Prevalence of pathogenic variants in DNA damage response and repair genes in patients undergoing cancer risk assessment and reporting a personal history of early-onset renal cancer

Pathogenic variants (PVs) in multiple genes are known to increase the risk of early-onset renal cancer (eoRC). However, many eoRC patients lack PVs in RC-specific genes; thus, their genetic risk remains undefined. Here, we determine if PVs in DNA damage response and repair (DDRR) genes are enriched in eoRC patients undergoing cancer risk assessment. Retrospective review of de-identified results from 844 eoRC patients, undergoing testing with a multi-gene panel, for a variety of indications, by Ambry Genetics. PVs in cancer-risk genes were identified in 12.8% of patients—with 3.7% in RC-specific, and 8.55% in DDRR genes. DDRR gene PVs were most commonly identified in CHEK2, BRCA1, BRCA2, and ATM. Among the 2.1% of patients with a BRCA1 or BRCA2 PV, < 50% reported a personal history of hereditary breast or ovarian-associated cancer. No association between age of RC diagnosis and prevalence of PVs in RC-specific or DDRR genes was observed. Additionally, 57.9% patients reported at least one additional cancer; breast cancer being the most common (40.1% of females, 2.5% of males). Multi-gene testing including DDRR genes may provide a more comprehensive risk assessment in eoRC patients. Further validation is needed to characterize the association with eoRC.


Control population in ExAc and gnomAD.
To compare the frequency of DDRR gene PVs found in the study to that in the general population, our results were compared to the Exome Aggregation Consortium (ExAc) dataset of largely unrelated ~ 60,000 whole exome sequencing results, and to the Genome Aggregation database (gnomAD) dataset consisting of ~ 125,000 exomes and ~ 15,000 genomes 27,28 . These datasets are the most commonly used genomic data at the population-level.
ClinVar analysis. ClinVar (https ://www.ncbi.nlm.nih.gov/clinv ar/), a database of medically relevant gene variants, was used to investigate all PVs in this study (retrieved on February 4, 2020). PVs that were not reported in ClinVar were noted as 'not reported'. Associated conditions for each PV were categorized into hereditary cancer predisposing syndrome(s), condition(s) related to renal cancer, and any other condition(s). To further elucidate any PVs related to renal cancer, the search term "renal cancer" was queried, and the results were noted as "associated with ClinVar search term 'Renal Cancer.'".
Statistical analysis. To identify potential correlations between PVs and characteristics such as tumor pathology, additional primary tumor type, and age of diagnosis, genes were combined into pathways/groups of interest, histology's were grouped, and cancer types were grouped. Each individual was categorized as having a variant in one of the genes within the group or no variant in the group. Gene categories were used for comparison of RC diagnosis with a DDRR or an RC-specific gene.
We also tested the hypothesis that different gene groups are associated with age at RC diagnosis. We used the median age of RC diagnosis in the study cohort (48 years), and studied PVs in patients < 48 years or ≥ 48 years of age. To test the association between the presence of PVs, and age of RC diagnosis, two-sided Fisher's exact tests were used, and p-values ≤ 0.05 were considered significant. Odds ratios (OR) were calculated to determine the odds that an outcome had occurred given a particular variant, compared to the odds of the outcome occurring in the absence of that variant in the population tested. Finally, we queried the SEER database for patients under 60 years old with RC to compare the distribution of their clinical characteristics (where available) to those in our study cohort 22 .
Due to the evolving nature of the panels during the course of this study, each version included a different total number of genes, and analysis of each gene is based on the number of individuals whose test included that gene (Table S1). Only data from 491 individuals was considered for comparison of individuals with RC-specific genes compared to those with variants in genes not typically associated with RC, as the other individuals did not have all 49 genes analyzed. For statistical comparisons analyzing correlations between specific genes with various characteristics, all individuals who had been tested for that specific gene were included.
To identify potential correlations between PVs and characteristics such as tumor pathology, additional primary tumor type, and age of diagnosis, RC-specific genes, other cancer-associated genes, and DDRR genes were combined into groups, and histologies were grouped. The categories for comparison of PVs and patient characteristics are as follows: www.nature.com/scientificreports/ requirement to obtain written patient informed consent was waived. A Data Use Agreement, and Materials Transfer Agreement was established between Ambry Genetics and Fox Chase Cancer Center. The FCCC Institutional Review Board (IRB) provided study oversight and approval (protocol number 14831). Ambry Genetics provided de-identified results for the study. All methods were performed in accordance with the relevant guidelines and regulation of the approved study.

Results
Patient characteristics. We first benchmarked the eoRC study cohort to the reported incidence of RC in SEER data for the general US population to provide context. In the study cohort, 40% of cases were between 50-59 years of age, and median age of diagnosis was 48 years. As expected, a higher percentage of RC cases were diagnosed between 20-44 years of age as compared to patients ≤ 60 diagnosed with RC in the general US population (SEER) (35%, versus 21.9%) (Fig. 1A). The study cohort was 67.1% female and 32.9% male (Fig. 1B, Table 1), versus 34.8% female and 65.2% male for the general US population prevalence of RC diagnosed ≤ 60 (Fig. 1B). Race/ethnicities in study cohort were 65.6% Caucasian, 5.8% African American/Black, 5.3% Ashkenazi Jewish, 7.6% Hispanic, 0.5% other, and 5.5% unknown ( Table 1). The tumor pathologies reported varied ( Fig. 1C and Table 1). Clear cell constitutes 44.5% of all RCs in SEER, and was the most commonly reported histology in the eoRC cohort (243/844 = 28.8%). Renal cell (not defined, but likely to predominantly reflect clear cell) was also common (168/844 = 19.9%, Fig. 1C and Table 1). Papillary and chromophobe histology were each identified in ~ 4-5% of the individuals (38/844 = 4.5% and 40/844 = 4.7%, respectively). Other histologies were identified rarely, but included Wilms tumor (19/844 = 2.3%) and oncocytoma (6/844 = 0.7%). For 34.7% of patients, the RC subtype was unknown.
High incidence of other cancers in study cohort. 57.9% (n = 489/844) of the cases in the study cohort reported at least one additional primary cancer (Fig. 1D, Table 1, Table S2). Each of the primary cancer types is also represented at a higher level in the study cohort than in the general US population as reported by the SEER database (Fig. 1D). For female-specific cancers, 40.1% of females (227/566) also had breast cancer, in comparison to the 4.3% breast cancer rate in women ≤ 60 in the general population (SEER) ( Fig. 1D and Table S2). The rate of additional primary cancer in the study cohort (57.9%) is much higher than the rate of each cancer type observed in SEER cases with eoRC (21.6%) (Fig. 1E). Finally, 784 patients out of 844 reported a family history of cancer, and of these 784 patients, 196 (24.7%) specifically reported at least one family member with RC (Table 1).

Multi-gene cancer panel testing identifies PVs in DDRR genes in the study cohort. The most
common gene with PVs identified in the eoRC patients was the DDRR gene CHEK2 (19/844, 2.25%, Fig. 2A, Table S3 and S4), consistent with a recent report by Carlo et al. 16 Of patients with CHEK2 PVs, 47.3% (n = 9/19) had a common, highly damaging variant (c.1100delC, p.Thr367Metfs) that is known to be associated with an increased risk for breast, prostate, colorectal and thyroid cancers (Table S4) [34][35][36][37] .
After CHEK2, PVs were most frequently observed in the DDRR genes BRCA2 (10/815, 1.23%), ATM (9/844, 1.07%) and BRCA1 (7/815, 0.86%) (Table S3). We compared the overall frequency of PVs in CHEK2, BRCA1, BRCA2, and ATM to the control population in ExAc and gnomAD, representing individuals sequenced for disease-specific and population genetic studies 27,28 . Overall, PVs in each of these genes were more common in the study cohort versus the control populations ( Fig. 2B,C, Table S5A). An outlier was the moderate risk CHEK2 c.470T>C p. I157T PV 38 identified in 5 individuals in the study cohort, which was higher in the controls (gno-mAD-OR, 0.60; 95% CI, 0.234-1.433; ExAc-OR, 0.72; 95% CI, 0.282-1.74). We compared the prevalence of all PVs in DDRR genes presented in Table S4, from 844 cases, to controls from gnomAD 23 . We found ~ 4.8-fold enrichment of PVs in DDRR genes in the study cohort versus the controls in gnomAD (8.4% vs. 1.8% respectively, Table S5B; each DDRR gene was corrected for number of patients assessed).
Cancer risk with MUTYH (DDRR gene) has only been defined for individuals with homozygous or compound heterozygous PVs, but not for heterozygous carriers 39 . We identified 17/844 individuals with MUTYH PVs, out of which 16/17 were heterozygous carriers and only 1/17 was compound heterozygous. Only the individual with compound heterozygous MUTYH PVs was counted in the full study cohort (n = 844, Table S3 and Fig. 2A). Similar to MUTYH, cancer risk from the FH (RC-specific gene) c.1431_1433dupAAA, p.K477DUP variant is currently considered to be pathogenic only in the compound heterozygous or homozygous state 40 . We identified 2 RC patients who were heterozygous carriers of this specific FH variant (Tables S3 and S4).
The overall gene variation rate in the full study cohort (n = 844) is presented in Table S3. The full study cohort was not tested for all 49 genes. The largest panel was tested in the sub-cohort of 491 cases, and consisted of 49 genes, which included 15 RC-specific genes, and 34 other-cancer associated genes including 19 DDRR genes (Table S1). Here, 12.8% (63/491) of cases had PVs. PVs were identified in one or more of the 16 genes not typically associated with RC in 9.16% cases (n = 45/491, Table S6), versus 3.7% (n = 18/491) with a PV in RC-specific genes (Fig. 2D, Table S6). Of the 16 genes not typically associated with RC, 12 were in DDRR genes (8.55%, n = 42/491 or 66.7%, n = 42/63). Among the 491 patients, 2 patients were found to have PVs in two genes. One patient had PVs in two DDRR genes (BRCA1 and MUTYH het), and the other patient in a RC-specific gene and a DDRR gene (SDHB and MUTYH het) (Table S4). There was no MUTYH or FH compound heterozygous or homozygous PV in the sub-cohort of 491 cases. , and the remaining cases bore PVs in non-DDRR genes associated with cancers other than RC (Fig. 2E). Next, we performed similar analysis as described above for patients who presented with eoRC plus one or more additional cancers. Among the 261 patients who presented with eoRC and at least one additional cancer, Table 1. Demographics and clinical characteristics of RC patients in the Ambry Genetics study cohort. Demographics and clinical characteristics of the RC cases in the study cohort were compared to those of RC (from birth to age 60) in the SEER data. Personal and family history of cancer were reported for the cases in the study cohort. *For family history of renal cancers, numbers include only those who reported on cancer history (n = 793). nr not reported. SEER data included 149 types of renal cancer histologies, not all were represented in dataset; "other" based on other category from Ambry cohort. Family histories as self-reported on the intake form/medical records and have not been validated. Overall, these data suggest that DDRR gene PVs are enriched similarly in individuals diagnosed with eoRC alone or eoRC plus at least one additional primary cancer, but that the frequency of PVs in DDRR genes, in either group, exceeded that in the control populations tested (gnomAD/ExAc) (Fig. 2, Table S5A). The specific PVs identified were similar in frequency to those identified in the full patient cohort (n = 844), with CHEK2 the most represented DDRR genes (Fig. 2). To gain additional insight into the prevalence of these PVs in cancer patients, we surveyed ClinVar (https ://www.ncbi.nlm.nih.gov/clinv ar/), and found that multiple PVs from this study (Table S4) have been reported in hereditary cancer predisposing syndromes (HCPS, summarized in Table S7). HCPS reflects a pattern of cancers in a family characterized by earlier onset, with individuals not necessarily having the same tumor and/or having more than one primary tumor, and having tumors that are more likely to be multicentric. RC patients with BRCA1 or BRCA2 PVs. Notably, 1.2% (10/815) of the eoRC cases had a BRCA2 PV, and 0.9% (7/815) RC cases had a BRCA1 PV ( Table 2, Table S3). This included 1.7% (n = 6/355, Table 2) of the cases who presented with only eoRC. Interestingly, despite the fact that the cohort was 67.1% female, 47.1% (8/17) of the detected BRCA1 and BRCA2 PVs were in males (

No correlation between age of RC diagnosis and type of PV in RC.
To determine if identification of specific classes of germline PV correlated with age of diagnosis in this cohort, genes were divided into four broad (overlapping) categories: all genes in the panel, RC-specific genes, non-RC genes (including DDRR genes) and DDRR genes (see "Methods"). The groups were compared to median age at first RC diagnosis of < 48 or ≥ 48 years. Given the invariable early-onset of Wilms tumor, the 20 individuals with this diagnosis were excluded from the analysis. Within this eoRC cohort, there was no significant association between age at diagnosis of RC and the type of PV for any of the four broad categories above (Fig. 3A).

Correlation of renal histologies with PVs in specific genes.
Of the 243 clear cell cases in this cohort, 13.6% (33/243) had a PV, of which 2.9% were RC-associated PVs. Similar findings were observed for the cases described as renal cell carcinoma, 13.1% (22/168) had a PV, of which 2.4% were RC-associated. DDRR gene PVs were found in 24/243 (~ 10%) of clear cell cases, and in 16/168 (9.52%) of renal cell cases. Figure 3B,C contrast the findings in clear cell and renal cell histology with the other non-clear cell histologies.

Discussion
This study for the first time demonstrates that PVs in multiple DDRR genes occur in patients with eoRC. Importantly, this study found that DDRR gene PVs were represented both in cases diagnosed with eoRC and additional cancers, and also cases diagnosed with eoRC alone. Comparison with a large control population indicated that germline PVs in DDRR genes were more common in this study cohort than in the control population, although further studies are required to confirm this finding and estimate the penetrance of PVs in DDRR genes for eoRC. We also found that germline testing using an RC-specific panel would have identified only 3.7% (18/491) of the RC cases with actionable PVs according to the NCCN recommended screening or management guidelines, compared to the 9.16% (45/491) additional cases identified with the expanded panels.
The most common gene with PVs identified in the patients in this study was the DDRR gene, CHEK2 (19/844, 2.25%). This is consistent with recent reports by Carlo et al. and Huszno et al. 15,16 . While evidence is mounting that CHEK2 PVs may increase risk for RC, in this study we did not consider CHEK2 as a gene typically associated with RC as it is not currently included on RC panels and would fail to be included in testing in many cases. In addition, limitations of the previous studies and the analysis reported here together indicate that larger studies with appropriate controls are needed before confirming that CHEK2 indeed confers a risk for RC.
Identification of germline DDRR gene PVs can have specific implications for the proband and the family. For example, 1.7% of cases diagnosed with eoRC alone had PVs in BRCA1 or BRCA2, but not all of these cases had a family history strongly indicative of HBOC syndrome. This is important because identification of a BRCA PV can potentially change medical management; for instance, PARP inhibitor therapy is effective in tumors with BRCA PVs, including non-breast tumors 41,42 . Also, screening and prevention of HBOC-syndrome cancers would likely be increased significantly in the proband and in family members found to have the same PV. Further, many of the specific PVs identified in this study have been annotated as relevant to various HCPS, emphasizing their role in the development of multiple cancer types. Our results support broader panel testing as a way to identify unexpected high-penetrant PVs in eoRC patients, when there is a personal or family history of additional cancers (especially an HBOC-syndrome cancer).  www.nature.com/scientificreports/ In RC, a number of germline PVs have been associated with treatment response. For example, bevacizumab with everolimus or erlotinib were added as treatment options for RC cases with germline PVs in FH 8,43 . Currently, the clinical significance of PVs in DDRR genes is not clinically defined for RC. There is an urgent need to study the biological impact of PVs in DDRR genes in renal tissue. Such work may also lead to improved understanding of RC pathogenesis. Studies are in progress to assess cancer risk in different tissue types, and response to treatment due to a germline defect in DDRR genes 44 . A recent study showed that VHL inactivation in RC led to reduced expression of DDRR genes (such as BRCA1 and BRCA2), and thereby increased sensitivity to PARP inhibitors 45 . These results indicate that RC tumors with DDRR gene vulnerabilities may be responsive to PARP or other DDRR gene inhibitors under development. Ongoing clinical trials are assessing the effect of PARP inhibitor, olaparib, in patients with somatic DDRR gene variants in the setting of metastatic RC (NCT03786796). Finally, it is also important to estimate the penetrance of DDRR gene PVs to clinically define RC risk. These studies will assist in genetic counseling of RC patients and their families.
The limitations of this study include the following: this is a relatively small cohort, and not all cases were tested for all 49 genes. The cohort is not representative of all individuals with RC, as the individuals reported in this study likely had clinical characteristics (e.g. high rate of additional primary cancers) or family history that led to the expanded panel testing. In the study cohort, females were overrepresented, even though more males are typically diagnosed with RC (Fig. 1B) 46 . This difference may reflect the observation that women are more likely to pursue genetic testing than men, or the fact that 34.4% of cases also had a diagnosis of breast, ovarian, or uterine cancer. Alternatively, men diagnosed with RC might be considered high-risk due to smoking or other environmental factors that lead their physician to be less suspicious of a hereditary component. A large percentage (34.7%) of tumors from the study cohort was listed as "unknown subtypes", limiting comparison of PVs and RC histology types. Finally, matched (such as age and gender) comparisons were not possible using the large publicaly available control population (ExAC and gnomAD databases), and we made no adjustment for population stratification. Differences in the study cohort, and the large publicly available control population (ExAC and gnomAD databases) in ascertainment strategies and data collection (i.e. bioinformatic pipeline for variant calling/filtering, sequence coverage, race/ethnicities) prevent us from making any conclusions about the relationship between the PVs and RC risk 28,[47][48][49] . The comparisons performed in this manuscript were not adjusted for multiple testing.

conclusions
This study is the first to indicate a role for PVs in multiple DDRR genes in eoRC. These results need to be validated in other large data sets. Additionally, to fully elucidate the biological relevance of DDRR genes to RC, family and functional studies are needed as a next step to quantify the associated risks. Red bars; DDRR genes, yellow bars; other-cancer associated genes; blue bars; RC genes. APC variants identified in this study were all the moderate risk c.3920T>A, p.I1307K variant, and 5 of the 19 CHEK2 variants were the moderate risk c.470T>C, p.I157T variant. *The individual with MUTYH was a compound heterozygote with two PVs. The data is presented as percent rather than 'n' due to the fact that not all 49 genes were tested for all patients in the full study cohort of 844 individuals. The percent adjusts for the number of individuals that were tested for each gene. The 'n' values are listed in Supplemental Table 3. (B,C) Odds of finding PVs in ATM (pink circle), BRCA1 (black circle), BRCA2 (orange circle), and CHEK2 (blue circle) from study cohort versus control population, ExAc (B) and gnomAD (C). Data is presented as log10 odds ratio (OR), and log10 confidence intervals. Dotted black line; association with outcome i.e. OR > 0 is enrichment in study cohort, OR = 0 no difference in cohorts. PVs not found in gnomAD or ExAc are indicated by the absence of any data or PVs not listed from Supplemental Table 4 were not found in the control population. Note: Computation of proportion or burden of individuals with all PVs in a specific gene(s) in the control population cannot be accurately performed as all PVs have not been defined, and while there is some agreement on which variants in a specific gene(s) are currently considered PVs, this is not true for all variants in that gene (as referenced in ClinVar, https ://www. ncbi.nlm.nih.gov/clinv ar/). (D) Cases with germline PVs in the cohort tested for all genes in the study (n = 491, 49 genes). The total 'n' is listed in Supplemental Table 6. (E) Individuals with germline PVs who were diagnosed with only RC (n = 230/491). The total 'n' is shown above the bar as PV per gene. (F) Individuals with germline PVs who were diagnosed with RC plus at least one additional primary cancer type (n = 61/491). The total 'n' shown above the bar as PV per gene. The color scheme as in (A). In (D-F), to remain consistent between graphs, the data is presented as percent rather than 'n' even though all 49 genes were tested for all individuals represented in these graphs. www.nature.com/scientificreports/  The total number of individuals with a PV and percent PVs per gene category is shown above the bar. (B,C) Includes counts from both homozygous and heterozygous carriers of MUTYH, and carriers of a FH variant that is currently considered to be pathogenic only in the compound heterozygous or homozygous state.