Background

Early onset renal cell carcinoma (eoRCC) in patients under the age of 60 has been increasing in frequency over the past decade [1]. In the United States alone, the most recent analyses report a range of 3.0% annual increase in RCC incidence among individuals aged 45–49 years to as high as a 6.2% increase in incidence among those aged 25–29 years [1]. EoRCC is in some cases linked to pathogenic germline variants (PGVs) in genes associated with RCC familial syndromes (VHL, MET, FLCN, TSC1, TSC2, FH, SDHx, PTEN, BAP1) [1,2,3]; these genes are also often somatically mutated in sporadic RCC cases [2,3,4,5]. Identification of a PGV in defined RCC familial syndrome genes guides clinical recommendations for surveillance, often improving survival due to early diagnosis of eoRCC. However, in recent work we found that only ~ 3.7% of eoRCC patients undergoing cancer risk assessment report a PGV in the currently defined RCC familial syndrome genes [6], reflecting the fact that the majority of eoRCC cases remain genetically not well characterized. Currently, there are no National Comprehensive Cancer Network (NCCN) guidelines for detection, prevention, or risk reduction in individuals who present with an eoRCC but lack a PGV in a familial RCC gene [7].

Recently, we reported that a significant subset of eoRCC patients undergoing cancer risk assessment carry PGVs in DNA damage response and repair genes (~ 8.55% vs. 3.7% in familial RCC genes) [6]. Similarly, Carlo et al. reported an increased prevalence of PGVs in DNA repair genes in advanced clear cell and non-clear cell renal cancer patients [3, 8]. Although PGVs in DNA repair genes are not currently defined by clinical testing guidelines as increasing risk of RCC, these recent studies suggest a potential role of defective DNA repair pathways in eoRCC carcinogenesis that could also lead to novel therapeutic options for RCC patients. Owing to the rising incidence of eoRCC and limited genetic data in younger RCC patients, we performed germline and tumor whole exome sequencing (WES) and functional assays on biospecimens from high-risk eoRCC patients diagnosed before 60 years of age, who were negative for PGVs in familial RCC syndrome genes and had a family history of RCC and/or other familial cancers. Our results suggest that constitutional defects in DNA repair underlie at least a subset of eoRCC cases. Screening patient lymphocytes to identify genotype–phenotype associations via functional assays may provide insight into the mechanism of carcinogenesis for a subset of genetically undiagnosed eoRCCs.

Results

eoRCC patients at the Fox Chase Cancer Center (FCCC) and family history of cancer

We analyzed the personal and family history of the probands in a cohort of 22 eoRCC patients. Multiple probands (6/22, 27%) had a second primary cancer, with breast cancer diagnosed in 3 probands (3/22, 14%) prior to diagnosis of RCC (Fig. 1A-E and Table 1). Here, 73% (n = 16/22) of probands had a family history of RCC, with 50% (n = 11/22) of probands having a first-degree relative with RCC. Intriguingly, 64% (n = 14/22) of probands had a family history of cancers of the prostate, bladder, and thyroid, and melanoma, which have been associated with an RCC diagnosis [9].

Fig. 1
figure 1

Select pedigrees from the eoRCC patient cohort and enrichment of predicted pathogenic variants in DNA repair genes in the cohort. A-E. Pedigrees of eoRCC patients with variants in: A— POLD1 and POLH; B— POLE; C— ATM; D— RRM2B and BCL2L1; E— OGG1, NEIL3 and UBR5. F. Summary of variants in genes and pathways, identified in the cohort. In color—number of variants identified for each gene. For detailed information, see Supplementary tables 1 and 2

Full size image
Table 1 eoRCC patient characteristics, genomic findings, and family history
Full size table

Analysis of whole-exome sequencing data reveals enrichment of germline variation in DNA repair and replication genes in eoRCC patients

We performed WES on lymphocyte DNA from the 22 eoRCC probands, which we analyzed to detect candidate variants in genes included in a candidate gene list (n = 613) that was developed in our prior studies ([10]; see (Supplementary Table 1) and Supplementary Materials and Methods). Here, our intention was to analyze whole exomes of the probands with an expanded list of genes beyond the targeted set of genes on clinical germline panels with the following justification:

  1. a)

    genes involved in genome stability (using Gene Ontology terms such as DNA repair, DNA replication, DNA damage checkpoints, cell cycle, mitotic machinery, replication stress, DNA damage response, chromatin remodeling) would be important for hereditary cancer risk. This is in line with recent work from our group and others in renal cancer [3, 6, 11, 12]. b) an expanded network of genes relevant to renal cell biology (such as cellular metabolism) and genes somatically mutated in RCC that might be relevant for eoRCC-predisposition [5, 13, 14]. For example, there are several genes that are relevant to RCC biology (e.g., PBRM1 [15, 16] SETD2 [15, 17] that are not tested as hereditary risk genes as there is currently no evidence to suggest they impact cancer risk.

Novel candidate variants were stringently defined as those that are predicted to disrupt protein function by the consensus of at least 4 protein predictor algorithms; are rare (gnomAD allele frequency < 0.01); and are nonsynonymous variants (frameshifts, stop gains, and splicing) (see Supplementary Methods for variant prioritization). After applying American College of Medical Genetics (ACMG) criteria [18], we identified candidate variants in 17/22 eoRCC patients in the study, yielding a total of 41 variants in 38 genes (Table 1, and Supplementary Table 2). Gene Ontology analysis confirmed that the candidate variants were enriched in DNA repair and replication pathway genes (Fig. 1F and Supplementary Fig. 1, WebGestalt) [19]. Here, 10 patients (46%; 10/22) had 17 candidate variants in 14 genes currently associated with hereditary cancers across major organ systems (ATM, BRCA2, POLD1, POLE, FH, MITF, MSH3, MUTYH, PDGFRA, RET, SDHB, SMARCA4, SMARCE1, TSC2). Only 4 patients had candidate variants in RCC familial syndrome genes (4/22 – FH, MITF, SDHB, TSC2). Finally, a total of 14 patients (64%; 14/22) had candidate variants from our expanded candidate gene list, from genes not currently defined as RCC-predisposing.

Among the DNA repair-associated genes, candidate variants were found in BRCA2 (Pt #2, Table 1, Supplementary Table 2) and in ATM (2 variants in Pt #14, Table 1, Supplementary Table 2). In addition, 5 candidate variants in DNA replication-repair genes (4/22 patients; 18%, Table 1, Supplementary Table 2, POLD1 and POLH (Pt #1), POLE (Pt #2), POLK (Pt #16), and RRM2B (Pt #4)). Pt #1 had candidate missense variants in PolD1, a catalytic subunit of the replicative DNA polymerase, Pol δ, and in the translesion synthesis DNA polymerase, Pol η. POLD1 G2275A p.V759I is in a highly conserved region of PolD1 subunit of the Pol δ (coded by POLD1 gene) [20] and occurs at a high allele frequency in the Ashkenazi Jewish population (0.0213) reported in gnomAD (versus 0.0018 in the complete gnomAD dataset [21]). POLH G626T (p.G209V) is in the Pol η catalytic core [22, 23]. Pt #2 has a candidate stop-gain G4872A (p.W1624X) variant in the POLE gene, coding replicative DNA polymerase Pol ε, in the conserved C-terminal domain [24]. Pt #4 had a splice site (intronic) variant in RRM2B, coding a subunit of p53-inducible ribonucleotide reductase, which performs de novo conversion of ribonucleotide diphosphates into the corresponding deoxyribonucleotide diphosphates for DNA synthesis [25], in genome position #103,237,248 (chromosome 8q23). Pt #16 had a missense variant in the highly conserved N-terminus domain of another translesion synthesis polymerase, POLK G85A (p.E29K); this variant has been described previously as compromising enzyme activity [26]. ACMG classification and ClinVar evidence are presented in Supplementary Table 2, with most variants currently classified as variants of uncertain significance (VUS).

Primary lymphocytes from eoRCC patients have reduced capacity to suppress DNA double strand breaks (DSBs)

To begin to assess the functional effect of candidate variants in genes linked to DNA replication and repair, we assessed the numbers of γ (phospho)-H2AX foci (a marker of DSBs, [27]) in patient peripheral blood monocytes (PBMCs) at baseline and after treatment with the DNA polymerase inhibitor aphidicolin (Fig. 2A). In PBMCs from both matched cancer-free controls (by age and gender) and eoRCC patients, aphidicolin significantly elevated the number of γH2AX foci; however, aphidicolin-treated cells from eoRCC patients had markedly higher numbers of γH2AX foci than those from similarly treated controls on treatment, indicating reduced DSB repair mechanism in eoRCC patient cells (Fig. 2A, P < 0.001). In complementary work, we tested whether the genes bearing candidate variants were specifically needed to suppress DNA DSBs in RCC cells. For this, we used siRNA to deplete the POLD1, POLE, POLH, POLK, RRM2B, and ATM genes in the Caki RCC cell line. For each gene, knockdown significantly increased γH2AX foci relative to control (Supplementary Fig. 2) further supporting a role for these proteins in DSB repair in renal cells.

Fig. 2
figure 2

Cell-based functional analysis revealed defects in DNA repair and DNA replication in lymphocytes from eoRCC patients. A. γH2AX foci immune fluorescence staining in primary PBMCs from eoRCC patients versus matched controls, at baseline or post treatment with aphidicolin (2 h). PBMCs from patients showed statistically significant elevation of γH2AX foci post treatment with aphidicolin. Data were normalized and are presented as percent of positive γH2AX foci. B. Representative Western blots showing expression of PolD1, Pol η, Pol ε and Pol κ in EBV-transformed cell lines carrying variants versus matched controls (without the variants). Data quantification was performed based on 3 independent biological repeats, technical repeats are presented on gels. C-E. Relative viability of EBV-transformed cell lines was assessed by CTB assay at baseline or after treatment with aphidicolin or UV. Data were normalized to CTB values for controls and are presented as percent cellular viability for POLD1/POLH (C), POLE (D), and POLK (E) cell lines. Data from 3 independent biological repeats are presented. F-G. Difference in DNA replication fork elongation/restoration in EBV-transformed cell lines (F— POLD1, POLH; G— POLE lines) at the baseline and post replications stress was assessed using DNA fiber assay. At baseline the EBV-transformed cells were labeled with IdU for 20 min, for fork restoration cells then were treated with 100 µM aphidicolin or 1 uM MNNG for 2 h, and then labeled with CldU for 40 min. For all conditions, post labeling, cells were lysed, and DNA fibers stretched onto glass-slides, fixed, denatured, blocked, and stained with corresponding antibodies. Fiber images were captured using the Nikon TS2R Inverted Microscope and analyzed in ImageJ software. Data for 3 independent repeats are presented as IdU tract length or CldU/IdU tract length ratio. For all graphs: *** for p < 0.001, ** for p < 0.01, * for p < 0.05 and NS for p > 0.05, unpaired, non-parametric t-test, Mann–Whitney criteria

Full size image

Patient-derived cell lines with candidate PGVs in DNA polymerases exhibit DNA replication defects

We prepared EBV-transformed cell lines from the primary lymphocytes of 3 patients bearing candidate variants (henceforth referred to as the POLD1/POLH cell line, POLE cell line, and POLK cell line) and from several age-and gender-matched cancer-free controls. The POLD1/POLH cell line had significantly reduced levels of the Pol η; for the other candidate variants, the level of the polymerase bearing the candidate variant was not affected (Fig. 2B). Cell Titer Blue (CTB) cellular assays showed significantly better viability than control-derived cell lines when treated with aphidicolin, or with ultraviolet light (which causes bulky adducts in DNA), suggesting that cell lines from patients had better ability to tolerate DNA damage (Figs. 2C-E). Such increased viability in the context of alterations in polymerases has been reported in a number of studies [28,29,30].

Analysis of cell cycle did not show any significant differences in patients and matched control cell lines (Supplementary Fig. 3A).

To further expand on these cell-based findings, we used a DNA fiber assay (see Supplementary Methods) and directly compared DNA replication in patient-derived versus control cells, either untreated or following treatment with aphidicolin or with a DNA-alkylating agent, 1-Methyl-3-nitro-1-nitrosoguanidine (MNNG) (Fig. 2F-G Supplementary Fig. 4A-D). The POLD1/POLH cell line and the POLE cell line exhibited a significantly lower rate of DNA replication in untreated cells (~ 1.4-fold decrease, p < 0.001 for the POLD1/POLH cell line, and ~ 1.9-fold decrease, p < 0.05 for the POLE cell line versus controls). We also observed a significantly lower replication fork recovery after 2 h treatment with aphidicolin (~ 1.44-fold decrease, p < 0.001 for the POLD1/POLH cell line, and ~ 1.88-fold decrease, p < 0.01 for the POLE cell line versus controls) (Fig. 2G-H). Intriguingly, the POLD1/POLH cell line showed defective replication fork restoration (~ 1.2-fold decrease, p < 0.001 versus control line) 2 h post-treatment with MNNG (Fig. 2G). The POLK cell line did not show any defects in DNA replication and replication recovery under the conditions tested (Supplementary Fig. 4A). A complete summary of results for the DNA polymerase variants is provided in Supplementary Table 3.

Altered enzymatic activity of Pol δ and Pol η variant proteins

Among the candidate variants detected in polymerases, the Pol κ variant E29K has previously been biochemically shown to possess not only a significantly reduced catalytic efficiency but also reduced replication fidelity [26]. E29K is in a conserved region of the Pol κ N-terminus, the N-clasp subdomain (1–32 aa), which is essential to maintaining the stability of the open conformation of the Pol κ active site [31]. Intriguingly, a previous study showed that deletion of the first 67 amino acids reduces Pol κ activity during translesion synthesis (TLS, i.e., replication by efficient bypass of bulky lesions in DNA) [32].

To directly test effects of the other candidate variants on polymerase activity, we first purified the polymerase delta (Pol δ) protein complex, with the PolD1 (POLD1), PolD2 (POLD2), PolD3 (POLD3), and PolD4 (POLD4) subunits from recombinant protein co-expressed in E. coli, and with preparations containing either wild type (wt) PolD1 or PolD1 V759I variant (Fig. 3). Both the wt and the variant-containing Pol δ complexes extended a Cy3-labeled DNA primer-template; however, the V579I variant complex had significantly less robust polymerase activity than the wt complex (Fig. 3A, p < 0.001). Furthermore, when Pol δ complexes containing PolD1 wt or PolD1 V759I proteins were mixed in a ratio of 1:1, the appearance of the extended primer-template was significantly more robust than the variant alone but significantly less robust than the wt alone. This result suggests that the variant is not only impaired for function but has a partial dominance over the wt in this assay (Supplementary Fig. 5, p < 0.001).

Fig. 3
figure 3

Structural and biochemical assays revealed altered enzymatic activities of the PolD1 and Pol η variants. A. Pol δ complex and primer extension assay. On the left—representative gel image of purified wt and variant Pol δ protein complexes, containing 4 subunits: PolD1 (125 kDa), PolD2 (50 kDa), PolD3 (66 kDa) and PolD4 (12.5 kDa). Center and right – Pol δ complex primer extension assay with quantification. Representative gel image showing reactions performed with 20 nM Cy-3 labeled DNA-duplex template (SA#1), 20 nM of indicated proteins and 500 uM dNTPs. PolD1 V759I complex extended DNA-template less efficiently comparing to wt protein complex. Data for 3 independent repeats are presented. B. Pol η and primer extension assay. On the left—representative gel image of purified wt and variant Pol η catalytic cores (432 amino acids), molecular weight ~ 56 kDa. Center and right – Pol η primer extension assay with quantification. Representative gel image showing reactions performed with 20 nM Cy-3 labeled DNA-duplex template (SA#1), 20 nM of indicated proteins and 500 uM dNTPs. Data for 3 independent repeats is presented. C. Pol η lesion (8-oxoG) bypass assay with quantification. Representative gel image showing reactions performed with 20 nM Cy-3 labeled DNA-duplex template with 8-oxoG in the position, opposite to 3`-OH group (SA#4), and template of the same sequence without lesion (SA#3), 20 nM of indicated proteins and 500 uM dNTPs. Data for 3 independent repeats are presented. For A-C: *** for p < 0.001, ** for p < 0.01, * for p < 0.05 and NS for p > 0.05, unpaired, non-parametric t-test, Mann–Whitney criteria. All template sequences may be found in Supplementary Table 4. D-E. Homology modeling structures using yeast protein templates for human PolD1 V759I (D) and for human Pol η G209V (E)

Full size image

Pol η is a low fidelity polymerase, which contributes to its ability to perform TLS [33]. Hence, G209V variant and wt Pol η prepared in E. coli were assessed for their ability to extend labeled DNA primer-template duplexes (Figs. 3B-C). In the absence of DNA damage (e.g., in a normally base-paired template), the wt and variant proteins both extended the template (Fig. 3B, p > 0.05), but the observed bands suggest higher processivity for the variant on template without lesions compared to wt (Fig. 3B). To evaluate repair of DNA damage, TLS activity was also tested using a template containing an 8-oxoGuanine (8-oxoG) DNA lesion [33]. Pol η wt bypassed the 8-oxoG lesion robustly compared to the Pol η variant (p < 0.05, Fig. 3C), suggesting better processivity for the wt protein on template with DNA lesion [33].

Finally, biochemical analysis of a purified Pol ε variant, W1624X, was not performed as it is a stopgain variant in the C-terminal domain or CTD, truncating 662 amino acids of the protein. The CTD region is not well-studied, but is thought to be essential for stability of the Pol ε holoenzyme [34].

Structural modeling of DNA polymerase variants in eoRCC suggests impact on polymerase function

The PolD1 V759I variant is located two amino acids away from residue D757 (Fig. 3D, in dark green). In the PolD1 active site, D757 coordinates the Mg2+ ions (neon green spheres) required for DNA synthesis and plays a direct role in the catalytic mechanism and binding of DNA [20]. Structural modeling indicates that a substitution of the valine (V) 759 to isoleucine (I) could plausibly alter the position of D757 and disrupt the efficiency of DNA polymerization. To further understand the structural changes that each polymerase variant might induce, we calculated the change in stability of each amino acid substitution as described in Supplemental Methods and Supplementary Table 4. Interestingly, the PolD1 I759 yielded a mild stabilization (ΔΔG of -1.36 kcal/mole) relative to the wildtype V759. The I759 residue makes twice the number of hydrophobic contacts as the wildtype V759 with the long helix below. The variant could lock this strand in an overly rigid position that compromises DNA polymerization steps that involve flexibility [35, 36], consistent with the biochemical results observed in Fig. 3A.

The Pol η G209 residue is in the catalytic core of the polymerase (Fig. 3E, residues 1–432, colored green) at a position often called the C-cap, i.e. the residue in this position is proximal to the C-terminal end of the α-helix (in orange) [37]. Typically, valine, threonine, and isoleucine residues are not preferred in the C-cap, due to poor solvation at the C-terminus of the helix when the side chains are bulky [37]. Structural modeling of the G209V substitution showed that the valine with a bulkier side chain could not only alter the stability of the α-helix, but also the nearby β-strands of the catalytic active site (in pink). Rosetta modeling shows the variant G209V is capable of making 5 hydrophobic contacts across the cleft with R24, which is just downstream from the key residue D13, which coordinates active site metals that are central to the catalytic mechanism and the binding of the incoming NTP. The predicted change in stability of the Pol η G209V variant revealed a significant destabilization (ΔΔG of 5.67 kcal/mole) relative to the wildtype G209, consistent with lower levels observed in protein (Fig. 2B, Fig. 3B (left gel), and Supplementary Table 4). Thus, a significant destabilization in a relatively rigid region may impact enzyme function, as supported by [35, 36]. Interestingly, the activity of the G209V compares well with wt for a normal primer template and might even be more processive (Fig. 3B, middle gel). However, the variant appears defective for TLS when the template contains an 8oxoG (Fig. 3C). Consideration of these data together shows that changes in stability and conformation may be subtle and even result in an alteration of substrate preference. Thus, a careful combination of both computational and biochemical methods is required to gain a clear understanding of the role a given polymerase variant might have in DNA replication, repair, and possibly cancer initiation and/or progression.

EoRCCs carrying candidate PGVs in DNA polymerases are hypermutant and microsatellite stable (MSS)

To extend these functional tests, we next explored tumor mutation burden (TMB) in tumors from RCC patients in TCGA and from the FCCC eoRCC patients. Previous studies have shown that colorectal and endometrial tumors carrying mutations in POLE exonuclease domain (ExoD) and in POLD1 exhibit a high burden of mutations, are typically MSS but few cases with microsatellite instability (MSI) have been reported, and do not exhibit loss of heterozygosity (LOH) [20, 38,39,40,41,42,43,44,45,46]. RCCs are typically non-hypermutated, with an average TMB of ~ 1 mut/Mb [47]; however, rare hypermutated (≥ 10 mut/Mb) and rarer ultra-hypermutated (> 100 mut/Mb) RCCs carrying polymerase mutations, with or without MSI, have been reported [47]. Analysis of TCGA renal tumor data found that several genes that were mutated in our study (Fig. 4A), including DNA polymerase genes, were somatically mutated in TCGA hypermutant clear cell RCCs (ccRCCs) (Fig. 4A). We analyzed the MSI/MSS and TMB status of tumors from the FCCC eoRCC patients. Both tumors were reported as MSS from clinical testing and were hypermutated (POLD1/POLH tumor- 12.85 mut/Mb, POLE tumor- 14.44 mut/Mb) (Fig. 4B). We analyzed mutational signatures from the POLD1/POLH tumor and the POLE tumor. In single-base substitutions (SBS), signatures SBS5 (clock-like aging signature) was the dominant signature in both tumors; this signature is typically observed in all tissues in the body, normal and tumor, and not associated with polymerase exonuclease defects. Notably, we did not observe SBS signatures (SBS10 subtypes) associated with exonuclease domain mutations, which was expected as the candidate variants in POLE and POLD1 are not within the exonuclease domain. Intriguingly, analysis of doublet-base substitutions (DBS) revealed presence of signature DBS3; DBS3 is typically associated with SBS10 signature but is not always observed in tumors with SBS10 signatures and/or exonuclease domain mutations [48, 49]. We also confirmed that the tumors from the FCCC eoRCC patients did not exhibit LOH of the polymerase genes, as has been observed in other polymerase-mutated tumors [47, 50] (Fig. 4C).

Fig. 4
figure 4

Renal tumors carrying polymerase variants showed high TMB, MSS, and no LOH. A. Percent alteration frequency in 897 tumors from TCGA in different histological types of RCC: chromophobe (n = 66), ccRCC—clear cell renal cell carcinoma (n = 538), ccRCC (hyper)—hypermutated samples (n = 12), papillary (n = 293). B. TMB and MSS data are presented for Pt #1 (POLD1 V759I, POLH G209V) and Pt #2 (POLE W1624X). C. Tumor and normal Sanger sequencing for variants in Pt #1 (POLD1 V759I, POLH G209V) and Pt #2 (POLE W1624X) showing no LOH. Arrows show variants of interest on sequencing tracks

Full size image

To expand the analysis of DNA polymerases in RCC, we next modeled the structural consequences of somatic PolD1 and Pol ε variants in hypermutated ccRCCs from TCGA. Supplementary Table 5 shows the predicted changes in stability for 20 different PolD1 variants and 23 Pol ε variants from hypermutated ccRCCs in TCGA. A broad range of both stabilizing and destabilizing variants was found in all domains of each of these polymerases. Figure 5 shows these variants in relation to known pathogenic variants in POLD1 (Fig. 5A-E) and POLE (Fig. 5F-I). PolD1 R823G/L is in a β-sheet of the polymerase domain close to the DNA (Fig. 5C). The substitution of a positively charged arginine (R) to a hydrophobic glycine (G) or leucine (L) could destabilize the β-sheet and impact DNA binding. PolD1 D893N is positioned close to the DNA and a variant in this region may destabilize DNA binding or position for DNA–protein interactions (Fig. 5D). PolD1 P151S is in a β-sheet in the ExoD and the change from proline (P) to serine (S) could destabilize the β-sheet geometry (Fig. 5E). P151 is close to a known cancer driver mutation [47, 51], E245K, in the unfolded region of the ExoD (Fig. 5A).

Fig. 5
figure 5

Structure mapping of the novel PolD1 and Pol ε variants from hypermutated ccRCCs in TCGA. A-E. DNA-bound PolD1 3D-model was refined from PDB:3IAY. The colored functional domains are exonuclease (light blue, residues 131–477) and polymerase (green, residues 550–978). A. Red spheres represent known cancer drivers. B. Blue spheres represent variants of uncertain significance in ccRCC. C-E. Fragments of PolD1 model showing variants: A810S and R823G (C), D893N, A810S, and R978C (D), P151S and E245K (E). F. DNA-bound N-terminal domain of Pol ε was refined from PDB: 4M8O. The colored functional domains are N-terminal subdomain (dark grey, residues 31–281), exonuclease (wheat, residues 282–527), polymerase (light pink, palm: 528–950; cyan, fingers: 769–833; lime, thumb: 951–1186). Red spheres represent known cancer drivers (structure above). Blue spheres represent variants of uncertain significance in ccRCC (structure below). G. A fragment of Pol ε model showing variant P696L. H. 3D-model of whole-length Pol ε (without DNA). Structure was refined as described in Methods based on [34]. The colored functional domains are N-terminal subdomain (dark grey, residues 31–281), exonuclease (wheat, residues 282–527), polymerase (light pink, palm: 528–950; cyan, fingers: 769–833; lime, thumb: 951–1186), C-terminal domain (light grey, residues 1308–2222). Blue spheres represent variants of uncertain significance in ccRCC. I. Fragments of Pol ε model showing variants S803L and F753

Full size image

Pol ε P696R is in the palm region of the polymerase domain, which is highly conserved among replicative polymerases (Fig. 5G). Arginine (R) has a very large positively charged side chain when compared to smaller proline (P), suggesting this variant may disrupt the polymerase structure and impact DNA synthesis. Pol ε S803L and F753L are in the flexible region of the polymerase domain or the fingers (in cyan, Fig. 5I). This finger region shifts (27° tilt) on DNA binding [52], and thus plays an essential role in polymerase function. S803 is close to the positively charged lysine (K) 431 in the ExoD, and serine (S) is polar and a smaller residue than the hydrophobic leucine (L). Pol ε S803L is near a site of known cancer driving mutations, C810 [47], suggesting that specific alterations in this α-helix could impact polymerase functioning. F753L is on the border of the fingers and the polymerase domain, close to the ExoD and could be important in the coordination of these regions with or without DNA binding. Finally, several variants were found in the C-terminal domain (Fig. 5H, in light grey), which is currently not well-studied, but is known for stabilizing the Pol2 (human Pol ε) complex in yeast [34].

Discussion

In this study, we focus on analysis of candidate variants in DNA repair and replication genes in probands with RCC diagnosed prior to 60 years of age who were undergoing cancer risk assessment at our cancer center and who tested negative for RCC familial syndrome genes. We applied a well-curated pipeline of candidate genes in genome stability, metabolism, metabolic stress, normal renal function, RCC biology, and chromatin remodeling to germline WES data from eoRCC patients. Gene Ontology analysis confirmed that the identified candidate variants were enriched in DNA repair and replication genes. Intriguingly, we found that many eoRCC patients exhibit defects in suppression of DSBs in their primary PBMCs, with PBMCs from eoRCC patients exhibiting higher γH2AX foci than matched cancer-free controls in response to DNA damage. Direct knockdown of some of these candidate variant genes in Caki RCC cell line also led to increased γH2AX foci. Genes with candidate variants were found to be mutated in sporadic RCCs, with specific enrichment of alterations in DNA polymerases (POLE, POLD1, POLH1, and POLK) and BRCA2 in hypermutant RCCs in TCGA. Importantly, detailed analysis of candidate variants in DNA polymerase genes from the FCCC eoRCC patients confirmed the damaging nature of the candidate variants and suggested a mechanistic basis for association of these variants with observed defects in DNA repair and replication. Several PolD1 and Pol ε variants from the hypermutant RCCs in TCGA were proximal to the catalytic center or substrate binding regions and will benefit from similar future biochemistry experiments shown in this study.

This work complements a number of recent studies indicating inherited defects in DNA replication machinery may increase cancer risk. Candidate variants in the ExoD of POLE and POLD1 predispose to cancer and exhibit a strong mutagenic effect, however, the role of non-ExoD variants in mutagenesis and cancer risk has been controversial [39, 41, 53]. A recent study reported that the POLD1 candidate variant (p.V759I, in Pt #1) is frequently present in the Ashkenazi Jewish population and proposed this gene as a founder mutation [2]. Mertz et. al. have demonstrated a strong mutator effect of the PolD1 polymerase domain variant R689W in human cells [54]. Barbari et. al have recently shown that defective proofreading is not the only important determinant of variant pathogenicity; polymerase fitness is also a key factor [55]. Several TLS polymerases including Pol η and Pol κ, are important for preventing accumulation of single strand DNA gaps, and the replication of DNA fragile sites. Owing to their high error-propensity, TLS polymerases are likely to contribute to oncogene-induced mutagenesis [56, 57].

It is likely that the candidate variants in DNA replication and repair genes detected here interact with other germline variants to impact eoRCC risk. In this study, Pt #1 also harbored candidate variants in MTOR, PARP1, FH, MCM2, suggesting the POLD1 and POLH variants assessed here may act together with other variants to augment the DNA repair defects observed. Pt #2 with a POLE variant also harbored candidate variants in BRCA2, PDGFRA. In other studies, hypermutant ccRCCs also carried mutations in TP53, PTEN, VHL, and UNC5C [58, 59]. POLE and POLD1 are currently not considered classical tumor suppressor genes, and LOH is typically not observed in tumors. Increased mutation frequency is observed with a heterozygous mutation in these replicative DNA polymerases, but only the homozygous mice have increased susceptibility to cancer; suggesting that there are other additional factors important for carcinogenesis [38, 39, 41, 60]. It is possible that defects in polymerase genes impact cancer risk by affecting biological processes beyond DNA replication and repair. Conversely, some familial RCC genes (such as FH, VHL, PBMR1, and SDHx) have also been implicated in suppression of DSBs and in replication stress [16, 61,62,63], based on mechanisms that are not well understood.

It is important to note that while the family history of the high-risk probands in this study is suggestive of underlying genetics [64], clinical testing for RCC familial syndrome genes did not yield any actionable PGVs according to current NCCN recommended guidelines. Our results suggest that in the absence of PGVs in RCC familial syndrome genes or phenotypic features of familial RCC or family history of RCC, a comprehensive assessment of general cancer predisposition genes, including DNA repair genes, may be beneficial. RCC may be one type of cancer induced by mutations, rather than the sole type of cancer. This is supported by the pedigree data in this study as multiple probands (27%) had at least one additional primary cancer with breast cancer being the most common additional primary cancer (14%). Here, 64% of probands had an extensive family history of cancers of the prostate, bladder, and thyroid, and melanoma, all of which have previously been associated with RCC diagnosis [9]. In fact, PGVs in DNA repair genes have been reported as risk factors for bladder, skin, thyroid, and prostate cancers [65,66,67,68,69,70]. A recent retrospective analysis of the Swedish Cancer Registry showed that ~ 10% of RCC patients develop another second primary cancer, and this is currently thought to be independent of the primary RCC, suggesting broader cancer predisposition [71], compatible with PGVs in genes affecting DNA repair.

Besides genetic screening, these data suggest the value of functional assessment for the families of individuals with eoRCC. In this study, the majority of eoRCC PBMC biospecimens samples exhibited elevated γH2AX levels and candidate variants in DNA repair genes. Our data supports a potential role of germline variation in DNA repair/replication leading to suboptimal encoding protein activity, and genome instability. Overall, these data suggest that assays of γH2AX foci in normal cells, supporting germline variation in DNA repair/replication genes, could be a potential tool for the identification of individuals with genetically unexplained eoRCC. As these defects could be detected in normal cells, it could lead to the identification of individuals in need of cancer risk assessment. This is especially relevant because case–control studies suggest that an elevated familial RCC risk may be multifactorial, and or due to an interaction of the heritable genetics and the shared environment [64]. It is possible that defective DNA repair in the heterozygous state could be a recessive heritable factor that when combined with other RCC risk factors may jointly increase the risk of eoRCC.

Currently, the therapeutic significance of DNA repair genes is not clinically defined for RCC. Evidence is emerging that PARP inhibitors could be therapeutics of choice in RCCs that may not carry mutations in the classical BRCA genes, but which have other defects in DNA repair, with recent clinical trials assessing the use of PARP inhibitors in RCC [72, 73]. Hence, there is a critical need to not only understand the biological impact of defective DNA repair in renal tissue but to also define risk of RCC due to a germline defect in DNA repair genes. A limitation of this study is a relatively small cohort which is not representative of all individuals with eoRCC (i.e. as the cohort subjects had clinical characteristics and/or family history to support germline testing). The candidate variants identified, and the functional assays suggest a mechanistic basis, but more studies are needed to conclude impact on RCC risk. Further work in a larger and more diverse (by race/ethnicity) patient population is clearly of interest for the future.

Conclusions

The results presented here suggest that constitutional defects in DNA repair such as DNA replication repair underlie a subset of eoRCC cases. Screening patient lymphocytes to identify these defects may provide insight into mechanisms of carcinogenesis in a subset of genetically undefined eoRCCs. Evaluation of DNA repair defects may also provide insight into the cancer initiation mechanisms for subsets of eoRCCs and lay the foundation for targeting DNA repair vulnerabilities in eoRCC.

Materials and methods

eoRCC patient population, and peripheral blood DNA analysis

Case-only eoRCC probands that underwent clinical germline genetic testing between 2010–2016 were included in this study (n = 22). Patients were followed by the Genitourinary Program at the Fox Chase Cancer Center and had undergone evaluation for inherited cancer risk at the FCCC Family Risk Assessment Program (RAP). Blood samples were banked in the FCCC Biosample Repository Facility under broad informed consent for research and deidentified. Each participant had a strong family cancer history as shown in Table 1, with either multiple first-degree or second-degree relatives with RC, RC-associated cancers, or other cancers. The mean age at eoRCC diagnosis was 48 years (range 36–59 years). No pathogenic mutations were identified from sequencing the following RC-specific genes: VHL, MET, FLCN, TSC1, TSC2, FH, SDHx, PTEN and BAP1. The patients reported here were self-reported white, non-Hispanic. Family histories were obtained by trained licensed genetic counselors and verified by attending physicians. All patient data obtained were de-identified and included family history of cancer, genetic test results, personal history of cancer(s), presence of multifocal tumors, cancer subtype/stage. Any de-identified personal or family history information including sex, ethnicity/race, age of cancer diagnosis, tumor histology, history of additional personal cancer, and history of family cancer and types was reported first as summarized data and later as de-identified individual case reports. See Supplementary Methods for more methods.