Genome-First Determination of the Prevalence and Penetrance of Eight Germline Myeloid Malignancy Predisposition Genes: A Study of Two Population-Based Cohorts

Myelodysplastic syndrome and acute myeloid leukemia are most commonly sporadic; however, it is increasingly recognized that up to 10% of individuals with a myeloid malignancy may carry a germline susceptibility. The genome-first approach provides a unique opportunity to quantify the prevalence of pathogenic/likely pathogenic (P/LP) germline variants in myeloid malignancy predisposition genes (gMMP) in large population-based exome sequencing datasets and correlate with cancer development. These genes have both variable penetrance and expressivity and confer increased risk myeloid malignancy (MM).

Genes of interest (ANKRD26, CEBPA, DDX41, ETV6, GATA2, MECOM, RUNX1, SRP72) were analyzed from the Geisinger MyCode DiscovEHR (n=170,503) and the UK Biobank (UKBB, n=469,595) cohorts. Variant curation was restricted to exonic nonsynonymous or splice single nucleotide variants (SNV) and small insertions/deletions with MAF <0.01 and a VAF ≥ 0.35 and ≤ 0.65. Variant classification was done using ClinVar and InterVar followed by manual curation. Phenotypes were analyzed in two groups: all hematological malignancies (HM) and the subgroup of MM.

Overall, there was 1:448 to 1:602 chance of having a gMMP P/LP variant in one of the eight genes. When grouping all genes, prevalence was less than 0.22% across both cohorts, but the penetrance was high, with 1:19 to 1:26 heterozygotes having an HM (5.2-7.8%). GATA2, RUNX1, and ETV6 had low prevalence, but particularly high penetrance with 1:4 heterozygotes having an HM. Conversely, DDX41 showed higher prevalence but lower, albeit elevated, penetrance in both cohorts (prevalence: 1:576 to 1:870; penetrance: 1:19 to 1:39). MECOM and SRP72 heterozygotes were more common than the GATA2, RUNX1, and ETV6 group, yet low penetrance was observed. Despite good sequencing coverage across the region, no ANKRD26 P/LP variants were identified within the 5’ UTR regulatory region, the previously described mechanism of disease for this gene.

Cross-sectional analyses showed an increased risk of MM in both DiscovEHR and UKBB and an increased risk of HM in UKBB across all P/LP heterozygotes compared to non-heterozygotes in the cohort. Specifically, in DiscovEHR, significantly increased risk for developing a MM for all heterozygotes was seen with an adjusted odds ratio (OR) of 4.6 [95%CI 2.1-9.7 p<0.0001]). Additionally for both HM and MM a significantly increased risk was observed for RUNX1, GATA2, and ETV6 heterozygotes on a per gene level and the three genes combined (combined OR 4.0 [95% CI 1.7-9.5, p=0.0016]). In UKBB, a significantly increased risk for developing HM and MM exists for heterozygotes of all genes (HM OR 2.6 [95% CI 2.0-3.5, p=3.6x10^-12], and particularly DDX41, RUNX1, and ETV6 (HM OR 2.7, 66, 11, respectively). Time-dependent Kaplan-Meier survival analyses showed heterozygotes with HM and MM were statistically less likely to survive compared to non-heterozygotes in both cohorts. For GATA2, ETV6, and RUNX1 heterozygotes, there is increased risk and earlier onset of all cancers. For DDX41 heterozygotes in both cohorts, there is a significant increase in all-cause mortality, despite the lower penetrance observed. Cumulative risk for all-cause mortality was determined across nine decades of life in both cohorts, with heterozygotes in DiscovEHR having a 37% (95% CI 23-48) and in UKBB at 25% (95% CI 19-30) probability of dying by age 80.

This study represents the first use of the genome-first approach to comprehensively quantify the prevalence of eight gMMP genes and measure the penetrance of hematological cancer risk. We report that these predispositions are more common that previously appreciated, and these individuals have a significantly increased risk of hematological malignancy and decreased overall survival. Similarities for DDX41 frequency and penetrance values in both DiscovEHR and UKBB provide noteworthy evidence supporting DDX41 as a major driver in myeloid malignancy germline predisposition. Predispositions due to the important hematopoietic transcription factors GATA2, RUNX1 and ETV6, while rare, are highly penetrant for hematological malignancy and have significantly decreased overall survival. The data presented here demonstrate that the genome-first approach effectively determine hereditary myeloid malignancy risk in the general population and provides evidence to better define this increasingly acknowledged group of syndromes.