Molecular Characteristics Influencing Relapse Risk in Low-Risk Pediatric Acute Myeloid Leukemia

The Children’s Oncology Group (COG) clinical trial AAML1031 decreased cytarabine (AraC) exposure from 45.8 gm/m² to 21.8 gm/m² for acute myeloid leukemia (AML) patients with low risk (LR) cytomolecular/response features. This 52% decrease in cumulative AraC exposure resulted in significantly decreased survival (PMID 34472213); however some patients were cured at this lower exposure. Currently available data do not distinguish between patients who did/did not experience relapse with de-escalated AraC therapy. To address this data gap, we analyzed RNA expression data available in CAVATICA to identify genes and pathways differentially expressed between patients with/without AML relapse in the LR subpopulation.

Our study cohort included eligible patients with LR AML treated on AAML1031 who had complete remission (CR) by the End of Induction II (EOI2) and had available RNA-seq data. AAML1031 defined the LR cytogenetics (LR-cyto) group as patients with inv(16), t(8;21), NPM1 and CEBPα. The remaining LR patients had uninformative cytomolecular features and were measurable residual disease (MRD) negative at the end of Induction I, categorized as standard-risk cytogenetics (SR-cyto). Patients were followed from the end of Induction II to relapse, other off-study event, or completion of planned follow up at 10 years post enrollment.

The short read RNA-Seq files were generated through the COG TARGET AML initiative and deposited on the CAVATICA platform. We performed gene quantification (StringTie 2.1.3), fusion partner detection (Arriba 2.2.1), and differential gene expression (DEG) analysis (DESeq2 3.19), with up-regulated genes defined as having a > 2-fold change and adjusted p-value < 0.05. QIAGEN Ingenuity Pathway Analysis was performed to identify signaling and metabolic pathways.

We obtained gene abundance values for 313 LR-cyto and 277 SR-cyto patients out of the 697 LR AML patients who had CR at EOI2. The demographic characteristics were comparable between LR AML patients with RNA-Seq data (n = 590) and the remaining eligible patients without RNA-Seq data (n = 107).

For the LR-cyto patients, we further stratified them into MRD positive (MRD+) (n = 37) and MRD negative (MRD-) (n = 276) subgroups. Twenty-two (59%) MRD+ patients experienced relapse, compared to 75 (27%) in the MRD- subgroup. There were 77 up-regulated genes in the MRD+ subgroup when comparing relapse to non-relapse samples, with B cell development being the most significantly altered pathway. In the MRD- subgroup, 27 different genes were up-regulated, with response to elevated platelet cytosolic Ca2+ as the most altered pathway.

For the SR-cyto subgroup, we further stratified patients into KMT2A (n = 151) and non-KMT2A subgroups (n = 126). Ninety-five (63%) KMT2A patients relapsed, compared to 70 (56%) in the non-KMT2A subgroup, aligning with the known poor prognosis associated with KMT2A rearrangement. We further identified fusion partner genes for KMT2A patients. The three most frequent fusion partner genes were MLLT3 (n = 52, 34%), MLLT10 (n = 20, 13%), and AFDN (n = 12, 8%). Additionally, 9 (6%) patients had more than one fusion partner with KMT2A. We identified 247, 193 and 290 up-regulated genes overall in SR-cyto, KMT2A and non-KMT2A subgroups, respectively, when comparing patients relapsed versus those who did not. Sixteen genes were shared among all three groups, with the expression and translocation of olfactory receptors being the most activated pathway in all three.

Our analyses identified specific relapse-associated genes and pathways, demonstrating potential immunotherapy antigens in LR de novo AML subpopulations. Work is ongoing to validate these analyses and identify transcriptomic predictors of relapse in this patient population.