Dynamic Changes in Clonal Clearance with Decitabine Therapy in AML and MDS Patients

To determine how AML subclonal architecture changes during decitabine treatment, and whether specific mutations might correlate with sensitivity vs. resistance to decitabine, we performed exome sequencing at multiple time points during single agent decitabine therapy. We enrolled 69 patients with either AML (age ≥ 60, or with relapsed/refractory disease, N = 45) or MDS (N = 24) on a phase I clinical trial. All subjects were treated with decitabine 20 mg/m² on days 1-10 of 28 day cycles. With a median follow-up of 13.7 months, the intention to treat clinical response (complete remission with or without complete count recovery: CR/CRi) is 40%, with survival correlating with response (median survival – CR/CRi: 583 days; partial response/stable disease (PR/SD): 260 days; progressive disease (PD) or failure to complete cycle 1: 36 days, p < 0.0001). We performed exome sequencing on unfractionated bone marrow cells at diagnosis (day 0), cycle 1 day 10, cycle 1 day 28, cycle 2 day 28, and, when possible, during remission and at clinical relapse/progression. We have completed sequencing analysis for the first 34 cases (outcomes: 5 CR, 15 CRi, 3 PR, 8 SD, and 3 PD). Several important themes have emerged, as follows:  1) We correlated mutation status at diagnosis with clinical response. All six patients with TP53 mutations obtained clinical CR or CRi, and exome analysis demonstrated near complete elimination of the TP53-associated founding clones by the end of cycle 2 (p < 0.03). Long-term outcomes were similar in these patients compared with other patients who achieved CR/CRi: four patients relapsed after 8, 9, 10, or 17 cycles; 1 patient is doing well post-transplant; and one patient died of an infectious complication after cycle 2. No other mutations were significantly associated with clinical response or with consistent mutation clearance. 2) We observed a reduction in blast counts, which preceded mutation elimination in fourteen cases with CR, CRi or PR. This suggests that decitabine may induce morphological blast differentiation in vivo prior to mutation elimination.  3) In eight of nine cases with a clinical response followed by relapse, clinical progression was associated with expansion of a pre-existing subclone. We have not yet observed any recurrent mutations that reliably predict whether a subclone will contribute to relapse. Intriguingly, in two of these cases, the relapse-associated subclone was detectable at diagnosis and was eliminated more slowly than the founding clone mutations, suggesting that this subclone harbored intrinsic decitabine-resistance.  4) Complete remission can occur with concomitant non-malignant, clonal hematopoiesis. In three cases with a CR, a new clonal population was selected for during the remission. In two of these cases, there were no shared mutations between the founding clone and the emergent, non-malignant, clonal hematopoiesis, suggesting that these clones were unrelated.  5) Mutational architecture is generally stable, but differential chemo-sensitivity can be detected even between subclones in the same patient. In ten cases with PR or SD, we observed minimal shifts within the mutational burden over the course of eight weeks, suggesting that “clonal drift” is a relatively slow process. However, in four cases with SD, what appeared clinically to be simple persistent disease was in fact a dynamic elimination of one subclone, and its replacement by a different subclone. Similarly, in three cases with CRi, we observed rapid clearance of a subclone with slower clearance of the founding clone, again suggesting differential chemo-sensitivity among subclones.  6) Finally, we correlated pharmacologic markers with clinical outcomes. We observed no correlation between steady-state plasma decitabine levels and clinical responses. Using Illumina 450k methylation arrays, we observed a correlation between response and the extent of decitabine-induced hypomethylation in total bone marrow cells that persisted on cycle 1 day 28 (p < 0.01), but not on cycle 1 day 10 (p < 0.1). In summary, these data reveal that response to decitabine is associated with morphologic blast clearance before mutations are eliminated, that relapse is associated with subclonal outgrowth that may be identified early in the treatment course, and that TP53 mutations may be predictive of rapid clinical responses, although, like most responses to decitabine, these are not necessarily durable.