Impact of Pathogenic Germ Line Variants in Adults with Acquired Bone Marrow Failure Syndromes Vs. Myeloid Neoplasia

Background. While childhood leukemia and bone marrow failure disorders always raise suspicions of a familial trait, recognition of germ line (GL) contribution to the pathogenesis of adult presentation of these disorders may be difficult because of the absent and unreliable family history, long disease anticipation and incomplete penetrance due to various factors, including competing mortality and interaction with environmental factors. The search for such GL alterations may involve unbiased whole exome sequencing or genome wide analysis studies approaches or, as we conducted here, rely on identification of GL variants in a rationally selected panel of genes previously involved in leukemia and bone marrow failure (BMF).

Methods. We hypothesize that, in analogy to pediatric disease, pathogenic GL alterations may exist also in a subset of adult patients with BMF and myeloid neoplasia (MN). Such lesions may contribute to the clinical features and the preferential occurrence of somatic hits leading to disease manifestation. We analyzed genomic data of 350 patients with BMF syndromes (AA, PNH, AA/PNH) and a cohort of 2,827 patients with MN (MDS, MDS/MPN, and AML) disorders. Our analysis included variant calling, stringent exclusion of artifacts and selection of GL variants (GLVs). These were then subjected to a bioanalytic pipeline and classified as tier-1, tier-2 and variants of uncertain significance according to their potential pathogenic importance. Tier-1 was defined as structurally predicted pathogenic mutations with known disease association, frameshift/nonsense mutations and highly recurrent missense mutations with low frequency in the general population. Tier-2 was defined as missense mutations with a high general population frequency (≥.01 but <1%). GL confirmation was performed in 33% of all cases. When we compared BMF vs. MN we found 42 vs. 202 Tier-1 GLVs and 309 vs. 962 Tier-2, respectively.

Results. In the BMF cohort, 10% had tier-1 and 44% had tier-2 GLVs. Remarkably, 27 patients (pts) (7%) had concomitant tier-1 and tier-2 mutations in different genes, among those, 5 were biallelic (SBDS, FANCM, BARD1, TERT, PALB2) and the rest were compound heterozygous. When we focused on tier-1 lesions, FA GLVs were most prevalent in this cohort comprising 18/350, including FANCA (n=3), BRCA1 (n=2) BRCA2 (n=2) and 11 other FA gene (affected once). In addition, we found also 60 tier-2 FA gene mutations. Surprisingly, telomerase-associated GLVs fulfilling tier-1 criteria were found only in 7 patients [CTC1 (n=4), TERT, TINF2, and WRAP53 (1 each)] and tier-2 in 18 patients. The rest of the tier-1 GLVs were [NF1 (n=4), SBDS (n=3) SAMD9L (n=2), BARD1, BCOR, BLM, CBLC, CHEK2, MLH1, MRE11A, XRCC3 (1pt each). Further analysis of the BMF cohort assessing disease outcomes revealed that among 40 pts, who progressed to MDS, 25% had tier-1 and 27% had only tier-2 GLVs. Reverse analysis showed that 27% of carriers of tier-1 GLVs progressed to MDS/AML as compared to only 10% of negative cases (P<.01), but no difference was found in terms of the frequency of tier-2 variants. Exclusion of patients with PNH clone did not enrich for the carrier status of GL tier-1/2 variants. No specific variants were associated with the PNH clone or the highly expanded PNH clone (manifest PNH). Adult MN was used as a comparative cohort: 7% of cases had tier-1 GLVs, while tier-2 were present in 25% of cases while 3% had concomitant tier-1/-2 GLVs. Patients with tier-1 GLVs were on average younger (median 65 years vs. 69 in those without any GLVs, P=0.008). Tier-1 variant distribution per gene family was as follows; 57/202 in FA, 45/202 in MN-predisposing genes (e.g. RUNX1, DDX41, CEBPA, SAMD9, SAMD9L), 27/202 DNA damage repair genes (e.g. ATM, RAD50, MSH2) and 26/202 GLVs in various other cancer predisposing genes (e.g. APC or CUX1). Biallelic GL hits were found in e.g. CHEK2, FANCI, FANCL. We further analyzed the impact of GLVs on the somatic genotype of evolving leukemias. For instance, biallelic GL/somatic variants were common for some genes (e.g. DDX41, SAMD9, SAMD9L, CEBPA, TP53) while other correlated with functionally seeming unrelated mutations.

In summary. Our analysis indicates that a significant fraction of adult patients with BMF are carriers of predisposition variants with broad clinical and social implications.