Signatures of selection in mammalian clock genes with coding trinucleotide repeats: Implications for studying the genomics of high‐pace adaptation

Abstract Climate change is predicted to affect the reproductive ecology of wildlife; however, we have yet to understand if and how species can adapt to the rapid pace of change. Clock genes are functional genes likely critical for adaptation to shifting seasonal conditions through shifts in timing cues. Many of these genes contain coding trinucleotide repeats, which offer the potential for higher rates of change than single nucleotide polymorphisms (SNPs) at coding sites, and, thus, may translate to faster rates of adaptation in changing environments. We characterized repeats in 22 clock genes across all annotated mammal species and evaluated the potential for selection on repeat motifs in three clock genes (NR1D1,CLOCK, and PER1) in three congeneric species pairs with different latitudinal range limits: Canada lynx and bobcat (Lynx canadensis and L. rufus), northern and southern flying squirrels (Glaucomys sabrinus and G. volans), and white‐footed and deer mouse (Peromyscus leucopus and P. maniculatus). Signatures of positive selection were found in both the interspecific comparison of Canada lynx and bobcat, and intraspecific analyses in Canada lynx. Northern and southern flying squirrels showed differing frequencies at common CLOCK alleles and a signature of balancing selection. Regional excess homozygosity was found in the deer mouse at PER1 suggesting disruptive selection, and further analyses suggested balancing selection in the white‐footed mouse. These preliminary signatures of selection and the presence of trinucleotide repeats within many clock genes warrant further consideration of the importance of candidate gene motifs for adaptation to climate change.


| 7255
PRENTICE ET al. underlying basis for this question lies in whether sufficient standing genetic variation or sufficient rates of molecular evolution (Barrett & Schluter, 2008;Hedrick, 2013) occur at key genes to keep pace with climate change. Ultimately, the rate of adaptive evolution in relation to the rate of climate change will contribute to the demographic effects of climate change on organisms and ecosystems (Bradshaw & Holzapfel, 2010;Bronson, 2009). This makes the characterization of adaptive genetic variation critical, in order to allow for a better understanding of the evolutionary potential and responses of species to environmental stressors (e.g., Harrisson, Paylova, Telonis-Scott, & Sunnucks, 2014). Further, understanding how standing genetic variation and genomic elements operating at higher rates of change contribute to adaptability will be important to estimate the relative roles of genetics, plasticity, and epigenetics in defining the "response capacity" or "adaptive potential" of species. Mammals are a particular taxonomic group whose vulnerability to climate change may be underestimated (Schloss, Nuñez, & Lawler, 2012), and as a result, there is a recognized need to identify and characterize mammalian genes responding to climate change (Dawson, Jackson, House, Prentice, & Mace, 2011;Franks & Hoffmann, 2012).
The seasonal timing of life-history events is often under the influence of selection as such events are frequently influenced by environmental cues (O'Malley, Ford, & Hard, 2010). Individuals that can anticipate the optimal timing of season-specific activities (e.g., migration, reproduction) are predicted to demonstrate higher fitness, as they are able to exploit the most favorable resources throughout the year. Photoperiod is one such environmental cue that is often used to determine the optimal timing of life-history strategies in species occupying seasonal environments (Bradshaw & Holzapfel, 2008).
Species respond to photoperiod cues via their circadian clocks, molecular oscillators that sense and respond to changes in photoperiod by triggering various effects including hormone secretions in mammals (Goldman, 2001). In fact, the negative relationship observed between day length and amplitude of the circadian pacemaker may be the cause of latitudinal clines often observed in the timing of seasonal events of many species (Pittendrigh, Kyner, & Takamura, 1991). Heritability of photoperiod responsiveness has been observed in mammals (Bronson, 2009;Heideman, Bruno, Singley, & Smedley, 1999;Lynch, Heath, & Johnston, 1981), particularly at higher latitudes where dependence on photoperiod increases with variance in day length and cues circannual seasonal changes (Bradshaw & Holzapfel, 2010). Thus, it has been argued that climate change is likely to introduce significant reproductive challenges for species inhabiting higher latitudes that rely on photoperiod to cue breeding, because at such latitudes an uncoupling of the phase relationship between environmental conditions and photoperiodic cues can occur (Milligan et al., 2009). Further, as range redistributions proceed due to shifts in temperature, species may be exposed to novel photoperiods. As species can track shifts in temperature through range redistributions, their persistence will more critically require the adjustment of photoperiod responses rather than thermal tolerance (Bradshaw & Holzapfel, 2006). Clock genes are thus one category of functional genes likely critical for adaptation to shifting seasonal conditions and novel environments (Kondratova, Dubrovsky, Antoch, & Kondratov, 2010).
The candidate gene approach has been used empirically to identify patterns of adaptive genetic variation and disentangle such patterns from neutral genetic population structure (DeFaveri, Jonsson, & Merilä, 2013;Hemmer-Hansen, Nielsen, Frydenberg, & Loeschke, 2007;Limborg et al., 2012;O'Malley et al., 2010). Candidate genes are selected based on known physiological functions perceived to be of relevance to the study species. This approach is supported where highly divergent allele frequencies are found, more often in genes with functions related to adaptive processes potentially under selection.
Such genes are dissimilar to neutral regions of the genome, which are not expected to vary among populations experiencing high rates of gene flow. Here, we use the candidate gene approach to examine specific motifs in targeted functional genes, specifically coding trinucleotide repeats.
The mutational mechanism of cTNR structures has been associated with the purity of the repeat structure itself, where purer repeats are more likely to undergo further slippage (Kruglyak, Durrett, Schug, & Aquadro, 1998). This may be of adaptive value by generating phenotypic variation upon which selection can act (Kashi & King, 2006;Laidlaw et al., 2007). These repeats also offer the potential for high mutation rates (Gemayel et al., 2010(Gemayel et al., , 2012, allowing for the rapid generation of novel alleles on the scale of contemporary adaptive evolution. This is particularly important in genes required for adaptation to climate change, in addition to a reliance on plasticity, regulatory elements (Bozek et al., 2009), and epigenetic effects (Ripperger & Merrow, 2011 (Johnsen et al., 2007) and fish (O'Malley et al., 2010) in addition to variation corresponding to earlier egg laying (Liedvogel, Szulkin, Knowles, Wood, & Sheldon, 2009). Further, the involvement of clock genes in seasonal entrainment has been demonstrated in mammals (Hazlerigg, Ebling, & Johnston, 2005), suggesting that cTNRs within these genes may play a role in seasonally fine-tuning the circadian characteristics of species inhabiting higher latitudes. Collectively, these results suggest that environmental factors correlated with latitude (e.g., photoperiod) may be driving selection at cTNRs within clock genes that are critical for the seasonal adaptation of life-history strategies. Thus, the characterization of cTNR structures in a range of other vertebrate species offers the potential to use the properties of microsatellite repeats (Press, Carlson, & Queitsch, 2014) to understand the genomics of adaptation.
Coding trinucleotide repeats have been observed in clock genes that facilitate the regulation of reproductive timing and social behaviors (Johnsen et al., 2007;Liedvogel & Sheldon, 2010;Liedvogel et al., 2009), genes associated with neuroprocesses (Whan et al., 2010), developmental homeobox genes, and transcription factors (Mularoni et al., 2010). For closely related species, their higher rates of mutation propose a mechanism for the convergence of pole-ward allele sizes following climate-induced range expansion. This is highly relevant to mammals, as many closely related species have evolved in complex and often isolated refugium patterns north or south of ice sheets during the Pleistocene (e.g., Shafer, Cullingham, Côté, & Coltman, 2010), thus allowing for the evolution of allelic repeat motifs specific to differential climatic conditions within an otherwise presumably conserved gene sequence.
Our objectives for this study were twofold. First, as wild mammal species have been infrequently characterized at clock genes for the presence of cTNR motifs, we wanted to characterize cTNRs within several candidate clock genes in a wide range of mammal species. Second, to evaluate the potential of clock genes for adaptation to differential latitudes, we compared three north-south congeneric species pairs at a selection of clock genes to determine the prevalence of cTNR repeats and levels of polymorphism. Due to the importance of such genes for circadian and circannual rhythms of mammal species, we hypothesized that clock genes are under selection in mammal species occurring along latitudinal clines. To test our hypothesis, we compared closely related species pairs adapted to different climatic niches and separated along latitudinal gradients at varying spatial scales. If clock genes are under selection in our study species, we expect to observe at least one of the following: clines in allele frequencies within or between species pairs, differentiation of allele frequencies between species pairs, departures from Hardy-Weinberg equilibrium (HWE), divergent patterns of differentiation (F ST ) between neutral microsatellites and each candidate cTNR, and/or identification of our cTNR loci as outliers in comparison with neutral genetic population structure.

| Characterization of candidate clock genes in mammal species
Before testing for selection, we wanted to characterize the presence and abundance of cTNRs in 22 candidate clock genes of mammal species. We selected the genes AANAT, ARNTL, ARNTL2, CLOCK, CRY1, CRY2,CSNK1A1,CSNK1D,MTNR1A,MTNR1B,NR1D1,NR1D2,PER1,PER2,PER3,RORA,RORB,RORC,RXRA,RXRB,TIMELESS,and TIPIN (Table 1) and used the Geneious (version 6.1.7, Biomatters, Auckland, NZ) databank search function to search GenBank for sequences of each clock gene across all species. We extracted the coding sequence of each clock gene in a total of 68 mammal species, excluding humans, and used the Geneious plug-in Phobos (Mayer, Christoph, Phobos 3.3.11, 2006 to search for tandem repeats. We defined our search criteria to locate repeat units that were 3 bp long and ≥9 bp (3 units) in length. Once repeats were located, we extracted information regarding the total repeat length, percentage perfection (purity), repeat unit type (e.g., CAG or polyglutamine) and the sequence of the repeat, and calculated metrics of repeat abundance and purity across all mammal species at each candidate clock gene. We estimated the total number of repeats found, the total number of pure (i.e., 100% perfection) repeats, the total number of repeats over 5 units (15 bp) long, the total number of pure repeats over 5 units long, and the species for which the longest repeats were observed at each gene. We also explored the relationship between repeat length and repeat purity of cTNRs across the candidate genes we surveyed in mammals by conducting a Spearman's rank correlation in R (R Core Team 2016).

| Study systems for investigating selection
To evaluate whether we could detect signatures of selection at candidate clock genes in natural systems, we assessed three pairs of congeneric species: Canada lynx and bobcat (Lynx canadensis and L. rufus), northern and southern flying squirrel (Glaucomys sabrinus and G. volans), and white-footed and deer mouse (Peromyscus leucopus and P. maniculatus). Each of these species pairs had a northern distributed species and a southern congener that is expanding northwards and increasing range overlap with its sister species. Although all of these species are widely distributed and exhibit high rates of intraspecific gene flow (e.g., Garroway, Bowman, Holloway, Malcolm, & Wilson, 2011;McKay, 2016;Row et al., 2012), both theoretical (Charlesworth, Nordborg, &Charlesworth, 1997) andempirical (DeFaveri et al., 2013) studies support the prediction that selection can maintain adaptive divergence at critical loci despite the rest of the genome being homogenized via gene flow. Thus, the evolutionary histories and distributional patterns of these species pairs provide a good opportunity to survey candidate genes associated with climate change in non-model organisms.

| Sample collection and strategy
The spatial scale of sampling for each species pair varied, and samples were obtained from a variety of sources. The Canada lynx and bobcat analysis was continental, incorporating the entire range of both species (Figure 1), and the area of range overlap at the southern extent of Canada (Koen, Bowman, Lalor, & Wilson, 2014 were selected from each Canadian province and territory (excluding Nunavut), and Alaska, USA, to obtain an even representation of individuals across their geographic range (N = 38, Table 2). Similarly, approximately three bobcat samples were selected from across the United States representing each of the genetic clusters identified by Reding, Bronikowski, Johnson, and Clark (2012). An additional three bobcat samples were selected from each of the Canadian provinces where there was a harvest for bobcat (N = 52, Table 2 Table 3). Some of these samples were the same as those used by Garroway et al. (2010Garroway et al. ( , 2011.  (Table 4).

| DNA extraction
As both the Lynx and Glaucomys samples were extracted for prior work, the Peromyscus samples were the only samples that required DNA extraction (described in Appendix S1).

| Selection, amplification, and genetic profiling of candidate clock gene fragments
The candidate gene fragments amplified for Canada lynx and bobcat, northern and southern flying squirrels, and white-footed and deer mice were nuclear receptor Rev-erbα (NR1D1), CLOCK, and PER1, respectively. The NR1D1 gene is a nuclear receptor that links circadian rhythms to transcriptional control of metabolic pathways and has been documented to play an important role in establishing and maintaining circadian body temperature rhythms of cold tolerance (Everett & Lazar, 2014;Gerhart-Hines et al., 2013). The CLOCK gene is a critical component of the circadian pathway, and the polyglutamine (PolyQ) motif within the CLOCK gene has been shown to play a role in regulating gene transcription (Darlington et al., 1998) and altering the corresponding circadian phenotype (Vitaterna et al., 1994). Specifically, Vitaterna et al. (2006)

| Analyses for signatures of selection
Initially, genotype distributions were assessed for all species to determine whether private alleles occurred in either species of each pair, where private cTNR alleles may indicate the differential evolution of or selection on cTNR alleles in northern versus southern closely related species. Bobcat samples were excluded from the remainder of the analyses due to low sample sizes preventing intraspecific analyses for selection in bobcats.
We used GenAlEx version 6.5 (Peakall & Smouse, 2006 to calculate allele frequencies, and observed (H O ) and expected (H E ) heterozygosity counts. We used Genepop version 4.2 (Raymond & Rousset, 1995;Rousset, 2008)  Population designations were determined differently for each species. Canada lynx are considered nearly panmictic across their range F I G U R E 2 Locations of sample sites of northern flying squirrels (Glaucomys sabrinus) and southern flying squirrels (Glaucomys volans) in Ontario, Canada. Shapes of points represent the species that was trapped at each site (northern flying squirrels, southern flying squirrels, and both species represented by squares, triangles, and circles, respectively). Outlined in red is the perimeter of Algonquin Provincial Park. The inset map in the top left corner shows an overview of the sampling area within North America with the ranges of northern and southern flying squirrels shown in gray and blue, respectively. Sampling site labels correspond to sample sizes in Table 3, and site-specific groupings are outlined by polygons  all sites within the "southern Kawartha" region (five sites; N = 15), and all sites within the Aurora region (three sites; N = 8) ( Figure 2).
All other sites were evaluated separately (Table 3). For southern flying squirrels, we grouped all sites just south of Algonquin Provincial Park (three sites; N = 27), keeping all other sites independent for analysis ( Figure 2, Table 3).  Table 4).
We tested for evidence of selection using a coalescent-based approach (Beaumont & Nichols, 1996) implemented in LOSITAN (Antao, F I G U R E 3 Locations of sample sites for deer mice (Peromyscus maniculatus) and white-footed mice (Peromyscus leucopus) in Ontario, Canada. Shapes of points represent the species that was trapped at each site (white-footed mice, deer mice, and both species represented by triangles, squares, and circles, respectively). Outlined in red is the perimeter of Algonquin Provincial Park. The insert in the top left corner shows an overview of the sampling area within North America with the ranges of white-footed and deer mice in blue and gray, respectively. Sampling site labels correspond to sample sizes in Table 4, and site-specific groupings are outlined by polygons where we first ran a simulation to remove potentially selected loci prior to computing the initial mean F ST , upon which putative adaptive loci were identified. We also selected the option to "force mean F ST ," in which LOSITAN will attempt to approximate a more precise F ST by running a bisection over repeated simulations. We ran 50,000 simulations at a 95% confidence interval and selected a stepwise mutation model, which is commonly used to describe STR markers (Antao et al., 2008;Fan & Chu, 2007). All other parameters were left at the recommended default settings. In cases where we identified a signature of selection at any locus, two additional independent tests were conducted on the same dataset for confirmation (i.e., a true signature of selection would be expected to persist in 3/3 tests). We also removed populations iteratively with replacement from each analysis to assess the sensitivity of our results to the exclusion of individual populations.

| Characterization of candidate clock genes in mammal species
In general, cTNR repeats within the coding regions of our candidate clock genes were relatively abundant (   Table 5). For the 1,791 Canada lynx samples that were analyzed for evidence of selection, none of the groups deviated from HWE at the NR1D1 locus. However, observed homozygosity was often slightly higher than expected, and observed heterozygosity slightly lower than expected (Table S2). For the neutral marker dataset, only the Yukon population significantly deviated from HWE at the Fca441 locus (p = .001).

| Glaucomys species and the CLOCK gene
A polyglutamine repeat motif (PolyQ) was successfully amplified in northern and southern flying squirrels. A total of nine alleles were observed at the CLOCK locus between the two species, which had largely overlapping allelic ranges; six of the nine alleles were found in both species. Two of the remaining three observed alleles were found solely in northern flying squirrels, and the third solely in southern flying squirrels; however, the frequencies of these three alleles were low (Table 6) however, many of the alleles present at this site were found at low frequencies (Table 6) Patterns of observed and expected homo-and heterozygosity were inconsistent across sites for both species. In some locations, common heterozygous genotypes were observed more than expected, whereas in others, the same genotypes were observed less than expected (Tables S3-S4).

| Characterization of candidate clock genes in mammal species
We observed a number of trends with respect to the presence, abundance, and purity of cTNRs in a range of mammal species. Repeats T A B L E 6 Allele frequencies of the coding trinucleotide repeat marker within the CLOCK gene in northern flying squirrels (Glaucomys sabrinus) and southern flying squirrels (Glaucomys volans) sampled in Ontario, Canada were generally abundant across most of our candidate clock genes.
Due to the overall short length, general abundance, and consistent purity of repeats that were 4 units (12 bp) or smaller, we suggest that the most promise for identifying selection will be found in repeats that surpass this threshold.
We found that longer repeats were sometimes more impure than shorter repeats; however, this trend was inconsistent across genes, and the overall relationship was not significant. Thus, the tendency for there to be a negative relationship between repeat purity and length may be tempered by variability among genes. Interestingly, impurity has been reported to significantly affect the stability of repeat structures (Pearson et al., 1998). A decrease of up to several orders of magnitude in the overall mutability of repeat fragments has been reported when impurities are present within shorter repeats, in multiple numbers, or near the center of the repeat unit (Ananda et al., 2014).
This suggests that impurities within exonic repeat structures may be selected for or against. For some genes, increased mutability and variation in repeat length might favor selection for pure repeats, whereas for other genes, a more stable repeat structure might be selected for to reduce maladaptation and essentially "lock-in" favorable geno-and phenotypes within an optimal functional range.
We also found that the longest repeats were in domestic rodent species, suggesting that domestics may experience elevated mutation rates in cTNRs (see also Laidlaw et al., 2007). Similarly, we found a large repeat in the naked mole rat, a blind species which does not rely on photoperiodic cues. It is possible that the expansion of cTNRs in both domestic and wild species that do not use photoperiod to cue life-history events could be caused by the lifting of evolutionary constraints on repeat size; however, this idea remains largely unsupported. Further, the considerable purity of these long repeats supports the claim that increased purity is the mechanism by which cTNRs mutate (Ananda et al., 2014;Gemayel et al., 2012;Kruglyak et al., 1998).
Thus, it is plausible that the removal of selective constraints, due to domestication or nonreliance on photoperiod, has removed the necessity for stable repeat structures to avoid maladaptation and thus facilitated the expansion of cTNRs in these species.

| Evidence of selection in candidate clock gene cTNRs of North American mammal species
We were able to detect signatures of selection at several candi- We found a complete divergence of alleles at the NR1D1 locus between Canada lynx and bobcat, supporting the role of selection in the separate Pleistocene evolution of closely related species, and subsequent adaptation of Canada lynx and bobcat to more northern and southern climatic habitats, respectively. While divergence times of these species may account for the divergence of alleles at the NR1D1 locus, the same pattern is not found in any of the neutral loci, whose allelic ranges are largely shared between species, thus supporting the role of selection in maintaining divergence at the NR1D1 locus. analyses, suggesting that this southern population is largely driving the signature of selection that we detected. It should be noted, however, that neutral genetic differentiation was also highest in pairwise comparisons including the "south of the St. Lawrence River" population (Koen et al., 2015).
Although CLOCK alleles were largely shared between northern and southern flying squirrels, there was evidence of differentiation in allele frequencies of some of the more commonly observed alleles.
This suggests that different alleles may be selectively favored over others in accordance with the differing habitats of the northern vs.
southern species. In addition to the divergence of allele frequencies, the greater genetic diversity of CLOCK alleles found in the northern flying squirrel may allow for greater fine-tuning capabilities in northern flying squirrel life-history strategies to cope with the more severe seasonal changes in their northern environment. We identified the CLOCK gene as within the range of balancing selection in LOSITAN for both northern and southern flying squirrels when the most southern (for both species) and northern (for northern flying squirrel only) sites were removed from the analysis, suggesting that these geographically "extreme" groups were influencing the signal of selection in both species.
As in the flying squirrels, differences in allele frequency distributions of the white-footed and deer mouse indicate potential differential selection between the two closely related species. Further, in the deer mouse, deviations from HWE and an excess of observed homozygotes at the most prevalent PER1 alleles in both large-and small-scale analyses of sites within and surrounding Algonquin Provincial Park suggest disruptive selection in this area, where small-scale changes in environmental features (e.g., microhabitats) may drive the selection of particular alleles in slightly different environments. These results would be predicted if there were mice with predispositions to breeding at different times of the year, essentially "isolation by time" (Hendry & Day, 2005). Pairwise estimates of genetic differentiation at our set of neutral microsatellites and the PER1 locus showed contrasting patterns for white-footed and deer mouse, indicating that diverse processes may be influencing the two closely related species differently.
We found that the northern residing species had lower allelic diversity at the PER1 locus, contrary to our expectations that we should find higher allelic diversity to allow for greater seasonal fine-tuning capabilities (as observed in flying squirrels). However, the sampled gradient used here is small with much of the sampled area being inhabited by both species; a wider latitudinal gradient may clarify these results.
Lastly, a persistent signature of balancing selection was observed in the white-footed mouse when the Algonquin Provincial Park site (the most northern sampled site) was removed from the LOSITAN analyses.
F I G U R E 8 Estimates of F ST at five neutral microsatellite loci (black points) and the PER1 cTNR locus (red points) estimated by (a,c) locus and (b,d) pairwise population comparisons at the "trapping site-specific" small scale (a,b), and "regional" large scale (c,d) for deer mice (Peromyscus maniculatus) sampled within Ontario, Canada. The x-axis is ordered by increasing genetic differentiation of neutral loci, ending with estimates of mean F ST (across loci and pairwise comparisons) for neutral and cTNR loci with standard error bars. Population abbreviations are as follows: APP (Algonquin Provincial Park), sAPP (south of Algonquin Provincial Park). Numbers beside the "APP" label represent specific sites in Algonquin Provincial Park (1-5). All other population labels are written in full Combined, our set of analytical approaches was able to detect signatures of selection in the cTNRs of candidate clock genes in an array of North American mammals along latitudinal clines. Our results suggest that these techniques are useful for surveying candidate genes in non-model species and that cTNRs are interesting markers to investigate in reference to mammalian adaptation. Further, the diversity of analyses used to detect selection in these genetic markers suggests that testing for indications of selection is best when multiple approaches are implemented that are able to detect different types of selective processes (positive, balancing, divergent, etc.). While the signatures we detected in our datasets point to the potential for adaptive differences at these coding motifs, our findings support the need for more extensive characterization of populations (e.g., assessing correlates with environmental variables).

| Limitations on the detection of selection
The influence of sample size on the accurate detection of selection signatures is important to consider given some of our datasets.
Lachance (2009)  F I G U R E 9 Estimates of F ST at five neutral microsatellite loci (black points) and the PER1 cTNR locus (red points) estimated by (a,c) locus and (b,d) pairwise population comparisons at the "trapping site-specific" small scale (a,b), and "regional" large scale (c,d) for white-footed mice (Peromyscus leucopus) sampled within Ontario, Canada. The x-axis is ordered by increasing genetic differentiation of neutral loci, ending with estimates of mean F ST (across loci and pairwise comparisons) for neutral and cTNR loci with standard error bars. Population abbreviations are as follows: APP (Algonquin Provincial Park), sAPP (south of Algonquin Provincial Park), SLINP (St. Lawrence Islands National Park). Numbers beside the "sAPP" and "Kawartha" labels represent specific sites south of Algonquin Provincial Park (1-3) and Kawartha (1-2), respectively. All other population labels are written in full Raquin et al., 2008). Most genes, on the other hand, act in pleiotropy (Harrisson et al., 2014), which produces more modest increases in allele frequencies over multiple loci (Hermisson & Pennings, 2005) and is less likely to affect the patterns of divergence between populations. Indeed, LOSITAN does not address the issue of non-linearity of F ST estimates that approach zero and, thus, is unlikely to detect low-F ST outliers when selection is not strong (Antao et al., 2008). In such cases, the detection of weak, polygenic selection requires much larger sample sizes than those required to detect hard sweeps.
It is also possible that cTNR motifs may not be the specific regions under selection in adaptive genes but are rather linked to other genes or regulatory elements under selection. It has been suggested that multilocus metrics of linkage disequilibrium (LD) are better suited for the detection of selection, as selection will result in LD adjacent to the selected locus (Ennis, 2007). It also has the added benefit of bearing the footprint of past selection (Lachance, 2009), which can provide important information on the genetic responses of species under past environmental change. Even if cTNRs are not the markers under selection, their increased variability and linkage to coding SNPs and regulatory elements could still be an important proxy for detecting candidate adaptive genes through haplotype profiling and analyses.
Although our approach was successful at identifying putative patterns of adaptive genetic divergence in our range of mammal species, further work is required to fully realize the mechanisms underlying the observed functional variation in these species, as it has been demonstrated that the mechanisms underlying phenotype-genotype correlations can differ between closely related species (Rosenblum, Römpler, Schöneberg, & Hoekstra, 2010). The general complexity of molecular mechanisms coupled with the pleiotropic effects of many genes argues for caution in interpretation of our results; however, we feel that we have provided support for the importance of cTNRs as targets of natural selection and adaptation in wild populations.

| Potential adaptive importance of cTNR loci
In the recent past, there has been an advancement in empirical studies implicating climatic and environmental gradients in the generation and maintenance of adaptive genetic diversity through selection, even in the face of ongoing gene flow between populations (DeFaveri et al., 2013;Fang et al., 2013;Watanabe, Kazama, Omura, & Monaghan, 2014). Clock genes in particular have been demonstrated to be important targets of selection, as they are likely to provide a means by which species can adapt to seasonal changes or adjust to novel environments (Kondratova et al., 2010). Our utilization of a candidate gene approach allowed us to identify and target cTNRs within clock genes characterized in closely related model organisms for which functional roles have been identified. Not only did this type of approach make optimizing gene fragments easier by facilitating primer design, but it also allowed us to use prior knowledge of gene function to develop a priori hypotheses regarding the environmental and climatic factors that may be driving selection. Thus, this methodology is useful in determining cTNR-containing genes that are good candidates for environmental association studies that correlate environmental variants with adaptive genetic variability in a spatially explicit framework (e.g., latent factor mixed models, LFMM; Frichot, Schoville, Bouchard, & François, 2013). This will lead us one step closer to being able to accurately characterize genotype-phenotype associations in wild populations.
Gene motifs, specifically cTNRs demonstrating both genetic and epigenetic characteristics (Gemayel et al., 2010(Gemayel et al., , 2012, may provide high-pace adaptive capabilities, making them ideal targets for mitigating the decline of species at risk through the identification of adaptively significant populations. A critical development in modeling a species' natural resilience (Dawson et al., 2011) and implementing solutions (e.g., Thomas et al., 2012) is mapping and promoting environments to maintain critical standing adaptive genetic variation and the potential generation of novel adaptive alleles; cTNRs offer the potential to support both of these objectives. We present here a methodology by which we were able to identify cTNRs that are potentially the targets of natural selection in a range of mammal species, a taxonomic group underestimated in terms of vulnerability to climate change. Variance at cTNR motifs in other genes may provide a mechanism for rapid evolutionary responses to a range of other phenotypes. Thus, the abundance of cTNR repeats in functional gene classes including but not limited to clock, immunity, behavioral, morphological, and stress-axis genes translates into a resource list of hundreds of candidate genes that can be used in the search for rapidly evolving motifs associated with adaptation in wild species.