Investigating suspected germline STAG2 variants in myeloid neoplasms

Abstract BACKGROUND: Among cohesin complex genes, STAG2 is the most frequently mutated, prognostically unfavorable gene in high-risk MDS and AML. While we and others have previously studied somatic (SM) mutations in cohesin genes, the presence of germline (GL) mutations in MDS and AML remains underexplored. GL mutations in cohesin components cause rare developmental disorders known as cohesinopathies, characterized by growth and cognitive impairments, limb anomalies, and craniofacial dysmorphisms. Emerging evidence suggests that cryptic GL predisposition may contribute more significantly to the development of myeloid neoplasms (MN) than previously recognized, even in adults without a family history of cancer or developmental disorder. However, to date, there is no definitive evidence that STAG2 germline variants constitute a bona fide cancer-predisposing gene. METHODS: We have previously described the mutational landscape of STAG2 mutations within our cohort of patients diagnosed with MN (n=1,967). In this cohort we identified patients with single (N=215) and multiple (>1; N=38) STAG2 hits, the latter potentially representing branching clonal evolution and evolutionary pressure conveyed by underlying GL or SM founder mutations. A small subset of patients within our STAG2 MN cohort were found to harbor nonsense variants previously known to be pathogenic via genomic sequencing in patients with cohesinopathies (p.R1012, p.R1033, p.R146). We hypothesized that the presence of these variants might be instructive into their relative effects on the STAG2 protein or provide evidence for a possible previously unknown cancer predisposition syndrome. We assessed clinical phenotypes, population-level frequency, and variant associations using large biobanks (FinnGen, AllofUs, UK Biobank). EMR review provided detailed clinical context. GWAS was performed on UK Biobank using ICD-10 codes for MN and standard variant QC (INFO ≥ 0.9, MAC ≥ 100, MAF ≥ 1%, call rate ≥ 90%). RESULTS: Overall, our cohort identified 8 patients with potential germline STAG2 variants of interest: p.R1012 (n=3), p.R1033 (n=4) and p.R146 (n=1). Patients in this cohort were mostly male and white (5/8). The median age of diagnosis of MN was highest for the p.R1033 group (76.5y) and lowest for the p.R1012 group (68y). Review of medical history and imaging studies did not reveal any findings that may point to features of cohesinopathy. Most patients (5/8) had remote evidence of non-myeloid malignancy including two with a family history of “blood disease”. The most common diagnosis of MN was MDS (5/8) and only two patients progressed to AML. Across the groups, the percentage of bone marrow (BM) blasts at diagnosis varied widely between 0%-22%. Further review of NGS findings showed that majority of patients (6/8) had STAG2 VAF <50% (range of 6.7%-68.7%). Cytogenetics were normal in all except two patients having t (11;14) (q13; q32) and +13+21. In addition, genes involved in regulating gene expressions such as ASXL1 (6/8), TET2 (5/8), and SRSF2 (5/8) emerged as the most frequent co-mutational partners, similar to the pattern observed in the multi-hit cohort. Consistent with prior studies, no patient had NGS findings showing co-mutation with other members of the cohesin complex genes. Assessment of the variants in genomic tools affirmed their pathogenicity but were conflicting on the GL likelihood of these variants. Briefly, COSMIC reported SM propensity while ClinVar and OMIM reported GL and findings in VarSome were mixed. None of these variants were detected in the UK Biobank, FinnGen or All of Us cohorts. Further GWAS of the UK Biobank cohort revealed ultra-rare, high-penetrance germline signals in three genes: ASXL1, DDX41, and DNMT3A (OR>15, INFO = 1.0) but none in STAG2 or any other myeloid-related genes on the X chromosome. CONCLUSIONS: We describe a cohort of MN patients with suspected GL STAG2 variants. Their absence in population-scale GL datasets and VAF <50% in most cases argue against true germline origin. Biochemical validation and functional modeling of “two-hit” STAG2 mutations are underway and will explore mechanisms of bi-allelic loss, hypomorphic-to-amorphic transition, or somatic gene rescue. These findings may offer insight into the molecular consequences of multi-mutant STAG2 and inform future classification of GL predisposition in MN. ACKNOWLEDGEMENT: Our sincere thanks to the participants and researchers of the UK Biobank, AllofUs and FinnGen.