Characterization of Potential Melanoma Predisposition Genes in High-Risk Brazilian Patients

Increased genetic risk for melanoma can occur in the context of germline pathogenic variants in high-penetrance genes, such as CDKN2A and CDK4, risk variants in low- to moderate-penetrance genes (MC1R and MITF), and possibly due to variants in emerging genes, such as ACD, TERF2IP, and TERT. We aimed to identify germline variants in high- and low- to moderate-penetrance melanoma risk genes in Brazilian patients with clinical criteria for familial melanoma syndrome. We selected patients with three or more melanomas or melanoma patients from families with three tumors (melanoma and pancreatic cancer) in first- or second-degree relatives. Genetic testing was performed with a nine-gene panel (ACD, BAP1, CDK4, CDKN2A, POT1, TERT, TERF2IP, MC1R, and MITF). In 36 patients, we identified 2 (5.6%) with germline pathogenic variants in CDKN2A and BAP1 and 4 (11.1%) with variants of uncertain significance in the high-penetrance genes. MC1R variants were found in 86.5%, and both red hair color variants and unknown risk variants were enriched in patients compared to a control group. The low frequency of germline pathogenic variants in the high-penetrance genes and the high prevalence of MC1R variants found in our cohort show the importance of the MC1R genotype in determining the risk of melanoma in the Brazilian melanoma-prone families.


Introduction
Melanoma is a multi-factorial skin cancer and emerges from an interaction between environmental exposure and genetic susceptibility.Ultraviolet radiation (UVR) is the major known etiologic agent associated with melanoma [1].The risk determined by UVR depends on geographic latitude, cumulative and exposure pattern, and history of sunburn, which is related to specific phenotypic features such as fair skin, red/blond hair, light-colored eyes, propensity to burn, multiple nevi, and presence of atypical nevi [1,2].The genetic basis of melanoma development stems from random acquired mutations and the accumulation of genomic changes within melanocytes.At the same time, melanoma susceptibility is also determined by the presence of heritable germline variants [3].Patients with a personal and or a family history of melanoma are the ones at highest risk for developing the disease [2].Approximately 5-10% of melanoma cases occur in a familial context, which implies that an inherited germline pathogenic variant (GPV) in cancer-predisposing genes, a shared environmental exposure, or both concur with melanoma pathogenesis [4,5].Outside the familial context, multiple primary melanomas (MPM) might also be related to genetic factors such as germline de novo pathogenic variants in high-penetrance genes and multiple variants in moderate-penetrance genes [6].
Cyclin-dependent kinase inhibitor 2A (CDKN2A) was the first high-penetrance melanoma gene identified and encodes two distinct proteins (p16 and p14ARF), both acting in cell cycle regulation [7].In general, CDKN2A GPV is found in 10 to 40% of melanoma-prone families [8,9] and is associated with an increased number of family members affected early age at melanoma diagnosis, MPM, and pancreatic cancer [10].
The second melanoma predisposition gene described was CDK4, an oncogene that also plays an important role in cell cycle regulation.The phenotype of CDK4-mutated families is indistinguishable from CDKN2A-mutated ones [11].Only 20 families have been identified with GPVs in the CDK4 gene to date [12,13].
Melanocortin 1 receptor (MC1R) and microphthalmia-associated transcription factor (MITF) are considered moderate-penetrance melanoma genes.MC1R is crucial in the control of human pigmentation and is a highly polymorphic gene.MC1R variants are associated with an increased risk of developing melanoma, especially the R variants, which are highly associated with red hair color phenotype [14].Melanoma risk determined by MC1R R variants varies from two to three times per R allele and is higher in the familial context when compared with the general population [14].MITF regulates melanocyte development and differentiation.The p.E318K variant has been associated with high nevi count, fair skin, and non-blue eye color and predisposes to both melanoma and renal cell carcinoma [14,15].
More recently, other high-penetrance melanoma genes have been identified, such as breast cancer 1 (BRCA1)-associated protein 1 (BAP1), with definitive evidence for association with melanoma and other tumors and other potential melanoma-predisposing genes related to telomere maintenance-protection of telomeres 1 (POT1), ACD, TERF2IP, and telomere transcriptase reverse (TERT) [16][17][18], which still lack validation from other studies.To date, these aforementioned genes have not been studied in Brazilian patients.Along with CDK4, they correspond to less than 3% of melanoma-prone families in the studied populations around the world; nevertheless, the majority of cases remain without known genetic etiology [4,14,18].
Familial clustering of melanoma and MPM cases are referred to as familial or hereditary melanoma.In countries with low to medium incidence of melanoma, clinical criteria for familial or hereditary melanoma include patients with two primary melanomas and families with two melanoma cases and/or one pancreatic carcinoma and one melanoma, in first-or second-degree relatives ("rule of two") [19].In areas with moderate to high melanoma incidence, the recommendation is to consider the "rule of three"-patients with three primary melanomas and families with three melanoma cases [19].
In Brazil, the incidence of melanoma varies widely according to the different geographic regions, mainly due to the population's ancestry, latitude, and UVR emission during the summer months [20,21].The World Health Organization (WHO) shows, for 2020, an estimated melanoma incidence rate in Brazil of 2.7 (female) and 3.1 (male) per 100,000 [22].The Brazilian National Cancer Institute estimates for 2020 a melanoma incidence rate that varies from 1.36 to 6.78 per 100,000 for the southeast region and 3.53 to 10.05 per 100,000 for the southern region, the two Brazilian regions with the highest melanoma incidence, achieving incidence rates similar to European countries such as Spain [22,23].
Data regarding the frequency of germline mutations in the CDKN2A gene for Brazilian individuals fulfilling clinical criteria for familial melanoma are scarce and differ according to the region studied and the adopted criteria, varying from 0% to 14% [21,[24][25][26].In a study from southeast Brazil, a mutational rate of 4.5% in CDKN2A was observed in a cohort of 22 probands [24].Another study from south Brazil found 0% (0/33) of CDKN2A GPV in unrelated patients selected according to the "rule of two" [21].In a previous study of our institution, 14% of GPV carriers in CDKN2A were detected in a cohort of 59 unrelated patients from the southeast region of Brazil [25].The clinical criteria used were also the "rule of two".However, when it was considered that families presenting both MPM and multiple relatives were affected, the mutation rate of CDKN2A rose to 36% [25].
Besides CDKN2A, other genes potentially associated with familial melanoma have also been investigated in Brazilian melanoma-prone families, such as CDK4, MITF, and TERT promoter variants [24,25,27].No pathogenic variant in CDK4 was identified so far [21,24,25].Both TERT promoter mutation at −57 bp from the start site and the MITF p.E318K variant, which are, respectively, high-and moderate-penetrance variants, were also assessed in 48 unrelated probands negative for CDKN2A GPV, detecting only one patient with the MITF GPV p.E318K [27].
The purpose of this study was to perform a mutational analysis of the seven candidate melanoma susceptibility genes (ACD, BAP1, MC1R, MITF, POT1, TERF2IP, and TERT) in high-risk melanoma patients (familial melanoma and MPM) from southeastern Brazil, besides the CDKN2A and CDK4 genes.Additionally, we aimed to verify the association among patient's GPV status, their clinical features, pathological characteristics of their tumors, and the presence of other malignancies, both in patients and their relatives.

Results
A total of 37 patients (MH1-MH12 and MH14-MH38), 19 women and 18 men were enrolled in this study, all, except MH31, with three or more melanomas or two or more first-or second-degree relatives with melanoma or pancreatic cancer ("rule of three").MH31 had only one first-degree relative with melanoma and a personal history of cutaneous melanoma, kidney tumor, GIST (gastrointestinal stromal tumor), and multiple skin carcinomas.Additionally, his twin brothers had both melanoma and kidney tumors.
Regarding the series, the mean age at diagnosis was 40.1 years (15-69 years), with a significant prevalence of phototypes I and II (89.2%), blue and green eyes (54%), and red and blond hair (59.4%).Sunburn in childhood was reported in 89.2% of the cases, and a high density of freckles on the back was found in 73% of the probands.Approximately 50% of the cases had a total nevus count greater than 100, and 40% fulfilled the AMS criteria.The 37 probands had a total of 148 primary melanomas, with a predominance of the superficial spreading histopathological subtype (85.8%), a higher incidence on the trunk (52.7%), and a higher prevalence of in situ and thin melanomas (96.7%).Nevus-associated melanoma was found in 50.7% of the cases, but in 16.9% of the tumors, this information was not available.
Compared to FM, MPM patients were younger (35.6 vs. 43.5, p = 0.047) and showed a higher median number of total nevi (145 vs. 74, p = 0.04) (Table 1).The anatomical site distribution of tumors was different between the two groups: the incidence of melanomas located in the head/neck and trunk was higher in the MPM group, and the incidence of melanomas in limbs (both upper and lower) was higher in FM patients (p = 0.04).No differences were found concerning phenotype (eye and hair color, phototype, and AMS), freckles density, history of sunburn, and other tumor pathological features.Regarding genetic analysis, we included the 36 patients who had three or more primary melanomas, or at least three melanomas diagnosed in the family, including melanoma and pancreatic cancer in relatives (the proband MH31 was not considered only for these analyses).Overall, we detected five patients with potentially relevant genetic variants in the high-penetrance genes studied.Their pedigrees are shown in Figure 1.Two patients (2/36-5.6%)showed GPV in CDKN2A and BAP1 genes, and four patients (4/36-11.1%)showed prioritized variants of uncertain significance (VUS) in ACD, POT1, and TERT genes.Detailed information on all variants identified is listed in Table 2.
The patient harboring the GPV in the CDKN2A gene (MH11) was also detected with one of the ACD VUS and a non-RHC variant of MC1R (r/0).The patient has fair skin, a history of sunburn both in childhood and adulthood, total nevi count < 100, no AMS phenotype, a personal history of multiple melanomas early diagnosed (30 years old), a mother also with multiple melanomas, and a family history of pancreatic, breast and colon cancer (Figure 1A).The p.Glu419Lys variant in the ACD gene is not described in ClinVar and is classified as LB by Varsome and as VUS by Franklin.In silico analysis showed conflicting data, with a predominance of benign predictions.
The proband carrying the GPV in the BAP1 gene (MH20) also showed an RHC variant of MC1R (R/0) and had multiple nevi, fair skin, and a history of sunburn in childhood, as well as multiple melanomas diagnosed at an early age.The patient does not have a family history of melanoma but has two family members diagnosed with mesothelioma.The patient's sister, still asymptomatic, also carries the same BAP1 variant (Figure 1B).
The other case where ACD VUS was found (MH32) has a personal history of multiple melanomas and a family history of melanoma and breast cancer (Figure 1C) and showed a non-RHC variant of MC1R (r/0), in addition to fair skin, light hair, and a history of childhood sunburn.The p.Ser321Leu ACD variant is classified as VUS by ClinVar and Franklin and as LB by Varsome, presenting conflicting data from in silico analysis with the majority of pathogenicity predictors indicating benignity.The patient that carries the TERT VUS (MH28) shows a high density of freckles, a personal history of multiple melanomas and colloid thyroid goiter, and a family history of melanoma with multiple family members affected (Figure 1D).A variant of unknown risk of MC1R (u/0) was identified.The p.Glu441del TERT variant has conflicting data both in ClinVar and pathogenicity functional studies.Twelve laboratories classify it as LB/B, three as VUS, and one as P. The Varsome tool classifies this variant as VUS and Franklin as B.
The proband where the POT1 VUS was detected (MH16) also showed two MC1R variants (R/u), fair skin, light hair, a history of sunburn in childhood, a personal history of six primary melanomas and thyroid cancer, and a family history of melanoma (Figure 1E).The p.Phe106Leu POT1 variant has not been described in ClinVar and is classified as VUS by Varsome and Franklin, showing conflicting data from in silico analysis with the majority of pathogenicity predictors indicating pathogenicity.
Of all 37 probands, 86.5% (32/37) showed at least one MC1R variant.The R alleles (R151C, R160W, D84E, D294H, and R142H) were detected in 54.1% of the patients (20/37), with the R151C and R160W being the most prevalent variants (27% and 18.9%, respectively).Fifteen patients (40.5%) showed r alleles (V60L, V92M, I155T, and R163Q).V60L was the most prevalent r variant, found in 24.3% of the probands.Only five patients were detected with u variants (13.5%).Nine patients carried two allelic variants.Probands were classified as R/R, R/r, R/u, R/0, r/r, r/u, r/0 or u/0 according to the allelic combination.Two patients were detected with one homozygote R and r variant and were also classified as R/R and r/r, respectively.Five probands were MC1R wild type and were classified as 0/0.MC1R allelic data are shown in Table 3.No differences were found between FM and MPM patients concerning the MC1R variants detected (p = 0.748 R variants, p = 0.080 r variants, and p = 1.00 u variants).The R151C variant was associated with blue/gray eyes (p = 0.042) and red/blond hair (p = 0.006), where 81.8% of carriers had blue or gray eyes, and all carriers were blond or redhead.No associations were found between the MC1R variants and sex, age at diagnosis, other phenotypic characteristics, number of primary melanomas, or tumors histopathological characteristics.

Discussion
This study characterizes a cohort of 37 high-risk melanoma patients (probands with strong family history and multiple primary melanoma individuals) according to phenotype, histopathological tumor features, pedigree, and mutational status of the main melanoma susceptibility genes.
Our patients had an earlier age at their first melanoma diagnosis (mean of 40.1 years) than the general population, with a higher prevalence of thin melanomas (96.7%), predominance of the superficial spreading histological subtype (85.8%), and a large number of patients with a history of childhood sunburn (89.2%), as described in other studies that characterized FM patients [28,29].Patients with multiple melanomas alone without a family history of melanoma (MPM) had a significantly lower average age at diagnosis of the first melanoma than those with a family history of melanoma, but no data were found in the literature to corroborate or justify this finding.The frequency of melanomas located in limbs was higher in FM patients.Although some studies showed the trunk as the most frequent location of cutaneous melanoma in both sexes, a recent study has presented a higher frequency of melanomas in females in the limbs [30], which could justify our findings since the FM group showed a higher frequency of females (Table 1).
The frequency of GPVs found in the high-penetrance genes studied was low (5.6%-2/36; 4.0% for FM group-1/25 and 9.1% for MPM group-1/11), despite the selection criteria adopted ("rule of three") and the number of melanomas found in the probands and their relatives.These results differ from the literature and previous Brazilian studies (0-14%), including one from our cancer center (36%-4/11, when considering FM + MPM) [9, [31][32][33].This difference is likely to be due to the overall small sizes of ours and previously described cohorts.
Regarding the clinical phenotypes of GPV carriers, the case that shows the GPV in CDKN2A (MH11) exhibits the typical phenotype described for CDKN2A germline mutations-early age at diagnosis, multiple primary melanomas, multiple family members affected, and association with pancreatic cancer [9,10].The proband carrying the GPV in the BAP1 gene (MH20) was also diagnosed with cutaneous atypical spitzoid tumors, BIMTs (BAP1-inactivated melanocytic tumors), which are considered clinical markers of BAP1 germline mutation.The detailed description of the clinical, dermoscopic, confocal microscopy, and histopathological aspects of the skin tumors of this patient (melanomas and BIMTs), in addition to the genetic aspects, have been published recently [34].
The POT1 VUS found in proband MH16 is located in the DNA-binding domain 1 (OB1), in which other pathogenic missense variants have been described [35].Recent studies identified POT1 germline variants in melanoma-prone families with cases of thyroid cancer, as seen in this case [36,37].This proband showed two MC1R variants (R/u).It is described that patients with two MC1R variants have a higher melanoma risk compared to those with single variants [38].
Many negative cases for the high-penetrance genes studied have multiple primary melanomas, and sometimes also multiple family members affected or relatives with multiple melanomas, featuring high-risk melanoma-prone families of great interest in deepening the genetic study by extension of the genetic panel according to the other tumors found in the family and whole-exome sequencing.In this sense, recent studies have attempted to identify new melanoma-predisposing genes through comprehensive genomic studies, but although interesting candidates have been recognized, no strong validated new genes have been described so far [39].It is also possible that in some of these families, the increased risk for melanoma may be related to the MC1R genotype combined with environmental exposure since melanin synthesis is related to melanoma progression.MC1R risk variants, especially R variants, determine an imbalance between eumelanin and pheomelanin, with an increase in the latter.Pheomelanin provides a pro-oxidative cellular environment and induces DNA damage, contributing to melanomagenesis [40].
A recent study carried out in Turkey [41] corroborated the importance of the MC1R genotype in countries with a low incidence of melanoma and mixed population in determining the melanoma risk.
The main limitation of our study was the small size of our cohort, determined by funding constraints, supplying us with no sufficient statistical power in some analysis, especially when comparing the two subgroups FM × MPM and MC1R variants assessment.This limited size is an important factor that impacts the generalizability of our findings.Moreover, the absence of a comparison with a control group for the other genes beyond MC1R, due to the limited number of positive patients, hindered our ability to definitively establish the overall role of these genes in melanoma risk in the Brazilian population.Still, for MC1R, our comparisons from our cohort to a population-matched control group showed an increased frequency of R and u variants in the patients' group.
Our patients showed a high prevalence of MC1R variants (86.5% total; 88.5% in FM and 81.8% in MPM), with the R151C (R) and V60L (r) being the most frequent variants detected (27.3% and 24.3%, respectively).A recent polled analysis showed that the presence of any MC1R variant confers a 60% higher risk for cutaneous melanoma to carriers when compared to noncarriers [42].RHC variants were identified in over 50% of all probands, with a significant increase when compared to the control group, corroborating their high prevalence in melanoma patients [43].In our study, the R151C variant was associated with light hair (red and blond) and blue/gray eyes, supporting the role of RHC variants in the determination of light phenotypic complexion [44].
An interesting case was the MH29 proband (Table 2), who developed multiple primary melanomas diagnosed at an early age and had a first-degree relative also with multiple melanomas.No variant in the high-penetrance genes studied was found, but two RHC variants of the MC1R gene (R/R) could explain, partly, the high-risk profile for melanoma seen in this proband.A recent study reported that MC1R genotype and nevi number concur synergistically to melanoma risk, and the R/R genotype combined with high nevi count results in a deeply high-risk outline for melanoma [45].
The MH01 proband (Table 2) also had multiple primary melanomas, and the first tumor was diagnosed before the age of 40.His brother also has multiple primary melanomas in addition to non-melanoma skin cancer and thyroid carcinoma.The proband showed no variant in the high-penetrance genes studied.Unlike the proband MH29, this patient presented only one u variant (p.Gln23Ter-unknown risk) of MC1R (u/0).Despite presenting an unknown risk, the p.(Gln23Ter) variant is a loss-of-function (LoF) variant, and MC1R LoF variants have already been associated with the risk of MPM [46].
Moreover, this group of MC1R variants (u alleles) showed a higher prevalence in our cohort when compared with the control group and may characterize high-risk melanoma patients in the Brazilian population.

Patients' Selection
Melanoma patients recruited between 2016 and 2019 from the Familial Melanoma Clinic at the Skin Cancer Department of A.C. Camargo Cancer Center, São Paulo, Brazil, with 3 or more melanomas or 2 or more first or second-degree relatives with melanoma or pancreatic cancer ("rule of three") were included in the study.All participants provided written and informed consent, and this study was approved by the ethics committee of the A.C. Camargo Cancer Center (2076/15).
A standardized questionnaire containing phenotypic characteristics, patients' demographic data, and tumor features was used, including sex, age at diagnosis, eye and hair color, skin phototype, total nevi count, presence of atypical mole syndrome (AMS) phenotype, freckles density, history of sunburn, Breslow thickness, tumor location, histological subtype, tumor-associated nevus, staging, proband and relatives' number of melanomas, and proband and relatives' presence of other tumors.Patients were classified as FM (familial melanoma = family history of melanoma or pancreatic cancer) or MPM (multiple primary melanomas = without family history of melanoma or pancreatic cancer).

Germline Genetic Analysis
Genomic DNA was obtained from blood leucocytes or saliva and then was subjected to a library preparation protocol according to an Ion AmpliSeq™ Library Kit 2.0 and sequenced with the Ion Proton Platform (Thermo Fisher Scientific, Waltham, MA, USA).The entire coding region of nine melanoma predisposition genes (ACD, BAP1, CDKN2A, CDK4, MC1R, MITF, POT1, TERF2IP, and TERT) were analyzed using a custom Ion Ampliseq™ Panel (Thermo Fisher Scientific, Waltham, MA, USA).
The Torrent Variant Caller tool 5.0-13 (ThermoScientific) was used to perform the alignment and the variant calling.The identified variants were noted and filtered using the VarSeq software (Version 1.8, Golden Helix, Bozeman, MT, USA).Variants with the following quality criteria were selected: base coverage ≥50× and frequency of the variant allele ≥0.25.The detected variants were evaluated for their classification in databases of clinical classification of variants (ClinVar, LOVD) and population allele frequency (gnomAD, 1000G, ESP, and ABraOM) using the Varsome tool.The variants were classified according to the criteria suggested by the American College of Medical Genetics (ACMG) [47].

Statistical Analysis
In assessing the association between categorical variables, the chi-square test or Fisher's exact test was used when appropriate.To compare the two groups (FM and MPM) and the MC1R variants according to numerical variables, the Mann-Whitney nonparametric test was applied.
A comparison of the MC1R variant frequencies found in our cohort with a control group was performed using a logistic regression model to calculate odds ratios (OR) with confidence intervals (CIs) of 95%.The control group used was the database of the Brazilian Online Archive of Variants (AbraOM), which represents a cohort of elderly Brazilians (over 60 years of age) with extensive phenotyping based on the SABE study census (Health, Wellness and Aging) initiated in 1999, enclosing data from São Paulo city.As approximately 10% of the Brazilian population lives in the metropolitan region of São Paulo, this is considered a representative sample of the national population.It includes 609 individuals, 216 men and 393 women, 517 self-declared white, 156 black, 12 yellow, 2 indigenous, 40 self-declared as "others", and 52 without definition [49].
The level of significance adopted was 5%, and statistical analyses were performed using IBM SPSS software version 25.

Conclusions
A low frequency of high-penetrance gene variants and a high prevalence of MC1R variants found in our cohort emphasize the likely importance of MC1R variants in determining melanoma risk in our population.The higher frequency of RHC and u MC1R variants in our cohort compared to the control group and the presence of two MC1R variants demonstrates that, to a certain extent, the increased melanoma risk of these individuals and families may be related to MC1R genotype combined to pigmentation phenotype, behavior of risk (UVR exposure) and the presence of multiple nevi.Future studies should seek to investigate how

Figure 1 .
Figure 1.Pedigrees of the probands that showed genetic variants in the high-penetrance genes.(A) Pedigree of FM patient (MH11) carrying the GPV in CDKN2A, VUS in ACD gene and classified as r/0 for MC1R.(B) Pedigree of MPM patient (MH20) carrying the GPV in BAP1 and classified as R/0 for MC1R.(C) Pedigree of FM patient (MH32) carrying the other VUS in ACD gene and classified as r/0 for MC1R.(D) Pedigree of FM patient (MH28) carrying the VUS in TERT gene and classified as u/0 for MC1R.( ) Pedigree of FM patient (MH16) carrying the VUS in POT1 gene and classified as R/u for MC1R.

Figure 1 .
Figure 1.Pedigrees of the probands that showed genetic variants in the high-penetrance genes.(A) Pedigree of FM patient (MH11) carrying the GPV in CDKN2A, VUS in ACD gene and classified as r/0 for MC1R.(B) Pedigree of MPM patient (MH20) carrying the GPV in BAP1 and classified as R/0 for MC1R.(C) Pedigree of FM patient (MH32) carrying the other VUS in ACD gene and classified as r/0 for MC1R.(D) Pedigree of FM patient (MH28) carrying the VUS in TERT gene and classified as u/0 for MC1R.(E) Pedigree of FM patient (MH16) carrying the VUS in POT1 gene and classified as R/u for MC1R.

Table 2 .
Description of germline variants detected in known and candidate melanoma genes.