Role of POLE and POLD1 in familial cancer

Purpose Germline pathogenic variants in the exonuclease domain (ED) of polymerases POLE and POLD1 predispose to adenomatous polyps, colorectal cancer (CRC), endometrial tumors, and other malignancies, and exhibit increased mutation rate and highly specific associated mutational signatures. The tumor spectrum and prevalence of POLE and POLD1 variants in hereditary cancer are evaluated in this study. Methods POLE and POLD1 were sequenced in 2813 unrelated probands referred for genetic counseling (2309 hereditary cancer patients subjected to a multigene panel, and 504 patients selected based on phenotypic characteristics). Cosegregation and case–control studies, yeast-based functional assays, and tumor mutational analyses were performed for variant interpretation. Results Twelve ED missense variants, 6 loss-of-function, and 23 outside-ED predicted-deleterious missense variants, all with population allele frequencies <1%, were identified. One ED variant (POLE p.Met294Arg) was classified as likely pathogenic, four as likely benign, and seven as variants of unknown significance. The most commonly associated tumor types were colorectal, endometrial and ovarian cancers. Loss-of-function and outside-ED variants are likely not pathogenic for this syndrome. Conclusions Polymerase proofreading–associated syndrome constitutes 0.1–0.4% of familial cancer cases, reaching 0.3–0.7% when only CRC and polyposis are considered. ED variant interpretation is challenging and should include multiple pieces of evidence.


Tumor mutational analysis
Whole exome sequencing of DNA extracted from the tumors and the corresponding blood samples was carried out in a Hi-Seq2000 (Illumina) with a mean coverage per sample of >70x for FFPE and >60X for fresh tumor tissue, after library preparation using the Agilent Sure Select Human All Exon v5 kit. Sequencing was performed at Centro Nacional de Análisis Genómico (CNAG, Barcelona, Spain). After quality control assessment using FastQC software, WES raw data were pre-processed with Trim Galore! (v0.4.0) for adapters and bad quality reads removal. Sequence alignment against reference human genome (hg19/GRCh37) was carried out with BWA (v.0.7.15) 2 and alignments were processed for duplicates' removal and base recalibration according to GATK best practices (https://gatk.broadinstitute.org). Variants were called with MuTect2 following the GATK4 pipeline. 3 Variants identified in the patient's blood DNA were eliminated for the analysis of somatic mutations in the tumor. Variants present in at least 10 reads, minimum Phred quality of 20, and minimum 10% of tumor allele frequency, were considered for subsequent analyses. The contribution of COSMIC mutational signatures 4,5 was calculated with DeconstructSigs 6 using the R package. Total mutation burden was estimated by considering single nucleotide variants (SNV) from exonic regions. Tumor samples developed by carriers of the pathogenic POLE p.L424V and POLD1 p.D316G and p.D316H variants, previously identified by our group, 7,8 were used as ED mutation-positive controls. The absence of somatic POLE and POLD1 mutations was confirmed in all tumors. TCGA and COSMIC tumor sequencing data from samples harboring somatic POLE/POLD1 ED variants and affecting the same amino acids as the germline variants identified in this study, were also analyzed following the same workflow.

Case-control study
The frequencies of rare variants in POLE and POLD1 were assessed in a population-based multi case-control study (MCC-Spain, 9 https://www.mccspain.org/). Cases and controls, all of them from Spain, had been genotyped with the Illumina Infinium Human Exome Bead Chip array, which includes rare variants in coding regions identified in the 1000 Genomes Project. For the purpose of our study, we analyzed genotyping data from the CRC case-control series, which includes 1,336 CRC patients and 2,744 cancer-free controls, and from the breast cancer case-control study, which includes 1,138 breast cancer patients and 1,240 controls. Fisher's exact test was used to compare genotype frequencies between cases and controls.

Classification of germline POLE and POLD1 exonuclease domain missense mutations
The following specific pieces of evidence were incorporated to the ACMG/AMP guidelines for the classification of POLE and POLD1 ED variants (shown in detailed in Supplementary Table S3). PM1 was used if the affected residue is located at the exonuclease domain and within the DNA binding cleft. REVEL in silico modeling program was used to predict pathogenicity, being PP3 evidence applied when REVEL score was ≥0.35, and BP4 if REVEL score <0.30. PS3_supporting or moderate was applied, when a mutator phenotype in S. pombe was observed at moderate (++, p=0.01-0.001) or high levels (+++, p<0.001), respectively. BS3_supporting was considered when no mutator effect was observed in yeast (p>0.05). For cases with WES data available from a carrier's tumor, mutation burden and mutational signature was applied as PP4 (following the recommendations from Walsh et al. 10 ), always in absence of a somatic ED (suspected) pathogenic variant. For variants without sequencing data from carriers' tumors, TCGA/COSMIC sequencing data from tumors harboring the same ED variant was used to obtain mutation burden and signature data, also making sure that no other POLE/D1 ED (suspected) pathogenic variant was present in the analyzed tumor. PP4_strong was applied when the tumor showed hyper/ultramutation and POLE/D1-associated mutational signature 10. PP4_moderate was applied for MSI tumors with hyper/ultramutation and POLE/D1-associated mutational signatures 14 or 20 (depending on the affected polymerase). In contrast, benign evidence BP5 (variant found in a case with an alternate molecular basis for disease) was considered when an MMR-proficient tumor was not hyper/ultramutated or when a DNA repair-proficient tumor was hypermutated but did not display mutational signatures 10, 14 or 20. This BP5 was directly applied when a tumor harboring the somatic variant was analyzed, however, when the analysis involved tumor(s) developed by germline variant carriers, BP5 was applied when the conditions were fulfilled by at least two tumors, in order to minimize the possibility of a false classification due to the presence of a phenocopy.
Additional supporting benign evidence BS1 (allele frequency greater than expected for disorder) was considered if MAF>0.1% in any gnomAD v2.1.1 non-cancer sub-population, and BS2_supporting, when the variant was observed in ≥10 healthy adult individuals (above 60 years of age) for a dominant heterozygous disorder (not applied when BS1 was considered). Table S1. Phenotypic description of the cohorts included in the study. None of the patients carried germline pathogenic variants in the known high-penetrance cancer genes associated with the patient and/or family's phenotype; i.e., breast and/or ovarian cancer patients had no germline pathogenic variants in BRCA1 or BRCA2; CRC or associated tumors (endometrial, gastric, small intestine) were MMR proficient (no Lynch syndrome); and Li-Fraumeni-suspected patients did not carry germline pathogenic variants in TP53.

Supplementary
a Families/patients fulfilling the classical criteria for HNPCC (Amsterdam or Bethesda) were not included in this cohort. b The polyps were either adenomas or hyperplastic polyps. Classic/attenuated colonic adenomatous polyposis had been previously studied (Bellido et al. 7 ) and were not included in this study.   (p), family from the prospective cohort; (r), family from the retrospective cohort; HG-TA, tubular colorectal adenoma with high-grade dysplasia.

Supplementary
Supplementary Figure S3. Pedigrees of the families carrying POLD1 exonuclease domain germline variants. LG-TVA, tubulovillous colorectal adenoma with low-grade dysplasia.