Harnessing the immune system after allogeneic stem cell transplant in acute myeloid leukemia

Allogeneic stem cell transplantation (allo‐SCT) is the most successful and widely used immunotherapy for the treatment of acute myeloid leukemia (AML), as a result of its anti‐leukemic properties driven by T cells and natural killer (NK) cells, leading to a graft‐vs‐leukemia (GVL) effect. Despite its essential role in AML treatment, relapse after allo‐SCT is common and associated with a poor prognosis. There is longstanding interest in developing immunologic strategies to augment the GVL effect post‐transplant to prevent relapse and improve outcomes. In addition to prophylactic maintenance strategies, the GVL effect can also be used in relapsed patients to reinduce remission. While immune checkpoint inhibitors and other novel immune‐targeted agents have been successfully used in the post‐transplant setting to augment the GVL effect and induce remission in small clinical trials of relapsed patients, exacerbations of graft‐vs‐host disease (GVHD) have limited their broader use. Here we review advances in three areas of immunotherapy that have been studied in post‐transplant AML: donor lymphocyte infusion (DLI), immune checkpoint inhibitors, and other monoclonal antibodies (mAbs), including antibody‐drug conjugates (ADCs) and ligand receptor antagonists. We also discuss additional therapies with proposed immunologic mechanisms, such as hypomethylating agents, histone deacetylase inhibitors, and the FLT3 inhibitor sorafenib.


| INTRODUCTION
Allogeneic stem cell transplantation (allo-SCT) is the most successful form of immunotherapy for acute myeloid leukemia (AML). The success of allo-SCT hinges on the GVL effect, the ability of the donor immune system to recognize and attack recipient leukemia cells. First suggested by a mouse transplantation model in 1956, 1 this GVL effect was applied clinically in 1965 when allo-SCT was used to treat a patient with acute lymphocytic leukemia (ALL). 2 Allo-SCT is now known to cure many patients with AML; however, post-transplant relapse remains the most frequent cause of death, with relapse rates of 37% following reduced-intensity conditioning (RIC). 3 The GVL effect is critical for preventing relapses, but stimulating donor T lymphocytes leads to both acute and chronic GVHD that can be fatal and cause long-term morbidity. This paradox has stymied the universal adoption of treatments designed to prevent relapse by stimulating the GVL effect after allo-SCT. However, three factors will play an important role in the broader use of prophylactic post-transplant therapies to harness the immune system in the coming years: an increased ability to identify AML patients at the highest risk for post-transplant relapse, a better understanding of the mechanisms underlying those relapses, and the proliferation of novel agents and cell-based therapies with immunologic mechanisms.
A variety of pre-and post-transplant factors can identify AML patients at the highest risk for disease relapse after allo-SCT. Pretransplant factors that confer an increased risk of relapse include measurable residual disease (MRD) by flow cytometry prior to transplant (cumulative incidence of relapse [CIR] at 3 years 63% vs 22% in MRDnegative patients), 4 a complex karyotype at diagnosis (CIR at 2 years 46% among patients in CR1 at transplant), 5 or a FLT3-ITD mutation (CIR at 2 years 30% vs 16% for FLT3-ITD wild type patients). 6 One of the simplest methods for predicting relapse is the disease risk index (DRI), which relies on disease risk (ie, cytogenetics in AML) and disease status at the time of transplant to risk stratify patients for OS and PFS, which is driven largely by risk of relapse. 7 Post-transplant factors that predict an increased incidence of subsequent relapse include peripheral blood T lymphocyte donor chimerism ≤85% from day +90 to +120 (CIR at 3 years 29% vs 15% for donor chimerism >85% in patients transplanted in CR1/CR2), 8 and the presence of MRD by flow cytometry (CIR at 1 year ≥75% in patients with detectable MRD after allo-SCT). 9 Post-transplant chimerism has been used to select patients for preemptive therapy, including the early withdrawal of immunosuppression, donor lymphocyte infusion (DLI), and azacitidine. 10,11 However, the majority of patients who relapse after suggesting a second, distinct mechanism of immune evasion. 14 This finding suggests that selective immune checkpoint blockade may represent the ideal therapeutic modality to reverse the immune dysfunction that is driving relapse in these patients. Thus, recent research highlights the unique and distinct mechanisms of immune evasion underlying post-allo-SCT AML relapse, which may allow therapy to be tailored to overcome these mechanisms of relapse.
In the current review, we will outline advances in post-transplant therapy for the treatment of AML by focusing on three domains: DLIs; immune checkpoint inhibitors; and monoclonal antibodies (mAbs), including antibody drug conjugates and ligand receptor antagonists.
An overview of these therapies and relevant targets, where applicable, is listed in Table 1 and shown in Figure 1. A summary of notable studies with evidence for these therapies in post-transplant AML can be found in Table 2.
We also discuss other therapies with proposed immunologic mechanisms, such as hypomethylating agents, FLT3 inhibitors, and histone deacetylase inhibitors. While our goal is to focus on prophylactic maintenance strategies to prevent relapse, we also include data on the efficacy of immunotherapies in relapsed post-transplant patients due to the relative paucity of clinical trials in the latter population. In addition to highlighting novel therapies, we seek to illustrate how our increased understanding of the risk of relapse and the mechanisms of relapse will allow us to better select patients for prophylactic post-transplant therapies, thus minimizing the risk of additional toxicities in those patients who are less likely to relapse while maximizing the benefit of the selected therapies in those patients at the highest risk of relapse. Post-transplant AML relapse is driven in part by loss of leukemicspecific T cells and loss of T cell function. DLI has been found to increase anti-leukemic T cells, but also reverses T cell exhaustion through increased IFN-γ production and reduced T cell inhibitory receptors, such as programmed cell death protein 1 (PD-1) and T cell immunoglobulin and mucin domain 3 (Tim-3). 15 As with allo-SCT, the major obstacle with DLI is to maximize the GVL effect while minimizing GVHD. Canine studies in the 1970s illustrated the importance of delaying DLIs for at least 2 months after transplant in order to prevent fatal GVHD. 16 This allowed for the first clinical use of allogeneic lymphocytes in a patient with relapsed, post-transplant chronic myeloid leukemia (CML), who achieved remission following DLI. 17 Since then, DLI as the sole therapy for relapsed myeloid malignancies was shown to yield remission in~80% of patients with relapsed chronic phase CML, but just 29% of patients with AML or MDS with very limited long-term survival even among responding patients. 18

| DLI for relapsed AML
A number of strategies have been utilized to augment the efficacy of DLI, including pairing it with conventional AML therapy or T A B L E 1 Overview of immunotherapies for acute myeloid leukemia after allogeneic stem cell transplantation (allo-SCT) prior to DLI can dramatically improve the durability of responses. 4 Lymphodepleting therapy, such as fludarabine and cyclophosphamide, represents an alternative to disease-specific therapy that can alter the immunologic milieu to enhance the efficacy of DLI and yields a response rate of 49% in patients with relapsed, non-CML hematologic malignancies. 20 However, retrospective studies have also shown an increased incidence of acute GVHD, particularly GVHD of the lower GI tract, among patients who receive lymphodepleting chemotherapy prior to DLI. 21 Given the known benefit of GVHD in enhancing the GVL effect, there is a clear rationale for the use of lymphodepletion prior to DLI. Ultimately, both disease-specific chemotherapy and lymphodepletion given prior to DLI for relapsed AML may increase both response rates and/or response durability, but prospective, randomized trials are needed.

| DLI after T-cell depleted (TCD) allogeneic stem cell transplantation (allo-SCT)
There is also interest in using DLI after TCD allo-SCT. While TCD grafts can reduce GVHD and may obviate the need for GVHD prophylaxis, 22 they are also associated with increased rates of graft failure and relapse. 23,24 One strategy for dealing with such relapse is DLI. In a study assessing the role of DLI in 51 patients with relapsed AML or MDS after TCD allo-SCT, treatment with DLI resulted in a 5-year overall survival of 40% with a 5-year relapse rate of 69% and cumulative GVHD incidence of 45% at 5 years. 25 Some have sought to enhance the GVL effect of DLI by removing CD25+ regulatory T cells. In a phase I study involving 21 patients with relapsed hematologic malignancies after transplant (15 with AML), compared to unmanipulated DLI, CD25+ depleted DLI was associated with a better response rate and improved event-free survival, although no significant differences were found in the AML population. 26

| Prophylactic DLI for AML
While DLI has more limited efficacy in treating relapsed posttransplant AML, recent studies suggest DLI may successfully prevent relapse after allo-SCT. In a retrospective review comparing outcomes in 46 high-risk AML patients who received DLI, while in CR after allo-SCT, only 22% of patients relapsed (compared with 53% in the control group), and overall survival was 67% (compared with 31% in the control group) with a median follow-up of 7 years. 27 More recently, a registry-based matched-pair analysis of 89 pairs (65 with AML) investigating the role of prophylactic DLI after allo-SCT found no significant difference in a standard-risk cohort, but a reduced rate of relapse (31% vs 46% in controls) and improved 5-year overall survival (70% vs 40% in controls) among patients with high-risk AML, defined as unfavorable cytogenetics or transplant not in CR. Prophylactic DLI was also associated with non-significant trend toward more chronic GVHD. 28

| Prophylactic DLI with hypomethylating agents
Another method of increasing DLI efficacy is to combine it with hypomethylating agents. In a phase II study of 30 patients with high-risk AML (n = 20) and MDS (n = 10) treated with prophylactic azacitidine (AZA) followed by DLI after allo-SCT, the 2-year OS was 66% with cumulative incidence of relapse 28%, and cumulative incidence of acute GVHD 32%. While these outcomes are impressive in this highrisk, post-transplant population, approximately half of the patients who were identified as high-risk prior to transplant were unable to actually enroll in the study, primarily due to GVHD (15 patients) and early relapse (eight patients) at the time of planned enrollment. This issue confounds many post-transplant studies, as outcomes data are greatly biased by the selective inclusion of patients who survive the early post-transplant period without significant GVHD or relapse. 29

| Prophylactic DLI after TCD allo-SCT
Prophylactic DLI has also been used after TCD allo-SCT in an effort to improve engraftment and immune reconstitution and thus augment the GVL effect. In a study of 62 patients with AML or MDS, 5-year overall survival was 80% with event-free survival of 65%, and cumulative incidence of GVHD 31% at 5 years. 25 While the optimal timing of prophylactic DLI after TCD allo-HSCT is not established, DLI given less than 6 months after transplant may be associated with increased risk of GVHD. 30

| DLI dosing
The optimal dosing and number of DLIs remain to be determined; however, initial doses >1.0 × 10 7 cells/kg have been associated with an increased risk of GVHD without improved outcomes. 31,32 Many recommend an individualized approach to DLI, such as adjusting the dose based on the type of transplant 27 or adjusting the total number of DLI based on the risk of relapse. 33 Perhaps the most common method is dose escalation, or increasing doses with each subsequent DLI in order to improve outcomes. 34 Though a clearly defined endpoint is not always possible, some continue dose escalation until the patient develops GVHD to offer the best opportunity for clinical response. 35

| Second allo-SCT for relapsed AML
An alternative to DLI is second allo-SCT. In a retrospective study of 179 patients with relapsed acute leukemia after allo-SCT (72 with AML), treatment with a second allo-HCT was associated with 25% overall survival at 2 years. 36 In a retrospective study from the Acute 3 | NK CELL-BASED THERAPIES

| Introduction to NK cell-based therapies
In addition to T cells, NK cells play a critical role in providing the antitumor effect of allo-SCT. This may be of particular importance in AML, where leukemic cells are more sensitive to NK-mediated cytotoxicity than solid tumors. 39 The NK cells belong to innate lymphoid cells and are able to distinguish "self" from "non-self" through the interaction of for further studies to determine optimal dosing and timing of NK cell therapy. 45 One approach to increase NK cell activity is administering interleukin-2 (IL-2). In a phase II study involving eight patients (six with AML, two with MDS) with relapsed or persistent myeloid malignancy after allo-SCT, treatment with NK cells and IL-2 infusions resulted in two patients (one with AML and one with MDS) achieving complete response, with both relapsing less than 2 months later.
There were no incidents of GVHD. Of note, the authors experienced difficulty detecting NK cells after transfer, indicating a lack of persistence of NK cells, which could be due to inadequate numbers of infused cells, inhibition from Tregs, or limited cytokine availability. 46

| NK cells for AML after allo-SCT: Current clinical trials
As shown in Table 3

| Introduction to immune checkpoint inhibitors
Immune checkpoint inhibitors are increasingly being used for the treatment of hematologic malignancies including the post-transplant setting. 47 The theory is that by blocking immune checkpoints-such as the interaction of cytotoxic T-lymphocyte-associated protein-4 (CTLA-4) and programmed death-1 receptors (PD-1) with their respective ligands-immune checkpoint inhibitors can prevent tumors from evading the donor immune system. There is also evidence that checkpoint blockade may enhance NK cell activity. 48,49 The major concern with using immune checkpoint inhibitors after allo-SCT is the increased risk of GVHD.

| CTLA-4 blockade in AML
Ipilimumab is a monoclonal antibody against CTLA-4, a molecule that competes with CD28 for binding to B7 on antigen presenting cells in the 6 patients who received ipilimumab at 3 mg/kg. Of the 22 patients who received ipilimumab at 10 mg/kg, five patients achieved a complete response (23%), including all three with leukemia cutis, one with myeloid sarcoma, and one with smoldering MDS developing into AML. Among these 22 patients, dose-limiting toxic effects included three cases of GVHD (two cases chronic GVHD of the liver, one case acute GVHD of the gut), as well as three cases of immune-related adverse events (IRAEs)-thrombocytopenia, colitis, and pneumonitis-including one death. 50 While there is clear evidence for the efficacy of CTLA-4 blockade following allo-SCT in AML, the risks of GVHD exacerbation and IRAEs are currently too significant for prophylactic use.

| PD-1 blockade in AML
In addition to CTLA-4 inhibition, PD-1 has also been targeted in the post-transplant setting, as it has been implicated in one of the common immunologic mechanisms of relapse by Toffalori et al. 14 In a phase I/Ib study of nivolumab in 28 patients with relapsed hematologic malignancy after allo-SCT (11 with AML), two of six patients treated with nivolumab at 1 mg/kg developed dose-limiting IRAEs.
Subsequent dose reduction to 0.5 mg/kg was well-tolerated in an initial cohort of eight patients. However, upon enrollment of 14 more patients, accrual was terminated due to four cases of dose limiting toxicity, including two cases of grade III acute GVHD (liver and gut) resulting in two deaths. Of the 19 evaluable patients treated at nivolumab 0.5 mg/kg, the overall response rate was 16%, including partial response in one patient with AML. 51 In another study, 11 patients with recurrence of hematologic malignancy after allo-SCT (eight with AML) were treated with pembrolizumab for up to 2 years.
In all, seven patients experienced immune related adverse events with three dose-limiting toxicities and one treatment-limiting toxicity. Of the four AML patients evaluable for response, two had stable disease and two experienced disease progression. This study is ongoing. 52   lymphoma (89 of 176). 54 Thus, while the safety of post-transplant checkpoint inhibition remains a major concern, this should be inter-  Table 4). thus, targeting these antigens could lead to significant myelosuppression. 57

| ADCs and ligand-receptor antagonists in AML
The most frequently targeted antigen in AML thus far is CD33, an antigen that is variably expressed in leukemic blasts in 85-90% of patients with AML. 58 The most notable drug to target CD33 is gemtuzumab ozogamicin (GO), a humanized anti-CD33 IgG4 antibody conjugated to calicheamicin. Though its role in the post-transplant setting is still being determined, there at least two case reports of GO successfully treating isolated extramedullary AML relapse after allo-SCT. 59,60 Of note, hepatotoxicity-notably, venoocclusive disease (VOD)-has been a particularly concerning side effect that may be more common in patients who undergo allo-SCT. 61 GO has also been used in combination with AZA as maintenance therapy after allo-SCT.
pro-inflammatory cytokine IL-2 to the bone marrow and extramedullary AML sites, leading to both NK cell and CD8 + T cell expansion with associated tumor eradication. 66 In a small study involving four patients with relapsed AML after allo-SCT, treatment with F16-IL2 combined with low-dose cytarabine resulted in one patient with disseminated extramedullary AML achieving a complete metabolic response lasting for 3 months, and two patients achieving blast reduction with transient molecular negativity. Significant toxicity in the form of cytokine release syndrome developed in two of the four patients, though this was effectively managed with high-dose glucocorticoids. Ultimately both patients tolerated re-exposure to lower doses of F16-IL2. 67 Daclizumab is a mAb that binds CD25, the interleukin-2 receptor alpha chain (IL-2Ra) on T cells. Prior to the discovery of CD25 + FOXP3 + regulatory T cells, there was some thought that daclizumab may prevent acute GVHD by blocking T cell activation.  Table 5 72 The use of FLT3 inhibitors has been particularly useful in patients with the FLT3-ITD mutation, which carries an unfavorable prognosis. 73 However, these therapies inhibit multiple off-target receptor tyrosine kinases, which may contribute to their anti-leukemic effects and potential toxicities. This is particularly true of the firstgeneration FLT3 inhibitors, such as sorafenib and midostaurin. 74

| FLT3 inhibitors in AML
Sorafenib has been found to increase IL-15 production in mutant FLT3-ITD + leukemia cells, which enhances the GVL effect by augmenting CD8 + CD107a + IFN-γ − + T cells with features of longevity-that is, high levels of Bcl-2 and low levels of PD-1. Sorafenib also increases IFN-γ production, particularly in responders, 75

| Hypomethylating agents in AML
As demonstrated above, hypomethylating agents such as AZA and DAC are frequently used in combination with other immunotherapiesnot only for their direct cytotoxic effects but also their immunomodulatory properties. Anti-tumor effects of HMAs involve increased tumor recognition by upregulation of tumor cell antigens-notably, cancer testis antigens (CTAs) 80,81 -as well as enhanced T cell reactivity by increased expression of HLA class I antigen and co-stimulatory molecules. 82,83 Both AZA and DAC have also been shown to demethylate the FOXP3 promoter, leading to increased FOXP3 expression and expansion of regulatory T cells. 84,85 Animal models suggest Tregs may reduce GVHD by suppressing early expansion of alloreactive donor T cells without suppressing T cell activation, thus preserving the GVL effect. 86 Of note, there is also evidence to suggest differential use of T cell-mediated cytolytic pathways: while GVHD is mediated primarily by the FasL effector pathway, GVL responses rely largely on the perforin pathway. 87 As previously discussed, low levels of donor chimerism have been associated with increased risk of relapse. 8 The RELAZA study investigated the use of standard doses of azacitidine (75 mg/m 2 over 7 days) in AML/MDS patients judged to be at risk of imminent relapse based on declining post-transplant CD34 chimerism, and showed that this strategy restored donor chimerism in the majority of patients. Although most patients eventually suffered a hematologic relapse, this may have been delayed by therapy. 10 More recently, a randomized controlled trial sought to determine the role of AZA maintenance monotherapy in 187 patients with AML or MDS after allo-SCT; however, the study was closed early due to slow accrual over 8 years. While relapse free survival was comparable between those who received AZA and those who did not, stratification by number of AZA cycles received showed a trend toward improved relapse free survival in patients receiving more cycles of AZA therapy. 88 In addition to injectable azacitidine, oral azacitidine (CC-486) has also been used for post-transplant maintenance therapy. In a phase I/II dose-finding study involving 30 patients with AML or MDS (26 with AML) in CR after allo-SCT, maintenance therapy with CC-486 was associated with relatively low 1-year rate of relapse or disease progression at 21% of evaluable patients (6/28), as well as 10% grade III acute GVHD. 89 Given the potential reduction in GVHD provided by AZA-mediated expansion of Tregs, AZA has been used in combination with the immunomodulatory agent lenalidomide, which was found to be a well-tolerated salvage therapy in post-transplant relapsed AML. 90 Others have suggested combining hypomethylating agents with BCL-2 inhibitors, such as venetoclax, which may sensitize AML cells to the effects of hypomethylating agents. 91 This combination will be investigated as a maintenance therapy for high-risk, post-transplant AML in an upcoming phase II trial (NCT04128501). Finally, azacitidine and cytarabine have been found to induce expression of the MHC class II antigen HLA-DR in leukemic cells of patients with AML, 92 possibly overcoming a significant immunologic mechanism of post-transplant AML relapse following non-haploidentical transplantation.

| Histone deacetylase inhibitors in AML
Histone deacetylase inhibitors (HDACis), such as panobinostat, have also been used to treat AML after allo-SCT. In addition to inducing apoptosis and differentiation, HDACis are thought to possess important immunomodulatory effects, including increased expression of tumor antigens, MHC class I and II molecules, costimulatory molecules, and NK cell-activating ligands. [93][94][95][96] In an open-label, multicenter phase I/II trial, prophylactic panobinostat after allo-SCT was given to 42 patients with high-risk AML or MDS (37 with AML, 67% with active disease at the time of transplant) to determine the maximum tolerated dose. At 2 years, only 20% of patients had relapsed and 7% developed grade III acute GVHD. Of note, 18 patients (43%) received a median of two DLIs, as DLIs were allowed throughout the study at the discretion of the treating physician. 97

| DISCUSSION
Despite the ability of allo-SCT to successfully treat AML, posttransplant relapse remains a significant concern. Various immunotherapies are now being used for the treatment of relapsed disease after allo-SCT, but these same approaches may also prevent relapse. Given the high risk of relapse in a select subset of AML patients, the prophylactic use of certain therapies after allo-SCT may be especially beneficial in these patients. With the exception of those patients with a history of severe acute GVHD, severe infection, or need for immunosuppression aside from GVHD prophylaxis (eg, active autoimmune disease requiring treatment), there are no specific contraindications.
As with any therapy, the increased risk of complications among older patients, and those with comorbidities, must be balanced against the likelihood of post-transplant relapse in determining the risk-benefit of maintenance therapies. Since the median time to relapse after transplant in AML is 7.5 months, 98 therapies aiming to reduce the risk of relapse should be started relatively soon after transplant.
As both our understanding of the mechanisms underlying AML relapse following allo-SCT and our arsenal of therapeutic options continue to expand, our ability to harness the immune system after allo-SCT should become increasingly nuanced. A number of therapies highlighted in this review-including DLI, immune checkpoint inhibitors, FLT3 inhibitors, and hypomethylating agents-possess the theoretical ability to reverse or prevent recently characterized immunologic mechanisms of AML relapse following allo-SCT. Historically, the challenge has been to enhance the powerful graft vs leukemia effect without augmenting graft vs host disease. The post-transplant use of DLI and immune checkpoint inhibitors, as highlighted in this review, illustrate this difficult compromise. However, with our improved understanding of the immunologic mechanisms underlying relapse, it may now be possible to select prophylactic maintenance therapies that overcome these mechanisms without exacerbating GVHD.
An additional limitation in AML has been the lack of ideal target antigens, which hinders the use of ADCC, T cell engaging techniques, and CAR (chimeric antigen receptor) T-cells. Nonetheless, there are numerous ongoing trials investigating the role of CAR T-cells for AML, 99 with major targets including CD33, 100 CD123, 101 CLL-1, 102 and FLT3. 103 Similarly, T-cell engaging techniques are now being used in AML, 104 mostly targeting CD33 105 and CD123. 106 The role of these therapies in the post-transplant setting remains to be determined.
Ultimately, our enhanced understanding of the mechanisms underlying post-transplant relapse could help us to identify novel antigens and treatment strategies needed to overcome these mechanisms, prevent relapse, and treat relapsed disease.

AKNOWLEDEGMENTS
The authors would like to thank Brent Rane for his help creating the figure in this paper.