During mammalian evolution genes involved in diet and immune response have been preferential targets of positive selection , highlighting the role of nutrient availability/preferences and pathogens as powerful selective forces. The protein product of NPC1 plays a central role in lipid metabolism, as it acts as a cholesterol transporter and its transcription is regulated by the SREBP pathway . Conversely, the gene does not participate in immune response, but is exploited by members of the filovirus family as an intracellular receptor that mediates the late steps of viral invasion [3–5]. Evidence has indicated that genes directly involved in antiviral response or acting as viral receptors (for example, HAVCR1, CD4) display domains evolving under positive selection as the result of a genetic conflict with extant or extinct viral species [38–46]. Positive selection at these host genes may result from adaptation either to increase viral recognition and restriction efficiency or to avoid binding of specific viral components. Our evolutionary analysis in mammals indicated a predominant role of purifying selection in driving the evolution of NPC1 but also identified few positions that have been targeted by positive selection. Specifically, maximum-likelihood ratio tests indicated that three residues in the N-terminal portion of luminal loop 2 evolved under positive selection; these codons are located in close proximity to each other, and selection was confirmed by an independent REL analysis. PAML also identified one positively selected site in luminal loop 1, but this was not supported by REL, suggesting that it may represent a false positive, as the M8 model has been shown to be more prone than REL to false positive results when a relatively high number of sequences (species) is used for analysis . These results suggest that the selective pressure responsible for positive selection in NPC1 stems from pathogens rather than from dietary changes. Indeed, a recent study has indicated that luminal loop 2 is necessary and sufficient to bind filovirus GP1 protein directly and to mediate productive infection ; the authors were able to map the GP1 residues involved in engaging loop 2 and determined that they are conserved among filoviruses . This observation, together with evidence showing that NPC1 is required for infection of both human and rodent cells by distantly related viral species, strongly suggests that the cholesterol transporter is a necessary factor for most members of the Filoviridae family [3–5]. These pathogens display a wide host range in mammals  and are thought to have affected vertebrates for millions of years, as testified by the detection of filovirus-derived elements in the genome of both eutherians and marsupials . Thus, we suggest that the positively selected sites we identified in luminal loop 2 evolved in response to a host-filovirus arms race and might represent relevant residues in mediating GP1 binding.
Population genetic analysis of NPC1 in humans revealed no evident signature of natural selection in loop 2 or any other gene region, although we cannot exclude that weak or geographically-restricted selective events have acted on the gene. With respect to filovirus infection, this might not be surprising as the known human pathogens Ebola and Marburg viruses are highly virulent agents that rapidly kill infected individuals, a feature that possibly limits their spreading in human populations  and makes them unlikely candidates to play a role as selective agents. Genetic diversity in human NPC1 has nevertheless been recently associated with metabolic dysfunction, this association being based on the central role of the gene in lipid trafficking. Specifically, the His215Arg (rs1805081) variant in luminal loop 1, which is involved in cholesterol binding, was shown to associate with obesity in populations of European descent [9, 10]. It has been proposed that alleles responsible for obesity and T2D might have evolved as 'thrifty' variants in ancient populations [51, 52]. In line with this hypothesis, selection signatures have been detected for a few polymorphisms associated with these conditions [53, 54], although this does not seem to be the case for NPC1. Nonetheless, inspection of nonsynonymous SNPs located in the gene revealed that, in addition to the above mentioned variant in loop 1, a polymorphism (Ile642Met, rs1788799) in the SSD domain segregates at relatively high frequency in human populations and affects an isoleucine residue which is conserved in all the mammals we analyzed.
We thus reasoned that this SNP might affect NPC1 function and modulate metabolic phenotypes. We tested this hypothesis in a large cohort of subjects from Saudi Arabia, a region where the prevalence of obesity and T2D is very high [55–57]. The previously described association between rs1805081 and obesity [9, 10] was not replicated in the Saudi sample, although the relatively lower minor allele frequency (MAF) of the variant in this population (12%) compared to Europeans (ranging from 25% to 40%) might have limited our detection power. No effect on BMI or obesity was detected in the Saudi cohort for the Ile642Met variant either. Similarly, the role of the His215Arg variant in predisposing to obesity was not observed in a cohort of Chinese children , although a possible interaction between this (and other) variant and sedentary behavior has been described in a population of the same ethnicity . Recently, a meta-analysis of rs1805081 on obesity risk in Europeans also revealed a weak effect of the polymorphism on body fat percentage, but not on BMI or on the odds of being obese . One possibility to explain these contrasting results is that variants in NPC1 interact with environmental cues, as suggested by the Chinese study  and possibly with additional genetic factors. This seems to be the case for Npc1
mice: these animals develop increased adiposity and metabolic disturbances but the phenotype depends on both fat intake and genetic background [7, 59]. These animals also present with increased fasting plasma glucose levels, glucose intolerance, and insulin resistance, indicating a T2D phenotype [7, 59]. Somehow in contrast with these results, a recent study indicated that heterozygosity for a hypomorphic Npc1 mutation on the C57BL/6J 'metabolic syndrome' genetic background protects old male mice, but not females, from weight gain . Overall, these observations suggest that Npc1 genetic variation interacts with diet, sex and with one or more gene(s) in modulating metabolic phenotypes.
A possible association between the two NPC1 variants and T2D was analyzed in the Saudi cohort. Overweight and obesity are strong risk factors for the development of T2D; genetic susceptibility is nevertheless believed to play a stronger role in non-obesity related T2D . Thus, we verified the effect of rs1805081 and rs1788799 on diabetes susceptibility by taking BMI into account; a significant association was detected between rs1788799 and T2D, with a predisposing role for the derived 642Met allele.
Several metabolic traits are sexually dimorphic in humans and/or show sex-specific heritability linked to the autosomes . Thus, it was suggested that variants with a sex-specific effect might be difficult to detect without separating the sexes or modeling for gender-based differences . Testing for interaction with sex in our cohort indicated the presence of a significant effect; stratification of the population on the basis of gender revealed that the association is driven by male subjects. This was even more evident when haplotype analysis using the two coding variants was performed. Notably, two major haplotypes showed an opposite effect on T2D susceptibility in men only, and the effect was evident in both obese and non-obese individuals. An interaction between gender and genetic factors has been described for some other genes involved in T2D [63–66]; the reasons underlying these sex-specific events remain to be elucidated and might include a role for sex hormones, epistatic effects with X-linked variants, or differences in dietary habits and lifestyle between the sexes that, in turn, interact with the genetic status.
Further analyses on plasma lipid levels showed the presence of different associations with NPC1 haplotypes in men and women. Nonetheless, these effects were generally weak and should be interpreted with caution. The stronger effect was detected for triglyceride levels. Thus, in men a minor haplotype unrelated to T2D susceptibility was found to associate with higher levels, whereas in women the two major haplotypes that predispose or protect men from diabetes were found to be associated with higher and lower triglyceride levels, respectively.