A plasma-based test for the measurable residual disease informs the response to treatment in AML patients

Abstract Background: Allogeneic hematopoietic cell transplantation (alloHCT) remains the only potentially curative treatment for many patients with AML or MDS. However, relapse after alloHCT continues to be the major cause of treatment failure. Despite its critical prognostic importance, routine measurable residual disease (MRD) monitoring after alloHCT is underutilized due to limitations of existing assays, which often rely on invasive bone marrow (BM) biopsies and have limited sensitivity. To address these limitations, we developed a personalized, plasma-based, cell-free DNA (cfDNA) assay, termed v96, that uses a panel of up to 96 patient-specific leukemia mutations. We then used this assay to evaluate MRD in AML/MDS patients undergoing alloHCT. Methods: We performed a longitudinal analysis of 185 serial plasma, peripheral blood leukocytes and BM samples from 31 patients (29 AML, 2 MDS; median age 63 years, range 35–75) who underwent alloHCT in first complete remission. Whole-genome sequencing (WGS) of diagnostic and matched remission samples was used to identify up to 96 leukemia-specific mutations in each patient. We then analyzed these mutations in both the Watson and Crick strands (duplex next generation sequencing) in serial plasma, BM and peripheral blood leukocytes samples obtained at various times before and after alloHCT. Results: Pre-alloHCT samples could be evaluated by v96 in 13 patients. MRD was detected in 13.3% (2/13) by multiparameter flow cytometry but in 100% by v96. Inclusion of a large number of mutations in the v96 was critical to its sensitivity; MRD could be detected in only 5 of the 13 patients when driver gene mutations alone (median of 1 mutation per patient) were evaluated through the same highly sensitive duplex sequencing assay. At 3 months post-alloHCT, inclusion of more mutations in the assay was also essential for high sensitivity: with v96, 69.2% (9/13) patients had MRD, while only 23.1% (3/13) had MRD based on analysis of the driver gene mutations. We then compared the sensitivity of MRD in paired BM and plasma samples from 20 patients using v96. MRD was detectable in 85.0% (17/20) of plasma samples versus 65.0% (13/20) of matched BM aspirates. Importantly, plasma cfDNA detected MRD in 4 patients whose BM samples were negative with v96. Overall, MRD burden was higher in plasma than in BM (median mutant allele frequency (MAF) of 1.14e-05 vs. 3.47e-06, respectively (p=0.01)). We then evaluated the prognostic value of the v96 assay in plasma. Prior to alloHCT, patients with a MAF > 0.01 (n=5) had a much higher 2-year cumulative incidence of relapse (CIR) than patients with a MAF≤0.01 (n=16) (80% (4/5) vs. 0% (0/16), respectively (p<0.0001)). At 3 months post-alloHCT, v96 assay positivity was also highly correlated with CIR: 67% (4/6) of v96-positive patients had a CIR while only 5.5% (1/18) of v96-negative patients had a CIR (p<0.001). Among patients who remained relapse-free, MAF levels assessed with v96 consistently declined over time, with 35% (8/23) achieving clearance by 12 months post-HCT. None of the 8 patients who cleared MRD via the v96 assay relapsed, while 2 of the 15 patients who did not clear MRD eventually relapsed. Conclusion: We developed a personalized assay for MRD in patients with AML or MDS. Its sensitivity was considerably higher with cell-free DNA from plasma than with DNA from bone marrow. The evaluation of multiple mutations, not simply driver gene mutations, was also critical to its sensitivity. These results have important clinical implications, enabling real-time post-alloHCT monitoring without the need for invasive procedures. The v96 approach may represent a major step toward precision post-alloHCT care in patients with AML or MDS.