In this study, we combined population genetics, bioinformatics, and experimental approaches to identify a cis-regulatory element harboring a SNP (rs11868112) associated with a strong signal of selection identified in a genome-wide study. The allele frequencies at this SNP are strongly correlated with latitude and winter temperature variables. A re-sequencing survey did not identify additional SNPs that are in strong LD with rs11868112 and that are likely to have functional effects. Because SNP rs11868112 lies within a predicted POU2F1 binding site located close to a RARA binding site identified by ChIP-chip, we hypothesized that this SNP influences the transcriptional response to RA. Consistent with this hypothesis, we showed that POU2F1 and RARA do bind to the genomic region spanning and near SNP rs11868112, respectively. Furthermore, reporter gene assays suggest that this region functions as a RA-dependent enhancer and that the allele status at rs11868112 affects enhancer activity. Although we cannot conclusively identify the target gene of this enhancer, RPTOR appears to be a strong candidate because it is induced by the selective RARA agonist AM580 in two different cell lines. The fact that the time of differential RPTOR expression in response to RA treatment coincides with the time of allele-specific down-regulation in reporter gene assays further supports this proposal. Overall, these results provide an example of how a selection signal can identify a functional SNP and suggest a role for the regulation of RPTOR expression in human adaptations to different environments.
Despite the clear signal of selection given by the correlation between allele frequency and climate variables, standard neutrality tests did not detect a significant departure from expectations. This included tests of the frequency spectrum as well as haplotype homozygosity (as assessed by the extended haplotype homozygosity 
or integrated haplotype score statistics 
). However, these tests are powerful when selection acts on a new rather than an existing mutation 
and when selection acts on a dominant or codominant allele 
. In the case of SNP rs11868112, the beneficial allele in cold climates segregates at appreciable frequencies in sub-Saharan African populations (8%–43%), thus suggesting that this variant predates the dispersal of human populations out of Africa and that this allele was neutral before becoming advantageous when humans moved to colder climates. Modeling studies have shown that under these circumstances standard neutrality tests have inadequate power to detect a signal of selection. For example, simulations of a model of directional selection on standing variation determined that, if a neutral allele occurred at frequency greater than 5% prior to becoming advantageous, virtually no signature is expected on the frequency spectrum, on patterns of linkage disequilibrium and on polymorphism levels 
. Given the relatively high frequency of the derived allele at rs11868112 in sub-Saharan African populations, it is therefore not surprising that we detected a signature of natural selection only by using the climate correlation approach.
The molecular function of RPTOR
is consistent with a role in local adaptations. This is because the MTORC1 complex, which contains RPTOR as a critical component, integrates environmental signals to regulate cell growth, metabolism and survival (
and references therein). However, given the diversity of biological processes regulated by MTORC1 and by retinoids, it is difficult to pinpoint the function of RPTOR that is the most likely target of selection. One possibility is that SNP rs11868112 influences the regulation of energy metabolism and mitochondrial function; under this model, the derived allele would have conferred a selective advantage by increasing thermogenesis during the dispersal to progressively colder climates. This scenario is supported by the fact that the MTOR pathway controls mitochondrial function, especially ATP synthetic capacity 
, directly 
and indirectly 
. Tissue-specific knockouts (KOs) of raptor
have provided further support for the critical role of mTORC1 signaling on whole body metabolism. Adipose-specific raptor
KO mice are resistant to diet-induced obesity, due to an increased mitochondrial uncoupling in white adipose tissue 
. Because mitochondrial uncoupling is an important mechanism for generating body heat, it is possible that the selective advantage conferred by SNP rs11868112 is due to its effect on thermogenesis and cold tolerance. Adaptations to cold climates are evident in the geographic distributions of many traits. For example, significant correlations exist between body mass and temperature 
, consistent with the long standing hypotheses that variation in body size and proportions are adaptations to maintain temperature homeostasis 
. Furthermore, there is evidence that human metabolism has been shaped by adaptations to cold stress from studies of arctic populations, which exhibit elevated basal metabolic rates compared to non-indigenous populations 
Although these geographic patterns of human phenotypes are well established, the genetic factors contributing to these adaptations are only partially understood. Genetic variants that can increase mitochondrial uncoupling efficiency might be advantageous in cold climates and thus may have been selected during human migrations northward. The derived T allele rs11868112, which is associated with lower RA-dependent transcription levels, may result in increased mitochondrial uncoupling in adipose tissue and, hence, higher thermogenesis. Consistent with the hypothesis that this allele confers resistance to cold stress, it increases in frequency with decreasing winter temperatures in worldwide population samples, and it is relatively rare in the equatorial populations of sub-Saharan Africa and South East Asia (see ). Although a role for RA in the transcriptional induction of RPTOR
was not previously reported, retinoids have been long known to regulate thermogenesis and energy expenditure through their effects on the expression of the major uncoupling protein gene, UCP1
, in brown adipose tissue (BAT) 
. This protein plays a key role in nonshivering thermogenesis, which is the main mechanism for heat generation in human infants. As with the RPTOR
gene, a distal enhancer upstream to the proximal UCP1
promoter was found to contain a RAR response element, which mediates its transcriptional induction 
. Therefore, our finding that the transcription of the RPTOR
gene is induced by RA treatment in liver and breast epithelial cells is consistent with the known function of RAR in regulating the expression of another gene with a clear role in thermogenesis. More recently, experiments in mice identified a role for retinoids in adipose tissue remodeling and, more specifically, in the acquisition of BAT-like properties in white adipose tissue 
. These findings further support the notion that the biological functions of RAR include the regulation of energy balance and thermogenesis. Moreover, our observation that RPTOR
is induced by retinoids provides a possible mechanistic link connecting the action of retinoids in adipose tissue remodeling and the finding that the adipose-specific raptor
KO exhibits increased mitochondrial uncoupling (i.e. a typical BAT property) in white adipose tissue.
Another possible explanation for the signal of selection observed at SNP rs11868112 may be related to the role of the MTOR pathway in the regulation of the immune response. Indeed, the MTOR pathway plays multiple roles in immunity, especially in the activation and proliferation of T cells 
, and has been implicated in the etiology of autoimmune disorders, such as systemic lupus erythematosus (SLE) 
. Moreover, the specific inhibitors for this pathway, rapamycin and its derivatives, can decrease proliferation of T lymphocytes and are used as immunosuppressant to avoid allograft rejection 
or to treat autoimmune patients 
. Since pathogen diversity decreases with latitude mainly as a result of climatic factors 
, the optimal level of immune response is also expected to vary according to latitude and climate. Therefore, we hypothesize that the increase in the frequency of the derived T allele with decreasing temperatures is due to selective pressures acting on the MTORC1 function in the regulation of the immune response. Under this scenario, the decrease in RPTOR
expression associated with the T allele at high latitudes could reflect a shift to maintain the appropriate balance between pathogen pressures and immune response, with an exaggerated immune response possibly resulting in increased risk for autoimmune diseases.
It might be argued that the modest (~20%) decrease of RPTOR
expression associated with the T allele is not sufficient to generate significant phenotypic and fitness differences between arctic and tropical populations. One possibility is that RPTOR and the MTOR pathway are located at the top of the signaling cascade 
, therefore, a subtle change in its expression can have major consequences. Alternatively, as observed for most susceptibility SNPs identified through GWAS of common diseases, rs11868112 may be just one of many SNPs with small effects on the phenotypes that are adaptive in different climates.