Dissecting the immune architecture of AML and its influence on therapeutic response identifies patients for whom immunotherapy will be most beneficial.

Single-cell and spatial multimodal technologies have propelled discoveries of the solid tumor microenvironment (TME) molecular features and their correlation with clinical response and resistance to immunotherapy. Computational tools are incessantly being developed to characterize tumor-infiltrating immune cells and to model tumor immune escape. These advances have led to substantial research into T-cell hypofunctional states in the TME and their reinvigoration with T-cell–targeting approaches, including checkpoint inhibitors (CPIs). Until recently, we lacked a high-dimensional picture of the acute myeloid leukemia (AML) TME, including compositional and functional differences in immune cells between disease onset and postchemotherapy or posttransplantation relapse, and the dynamic interplay between immune cells and AML blasts at various maturation stages. AML subgroups with heightened interferon gamma (IFN-γ) signaling were shown to derive clinical benefit from CD123×CD3–bispecific dual-affinity retargeting molecules and CPIs, while being less likely to respond to standard-of-care cytotoxic chemotherapy. In this review, we first highlight recent progress into deciphering immune effector states in AML (including T-cell exhaustion and senescence), oncogenic signaling mechanisms that could reduce the susceptibility of AML cells to T-cell–mediated killing, and the dichotomous roles of type I and II IFN in antitumor immunity. In the second part, we discuss how this knowledge could be translated into opportunities to manipulate the AML TME with the aim to overcome resistance to CPIs and other T-cell immunotherapies, building on recent success stories in the solid tumor field, and we provide an outlook for the future.

Purpose: The stromal and immune bone marrow (BM) landscape is emerging as a crucial determinant for acute myeloid leukemia (AML). Regulatory T cells (Treg) are enriched in the AML microenvironment, but the underlying mechanisms are poorly elucidated. Here, we addressed the effect of IFNγ released by AML cells in BM Treg induction and its impact on AML prognosis. Experimental Design: BM aspirates from patients with AML were subdivided according to IFNG expression. Gene expression profiles in INFγhigh and IFNγlow samples were compared by microarray and NanoString analysis and used to compute a prognostic index. The IFNγ release effect on the BM microenvironment was investigated in mesenchymal stromal cell (MSC)/AML cell cocultures. In mice, AML cells silenced for ifng expression were injected intrabone. Results: IFNγhigh AML samples showed an upregulation of inflammatory genes, usually correlated with a good prognosis in cancer. In contrast, in patients with AML, high IFNG expression was associated with poor overall survival. Notably, IFNγ release by AML cells positively correlated with a higher BM suppressive Treg frequency. In coculture experiments, IFNγhigh AML cells modified MSC transcriptome by upregulating IFNγ-dependent genes related to Treg induction, including indoleamine 2,3-dioxygenase 1 (IDO1). IDO1 inhibitor abrogated the effect of IFNγ release by AML cells on MSC-derived Treg induction. In vivo, the genetic ablation of IFNγ production by AML cells reduced MSC IDO1 expression and Treg infiltration, hindering AML engraftment. Conclusions: IFNγ release by AML cells induces an immune-regulatory program in MSCs and remodels BM immunologic landscape toward Treg induction, contributing to an immunotolerant microenvironment. See related commentary by Ferrell and Kordasti, p. 2986

Breast cancer bone metastasis is currently incurable, ~75% of patients with late-stage breast cancer develop disease recurrence in bone and available treatments are only palliative. We have previously shown that production of the pro-inflammatory cytokine interleukin-1B (IL-1B) by breast cancer cells drives bone metastasis in patients and in preclinical in vivo models. In the current study, we have investigated how IL-1B from tumour cells and the microenvironment interact to affect primary tumour growth and bone metastasis through regulation of the immune system, and whether targeting IL-1 driven changes to the immune response improves standard of care therapy for breast cancer bone metastasis. Using syngeneic IL-1B/IL1R1 knock out mouse models in combination with genetic manipulation of tumour cells to overexpress IL-1B/IL1R1, we found that IL-1B signalling elicited an opposite response in primary tumours compared with bone metastases. In primary tumours, IL-1B inhibited growth, by impairing the infiltration of innate immune cell subsets with potential anti-cancer functions but promoted enhanced tumour cell migration. In bone, IL-1B stimulated the development of osteolytic metastases. In syngeneic models of breast cancer, combining standard of care treatments (Doxorubicin and Zoledronic acid) with the IL-1 receptor antagonist Anakinra inhibited both primary tumour growth and metastasis. Anakinra had opposite effects on the immune response compared to standard of care treatment, and its anti-inflammatory signature was maintained in the combination therapy. These data suggest that targeting IL-1B signalling may provide a useful therapeutic approach to inhibit bone metastasis and improve efficacy of current treatments for breast cancer patients.

Acute myeloid leukemia (AML) is a genetically heterogeneous hematological malignancy with variable responses to chemotherapy. Although recurring cytogenetic abnormalities and gene mutations are important predictors of outcome, 50% to 70% of AMLs harbor normal or risk-indeterminate karyotypes. Therefore, identifying more effective biomarkers predictive of treatment success and failure is essential for informing tailored therapeutic decisions. We applied an artificial neural network (ANN)–based machine learning approach to a publicly available data set for a discovery cohort of 593 adults with nonpromyelocytic AML. ANN analysis identified a parsimonious 3-gene expression signature comprising CALCRL, CD109, and LSP1, which was predictive of event-free survival (EFS) and overall survival (OS). We computed a prognostic index (PI) using normalized gene-expression levels and β-values from subsequently created Cox proportional hazards models, coupled with clinically established prognosticators. Our 3-gene PI separated the adult patients in each European LeukemiaNet cytogenetic risk category into subgroups with different survival probabilities and identified patients with very high–risk features, such as those with a high PI and either FLT3 internal tandem duplication or nonmutated nucleophosmin 1. The PI remained significantly associated with poor EFS and OS after adjusting for established prognosticators, and its ability to stratify survival was validated in 3 independent adult cohorts (n = 905 subjects) and 1 cohort of childhood AML (n = 145 subjects). Further in silico analyses established that AML was the only tumor type among 39 distinct malignancies for which the concomitant upregulation of CALCRL, CD109, and LSP1 predicted survival. Therefore, our ANN-derived 3-gene signature refines the accuracy of patient stratification and the potential to significantly improve outcome prediction. Key Points Machine-learning approaches identified a parsimonious gene-expression signature that predicts risk in newly diagnosed AML. The 3-gene PI could be used to refine the accuracy of patient stratification and outcome prediction in routine clinical practice. Visual Abstract