The SNP rs67047829 was found in the present study to be associated with obesity in males and females together or in males alone, using a heterozygote model adjusted for year of birth (and interaction with year of birth not significant). Note that positive results from other models disappeared when adjustment for interaction with year of birth was included (Supplemental_file_OR1) and therefore no relevant conclusions could be drawn from other models.
For the heterozygote model the fact that year of birth did not disturb these results is indicated by the lack of interaction in the regressions and also from Table 1: median age from the AA genotype was slightly lower than for the other genotypes for both sexes together, but mean age was slightly higher for males alone.
The individuals' years of birth distributed between the genotypes in the heterozygote model in a way which gave a non-significant interaction adjustment i.e. that year of birth could be discounted as giving the effect observed. It should not be concluded from this that the heterozygote model is the best model to describe the effects of genotype on obesity (which is obvious from Fig. 2), only that for this model the distribution of years of birth was conducive to maintain the result in the data analysed. It is well known, and has be shown from this data,10 that obesity in the population has increased with time, and more data from a large cohort at the same age might be needed to extend this result, in order to avoid further interaction with year of birth, to other models (but see discussion of ethnic background below).
Note also it appears from the BMI distributions (Fig. 2) that results for males and females were rather similar, with BMI decreasing from GG to GA to AA, but significance disappeared for females: possibly indicating insufficient data for this sub-group.
The final p value for genotype:BMI (adjusted for year of birth but not interaction) for males and females together was 0.00145 and for males alone was 0.00349. Although these are low note they do not survive Bonferroni (0.05/137 = 0.000365) or false discovery rate correction (see Supplemental_Table_S3) and the results remain as "fishing trip" results which await confirmation from a larger study or from a different genetic background (see below).
The SNP rs67047829 (ERV3-1:NC_000007.14:g.64992360[G > A]; NM_001007253.4:c.667[C > T]; NP_001007254.2:p.Arg223Ter) A allele provides a premature termination codon in the SU domain of the "Endogenous retrovirus group 3 member 1, envelope" (UNIPROT short name: ERV3-1 envelope protein, ERV3-1env), probably, but not necessarily, resulting in degradation of mRNA before a protein can be produced. If ERV3-1env is normally responsible for stimulating adipose cell proliferation, then removal via this SNP might confer a protective effect.
ERV3-1env is composed of two major domains, SU and TM (which split from each other during processing and are held together by non-covalent bonds giving a heterodimer). In a virus, SU would mediate receptor recognition whereas TM would be a transmembrane domain. If ERV3-1 with the rs67047829 A allele did produce a protein then this would be without the TM domain and would have 222 rather than 604 amino acids.
Numerous retroviral introgressions into primate or human nuclear DNA have occurred throughout evolution, to create the class of human endogenous retroviruses (HERV). HERVs, along with long terminal repeat (LTR) elements, constitute nearly 8% of the human genome.17 To produce an intact virus, a HERV genome would need active gag, pol and env genes flanked by LTRs.18 However, although most HERV remnants (which are still referred to as human endogenous retroviruses) are now virally inactive, protein expression often still occurs and HERVs have been associated with several autoimmune diseases.19
The ERV3-1 envelope protein (ERV3-1env) has not been found to compose an element of a virus and it is well-established that it has lost its fusogenic properties.20 However it is recognised that, as its open reading frame has been conserved through 30 million years of primate evolution, and as full-length proteins (from four exons) are expressed in many tissues,21 it likely has a biological function beneficial to the host.22 In a placental trophoblast model with BeWo cells stably transfected with ERV3-1env, beta-human chorionic gonadotrophin (hCG) expression, which positively regulates the cell cycle,23 was increased; cyclin B expression, which promotes cell cycling, was reduced; while p21 expression, which negatively regulates the cell cycle, was up-regulated.24
Of probable direct relevance to the result found here is the fact that, although ERV3-1env mRNA is found in all tissues, it is highly expressed in adipose tissue. According to the databases found in the Human Protein Atlas (https://www.proteinatlas.org 21), adipose tissue had the highest ERV3-1env mRNA expression in one database (number of tissues: 45) and was found in the top four mRNA expression levels in the other databases (note that, conversely and rather strangely, protein expression was found to be low in adipose tissue). (As an aside it is interesting to note that Luteinising hormone plays a cardinal role in androgen production; in three databases the highest expression of ERV3-1env was found in adrenal glands, and a combination of LH/human chorionic gonadotropin (hCG) was found to directly stimulate T-I cell proliferation.)
Lastly, the SNP rs67047829 A allele shows high ethnic dependence with a prevalence of 4% in people of relatively-recent African descent, 9% with European descent and 19% with Asian descent,25 and analysis of people with Asian descent might possibly provide an easier route to the elucidation of relationships between this SNP and obesity phenotypes.
In conclusion, this fishing study of premature termination codons using a large sample of the Polish population (n = 5757) has allowed discovery of a possible association with a single nucleotide polymorphism, rs67047829 in ERV3-1, a gene known to be highly expressed as mRNA in adipose tissue, with obesity. This pretermination SNP is highly likely to change ERV3-1 envelope protein expression and activity, although it is not known whether remnant proteins would be active or not. The association found with body mass index, independent from date of birth, indicates a possible protective effect against obesity. Further study involving an even larger cohort or from another (e.g. Asian) ethnic group is needed to confirm and extend this result and in vitro studies could determine whether remnant protein survives in adipose cell cytoplasm or is destroyed by nonsense-mediated mRNA decay. Functional studies are also needed.