For this retrospective cohort study, we collected 527 patients with uterine malignancies diagnosed at nine Czech health care centers (General University Hospital in Prague, Masaryk Memorial Cancer Institute, AGEL Laboratories, Gennet, GHC Genetics, University Hospital Pilsen, Pronatal, Palacky University Olomouc) and the Bank of Clinical Samples (First Faculty of Medicine). The full list of all participating institutions is provided in the Table SI. Patients were enrolled between 2011–2021 and were Caucasians of the Czech origin. The clinicopathological characteristics (Table I) revealed that endometrial cancers (EC; 89.7%) were the dominant type of collected uterine malignances, therefore the whole cohort of patients with uterine malignancies will be hereafter referred to as ‘EC patients’. Deficient MMR, microsatellite instability and MLH1 hypermethylation statuses were not available. We divided patients according to national indication criteria for germline genetic testing of LS and/or HBOC patients:

Breast cancer or ovarian cancer (C50/C56)-national indication criteria for germline genetic testing [HBOC criteria; cancer diagnoses (C##) correspond to the International Classification of Diseases 10; available at https://icd.who.int/browse10/2019/en#/C00-C97]. Personal history: i) patient is diagnosed with C50 <45 years or <50 years, if family history is unknown; ii) patient has bilateral C50 with the age of diagnosis of the first one <50 years and of both <60 years; iii) patient is diagnosed with triple negative C50 ≤60 years; iv) patient is a male diagnosed with C50; v) patient is diagnosed with either C56, C57 or C48; vi) patient has a duplicity od C50 and C25 regardless of age. Family history: i) patient and two relatives are diagnosed with C50; ii) patient and one relative are diagnosed with C50 <50 years or both C50 <60 years (patient included); iii) patient and a direct relative (parent, sibling, child, alternatively mother or father's sister) are diagnosed with either ovarian cancer, fallopian tube or primary peritoneal tumor, triple negative C50/medullar C50, male relative diagnosed with C50, pancreatic cancer, prostate cancer with Gleason score ≥7 or primary metastatic C61.

Colorectal cancer or EC-national indication criteria for germline genetic testing (LS criteria): i) Age of diagnosis <50 years; ii) proven microsatellite instability <60 years; iii) patient has a concurrent diagnosis linked to LS (colorectal cancer, stomach cancer, pancreatic cancer, ovarian cancer, small intestine cancer, ureter cancer, renal pelvis cancer, bile tract cancer, glioblastoma); iv) patient and one first degree relative have diagnoses linked to LS <50 years; v) patient and two second degree relatives have diagnoses linked to LS regardless of the age of diagnosis; and vi) patients with colorectal cancer and more than ten adenomas/polyps.

Of all patients 151/527 (28.7%) met only LS genetic testing criteria, 16/527 (3.0%) met only HBOC criteria, and 82/527 (15.6%) met both these criteria. A total of 278/527 (52.7%) patients would not be indicated for germline genetic testing.

The study was approved by Ethics Committees of participating institutions. Written consent for the research analysis was obtained from all participants. Clinicopathological information was collected during genetic counselling or retrieved from patients' record.

Two sets of population-matched controls (PMC) were used for comparisons with analyzed EC patients. First, used as a reference for genetic variant prioritization, included 777 non-cancer volunteers aged >60 years that were analyzed identically with EC patients as described previously (19). Second group, used in case-control analyses, included 1662 PMC analyzed as described previously (20). Briefly, the unselected controls (1,170 males and 492 females; median age 57 years, range 18–88 years) were unrelated individuals analyzed by whole-exome sequencing by National Center for Medical Genomics (https://ncmg.cz/) for various noncancer conditions.

Genomic DNA was isolated from peripheral blood collected at the time of enrollment in each respective center. DNA samples were analyzed by NGS using a custom-designed CZECANCA panel as described previously (21) with minor modifications reflecting recent technological updates. These modifications included a new probe synthesis HyperDesign (Roche) improving target coverage for all 226 genes (the sequence capture panel development is shown in detail on the panel web page: http://www.czecanca.cz/eng/panel.html, and full list of genes targeted in this project is described in Table SII). Further modifications included usage of cheaper and faster enzymatic fragmentation replacing ultrasound DNA fragmentation, preparation of DNA libraries using recently introduced KAPA HyperPlus Library Preparation kit (Roche; according to the manufacturer's instruction) and Illumina NextSeq500 sequencing. Resulting NGS data were processed by an in-house bioinformatics pipeline as we described previously (21). Briefly, SAM files were generated from FASTQ using NovoAlign v2.08.03 (http://www.novocraft.com/products/novoalign/) and transformed into BAM by Picard tools v1.129 (https://broadinstitute.github.io/picard/). The Genome Analysis Toolkit v3.8.1 (https://software.broadinstitute.org/gatk/) (22) was used to prepare variant-call format, annotated by SnpEff v4.3 (http://pcingola.github.io/SnpEff/). Identification of medium size indels was performed by Pindel v0.2.5a7 (http://gmt.genome.wustl.edu/packages/pindel/) and copy number variations (CNV) were detected using CNV kit v0.7.4 (https://pypi.python.org/pypi/CNVkit).

All 226 analyzed genes were divided into 19 known EC-predisposition genes described by NCCN guidelines or reviewed by Spurdle et al (5) and 207 other ‘candidate’ genes. Five genes associated with LS (MLH1, MSH2, MSH6, PMS2, EPCAM) and 14 genes associated with HBOC (ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CHEK2, NF1, PALB2, PTEN, RAD51C, RAD51D, STK11, TP53) were considered as the EC-predisposition genes. The remaining 207 candidate cancer-susceptibility genes included those that have been episodically associated with EC predisposition (incl. APC, MUTYH, NBN, POLD1, POLE; Table SII) (5).

Genetic variants found in patients were filtered, excluding variants: i) with low sequencing quality (q<150); ii) with a high minor allele frequency (MAF >0.001) in population databases (gnomAD https://gnomad.broadinstitute.org/, Exome Sequencing Project https://evs.gs.washington.edu/EVS/, 1000 Genomes Project http://www.internationalgenome.org/) (23–25) unless classified as pathogenic/likely pathogenic (P/LP) in the ClinVar database (https://www.ncbi.nlm.nih.gov/clinvar/) (26); iii) present with frequency higher than 0.5% in a group of 777 PMC, except for variants with P/LP ClinVar classification; iv) in untranslated region, intronic outside of consensus splice sites, synonymous and insertion/deletions not resulting in a frameshift unless classified as pathogenic/likely pathogenic (P/LP) in the ClinVar database; v) classified as benign/likely benign in ClinVar with at least two-star rating; vi) low risk variants in BRCA2 (c.9976A>T; p.Lys3326Ter) and in CHEK2 (c.470T>C; p.Ile157Thr).

Resulting set of variants was evaluated according to the ACMG (American College of Medical Genetics) recommendations (27). Variants mentioned in ClinVar as a single submitter or with a conflicting interpretation of pathogenicity were categorized as variants of uncertain significance (VUS). Whole gene duplication and truncating variants localized in the last exon were considered VUS, unless they were classified as P/LP in ClinVar. All PV were inspected in Integrative Genomics Viewer or confirmed using Sanger sequencing or multiplex ligation-dependent probe amplification analysis (MRC Holland). Confirmed PV were submitted to ClinVar database.

The frequencies of PV in EC patients were compared with the frequencies of PV in a group of 1662 unselected PMC. Odds ratios (OR) with 95% confidence intervals (CI) were calculated for EC patients carrying found germline PV using 2×2 contingency table. The χ² or Fisher's exact tests were used for the calculation of P-values (considered significant when P<0.05). Differences in age at diagnosis were analyzed by one-way ANOVA followed by Tukey-Kramer's test. Statistical analysis was performed using the R language v4.1.

We performed germline genetic testing in 527 Czech EC patients including 249 individuals fulfilling LS, HBOC, LS + HBOC indication criteria and 278 individuals not fulfilling any criteria for germline genetic testing. Germline PV were significantly more frequent in patients (118/527; 22.4%) than in population-matched controls (290/1662; 17.4%; P=0.011).

PV were found in 12 (out of 19 tested) EC-predisposition genes (Table II). Frequency of these variants was more than four-times higher in EC patients (60/527; 11.4%; Table SIII) than in PMC (46/1662; 2.8%; P=9.7×10⁻¹⁶). PV in LS genes were found in 27/527 (5.1%) patients (half of them were MSH6 PV carriers) and in 4/1662 (0.25%) controls and they represented the strongest genetic risk factor for EC development (OR=22.4, P=1.8×10⁻¹⁷). Interestingly, no PMS2 PV were observed among patients. BRCA1, BRCA2 and CHEK2 were the most frequently mutated HBOC genes, their PVs conferred significantly increased EC risk for female carriers. However, this risk was lower in comparison to LS genes (ranging from high EC risk in BRCA2 to moderate EC risk in BRCA1 and CHEK2, respectively). PV in the remaining 11 HBOC genes were not identified or did not differ significantly from PMC (Table II). Two carriers harbored coincidental mutations in MLH1/BRCA1 and MSH2/ATM, respectively.

Among EC patients indicated for germline genetic testing according to the above-mentioned criteria, the proportions of PV carriers fulfilling criteria for LS, HBOC, and both conditions were similar (16.6, 18.8, and 18.3%, respectively). These proportions were approximately three-times higher than in EC patients not fulfilling any criteria for germline genetic testing (6.1%; Fig. 1). As expected, the highest proportion of PV in LS genes (11.3%) was detected in a subgroup of patients fulfilling criteria only for LS testing. Similarly, patients meeting solely the HBOC testing criteria had the highest frequency (18.8%) of PV in HBOC genes. Even though the overall percentage of PV carriers differed between subgroups of patients meeting both LS + HBOC genetic testing criteria (18.3%) and not fulfilling any criteria (6.1%), the ratio of carriers of PV in LS:HBOC genes in these two subgroups was similar (5:11 vs. 5:12; Table II, Fig. 1). On the other hand, highly penetrant genes (MLH1, MSH2, BRCA1) were predominantly affected in the subgroup fulfilling both criteria, whereas the subgroup of non-indicated patients was characterized by PV in less penetrant genes (MHS6, ATM).

Moreover, among non-indicated patients we found 2 PVs in HBOC genes in subset of 41 patients with double primary EC and breast cancer (BC; 1×ATM, 1×BRCA1; 2/41; 4.9%) and 3 PVs in HBOC genes in subset of 31 patients with EC and BC in family cancer history (2×BRCA2, 1×CHEK2; 3/31; 9.7%).

The overall prevalence of PV in remaining candidate genes (identified in 48 out of 207 genes) was significantly higher in EC patients (66/527; 12.5%) compared to controls (139/1662; 8.4%; P=0.004; Table SIV). Eight EC patients (and no PMC) carried a coincidental PV in EC-predisposition and candidate genes. Excluding all 60 carriers of PV in EC-predisposition genes, the frequency of PV carriers in other candidate genes was still significantly higher in 467 EC patients (N=58; 12.4%) in comparison to 1616 PMC (N=139; 8.6%; P=0.01). The most frequent PV were found in MUTYH (monoallelic PV in 5/467, 1.1%) and FANCA (4/467; 0.8%). Their frequencies, however, did not differ from that in PMC (MUTYH−18/1616, 1.1%; FANCA−10/1616, 0.6%).

Interestingly, three patients carried germline truncating variant in the genes coding for DNA polymerases (two in POLE and one in POLD1) that have been linked to EC-predisposition previously (5). In contrast, only one POLE and no POLD1 mutation was detected in PMC. Thus, the overall frequency of PV in DNA polymerases was significantly higher in EC-predisposition gene negative patients (3/467; 0.6%) than in PMC (1/1616; 0.06%; OR=10.44; 95% CI 1.08-100.51; P=0.012).

Regarding subgroups of patients based on indication criteria for genetic testing, the frequency of PV in candidate predisposition genes (after excluding the carriers of PV in EC-predisposition genes) was significantly higher in patients fulfilling both indication criteria for LS + HBOC (14/67; 20.9%) in comparison to subgroup of patients fulfilling no genetic testing criteria (28/261; 10.7%; P=0.04, Table SIV). The frequencies of PV in patients meeting indication criteria for LS only and HBOC only did not differ significantly (14/126; 11.1 and 2/13; 15.4%, respectively).

The median age at EC onset was significantly lower only in patients with PV in LS genes compared to non-carriers (51.0 vs. 61.4 years, P=0.01, Fig. 2A).

Concerning the histology subtypes (Fig. 2C), the overall frequency of PV in EC-predisposition genes was similar in patients with endometrial carcinoma to those with sarcoma subtypes (39/349, 11.2% and 4/40, 10.0%; respectively); however, no carrier of PV in LS gene was diagnosed with sarcoma. Interestingly, two out of eight patients diagnosed with precancerous EIN (endometrial intraepithelial neoplasia) carried a PV in BRCA1. Unfortunately, the histologic subtypes of endometrial carcinomas other than endometrioid were rarely represented, thus the frequencies of PVs in these subgroups cannot be calculated and compared.

Analysis of patients with second primary tumors (Fig. 2B) revealed that the highest frequency of PV in EC-predisposition genes was found in patients with 3 primary tumors and in patients with second primary colorectal cancer (CRC). The proportion between carriers of PV in LS and HBOC genes respected the corresponding indication criteria: the carriers of LS gene variants were enriched in patients with EC + CRC and 3 primary tumors. Accordingly, all 13 patients with 3 primary tumors developed either CRC (N=5) and/or ovarian cancer (OC; N=10). The carriers of PV in HBOC genes were more frequent in patients with EC + OC and EC + BC.

When considering family cancer history (Fig. 2D), the highest frequency (reaching 40%) of PV in EC-predisposition genes were found in small subgroups of patients with family history of multiple primary tumors and family history of ovarian tumors. Not surprisingly, predominant tumor types in a family were in concordance with the elevated frequencies of PV in LS or HBOC genes.

The prevalence of carriers of PV in candidate predisposition genes did not differ from that of non-carriers in any of the clinicopathological categories.

The information about immunohistochemistry and microsatellite instability in EC tumor specimens was unavailable.