Acute Myeloid Leukemia (AML) is a malignancy of hematopoietic stem/progenitor cells. Although immune checkpoint blockade has been effective for some cancers, this approach has had limited efficacy in AML patients. A better understanding of the mechanisms contributing to the immune escape of AML at presentation may lead to new insights for improving mechanism-based immunotherapies.

We therefore performed Cytometry by Time Of Flight (CyTOF) with 42 antibody markers to define the immune microenvironment in bone marrow samples from adult de novo AML patients at presentation, with healthy donor bone marrow samples as comparators. We evaluated a cohort of 59 de novo AML patient samples, carefully chosen to represent common AML mutations, and responses to standard-of-care treatment regimens. This cohort had complete clinical and mutational annotation available. Bulk RNA sequencing of all AML bone marrow samples was performed to define the transcriptomic features of the dominant AML clone in each sample.

Compared to healthy donors, AML bone marrow samples were significantly enriched for differentiated lymphocyte populations (for B, T, and NK cells). This finding was most prominent in T cell subsets, where AML samples had significantly more memory CD4+ and CD8+ cells (e.g. 20.3% of CD8+ cells were classified as Effector Memory T cells re-expressing CD45RA/”TEMRA” in HD marrow, compared to 35.8% in AML samples, P=0.01). Th1 cells (as a subset of CD4+ T cells) were also more abundant in AML samples (HD=3.2%, AML=12.9%, P<0.001). However, we also found that T cell phenotypes are highly variable in AML samples.

Using unsupervised clustering of B and T cell subsets, we identified one group of cases (21/59) where the T cell phenotype is “naïve“ (i.e. similar to healthy donors), and other samples (13/59) that contained mostly terminal effector T cells (TEMRA CD4+ and CD8+ T cells, as well as CD56+CD57+CD8+ T cells). We hypothesized that these phenotypes may represent different mechanisms of immune escape: for example, AML samples with naïve lymphocytes may lack neoantigens, or are immunologically silent because of other mechanisms. In contrast, the T cells of AML samples with large numbers of terminal effector T cells are likely recognizing AML-specific neoantigens, but the T cells fail to clear the tumor, perhaps due to AML-cell intrinsic resistance mechanisms. To test this hypothesis, we compared transcriptomic data from patients in the ”naïve“ and ”terminal effector“ groups. We found that samples from the ”terminal effector“ group had increased expression of the anti-apoptotic gene BCL2 compared to the “naïve” group (1.7 fold, p=0.01), which has an established role in resistance to granzyme B-mediated killing mediated by activated T or NK cells. IFNGR2, encoding for a subunit of the Interferon Gamma Receptor, is expressed at lower levels in AMLs with the ”terminal effector“ phenotype vs. the “naïve” phenotype (1.75 fold, p=0.03), potentially preventing the effector function of IFNG secreted by T cells, representing an alternative resistance mechanism.

We also evaluated the mutational landscapes of these sample groups, and found that patients with activating KRAS mutations (5/59) were enriched for naïve cells in both CD4+ and CD8+ compartments. KRAS mutations are almost always subclonal in AML, but its ability to facilitate immune escape via a “trans” mechanism could potentially explain why it is advantageous for AML progression. Finally, we also evaluated the immunological phenotypes of patients with stable complete remissions (“CRs” lasting >2 years with chemotherapy alone) vs. patients who were refractory to two rounds of intensive induction therapy (i.e. Primary-Refractory AML = P-R). The P-R AML samples had increased terminally differentiated CD56+CD57+ CD8+ T cells (CR=5.2%, P-R=13.9%, P=0.02), and decreased CD4+ Th2 cells (CR=15.5%, P-R=7.02%, P=0.001), compared to CR patients. These data suggest that T cell immunity may play an important role in the effectiveness of chemotherapy, corroborating previous studies from our lab (PMID: 34845035). In summary, we developed a “gold-standard” CyTOF dataset that identified heterogenous mechanisms of immune escape by AML cells; we will use these data to design mechanism-based approaches to improve AML immunotherapy.

This content is only available as a PDF.
2025
Sign in via your Institution