Does the Immune System Matter in Treatment for AML?

Dr. Michaelis is the principle investigator of the “Safety Study of MGD006 in Relapsed/Refractory Acute Myeloid Leukemia (AML) or Intermediate-2/High Risk MDS” for Medical College of Wisconsin (www.clinicaltrials.gov/ct2/show/study/NCT02152956).

Acute myeloid leukemia (AML) is a heterogeneous disease, and therefore categorization and patient risk stratification are necessary to identify patients who are predicted to have a favorable versus poor response to standard treatment, or who may benefit from alternative targeted therapy or allogeneic bone marrow transplantation. As our understanding of the genomic landscape of AML moves forward and immunotherapies become available, we may need to further refine our ability to predict disease and treatment response beyond the current genomic European Leukemia-Net (ELN) guidelines. From experience treating solid organ malignancies with immune checkpoint inhibitors, we know that response to anti-programmed cell death 1/programmed death ligand 1–targeted immunotherapy is more common in patients with tumors that have immune-inflamed gene expression profiles.^1,2  Investigation of the immune tumor microenvironment (TME) of AML may allow identification of molecular predictions of immunotherapeutic benefit and guide treatment decisions.

Dr. Vadakekolathu and colleagues have delved into the complex immunologic profiles of the AML TME using immune gene expression profiling to reveal IFN-dominant, adaptive, and myeloid mRNA profiles, similar to those seen in patients with melanoma. When used in an unbiased manner, these gene profiles can dichotomize AML into immune-infiltrated and immune-depleted subtypes. AML cases with immune-infiltrated profiles were characterized by higher expression of IFN-stimulated genes and T-cell related functions. Highly multiplexed digital spatial profiling of immune-oncology protein expression in a small subgroup of patients showed deregulation in a particular gene signature including programmed death ligand 1, FoxP3, GZMB, PTEN, and BCL-2. This signature occurred in patients with predominantly CD3-rich immune-infiltrated AML and correlated with adverse ELN cytogenetic features and poor clinical outcomes. While immune subtypes improved the accuracy of ELN cytogenetic risk categories, this was context dependent. Immune-infiltrated TME predicted a longer relapse-free survival (RFS) and overall survival (OS) in favorable-risk disease; however, in adverse-risk disease it predicted worse clinical outcomes. Immune subtype alone was not able to stratify survival in patients with intermediate-risk AML; however, it seemed to improve the predictions of Nucleophosmin 1 (NPM1) and FMS-like tyrosine kinase 3 mutational status on survival. Further, identification of a set of 21 differentially expressed immune genes between high-risk and favorable-risk AML was able to separate ELN intermediate-risk patients into high and low gene expression values. Intermediate-risk patients with low expression of the 21 genes closely resembled the favorable-risk patients; conversely, high expression was similar to adverse-risk patients and associated with worse RFS and OS. Finally, high levels of immune infiltration and IFN signaling molecules were seen in patients with TP53 mutations who are known to have a particularly poor prognosis.

The clinical relevance of these findings may be exemplified by three scenarios (Figure). First, the immune subtype of AML seems to be important for selecting patients for allogeneic hematopoietic stem cell transplantation (aHSCT). Patients with immune-infiltrated AML at baseline showed a significantly longer OS to those that had chemotherapy alone, and this is presumably due to the beneficial graft-versus-leukemia effect. Interestingly, patients with immune-depleted AML did not show a survival benefit of aHSCT. Secondly, IFN-γ-related genes in immune-infiltrated AML cases improved the prediction of chemotherapy resistance with IFN-dominant mRNA profiles showing higher rate of primary induction failure and early relapse. The immune landscape may therefore identify patients with AML who would not benefit from traditional “3+7” cytarabine and anthracycline treatment. Finally, immune subtyping of AML may predict response to novel therapies, particularly immunotherapy. Flotetuzumab is a CD3 × CD123 dual affinity retargeting molecule that seems to have some clinical activity in AML. Gene expression profiling was performed on 30 patients with relapsed or refractory AML enrolled in the CP-MGD006-01 clinical trial (NCT #02152956). The authors identified that clinical responses to flotetuzumab were seen in 92 percent of patients with evidence of an immune-infiltrated AML TME. Here, the tumor inflammation signature (TIS) score was a strong predictor of antileukemic response to this immunotherapy, with an area under the receiver operating characteristic value of 0.847.

Collectively, these findings identify an immune landscape of AML and add depth to the predictive models of disease response and prognosis. One of the key strengths of this study is the use of a large number of primary patient samples (n > 440) with longitudinal response data and independent validation with publicly available transcriptomic and clinic data (e.g., HOVON, Beat AML, and TCGA AML cohorts). While TME-immune profiling is not in a position to replace the current genomic classifications, it may become a useful adjunct to extend response predictions and treatment algorithms.

We wish to thank Prof. Rutella for his review and helpful comments.