Dissecting the Immune Landscape of Acute Myeloid Leukemia

Abstract

Acute myeloid leukemia (AML) is a molecularly heterogeneous hematological malignancy with variable response to treatment. Recurring cytogenetic abnormalities and molecular lesions identify AML patient subgroups with different survival probabilities; however, 50⁻70% of AML cases harbor either normal or risk-indeterminate karyotypes. The discovery of better biomarkers of clinical success and failure is therefore necessary to inform tailored therapeutic decisions. Harnessing the immune system against cancer with programmed death-1 (PD-1)-directed immune checkpoint blockade (ICB) and other immunotherapy agents is an effective therapeutic option for several advanced malignancies. However, durable responses have been observed in only a minority of patients, highlighting the need to gain insights into the molecular features that predict response and to also develop more effective and rational combination therapies that address mechanisms of immune evasion and resistance. We will review the state of knowledge of the immune landscape of AML and identify the broad opportunity to further explore this incompletely characterized space. Multiplexed, spatially-resolved immunohistochemistry, flow cytometry/mass cytometry, proteomic and transcriptomic approaches are advancing our understanding of the complexity of AML-immune interactions and are expected to support the design and expedite the delivery of personalized immunotherapy clinical trials.

Keywords: acute myeloid leukemia; bispecific antibodies; immune checkpoint blockade; immunotherapy; tumor immunological microenvironment.

Figure 1

Therapeutic targeting of immune suppression in the Acute myeloid leukemia (AML) tumor immunological microenvironment (TME). Microenvironmental soluble factors, such as interferon (IFN)-γ produced by cytotoxic T cells, promote leukemia cell proliferation [85], instigate immune suppressive mechanisms, including the induction of indoleamine 2,3-dioxygenase-1 (IDO1), and mediate resistance to genotoxic damage [36]. IDO1 inhibitors such as epacadostat, indoximod and navoximod have entered the clinical arena for patients with advanced solid tumors [86,87]. Nitric oxide (NO)-releasing aspirin (nitroaspirin) interferes with the inhibitory enzymatic activities of arginase-2 (ARG2) and NO synthase expressed in myeloid cells and has been administered orally to normalize the immune status of tumor-bearing mice [88]. In AML patients, DNAM-1, an activating receptor for NK cells and T cells, is reduced and its ligands CD155 and CD112 are increased, indicating a tolerogenic phenotype [89]. Shedding of CD137L leading to increased serum levels correlates with worse prognosis and may constitute an immune suppressive circuit in AML [90]. CD200 (OX2) is a negative regulator of T-cell function that is frequently increased in AML and is associated with poor prognosis. CD200R immunomodulatory fusion proteins (IFPs) with the cytoplasmic tail replaced by the signaling domain of the costimulatory receptor CD28 have been recently engineered [91,92]. Adoptive therapy with CD200R-CD28-transduced leukemia-specific CD8⁺ T cells has been shown to eradicate murine AML more efficiently than wild-type T cells. Antibodies targeting CD47, an inhibitory receptor preventing phagocytosis of AML cells [93], are currently being tested in a phase I, dose-escalation clinical trial (ClinicalTrials.gov Identifier: NCT02678338). Antibodies targeting Tim-3 [94] are under evaluation in combination with hypomethylating agents and immune checkpoint blockade (ICB) for patients with AML and high-risk myelodysplastic syndrome (MDS) (ClinicalTrials.gov Identifier: NCT03066648). Blue boxes denote therapeutic strategies already in the clinic. Yellow boxes highlight therapeutic strategies that have been evaluated pre-clinically. ARG2 = arginase-2; Gal-9 = galectin-9; NK = natural killer.

Figure 2

Identification of prognostic biomarkers in the AML TME. PREdiction of Clinical Outcomes from Genomic profiles (PRECOG) is a pan-cancer resource supporting the identification of prognostic genes in public datasets of human malignancies [104]. A machine-learning tool, known as CIBERSORT [102], can be applied to PRECOG data to comprehensively map compositional differences in tumor-infiltrating leukocytes (22 distinct subsets) in relation to patient outcome. (Panel A) shows hierarchical clustering (Euclidean distance; complete linkage) of CIBERSORT-inferred immune cell type fractions in a broad spectrum of hematological malignancies (1957 samples), including AML. Data were analyzed using Morpheus (Broad Institute, MA; https://software.broadinstitute.org/morpheus/). Red denotes an association with shorter survival times, whereas blue indicates an association with better clinical outcomes. Each column represents an immune cell type and each row represents a disease type. (Panel B) shows a similarity matrix (Pearson correlation) of CIBERSORT-inferred immune cell type fractions in hematological malignancies. This unbiased approach could support the identification of co-expression patterns of specific immune cell populations in the TME, thus providing unique insights into the immuno-biology of hematological malignancies and accelerating the delivery of personalized immunotherapy approaches. BCP-ALL = B-cell precursor acute lymphoblastic leukemia; CLL = chronic lymphocytic leukemia; BL = Burkitt lymphoma; DLBCL = diffuse large B-cell lymphoma; FL = follicular lymphoma; MM = multiple myeloma.

Figure 3

Selection of immunotherapy approaches in AML with inflamed versus non-inflamed TMEs. Messenger RNA (mRNA) profiles and spatially-resolved expression of immune checkpoints could be integrated with conventional AML prognosticators, such as patient age, presenting white blood cell count and ELN cytogenetic risk, to stratify patients into categories with different survival probabilities. Patients with T-cell inflamed profiles, indicative of adaptive resistance-driven immune dysfunction, could be considered for immunotherapy approaches that incorporate IDO1 inhibitors [86], either as monotherapy or in combination with PD-1/PD-L1 ICB [87], or other immunotherapy agents that deliver an activation signal to T cells, including CD3 × CD123 DART proteins [9], and/or revert MDSC- and Treg-mediated immune dysfunction in the TME. In contrast, AML cases with a non-T-cell inflamed TME, and/or blast cells lacking IFN-γ responsiveness as a result of abnormalities in intracellular signaling pathways, could be candidates for therapeutic strategies that enhance T-cell trafficking into the BM (STING agonists, β-catenin inhibitors [31]) and/or passive immunotherapy approaches such as the infusion of leukemia antigen-specific T cells or CD123-CAR T cells [115]. Pharmacological approaches, including the use of hypomethylating agents, could enhance T-cell infiltration to the BM, thus converting a “cold” TME into a “hot” TME [110]. IDO1 = Indoleamine 2,3-dioxygenase-1; L-TRP = l-tryptophan; 1MT = 1-methyl-tryptophan; CAR = chimeric antigen receptor; TAM = tumor-associated macrophage; Treg = regulatory T cell; MDSC = myeloid-derived suppressor cell; WT1 = Wilms’ tumor 1; PRAME = preferentially expressed antigen in melanoma. Green arrows denote stimulation; red arrows denote inhibition.