Genetic Risk Factors for Early-Onset Merkel Cell Carcinoma

Key Points

Question

What genetic risk factors are associated with early-onset Merkel cell carcinoma (MCC)?

Findings

In this case-control study of 1012 individuals (37 with early-onset MCC, 45 with later-onset MCC, and 930 unrelated controls), 7 patients (19%) with early-onset MCC were positive for germline disease variants in genes causing cancer predisposition syndromes (ATM, BRCA1, BRCA2, TP53, and MAGT1), whereas 0% of patients with later-onset MCC and 9 controls (1%) carried such variants.

Meaning

Germline variants in genes associated with cancer predisposition are significantly associated with early-onset MCC; accordingly, genetic counseling and testing should be considered for patients with MCC who are younger than 50 years.

This case-control study examines genetic risk factors associated with early-onset Merkel cell carcinoma.

Abstract

Importance

Merkel cell carcinoma (MCC) is a rare, aggressive neuroendocrine skin cancer. Of the patients who develop MCC annually, only 4% are younger than 50 years.

Objective

To identify genetic risk factors for early-onset MCC via genomic sequencing.

Design, Setting, and Participants

The study represents a multicenter collaboration between the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS), the National Institute of Allergy and Infectious Diseases (NIAID), and the University of Washington. Participants with early-onset and later-onset MCC were prospectively enrolled in an institutional review board–approved study at the University of Washington between January 2003 and May 2019. Unrelated controls were enrolled in the NIAID Centralized Sequencing Program (CSP) between September 2017 and September 2021. Analysis was performed from September 2021 and March 2023. Early-onset MCC was defined as disease occurrence in individuals younger than 50 years. Later-onset MCC was defined as disease occurrence at age 50 years or older. Unrelated controls were evaluated by the NIAID CSP for reasons other than familial cancer syndromes, including immunological, neurological, and psychiatric disorders.

Results

This case-control analysis included 1012 participants: 37 with early-onset MCC, 45 with later-onset MCC, and 930 unrelated controls. Among 37 patients with early-onset MCC, 7 (19%) had well-described variants in genes associated with cancer predisposition. Six patients had variants associated with hereditary cancer syndromes (ATM = 2, BRCA1 = 2, BRCA2 = 1, and TP53 = 1) and 1 patient had a variant associated with immunodeficiency and lymphoma (MAGT1). Compared with 930 unrelated controls, the early-onset MCC cohort was significantly enriched for cancer-predisposing pathogenic or likely pathogenic variants in these 5 genes (odds ratio, 30.35; 95% CI, 8.89-106.30; P < .001). No germline disease variants in these genes were identified in 45 patients with later-onset MCC. Additional variants in DNA repair genes were also identified among patients with MCC.

Conclusions and Relevance

Because variants in certain DNA repair and cancer predisposition genes are associated with early-onset MCC, genetic counseling and testing should be considered for patients presenting at younger than 50 years.

Introduction

Merkel cell carcinoma (MCC) is a rare neuroendocrine skin cancer with approximately 3000 US cases per year.¹ Approximately 80% of MCC tumors are driven by integrated Merkel cell polyomavirus (MCPyV).² Additional risk factors include UV-mediated DNA damage, immunosuppression, and advanced age.³ The median age of MCC diagnosis is 76 years.⁴ Early-onset MCC—defined here as MCC occurrence in individuals younger than 50 years—accounts for only 4% of cases.^5,6,7,8 Understanding the etiology of MCC is crucial to improving treatments and prevention. Because patients with germline cancer predispositions often develop cancer at younger ages,^9,10 we set out to determine whether patients with early-onset MCC harbor germline cancer risk variants.

Methods

Sample Collection

Early-Onset MCC Cohort

Patients with histopathologically confirmed MCC were prospectively enrolled between January 2003 and May 2019 (time of data lock) in an institutional review board–approved repository at the University of Washington with written informed consent (eMethods in Supplement 1).

MCPyV testing was performed using a combination of immunohistochemical staining with the CM2B4 antibody to detect large T antigen in formalin-fixed, paraffin-embedded tumor sections and/or the AMERK test for circulating antibodies against small T antigen. A patient’s tumor was considered MCPyV positive if either test result was positive, and negative if both test results were negative.¹¹

Control Cohort

Controls consisted of 930 participants (eFigure 2 in Supplement 1) enrolled in the National Institute of Allergy and Infectious Disease (NIAID) Centralized Sequencing Program (CSP) who did not have MCC and were undergoing evaluations for reasons other than familial cancers or MCC. Participants were referred September 2017 through September 2021 from a diverse set of NIH research protocols studying immunological, neurological, and psychiatric disorders. Phenotype curation was completed using standardized Human Phenotype Ontology (HPO) terminology.¹² All participants provided written informed consent and the study was approved by the NIAID institutional review board (NCT03206099).

Later-Onset MCC Cohort

Patients from the University of Washington repository diagnosed with MCC when older than 50 years whose germline DNA had undergone whole-exome sequencing (WES) as matched controls for tumor mutational analysis (eFigure 3 in Supplement 1).¹³

Genome Sequencing

Genome sequencing (GS) for patients with early-onset MCC was performed at the NIAMS sequencing core on peripheral blood DNA. Libraries were prepared using the Illumina DNA PCR-free Library Prep kit (Illumina, 20041795) following manufacture’s instructions (Illumina, version 03; 1000000086922). Briefly, bead-linked transposomes PCR-free (BLT-PF) were used to tagment 500 ng of genomic DNA. Following a posttagmentation clean-up, Index 1 (i7) and Index 2 (i5) adapters were ligated to each sample. Before sequencing, final libraries were pooled and quantified by Qubit ssDNA (single-stranded) assay kit (ThermoFisher, #Q10212) to ensure optimal cluster density. Samples were sequenced on the Illumina NovaSeq 6000 on S2 or S4 flow cells using the version 1.0 NovaSeq reagent kit with VP10 custom sequencing primer. All libraries were sequenced to generate 151 bp paired-end reads with 10 bp each for i7 and i5 indexes.

GS for controls was performed at the Human Genome Sequencing Center Clinical Laboratory at Baylor College of Medicine using extracted DNA. Libraries were prepared using KAPA Hyper PCR-free library reagents in Beckman robotic workstations. Briefly, 750 ng of DNA was sheared into fragments of approximately 200 to 600 bp using the Covaris E220 system (96 well format; Covaris, Inc) followed by purification of the fragmented DNA using AMPure XP beads. A double size selection step was used, with different ratios of AMPure XP beads, to select a narrow size band of sheared DNA molecules for library preparation. DNA end-repair and 3′-adenylation were then performed in the same reaction followed by ligation of the Illumina unique dual barcodes adapters (Cat# 20022370) to create PCR-Free libraries, and the library was run on the Fragment Analyzer (Advanced Analytical Technologies, Inc) to assess library size and presence of remaining adapter dimers. This was followed by quantitative polymerase chain reaction (qPCR) assay using the KAPA Library Quantification Kit (KK4835) using their SYBR FAST qPCR Master Mix to estimate the size and quantification. Samples were sequenced on the Illumina NovaSeq S4 flow cells to generate 150 bp paired-end reads. As a quality control measure, each sample was also single-nucleotide variation (formerly, single-nucleotide polymorphism [SNP]) array genotyped with the Fluidigm SNPtrace Panel.¹⁴ SNPtrace genotyping results were compared with the GS data to ensure correct sample identification and to assess sequencing quality.

Sequence data for both the cases and controls were processed using the same pipeline. Specifically, FASTQ files were aligned with BWA-MEM (version 0.7.17)¹⁵ to the GRCh38 human genome reference, PCR duplicates were marked using Samblaster (version 0.1.25),¹⁶ and BAM files were sorted using Samtools (version 1.11).¹⁷ Sorted, deduplicated BAM files were then processed following the GATK4 Best Practices¹⁸ and using GATK (version 4.1.9.0). In summary, BAM files underwent base call quality score recalibration (BQSR), single-sample gVCFs were generated using the HaplotypeCaller, joint genotyping was performed using GenotypeGVCFs, and variants were then processed using Variant Quality Score Recalibration (VQSR) and genotype refinement following the GATK recommended settings.

Exome Sequencing

Existing exome sequencing data were reanalyzed for this study.¹³ Exome capture was performed on peripheral blood DNA using the 2.1M NimbleGen exome reagent (Roche NimbleGen) and 75 base-paired end sequencing was done on an Illumina 2000 (Illumina). We used the OpenOmics Genome-Seek pipeline as described.¹⁹ Briefly, following this pipeline, set to call variants in exome samples, we trimmed the short-read data using fastp to remove adapter sequences and low-quality sequences (default parameters).²⁰ Following trimming, paired-end reads were mapped to the hg38 human reference genome using BWA-MEM2 (parameters: -K 1000000000 -M, BWA-MEM2 reference).²¹ Germline variants were called using DeepVariant, a deep learning-based variant caller.²² We used VEP²³ and Slivar²⁴ to tag variants associated with different genes and to remove common variants reported in gnomAD.²⁵

Molecular Genomics Analyses

Clinical-grade analysis was carried out using seqr, an enhanced analysis tool developed by the Broad Institute and further customized for the NIAID CSP.²⁶ Pathogenic and likely pathogenic variants with established gene-disease associations were curated and interpreted according to the American College of Medical Genetics and Genomics (ACMG) guidelines.^27,28 Variants without an established gene-disease association were classified based on the guidelines of the ACMG and the Association for Molecular Pathology.²⁸

Genetic analyses were conducted using 3 curated gene lists: genes associated with inborn errors of immunity (IEI), nuclear DNA repair genes, and hereditary cancer genes (Gene Lists in Supplement 2). For each list, genes were assessed for the presence of variants with apparent loss of function (frameshift, nonsense, and essential splice site variants), as well as missense, extended splice site, and in-frame deletions and duplications. Variants in IEI genes were classified as pathogenic if also associated with conditions of immunodeficiency, whereas variants in DNA repair and hereditary cancer genes were classified as pathogenic if also associated with conditions of cancer predisposition. Novel variants were classified as having a minor allele frequency of 0 in the curated Human Gene Mutation (HGMD) and gnomAD databases (date of search February 2022).

We first identified rare pathogenic and likely pathogenic variants in inborn errors of immunity genes previously described in the curated International Union of Immunological Societies (IUIS) gene list.^29,30 With the exception of 1 patient with a MAGT1 variant, there were no other individuals with a molecular diagnosis of any immune-related disease. Accordingly, further analysis on immunodeficiencies was not performed. However, during the analysis of the IUIS genes, we identified many variants in DNA damage repair genes associated with cancer predisposition; as such, we analyzed variants in a curated list of genes associated with nuclear DNA damage repair.^{31,32,33,34,35,36,37,38,39,40,41} Lastly, to investigate cancer predisposition genes not previously accounted for in the IUIS and DNA damage repair genes lists, we analyzed a gene list based on the ACMG list for reporting secondary findings⁴² and publicly available direct-to-consumer hereditary cancer risk panels. All individuals who were still living and found to have a cancer predisposition syndrome or a genetic immunodeficiency were notified and referred to a genetic counselor for clinical evaluation.

Statistical Analyses

Peddy (version 0.4.6) was used to perform principal component analysis to generate ancestry principal components from germline variation.⁴³ The 1000 Genomes phase 3 reference panel of 2504 individuals was used to generate predicted ancestry,⁴⁴ and principal components 1, 2, and 3 were plotted using the ggplot2 package in R statistical software (R Foundation).⁴⁵

Cohort demographics were assessed using descriptive statistics. Demographics were compared using a nonparametric independent sample t test to compare the medians for age, χ² test for differences in sex and immunosuppression, and Fisher exact test for differences in MCC stage between cohorts. Multiple logistic regression adjusting for sex, age, and the first 10 ancestry principal components as covariates was used to test the association between MCC case-control status and having a cancer-predisposing pathogenic or likely pathogenic variant in ATM, BRCA1, BRCA2, TP53, and MAGT1 using the following model: ~ sex + age + PC1 + PC2 + PC3 + PC4 + PC5 + PC6 + PC7 + PC8 + PC9 + PC10 + has P/LP variant. Age corresponded to age at MCC diagnosis in patients with MCC. Age at study enrollment was used for controls. All statistical analyses were completed using R statistical software (version 4.1.3, R Foundation).⁴⁶ Scripts for analyses are available upon request.

Results

To identify genetic risk factors for early-onset MCC, we performed GS on patients diagnosed with MCC when younger than 50 years.^7,8 Participants were identified by searching the MCC database at the University of Washington, wherein 78 of 1637 participants (5%) presented at age younger than 50 years, and 37 of these had peripheral blood mononuclear cells (PBMC) available for sequencing (Figure). Consistent with patient-reported race data (not shown to preserve patient privacy), a principal component analysis of ancestry revealed that most participants were of European genomic ancestry (eFigure 1 in Supplement 1). Overall, the cohort included 20 men (54%) and 17 women (46%) (Table 1), and the median (range) age of diagnosis was 45 (18-49) years. Of the 26 patients tested for MCPyV, 25 (96%) had positive tumors, suggesting that young patients with MCC have a higher rate of virus positive tumors. Overall, 32 participants (87%) lacked a history of immunosuppression (Table 1). Tumor primary site and stage are listed in eTable 1 in Supplement 1.

Figure. Flow Diagram of Patients With Histopathologically Confirmed Early-Onset Merkel Cell Carcinoma (MCC) Included in the Analysis.

Table 1. Early-Onset MCC, Later-Onset MCC, and Control Cohort Demographics and Clinical Data Summary.

Characteristic	No. (%)			P value
	Early-onset MCC (N = 37)	Later-onset MCC (N = 45)	Control cohort (N = 930)	Early-onset vs later-onset MCC	Early-onset vs control cohort
No. of participants	37	45	930
Agea
0-17	0	0	73 (7.8)	NA	.71
18-29	5 (13.5)	0	117 (12.6)
30-39	3 (8.1)	0	162 (17.4)
40-49	29 (78.4)	0	206 (22.2)
50-59	0	17 (37.8)	216 (23.2)
60-69	0	12 (26.7)	113 (12.2)
70-79	0	11 (24.4)	38 (4.1)
≥80	0	5 (11.1)	5 (0.5)
Sex
Female	17 (45.9)	19 (42.2)	499 (53.7)	.74	.45
Male	20 (54.1)	26 (57.8)	431 (46.3)
Immunosuppression
No	32 (86.5)	39 (86.7)	NA	.98	NA
Yesb,c	5 (13.5)	6 (13.3)	NA
MCC stage
I/II	17 (45.9)	15 (33.3)	NA	.21	NA
III	19 (51.4)	28 (62.2)	NA
IV	1 (2.7)	2 (4.4)	NA
Tumor MCPyV status
No. tested	26	37	NA	NAd	NA
Positive	25 (96.2)	20 (54.1)	NA
Negative	1 (3.8)	17 (45.9)	NA

Abbreviations: MCC, Merkel cell carcinoma; NA, not available.

^a Age at diagnosis in MCC cohorts and age at enrollment in control cohort.

^b Immunosuppression in the early-onset cohort includes solid organ transplant (n = 3) and autoimmune disease (n = 2).

^c Immunosuppression in the later-onset cohort includes autoimmune disease (n = 4) and chronic lymphocytic leukemia (n = 2).

^d Later-onset cohort was stratified by Merkel cell polyomavirus status.

We performed an unbiased analysis of GS data. Because MCC is immunogenic, we focused on the curated IUIS gene list^29,30; after identifying variants in DNA repair genes, we further focused on a list of nuclear DNA damage repair genes.^{31,32,33,34,35,36,37,38,39,40,41} Of the 37 patients, 7 (19%) were found to be heterozygous or hemizygous for well-described pathogenic or likely pathogenic variants in genes associated with cancer predisposition (Table 2). Among these 7 patients, we detected a variant in MAGT1 (associated with immunodeficiency and increased lymphoma risk⁴⁷) in 1 patient and variants associated with heritable cancer predisposition syndromes, including inherited risk for breast, ovarian, and other cancers in 6 patients (ATM = 2, BRCA1 = 2, BRCA2 = 1, and TP53 = 1). Notably, ATM, BRCA1, BRCA2, and TP53 all function in DNA repair. We identified additional DNA repair gene variants including 16 patients (43%) with variants not found in general population databases and 3 patients (8%) who were heterozygous carriers for cancer-predisposing pathogenic/likely pathogenic variants (eTables 2 and 3 in Supplement 1). Because carrier states and novel variants in these additional DNA repair genes are not known to cause cancer predisposition, it is unclear if they contributed to MCC risk. Six patients had variants in multiple genes associated with cancer predisposition or DNA repair (eTable 4 in Supplement 1).

Table 2. Pathogenic and Likely Pathogenic Variants in Genes That Predispose to Cancer Associated With Early-Onset Merkel Cell Carcinoma.

Patient identification	Gene	Coding	Protein	Zygosity	Transcript (NM#)	P/LP	Human Gene Mutation database accession number	Sex
W892	MAGT1	c.712C>T	p.Arg238Ter	Hemi	NM_001367916	P	CM152374	M
W754	ATM	c.790del	p.Tyr264fs	Het	NM_000051.4	P	CD991594	M
W864	ATM	c.378del	p.Asp126fs	Het	NM_000051.4	P	CD1716656	F
W771	BRCA2	c.5946del	p.Ser1982fs	Het	NM_000059.4	P	CD961857	M
W214	BRCA1	c.4065_4068del	p.Asn1355fs	Het	NM_007294.4	P	CD941619	F
Z1395	BRCA1	c.2071del	p.Arg691fs	Het	NM_007294.4	P	CD982486	F
W112	TP53	c.527G>T	p.Cys176Phe	Het	NM_000546.6	LP	CM1619982	M

Abbreviations: Het, heterozygous; hemi, hemizygous; P, pathogenic for cancer predisposition; LP, likely pathogenic for cancer predisposition; M, male; F, female.

In addition, to assess the association of variants in hereditary cancer genes not included in the IUIS gene list or nuclear DNA repair gene list, we curated a list of genes based on the ACMG secondary findings gene list and publicly available direct-to-consumer hereditary cancer risk panels (gene lists in Supplement 2). Analysis of this hereditary cancer gene list identified no additional variants pathogenic or likely pathogenic for cancer predisposition in this cohort.

To assess whether cancer-predisposing pathogenic/likely pathogenic variants in ATM, BRCA1, BRCA2, TP53, and MAGT1 were significantly enriched in our early-onset MCC cohort, we performed GS analysis using the same variant classification guidelines in a control cohort of 930 unrelated individuals. Controls were undergoing GS for reasons other than familial cancers or MCC and were not significantly different from case participants with respect to age and sex (Table 1). Variants in the 5 genes that were pathogenic or likely pathogenic for cancer predisposition were observed in 9 control individuals (1%) (eTable 5 in Supplement 1). Specifically, 4 ATM (0.4%), 2 BRCA1 (0.2%), 2 BRCA2 (0.2%), and 1 MAGT1 (0.1%) variants were identified. Multiple logistic regression adjusting for sex, age, and the first 10 ancestry principal components as covariates (eTable 6 in Supplement 1) showed that early-onset MCC was significantly associated with cancer-predisposing pathogenic/likely pathogenic variants in ATM, BRCA1, BRCA2, TP53, and MAGT1 (odds ratio, 30.35; 95% CI, 8.89-106.30; P < .001). Sex, age, and ancestry PCs 1 to 8 and PC10 were not significantly associated with MCC case-control status. Ancestry PC9 was associated with P = .05, emphasizing the importance of adjusting for genetic ancestry when testing for associations with rare genetic variants.

To assess whether the MCC risk associated with germline mutations in these 5 genes was specific to early-onset MCC, we analyzed existing WES data from 45 patients diagnosed with MCC at age 50 years or older (later-onset MCC). Demographic information for the later-onset MCC cohort is available in eTable 7 in Supplement 1. No alleles pathogenic or likely pathogenic for cancer predisposition were detected in ATM, BRCA1, BRCA2, TP53, or MAGT1 in patients with later-onset MCC. The later-onset MCC cohort had 7 patients (15.6%) with DNA repair gene variants not found in general population databases and 9 patients (20%) were heterozygous for presumed loss of function variants in genes associated with nuclear DNA repair (eTables 8 and 9 in Supplement 1); however, none of these genes have disease alleles associated with increased cancer risk.

Discussion

To our knowledge, this is the first study to identify potential heritable risk factors for MCC. We discovered that 19% of patients with early-onset MCC were associated with variants in cancer predisposition syndrome genes. Germline variants in ATM, BRCA1, BRCA2, TP53, and MAGT1 were significantly more common among patients with early-onset MCC than controls and were not identified in patients with later-onset MCC. These variants are respectively associated with susceptibility to breast cancer, BRCA1- and BRCA2-associated hereditary breast and ovarian cancer (HBOC) syndrome, Li-Fraumeni syndrome, and X-linked immunodeficiency with magnesium defect, Epstein-Barr virus infection, and neoplasia (XMEN) disease. On clinical review, the patients with MCC with these variants all had personal or family histories consistent with the associated genetic syndromes. It is noteworthy that MCC has not been considered to run in families, nor has it been associated with well-studied syndromes such as HBOC. It is likely that the association between MCC and familial cancer risk variants has gone undetected owing to the rarity of MCC, infrequent genetic testing of patients with MCC, and ascertainment bias in family history-based studies of hereditary cancer syndromes.⁴⁸

Some evidence suggests that the incidence of early-onset MCC is decreasing⁸; however, other research has shown that the overall incidence of early-onset cancers is on the rise.⁷ Although the factors responsible for this increase have yet to be elucidated, changes in lifestyle, environmental exposures, and the presence of germline variants in hereditary cancer genes are all implicated in this alarming trend.⁷ The high frequency of DNA repair gene variants identified among early-onset and later-onset patients with MCC suggests that DNA repair defects may increase MCC risk. This association between DNA repair defects and MCC is further strengthened by the lack of cancer-predisposing pathogenic and likely pathogenic variants in genes unrelated to DNA damage repair. In addition, prior studies investigating the genomics of MCC tumors have reported somatic sequence variants in DNA repair genes including ATM, BRCA1, BRCA2, and TP53.^{13,49,50,51,52} These results support a potential role for DNA repair defects in MCC carcinogenesis and are consistent with a model where germline variants in these genes predispose to MCC. Importantly, germline variants in DNA repair genes can increase the risk for other rare, early-onset cancers.^10,53

The exceptionally high rate of virus-positive tumors (96%) we observed among patients with early-onset MCC is reasonable when considering that UV-signature DNA mutations drive virus-negative MCC pathogenesis.^3,4 Younger individuals typically have less accumulated skin damage from UV exposure. These findings also suggest that defective DNA repair may potentiate MCPyV integration events to promote virus-positive MCC formation. Indeed, augmented DNA damage has been associated with the integration of other human oncoviruses.⁵⁴ The fact that MCPyV T antigens can inhibit DNA repair may further enhance this process in virus-infected cells.⁵⁵ Elucidating the specific mechanisms whereby these genetic variants increase MCC risk will require further investigation.

Limitations

Our results should be interpreted in the context of several limitations. First, given the rarity of early-onset MCC, the cohort was small. Consequently, estimating a precise effect size for genetic variants on MCC risk is difficult. Moreover, on account of the cohort size, this study was underpowered for rare variant association testing. Accordingly, we restricted the case-control analysis to a clinical-grade analysis for the 5 genes identified in early-onset cohort. Second, the control participants were clinically heterogeneous and may not represent a population-based sample. Third, the later-onset MCC analysis was done using existing WES rather than GS. Nonetheless, these results support an association between heritable cancer predisposition variants and increased risk for early-onset MCC.

Conclusions

This case-control study found that 19% of patients with early-onset MCC had variants in genes associated with heritable cancer predisposition, suggesting that these patients and their families may have a higher risk for other cancers. Accordingly, genetic counseling and cascade testing of potentially affected family members should be considered for patients diagnosed with MCC at age younger than 50 years, with a focus on detecting pathogenic variants in ATM, BRCA1, BRCA2, TP53, MAGT1, or DNA repair genes. Similarly, a personal or family history of early-onset MCC should be considered relevant when evaluating patients for heritable cancer syndromes. Future genetic testing of patients with MCC will allow for additional understanding of how germline variants influence MCC risk.

Supplement 1.

eMethods

eFigure 1. Genomic ancestry of patients with early-onset MCC

eFigure 2. Flow diagram of control cohort included in the analysis

eFigure 3. Flow diagram of patients with histopathologically-confirmed later-onset Merkel cell carcinoma (MCC) included in the analysis

eTable 1. Demographics of patients with early-onset MCC

eTable 2. Novel† germline variants in genes associated with nuclear DNA repair identified in patients with early-onset MCC

eTable 3. Heterozygous carriers of presumed loss of function variants in genes associated with nuclear DNA repair identified in patients with early-onset MCC

eTable 4. Patients with early-onset MCC listed in Table 2 and/or eTable 2 with more than one variant in genes that predispose to cancer or genes involved in DNA repair

eTable 5. Pathogenic and likely pathogenic germline variants in the five cancer predisposition genes associated with early-onset MCC (ATM, BRCA1, BRCA2, TP53, and MAGT1) observed in controls

eTable 6. Logistic regression model for case-control analysis testing for association with variants in ATM, BRCA1, BRCA2, TP53, and MAGT1

eTable 7. Demographics of patients with later-onset MCC

eTable 8. Novel germline variants in genes associated with nuclear DNA repair identified in patients with later-onset MCC

eTable 9. Heterozygous carriers of presumed loss of function variants in genes associated with nuclear DNA repair identified in patients with later-onset MCC

Supplement 2.

Gene Lists

Supplement 3.

Data Sharing Statement