Variation in cranial shape
Principal component analyses (PCA) (nlateral = 452; nventral = 420) revealed that the first two principal component (PC) axes account for over half of the shape variation in both lateral and ventral cranial views (PC1lateral = 43.0 %, PC2lateral = 12.8 %; PC1ventral = 50.2 %, PC2ventral = 9.5 %; Fig. 3a–f). The remaining PCs each explained less than 7 % of variation. These results, and their relation to the digging modes known for Thomomys taxa, are summarized in Fig. 3g.
The lateral cranial PC1 divides the two Thomomys subgenera: the larger-bodied subgenus T. Megascapheus taxa (reds and yellows) have more procumbent incisors (Fig. 3a, c), and the smaller subgenus T. Thomomys taxa (blues and greens) have less procumbent incisors (Fig. 3a, c). The distribution of taxa along ventral cranial PC1 is similar to their distribution along the lateral cranial PC1 with larger body size corresponding to deeper, more robust skulls (Fig. 3d, f) and smaller body size corresponding to more gracile skulls (Fig. 3d, f). On the PCA graphs of lateral and ventral cranial shape, subspecies of T. (Thomomys) talpoides (blues) and T. (Megascapheus) bottae canus (light yellow; note that this taxon is an unfortunately named member of the T. M. Townsendii clade) occupy the intermediate morphospace between the two subgenera (Fig. 3c, f). In summary, the lateral and ventral PC1 axes appear to reflect allometrically increased incisor procumbency and cranial robusticity, respectively.
The lateral cranial PC2 distinguishes between the taxa of the two subgenus Megascapheus clades, T. M. Bottae (reds) versus T. M. Townsendii (yellows) (Fig. 3c, f). This divergence is based on T. M. Bottae taxa (reds) having distally displaced infraorbital foramens relative to T. M. Townsendii taxa (yellows) in addition to increased procumbency (Fig. 3b). Furthermore, the lateral PC2 almost completely separates the two smallest T. (T.) talpoides subspecies (blues) from the rest of subgenus T. Thomomys clade (greens). Similar to the T. Megascapheus divergence, this divergence within the small-bodied subgenus T. Thomomys is associated with increased procumbency and a distally displaced infraorbital foramen (Fig. 3b, c). The distal movement of the infraorbital foramen in both large and small taxa suggests that allometry-independent cranial rearrangement may underlie increased procumbency along this axis.
The ventral cranial PC2 also distinguishes between the taxa of the two subgenus T. Megascapheus clades (reds versus yellows; Fig. 3e, f). This divergence in the ventral view is based on the greater size of ventral muscle attachment sites in T. M. Bottae taxa (reds) versus a dorsal-shift in the orientation of the foramen magnum in T. M. Townsendii taxa (yellows) (Fig. 3e). The largest taxa in the genus, T. (M.) townsendii, has the most dorsally shifted foramen magnum (Fig. 3e). Unlike the lateral PC2 results, however, procumbent T. Thomomys taxa (blues) do not diverge from non-procumbent sister taxa (greens) in either foramen magnum location or ventral muscle attachment sites (Fig. 3f). In summary, the distribution of taxa along lateral cranial PC2 do not always correspond to their distribution along ventral cranial PC2.
Variation in cranial allometry
The MANCOVAs confirmed a significant association between size and shape in both cranial views (Plateral = 0.001; Pventral = 0.001). Size explained about 20 % of the shape variation while phylogenetic affiliation (“taxa”; Fig. 1) explained about 38 % (Table 1). Furthermore, the MANCOVAs confirmed significant interaction terms between size and taxa for both cranial views (Plateral = 0.001; Pventral = 0.001; Table 1; Fig. 4b, e), indicating that allometric slopes differ between taxa. Subsequent pairwise slope tests (Fig. 4b, e) showed more instances of significant taxa-specific allometric slope divergences in the ventral view than in the lateral view. These results suggest that the allometric relationship between size and incisor procumbency is less labile than that between size and skull robustness (indicated by fewer significant values in Table 2
versus Table 3, Fig. 4b
versus e). In other words, the allometric patterning of ventral view muscle attachment sites and of foramen magnum location appears to have greater variation than the allometric patterning of incisor procumbency. This is consistent with the results from the cranial shape PCAs.
The allometric plots visualizing the variation detected in the MANCOVAs reveal substantial differences in allometric slopes and in intercepts between taxa (Fig. 4). The significant slope differences preclude tests for intercept differences; however, both lateral and ventral view plots show that many taxa separate along the y-axis such that intersections would not occur in the biologically possible morphospace (Fig. 4b, e). For example, the slopes of T. (M.) b. navus (red) and T. (M.) b. canus (light yellow) would intersect well outside the size range of this rodent genus (Fig. 4b, e).
Upward shifts in y-intercepts appear to identify the tooth-digging taxa compared to claw-digging relatives (Fig. 3g, Fig. 4b, e; Fig. 5): tooth-digging T. M. Townsendii clade taxon, T. (M.) townsendii (dark yellow) versus claw-digging sister taxon, T. (M.) b. canus (light yellow); tooth-digging T. M. Bottae clade taxa, T. (M.) bottae navus, T. (M.) b. leucodon, and T. (M.) b. saxatalis (reds) versus claw-digging sister taxon, T. (M.) b. laticeps (pink); and finally the two tooth-digging subgenus T. Thomomys subspecies of T. (T.) talpoides (blues) versus claw-digging sister taxa, T. (T.) mazama and monticola (greens) (Fig. 4e). Subgenus T. Thomomys taxon, T. (T.) talpoides quadratus (dark blue) stands out by displaying more robust crania than much larger individuals of subgenus T. Mesgascapheus Townsendii taxon, T. (M.) b. canus (yellow) (indicated by parallel slopes in Fig. 4d, e). The lateral cranial view shows the same pattern of y-intercept upward shifts for tooth-digging in T. Megascapheus taxa but not in the subgenus T. Thomomys taxa (indicated by overlapping blues and greens in Fig. 4b).
Taxa within a clade rarely exhibit significantly different allometric slopes, however, three notable exceptions occur. First, a significant difference between the two T. Megascapheus Townsendii taxa appears to reflect the nearly flat allometric slope of T. (M.) townsendii (dark yellow) in contrast to the steeper slope of sister taxon T. (M.) b. canus (light yellow), seen in both views (Tables 2 and 3; Fig. 4b, e). In support of this, the allometric slope of T. (M.) townsendii significantly differs from all other taxa in the ventral view (Table 3, Fig. 4e). A similar effect appears to occur within T. Megascapheus Bottae taxa in the significantly different lateral cranial allometric slopes of T. (M.) b. navus (light red) and T. (M.) b. saxatilis (dark red); these are also the two most procumbent taxa in the genus (Fig. 4b).
The third instance of within-clade significant allometric slope divergence involves the two subspecies of T. (T.) talpoides (Table 3; Fig. 4e). In the lateral cranial view, procumbency appears to have less relation to centroid size for T. (T.) talpoides quadratus (dark blue) (indicated by significant differences with larger taxa in Table 2 and a flat slope in Fig. 4b), while the ventral cranial view, this taxon shows similar allometric patterns of skull robustness to other Thomomys taxa (non-significant values in Table 3; parallel slope to other taxa in Fig. 4e). By contrast, the ventral crania of T. (T.) talpoides fisherii (light blue) appear less robust with increasing centroid size (significant values in Table 2; negative slope in Fig. 4e), while its allometric relationship with procumbency is similar to that of the other clades (non-significant values in Table 3; parallel slope in Fig. 4b).
Splitting the dataset into regional subpopulations (Fig. 4c, f) reveals substantial diversity in allometric slopes which clearly contribute to differences in taxa-specific allometric slopes. The sample sizes for these subpopulations, however, are possibly too small for a confident identification of significantly different allometric slopes (Additional file 3: Table S3).
Variation in cranial shape in relation to soil
The PCAs of cranial shape colored by soil type demonstrate that the presumed ancestral taxa, T. (T.) mazama and T. (T.) monticola (greens) occupy the softest soils in the region (Fig. 6a–f). Their sister taxa, however, (the two subspecies of T. (T.) talpoides [blues]) appear to inhabit soils of hardness comparable to the larger subgenus T. Megascapheus gophers (reds and yellows) (Fig. 6a–f). This is particularly visible in the lateral cranial PCAs (Fig. 6a–c). The least procumbent T. Megascapheus taxon, T. (M.) b. canus (light yellow), however, appears to inhabit soils of hardness similar to its more procumbent and robust sister taxon, T. (M.) townsendii (dark yellow, Fig. 6a–f).
Variation in humeral shape and its relation to soil
From the humeral views (nanterior = 73), only the first two anterior PC axes explained a meaningful amount of shape variation (PC1anterior = 31.8 %, PC2anterior = 15.0 %; Fig. 5a–d). The results from the lateral humeral view did not provide conclusive information beyond the anterior humeral view (Additional file 7: Supplementary results: lateral humeral view).
PC1 of anterior humeral shape captured deltoid process size, an important muscle attachment site, relative to the lateral epicondyle (Fig. 5a). The anterior humeral PC2 captured the distance of the deltoid process from the humeral head, the size of the medial epicondyle relative to the articular surface, and the orientation of greater tuberosity (Fig. 5b). All of these shape changes associated with more positive values of PC2 increase the mechanical advantage for digging. Along PC1, the larger subgenus T. Megascapheus taxa (reds and yellows) tend to have more robust humeri with larger deltoid processes as compared to the smaller subgenus T. Thomomys (blues and greens) (Fig. 5a, c). The claw-digging T. Megascapheus Townsendii taxa, T. (M.) b. canus scores highly along both axes, meaning all of the claw-digging muscle attachment sites highlighted above have increased in relative size.
In comparison to cranial shape, the MANCOVA for humeral shape and size showed that a smaller proportion of shape (less than 8 %, compared to over 20 % in both cranial views) is explained by size (Table 1). Tests for differences in static allometry between taxa were all non-significant (Table 1). Interestingly, while the tooth-digging T. M. Bottae taxa have slightly larger skulls than subgenus Thomomys taxa (Fig. 4), these two clades have humeri of overlapping centroid sizes (Additional file 8: Figure S2).
The first quadrants of the PCAs of humeral shape colored by soil type show that the most derived humeral shape belongs to the non-procumbent taxon T. (M.) b. canus (light yellow), which nonetheless inhabits hard soils (Fig. 6h, i). The least procumbent T. M. Bottae clade taxon, T. (M.) b. laticeps is found in similar soils as T. (M.) b. canus and the two taxa converge on a larger deltoid crest (Figs. 5, 6 and 7). The second quadrant captures a humeral shape associated with non-procumbent taxa inhabiting soft soils (Fig. 6g–i). The two bottom quadrants appear to capture a humeral shape of taxa with the ability to dig with their teeth (Fig. 6g–i). The smallest species in the genus, T. (T.) talpoides and the largest species in the genus, T. (M.) townsendii are intermediate between the humeral shape of species known to specialize in claw-digging and those that have adaptations for tooth-digging (Fig. 6; Table 4).