Use of Big Data to Estimate Prevalence of Defective DNA Repair Variants in the US Population

Abstract

Importance

Wide use of genomic sequencing to diagnose disease has raised concern about the extent of genotype-phenotype correlations.

Objective

To correlate disease-associated allele frequencies with expected and reported prevalence of clinical disease.

Design, Setting, and Participants

Xeroderma pigmentosum (XP), a recessive, cancer-prone, neurocutaneous disorder, was used as a model for this study. From January 1, 2017, to May 4, 2018, the Human Gene Mutation Database and a cohort of patients at the National Institutes of Health were searched and screened to identify reported mutations associated with XP. The clinical phenotype of these patients was confirmed from reports in the literature and National Institutes of Health medical records. The genetically predicted prevalence of disease based on frequency of known pathogenic mutations was compared with the prevalence of patients clinically diagnosed with phenotypic XP. Exome sequencing of more than 200 000 alleles from the Genome Aggregation Database, the National Cancer Institute Division of Cancer Epidemiology and Genetics database of healthy controls, and an Inova Hospital Study database was used to investigate the frequencies of these mutations in the general population.

Main Outcomes and Measures

Listing of all reported mutations associated with XP, their frequencies in 3 large exome sequence databases, determination of the number of patients in the United States with XP using modeling equations, and comparison of the observed and reported numbers of patients with XP with specific mutations.

Results

A total of 156 pathogenic missense and nonsense mutations associated with XP were identified in the National Institutes of Health cohort and the Human Gene Mutation Database. The Genome Aggregation Database provided frequency data for 65 of these mutations, with a total allele frequency of 1.13%. The XPF (ERCC4) mutation, p.P379S, had an allele frequency of 0.4%, and the XPC mutation, p.P334H, had an allele frequency of 0.3%. With the Hardy-Weinberg equation, it was determined that there should be more than 8000 patients who are homozygous for these mutations in the United States. In contrast, only 3 patients with XP were reported as having the XPF mutation, and 1 patient was reported as having the XPC mutation.

Conclusions and Relevance

The findings from this study suggest that clinicians should approach large genomic databases with caution when trying to correlate the clinical implications of genetic variants with the prevalence of disease risk. Unsuspected mutations in known genes with a predisposition for skin cancer may be responsible for some of the high frequency of skin cancers in the general population.

This molecular epidemiologic study examinsd 3 large exome sequence databases totaling more than 200 000 alleles to correlate disease-associated allele frequencies with the expected and reported prevalence of clinical disease.

Key Points

Question

Do databases of exome sequences reliably correlate with the prevalence of individuals with defective DNA repair?

Findings

In this molecular epidemiologic study examining 3 large exome sequence databases totaling more than 200 000 alleles, unexpectedly high frequencies were found of 2 mutations associated with xeroderma pigmentosum in DNA repair genes (XPF [ERCC4] p.P379S, 0.4% and XPC p.P334H, 0.3%). These frequencies estimate the presence of more than 8000 people with xeroderma pigmentosum in the United States with these mutations, yet only 4 individuals were clinically identified in this study.

Meaning

Unsuspected mutations in known genes with a predisposition for skin cancer may be responsible for some of the high frequency of skin cancers in the general population.

Introduction

Xeroderma pigmentosum (XP) is a rare, autosomal recessive disorder associated with mutations in genes involved in the nucleotide excision repair pathway. Xeroderma pigmentosum can be caused by mutations in 7 nucleotide excision repair genes (XPA-XPG [XPA, OMIM 611153; XPB excision repair, complementing defective, in Chinese hamster, 3, ERCC3, OMIM 133510; XPC, OMIM 613208; XPD, excision repair, complementing defective, in Chinese hamster, 2, ERCC2, OMIM 126340; XPE, DNA damage-binding protein 2, DDB2, OMIM 600811; XPF, excision repair, complementing defective, in Chinese hamster, 4, ERCC4, OMIM 133520; XPG, excision repair, complementing defective, in Chinese hamster, 5, ERCC5, OMIM 133530]) and in the POLH gene (POLH XP variant, polymerase, DNA, ETA, OMIM 603968), a bypass polymerase, and is characterized by the inability to repair UV-induced and chemically induced DNA damage.¹ Patients with XP are usually clinically identified owing to severe sun sensitivity after minimum sun exposure, freckle-like (lentiginous) pigmentation on sun-exposed skin before 2 years of age, and premature development of many skin cancers. Some patients (approximately 20% in the United States) also have early onset of progressive neurologic degeneration.^1,2 Patients with XP younger than 20 years have a 10 000-fold increased risk of sunlight-induced skin cancer.² The frequency of XP in western Europe is clinically estimated to be about 1 in 1 million individuals.³ Clinical diagnosis is often confirmed by DNA sequencing of the genes known to cause XP.

With the growing emphasis on using genomic sequencing to assess risk of disease, we investigated the frequency of XP using data from large genomic databases. To identify and analyze the prevalence of recessive germline mutations associated with XP, we mined 4 databases: the Human Gene Mutation Database,⁴ the Genome Aggregation Database (gnomAD)⁵; an independent database from the National Institutes of Health (NIH), National Cancer Institute Division of Cancer Epidemiology and Genetics (DCEG)^6,7,8; and an Inova Hospital Study database.⁹ We also examined our cohort of patients at the NIH with XP.² We sought to compare the genetically estimated prevalence of disease, based on the frequency of known pathogenic mutations in heterozygous carriers from the general population, with the reported prevalence of the clinically observed and diagnosed XP phenotype. The clinical pathogenicity of the mutations associated with XP were confirmed by identifying patients in the literature who harbored these mutations, who received a diagnosis of clinical XP, and who underwent a laboratory assessment confirming reduced DNA repair.

We found high allele frequencies for several XP genetic variants. This finding suggests that there should be thousands of patients who are homozygous for missense or nonsense mutations in XPF, XPC, and other genes causing XP and should, consequently, express the clinical XP phenotype. It could also suggest that our understanding and interpretation of the variants associated with these human diseases is incorrect. Could these genetic variants be a risk factor in some patients with multiple skin cancers with subclinical features of XP?

Methods

From January 1, 2017, to May 4, 2018, we searched the Human Gene Mutation Database Professional⁴ and our cohort at the NIH² to identify all germline mutations with a reported association with a clinically recognized XP phenotype. The patients at the NIH were examined under a protocol approved by the National Cancer Institute Institutional Review Board. All patients and/or parents provided written informed consent.

The Human Gene Mutation Database includes mutations from 7791 genes representing most of the published mutations responsible for human inherited diseases. The Human Gene Mutation Database does not include somatic or mitochondrial mutations. We focused exclusively on the missense and nonsense mutations to estimate the frequency of XP because those mutations cause changes in DNA that result in an amino acid change or a premature stop codon and are likely to be pathogenic. Also, a substantial portion of these mutations had allele frequencies listed in gnomAD.⁵ GnomAD provides allele frequencies of human genomic variants and mutations and includes 123 000 exome sequences from 8 different world subpopulations. We used gnomAD to represent the general population for our determinations.

We investigated the consistency of some of these allele frequencies by analyzing control data from a separate independent exome sequencing database from the National Cancer Institute DCEG database. Data on 998 controls from 2 cohort studies (the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial⁸ and the Cancer Prevention Study of the American Cancer Society⁶) and 1 case-control study (the Environment and Genes in Lung Cancer Etiology⁷) were available for the present study. In addition, we evaluated data from 681 healthy individuals in the Inova Hospital Study.⁹ Controls were cancer free at the time of enrollment. We used the Hardy-Weinberg equation¹⁰ (https://www.nature.com/scitable/definition/hardy-weinberg-equation-299) to make estimates for the prevalence of expected clinical disease: p² + 2pq + q² = 1, where p is the population frequency of the major allele and q is the population frequency of the minor allele. Thus, p + q = 1, and q² represents the frequency of the homozygous minor allele. Because XP is a recessive disorder requiring 2 pathogenic mutations, we focused on estimating the number of affected homozygotes (q²).

Results

Mutations Associated With XP

We identified a total of 156 missense and nonsense mutations associated with XP (Table 1; eTable 1 in the Supplement) in 8 XP genes (XPA-G and POLH) using the Human Gene Mutation Database⁴ and our patients at the NIH.² The XPD (ERCC2) gene had the most mutations associated with XP (32 reported), while XPB (ERCC3) had the least mutations associated with XP (2 reported).

Table 1. Missense and Nonsense Mutations and Allele Frequencies Associated With Xeroderma Pigmentosum (XP) Disease.

Complementation Group (Genes)	No. of Different Mutationsa	No. of Different Mutations With Frequencies Listed in gnomADb	No. of Alleles Detected in gnomADb	% of Alleles Detected in gnomAD (q)c	Estimated Frequency of Affected Individuals in gnomAD (q2)c
XPA (XPA)	16	6	146	0.053	2.83 × 10−7
XPB (ERCC3)	2	2	3	0.001	1.48 × 10−10
XPC (XPC)	27	12	867	0.319	1.02 × 10−5
XPD (ERCC2)	32	16	446	0.162	2.62 × 10−6
XPE (DDB2)	10	0	0	0	0
XPF (ERCC4)	12	6	1267	0.457	2.09 × 10−5
XPG (ERCC5)	26	9	39	0.015	2.40 × 10−8
XP variant (POLH)	31	14	330	0.120	1.44 × 10−6
Total	156	65	3098	1.13	3.55 × 10−5

Abbreviation: gnomAD, Genome Aggregation Database.

^aFrom the Human Gene Mutation Database and the National Institutes of Health cohort (eTable 1 in the Supplement).

^bSee eTable 2 in the Supplement.

^cSee Methods section.

Mutations in Exome Databases Associated With XP

With more than 200 000 sequenced alleles,⁵ gnomAD provided data for 65 of the 156 mutations associated with XP (41.7%) (Table 1; eTable 2 in the Supplement), including allele count, allele frequency, and number of homozygotes harboring these genomic variations. Ninety-one of the 156 mutations (58.3%) associated with XP that were not listed in gnomAD were each presumably less frequent than 1 in 200 000. The 65 mutations associated with XP that were listed in gnomAD had a total allele frequency of 1.13% (Table 1). They included mutations in all of the known genes associated with XP except XPE (DDB2). The 2 XP genes with most frequent mutations were XPF (ERCC4), at 0.457%, and XPC, at 0.319%.

XPF (ERCC4) mutations account for only 1.6% of 189 published cases of XP.¹¹ However, the XPF (ERCC4) gene contained 12 different mutations associated with XP. Seven of these mutations had data reported in gnomAD, including p.P379S and p.R799W, with an allele frequency of 4.05 × 10⁻³ for p.P379S and 4.48 × 10⁻⁴ for p.R799W (Table 2). In addition, gnomAD reported 4 people as homozygous for the XPF p.P379S mutation (Table 2; eTable 2 in the Supplement).

Table 2. Higher Frequency of XPF (ERCC4) and XPC Mutations in Genetic Databases Compared With Phenotypic XP Observed in the United States.

Complementation Group (Gene)	Mutation Associated With XP	Total No. of Alleles Sequenced	No. of Individuals	No. of Alleles	% of Alleles Reported in Database (q)	Estimated % of Homozygous Affected Individuals (q2)a	Total No. of Genetic Homozygotes Reported in Database	Total No. of Genetic Homozygotes Estimated in Databasea	rs No.b
gnomAD
XPF (ERCC4)	p.P379S	276 560	138 280	1122	0.41	1.65 × 10−5	4	2.28	rs1799802
XPF (ERCC4)	p.R799W	277 034	138 517	124	0.04	2.00 × 10−7	0	0.03	rs121913049
XPC	p.P334H	274 914	137 457	838	0.30	9.29 × 10−6	7	1.28	rs74737358
DCEG database (healthy controls)
XPF (ERCC4)	p.P379S	1996	998	10	0.50	2.51 × 10−5	0	0.03	rs1799802
XPF (ERCC4)	p.R799W	1996	998	1	0.05	2.51 × 10−7	0	0	rs121913049
Inova Hospital Study database (healthy controls)
XPF (ERCC4)	p.P379S	1362	681	7c	0.51	2.60 × 10−5	0	0.02	rs1799802
XPF (ERCC4)	p.R799W	1362	681	1	0.07	4.90 × 10−7	0	0	rs121913049
XPC	p.P334H	1362	681	8	0.59	3.48 × 10−5	0	0.02	rs74737358
US estimations
XPF (ERCC4)	p.P379S	NA	323 100 000d	NA	0.41e	1.65 × 10−5	3f	5298g	rs1799802
XPF (ERCC4)	p.R799W	NA	323 100 000d	NA	0.04e	2.00 × 10−7	11f	66g	rs121913049
XPC	p.P334H	NA	323 100 000d	NA	0.30e	9.29 × 10−6	1f	3002g	rs74737358
65 XP mutationsh		NA	NA	NA	1.13e	2.81 × 10−5	300 (US only)f	9007g	NA

Abbreviations: DCEG, Division of Cancer Epidemiology and Genetics; gnomAD, Genome Aggregation Database; NA, not available; XP, xeroderma pigmentosum.

^aCalculated using Hardy-Weinberg equation (see Methods section).

^bFrom the Single Nucleotide Polymorphism database (https://www.ncbi.nlm.nih.gov/projects/SNP/).

^cCalculated from the total number of alleles sequenced and the allele frequencies reported in Inova Hospital Study database.

^dIn US populaton.

^eIn gnomAD.

^fPhenotypic patients with XP reported in the world literature who are homozygous or heterozygous for indicated mutation.

^gTotal number of genetic homozygous patients with XP predicted in United States based on gnomAD.

^hMutations for XP in gnomAD (Table 1; eTable 2 in the Supplement).

We analyzed XPF (ERCC4) p.P379S and p.R799W in 2 independent databases. The frequencies of these mutations in the National Cancer Institute DCEG control population of 998 clinically healthy individuals were 5.0 × 10⁻³ for p. P379S and 5.0 × 10⁻⁴ for p.R799W, which were comparable to the frequencies found in the much larger gnomAD (Table 2). Similarly, the frequencies in the Inova Hospital Study database were 5.1 × 10⁻³ for p. P379S and 7.0 × 10⁻⁴ for p.R799W. No homozygotes were reported in these smaller cohorts.

One of the XPC mutations, p.P334H, had a frequency of 3.05 × 10⁻³ in gnomAD and 5.9 × 10⁻³ in the Inova Hospital Study database (Table 2; eTable 2 in the Supplement). These mutations were identified predominantly in individuals of African descent. All 7 individuals reported as homozygous for this mutation in gnomAD were of African ancestry.

Before estimating the expected number of patients, we tested the Hardy-Weinberg equation using the data reported in gnomAD and the other databases to estimate the number of homozygotes expected to be in each database. We estimated 2.28 XPF p.P379S homozygotes would be present in gnomAD, 0.03 XPF p.P379S homozygotes would be present in the DCEG database, and 0.02 XPF p.P379S homozygotes would be present in the Inova Hospital Study database. The actual reported number of homozygotes was 4 in gnomAD, 0 in the DCEG database, and 0 in the Inova Hospital Study database (Table 2). The comparability of the number of estimated homozygotes and observed homozygotes confirmed the reasonableness of using the Hardy-Weinberg equation to estimate the number of homozygotes present using the US population.

Estimations of XP Mutation Prevalence in the US Population

We used the allele frequencies of XPF p.P379S and the reported population of the United States (3.23 × 10⁸) (http://www.worldometers.info/) to estimate the number of genetically homozygous patients expected to express the XP phenotype. We estimated that there should be approximately 5300 patients who are homozygous for the XPF p.P379S mutation in the United States (Table 2). In contrast, only 3 patients who were homozygous or compound heterozygous for the XPF p.P379S mutation were identified as affected in the world literature (Table 3).^12,13,16 Using a similar approach, Fassihi et al¹² noted that this XPF p.P379S mutation was present at a frequency of 0.3% in the Single Nucleotide Polymorphism database. They state, “this estimate implies that there might be a significant cohort of UVR-sensitive individuals not recognized as having XP, but homozygous for this mutation.”^12(pE1242) We found a similar disparity for the second most common XP associated mutation, XPC p. P334H, which was estimated to be present in about 3000 people in the United States. Only 1 patient with this mutation was reported in the literature (Table 3).^19,20,21,22

Table 3. Clinical Phenotypes of Patients With Most Common XPF (ERCC4) and XPC Mutations.

Patient No./Sex/Age, y	Gene	Mutation 1	Mutation 2	Sun Sensitivity	Abnormal Freckling	Neurologic Degeneration	Skin Cancer	DNA Repair Defect	DNA Repair Assay	Source(s)
XP72BR/M/18	XPF (ERCC4)	p.P379S	Homozygous	Yes	Few lentigines	No	No	Yes	UDS, 35%	12
XP32BR/M/23	XPF (ERCC4)	p.P379S	p.R589W	Yes	No	No	No	Yes	UDS, 18%	12
XP7NE/M/28a	XPF (ERCC4)	p.P379S	Silent	Yes	Yes	NA	No	Yes	UDS, 30%	13
XP126LO/F/22	XPF (ERCC4)	p.R799W	p.T770Pfs*46	Yes	Yes	No	No	Yes	UDS, 13%	14, 15
XP24BR/F/48	XPF (ERCC4)	p.R799W	p.R589W	Yes	Few lentigines	Yes (adult onset)	No	Yes	UDS, 5%	12, 16
XP24KY/ Unknow/50	XPF (ERCC4)	p.R799W	537fs+7bp	Unknown	Unknown	Yes (adult onset)	Unknown	Yes	UDS, 7%	13
XP48DC/F/51 (deceased)	XPF (ERCC4)	p.R799W	p.F196Qfs*20	Yes	Yes	Yes (began at 24 y)	Yes (BCC)	Yes	HCR, reduced	17
C014TA (sibling)/F/51	XPF (ERCC4)	p.R799W	p.S459X	Yes	Yes	Yes (began at 30 y)	No	Yes	HCR, reduced	17
CO107TA (sibling)/M/51 (deceased)	XPF (ERCC4)	p.R799W	p.S459X	Yes	Unknown	Yes (began at 42 y)	No	Yes	HCR, reduced	17
UDP-7356/F/54 (deceased)	XPF (ERCC4)	p.R799W	p.S459X	Yes	Yes	Yes (chorea and ataxia at 34 y)	Yes (>20 BCC began at 16 y)	Yes	UDS, 21%	18
UDP-3675/F/60	XPF (ERCC4)	p.R799W	p.R589W	Yes	Yes	Yes (cognitive impairment at 46 y)	No	Yes	UDS, 26%	18
XP42RO/M/62	XPF (ERCC4)	p.R799W	Homozygous	Yes	Yes	Yes (began at 47 y)	Yes (9 at 27 y)	Yes	UDS, 20%	13, 14
XP62RO/ unknown	XPF (ERCC4)	p.R799W	Homozygous	Unknown	Unknown	Yes (late onset?)	Unknown	Yes	UDS, 20%	13
XP26BR/F/41a	XPF (ERCC4)	p.R799W	L291X	Yes	Yes	Mild	No	Yes	UDS, 15%	13
XP1MI/F/17 (deceased)	XPC	p.P334H	Homozygous	Yes	Yes	No	Yes	Yes	UDS, 29%	19-22

Abbreviations: BCC, basal cell carcinoma; HCR, plasmid host cell reactivation after UV damage; NA, not available; UDS, unscheduled DNA synthesis after UV damage.

^aA. Lehmann, PhD, C. Arlett, PhD, H. Fassihi, MD, written communication, September 2018.

Phenotypes of People With XP Mutations

The 15 patients with XP reported in the literature^{12,13,14,15,16,17,18,19,20,21,22} with at least 1 of 3 relatively frequent mutations associated with XP, XPF p.P379S, XPC p.P334H, and XPF p.R779W, were confirmed as receiving a clinical diagnosis of XP and having reduced DNA repair (Table 3). Their clinical features varied. Thirteen of these patients were reported to have marked sun sensitivity, 11 had abnormal freckling (lentigines), and 4 had skin cancer. Of the 3 patients with XP with the XPF p.P379S mutation, an 18-year-old man (patient XP72BR) was homozygous for p.P379S. He had several episodes of severe sunburn as a child. He had few lentigines, no skin cancer, and no neurologic or ocular abnormalities. His cells had decreased DNA repair (unscheduled DNA synthesis after UV damage [UDS], 35% of normal).¹² A 23-year-old man (patient XP32BR) was reported as being a compound heterozygous with XPF p.P379S and p.R589W. Since infancy, he had severe sunburns with minimal sun exposure, resulting in a diagnosis of XP at 7 years of age. He had a pale complexion and no neurologic abnormalities. His cells had decreased DNA repair (UDS, 18% of normal).¹² The third patient (patient XP7NE) was reported as harboring XPF p.P379S and a silent allele.^13,16 This patient was a 28-year-old man with decreased DNA repair (UDS, 30% of normal), a history of severe sunburns, numerous large freckles on the back of his hands and shoulders, and no skin cancer¹⁶ (A. Lehmann, PhD, C. Arlett, PhD, and H. Fassihi, MD, written communication, September 2018).

XPC p.P334H was the second most common mutation associated with XP. However, there was only 1 patient (patient XP1MI) reported in the literature with this mutation.^19,20,21,22 She was homozygous for the mutation, born with low birth weight and microcephaly and had reduced DNA repair (UDS, 13% of normal). This patient (with cell lines CRL 1333 and GM02096) was reported as a 17-year-old African American girl who had both XP and systemic lupus erythematosus.¹⁹ She was examined at the NIH at 18 years of age, where the diagnoses of both XP and systemic lupus erythematosus were confirmed. She had abnormalities of her skin typical for XP (hypopigmentation and hyperpigmentation), multiple skin cancers, and severe photodamage to her eyes. Her IQ was previously reported as being in the range of persons with intellectual disability (mental retardation), and it was noted that her IQ may have been affected by her limited vision and other conditions. The diagnosis of systemic lupus erythematosus was confirmed, with several swollen joints making it difficult for her to write and walk and with positive blood test results for lupus anticoagulant and antinuclear antibody. This patient died at 37 years of age.

Two of the 11 patients with XP and the XPF p.R799W mutation were reported as homozygous, and 9 were compound heterozygotes.^{12,13,14,15,16,17,18} All had reduced DNA repair. Three of these patients had skin cancer, and 9 had adult-onset severe neurodegeneration (Table 3).

We also found 6 other mutations associated with XP with frequencies greater than 10⁻⁴ in gnomAD (eTable 2 in the Supplement). Two mutations were in the XPA gene (p.R228X,¹² with a frequency of 3.44 × 10⁻⁴, and p.H244R,²³ with a frequency of 1.38 × 10⁻⁴), 2 in the XPD (ERCC2) gene (p.R616P,¹² with a frequency of 1.37 × 10⁻⁴ in gnomAD and a frequency of 5.0 × 10⁻⁴ in the DCEG database, and p.L461V plus p.V716-R730del,^12,24,25 with a frequency of 1.13 × 10⁻³ in gnomAD and a frequency of 2.9 × 10⁻³ in the Inova Hospital Study database), and 2 in the polH gene (for XP variant) (p.K535E,²⁶ with a frequency of 8.43 × 10⁻⁴ in gnomAD and a frequency of 5.0 × 10⁻⁴ in the DCEG database, and p.T692A,²⁷ with a frequency of 1.52 × 10⁻⁴ in gnomAD). These 6 mutations plus the 2 mutations in XPF (ERCC4) and the 1 mutation in the XPC gene have a combined population frequency of about 1% and account for 93% of the frequency of the 65 variant alleles associated with XP in gnomAD (eTable 2 in the Supplement).

Discussion

Extensive work has focused on the “genotype-phenotype correlation,” which usually has meant assessing phenotype and then looking for an underlying genotype. Much of our understanding of the association between clinical disease and underlying genetics has been based on this approach, and it has resulted in an assumption that certain mutations equate with specific clinical manifestations. However, to our knowledge to date, there are few data on the reverse approach: understanding the spectrum of clinical manifestations of variants and how they are associated with penetrance and expressivity, especially when those variants may cause subclinical manifestations of disease. Large genomic databases now allow for the estimation of the frequency of variant alleles. Measuring genotypes in a population and then looking for expected phenotypes have resulted in a surprising discordance: an absence of patients expected to be affected. This finding remains true even with the extensive data processing, variant calling, quality control, and filtering performed on the data in gnomAD.⁵

Our investigation provided unexpected information. The population estimations based on the allele frequencies of the 9 most frequent mutations associated with XP suggest that, based on prevalence of genotype, there should be substantially more cases of XP than have been identified clinically. One of the least common genes reported to cause XP, XPF (ERCC4), had mutations associated with XP with notably high frequencies in 3 independent databases (2 with verified normal populations). The frequency of the XPF p.P379S mutation in gnomAD suggests that there should be about 5300 patients in the United States who are homozygous for this mutation (Table 2). These numbers were estimated using the US population of 323 million. This estimation was vastly contrasted with only 3 patients being identified in the literature as harboring at least 1 XPF (ERCC4) p.P379S mutation (Table 3). Similarly, the XPC p.P334H mutation frequency in gnomAD would estimate that about 3000 individuals in the United States are homozygous for this mutation, and only 1 individual was reported in the literature (Table 2). The combined frequency of the 65 variants associated with XP listed in gnomAD is 1.13% (Table 1; eTable 2 in the Supplement). This frequency would imply that more than 9000 individuals are homozygous for these variants in the United States (Table 2). With the high frequencies of skin cancer in the general population, homozygous carriers of these mutations may represent a subgroup of individuals with a mild or subclinical XP phenotype. If these patients could be identified early, strict sun avoidance and surveillance measures could reduce the morbidity and mortality rates of skin cancer.

These reported clinical and laboratory observations demonstrated the clinically associated pathogenicity of these mutations. However, the clinical features of patients with these reported mutations challenge our approach to recognizing the clinical profile of XP, which currently is strongly associated with sun sensitivity and a 10 000-fold increased risk of skin cancer. Although the 14 patients with XPF (ERCC4) mutations ranged in age from 18 to 62 years, only 3 of them received a diagnosis of skin cancer. Neurologic degeneration in patients with XP with mutations in the XPA or XPD (ERCC2) genes typically has an onset in the first decade of life. In contrast, 9 of 11 patients harboring the XPF (ERCC4) p.R799W mutation had a different phenotype, with adult-onset neurologic degeneration with onset as late as 47 years of age (Table 3). Xeroderma pigmentosum should be considered in patients with sun sensitivity, premature development of many skin cancers (especially of more than 1 type), freckle-like (lentiginous) pigmentary abnormalities or poikilodermatous skin changes, and/or undiagnosed neurologic degeneration.¹

Limitations

There is a marked discrepancy between the data in these genomic databases and the reported clinical prevalence of recessive XP. It is possible that we are not identifying some homozygous patients owing to a mild or varied phenotype, incomplete penetrance of mutated alleles, modifier genes, fetal loss, or late onset of clinical manifestations such as neurodegeneration. The penetrance of XP-type mutations may depend on many factors, including skin type, the rate at which tumorigenic mutations (eg, TP53) accumulate, possible suppressor mutations, compensating levels of gene expression, and levels of patients’ sun exposure. Unfortunately, the DNA samples in gnomAD were collected in a manner that does not permit obtaining clinical information on donors with variations of interest. Mutations in the XPF endonuclease have been reported to result in diverse clinical manifestations, including Cockayne syndrome and Fanconi anemia in addition to XP.²⁸ These mutations might also cause an unidentified clinical phenotype with minimal skin involvement. In addition, we may have underestimated the number of mutations causing XP since we considered only missense and nonsense mutations in a single database of human mutations. Furthermore, we were not able to independently verify the loss of function for many of the mutations listed in the database.

Our understanding and interpretation of the variant(s) may be incorrect. It is possible that variants in other genes modify the phenotype so that the usual features of XP are not apparent. Studies of some dominantly inherited diseases have made similar observations. For autosomal-dominant Li-Fraumeni syndrome, Rana et al²⁹ found different clinical features in TP53-positive patients who were ascertained by multigene panel testing compared with single-gene panel testing. Furthermore, a higher than expected population prevalence of potentially pathogenic germline TP53 variants in individuals unselected for cancer history was reported by de Andrade et al.³⁰ Similarly, Kim et al³¹ reported that likely pathogenic variants of the DICER1 gene (associated with the dominant DICER1 cancer predisposition syndrome) are much more prevalent in the large Exome Aggregation Consortium database using Meta Support Vector Machine, Rare Exome Variant Ensemble Learner, and Combined Annotation Dependent Depletion than by considering only loss of function and previously published pathogenic variants.

Conclusions

It is important to follow up with patients who might possibly have genomic abnormalities to determine whether they will develop any clinical manifestations and what those manifestations might be.³² A new NIH precision medicine initiative study (All of Us [https://allofus.nih.gov/]) has a goal of obtaining clinical information as well as DNA sequences from 1 million volunteers of diverse genetic backgrounds and with differences in lifestyle, environment, and biology. We should approach current large genomic databases with caution when trying to estimate the clinical implications of genetic variants to determine the prevalence of disease risk. This caution is also relevant with prenatal testing and identifying disease-associated mutations that may never be expressed, even with recessive disorders.

Supplement.

eTable 1. 156 XP Disease Associated Missense or Nonsense Mutations in XPA-G genes from HGMD database and NIH cohort

eTable 2. 65 XP Mutations Listed in gnomAD Database