GWAS OF SKIN AGING IN KOREAN POPULATION 73 levels (Supplementary Figure 3) moreover, the difference appeared as a result of confi rm- ing the difference by genotype in grading scale of the pigmentation measurement value (p 0.001, Figure 2B). Although there were no strong signals for correlations with skin phenotypic changes, suggestive SNPs were identifi ed for moisture content, oil content, and skin sensitivity. Regarding moisture content, correlations were identifi ed for six SNPs, among which rs9873353 showed a strong correlation (β = –0.567, p = 1.47 × 10-6). However, there was no gene located near the SNP. We found fi ve SNPs for oil con- tent and two SNPs for skin sensitivity. The SNPs rs308971 (β = –0.325, p = 4.60 × 10-6) and rs7334780 (odds ratio = 0.635, P = 2.82 × 10-6) showed the highest association with the oil content and sensitivity of skin, respectively. When looking at the genes associated with these two SNPs, rs7334780 does not have a gene located nearby however, rs308971 is located in the SYN2 gene, which is known to be involved in a neurotransmitter pathway. In particular, its function is associated with the lipid membrane in which neurotransmitters are embedded. SKIN TISSUE EQTLS FOR SNPS IN THE GTEX PORTAL When the GWAS-identifi ed SNPs related to skin phenotype s were searched in the eQTL database (GTEx Portal, https://gtexportal.org/), differences in expression levels in skin tissues according to genotype were identifi ed in fi ve SNPs: rs7042102C T, rs34466224 G A, rs4653497T C, rs308971 G A, and rs9577919C T (Supplementary Figure 4). In sun-exposed skin tissue, rs7042102C T (p = 5.2× 0-12), rs34466224 G A (p = 1.1 × 10-7), and rs308971 G A (p = 1.0 × 10-13) showed signifi cant differences in expression according to genotype, and rs4653497T C (p = 1.7 × 10-4) and rs9577919C T (p = 4.9 × 10-5) showed altered expression levels associated with genotype in non–sun- exposed skin tissue. The eQTL expression levels for each SNP change gradually with the mi- nor allele homozygous genotype rs7042102C T, rs308971 G A, and rs9577919C T gradually increased in expression, and rs4653497T C gradually decreased in expression. The difference in rs34466224 G A expression level according to genotype was statisti- cally signifi cant, although it showed some deviation. DISCUSSION This study has numerous strengths compared with previous studies. First, most of the re- ported GWAS of skin phenotypes have been performed on pigmentation (25–27), sun sen- sitivity (28,29), and infl ammation (30), whereas this study analyzed representative skin traits, such as wrinkles, moisture content, pigmentation, oil content, and sensitivity. Second, the analysis of the association between skin phenotypes and genes in Koreans is considered important because no such analysis has ever been performed in this population. Third, the sample size of the Korean participants was based on a large research group that included more than 1,000 people, and the statistical verifi cation power was excellent. Because we generated an analytical dataset based on this group, our results are highly reli- able. Finally, our results are meaningful in that the measured skin phenotype values were not analyzed based on individual measurements, but instead were averaged according to the measurement equipment and the measurement site, and scored by tertile. This approach applies universally to the data cleansing method generated through the analysis of the results

JOURNAL OF COSMETIC SCIENCE 74 because universality can be achieved by applying tertile-divided scoring to the measured values, even when using different equipment or measuring different areas. There were also some limitations to our research. First, reproducibility was not verifi ed in our study because no replication study exists. However, our fi ndings provide scientifi c evi- dence, given the large sample size, that are considered to be reliable enough to include the OCA2 SNP (rs74653330) (31,32), which is known to be associated with pigmentation. As the results of these phenotypic studies show, there is a link between genetic variation and changes in skin characteristics which allows us to produce skin prediction models. If we can identify gene variants that affect skin phenotype changes, we can provide a per- sonalized guide to apply to lifestyle, diet, and makeup and help prevent and resolve skin problems using customized cosmetics. By providing relief and therapeutic intervention in the early stages of symptoms, cosmetic problems and medical costs can be reduced. These results can help us understand the personality of each individual’s skin and provide personalized cosmetics and skincare. In addition, we can also provide customized services based on genotype. We suggest that this study is a valuable foundation to produce cus- tomized cosmetics, wherein ingredients and materials may be added to the cosmetics for the prevention of skin damage or application of functional cosmetics, according to skin characteristic markers through genetic testing. WRINKLES Among the various skin traits examined, wrinkles were the primary skin phenotype show- ing the most obvious change. Generally, wrinkles appear naturally because of aging, which is an interesting fi eld in cosmetic dermatology because it is a phenotype that does not re- cover completely, unlike scars or disease (33). However, wrinkles are not only caused by natural aging but can also be caused or accelerated by external factors and internal abnor- malities (34). In particular, the internal immune response, triggered by environmental fac- tors, can produce chronic infl ammation and accelerate the occurrence of wrinkles (34). Aging results in the accumulation of the B-cell receptor, which increases the risk of chronic infl ammation stimulated through this receptor and its induced transcript, FCRL5 (35). FCRL3, a well-known member of the immunoglobulin receptor superfamily and one of several Fc receptor–like glycoproteins, has immunoreceptor tyrosine-based activation and inhibitory motifs in its cytoplasmic domain, which may play a role in the regulation of the immune system. According to our results, rs117381658 exists downstream of FCRL5 and can affect FCRL5 expression by forming a signifi cant SNP cluster related to chronic infl am- matory status (36). FCRL5 expression levels can affect the infl ammatory response and NF-κB, which can disrupt modulators of tissue homeostasis and affect skin aging (37). In addition, our study also identifi ed SNPs associated with wrinkles, such as rs1961184 (REEP3), rs1929013 (ADSS2), and rs7042102 (SPTLC1). The SNP rs1961184 is located downstream of REEP3 and has been reported to infl uence REEP3 expression patterns in the heart tissue. REEP3 is involved in GPCR signaling, affects the normal progression of mito- sis, and is reported to be related to aging (38). Although there are no reports of REEP3 di- rectly impacting skin aging, it is a potential marker of skin aging. The SNP rs1929013 is located downstream of ADSS2, which encodes an enzyme that converts inosine monophos- phate to adenosine monophosphate. Adenosine is a well-known factor involved in wrinkle improvement (39), and ADSS2 can regulate the function of adenosine (40). It is conceivable