Independent Prognostic Value of DNA-Based Minimal Residual Disease Assessment in Unselected Population-Based Cohort of Pediatric Acute Myeloid Leukemia

Monitoring of minimal residual disease (MRD) in childhood acute myeloid leukemia (AML) can assess response to treatment and predict relapse. Current MRD monitoring methods in childhood AML – flow cytometry and quantitation of fusion transcripts – have significant shortcomings and do not cover the entire spectrum of patients. We demonstrated (Lukes et al, Hemasphere 2020) the feasibility of MRD monitoring using DNA-based primary targets in selected subgroups of AML. Now we extended this approach to the unselected consecutive population of pediatric AML.

We included 133 children (0-18) out of 135 consecutively diagnosed with AML in Czechia (2012-2022), all treated on non-MRD-based protocols. For genetic characterization, we used algorithm consisting of screening for recurrent gene fusions and mutations by PCR and targeted-NGS followed by transcriptome sequencing (WTS). Targeted-NGS (or PCR) was performed to identify genomic fusion sequences.

130 patients were eligible for MRD monitoring (3 patients died early). MRD was preferentially monitored by DNA-based qPCR (alternatively by amplicon NGS or qRT-PCR). MRD levels were expressed relative to diagnosis.

Primary genetic aberration was found by targeted screening in 102 children; remaining 31 were further investigated using WTS. In 27 children, WTS identified rare, novel or atypical primary genetic aberrations not covered by the targeted screening.

Primary genetic aberration was found in 97% of cases (129/133). Majority of AMLs (81%) were classified into common subtypes: AML with KMT2Ar, PML::RARA, RUNX1::RUNX1T1, CBFB::MYH11, mutation (m) of GATA1 , CEBPA or NPM1. Recurrent but rare aberrations in pediatric AML were found in 16 patients: UBTFm, RUNX1m, HOXA10 translocation, KAT6A::CREBBP, KAT6A::LEUTX, DEK::NUP214, BCR::ABL1, NUP98::NSD1 and CBFA2T3::GLIS2. In 5 patients, fusion genes were found, previously described sporadically (SPFQ::ZFP36L216, XPO1::TNRC1817) or not at all in AML (ETV6::CTNNB1, FUS::FEV, ZEB2::RUNX1). Only secondary aberrations were identified in remaining 4 patients.

In 82 patients (of 84 examined), genomic fusion sequence was found and qPCR designed and used. Of 41 children carrying genetic aberrations other than fusions, in 29 quantification system was successfully implemented. Altogether, quantification system for MRD monitoring was established in 122 patients, with primary aberrations as DNA targets in 116 patients, reaching sensitivity of 10^-4 in 112 and 5x10^-4 in 4 patients. Four and two children were monitored with sensitivity of 10^-4 using secondary aberrations as DNA targets or by qRT-PCR-based quantification of fusion transcripts, respectively.

MRD clearance significantly differed among genetic subtypes: while GATA1m AML had the fastest response, in contrast to patients with prognostically favorable genetic subtypes - CBFB::MYH11, RUNX1::RUNX1T1, CEBPAm and NPM1m – who were treated predominantly on the standard risk (SR) arm of the AML-BFM 2012 Registry Protocol and none achieved molecular remission (mREM) at D28 (70% vs 0%, p < 0.0001). Significant proportion of patients (25-89% within individual subtypes) did not achieve mREM even after the last block of chemotherapy. The response of patients with KMT2Ar AML and AML classified into remaining subtypes (predominantly intermediate- and high-risk) was overall faster compared to prognostically favorable subtypes on SR arm (KMT2Ar D28 mREM 33% vs 0% at D28, p < 0.0001). KMT2A::MLLT10/3 had faster clearance then other KMT2Ar (mREM D28 44% vs. 0% in other KMT2Ar, p = 0.02).

In a multivariate analysis including (cyto)genetic risk and treatment, D28 MRD was the only significant predictor of outcome using both the 10^-3 (p = 0.006 for EFS and 0.012 for OS) and 10^-2 levels (p = 0.004 for EFS and 0.01 for OS) for stratification. Similarly, in an alternative multivariate analysis model, D56 MRD (positive at any level versus negative) was also the only significant predictor of outcome independent of risk and treatment (p = 0.005 for EFS and 0.01 for OS)

In summary, we present a strategy for MRD monitoring in pediatric AML that is technologically feasible, real-life applicable to vast majority of all patients, and has clear prognostic significance.

Supported by National Institute of Cancer Research No. LX22NPO5102 funded by the European Union–Next Generation EU, and AZV grant NU20-07-00322.