Analysis of 3346 patients from the combined Children's Oncology Group AML studies further characterizes RAM immunophenotype AML and reveals other associated phenotypes and genetic drivers.

Abstract

Abstract Introduction: RAM phenotype acute myeloid leukemia (RAM-AML), identified by flow cytometry at diagnosis, is associated with age under 2 years, positive MRD at end of induction I (EOI1), and poor outcomes. This phenotype is linked to the CBFA2T3::GLIS2 (CBF-GLIS) fusion and aberrant folate receptor alpha (FOLR1) expression. Given the rarity of this AML subset, full characterization of the phenotype remains limited. Clarification of the specifics of RAM phenotype may facilitate identification of ‘RAM-alike’ cases with similar but distinct phenotypes. To address this, we performed a combined analysis of flow cytometric data from COG studies AAML0531, AAML1031, and AAML1831. Methods: Patients enrolled in AAML0531, AAML1031, or AAML1831 were analyzed. Study details have been previously published. 1831 patients who were FLT3+ were excluded due to continued enrollment in Arm C of the trial. Immunophenotyping was centrally and prospectively evaluated by ΔN, as previously described. Initial filtering was performed using the MFI of canonical RAM antigens CD56, HLA-DR, and CD38 (n=159) followed by manual review. CBF-GLIS fusion status, and a panel of other AML fusions was assessed via RNAseq and/or RT-PCR. Results: Flow cytometric analysis revealed 64 patients with immunophenotypes consistent with RAM-AML. In addition to the established phenotype no patients expressed CD11b or CD36, despite FAB M7 association. RAM-AML universally expressed CD117 and were overwhelmingly CD33 positive (98%). 84% expressed CD34 on at least a subset of cells. CD13 was variable, and when positive often dimly expressed. All RAM-AML patients with data available (n=18) were FOLR1 positive (heterogenous expression, 78% moderate/bright). Data confirmed previously reported clinical features of RAM-AML, including age (median: 1.5 years) and poor outcomes (5-year EFS and OS 21% and 26%, respectively). Of 62 RAM-AML patients with available data, 59 (95%) had CBF-GLIS fusions. While the three CBF-GLIS negative patients were consistent with RAM phenotype, it is noted that FOLR1 surface expression was not available; one patient was significantly older than typical for RAM patients (12 years), one lacked expression of CD184, typically highly expressed in RAM-AML patients, and the final fusion negative patient had a CBF-GLIS positive twin sibling. Manual review revealed 46 ‘RAM-alike’ cases with similar but distinct phenotypes to RAM-AML. Compared to RAM-AML, these patients were significantly older and less frequently EOI1 MRD positive (p<0.001 for both). Although these patients had numerically improved EFS and OS compared to RAM-AML, the differences were not significant and overall outcomes remained poor (5-year EFS and OS 28% and 38%, respectively). Comparison to genetics in RAM-alike patients revealed recurrent gene fusions involving KMT2A (n=7), ETV6 (n=4), and FUS (n=13). Notably, 12/15 (80%) FUS::ERG patients across COG AML trials were found in the RAM-alike group. RAM-alike patients with FUS and ETV6 fusions had especially poor outcomes (5-year EFS, OS of 0%, 9.1% and 0%, 0% respectively). Given reports of ERG dysregulation in CBF-GLIS positive patients we investigated the immunophenotype of all FUS::ERG cases. We found that while many patients were RAM-alike there were common key differences, including frequent expression of CD11b and lack of FOLR1. FUS::ERG patients not in the RAM-alike cohort had phenotypic similarities, but expressed either normal CD45 or lacked CD56. While less consistent in presentation, the data indicate FUS::ERG may define an immunophenotypic subgroup related but distinct from RAM-AML. Conclusions: Analysis of the largest reported cohort of RAM-AML patients confirms the association with age, CBF-GLIS fusions, expression of FOLR1, and poor outcome. Our data supports refining the RAM phenotype to include lack of CD11b, expression of FOLR1, and presence of CD117 and/or CD33. Inclusion of these markers enhances diagnostic specificity and excludes non-hematopoietic tumors, such as neuroblastoma, which can appear similar to the original definition. Finally, analysis of RAM-alike patients with similar but distinct immunophenotypes revealed recurrent genetic fusions, most notably FUS::ERG, which demonstrated its own immunophenotypic characteristics. The phenotypic similarities in RAM-alike indicate that fusions involving functionally linked genes may result in similar immunophenotypes and clinical outcomes, possibly with treatment implications.