The ultimate and proximate mechanisms driving the evolution of long tails in forest deer mice

Abstract

Understanding both the role of selection in driving phenotypic change and its underlying genetic basis remain major challenges in evolutionary biology. Here, we use modern tools to revisit a classic system of local adaptation in the North American deer mouse, Peromyscus maniculatus, which occupies two main habitat types: prairie and forest. Using historical collections, we find that forest-dwelling mice have longer tails than those from nonforested habitat, even when we account for individual and population relatedness. Using genome-wide SNP data, we show that mice from forested habitats in the eastern and western parts of their range form separate clades, suggesting that increased tail length evolved independently. We find that forest mice in the east and west have both more and longer caudal vertebrae, but not trunk vertebrae, than nearby prairie forms. By intercrossing prairie and forest mice, we show that the number and length of caudal vertebrae are not correlated in this recombinant population, indicating that variation in these traits is controlled by separate genetic loci. Together, these results demonstrate convergent evolution of the long-tailed forest phenotype through two distinct genetic mechanisms, affecting number and length of vertebrae, and suggest that these morphological changes-either independently or together-are adaptive.

Keywords: Caudal vertebrae; Peromyscus maniculatus; convergence; local adaptation; parallel evolution; skeletal evolution.

Figure 1

Deer mouse geography and tail length variation. (A) Map of North America showing the roughly defined range of P. maniculatus. Broad‐scale forest (green) and prairie (tan) range limits are shown, following Osgood (1909) and Hall (1981). Each dot represents a collecting locale from which one to nine samples were obtained. Dot color represents the local GIS land cover‐defined habitat of the site (green = forest, tan = nonforest). Mice from all sample locales were used in subsequent genome‐wide capture analyses. The red outline for “x‐ray samples” indicates that we used samples from those locations for the comparison of vertebral number and length. (B) Box‐and‐whisker plot of tail/body length ratio variation among deer mouse subspecies from museum collections. Box color indicates local habitat in which samples were collected based on GIS land cover data for those subspecies; green = forest, tan = nonforest, gray = mixed. (“Mixed” indicates that populations from a single subspecies were captured in both forest and nonforest locations.) [Color figure can be viewed at wileyonlinelibrary.com]

Figure 2

Deer mouse population structure and gene flow. (A) Plot of the first two principal components calculated from a genome‐wide sample of 7396 high‐quality SNPs from populations ranging across the continent (sampling locales shown in Fig. 1A). Each dot (n = 80) represents a single individual. (B) Cladogram based on nuclear Maximum Likelihood phylogeny collapsed to high‐confidence clades. Node labels represent bootstrap support. In both panels, colors indicate local GIS land cover‐defined habitat (tan = prairie, green = forest, gray = mixed). [Color figure can be viewed at wileyonlinelibrary.com]

Figure 3

Tail length differences between habitats when accounting for genetic non‐independence. (A) Tail/body length ratios for each population predicted by a generalized linear‐mixed model taking genetic differentiation (F_ST) between populations into account. (B) Tail/body length ratios for prairie and forest individuals predicted by a similar model as in A, but with pairwise genetic relatedness (kinship coefficient [Manichaikul et al. 2010]) between individuals taken into account. Error bars represent 95% confidence intervals. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 4

Convergent tail vertebral morphology in eastern and western forest‐prairie population pairs. (A) Representative radiographs of deer mouse tails from the eastern and western population pairs. (B) Forest mice have more caudal vertebrae than prairie mice in the east and west. Vertical red lines represent medians. Note truncated axis. (C) Forest mice have longer vertebrae than prairie mice in the east and west. Dashed line represents the median length of the longest prairie vertebra; the segment of the forest tail with longer vertebrae is 4–16 and 4–15 for western and eastern populations, respectively. (D) Summary of tail vertebral differences between forest and prairie mice. Dots represent mean cumulative tail lengths. Dashed line is the cumulative length of the mean prairie tail with three extra vertebrae added, which represents an estimate of the maximum contribution of difference in number of vertebrae to the difference in total length. For C and D, “vertebral position” indicates the nth caudal vertebra, starting from the base of the tail. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 5

No significant correlation between number and length of caudal vertebrae in a laboratory intercross. Each point represents the number of caudal vertebrae and the length of the longest caudal vertebra measured from a radiograph of a parental type (A, P. m. nubiterrae (green), P. m. bairdii (tan), n = 12 of each subspecies), or first‐generation F1 (B, n = 10) or second‐generation F2 (C, n = 96) nubiterrae x bairdii hybrids. Ninety‐six F2 individuals allow 80% power to detect a correlation of r > 0.25. [Color figure can be viewed at wileyonlinelibrary.com]