Tumor phenotypes are dictated not only by the neoplastic cell component, but also by the tumor microenvironment (TME), which includes immune and inflammatory cells. Acute myeloid leukemia (AML) is the second most common leukemia of childhood. Although intensive multi-agent chemotherapy in conjunction with improved supportive care has increased survival rates to 70%, 30-40% of children with AML relapse and only one-third of them will survive to adulthood. Investigation of new therapeutic strategies for high-risk AML, including immunotherapy, remains a priority and is being actively pursued. Importantly, AML-induced immune suppression poses a fundamental barrier to successful immune-based interventions. We previously showed that in vitro treatment of primary AML cells with interferon (IFN)-γ induces functional indoleamine 2,3-dioxygenase-1 (IDO1), a molecule with potent immunosuppressive properties, in 50% of unselected cases (IFN-γ responders). With a median follow-up of 8 years, IFN-γ responders experienced a significantly shorter event-free survival and overall survival (OS).
Herein, we aimed to get insights into the immune landscape of childhood AML using multi-scale immune profiling strategies with the aim to identify molecular circuits that can be targeted to revert leukemia-induced immune dysfunction and improve clinical outcome.
We employed a novel high-throughput digital platform, the nCounter system (nanoString Technologies, Seattle, USA), and Optimized Multi-color Immuno-phenotyping Panels (OMIPs) to comprehensively characterize the bone marrow immune infiltrate in 34 children diagnosed with non-promyelocytic AML (17 males, 17 females; median age at diagnosis=10 years; 28 de novo AML, 4 infant leukemia, 1 secondary AML, 1 congenital leukemia; 7 patients with favorable-risk cytogenetics, 21 patients with intermediate-risk cytogenetics and 6 patients with unfavorable-risk cytogenetics). The RNA Pan-Cancer Immune Profiling Panel, which includes 109 cell surface markers for 24 immune cell types, 30 cancer testis antigens, and >500 immune response genes, allowed us to measure immune gene expression levels on the clinic-ready nCounter® FLEX platform (nanoString Technologies). Transcriptomic data were normalized and analyzed using the nSolver software package. Relapse-free survival (RFS) and OS were selected as covariates. For immune phenotyping studies, viable bone marrow mononuclear cells were labeled with fluorochrome-conjugated antibodies and then analyzed on a 10-color Gallios flow cytometer (Beckman Coulter Life Sciences, Buckinghamshire, UK) with standard settings.
Hierarchical clustering of immune gene expression data highlighted patient subgroups with heightened expression of T-cell and natural killer (NK)-cell function genes, as well as Toll-like receptor (TLR) genes, interleukin genes and tumor necrosis factor (TNF) family genes. Patients with high levels of CD8A mRNA also expressed IFNG, CXCL9, CXCL10, FOXP3 and negative checkpoints, including LAG3, CTLA4, IDO1 and CD279 (PD-L1), consistent with T-cell inflamed phenotypes that might underpin the establishment of adaptive immune resistance mechanisms of immune escape. In-depth phenotyping studies showed higher percentages of bone marrow-resident naïve B cells, plasmablasts, intermediate and non-conventional monocytes, plasmacytoid dendritic cells (DCs) and CXCR3-expressing Th1 cells at diagnosis compared with complete remission (CR). By contrast, both type 1 and type 2 myeloid DCs were more represented in the TME of patients with AML in CR. The frequencies of innate lymphoid cells (ILCs) and Vγ9+Vδ2+ T cells were not significantly different when comparing diagnosis and CR samples.
Relapses occurred in 71.4% of children with “inflamed” AML compared with 28.6% of patients with “non-inflamed” AML (p=0.012; two-sided Fisher’s exact test). Importantly, RFS was remarkably shorter in patients with inflamed compared with non-inflamed AML (median RFS=394 days vs. undefined; hazard ratio = 0.266; 95% CI=0.11-0.64; p=0.0039, log-rank test). A non-significant association with OS was also observed.
In conclusion, inflamed or “hot” AMLs are associated with shorter RFS and might be amenable to immunotherapy approaches tailored to the bone marrow TME, including PD-1/PD-L1 blockade and/or small-molecule IDO1 inhibitors.
Grant support: Roger Counter Foundation, Dorset, UK and Qatar National Research Fund (grant #NPRP8-2297-3-494).
Citation Format: Jayakumar Vadakekolathu, Gemma A. Foulds, Tasleema Patel, Stephen Reeder, Alan Graham Pockley, Sarah K. Tasian, Sergio Rutella. Immune gene expression profiling identifies predictors of relapse in childhood acute myeloid leukemia [abstract]. In: Proceedings of the AACR Special Conference on Tumor Immunology and Immunotherapy; 2017 Oct 1-4; Boston, MA. Philadelphia (PA): AACR; Cancer Immunol Res 2018;6(9 Suppl):Abstract nr B62.