The rs1421085 variant within FTO promotes thermogenesis and is associated with human migration

One lead risk signal of obesity–rs1421085 T > C within the FTO gene–is reported to be functional in vitro but lack of organismal evidence. Here, we established global and the brown-adipocyte specic locus-knock-in mice to recapitulate this homologous variant in humans and discovered that mice carrying the C-alleles showed increased thermogenic capacity and a resistance to high-fat diet-induced adiposity with enhanced FTO transcription, while FTO knockdown or inhibition effectively eliminated the increased thermogenic ability of brown adipocytes. In humans, the C-allele was associated with lower birthweight, and its allele frequency increases following the environmental temperature decreases. Cumulatively, these ndings identied rs1421085 T > C as a functional variant promoting whole-body thermogenesis and was associated with early human migration from hot to cold environments.


Introduction
Following genome-wide association studies (GWAS), enormous common genetic variants in the human genome have been unraveled and proved to be associated with increasing risk of diseases development 1 . However, few of these loci show biological evidence to support their causality with diseases 1,2 . One main challenge is that most positive signals are dispersed within the noncoding regions which are less conserved among different species, hindering further functional studies using genetically modi ed animal models 3,4 Despite successful applications of gene editing in human cells, functional examinations at the organism level to recapitulate the speci c variant in humans are still lacking 5 .
To date, the largest obesity-GWAS study has revealed ~ 1000 common polymorphisms, explaining ~ 6.0% variation of body mass index (BMI) 6 . Interestingly, the positive signals repeatedly occur nearby the known monogenic obesity genes, e.g., MC4R, LEPR and BDNF 4 . Notably, the strongest signals of the common variants are localized within the introns of the FTO gene and show consistent risk-effects across multiple ethnic populations 6,7 . However, there are no evidence on FTO mutations causing obesity in clinical settings 8 . Loss-of-function mutations of human FTO are lethal owing to serious developmental defects and growth retardation 9 . Higher mortality was also observed in global Fto knockout mice, and the survived mice exhibited a short stature and a lean phenotype 9 . In contrast, the adipocyte-speci c Fto knockout mice gained more body weight and showed adiposity under HFD compared with WT littermates 10 . In the past decade, the roles of FTO and its genetic variants in adiposity remained a topic of vigorous debate.
A turning point occurred when two research groups independently argued that certain loci (especially rs1421085) within the FTO variant cluster enhanced the transcription of several distal genes (including IRX3 and IRX5) but not FTO per se 11,12 . Very recently, the Nóbrega group reported the functional regulation between three additional functional FTO "risk variants" and the candidate genes by using a gene-editing animal model harboring 20 kb deletion of the orthologous obesity-associated interval in the Fto gene 13 ; while the Claussnitzer and Cox groups further reported a mouse model with a deletion of 82 bp containing the homologous rs1421085 locus in FTO which yield a mildly lean phenotype in mice under high-fat diet (HFD) challenge 14 . Although these genetic editions contain the deletion of the FTO rs1421085 loci, they might also delete other potential function elements around, hence a precise knock-in mouse model of the FTO genetic risk loci (e.g., rs1421085_C-allele) to speci cally recapitulate the human variants is still in need to understand the function of the speci c risk loci.
Towards this end, we established two global and one brown adipocyte-speci c single-nucleotide knock-in mouse models of the homologous human rs1421085 T > C variant, the lead functional variant in the FTO gene 12,15,16 (Fig. 1, A, B and K; Fig. 3A; g. S1, A-C), and found consistent enhancement of thermogenesis and resistance to HFD-induced obesity, at least in part, by increasing FTO expression. Interestingly, the Callele was associated with lower bodyweight of the human newborns, who have functional brown adipose tissue (BAT) as rodents do 17 . Furthermore, the C-allele frequency in multiple ethnic populations was closely related to ambient environmental temperature where they settled, with a stepwise increasing pattern from hot Africa shifting to cold Siberia.

Results
The homologous rs1421085 T > C variant promotes energy expenditure and resists HFD-induced adiposity Evolutionary conservation around the T allele of human rs1421085 and mouse homologous loci was evaluated, which showed a high similarity (Fig. 1A). To clarify the biological function of human rs1421085 T > C variant in vivo, a global knock-in mouse model carrying the homologous singlenucleotide mutation (the homozygous rs1421085_CC, termed KI cas9 ) was constructed using CRISPR/Cas9 system, with wild-type (WT, rs1421085_TT) littermates as controls ( Fig. 1B and g. S1A). No signi cant difference on body weight gain between the two groups was observed when fed a normal chow-diet (NCD) ( g. S2A). Unexpectedly, under a HFD (60% fat) challenge, male KI cas9 mice gained less body weight compared with WT littermates (Fig. 1C). In agreement with the lean phenotype, KI cas9 mice showed an improvement of blood glucose and lipid levels as well as decrease lipid storage in the liver ( g. S2B-H). KI cas9 mice also had reduced total fat mass content and less inguinal white adipose tissue (iWAT) ( Fig. 1D and g. S2I). Despite no obvious alterations in food intake, physical activity, or respiratory exchange ratio (RER) ( Fig. 1E; g. S2, J and K), KI cas9 mice exhibited increased O 2 consumption (Fig. 1, F and G) and CO 2 production (Fig. 1, H and I), indicating a higher basal energy expenditure (Fig. 1J). We validated the HFD challenge in the following KI cas9 generation and gained consistent results ( g. S2L). In addition, a similar lean phenotype was also observed in female KI cas9 mice ( g. S2, M-Q).
To further con rm the phenotypes and exclude the possibility of off-targeted editing by CRISPR/Cas9 method, we used the conventional gene targeting strategy with homologous recombination applied to mouse embryonic stem cell (ES) to construct a second global rs1421085_CC knock-in (termed KI ES ) mouse model ( Fig. 1K and g. S1B). Both male and female KI ES mice showed no difference in body weight as compared with control littermates under NCD ( g. S3, A and B), while showed less body weight gain, a decreased trend of fat mass and attenuated fat accumulation in liver than control littermates under HFD challenge (Fig. 1, L and M; g. S3, C-I). These results suggested that homologous rs1421085 T > C variant resisted HFD-induced adiposity.
The homologous rs1421085 T > C variant augments the thermogenic capacity of brown adipocytes Histological analysis showed no signi cant difference in the morphology of white adipose tissues between KI cas9 mice and WT littermates under HFD challenge (Fig. S2H). However, we observed that KI cas9 mice had more condensed and smaller adipocytes in BAT while WT mice showed a brown-to-white change under HFD ( Fig. 2A). Consistently, the expression of thermogenesis-related genes, including Ucp1, Pgc-1α, Prdm16, Cidea, Elov6, Cox7a and Cox8b, were signi cantly increased in BAT of KI cas9 mice ( Brown adipocyte-speci c knock-in of homologous rs1421085 T > C variant enhances thermogenesis and resists HFD-induced adiposity To test the hypothesis that homologous rs1421085 T > C variant in BAT might primarily contribute to the obesity-resistant phenotype, we next constructed another brown adipocyte-speci c "risk allele" knock-in model by crossing homologous rs1421085 oxp/ oxp (KI / ) mice with Ucp1-Cre mice (Ucp1-Cre; rs1421085 oxp/ oxp , termed Ucp1-KI / ) ( Fig. 3A and g. S1C). The potential interference effects of the oxp insertion per se on the expression of reported targets was excluded ( g. S4, A-E). Similar to the two global knock-in models, Ucp1-KI / mice showed no obvious change in body weight under NCD ( g. S4F), while displayed a resistance to HFD-induced obesity, indicated by a decrease of body weight and fat mass in comparison with control littermates (Fig. 3, B and C). Smaller lipid droplets and more mitochondria were observed in the BAT of Ucp1-KI / mice, along with a moderate improvement of lipid accumulation in liver (Fig. 3, D and E; g. S4, G-H). In consistence, an increased expression of Ucp1 and other thermogenesis-related genes was observed in BAT of HFD-fed Ucp1-KI / mice (Fig. 3, F-H). Taken together, with the phenotypes of the two global and one brown adipocyte-speci c knock-in models, we illustrated that the biological effect of the homologous rs1421085 T > C variant in BAT prominently contributed to the increased thermogenesis and the resistance to HFD-induced adiposity.

The homologous rs1421085 T > C variant increases thermogenesis via upregulating Fto expression
Several genes neighboring the rs1421085 locus are regarded as effective candidates of the FTO SNP cluster, e.g. IRX3/IRX5/IRX6 12 , RPGRIP1L 10,15,19,20 , and FTO 21,22 , among which IRX3 and IRX5 have been shown to increase in subcutaneous preadipocytes of C-allele carriers 12 . We unbiasedly and systemically evaluated the expression of all the above genes in the BAT of three mouse models (KI cas9 , KI ES and Ucp1-KI / ) and induced mature brown adipocytes of KI ES mice, compared with each control. An increase of Irx3 expression was detected in KI cas9 BAT ( g. S5A), and increased trends of Irx3 expression were also observed in other KI models ( g. S5, B-D). No signi cant difference in other candidates was observed (despite increased trends in certain context, like Rpgrip1l in KI ES BAT) ( g. S5, A-D). Importantly, moderate and sustainable elevation of Fto expression was detected in both mRNA and protein levels in each KI model (Fig. 4, A-C). Indeed, the 1 kb genome region containing the "risk allele" of rs1421085 (as well as rs9940128 and rs11642015) showed higher enhancer activity than the major allele 12,13 ; however, the downstream target-promoter of these variants has not been clari ed. In this regard, four reporter plasmids carrying the 1 kb human genome fragments centered on rs1420185 (T and C allele, respectively) inserted ahead of the human FTO promoter (hFTOp) sequence either in a forward direction (5' to 3') or reverse direction (3' to 5'), were constructed to investigate the direct enhancer effect of this variant on the activity of FTO promoter per se (Fig. 4D). When linked in a reverse direction, the major allele (T) 1-kb genome fragment enhanced the transcription activity of FTO promoter, and the "risk allele (C)" genome fragment further advanced the activity. This phenomenon did not occur once the fragments were arranged in a forward direction ahead of the promoter (Fig. 4E).
The biological role of FTO in brown adipocytes remains elusive 10 . We found that during the in vitro differentiation of brown adipocytes, Fto expression increased continuously, in parallel with the increase pattern of thermogenesis-related genes such as Ucp1 and Pgc-1α ( g. S6A). A strong positive linear correlation between Fto and Ucp1 mRNA expression was observed ( g. S6B). Of importance, Fto knockdown reduced the mRNA and protein expression of Ucp1 and other thermogenesis-related genes remarkably, and Fto de ciency impaired the thermogenic capacity of induced brown adipocytes in a dosedependent manner ( To better illustrate the role of FTO in functional BAT pads, and to avoid the secondary interference of growth defects observed in global Fto knockout mice 3 , we crossed Fto oxp/ oxp mice with Myf5-Cre mice 23 to delete Fto in BAT (Myf5-Cre;Fto oxp/ oxp , termed MFKO) ( g. S7A). Under HFD challenge, MFKO mice showed an increased body weight gain ( g. S7, B and C) and a decline of energy expenditure as compared with control mice ( g. S7, D-K). Of importance, ex vivo experiments showed that UCP1 protein expression decreased upon FTO deletion in the induced brown adipocytes from MFKO mice ( g. S7, L and M).
We further performed RNA immunoprecipitation followed by qPCR (RIP-qPCR) and agarose gel electrophoresis (AGE) tests, and observed that endogenous FTO protein bound to Ucp1 mRNA in mature brown adipocytes ( g. S8, A-C). We also observed an interaction of FTO protein with C/ebpα mRNA ( g. S8, B and D), consistent with previous ndings in NOMO-1 cells 24 . Moreover, FTO de ciency signi cantly reduced the stability of Ucp1 transcripts while merely affect that of C/ebpα, suggesting that Ucp1 was one main target of FTO protein in mature brown adipocytes ( g. S8, E and F). Importantly, the increased Ucp1 expression and enhanced OCR of KI ES brown adipocytes was greatly repressed with either silence of Fto gene or enzymic inhibition of FTO protein (Fig. 4, L-O; g. S8, G-H). We noticed an incomplete elimination of the differences between groups, which indicated that other downstream genes of the variant (like Irx3 12 ) might also be involved in the process 25,26 . Taken together, these ndings revealed that the effects of the homologous rs1421085 T > C variant on the increasing thermogenesis and Ucp1 expression depended largely on the FTO protein.
The rs1421085 T > C variant is associated with lower body weight of human infants and its allele frequency parallels with latitudes and ambient temperature of population distribution Rodents possess thermo-active BAT throughout life, whereas in humans BAT activation peaks after birth to combat the extrauterine coldness for survival and then vanishes with growth 27 . To evaluate the potential contribution of BAT degeneration on the adiposity-related outcomes of rs1421085 T > C variant, we surgically removed intrascapular BAT (iBAT) by iBAT excision (△iBAT) 28 in KI cas9 and WT littermates ( g. S9A). Without iBAT, KI cas9 mice lost adiposity resistance under HFD; on the contrary, they appeared to have more body weight increase, higher fat mass percentage, larger white adipocytes and more severe ectopic lipid deposition in the liver when compared to controls ( g. S9, B-F), in agreement with the adiposity traits of the adult humans carrying the "risk-allele".
Given the substantial biological effects of homologous rs1421085 T > C variant on murine BAT function and adiposity, we hypothesized that rs1421085 T > C variant may associate with a lower body weight in humans carrying functional BAT. Indeed, a few well-designed longitudinal studies have indicated that the association between FTO SNPs and body weight changes dynamically with growth 29 . Thus, we performed a meta-analysis to explore the effect of FTO polymorphism (including rs1421085 and other three linkage disequilibrium SNPs) on body weight/BMI. In agreement with previous ndings [30][31][32][33][34][35][36][37][38] , BMI difference between the effect (minor)-allele and the reference (major)-allele groups emerged since about 8 years old, as a higher average BMI was observed in the effect-allele group ( g. S10A); this difference disappeared when the included population was restricted to younger than 8 years old ( g. S10B). Importantly, we observed a mild but signi cant decrease of birth weight in the effect-allele group (P = 0.009, 95% CI: [-0.11, -0.02]; Fig. 5A). As only a few of studies described the histological and molecular characteristics of human fetal BAT 39 , we collected fetal BAT and con rmed the existence of functional BAT even at middle-gestational stage ( g. S11). In further, we found the co-localization of FTO and UCP1 protein in fetal brown adipocytes featured with strong UCP1 and PLIN1 staining ( g. S11), suggesting a potential role of FTO in fetal BAT development and human thermogenesis.
Although a strong positive association between FTO SNPs (including rs1421085) and obesity is con rmed across adult populations of diverse ancestry, a relatively weaker association and a lower frequency of "risk alleles" have been observed according to large-scale GWAS studies based on African populations 16,40,41 . To date, reasons for the discrepancy among African and non-African populations remains unclari ed. Herein, we assumed that environmental coldness, which activates BAT but rarely occurred in Africa, might contribute to the low frequency distribution of the FTO variants. To this end, we rst analyzed ambient environmental temperatures with the allele frequency of the rs1421085 T > C variant in various populations (of all the 1000 Genomes populations, see Methods), and observed a signi cantly inverse correlation in populations settled in Africa-Eurasian continents (P = 0.0001, R 2 = -0.33; Fig. 5B). The inverse correlation remained signi cant after adjusting Paleozoic temperature (which possibly affected BAT function in ancient humans) using two published models 42 (P < 0.0001, R 2 = -0.40, and P < 0.0001, R 2 = -0.39, respectively; g. S12, A and B). In addition, we analyzed the correlation between the variant frequency and the coordinates latitudes (one determinant factor of ambient environment temperature) of all populations and observed a more robust positive association (P < 0.0001, R 2 = 0.48) (Fig. 5C). Meanwhile, no statistically signi cant correlation was found between the variant frequency and longitudes or altitudes among populations ( g. S12, C and D).
Adaption to coldness outside Africa, especially during the last ice age, was a major challenge for early Homo sapiens migrating from East Africa 43 . Acclimatization through genetic accommodation during human migration is a prevailing theory, e.g. the EDAR V370A mutation selected by warm and humid Asia environment 44 , the unique EPAS1 haplotype structure selected by the hypoxic environment of the highaltitude Tibetan platens 45 . From this aspect, with modern humans leaving Africa, migrating, and nally settling on other continents, their ambient temperature/residential location shifted from a warm/low latitude to cold/high latitude. While referring to human migration route maps reported by previous studies 46 , we proposed a genetic diffusion map of the rs1421085 T > C variant frequency on the African and the Eurasian continents, that is, the frequency was lowest in East Africa, intermediate in East and South Asia, increased in Europe, and nally reached its zenith in Siberia (Fig. 5D).

Discussion
Immediately after the identi cation of the association between FTO variants and obesity, Palmer's group made an elegant and well-controlled comparison of the energy balance of FTO "risk allele" carriers versus non-carriers in 4-10 years old children 47 . An indirect calorimetry approach was applied to assess basal metabolic rate (BMR), while the doubly labeled water method (used as golden standard 48 ) was employed to measure total energy expenditure. Paradoxically, they revealed that the "risk allele" carriers had a higher BMR and total energy expenditure than non-carriers. Nevertheless, despite that covariate adjustment partially abolished the "risk allele" effects on BMR, the total energy expenditure remained signi cantly different between the two groups (an increase of 1160 kJ in "risk-allele" carriers). Unfortunately, these important ndings were ignored as subsequent studies overwhelmingly argued that the "risk allele" had no or a reduction effect on energy expenditure in adults 12,49 . The causal effects of the FTO risk allele at the organism level remain to be illusive. It is emphasized that humanized murine models are powerful tools to dissect mechanisms underlying the association of FTO variants with human phenotypes 50,51 . To our knowledge, to date, there is no report regarding murine models to exactly recapitulate human FTO polymorphisms with single-nucleotide substitution methods. Towards this end, we initially established two strains of global homologous rs1421085 T > C variant knock-in mice using CRISPR/Cas9 and conventional gene-targeting with homologous recombination strategies, respectively. A third strain of brown adipocyte-speci c homologous C-allele single-point knock-in mice was constructed to con rm its biological roles in BAT. In support of Palmer's ndings in children 47 , our study reported the FTO variant yielded an enhancement of thermogenic gene expression, oxygen consumption, and total energy expenditure of brown adipocytes both in vivo and ex vivo. Another important nding is that the rs1421085 T > C variant is indeed a pivotal genetic trigger modulating a line of downstream genes, among which Fto showed the most stably increased expression in "risk-allele" group. Furthermore, we provided a reliable evidence of FTO protein directly regulating Ucp1 expression both in vivo and in vitro, which was consistent with the obese phenotype observed in adipocyte-speci c Fto knockout mice 10 , pointing out the involvement of FTO in the function of the rs1421085 T > C variant. Except for Fto, we also validated an increased expression of Irx3, Irx5 [11][12][13] and Rpgrip1l 15 in certain contexts, supporting a temporal, spatial, and cell type-dependent regulation by the variant 13 . Furthermore, we have previously revealed that IRX3 overexpression promoted the transcriptional activity of Ucp1 and enhanced thermogenesis in brown adipocytes 25,26 , which might be one plausible explanation for the incomplete rescue of FTO knockdown or inhibition to the rs1421085 T > C variant's effects. We want to mention that these ndings in the knock-in models of this single-nucleotide variant is not completely the same as in the recently reported two mouse models harboring a deletion of 82 bp and 20 kb containing the same FTO variant, respectively 13,14 , which may be due to non-speci c deletion of other potential functional motifs in the region by their strategies. Our models that recapitulate the same human variant are more speci c for the evaluation of the pathophysiological changes produced by the homologous rs1421085 T > C variation at the organism level.
The rst available document about BAT deposition in infants was reported by Neumann in 1902 52 . BAT in human newborn infants produces heat by non-shivering thermogenesis to maintain body temperature.
When the infants were shifted from 34-35℃ (thermoneutral temperature) to 25-26℃ (cooling temperature), their rectal temperature dropped 0.1-0.5℃ and O 2 consumption almost doubled, accompanied by a relatively warmer skin temperature of the intrascapular BAT region than other sites 53 . BAT depletion or BAT functional defects are associated with the development of "cold syndrome", which threatens newborn infants' survival 54 . These phenomena gave rise to speculation that higher BAT content or activity would increase the survival of infants born in ambient cold temperatures. For example, the Last Glacial Maximum (LGM), when the mean temperature was 6.1℃ lower than nowadays 55 , challenged all species including modern humans. Some species adapted to the prehistoric climate successfully 56 , while others, e.g., archaic humans, Neanderthals, Denisovans, and Homo erectus, became extinct 57 . To date, factors that determine such successful adaptation are still poorly understood 58, 59 . We found that rs1421085 T > C knock-in increased the thermogenic capacity of mice. In this context, the functional FTO variant-rs1421085 T > C-has likely increased the survival odds of human newborn infants by increasing their thermogenesis. In consistent with this hypothesis, we found that the frequency of rs1421085 T > C was the lowest in East and Middle Africa, higher in East Europe, much higher in North Europe, and reached the highest in Siberia. Interestingly, the association between the FTO variants and obesity in European and Asian populations is not well-validated in Africans, and only mild effect was observed in a few studies 16,40,41 . Furthermore, a tight linkage disequilibrium among various obesityrelated GWAS positive signals that clustered within the rst intron of the FTO gene is observed in non-African populations, while this genetic architecture of the FTO loci breaks down in native Africans 16 . These ndings suggest that this genomic region has experienced genomic recombination and undergone genetic selection since modern humans descended from several branches of Homo sapiens populations that migrated out of Africa 40 . Several possibilities exist regarding the origin and distinct distribution frequency of the rs1421085 T > C variant between Africans and non-Africans. One hypothesis is that the variant arose occasionally in African populations, and after encountering extremely cold temperatures during the migration away from Africa, those newborns carrying the variant gained higher thermogenic capacity to thrive. The frequency subsequently increased stepwise, in parallel with migration, which started in the hot/low latitudes and gradually increased when moving to the cold/high latitudes. Another hypothesis claims that, in cold temperatures, an intense genomic recombination within the rst intron of the FTO gene induced random occurrence of rs1421085 T > C in a variety of populations, independently. Populations in Africa rarely encountered such cold exposure, thus the variant was merely observed. According to this hypothesis, the rs1421085 T > C variant rst presented in archaic humans after they settled in Eurasia, later the genomic fragment containing the variant possibly owed back to Africa. This process might have happened repeatedly 60 . It should be pointed out that all these possibilities warrant further genetic evidence of modern and archaic humans for con rmation or exclusion. Of note, a few nucleotide sequences centered to the rs1421085 locus were reported to be transcription factor (e.g., ARID5B and CUX1 12,15 ) -binding elements, showing a high conservation among species, even including zebra sh 12 . These results indicate the cis-elements might be essential for the expression of downstream targets, while our data support that the T > C alteration belongs to a gain-of-function mutation. In future studies, element disruption at the locus would be informative for us to understand the roles of wild-type genotypes in obesity and other pleiotropic outcomes. Despite that BAT brought evolutionary advantages for mammals to survive cold stress 61,62 , whether the thermogenic variation contributes to natural selection in the evolution of mammalian endothermy is also a matter of uncertainty 63 . It should be noted that, in additional to the rs1421085 T > C variant, other FTO variants also have functional impacts on the target genes 13,15 . The various haplotypes at convergence of different variants would lead to uctuation of the phenotypes, including energy expenditure and appetite 64 .
On the other hand, how to re-evaluate the predisposition of the "risk-allele" to adiposity in adult humans? According to the Early Growth Genetics (EGG) Consortium data, a shift pattern of the effect of the FTO rs9939609 T > A variant on BMI was observed in adolescents of different ages: with a positive association between the "risk allele" and BMI from 5.5 years old onwards; however, there was an inverse association below 2.5 years old 29 . In consistence, the genetic in uence of the same variant (rs9939609) on BMI was progressively stronger with increasing age (from 0.48 at 4y to 0.78 at 11y), as reported from a longitudinal large twin cohort of more than 7,000 children 65 , which further indicated an age-related characteristic of FTO variants in the regulation of adiposity. In humans, BAT diminishes rapidly in the rst decade of life 17,66 . To test whether this disappearance in uence the biological effect of Fto variant on adiposity phenotypes, we removed intrascapular BAT of both KI ES and WT mice, and successfully simulated the predisposition of the "risk allele" to obesity, as observed in humans. This interesting nding provided some potential underlying explanations for the phenotype shift from obesity-resistance to obesity-sensitivity of the rs1421085 T > C variant, such as the compensatory role of this variant in other tissues in addition to BAT (e.g., hypothalamus and WAT), the reduction of certain metabolic-protective secretory factors (e.g., BMP8b, FGF21 and IL-6) of BAT 67 , or the long-term consequence of altered feeding/eating habits at early life and following catch-up growth 68 . In a further speculation, other potential downstream targets of rs1421084 T > C, e.g., IRX3/IRX5/IRX6 12,13 and RPGRIP1L 15 , as well as the target of FTO per se 69 , could also participate in the compensatory predisposition to adiposity.

Conclusion
In summary, by collecting evidence from population studies as well as genetically modi ed murine models, we uncovered the possible biological roles and molecular basis of the rs1421085 T > C variant and postulated the rs1421085 T > C-Fto-Ucp1 axis in the regulation of BAT thermogenesis. These ndings may enhance our understanding regarding the possible "causative biological functions" of the "risk variants" derived from extensive GWAS data especially for those localized in non-coding regions and further remind us of the importance of genome-environment interaction, and the potential contribution of genetic variants in the process of adaptive evolution. Figure 1 The homologous rs1421085 T>C variant promotes energy expenditure and resists HFD-induced adiposity.     The rs1421085 T>C variant is associated with lower body weight of human infants and its allele frequency parallels with latitudes and ambient temperature of population distribution. (A) The Forest plot of the association between birth weight and human FTO rs1421085 polymorphism. Studies or subgroups included from previous studies and our in-home data were represented (see Methods). The size of the green box corresponding to each study was proportional to the sample size, while the horizontal line