Longitudinal Tracking of MDS Patients Using Next Generation Sequencing Provides a Predictive Measure for Azacitidine Response and AML Progression

Introduction:

Myelodysplastic syndromes (MDS) are clonal hematopoietic disorders characterized by dysplastic changes in one or more cellular lineages causing impaired bone marrow function. One third of patients diagnosed with MDS progress to secondary acute myeloid leukemia (sAML). These patients have significantly worse prognoses than de novo AML patients. Azacitidine (AZA), a hypomethylating agent is commonly used to treat MDS patients as a frontline therapy. Although its survival benefits over supportive care in a randomized trial has been demonstrated, the underlying genetics and clonal dynamics upon AZA response/AML progression have not been well examined. Using next generation sequencing (NGS) technology, we attempted to assess the clinical relevance of somatic mutations and their dynamics as they relate to AZA treatment in MDS patients using longitudinal samples.

Patients and Methods:

Ninety-five MDS patients (56 lower risk and 39 higher risk MDS based on the revised IPSS scoring system) were enrolled in this study. The median age of the 95 patients is 67 years (range of 31 Ð 84) and median follow-up duration was 747 days (range of 137-3328 days). We performed targeted deep sequencing (entire exon region of a panel of 84 myeloid genes, Agilent custom probe set) on 285 bone-marrow samples including the longitudinal samples taken at diagnosis (n=95) and post-AZA treatment, (median 4 cycles) as well as T-cell fraction (CD3+). We multiplexed and sequenced the samples using an Illumina Hiseq 2000. After read mapping and variant calling, hierarchical clustering, pathway and survival analyses were performed in R.

Results:

Targeted sequencing on the myeloid gene panel revealed 176 mutations in 68 patients (68/95, 71.6%) with a median of 2 mutations per patient (ranges 2-6). The average on-target coverage for 285 sequenced samples was 1205x. Twenty-five of 44 mutated genes were recurrently mutated. ASXL1 was the most frequently mutated in the cohort (21%), followed by TET2 (15%), DNMT3A (11%), and SRSF2 (11%). Mutated genes were then grouped into 8 biological pathways, defined in The Cancer Genome Atlas (TCGA) AML study. The most frequent biological pathway with mutated genes at diagnosis was DNA methylation (28.4%), followed by spliceosome (25.2%), chromatin modifiers (22.1%), myeloid transcription factors (TFs) (11.6%), activated signaling (11.6%), tumor suppressors (12.6%), and cohesin complex (6.3%).

When assessing the differences in patterns of variant allele frequency (VAF), we found significant VAF reduction in responders compared to non-responders (p = 0.007, repeated measures using general linear model, Figure A). Multivariate analyses revealed that mutation burden in different genes and biological pathways have distinct impact on AZA response, AML transformation, and overall survival. Higher bone marrow blast percentage (5%) was associated with all three measures (Figure B). Most significantly, mutations in activated signaling pathway genes are associated with AML progression (p=0.002). In addition, we could not detect decreased VAFs in activated signalling pathway genes even in responders (Figure C-D). Patients with SRSF2 mutations tend to respond to AZA (OR 14.084, p=0.003). Mutations in tumor suppressors (HR 4.825, p<0.001) and myeloid TFs (HR 3.070, p=0.020) were adverse prognostic factors in overall survival. Of interest, mutations in DNA methylation pathway were not independent prognostic factor for AZA response, AML transformation, or overall survival.

Conclusion:

These data and analyses show that reduction in mutation burden is correlated with AZA response. Mutations in different genes and biological pathways are associated with distinct clinical measures that tumor suppressors and myeloid TFs were identified as poor prognostic factors in terms of OS. Persistent mutation burden in activated signaling pathways is a strong predictor for AML transformation. In summary, longitudinal tracking of MDS patients using NGS may improve criteria for AZA response and early detection of AML progression.

Figure 1.

View largeDownload slide

Figure 1.

View largeDownload slide

Close modal