Immunologic Targets in AML

The established clinical efficacy of checkpoint inhibitors such as anti–programmed cell death protein 1 (anti–PD-1), anti–programmed death ligand 1 (anti–PD-L1), and anti-CTLA4 for many cancers confirms the central role that endogenous T cells play in the development of tumor immunity and that these responses can be converted into meaningful levels of tumor cell control and elimination in vivo. Because checkpoint inhibitors do not act directly on tumor cells or immune cells to generate new immune responses, the efficacy of these agents is primarily dependent on the presence of preexisting T-cell responses that have been neutralized by the tumor cells themselves, often through creating an immune-suppressive microenvironment that protects the tumor and inhibits effective T-cell immunity. Importantly, these endogenous T-cell responses are relatively tumor specific but in most instances, the exact specificity of these responses has not been established.

In this issue of Blood Cancer Discovery, Nelde and colleagues (1) use a sensitive mass spectrometry–based immunopeptidomics approach to characterize the antigenic landscape of acute myeloid leukemia (AML). Unlike approaches that predict immunogenic epitopes, this mass spectrometry–based approach characterizes and quantifies large numbers of peptides actually presented by HLA molecules in each sample. To develop a comprehensive profile of potential antigenic targets in AML, the investigators purified relatively undifferentiated leukemia stem/progenitor cells (LSC) as well as bulk AML cells from patients and characterized peptides eluted from both HLA class I and HLA class II molecules. Although the physical detection of HLA-bound peptides does not guarantee their immunogenicity, this approach avoids some of the limitations of epitope prediction based on exome sequencing. Overall, there was a great deal of individual variation in the number of antigenic targets expressed by each patient's leukemia cells but most peptides were eluted from both LSC and bulk AML blasts. However, it was also evident that each cell type expressed unique peptides. This likely reflects the expression of different genes as LSC undergo aberrant differentiation to produce the still relatively immature bulk population of AML blasts and highlights the need for an effective immune response to specifically target LSC as well as bulk AML cells. Based on these new findings, targeting antigens expressed only on LSC or bulk AML cells is not likely to be an effective immunologic strategy.

As documented in this report, many unique peptides were eluted from both HLA class I and HLA class II molecules and found to be derived from noncanonical cryptic peptides as well as mutation-derived neoepitopes. Further studies examined the immunogenicity of these targets and found that specific responses could be induced in healthy donor T cells, validating these peptides as potentially immunogenic targets. More importantly antigen specific memory responses directed against many of these peptides could be detected in AML patients. Most of these responses were mediated by CD4 T cells directed at peptides presented by HLA class II molecules. Intriguingly, the diversity of the HLA class II immunopeptidome was positively associated with overall survival and preexisting CD4 T-cell memory responses were associated with improved event-free survival in small cohorts of selected patients. These additional findings further supported the validity and clinical relevance of immune responses directed at these antigens.

Taken together, these studies by Nelde and colleagues (1) provide convincing data that primary AML cells are highly immunogenic and that endogenous T cells—primarily CD4 T cells—can recognize and respond to a variety of cryptic peptides as well as mutation-derived neoepitopes. Importantly, these findings support the clinical relevance of autologous immune responses to AML tumor cells in individual patients. As pointed out by the authors, these studies also lay the foundation for future clinical trials to evaluate the safety and efficacy of personalized multi-peptide vaccines based on the use of mass spectrometry–based immunopeptidomics to identify relevant immunogenic epitopes.

While being highly informative, these studies also bring up new issues that can be addressed in future laboratory studies. Although many patients with AML can achieve complete remission in response to chemotherapy, most patients eventually relapse. This indicates that the immune response detected in patients with AML is not sufficient to provide effective long-term control of malignant cells in vivo. Moreover, having demonstrated leukemia-specific T-cell memory responses in vivo, why don't these patients respond to checkpoint inhibition therapy? Notably, most of the patients studied in the present report (70%) had newly diagnosed AML. Comparison of immunopeptidome at time of initial diagnosis with similar analysis at time of relapse in multiple patients will provide information on the stability of the initial set of epitopes presented for immune recognition. It will also be important to determine which of the immunogenic peptides are presented clonally by all LSC and bulk AML cells and not by subclones that evolve in response to therapy. It is certainly possible that the HLA-presented immunopeptidome might change dramatically as relatively small chemotherapy resistant AML subclones present at diagnosis emerge as dominant clones at the time of relapse after chemotherapy or in response to immune selection. Even “AML driver” mutations in genes such as NPM1 may not be clonal and peptides derived from this AML neoantigen were only detected in bulk AML and not in LSC. Major HLA loss itself was not detected in the present studies, but may nevertheless occur in selective and subtle forms at time of relapse as a mechanism to alter antigen presentation and prevent immune recognition (2, 3). These types of changes may also explain resistance to checkpoint inhibitor therapy which has been primarily applied in patients with relapsed disease. In that setting, antigen-specific memory CD4 T cells directed at diagnostic AML cells will not be effective if the same epitopes are no longer expressed at time of relapse.

While the immune response mediated by autologous T cells may not be able to provide long-term leukemia control, allogeneic stem cell transplantation is an effective therapy for many patients with AML. In the allo-transplant setting, elimination of AML cells in vivo is primarily mediated by donor T cells exerting a graft-versus-leukemia (GVL) response. In this context, it will be informative to compare autologous T-cell responses with allogeneic GVL responses that are able to confer long-lasing effective control of AML in 40% to 60% of patients. Of course, GVL responses are not effective in all patients and mechanisms of resistance to donor T cells have been well documented (4). These include complete or partial genetic loss of HLA class I expression as well as downregulation of HLA class II gene expression (5, 6). In both instances, AML cells have responded to intense immunologic pressure mediated by donor CD8 and CD4 T cells to reduce expression of immunogenic epitopes. Donor T cells can also develop “exhaustion signatures” and GVL responses can sometimes be restored by infusion of additional donor T cells [donor lymphocyte infusion (DLI)] or discontinuation of immune suppression used to prevent graft versus host disease (GvHD; refs. 7, 8). Checkpoint inhibitor therapy has also demonstrated some responses in the setting of posttransplant relapse, particularly in patients with isolated extramedullary relapse (9). Detailed single-cell studies in these settings of alloimmunity have revealed the ability to reverse T-cell exhaustion in some patients and other mechanisms to elicit effective immune responses in some patients (7).

As with autologous AML T-cell responses, the repertoire of AML peptide epitopes recognized by donor T cells after allogeneic stem cell transplantation has not yet been fully characterized. It is likely that at least part of the allogeneic T-cell response is directed at some of the same types of immunogenic peptides identified in the current report. There is some evidence that this includes responses to AML-specific neoantigens. A recent report examining clinical outcomes in patients with NPM1-mutated AML found significantly lower relapse rates in patients with HLA alleles capable of presenting the NPM1c-mutant peptide (10). However, it is also likely that the GVL response, unlike autologous T-cell responses, is not limited to detect epitopes that distinguish LSC and bulk AML cells from normal hematopoietic stem cells in the same individual. Specifically, allogeneic GVL responses are also able to target minor histocompatibility antigens (miHA). miHA are derived from genetic polymorphisms that distinguish recipient and donor even if they are fully HLA-matched. These genetic differences, usually in the form of nonsynonymous single-nucleotide polymorphisms, do not affect gene function but can result in immunogenic peptides that are detected as foreign epitopes by donor T cells. When miHA are widely expressed in nonhematopoietic tissues of the recipient, the donor immune response generated against these antigens results in severe GvHD. But when miHA are expressed only in AML cells and normal recipient hematopoietic cells, the GVL response results in the elimination of both recipient hematopoiesis and recipient AML cells while simultaneously facilitating engraftment of donor hematopoietic stem cells. Unlike mutation-derived neoantigens and cryptic neoantigens in AML cells which may be subclonal, miHA are more likely to be clonal and expressed by all AML cells and all normal recipient hematopoietic cells.

Advanced technologies such as the mass spectrometry–based immunopeptidomics approach utilized by Nelde and colleagues (1) offer great promise in our ability to fully characterize and dissect effective tumor immune responses directed at primary tumor cells. As exemplified for AML in the current report, this approach can also be applied to other hematologic malignancies as well as solid tumors. Future studies can also be directed to identify mechanisms of resistance as well as to characterize effective immune responses and evaluate new clinical approaches designed to induce effective immunologic responses in vivo.