Adverse Outcomes in Acute Myeloid Leukemia Are Associated with Tumor Cell-Mediated Immunosuppression

Background: A normal karyotype is the most common cytogenetic finding in acute myeloid leukemia cells ("NK-AML") and has highly variable and difficult to predict outcomes at presentation. Refinements introduced with the ELN 2017 Criteria (presence or absence of mutations in NPM1, FLT3-ITD, and/or biallelic CEBPA mutations) have improved--but not completely solved--up-front risk classification. NK-AML patients with very long first remissions after chemotherapy only are rare, and their genetic and epigenetic characteristics have not yet been definitively evaluated, but could potentially provide clues regarding determinants that affect chemotherapy responses.

Results: Using exome sequencing, RNA-sequencing, and functional studies, we performed an in-depth characterization of 28 de novo adult NK-AML patients with >5 year, Long First Remissions (LFR) after chemotherapy only (7+3 with HiDAC consolidation). As comparators, we studied a well-matched group of 31 NK-AML patients who relapsed within 2 years (Standard First Remissions, SFR). None of the patients in either group received an allogeneic transplant in first remission. Bulk RNA sequencing and targeted investigation of cell type proportions and pathways expected to be relevant for chemotherapy sensitivity (i.e. cell cycle, DNA repair, or AraC responsiveness) revealed no significant differences between LFR and SFR AML samples. However, single cell RNA sequencing (sc-RNAseq) analyses of genetically-defined AML cells from 8 LFR and 7 SFR cases showed that SFR AML cells expressed significantly higher levels of stem cell markers, and of several immunomodulatory genes (MHC class II, CD200, MRC1). Combined sc-RNAseq and flow cytometry analyses of the AML bone marrow samples from LFR patients revealed well-preserved CD4 and CD8 populations that responded robustly to TCR-mediated activation, achieved through CD3/28 bead stimulation. In contrast, SFR AML marrows had reduced numbers of immune cells (particularly of CD4 Th1 cells) and their residual T cells expressed an exhaustion signature. TCR-mediated T cell activation was blocked in the SFR samples by the presence of the AML cells in the bone marrow samples: removing the AML cells released this blockade. Evaluation of T cell immune checkpoints and their putative ligands revealed LAG3 to be a leading candidate for immune suppression, because of its high expression in SFR T cells, and because the genes encoding its major ligands (MHC Class II) were highly expressed in SFR AML cells. Indeed, blocking the inhibitory receptor LAG-3 was able to reverse the AML cell-mediated T cell suppression of most SFR cases. To validate and extend these findings, we performed the T cell activation assay in an independent cohort of 50 AML cases representing all ELN risk groups. The ability of marrow derived CD4+ T cells to activate correlated strongly with ELN groups. Importantly, for the ELN intermediate patients, the immune activation assay stratified these patients in 2 groups with very different relapse free survival (RFS): the T cell "activators" had an RFS of 1200 days, while the "non-activators" had RFS of 146 days (Mantel-Cox Log-rank = 0.02). A multivariate analysis of covariates at diagnosis, including sex, age, ELN risk category, WBC count, bone marrow blast count at presentation, and mutation status (for DNMT3A, NPM1, FLT3, NRAS, WT1, TET2, CEBPA, PTPN11, IDH1, IDH2, RUNX1, RAD21, SMC1A, and SRSF2) identified none that were predictive of CD4 T cell activation status at presentation.

Conclusions: These data strongly suggest that most AML cases have an immunosuppressive phenotype at presentation, mediated by inhibitory T cell receptors (e.g. LAG3) via ligands that are expressed by AML cells (e.g. MHC Class II proteins). The absence of this immunosuppressive phenotype was strongly correlated with favorable clinical outcomes with standard chemotherapy. The ability of CD4 T cells to be activated via the TCR in the presence of AML cells at presentation (which indirectly measures AML cell-mediated immunosuppression) is a novel biomarker for risk assessment that is currently not captured by any other well-recognized covariate (especially for ELN intermediate risk cases), and should be further evaluated for its predictive potential in prospective clinical trials.