Abstract
Immunotherapy has revolutionised therapeutic approaches to fight cancer and, in certain diseases dramatically improves survival. Clinical responses to immune checkpoint blockade have in part been attributed to high mutational burden of tumours such as melanoma. High-risk acute myeloid leukaemia (AML) is defined by molecular and cytogenetic factors. AML has a low prevalence of somatic mutations and is predicted to have low immunogenicity. We aimed to determine how AMLs driven from different classes of oncogenes interact with endogenous anti-leukemic immune responses.
Methods and Results
We generated three oncogenically distinct models of AML: BCR-ABL+NUP98-HOXA9 (BA/NH9), MLL-AF9 (MA9), and AML1-ETO+NRASG12D (AE/NRAS), using retroviral transduced bone marrow transplanted into immune-competent, non-irradiated C57BL/6J (B6) mice or immune-deficient Rag2-/-γc-/- mice. Immunologic control of AML was dependent on the driver oncogene, as AE/NRAS AML was effectively controlled in B6, but not Rag2-/-γc-/-recipients, whereas survival of BA/NH9 AML recipients was similar between B6 and Rag2-/-γc-/-. MA9 AML had an intermediate phenotype (Figure 1A-C).
To examine the mechanisms underlying immune escape in AE/NRAS, AML from immune-deficient or immune-competent hosts, was passaged through immune-competent hosts. Prior exposure to an intact immune system dramatically accelerated disease progression of AE/NRAS AML in subsequent B6 recipients, but this was not seen in passage through Rag2-/-γc-/- recipients. This demonstrates specific, functional immunoediting of AML resulting in evasion of immune control.
Despite evidence of disease attenuation in immune competent hosts, functional immunoediting was not observed in MA9 AML. Antibody-mediated immune cell depletion experiments demonstrated that natural killer (NK) cells and T cells both contribute to the control AE/NRAS AML, whereas MA9 immune control was dependent on NK cells. As immunoediting was only seen in AE/NRAS model, this suggests that functional immunoediting in this model is primarily mediated by T cells.
To characterise the mechanisms regulating immunoediting, we integrated proteomic and transcriptional analysis of immunoedited and non-immunoedited AE/NRAS AML. There was strong correlation between increased protein expression and transcriptional regulation. There was distinct regulation of inflammatory pathways between immunoedited and non-immunoedited AML. Immunoedited AE/NRAS cells showed increased IFN-γ-dependent response signatures, consistent with direct targeting of the leukemic cells by the immune system. Transcriptional analysis also showed modulation of expression of immune checkpoint molecules including upregulation of suppressive molecules Tim-3 and CD39 and downregulation of activating ligand CD137L. These findings were confirmed by cell-surface flow cytometry. Immunoedited AE/NRAS downregulated RAS signalling transcriptionally, with coordinate activation of MYC targets.
In the murine AE/NRAS model, CD4+ and CD8+ T effector memory (TEM) cells (CD44+ CD62L-) demonstrated increased PD-1 expression compared to naïve mice. In addition, mice with high disease burden also had increased frequency of T cells co-expressing exhaustion markers PD-1, Tim-3 and LAG-3, consistent with suppression of the anti-leukemic effector immune response. To understand if these findings were relevant to AML in the clinic, we obtained single cell RNA-sequencing data from the CD45+ CD34- non-leukemic fraction of bone marrow in a patient with AML1-ETO AML at diagnosis compared to that in normal marrow. Single cell type classification and clustering using tSNE demonstrated remodelling of the immune microenvironment in AML with loss of NK cells, pre-B cells and skewing of T cell subsets. There was depletion of CD8+ TEM cells and greater proportions of CD4+ and CD8+ TEM cells expressing activation and exhaustion markers (IFN-γ, PD-1, LAG-3, TIM-3).
Conclusions
These data demonstrate that immune responses in AML are oncogene-specific and provide evidence that AE/NRAS AML cells undergo immunoediting over time in the presence of a competent immune microenvironment. Since AML is associated with alterations in T cell subsets, and changes in T cell activation and exhaustion states, these findings may inform translational strategies to use immunotherapies for patients with AML.
Smyth:Bristol Myers Squibb: Other: Research agreement; Tizona Therapeutics: Research Funding. Lane:Janssen: Consultancy, Research Funding; Celgene: Consultancy; Novartis: Consultancy.
Author notes
Asterisk with author names denotes non-ASH members.