Myelodysplastic syndromes (MDS) represent a heterogeneous group of myeloid malignancies characterized by cytopenia and increased risk of progression to acute myeloid leukemia (AML) driven by accumulated somatic mutations in hematopoietic stem cells (HSCs). While recurrent mutations are known to associate with adverse outcomes in myelodysplastic syndrome (MDS), 10% of MDS cases have no known genetic indicators for survival prognosis. Whole genome sequencing (WGS) detects comprehensive mutations on coding and non-coding regions that empowers discovery of novel genetic biomarkers.
We conducted WGS at 60x read depth on 494 MDS patient subjects and scanned genome-wide for novel somatic biomarkers associated with MDS post allogeneic hematopoietic cell transplantation (allo-HCT) outcomes. Forward and backward stepwise variable selection was performed, and important clinical covariables were integrated into multivariate Cox proportional hazard models for survival outcome association tests. To test for associations with overall survival (OS), we conducted burden tests of somatic variants by summing the number of somatic nonsynonymous (missense, nonsense, splice) mutations per gene and testing for association with OS. For non-coding regions, we applied a sliding window approach, summing somatic mutations within each 10kb window and testing for association with OS. Furthermore, to distinguish potentially novel biomarkers from mutations in known MDS prognostic genes, we re-ran our association analyses considering only those MDS subjects with no known mutations in previously-identified prognostic genes for post allo-HCT survival(including TP53, RAS, JAK2, TET2, EZH2, ETV6, RUNX1, DNMT3A and ASXL1).
Genome-wide analyses of somatic coding variants on OS identified two statistically significant associations (Figure 1A I); one at TP53 (hazard ratio [HR], 2.19; 95% confidence intervals [CI] 1.48-2.75; P = 1.51e-06) and the other at HCN2 (HR, 5.84; 95% CI 2.38-14.34; P=1.18E-04). When we restricted our analyses to those patients with no mutations in known prognostic genes (N=301), we found one statistically significant candidate at DDX11 (HR, 3.74; 95% CI 1.87-7.47; P=1.83E-04). When we expanded our analyses to include non-coding somatic variants, we observed five additional significant genes (Figure 1A II, IV) including CHD1 (HR, 7.31; 95% CI 3.33-16.08; P=7.45E-07) that was significantly associated with post allo-HCT OS in the patient subset with no known mutations. Meta-analyses of OS using TCGA lymphoma and AML/MDS data provided validation that DDX11 and CHD1 mutations are associated with poor overall survival in hematopoietic malignancies (data not shown).
Interestingly, many of the candidates from these analyses (e.g., DDX11, CHD1, RASGRF1 and ARHGEF7) are involved in DNA repair pathways. Gene set enrichment analyses confirmed that our association results are enriched a TP53-centered networks (p value: 0.0042, Figure 1B).
Besides the novel candidates, a known MDS prognostic gene TP53 was replicated by our nonbiased screening approach. Co-occurrence of TP53 mutation and complex karyotypes, as well as high VAF of TP53 mutations in our MDS cohort were highly associated with poor post allo-HCT OS (data not shown). MDS patients with TP53, DDX11 and CHD1 mutations were overrepresented in samples with complex karyotypes.
Overall, through WGS of samples obtained at the time of allo-HCT, we identified two novel genetic prognostic biomarkers, DDX11 and CHD1 mutations, which were associated with inferior post allo-HCT OS. These results once validated could contribute to personalized risk assessments of post HCT outcomes for MDS patients.
Saber: Govt. COI: Other.