Analysis of Even a Limited Number of Genes Indicates a Strong Inherited Component in Otherwise Typical Sporadic MDS

Somatic mutations are frequently found in patients with MDS. Germline (GL) alterations are less common, in part due to sequencing panels limited to mutations seen commonly in myeloid malignancies, with further exploration only pursued in patients with a clear indication of a familiar disease. It is common to have GL cases where family history is not as informative as we expect, given unreliable recollection by patients and premature mortality of family members. The recognition of pathogenic GL variants is difficult due to the tremendous numbers of inconsequential SNPs, unclear pathologic significance of most new/rare variants and the lack of paired tumor/GL DNA. GL variants may be affecting risk as i) predisposing factors to seemingly sporadic adult MDS or ii) known "leukemia genes" with variable penetrance or longer latency due to lesser functional impact than canonical defects seen in childhood. In addition, there is an ongoing debate as to the role of heterozygous variants of recessive traits serving as predisposition factors. Either way, GL variants may constitute a non-clonal first hit with a long latency until clinical manifestation.

To determine the percentage of MDS cases whose etiology may be related to GL mutations, we screened for GL alterations in two sets of selected genes, i) genes often affected by somatic mutations in MDS (n=65), ii) genes known to predispose to bone marrow failure, leukemia or other cancers (n=105). We focused on Tier 1 variants only.

We analyzed 766 patients diagnosed with MDS for the presence of GL variants in the previously described set of genes; a rational pipeline was developed. In brief, to determine which alterations were pathogenic, all variants with a VAF <30% were deemed somatic. From there, all variants with a CADD score less than 15 or with a benign annotation in ClinVar were removed. We then removed any remaining variants confirmed to be somatic in COSMIC or those tested as somatic by study of GL DNA. Questionable variants were validated by Sanger and targeted deep sequencing to exclude somatic lesions and technical artifacts. This strategy left a total of 829 alterations. We furthermore classified the variants as Tier 1 and Tier 2 variants. All frameshift/nonsense variants were classified as Tier 1, along with missense variants that were previously described as pathogenic or disease causative. A total of 283 variants were classified as Tier 1, and about half were frameshift/nonsense mutations.

Focusing on genes that are frequently somatically mutated in MDS, GL variants were most commonly in TP53 (n=18), NF1 (n=17), RUNX1 (n=12), CEBPA (n=12), and ETV6 (n=6). GL mutations in these genes have been previously described as predisposition factors to MDS evolution. No Tier 1 variants were found in spliceosomal, cohesion complex genes or TET2 or ASXL1.

DNA repair genes were frequently affected by Tier 1 variants; we found 24 FA, 16 ATR/ATM, and 25 BRCA1/2 variants. Genes associated with mismatch repair were also found to be recurrently mutated with 7 Tier 1 variants in MSH2/6. Most GL alterations in genes of the telomerase complex were classified as Tier 2 alterations, with the exception of a few frameshift alterations in POT1, DKC1, and TERT. We found 18 previously described DDX41 GL alterations, a number of already described CSF3R variants (N=20), and a reoccurring BARD1 nonsense mutation in 7 patients.

Focusing on biallelic lesions of the same genes, GL variants can serve as bona fide ancestral non clonal hits with a less then random clonal succession. For example, multiple mutations, one somatic and one GL, of the same gene can be seen in patients with DDX41, RUNX1, and CSF3R.

When assessing the frequency of GL truncating variants in the general population, we found DNA repair genes to be significantly more frequent in otherwise spontaneous adult MDS patients (p<.001).

Despite the challenge in identifying germline-initatied MDS because of competing mortality risks, genetic factors likely play a role in adult patients with otherwise typical seemingly spontaneous MDS/AML. Some alterations include less penetrant alterations of oncogenic genes while others imply a generalized predisposition consistent with a mutator phenotype, HRD or genomic instability. Given our ability to obtain NGS information about myeloid neoplasms, it is essential to delineate not only somatic but also GL predisposition mutations that contribute to an individual's risk and their impact on genealogy.