MC1R variants and associations with pigmentation characteristics and genetic ancestry in a Hispanic, predominately Puerto Rican, population

Skin cancer risk information based on melanocortin-1 receptor (MC1R) variants could inform prevention and screening recommendations for Hispanics, but limited evidence exists on the impact of MC1R variants in Hispanic populations. We studied Hispanic subjects, predominately of Puerto Rican heritage, from Tampa, Florida, US, and Ponce, PR. Blood or saliva samples were collected by prospective recruitment or retrieved from biobanks for genotyping of MC1R variants and ancestry informative markers. Participant demographic and self-reported phenotypic information was collected via biobank records or questionnaires. We determined associations of MC1R genetic risk categories and phenotypic variables and genetic ancestry. Over half of participants carried MC1R variants known to increase risk of skin cancer, and there was diversity in the observed variants across sample populations. Associations between MC1R genetic risk groups and some pigmentation characteristics were identified. Among Puerto Ricans, the proportion of participants carrying MC1R variants imparting elevated skin cancer risk was consistent across quartiles of European, African, and Native American genetic ancestry. These findings demonstrate that MC1R variants are important for pigmentation characteristics in Hispanics and that carriage of high risk MC1R alleles occurs even among Hispanics with stronger African or Native American genetic ancestry.

epidemiology studies of skin cancer to date have focused on non-Hispanic white populations, which potentially limits the generalizability of skin cancer risk information based on MC1R genotypes to at-risk populations with more diverse genetic ancestry, such as Hispanics.
The incidence of melanoma-the deadliest form of skin cancer-is rising among Hispanics 10,11 . Although Hispanics have a lower lifetime risk of melanoma than non-Hispanics, Hispanics are more likely to be diagnosed at a younger age, have higher disease morbidity, and experience late stage clinical presentation of melanoma leading to higher mortality rates [11][12][13][14] . Alongside increasing melanoma incidence, over 6000 new cases of non-melanoma skin cancers were diagnosed among Hispanics living in Puerto Rico in 2005, which represents about a 300% increase since 1974 15 . In addition to well-established factors such as unequal access to healthcare that influence disparities in health outcomes for Hispanics, poorer outcomes may also be impacted by public health efforts focusing on sun-sensitive phenotypes and a lack of patient and clinician awareness about skin cancer risk in Hispanics 16,17 . Prevention and screening advice that incorporates MC1R genotypes may improve skin cancer risk awareness and risk reduction among Hispanics 18 , but evidence on MC1R variants and their associations with pigmentation characteristics in Hispanic populations is limited.
As a prelude to conducting an intervention study among Hispanics to determine whether feedback of MC1R genotype (i.e. precision prevention) can affect change in skin cancer prevention behaviors, we first addressed some gaps in research evidence by conducting a pilot study to examine the prevalence of MC1R variants among Hispanics living in the Tampa Bay region of Florida, US and in Puerto Rico. These geographies were selected because of an ongoing federally-funded partnership initiative between Ponce Health Sciences University (PHSU) in Ponce, PR, and Moffitt Cancer Center (MCC) in Tampa, Florida, US, the overall goal of which is to improve cancer care outcomes for Hispanics in Puerto Rico and Florida. We further assessed associations between MC1R variation and traditional skin cancer risk factors and genetic ancestry in this population.

Materials and methods
Subjects and data collection. Three different sources were used to obtain samples and/or recruit study participants: the Puerto Rico Biobank (PRBB), located in Ponce, PR; a Community Participant Registry (CPR) covering Puerto Rico and Florida, US; and the Morsani Family Medicine clinics (MFMC) at the University of South Florida, in Tampa, Florida, US. All participant information, informed consent forms, and questionnaires were available in both Spanish and English. We also obtained genotype data from the 1000 Genomes project.
Puerto Rico Biobank (PRBB). The PRBB is a cancer tissue biobank housed at PHSU that was established as part of a PHSU-MCC partnership initiative. In order to contribute to the PRBB, participants were required to confirm their Puerto Rican heritage as indicated by having at least three Puerto Rican grandparents. Processes of informed consent, collection, processing, and storage of samples are published in detail elsewhere 19 . Briefly, we obtained de-identified stored peripheral blood samples from 122 healthy controls and 78 randomly selected cancer patients diagnosed with cancers other than melanoma (Supplemental Table 1) for isolation of DNA. Because MC1R is not known to be associated with cancers other than skin cancers, these cancer patients are considered representative of the general population. Information on diagnosis of non-melanoma skin cancers was unavailable from the PRBB. Ethics approval was obtained from the PHSU Institutional Review Board (IRB) Committee, and all research was performed in accordance with relevant regulations.
Community Participant Registry (CPR). The CPR was a resource also developed as part of the PHSU-MCC partnership and comprised Hispanic residents in Tampa Bay and Puerto Rico who had provided consent to be contacted about cancer prevention and control research studies. Invitation packets that included an information statement, consent form, and questionnaire on demographics and pigmentation characteristics (Supplemental Table S2) were mailed to 176 eligible individuals; 39 (22%) returned the signed informed consent and questionnaire. These individuals were then asked to provide a saliva sample via a mailed Oragene ® DNA (DNA Genotek) kit. Mailed kits included detailed instructions to promote maximize yield and minimize contamination of collected saliva. Ethics approval was obtained from Chesapeake IRB with subsequent continuing renewal approval by Advarra, and all research was performed in accordance with relevant regulations.

Morsani Family Medicine clinics (MFMC), University of South Florida.
To augment the number of Hispanics in our pilot study who lived in the Tampa Bay area, eligible patients (n = 105) at the MFMC were identified via clinical records and invited to participate at scheduled appointments. At the time of providing written informed consent, participants completed a questionnaire eliciting information on demographic variables and pigmentation characteristics (Supplemental Table S2) and provided a saliva sample using an Oragene ® DNA collection kit. Ethics approval was obtained from the Institutional Review Board of the University of South Florida, and all research was performed in accordance with relevant regulations.
1000 Genomes Project. Genotype data at the MC1R locus were extracted from VCF files for the 104 Puerto Rican (PUR) participants available from the 1000 Genomes Project data portal 20 .
Sample processing and genotyping. For DNA samples retrieved from the PRBB and obtained from participants recruited through the CPR and from the MFMC, we performed direct Sanger sequencing of the one exon coding region of MC1R to identify all existing variants 21 . We also genotyped 106 ancestry informative markers (AIM) that discriminate between Native American, African, and European ancestry. To maximize genetic information, SNPs with a large difference in allele frequency among ancestral populations were chosen. As well, representation across all 22 autosomal chromosomes was considered when selecting SNP markers. This AIM panel has been described previously 22 . AIM genotyping used a multiplex PCR coupled with single base extension  Puerto Rican and other 0 0 6 (7.  Continued methodology with alleles called using a Sequenom analyzer. Genotyping quality control for AIM was assessed using standard sample-level and SNP-level metrics. MC1R variants and genotype calls for the same 106 AIM were abstracted from 1000 Genomes genotyping data using VCFTools 23 .

MC1R variant categorization.
We categorized participants into three groups based on the number and type of MC1R variant(s) carried using an algorithm similar to that described in Hernando et al. 24 Participants in the low risk group did not carry any MC1R variant (consensus) or carried only variants that do not impact on receptor function, i.e. pseudoalleles, based on published functional analyses or as predicted from bioinformatical algorithms. Participants in the medium risk group carried only a single MC1R variant that results in partial loss of receptor function, i.e. "r" allele, based on published functional analyses or as predicted from bioinformatical algorithms. Participants in the high risk group carried either two "r" variants or carried at least one variant known to result in loss of receptor function based on published functional analyses or as predicted from bioinformatical algorithms.
Ancestry estimations. For each individual, global ancestry proportions were measured using the software Admixture v1.3 at a k = 3 and under a supervised model 25  Statistical analysis. MC1R minor allele frequencies were calculated. We compared MC1R risk categories by phenotypic skin cancer risk characteristics after dichotomizing phenotypic measures, and we used logistic regression models to determine odd ratios (OR) and corresponding 95% confidence interval (CI) with adjustment for gender. We compared MC1R risk categories by quartiles of European, African and Native American genetic ancestry and tested for differences in proportions using the Jonckheere-Terpstra test or the Cochran-Armitage trend tests. Quartiles of genetic ancestry were based on the overall participant sample (n = 315) that had successful ancestry genotyping. All analyses were conducted using SAS.

Results
From the PRBB, 193 (97%) samples were successfully genotyped for MC1R (72 cases, 121 controls) and demographic information was obtained for 167 (87%) participants (49 cases, 118 controls). From the CPR, 32 (82%) of the responders completed the questionnaire on demographics and self-reported pigmentation characteristics and provided a saliva sample, and 30 (77%) samples were successfully genotyped for MC1R. Eighty-eight (84%) participants from the MFMC completed the questionnaire on demographics and self-reported pigmentation characteristics and provided a saliva sample; 79 (90%) samples were successfully genotyped for MC1R. Flow diagrams summarizing participant enrollment, biosample collection and genotyping, and availability for analyses are given in Supplemental Figure S1. Characteristics of these individuals with successful MC1R genotyping are summarized in Table 1.   (20) 15% (12) 17% (12) European (mean percent, standard deviation) 68% (12) -59% (21) 72% (14) 68% (14) Native American (mean percent, standard deviation) 14% (7) -19% (16) 13% (7) 14% (8)  www.nature.com/scientificreports www.nature.com/scientificreports/ 89 (22%) were in the high risk group (Table 1). By definition, all individuals in the medium risk group were heterozygous carriers of an r allele. Among the 89 participants in the high risk group, two (2.3%) were heterozygous compound carriers of two R alleles, 20 (22.5%) carried one R and one r allele, 22 (24.7%) carried two r alleles (six in a homozygous state), and 45 (50.6%) carried one R allele. Table 2 summarizes the minor allele frequencies in individuals of sole Puerto Rican heritage, who comprise the majority (87%, n = 355) of participants in this study. Except for the D84E variant, each of the other nine most well-described risk variants (V60L, V92M, R142H, R151C, I155T, R160W, R163Q, D294H) was observed. Twenty-six observances of 12 rare non-synonymous or insertion/deletion variants were detected in our study participants, but none were novel. MC1R genotype and pigmentation characteristics. Among CPR and MFMC participants who completed a questionnaire capturing phenotypic information, most reported having darker phenotypic characteristics (Table 1). Associations among MC1R risk categories and pigmentation characteristics are given in Table 3. We noted a significant trend (p = 0.0004) across MC1R risk categories with tendency to burn: Hispanic individuals in the medium (OR: 3.4, 95% CI: 1.2-9.2; p = 0.017) and high (OR: 8.4, 95% CI: 2.5-28; p = 0.0005) MC1R risk categories were more likely to report skin that burned (including sunburn without blistering and severe burning with blistering) compared to those in the low MC1R risk category. A similar, but weaker, trend (p = 0.025) was noted with tendency to tan: Hispanic individuals in the medium (OR: 2.0, 95% CI: 0.61-6.7; p = 0.25) and high (OR: 4.4, 95% CI: 1.2-16; p = 0.025) MC1R risk categories were more likely to report skin that only mildly tanned at best compared to those in the low MC1R risk category. Only the trend for burning remained significant after Bonferroni correction for multiple comparisons. Although individuals in the medium (OR: 2.0, 95% CI: 0.75-5.5) and high (OR: 2.4, 95% CI: 0.76-7.8) MC1R risk categories were more likely to freckle (including very few, few, some, many, or very many) compared to individuals in the low risk category, associations were not statistically significant. We did not find an association between MC1R risk categories and eye or hair color, in part because

Discussion
This pilot study on the prevalence of MC1R variants in a Hispanic population in Tampa and Puerto Rico found that 56% of participants carried a MC1R allele(s) that placed them at elevated risk for skin cancer, the vast majority of which have been shown to increase the odds of melanoma, SCC, or BCC by at least 80% 6,8 . Our analyses demonstrate that MC1R variants are associated with some pigmentation characteristics in this overwhelmingly Puerto Rican sample, consistent with observations seen in European populations 26 . Among individuals reporting only Puerto Rican heritage, the proportion of participants categorized at elevated MC1R risk was comparable across quartiles of European ancestry, across quartiles of African ancestry, and across quartiles Native American genetic ancestry, indicating that those with greater African or Native American genetic ancestry carried MC1R risk alleles. Although our findings are limited by the overall and sub-group sample sizes and self-reported pigmentation characteristics, these results contribute novel evidence on the MC1R gene in an underserved population, which will inform future research studies. Currently, our understanding of MC1R variants and their association with skin cancer risk is based on studies predominately in non-Hispanic whites. However, some international studies in general population and melanoma family settings have demonstrated a range in the prevalence and variation of MC1R variants according to geographic location 27,28 . Studies conducted in Spanish populations have found approximately 50-70% of individuals carry at least one risk variant 24,29,30 , which is comparable to the combined prevalence of medium and high Many or very many 0 (0) 2 (4.9) 2 (9.5) www.nature.com/scientificreports www.nature.com/scientificreports/ risk alleles in our study. One study compared the association between melanoma and MC1R variants in German and Spanish populations, and the authors found significant differences in the frequency of, and risk attributable to, MC1R variants in the two populations 31 .
In our study, the overall proportions of genetic ancestry were consistent with those from published data, which demonstrate that Hispanic and Latino populations have high admixture of predominately European, African, and Native American ancestry 32 . We observed minimal variation in MC1R risk according to genetic ancestry among Hispanics of Puerto Rican heritage. Our findings suggest that MC1R variants are relevant to a diverse population and that even among populations with less European and stronger African or Native American genetic ancestry there may be carriers of alleles conferring elevated risk. Furthermore, these data reinforce the need for future research on the varied functional impact of MC1R variants according to population subgroups that are genetically and geographically diverse 31 .
Several studies among non-Hispanics have shown that genetic variation at MC1R is associated with melanoma risk independent of traditional phenotypic characteristics (e.g. hair and skin color) and that MC1R may even confer higher risk among individuals with a darker phenotype than among those with a lighter phenotype 21,27,28 . However, to the best of our knowledge, studies of associations between MC1R and skin cancer risk in Hispanic populations are lacking. Some studies have reported on the prevalence of some MC1R risk variants in Hispanic and Latinx sub-groups such as Mexican-American, Uruguayan and Brazilian, but most have either focused on populations at elevated risk of melanoma due to personal/family disease history 33,34 or they were conducted in the context of other diseases, such as depression 35,36 . One recent study set in New Mexico reported carriage of a medium or high risk MC1R variant in 63% of enrolled Hispanics 37 . Some studies have examined associations between MC1R variation and genetic ancestry 38,39 , but there appears to be limited research that also incorporates pigmentation characteristics in Hispanic populations. Our findings contribute novel evidence on the association between MC1R variation and pigmentation characteristics in Hispanics who were not selected to participate on the basis of their personal or family history of melanoma (or non-melanoma skin cancers).   Quartile cutpoints are based on the distribution of genetic ancestry observed the overall participant sample with successful ancestry genotyping (n = 315). b P-value is from the Jonckheere-Terpstra test comparing the proportion of participants in MC1R risk categories across quartiles of genetic ancestry. c P-value is from the Cochran-Armitage test for trend comparing the proportion of participants in a combined low and medium MC1R risk category to that in the high MC1R risk category across quartiles of genetic ancestry. d P-value is from the Cochran-Armitage test for trend comparing the proportion of participants in the low MC1R risk category to that in a combined medium and high MC1R risk category across quartiles of genetic ancestry.
The belief that darker skin pigmentation is infallibly protective against skin cancer can serve as a barrier to Hispanics undertaking, and receiving education in, prevention and early detection behaviors 16 . As expected, the majority of participants in our study who completed the questionnaire reported darker phenotypic characteristics. Among those who reported moderate to strong tanning ability, 53% carried either medium or high MC1R risk variants, and this group particularly stands to benefit from receiving information on their MC1R genotype as it may change their perceived skin cancer risk (from lower to higher). High uptake and interest in MC1R testing among Hispanics has previously been demonstrated 18 , and the impact of receiving such genetics-based skin cancer risk information on behavioral, psycho-social and ethical outcomes is being investigated in ongoing trials in Hispanic and broader population contexts 40,41 . Additionally, our findings indicate a need for further research on the pathways and attributable risk of MC1R variants in individuals with sun resistant phenotypes.
This pilot study demonstrated that risk information based on MC1R genetic variants may be relevant to a diverse, Hispanic population and could inform skin cancer risk assessment, prevention and early detection recommendations in this setting. Our findings further highlight the need to ensure that prevention and early detection recommendations are inclusive of populations with low risk phenotypic characteristics alongside general population strategies, and the need for research on the pathways and risk of MC1R variants in groups with diverse genetic ancestry and phenotypic characteristics.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.