Human local adaptation of the TRPM8 cold receptor along a latitudinal cline

Abstract

Ambient temperature is a critical environmental factor for all living organisms. It was likely an important selective force as modern humans recently colonized temperate and cold Eurasian environments. Nevertheless, as of yet we have limited evidence of local adaptation to ambient temperature in populations from those environments. To shed light on this question, we exploit the fact that humans are a cosmopolitan species that inhabit territories under a wide range of temperatures. Focusing on cold perception-which is central to thermoregulation and survival in cold environments-we show evidence of recent local adaptation on TRPM8. This gene encodes for a cation channel that is, to date, the only temperature receptor known to mediate an endogenous response to moderate cold. The upstream variant rs10166942 shows extreme population differentiation, with frequencies that range from 5% in Nigeria to 88% in Finland (placing this SNP in the 0.02% tail of the FST empirical distribution). When all populations are jointly analyzed, allele frequencies correlate with latitude and temperature beyond what can be explained by shared ancestry and population substructure. Using a Bayesian approach, we infer that the allele originated and evolved neutrally in Africa, while positive selection raised its frequency to different degrees in Eurasian populations, resulting in allele frequencies that follow a latitudinal cline. We infer strong positive selection, in agreement with ancient DNA showing high frequency of the allele in Europe 3,000 to 8,000 years ago. rs10166942 is important phenotypically because its ancestral allele is protective of migraine. This debilitating disorder varies in prevalence across human populations, with highest prevalence in individuals of European descent-precisely the population with the highest frequency of rs10166942 derived allele. We thus hypothesize that local adaptation on previously neutral standing variation may have contributed to the genetic differences that exist in the prevalence of migraine among human populations today.

Fig 1. Overview of the populations used and their allele frequencies for rs10166942, average temperature, and F_ST signatures.

(A) Geographic location of the 1KGP populations used, with the derived allele frequency of the rs10166942 allele in pie charts (T allele in color according to population), and their latitude. (B) In columns, annual mean temperature at the geographic location of each population, the level of F_ST-based population differentiation with YRI, the log₁₀ empirical P-value of this F_ST value, and the proportion of SNPs in the 65 kb target region with an empirical P-value lower than 0.05.

Fig 2. Correlation between latitude and derived allele frequency.

Correlation of the frequency of the rs10166942 T allele with latitude. The fitted function (dashed line) results for the 1KGP data from (A) the PGLS and (B) GLMM analysis. (C) Results of the best model in the GLMM analysis of the SGDP dataset. The fitted response is shown as gridded surface, and the dots represent the average frequency of the rs10166942 T allele per cell of the gridded surface. Points above the surface are filled, points below are open. The volume of the points corresponds to the number of populations per cell.

Fig 3. ABC analysis.

(A) Graphical representation of the three models (SSV, SDN, NTR) and their associated parameters. Birth of the allele and start time of selection are shown by black and red lines, respectively. The range of the prior distribution for time of selection start is depicted by a star and a blue line. A double headed arrow indicates population migration. (B) Posterior probabilities for each model and population. (C) Prior distribution of each parameter as a histogram. Posterior distribution of the SSV model parameters as a line for each population.