Our results show that in this European population fairer skin was linked to higher concentrations of 25(OH)D, in agreement with the vitamin D hypothesis for the evolution of skin pigmentation. We used genetic scores that are robustly associated with pigmentation traits and sun exposure measurements in order to reduce bias due to self-report or clinical assessment. Whilst age, sex and maternal education were all associated with one or more reported or assessed pigmentation phenotypes and also associated with 25(OH)D, they were mostly unassociated with our genetic scores, supporting the use of genetic scores as unbiased and unconfounded proxies for pigmentary features. In cohorts with genotypic data available but no sun exposure measurements or low quality sun exposure information we could use genetic scores to assess the relationship between sun exposure and an outcome of interest.
In our study population higher scores corresponded to higher skin reflectance (i.e. lighter skin colour), a greater likelihood of having freckles, a higher mole count, a skin that always burns and never tans, having experienced at least one bad sunburn, always covering the skin with clothing or sun block when out in the sun, and using sun lotion with a high sun protection factor (SPF). The skin colour score, for instance, explained ~3.5% of the variability in skin reflectance among ALSPAC children.
The association of the genetic scores with sun exposure phenotypes confirms previous results obtained in British men who participated in the Prostate Testing for cancer and Treatment (ProtecT) study, where the same pigmentation scores were used .
The pigmentation scores, in particular the skin colour score, were associated with 25(OH)D levels, with higher scores predicting higher serum 25(OH)D, despite being associated with behaviours that would reduce the impact of UVR exposure on vitamin D synthesis. However, the magnitude of the effect was small. Although darker children (i.e. those with lower pigmentation scores) had lower 25(OH)D levels on average, only 0.5% of children were vitamin D deficient in this population (serum 25(OH)D < 25 nmol/l ). These results do not concur with our findings in the ProtecT study which showed that British men with lighter skins –as suggested by their genetic scores- had lower circulating 25(OH)D , or with a study of British women in which lighter skin pigmentation (assessed by nurses using the Fitzpatrick scale) was also associated with lower levels of 25(OH)D . These individuals possibly used more abundant sun protection or restricted their exposure to the sun. Children not wearing enough or effective sun protection and/or spending more time outside than adults may explain the reverse association.
There were clear differences by sex in reported sun exposure patterns as well as in pigmentation traits that suggested that girls were to some extent darker than boys. Yet, although apparently more protected from sunburn, girls reported more consistently applying sun block (or having sun block applied on them) when outdoors than boys. Time spent outside in the sun was similar for boys and girls. Girls have been found to have darker skin than boys in pre-adolescence, with boys becoming darker than girls around puberty [2, 27]. In our study there were more girls than boys with a low mole count at 4–5 years of age but this pattern appeared to have reversed by 15 years. In adults, across a range of countries and ethnicities, men tend to be more pigmented than women [2, 28], however, lighter-skinned men were identified in a study carried out in four European countries, in a sample of European Americans , and amongst populations in the Netherlands, Belgium and the Basque Country [20, 21].
On the other hand, mean pigmentation genetic scores did not vary between sexes, suggesting that the observed gender differences in pigmentation phenotypes may be explained by environmental, hormonal or other genetic causes. For instance, it is well-known that cutaneous pigmentation is modulated by growth factors, cytokines and endocrine hormones, which stimulate melanocyte proliferation and⁄or melanogenesis [30, 31].
The association of the pigmentation scores with skin reflectance and having freckles was for the most part stronger in girls (with a higher reflectance and likelihood of freckling as the pigmentation score increases towards the lighter phenotype), accounting for about twice the phenotypic variance explained by boys’ scores. A possible reason for this finding may be sexual selection by men for lighter-skinned women during human evolution [27, 32]. In fact, a few of the SNPs included in the pigmentation scores are located in genes that have been targets of selection (i.e. MC1R, HERC2, TYR, ASIP) [33–35]. However, because we concurrently found lower reflectance values in girls, other factors could be influencing skin colour in the opposite direction in the pre-pubertal stages. A recent study on mate preferences among UK twins revealed a patent sexually dimorphic preference for skin colour but, unexpectedly, women preferred fair-skinned men whilst men showed a preference for women with olive skin. Although a familial environmental effect was uncovered, no genetic effects on mate preference for skin colour were evident in this study . These sex differences deserve to be examined more carefully in the future as they may have an impact on the onset and progression of melanoma . Alternatively, given our small sample size and the lack of strong evidence for statistical heterogeneity between girls and boys, the observed sex disparities may simply be due to chance.
Differences in sun exposure-related variables were also detected with respect to maternal educational achievement, such that children of better educated mothers were more likely to burn rather than tan, wear protective clothing and use sun block, as well as were more often exposed to the sun in Britain or abroad. These children were also found to have slightly lower 25(OH)D levels. These findings possibly reflect reporting bias since educated mothers are usually more aware of the risks and benefits posed by the sun, and can also afford sun protection and summer holidays. Similar results with respect to educational level were reported in an observational study that assessed the relationship between sun exposure habits, sun protection behaviour and readiness to increase sun protection in an adult primary health care population in Sweden. Subjects of lower educational attainment reported being sunburnt less frequently, used less sun protection and revealed a lower propensity to increase sunscreen use .
Some of the pigmentation and sun exposure variables were measured only once in children between birth and age 17 (latest data release on ALSPAC children at the time of analysis) and therefore, it was not possible to assess their variation with age or to examine the contribution of pigmentation genes to these exposures at different ages. Other variables were measured at time points that were close together so were not very informative about age-related variation either. Time spent outside was only analysed based on having holidays near the sea or abroad and not with basis on the children’s regular outdoor experience. Given that in the 90s, when ALSPAC children were growing up, there was less awareness of UVR exposure risks and the industry of sun protection was less developed -hence the low sun block SPFs available- parents’ behaviour towards their offspring sunlight exposure may have been different from that of present-day parents of young children . Therefore, findings from this study may not be fully generalisable to the present day.
Because some SNPs were associated with more than one pigmentation trait, they were included in more than one score and therefore, scores were correlated. Scores explained some of the variation in the non-specific traits, but the strongest association was seen with the specific phenotypes. Since we were mostly interested in assessing the influence of a lighter skin phenotype overall on sun exposure measurements and 25(OH)D levels, and not to unequivocally establish the effects of each specific pigmentation trait, the correlation between scores does not bias our conclusions. However, we acknowledge that to understand the role of specific pigmentation traits (i.e. skin colour, tanning and freckling) on the efficiency of vitamin D synthesis other approaches would be needed [40, 41].
We had limited statistical power to examine the association of some of the sun exposure variables with the genetic scores and with 25(OH)D levels, especially with respect to the interaction analyses, consequently the evaluation of these associations in larger studies is essential.