Commentary on: “A genome-wide association study in Caucasian women suggests the involvement of HLA genes in the severity of facial solar lentigines” by Laville et al., 2016

We have read with interest the study by Laville et al. that was published in the September 2016 issue of PCMR (Laville et al., 2016) and that represents the second genome-wide association study of facial solar lentigines. This study (SU.VI.MAX cohort) showed two novel genetic regions, both located on chromosome 6 [6p22 intergenic region (p = 1.6 9 10 -6) and 6p21 USP81P/HLA-C region (p = 2.5 9 10 -7)] but both not significantly associated with solar lentigines based on the genome-wide association study. Laville et al. reported, however, that one SNP in the 6p22 region was significantly associated with forehead lentigines in a recessive model (p = 1.4 9 10 -8). We were interested in the question of whether we could replicate these findings in our cohort of 2844 Dutch North Europeans (mean age 67 years, 53% women). This is a subgroup of the Rotterdam Study (RS), for whom facial pigmented spots (mainly solar lentigines) were quantified as the percentage of affected skin area using digital facial photographs (Jacobs et al., 2015). We tried to select Laville et al. top SNPs in our data. From the 6p22 region, the two top SNPs of Laville were available in the RS as imputed SNPs. These two SNPs were both not associated with pigmented spots in our data (rs9350204: p = 0.62 and rs9358294: p = 0.60). From the 6p21 region, none of the top SNPs (rs2853949, rs2844614, rs2844613, rs2524069, rs2853947, rs2524067, and rs2524065) reported by Laville were available in our data. We tried to identify SNPs that were in linkage disequilibrium (LD) with the SNPs reported by Laville et al. using data from the 1000 Genomes Browser. Using the latest genome build, we noted that the SNPs map ambiguously to different regions of the genome (for example: http://www. ensembl.org/Homo_sapiens/Variation/Explore?db=core;v= rs2853949;vdb=variation). Using the previous genome build (GRCh37), we downloaded pair-wise LD (R2), encompassing the region 6:31233540-631253539 that included the SNPs the authors reported, and we looked into the RS for associations. Of these, 39 SNPs were available in the RS database (Table S1). None of the SNPs were significantly associated with pigmented spots (p > 0.30), although the R2 indicating LD with rs2853949 for the 39 SNPs is poor to moderate (<0.36), or unknown (ST1). It should be noted that the region reported by Laville et al. is located at the major histocompatibility complex (MHC), which is one of the most variable regions in the human genome (De Bakker and Raychaudhuri, 2012; Traherne et al., 2006). Within the HLA, the regions of LD are large due to a low recombination rate and adaptive selection (Gourraud et al., 2014). Therefore, associations in these regions should be taken with caution, as mapping SNPs to these regions is problematic.
The Rotterdam Study cohort and the SU.VI.MAX cohort are somewhat different. SU.VI.MAX consists solely of young women (mean age 53 years), and solar lentigines were manually graded using a photographic scale. The genetic analysis in the Rotterdam Study was adjusted for skin color and in SU.VI.MAX for cumulative sun exposure. However, we do not think that these minor differences account for the complete lack of replication. More likely, the size of the discovery cohort (n = 500) is too small to detect a true positive finding (Ioannidis et al., 2009). As they describe in their article, their power to detect a SNP with an explained variance of 5% is only 35%. This power is even lower, because 5% explained variance is unrealistically high (Mccarthy et al., 2008). The low power is also illustrated by the fact that Laville et al. replicated the association between solar lentigines and the rs12203592 in IRF4 (Jacobs et al., 2015) with a p-value of only 0.01, whereas the p-value in the RS was 10 -27.
The finding of HLA-C (an immunity related gene) is an appealing finding, because immunity could well play an important role in the origin of solar lentigines. Nevertheless, this finding should be first confirmed by additional studies. Therefore, we advocate for replication of genome-wide association study findings whenever possible.