the models.

FIGURE 5.

Variation in 3rd caudal vertebra dimensions (a and b) and shapes (c and d) with respect to body length, between salt marsh (Reithrodontomys raviventris; black triangles) and western (R. megalotis; grey circles) harvest mouse specimens. Regression lines show the relationships between variables.

Caudal vertebra body lengths varied along the length of the tail differently between species (F _24,240 = 3.81, p = 0.0001; Table 4). In both species vertebral lengths increased rapidly along the proximal tail, reached a peak at cv08 and cv10, for WHM and SMHM, respectively, then steadily decreased in length caudally (Figure 6). Vertebrae cv16–cv25 were significantly longer in SMHM (F _1,10 = 1.00, p = 0.0435; Table 4), but there was no significant difference between species along vertebrae cv01–cv15.

TABLE 4.

Results of a nested RRPP ANOVA in a mixed within‐ and between‐subjects framework to test for differences in vertebral lengths along the tail and between species.

	df	SS	MS	Rsq	F	Z	Pr(>F)
Caudal vertebrae 1–25
Species	1	0.0000011	0.00000105	0.00003	0.0066	−1.27	0.887
Vertebra	24	0.0142	0.000593	0.453	167.44	30.24	0.0001
Specimen (species)	10	0.00159	0.000159	0.0507	45.01	15.28	0.0001
Species × vertebra	24	0.000324	0.000014	0.0103	3.814	5.41	0.0001
Residuals	240	0.000850	0.0000035	0.0270
Total	299	0.0314
Caudal vertebrae 16–25
Species	1	0.000039	0.000039	0.00450	0.307	1.10	0.136
Vertebra	9	0.00364	0.00041	0.424	281.21	24.75	0.0001
Specimen (species)	10	0.00126	0.000126	0.147	87.62	16.39	0.0001
Species × vertebra	9	0.000018	0.0000020	0.00204	1.36	0.789	0.216
Residuals	90	0.000130	0.0000014	0.0151
Total	119	0.00860
Caudal vertebrae 16–25, reduced model
Species	1	0.000126	0.000126	0.0147	1.00	1.67	0.0435
Vertebra	9	0.00666	0.000740	0.774	497.74	29.77	0.0001
Specimen (species)	10	0.00126	0.000126	0.147	84.88	16.14	0.0001
Residuals	99	0.00860
Total	119	0.00860

Note: Vertebra represents the positional identity of vertebrae (cv1–cv25; from cranial to caudal) treated as repeated measures for each specimen, with specimen nested within species.

FIGURE 6.

Craniocaudal lengths (standardized to body length) of caudal vertebrae (cv) 1–25 (of approximately 30–31), arranged from the cranial to the caudal end of the tail for salt marsh (Reithrodontomys raviventris; charcoal; n = 6) and western (R. megalotis; gray; n = 6) harvest mouse specimens.

3.3. Interspecific differences in phalangeal characteristics

There were no significant differences in phalanx dimensions or indices between species. However, the SMHM had a relatively greater digit III combined proximal and intermediate phalanx length (mean ± SE = 3.14 ± 0.110 [n = 9] and 2.88 ± 0.082 [n = 8] for SMHM and WHM, respectively; Figure S9) that was not significant (F _1,15 = 3.43, p = 0.084; Table S2) after removing the non‐significant effect of body length (F _1,1 = 0.032, p = 0.863).

4. DISCUSSION

Biomechanical specialization may contribute to SMHM survival in its challenging wetland habitat, as the mouse climbs to avoid tidal inundation and is able to swim for at least 2 h without rest (Fisler, 1965; Smith et al., 2014). Climbing locomotor capabilities specifically affect SMHM survival as the mouse climbs to avoid drowning, evade predation, and to nest above tidal inundation, which may offer increased protection for mouse pups (Fisler, 1965; Johnston, 1956; Smith et al., 2014). Previous work on live‐captured mice in the field has shown that the northern subspecies of the SMHM (Reithrodontomys raviventris halicoetes) has a relatively longer and thicker tail than the closely related WHM (R. megalotis) (Fisler, 1965; Shellhammer, 1984; Statham et al., 2021; Sustaita et al., 2018). We note that we did not see this difference here, owing to the small sample size, coupled with the fact that our skeletal measurements encompass the entirety of the tail, whereas our field measurements of tail length begin (proximally) at around the level of the 4th caudal vertebra. In other rodent species, a relatively longer tail has been shown to significantly counteract body rotational forces experienced during climbs better than shorter tails (Hager & Hoekstra, 2021). Decreasing body roll increases stability and decreases risk of falling, which in turn decreases predation and injury susceptibility (Cartmill, 1985). Here we examined the skeletal basis to these differences using micro‐CT and found differences in key dimensions, particularly in the depth of the 3rd caudal vertebra and in the lengths of the distal caudal vertebrae along the length of the tail, which may have implications for muscle attachments and bending moments (discussed below).

4.1. Interspecific differences in caudal vertebral dimensions

There was a significant, overall, difference between species when all of the variables known to underly climbing and scansorial behavior were considered collectively. The SMHM tended toward higher, positive values along PC1, indicating a wider 3rd caudal vertebra and longer (craniocaudally) 9th and 15th caudal vertebrae, and lower along PC4, reflecting a deeper 3rd caudal vertebra and a longer digit III proximal phalanx. The univariate results mostly corroborated these overall differences. After adjusting for differences in body length, SMHM had an absolutely and relatively (to vertebral length and width) deeper 3rd caudal vertebra. Although not statistically significant (due perhaps to the low power of the analysis owing to the small sample size), SMHM also had comparatively shorter 3rd caudal vertebra transverse process lengths.

Cross‐sectional shape (depth/width) and relative depth (depth/length) of the 3rd caudal vertebra were greater in SMHM and decreased with increasing body size. In Peromyscus species, allometric scaling of vertebral length, width, and depth was proposed to explain interspecific differences in the ranges of tail curvature (Hager & Hoekstra, 2021). Specifically, elongate, narrow vertebrae in the distal tail region of forest deer mice permit larger intervertebral bending angles than the relatively wider caudal vertebrae of prairie deer and California mice (Hager & Hoekstra, 2021). Larger mice in our study would be characterized by relatively elongate, narrow vertebrae suggestive of enhanced intervertebral bending capacity, upon which larger mice may arguably be more dependent during scansorial locomotion.

Climbing tetrapods (e.g., experimental lab mice conditioned to climb, arboreal chameleons, and new world monkeys) exhibit increased robustness of caudal vertebrae in terms of transverse process prominence relative to craniocaudal length (Byron et al., 2011; Luger et al., 2020; Organ et al., 2010; Russo, 2015; Schmitt et al., 2005). Contrary to our predictions, the WHM exhibited larger transverse processes on the 3rd caudal vertebra than the SMHM, and no discernable differences in this trait were observed on the 2nd vertebra. Transverse process length on the 3rd caudal vertebra was positively correlated with increasing body length (although this metric was not size dependent for the 2nd vertebra). In New World monkeys, transverse process breadth increases with increasing relative tail length (which itself is associated with prehensility; Horner, 1954; Organ, 2010; Russo, 2015). Increased transverse process length may increase the surface area available for musculature attachment in the coronal plane (Organ, 2010; Schmitt et al., 2005), so it makes sense that this feature should increase with body size. At the same time, however, the relatively longer transverse processes of WHM could limit the flexibility and speed (due to the increase in muscle moment arms) of tail movements. The deeper 3rd caudal vertebra of the SMHM may be indicative of larger neural and hemal processes, depending on the location of these attributes lengthwise on the craniocaudal axis. Spinous processes have been shown to increase with relative tail length in other climbing mammals, potentially offering increased mechanical advantage (Russo, 2015). The PCA also revealed comparatively longer 9th and 15th caudal vertebrae in the SMHM (PC1; Figure 4), but no discernable differences in their widths and depths. This is interesting given the in vivo differences in tail “diameter” observed between species (approximately 0.2–1.0 mm greater in SMHM; Fisler, 1965; Shellhammer, 1984; Sustaita et al., 2018; Statham et al., 2021), in a position along the tail corresponding roughly to the 9th caudal vertebra. This suggests that soft tissues (e.g., tendons, ligaments, and skin) makes up the observed difference in tail thickness between species.

4.2. Analysis of sequential caudal vertebral lengths

The number of caudal vertebrae ranged from 26 to 31 among individuals and between species, which is consistent with the range (26–31) reported for a closely related congener, R. montanus (Wilkins, 1986). However, unlike the deer mice studied by Kingsley et al. (2017), the number of vertebrae comprising the tail did not differ between SMHM and WHM. It appears that the disparity in tail length between the northern subspecies of the SMHM and WHM (Fisler, 1965; Shellhammer, 1984; Sustaita et al., 2018) is attributable to differences in the craniocaudal lengths of vertebra in the distal tail rather than an increase in the number of vertebrae (Figure 6). Ecomorphs of Peromyscus maniculatus found in forested environments exhibit longer tail lengths composed of craniocaudally longer vertebrae, compared to their terrestrial conspecifics (Kingsley et al., 2017). Longer proximal vertebrae may increase circumferential range during tail abduction, though craniocaudally shorter vertebrae may permit tighter coiling (Russo, 2015). That said, the SMHM tends to coil its tail in the mid to distal segments during climbs (Figure S10); therefore, a tighter arc of flexion in the proximal area may not be required. Less flexibility in the proximal region of the tail is customary of other climbing rodents such as deer mice, in which tail radius of curvature decreases distally, with forest subspecies exhibiting a lower radius of curvature overall (Hager & Hoekstra, 2021).

In concordance with Russo's (2015) description of the mammalian tail, the caudal extremity of the SMHM is defined by relatively short proximal vertebrae, which steadily progress to double in length in the transitional region of the tail. Caudal to the longest vertebra of the tail (the 10th, on average, in the SMHM), the vertebrae gradually shorten until they become minute bone fragments at the end of the tail (Figure 6). The WHM's tail vertebrae follow a similar pattern, increasing in length only to peak at approximately the 8th vertebra (on average) and then decreasing in length distally. In new world monkeys, the location of the longest vertebra accurately predicts prehensility, with prehensile species exhibiting their longest tail vertebra more caudally than their non‐prehensile comparators (Organ, 2010; Organ & Lemelin, 2009). Mammalian tail robustness is known to decrease in the distal tail region, and the longest vertebra indicates the start of this tail section (Russo, 2015).

4.3. Interspecific differences in phalangeal characteristics

Although the phalanx measurements showed little in the way of differences between species, some (such as digit III metacarpal and proximal phalanx lengths) loaded substantially on the PCA and contributed to overall differences between species. High phalangeal indices (intermediate/proximal phalanx, proximal phalanx/metacarpal and proximal phalanx/metacarpal length) contribute to elongated digits and are characteristic of climbing taxa (Mammalia: Ji et al., 2002, Kirk et al., 2008; Murinae: Nations et al., 2019; Rodentia: Samuels & Van Valkenburgh, 2008). Lengthened digits may increase gripping abilities, allowing mice to fully wrap digits around the substrate while navigating three‐dimensional spaces (Cartmill, 1985). Increased grip on supports during climbs allows for the generation of higher correctional torque when yaw, roll, and pitch forces impede balance (Cartmill, 1985; Lammers & Zurcher, 2011). The SMHM showed a greater combined digit III proximal and intermediate phalanx length. Longer proximal phalanges alone can be used to successfully discriminate between arboreal and terrestrial ecologies in mammals (Kirk et al., 2008).

4.4. Functional and ecological implications

Even in the absence of distinct morphological differences, species may exhibit specializations for particular modes of locomotion and substrates through behavioral means (e.g., Byron et al., 2011; Camargo et al., 2016; Granatosky et al., 2021). Thus, it is possible that the SMHM supplements morphological adaptation with specialized spatiotemporal gait and tail behaviors to maintain balance and climb efficiently. In arboreal rodents, examples of these specialized behaviors include: manipulation of foot positioning, velocity, stride length, stride frequency, and limb duty factor when climbing on structures of varying diameter (Camargo et al., 2016; Clemente et al., 2019; Karantanis et al., 2017). Studies are currently underway to obtain these data for salt marsh harvest mice and coexisting rodent species throughout the Suisun Marsh (Robles Martinez et al., 2023). Among other things, these data will allow us to narrow down specifically which caudal vertebrae the mouse coils during climbing, or which tail positions the mouse uses the most in the field.

An exploration of the limbs in full, their articular surfaces, and intervertebral joints may provide more information with regards to the biomechanical specialization of the SMHM (Cartmill, 1985). Climbing rodents are known to exhibit a 21% decrease in their talar head angle, a more rounded talus, lengthened humeri, and increased curvature of the 3rd metacarpal (Byron et al., 2011; Samuels & Van Valkenburgh, 2008). Rounded joints serve to increase joint mobility, whereas lengthened forelimbs help to bridge gaps during climbs (Cartmill, 1985; Hildebrand & Goslow, 2001). Additionally, climbing species exhibit elongate femora and specialized cranial articular caudal vertebral surfaces, the latter of which permit increased flexibility of the tail (Bou et al., 1987; Russo, 2015; Samuels & Van Valkenburgh, 2008). Specifically, cranial vertebral surfaces of the mammalian prehensile tail are characterized by a decrease in dorsoventral breadth, leading to an overall rounded shape (Russo, 2015). Reduced olecranon processes can also predict climbing adeptness and permit increased extension of the forelimb (Samuels & Van Valkenburgh, 2008). If the SMHM consistently climbs, mechanical loading and increased elbow flexion may lead to increased cranial curvature of the ulna and humerus (Henderson et al., 2017).

However, high tides, predation, and foraging may also force SMHM to swim (Fisler, 1965; Smith et al., 2014). The typical morphologies associated with semiaquatic and terrestrial forms of locomotion often conflict (e.g., the robust appendicular bones of semiaquatic vs. the elongate appendicular bones of arboreal rodents; Samuels & Van Valkenburgh, 2008). Specifically, increased surface area available for muscle attachment on the femur, olecranon process, and tibia is experienced by semiaquatic rodents (Samuels & Van Valkenburgh, 2008). In contrast, the appendages of arboreal species are more gracile in nature, exhibiting a reduction in femoral, olecranon, and tibial processes (Samuels & Van Valkenburgh, 2008). These appendicular differences between semiaquatic and arboreal rodents may also be relevant to axial and phalangeal structures. That said, it is possible that the diverse forms of locomotion used by the salt marsh specialist are facilitated by a jack‐of‐all‐trades body plan, rather than a distinctly aquatic, scansorial, or terrestrial morphology. If this is so, it may be difficult to discern functional skeletal differences in the SMHM compared to other marsh taxa, especially based on existing literature in which categorization of morphology and locomotor type is strictly defined.

5. CONCLUSION

We detected compelling differences in skeletal morphology between the endangered SMHM and congeneric, sympatric WHM, in spite of the small sample size owing to the limited number of specimens available, and their close phylogenetic relationship (Arellano et al., 2005; Statham et al., 2016). These differences might reflect the SMHM's specialization to the challenges of an obligate wetland existence, compared to the more generalized, upland‐dwelling WHM. Compared to WHM, the SMHM exhibited a relatively and absolutely deeper 3rd caudal vertebra, craniocaudally longer individual distal tail vertebrae, a non‐significant (but compelling) longer digit III combined proximal and intermediate phalanx length, and shorter 3rd caudal vertebra transverse process. These attributes may offer functional advantages during climbing, by increasing contact with the substrate via both manual and prehensile tail gripping tactics (Cartmill, 1985; Russo, 2015). Increased proximal vertebral depth, if a reasonable proxy of neural and hemal processes lengths, suggests an increase in the surface area available for muscle attachment (Luger et al., 2020; Shinohara, 1999). Longer (distally) caudal vertebrae, combined with shorter 3rd caudal vertebra transverse process, in the SMHM may be indicative of greater tail flexibility and speed in the coronal plane (Russo, 2015) which in turn, may increase the amount of supports available to the mouse while traveling through emergent marsh vegetation. The SMHM is a unique organism, able to employ ambulatory, arboreal, and semiaquatic locomotion and on‐going investigations of its ecology, climbing (i.e., gait components and tail usage), and swimming performance are required to better understand the functional consequences of their skeletal characteristics.

AUTHOR CONTRIBUTIONS

KMW and DS conceived the study; KWM, DS, and RBP secured the funding; LMB‐T provided specimens; MJS provided specimens and performed genetic species identification; RBP performed the micro‐CT scans; KMW built the volumetric models from micro‐CT scans; KMW and DS gathered the data; KMW and DS performed the statistical analyses and drafted the figures; KMW and DS drafted the manuscript; MJS, RBP, and LMB‐T contributed to, and edited, the manuscript.

CONFLICT OF INTEREST STATEMENT

Authors have no conflicts of interest to declare.

Supporting information

Figure S1.

Figure S2.

Figure S3.

Figure S4.

Figure S5.

Figure S6.

Figure S7.

Figure S8.

Figure S9.

Figure S10.

Table S1.

Table S2.

Table S3.

Table S4.

Table S5.

Text S1.

Woldt, K.M. , Pratt, R.B. , Statham, M.J. , Barthman‐Thompson, L.M. & Sustaita, D. (2024) Comparative skeletal anatomy of salt marsh and western harvest mice in relation to locomotor ecology. Journal of Anatomy, 245, 289–302. Available from: 10.1111/joa.14047

DATA AVAILABILITY STATEMENT

The data supporting the results of this manuscript are available through direct (and reasonable) request to the corresponding author.