Decoding Immune Interactions in Acute Myeloid Leukemia

Immunotherapy has transformed cancer treatment, offering new hope for patients with otherwise limited options. Despite its success across a variety of cancers, outcomes in myeloid malignancies, particularly acute myeloid leukemia (AML), remain suboptimal. Unlike other cancers, AML has a notoriously hostile bone marrow (BM) microenvironment, which is characterized by complex immune evasion mechanisms. This poses significant challenges in achieving long-lasting remission. Allogeneic hematopoietic stem cell transplantation (HSCT) is a cornerstone treatment for AML, providing a curative potential through the graft-versus-leukemia (GvL) effect, where donor immune cells target and eliminate residual leukemia. However, even with this intervention, relapse rates remain alarmingly high, with limited therapeutic options available for patients who relapse post-HSCT. Donor lymphocyte infusion (DLI) is one such salvage strategy, designed to reinstate immunological control by delivering immune cells from the original donor to reignite GvL activity. While DLI has proven highly effective in chronic myeloid leukemia (CML), with response rates nearing 80%, its efficacy in AML is significantly lower, averaging only 15–20%. This disparity underscores the unique challenges posed by AML, including its ability to suppress and evade immune responses through the modulation of the BM microenvironment. Furthermore, the mechanisms underlying these differing responses between AML and CML are poorly understood, leaving clinicians and researchers grappling with how to optimize DLI for broader success in AML. Current research in cancer immunotherapy often focuses on tumor-infiltrating immune cells, but little is known about how these cells behave within the specialized niches of the BM, particularly in AML. The immune ecosystem in AML is remarkably complex, involving dynamic interactions between various immune cells, such as T cells, natural killer (NK) cells, and myeloid cells, as well as the leukemic cells themselves. These interactions influence whether the immune system can mount an effective response or falls prey to exhaustion and dysfunction. Understanding these interactions is critical for identifying why some patients respond favorably to DLI while others do not.

To this note, new research paper published in Science Immunology and led by Professor Catherine Wu from the Dana-Farber Cancer Institute and Professor Elham Azizi from Columbia University employed cutting-edge single-cell transcriptomics, spatial profiling, and advanced computational modeling, to investigate the cellular and molecular dynamics that define successful and unsuccessful responses to DLI in AML. The goal was not just to understand these mechanisms but also to uncover actionable insights that could guide the development of more effective immunotherapies. This study represents a significant step toward personalizing treatment for AML patients and overcoming the barriers that currently limit immunotherapy’s potential in this challenging disease.

The team conducted a comprehensive investigation into the immune dynamics of the bone marrow microenvironment in patients with AML undergoing DLI therapy. Using advanced single-cell RNA sequencing (scRNA-seq), they analyzed over 348,000 individual transcriptomes from longitudinal bone marrow samples taken from 25 patients. This dataset allowed them to delve into the intricate cellular composition and immune interactions present in both responders (Rs) and nonresponders (NRs) to DLI. Through this approach, they identified distinct immune cell populations and their functional states that were directly associated with therapeutic success or failure. The authors mapped the trajectory of T cells, particularly focusing on CD8+ cytotoxic T lymphocytes (CTLs), which play a central role in antileukemia immunity. In responders, the researchers observed the clonal expansion of a specific subset of CD8+ CTLs characterized by high expression of ZNF683, a marker associated with cytotoxic and effector functions. These ZNF683+ CTLs exhibited increased expression of genes linked to cytotoxicity, such as GZMB and PRF1, and were shown to be highly reactive against AML cells. Interestingly, these cells originated primarily from the infused donor lymphocytes, confirming that DLI contributes directly to the reinvigoration of immune responses in responders. In contrast, nonresponders lacked such clonal expansions, and their CTLs displayed elevated levels of exhaustion markers like TIGIT, highlighting a failure to mount an effective antileukemia response. To complement their transcriptomic analyses, the researchers employed spatial transcriptomics and protein-based spatial profiling to examine the physical organization of immune cells within the bone marrow. In responders, they discovered that the bone marrow microenvironment was spatially organized into highly diverse immune niches. These niches were enriched with effector T cells, natural killer (NK) cells, and dendritic cells that appeared to work in concert to mediate effective GvL responses. The spatial proximity of these immune cells suggested that responders’ bone marrow supported robust cross-talk between different immune subsets. Conversely, nonresponders showed less diverse and more disorganized immune niches, with myeloid cells dominating the landscape and minimal interactions between cytotoxic and effector immune cells. These findings pointed to a breakdown in immune coordination as a key feature of treatment resistance. The researchers further validated these findings using a newly developed computational model called DIISCO, designed to analyze dynamic cell-cell interactions over time. This model confirmed that, in responders, the ZNF683+ CTLs acted as central hubs, orchestrating interactions with NK cells and other immune subsets to suppress leukemia cells effectively. The DIISCO model also revealed that these interactions became more robust and targeted following DLI, underscoring the dynamic nature of the immune response in successful treatment scenarios. Nonresponders, on the other hand, lacked this coordinated immune network, with interactions between T cells and other immune cells remaining weak or absent.

Another notable experiment involved comparing the immune cell states before and after DLI. In responders, the researchers observed a striking transition in T cells from inhibitory states to highly cytotoxic phenotypes after therapy. This shift was evident in both transcriptomic data and spatial organization, where post-DLI immune niches showed an increased prevalence of CD8+ effector memory T cells co-expressing markers like CD57 and granzyme B. Nonresponders did not exhibit such transitions; instead, their T cells remained in states associated with exhaustion and dysfunction, indicating an inability to adapt effectively to the therapeutic challenge. Additionally, the team conducted flow cytometry on peripheral blood samples from an independent cohort of 53 AML patients who underwent DLI. This analysis corroborated their findings, revealing that responders had a marked expansion of CD8+ T cells with effector memory and cytotoxic profiles in their peripheral blood, mirroring the dynamics observed in the bone marrow. This external validation strengthened the study’s conclusions and highlighted the systemic nature of the immune responses seen in successful DLI treatments.

In summary, Professor Elham Azizi and colleagues provided compelling evidence that successful treatment hinges on the coordinated activity of specific immune subsets, particularly ZNF683+  CTLs. These findings bridge a critical gap in our understanding of immunotherapy for AML, a disease historically resistant to conventional immune-based treatments. One of the most significant outcomes of the study is the identification of key immune players that dictate therapeutic success. The expansion of ZNF683+ CTLs in responders underscores their pivotal role in mediating GvL effects. This insight offers a potential biomarker for predicting patient outcomes, allowing clinicians to stratify patients based on their likelihood of benefiting from DLI. It also highlights a tangible therapeutic target—strategies that enhance the proliferation or activity of these CTLs could potentially improve outcomes for nonresponders. The study also emphasizes the importance of immune cell interactions within the bone marrow. In responders, the spatial proximity of CTLs, NK cells, and dendritic cells created a collaborative network that amplified antileukemia responses. This underscores the significance of not only the presence of specific immune cells but also their ability to coordinate effectively within the microenvironment. Future therapies could aim to engineer this level of spatial and functional coordination, perhaps through adoptive cell transfer or localized cytokine delivery. For nonresponders, the study’s findings point to immune exhaustion as a major barrier to success. Elevated expression of inhibitory markers such as TIGIT in CTLs suggests that these cells lose their ability to effectively combat leukemia. This provides a clear avenue for intervention: targeting inhibitory pathways using immune checkpoint inhibitors could reinvigorate these cells, restoring their cytotoxic potential. By addressing the dysfunction in nonresponders, this research paves the way for therapeutic strategies that could broaden the applicability of DLI. We think the implications of this work extend beyond basic research, offering practical solutions to improve patient outcomes. For instance, the findings could inform the design of combination therapies that pair DLI with immune modulators, such as checkpoint inhibitors or cytokine-based treatments, to enhance efficacy. Moreover, the study’s validation of ZNF683+ CTLs as a critical immune subset opens avenues for the development of diagnostic tools to monitor and predict patient responses in real time.