Recent Advances in Immune-Based Therapies for Acute Myeloid Leukemia

Abstract

Despite advancements, acute myeloid leukemia (AML) remains unconquered by current therapies. Evidence of immune evasion during AML progression, such as HLA loss and T-cell exhaustion, suggests that antileukemic immune responses contribute to disease control and could be harnessed by immunotherapy. In this review, we discuss a spectrum of AML immunotherapy targets, encompassing cancer cell-intrinsic and surface antigens as well as targeting in the leukemic milieu, and how they can be tailored for personalized approaches. These targets are overviewed across major immunotherapy modalities applied to AML: immune checkpoint inhibitors, antibody–drug conjugates, therapeutic vaccines, bispecific/trispecific antibodies, and chimeric antigen receptor (CAR)-T and CAR-NK cells.

Significance: Immune therapies in AML treatment show evolving promise. Ongoing research aims to customize approaches for varied patient profiles and clinical scenarios. This review covers immune surveillance mechanisms, therapy options like checkpoint inhibitors, antibodies, CAR-T/NK cells, and vaccines, as well as resistance mechanisms and microenvironment considerations.

Introduction

Acute myeloid leukemia (AML) is a heterogeneous disease characterized by an early myeloid differentiation block with immature blast expansion. Biologically, it is organized as a hierarchy that originates from leukemia stem cells (LSC), which are responsible for the initiation and maintenance of the disease and relevant clinical phenotypes, including chemo-resistance or immune-escape. Genetically, AML evolves from a hematopoietic stem/progenitor cell (HSPC) through a multistep process involving recurrent chromosome translocations and/or driver gene mutation. Mutations occur at variable frequencies within individual leukemias and lead to unique sub-clonal patterns and high inter- and intra-tumoral heterogeneity (ITH; ref. 1). Notably, ITH may evolve under environmental pressure [e.g., antileukemic treatments (2)], providing the substrate for establishing adapted phenotypes. ITH, including epigenetic, transcriptional, phenotypic, metabolic, and secretory heterogeneity (3), can itself modify the tumor microenvironment (TME).

The standard backbone treatment of AML for fit patients, unchanged since the 1970s, is the 7+3 regimen (7 days of cytarabine and 3 days of anthracycline infusion) followed by consolidation chemotherapy and/or allogeneic hematopoietic stem cell transplantation (allo-HSCT) in patients with high-risk of relapse. Lately, the definition of the AML genetic landscape has provided a list of driver genes for new drug development. Indeed, starting in 2017, the FDA has approved new drugs [e.g., the FLT3 inhibitors midostaurin, quizartinib, and gilterninb; gemtuzumab ozogamicin (GO) targeting CD33⁺ blasts; venetoclax targeting the pro-apoptoic BCL2 protein; glasdegib, a Hedgehog inhibitor and the IDH inhibitors ivosidenib, olutasidenib, and enasidenib] to treat newly diagnosed or refractory/relapsed (R/R) AML. These medications are revolutionizing initial treatment strategies for both fit individuals, who receive standard chemotherapy combined with targeted drugs based on their mutation profile [e.g., midostaurin for FMS-like tyrosine kinase 3 (FLT3)-mutated patients], and older or unfit patients. For the latter group, azacitidine or hypomethylating agents (HMA) in combination with venetoclax, either alone or alongside targeted drugs (such as magrolimab or gilteritinib, currently being evaluated in clinical trials), are emerging as new treatment standards (4). AML blasts, however, most often develop strategies to become resistant, highlighting the need for alternative and/or complementary therapeutic approaches. The groundbreaking achievements of immunotherapy in other tumor types make it a compelling candidate for investigation in the AML setting.

Is AML an Immune Target?

The first evidence that AML is an immune target came from observations of increased incidence of AML in patients with prolonged immune-suppressive regimens after solid organ transplantation (5), documentation of the graft versus leukemia (GVL) effect of the donor bone marrow in patients undergoing allo-HSCT (6), and identification of donor-derived T cells as one of the major drivers of the GVL effect. This work provided the rationale for the subsequent introduction of donor-derived lymphocyte infusion (DLI) to stimulate GVL for the prevention or treatment of relapsing AML after allo-HSCT (7). If the immune system controls AML growth, it is then plausible that during AML development leukemic blasts develop diverse strategies to evade the immune surveillance.

Unlike in solid tumors, genetic loss or transcriptional downregulation of HLA class I or II is uncommon in AML at diagnosis, whereas it becomes more prevalent after relapse, particularly following allo-HSCT. A seminal study demonstrated that 30% of relapsed AMLs after HLA-mismatched (haploidentical) transplants carried the genomic loss of the recipient-specific HLA haplotype, rendering donor T cells unable to recognize leukemic cells (8). A more recent report revealed high frequencies of HLA class I or II mutations, insertions/deletions, or losses in both mismatched (25%) or matched (50%) allo-HSCT (9). In the mismatched settings, both matched and mismatched HLA alleles were mutated, suggesting strong immune selection against alleles presenting the most immunodominant peptides. Notably, most patients without genomic HLA alterations show transcriptional downregulation of HLA class II and less frequently class I expression (9, 10), suggesting that impaired antigen presentation through HLA receptors is a critical mechanism for AML blasts to escape donor-derived T cells.

Donor-derived T cells recognize polymorphic peptides presented on patient cells via shared HLA molecules (minor histocompatibility antigens), which contribute to both graft versus host disease (GVHD) and GVL effects (11), suggesting a crucial role of HLA-restricted antigens in AML immune surveillance. Compelling evidence for antigen presentation as an immunosurveillance mechanism in AML comes from a recent study showing that LSC/AML blasts express HLA-restricted tumor antigens (nonmutated peptides, cryptic neoepitopes, and neoepitopes derived from common AML driver mutations) that can be recognized by preexisting memory T cells and correlate with clinical outcome (12).

Beyond restricting antigen recognition, AML utilizes additional strategies to evade the immune attack (Fig. 1). The AML bone marrow contains CD8⁺ T cells with exhausted and senescent phenotypes (13), whose number negatively correlates with overall survival (OS; ref. 14). Consistently, AML blasts can acquire features of immunosuppressive myeloid-derived suppressor cells (MDSC) and secrete molecules [e.g., arginase, indoleamine 2,3-dioxygenase (IDO), and NOS] that induce T- and NK-cell anergy (15). Notably, increased expression of the inhibitory molecules PDL1, CD276/B7-H3, and CD155/PVRL2 has been documented in 40% of AML cases after allo-HSCT (16).

Figure 1.

Main mechanisms of immune evasion of AML blasts. Blasts impair T and NK cell activity by exposing inhibitory T-cell ligands (e.g., PDL1 and CD86) or releasing the soluble form of NKG2DL, respectively. Moreover, they downregulate MHC-I/II exposure by hiding themselves from innate (DC and macrophage) and adaptive (T cells) immune cells. Macrophage-mediated clearance is prevented through the exposure of don’t eat me signals like CD47. AML blasts shape the immune microenvironment by releasing soluble factors such as IDO1, iNOS, adenosine, and arginase-1 (ARG1), which induce MDSC differentiation that in turn produces the same soluble factors with autocrine and paracrine activity, ultimately blocking T- and NK-cell antitumor activities. Different strategies have been developed to counteract blasts-mediated immune evasion: the main approach involves the usage of antibodies that target molecules involved in mediating AML immune evasion, such as mAbs against CD47, PD1/PDL1, and CTL4. Moreover, treatment with hypomethylating agents (HMA, e.g., 5-azacytidine and decitabine) has shown the ability to counteract HLAI/II downregulation in blasts. Lastly, targeting CD33 may have a dual cytotoxic activity against malignant cells and immunosuppressive MDSCs. This figure was created with Biorender.com.

Thus, AML blasts actively participate in the immune editing process, a potentially targetable vulnerability for immunotherapies, including immune checkpoint inhibitors (ICI), chimeric antigen receptor T (CAR-T) cells, monoclonal antibodies (mAb), and vaccinations.

Immune Checkpoint Inhibitors

ICIs are designed to block negative regulatory circuits between cancer and immune cells mainly in the TME. Consistently, significant clinical responses upon ICI treatment correlate with specific TME immune phenotypes. Because different TME changes are found at diagnosis, during treatment, or at relapse, identification of the most appropriate disease phase for treatment and accurate analysis of the TME are critical to predicting clinical response. Most studies on the interplay between ICI, TME, and cancer cells were conducted on solid tumors, where all the key cell players are spatially confined and their identity unambiguously defined. In AML, immune composition investigations are more challenging, because tissue organization is less defined and blasts share markers and functional properties with most innate immune cells, as they originate from the same progenitors.

CD8⁺ cytotoxic T-lymphocytes are critical for the establishment of antitumoral immunity and are suitable targets for ICIs. The extent of cytolytic infiltration is heterogeneous in AML, generally lower than in other hematological malignancies, and most frequently associated with complex karyotypes. By combining immune gene expression and digital-spatial profilings, AMLs were recently classified as either immune-depleted or immune-infiltrated, the latter characterized by RNA immune scores of highly immune-suppressed infiltrates (high expression of IFN-inducible immune checkpoints and immunotherapy targets as IDO1 and PDL1, secreted IFNγ, and TNF), chemotherapy resistance and enhanced probability of immunotherapy response (17). Similar results were obtained by direct phenotypic characterization of bone marrow immune cells, which showed increased frequencies of T cells expressing markers of exhaustion (regulatory T cell—Tregs) and/or immunosuppressive activity (VISTA, CTLA4, PD1, IDO1, TIM3, and LAG3) in patients with TP53 mutations, higher mutation rate and age, poor outcome, resistance to chemotherapy, and, most notably, increased benefit from allo-HSCT (18, 19).

Several clinical trials (Table 1) using ICIs approved for other cancer types have been initiated in the last years in R/R AML, AML at high-risk of relapse after allo-HSCT or other treatments, AML not eligible for allo-HSCT, or TP53- or FLT3-mutated AML. ICI alone showed, however, limited single-agent activity, with OS ranging from 10% to 20% (reviewed in ref. 20). Combinations with other therapies hold more potential. Anti-PD1 and CTLA4 antibodies are being tested in combination with antileukemic drugs that showed immunological effects in other tumor types, such as HMA (decitabine and azacitidine), which upregulate HLA and putative tumor antigens in different solid tumors, and the Bcl2-inhibitor venetoclax, which increases intratumoral effector T cells and ICI antitumor activity in preclinical models of colon cancer (21). As an example, nivolumab (an anti-PD1 antibody) and azacitidine gave an overall response rate in 56% of patients with elevated CD3⁺ T cells bone marrow infiltration in both HMA-naïve and pretreated patients (22). A more recent study using durvalumab (an anti-PDL1 antibody) in older patients, however, failed to demonstrate any additional clinical effect of azacitidine (23), evidencing an ongoing difficulty in predicting responders and the necessity to set up new biomarkers able to predict the clinical outcome. In alignment with this, Abbas and colleagues employed a single-cell-based multiplexed immune assay to establish a polyfunctional index for the CD4⁺ subset. This index serves as a predictive factor for the response of patients undergoing anti-PD1-based therapy (24).

Table 1.

Selected trials of ICI-based combination therapies.

Clinical trial	Target population	Response	Reported toxicity
NCT04284787 Phase II trial with Anti-PD1 Pembrolizumab in combination with AZA and venetoclax as frontline therapy	ND_AML and secondary AML in older patients not eligible for IC	Ongoing, not reported	Ongoing, not reported
NCT03390296 Multi-arm phase Ib/II trial Anti-PDL1 avelumab in multiple combination with AZA, venetoclax, PF-04518600, GO	R/R AMLs	Modest response in the avelumab-containing arms, except for one patient with extramedullary-only disease	No grade ≥3 related adverse events or DLTs were observed
NCT02768792 Phase II trial High-dose cytarabine followed by pembrolizumab (anti-PD1 Ab)	R/R AMLs	CRc rate: 38%; OS: 11.1 months. Higher OS (13.2 months) in patients with refractory/early relapse	Grade ≥3 irAE were rare (14%) and self-limiting
NCT02890329 Phase I trial Decitabine in combination with ipilimumab (IPI) (anti-CTLA4 antibody)	R/R AML and MDS in both post-HSCT and transplant-naïve settings	Highest CR/Cri response (64%) in patients with transplant naïve. Possible usage as bridge treatment to transplant	Grade ≥3 treatment-emergent adverse events. irAE. DLT at higher IPI doses (>10 mg/kg)
NCT02275533 randomized phase II trial Nivolumab (anti-PD1 Ab) as maintenance therapy	AML in first remission (CR/CRi) not candidate for HSCT	Modest OS improvement	Increased AEs and SAEs, though manageable
NCT04150029 Phase II, single-arm trial MBG453 (anti-TIM3 Ab) in combination with AZA and venetoclax	ND-AML ineligible for IC	Ongoing	TRAEs grade ≥3 in more than 25% of pts. AEs led to sabatolimab interruption in six patients
NCT02775903 Phase II trial AZA + durvalumab (anti PDL1 Ab)	Elderly patients with AML	The treatment was feasible but did not improve the clinical outcome	Durvalumab-related AEs were reported by 71.1% of patients. Azacitidine-related AEs were reported in about 80% of the patients
NCT04623216 Phase Ib/II trial Sabatolimab alone or in combination with AZA	AML in CR with positive MRD (MRD+) after alloSCT.	Preliminary data are promising with 30% of the pts at 400 mg still in CR after more than 1 year of treatment	Overall well tolerated. There were no cases of GvHD or immune-related AEs

Abbreviations: AE, adverse events; DLT, dose-limiting toxicity; irAE, immune-related adverse event; SAE, serious adverse event, TRAEs, treatment-related AEs.

Cytarabine sensitizes AML blasts to T-cell-mediated killing and was used in combination with pembrolizumab (an anti-PD1 antibody) in newly diagnosed and R/R AML, with good response rates in the latter (overall response rate, 46%; complete response rate, CRR, 38%; median overall survival, mOS, 11.1 months), which correlated with higher infiltration of bone marrow CD8⁺ T cells expressing TCF-1 after immunotherapy (25).

In murine models, rare dormant blasts with leukemia-initiating potential survive antileukemic immune responses by overexpressing PDL1 or CD80, suggesting that ICIs may also be exploited in the measurable residual disease (MRD)–positive setting. Nivolumab, as a single agent, was employed for the maintenance of high-risk patients with AML who responded to induction therapy and were ineligible for allo-HSCT, yielding only modest effects on MRD eradication and disease-free survival (26). In a more recent report, no benefit was appreciated in 80 patients treated with nivolumab upon first remission (27), suggesting a poor response and a general difficulty in identifying patients who would benefit from this type of treatment. These negative results have been recently confirmed in a study (NCT02275533) where nivolumab maintenance failed to improve progression-free survival and OS while increasing adverse events. Further studies are needed to stratify patients for immune checkpoint protein (ICP) expression at different phases of the disease and find the best ICI combinations.

Sabatolimab (MBG453) targets TIM3 on immune cells, leukemic blasts, and LSCs and is under evaluation for the treatment of high-risk MDS and newly diagnosed AML not eligible for intensive chemotherapy or HSCT (STIMULUS clinical trial). In a preliminary phase I/b study (NCT03066648) in combination with HMAs, sabatolimab showed good safety and tolerability and some durable clinical benefits (28). Another study is ongoing in AML (NCT04150029) in combination with HMA and venetoclax (23).

ICIs are also under development in the post-alloHSCT setting. Donor T cells recognize leukemic blasts (GVL) and also normal host cells triggering the GVHD. Loss of GVL because of immune escape, including increased expression of ICPs, is a key element for AML relapse. Notably, increasing frequency of exhausted T cells after allo-HSCT correlates with relapse (29), whereas relapsed AML blasts show increased expression of ICPs and/or downregulation of class II HLAs, the latter because of promoter DNA hyper-methylation of the class II regulator CIITA (16). Interestingly, loss of class II can be reverted by IFNγ or HMAs, which, however, also promote ICP expression, again underlying the need for careful evaluation of patient-specific immune escape mechanisms. Several trials have been initiated to investigate the impact of ICIs-based treatments in patients with AML after relapse or at high risk of relapse following allo-HCST. In most cases, anti-PD1 was used alone or in combination with anti-CTLA4 (NCT03600155, NCT02846376, NCT01822509, and NCT04361058), azacitidine (NCT04128020), or CD25/Treg-depleted DLI (NCT03912064). However, preliminary results showed only modest efficacy while exposing the risk of immune-mediated toxic effects and GVHD reactivation (grade 3/4 GVHD; ref. 30). Recently, a phase I study (NCT02890329) has shown promising results with the anti-CTLA4 ipilimumab combined with decitabine for relapsed MDS/AML in both post transplant and transplant-naïve settings (31).

ICIs’ side effects remain critical and challenge the design of new combinatorial studies (Table 1). In 20% to 40% of patients, ICIs lead to a variety of immune-related adverse events (irAE), sometimes difficult to recognize, involving any organ and/or system, and lasting for months after therapy discontinuation, with rates and intensity markedly higher in patients after HSCT (32). There is indeed an urgent need for empirical studies on the irAE impact to establish detailed guidelines for immunotherapy of AML in different treatment settings.

In conclusion, using ICIs alone in AML has shown limited success, likely because of the prevalent exhaustion/senescence phenotype of infiltrating CD8⁺ T cells. However, ongoing clinical trials in combination with other drugs, particularly demethylating agents, are yielding more promising results. Careful characterization of the immune composition and ICP expression remains crucial to improve AML patient risk-stratification and identify the most suitable and well-timed therapeutic approach for each patient.

Inhibitors of Macrophage Immune Checkpoints

Emerging evidence from preclinical models suggests a critical role of macrophages in the AML immune-control. AML blasts overexpress CD47, a don’t eat me signal that interacts with the signal regulatory protein alpha on macrophages to prevent phagocytosis (Fig. 1). High levels of CD47 are a poor prognostic factor in AML, whereas anti-CD47 blocking antibodies activate blast phagocytosis in preclinical models, leading to AML eradication (33). Further, although CD47 is expressed in cancer and normal cells, only cancer cells are targeted by macrophages, because of the co-expression of eat me signals (calreticulin, CRT) that favors macrophage recognition and clearance. Patient treatment with anti-CD47 antibodies (magrolimab) as a single agent yielded modest clinical responses, with the appearance of increased T-cell infiltrates in the few responders (34). A phase Ib trial was initiated to evaluate the combination of magrolimab and the anti-PDL-1 atezolizumab in patients with R/R AML (NCT03922477), which was discontinued because of the occurrence of adverse events and lack of clinical response.

Azacitidine induces CRT exposure on AML blasts leading to their almost complete clearance in xenograft models. The azacitidine–magrolimab combination was thus tested in a phase Ib trial (NCT03248479; Table 2) in patients with AML previously untreated and ineligible for induction chemotherapy, including 80% of TP53-mutant AML. The treatment was relatively well tolerated and induced 32% of CR with higher clinical efficacy and promising results in patients with TP53-mutated (CRR 40%; ref. 35). Unfortunately, two phase III trials examining the frontline combination of magrolimab and azacitidine in patients with AML and TP53 mutations (NCT04778397), or the triple combination azacitidine–magrolimab–venetoclax (NCT05079230), were discontinued because of futility, and the FDA placed on full clinical hold all magrolimab studies in MDS and AML (Table 2).

Table 2.

Selected trials of magrolimab-based combination therapies.

Clinical trial	Target population	Response	Reported toxicity
NCT04435691 Phase Ib/II trial Magrolimab + AZA in combination with venetoclax	ND-AML, both primary and secondary, and R/R AMLs	CR rate: 49% in ND_AML and 19% in patients with R/R AML	All had at least 1 AE and 90% had at least 1 grade 3 or higher AE. There were no immunologic AEs or discontinuations due to TRAEs
NCT05079230 Phase III trial Magrolimab + AZA in combination with venetoclax	Untreated AML ineligible for IC	Trial in progress
NCT04778397 Phase III trial Magrolimab + AZA vs. venetoclax + AZA or IC	Untreated TP53 mutant AMLs	Discontinued No survival benefits	Not reported
NCT03248479 Phase Ib trial Magrolimab + AZA	Untreated AML ineligible for IC. Most patients (82%) had TP53 mutations	CR: 32.2%. The median OS in TP53-mutant and wild-type patients were 9.8 months and 18.9 months, respectively	About 50% of treated patients had TRAEs. Overall, magrolimab with azacitidine was relatively well-tolerated

Abbreviations: AE, adverse events; TRAEs, treatment-related AEs.

Considering the reported clinical challenges of the available ICIs in terms of both toxicity and efficacy, it is critical to further investigate molecular mechanisms of AML blast immune-evasion to find alternative therapeutic approaches. For example, TGFβ1 secretion in the TME can suppress adaptive and innate immunity, whereas TGFβ inhibitory therapies can restore anticancer immunity in solid tumors (36) and AML preclinical models (37). Metformin favors AML blast clearance by activating the expression of molecules that induce NK and T-cell activities (NKG2DL and ICAM-1; ref. 38). Finally, the cell cycle status of cancer cells may influence their immunogenic properties, as shown by the intrinsic property of quiescent cells to escape immune surveillance (39), or the secretion by senescent cancer cells of soluble factors favoring immune infiltration and anticancer immune responses (40). Thus, inhibition of TGFβ signaling, prevention of quiescence, or induction of senescence may represent further strategies to promote tumor immune clearance.

Inhibition of MDSC Immune-Suppressive Functions

MDSCs and multipotent mesenchymal stem cells are major components of the immunosuppressive TME of AML. They impair T-cell activation through exposure to PDL1, or induce apoptosis of T cells by secretion of arginase-1, nitric oxide, reactive oxygen species, or IDO (Fig. 1). Patients with AML show increased numbers of MDSCs (CD33⁺, CD11b⁺, and HLA-DR^low/neg) that remain high in patients with poor response to chemotherapy. Notably, AML-derived extracellular vesicles induce monocytes to acquire the MDSC phenotype, whereas patients with higher MDSCs-like blasts show a worse prognosis (41).

Because MDSCs express high levels of CD33, treatment with bispecific CD33/CD3 antibodies would target both MDSCs and CD33⁺ blasts, favoring antitumor T-cell activation in different tumor types, including AML and MDS (42, 43). A phase I/II clinical trial has been recently initiated to evaluate a novel tri-specific molecule (GTB-3550 TriKE, see “Monoclonal Antibodies, Antibody-Drug Conjugates and Bispecific Antibodies”) for the treatment of CD33⁺ high-risk MDS, R/R AML or advanced systemic mastocytosis (NCT03214666). MDSC-based cellular therapies showed beneficial effects in the allo-HSCT setting in murine models, by counteracting GVHD without affecting GVL (44), suggesting a role for MDSC-based cellular therapies in the prophylaxis or treatment of GVHD post-allo-HSCT (45). Further studies in AML preclinical models are required to elucidate the impact of different approaches to MDSC suppression in different clinical settings.

Adoptive Cell-Based Immunotherapy

Given the growing established role of T or NK cells in mediating the GVL of allo-HSCT, in recent years different types of T- or NK-based cellular therapies have been developed and tested in AML preclinical or clinical settings.

Nonengineered T-cell Infusion

DLI infusion has been employed to trigger GVL for managing or preventing relapsed AML post-allo-HSCT (7). However, the prognosis of DLI-treated patients remains dismal (2 years OS 25%), and the benefits come with the burden of toxicity such as GVHD. Some novel targeted and immunomodulatory agents (e.g., HMA, FLT3, and BCL2 inhibitors), have been used in combination with DLI, aiming to enhance the GVL effect (46). Moreover, manipulation of the DLI graft through cell selection (e.g., donor NK cells) or cell engineering (e.g., donor CAR-T cells) has shown potentially superior antitumor effects but less GVH effect than conventional DLI in clinical trials, as discussed in the upcoming sessions. On the other hand, although the infusion of tumor-infiltrating lymphocytes has shown feasibility and efficacy in solid tumors, to the extent of being recently approved by the FDA for the treatment of advanced melanoma (47), it remains unclear whether this approach could be effective for the treatment of hematological malignancies, including AML. In particular, the lack of tumor-infiltrating lymphocyte research in AML treatment may be attributed to the disruptions in T-cell function that result from leukemic disease (e.g., T-cell exhaustion, increase of regulatory T cells, and production of soluble factors that inhibit T-cell proliferation; ref. 48).

TCR-T-cell–Based Immunotherapy

T-cell receptor (TCR)-T cell–based therapy involves engineering T cells to express specific TCRs targeting tumor peptides through MHC-dependent mechanisms. Various types of target antigens have been studied and validated for AML (49): (i) lineage-restricted antigens (e.g., Wilms tumor 1, WT1), (ii) overexpressed antigens (e.g., telomerase reverse transcriptase, TERT), (iii) cancer-testis antigens (e.g., preferentially expressed antigen in melanoma, PRAME, and NY-ESO-1), (iv) minor histocompatibility antigens (polymorphisms differing between transplant donor and recipient, e.g., HA1 and HA2), and (v) neoantigens derived from nonsynonymous somatic mutations (e.g., NPM1). TCR-T-cell products targeting these antigens have demonstrated promising effectiveness in preclinical studies. However, progression to early-phase clinical trials is limited by the requirement of precise HLA matching, significantly restricting patients’ recruitment. Despite this limitation, WT1 TCR-T-cell therapies have entered in clinical stage, showing successful in vivo expansion and persistence of the infused cells (50). The adoptive transfer of WT1 TCR-T cells in 12 poor-risk patients with AML with no evidence of disease 28 days post-HSCT led to disease control (event-free survival 100% at a median of 44 months postinfusion) with an excellent safety profile (one reported case of severe acute GVHD and two cases of chronic GVHD; ref. 51). Other TCR-T cells are under evaluation in different phase I/II trials (e.g., HA1 and HA2 TCR-T in the NCT03326921 and NCT05473910 and PRAME TCR-T in the NCT03503968) showing preliminary efficacy (52).

CAR-T-cell Therapies

CAR-T cells are autologous T cells genetically engineered to express on their surface a chimeric antigen-targeting receptor linked to an intracellular T-cell signaling domain. Upon reinfusion, CAR-T cells bind to their specific cancer antigens, triggering an intracellular signaling cascade leading to T-cell activation, proliferation, and elimination of target cells, either directly or through activation of other immune cells. CAR-T cells have changed the treatment paradigm for R/R B-lineage acute lymphoblastic leukemia, non-Hodgkin lymphomas, and multiple myeloma. However, CAR-T-cell therapy has not been successful in AML so far.

AML blasts express several surface antigens that have been proposed as CAR-T-cell targets, such as CD33, CD123, CD70, and CD47. Ideal AML targets should be highly expressed on LSCs and/or leukemic blasts, essential for their survival, yet poorly or not expressed by normal HSPCs, myeloid progenitors, and activated T cells, as well as on other cell types. A few antigens have been investigated for AML CAR-Ts, all of which partially meet these requirements (Fig. 2).

Figure 2.

Targets of CAR-T cells in AML. Numerous targets reported in preclinical models are shown on the bottom/left side. Targets under investigation in clinical trials (listed in Table 3) are shown on the top/right side. PR1 and WT1 are depicted as peptides on the cell surface complexed with HLAs. This figure was created with Biorender.com.

CD123, the IL3 receptor α subunit, is more expressed by LSCs and AML blasts than normal HSPCs. In xenograft models, anti-CD123 CAR-Ts (CART123) exhibited potent antitumor activity with marked myelotoxicity. Surprisingly, CARs with VH and VL chains from different CD123-specific mAbs presented lower off-tumor toxicity. The first phase I trial that used the allogeneic product UCART123 was suspended after the death of one patient 9 days postinfusion, because of cytokine released syndrome (CRS) and vascular toxicity (NCT04106076). When reevaluated in the AMELI-01 trial, UCART123v1.2 showed clinical activity in 4/16 of the patients (53). To reduce off-target toxicities, CAR-Ts electroporated with anti-CD123 CAR mRNA were evaluated (NCT02623582), showing no clinically apparent vascular, neurological, or hematological toxicity, and lacking antitumor effects. Subsequent trials relied on the use of lentivirally transduced CAR-T targeting CD123. In phase I trials NCT03114670 and NCT02159495, CART123 showed reversible and manageable side effects, with some evidence of clinical benefit. In particular, among seven highly pretreated refractory patients with AML, two patients achieved CR and proceeded with a second allo-HSCT, whereas three showed blast reduction (NCT02159495). Lastly, a recent preclinical study demonstrated that azacitidine pretreatment increases AML blasts immunogenicity, improving cytotoxicity and efficacy of anti-CD123 CAR-Ts, paving the way for clinical trials combining HMAs with anti-CD123 CAR-Ts (54).

CLL-1, or human C-type lectin-like molecule-1, is highly expressed in more than 80% of patients with AML and, like CD123, less expressed in normal HSPCs. CAR-Ts against CLL-1 showed low toxicity and induced CR in 3/4 pediatric patients with R/R AML (NCT03222674) and a 70% CR rate in a phase I trial on 10 adult patients (55). Lastly, bicistronic CAR-Ts targeting CD123 and CLL-1 showed appreciable anti-AML activity in murine models. Moreover, 123CL CAR-Ts can be eliminated by a natural safety switch without targeting normal HSPCs (56).

CD33, a transmembrane protein of the sialic acid-binding immunoglobulin-like lectin family, is expressed on the surface of myeloid blasts in approximately ∼90% of patients with AML, as well as on normal myeloid cells downstream of the common myeloid progenitor. The monoclonal immune-conjugate anti-CD33 (GO), in clinical use since 2000 with some benefits in selected AML subtypes (see “Monoclonal Antibodies, Antibody-Drug Conjugates and Bispecific Antibodies”), was used to engineer anti-CD33 CAR-Ts, which were shown to possess in vitro and in vivo anti-AML activity in several preclinical studies. In a phase I trial on refractory AML (NCT01864902), anti-CD33 CAR-Ts induced significant clinical responses followed by profound myeloablation and rapid disease progression. Currently, there is only one actively recruiting phase I/II trial with anti-CD33 CAR-Ts in children and young adults with R/R AML (NCT03971799). Expression of CD33 by normal myeloid progenitors represents a major obstacle to the use of anti-CD33 CAR-T, unless combined with allo-HSCT.

FLT3 receptor, although present on normal HSPCs, is expressed by AML blasts and is essential for their survival and proliferation, especially with activating mutations such as internal tandem duplications or tyrosine kinase mutations. Preclinical studies on FLT3-mutated AML showed that FLT3 inhibitors induce FLT3 upregulation, exposing blasts to the recognition and elimination by anti-FLT3 CAR-Ts. Though this therapeutic approach is potentially myeloablative and may require reconstitution of the hematopoietic system (57), its usage followed by allo-HSCT may be a viable strategy for high-risk AML. The clinical feasibility of anti-FLT3 CAR-Ts is currently investigated in the NCT05023707 and NCT05017883 trials.

Other CAR-Ts targeting AML-antigens are also expressed by myeloid and lymphoid progenitors or monocytes. CAR-Ts against CD70, a ligand for CD27, were well tolerated in preclinical studies, yet lacked significant antileukemic activity (58). Antileukemic activity increased when engineered to stabilize CD70-binding and combined with azacitidine, which increases CD70 density on target cells. CAR-T cells against the folate receptor β showed significant antileukemic activity in preclinical settings. Being folate receptor β highly expressed in monocytes, these CAR-Ts should be carefully evaluated for off-tumor toxicity (59). CAR-Ts against the Lewis receptor Y, a carbohydrate tumor-associated antigen related to blood group members, gave encouraging results in both preclinical and clinical settings (60). CAR-Ts against CD7, a transmembrane glycoprotein expressed by T and NK cells, but absent on healthy myeloid cells, were gene-edited to lose CD7, inducing prolonged mice survival without signs of toxicity (61). Unlike antibody-based CARs, NKG2D-CAR specificity is conferred by the NKG2D receptor. NKG2D is expressed by cytotoxic lymphocytes and recognizes oncogenic stress-induced ligands MIC-A and MIC-B on neoplastic cells, triggering antitumor responses. Despite successful preclinical studies, a phase I trial on 12 patients with AML and MDS showed objective responses in only one AML patient (NCT02203825; ref. 62). CAR-Ts against CD38 and CD44v6 glycoproteins (63, 64); TCR-mimicking CARs recognizing HLA-complexed peptides of proteinase 1 and WT1 (65), leukocyte immunoglobulin-like receptor-B family LILRB4 (66) and NOT-gated approach for CD93 (67) also showed encouraging results in preclinical models.

These approaches, despite their potential and the numerous ongoing trials (Table 3), have been, so far, of limited clinical relevance, due to specific disease-related CAR-T challenges (e.g., heterogeneous disease biology, lack of a unique targetable antigen, and immune exhaustion).

Table 3.

International CAR-T therapy clinical trials for patients with AML.

Target	Clinical trial
ADGRE2	NCT05463640* Phase I, NCT05748197 (ADGRE2/CLEC12) Phase I
CD19	NCT03896854 and NCT04257175, Phase I and II
CD327/Siglec-6	NCT05488132 Phases I and II
CD38	NCT05239689 Early phase I, NCT03291444 Phase I
CD4	NCT06197672 Phase I
CD7	NCT05454241 Phase II, NCT05377827 Phase I, NCT05907603 Early Phase I
CD70	NCT04662294 Early phase I, NCT05907603 Early Phase I
CLEc12a (CD371)	NCT06017258 Phase I
IL1RAP	NCT04169022**, NCT06281847* Phase I and II
NKG2D ligands	NCT04658004* Early phase I, NCT02203825 Phase I
FLT3	NCT05017883, NCT05023707 Phase I and II, NCT05445011 Phase I, NCT05432401 Early phase I
CD33	NCT04835519 Phase I and II, NCT06326021 Phase I, NCT05672147 Phase I, NCT05984199 Phase I, NCT05942599 Phase I (BE-CAR33), NCT05945849 Phase I (with CD33KO HSCs), NCT05105152 Phase I, NCT04835519 Phase I and II, NCT05473221* Phase I, NCT05445765* Phase I, NCT03291444 Phase I
CLL1	NCT05467202* Phase I, NCT04884984 Phase I, NCT04923919 Early Phase I, NCT04219163 Phase I, NCT05252572 Early phase I
CD123	NCT02159495*** Phase I, NCT04318678*** Phase I, NCT03585517** Phase I, NCT04230265 (UniCAR02-T) Phase I, NCT04265963 Phase I and II, NCT04272125 Phase I and II, NCT04599543* Early Phase I, NCT04803929 Phase I, NCT05949125 Phase I (Allo-RevCAR01-T), NCT06125652 (Tim3) Phase I and II, NCT03190278 Phase I (UCART123v1.2)
CLL1/CD33	NCT05467254* Phase I
CLL1, CD33, CD38 and/or CD123	NCT05995041 Phase I (Universal CAR-T)

Target population: R/R AML, except for the NCT04169022, which has an arm with patients with AML at diagnosis. Trials are ongoing (data not reported), unless specified: *not yet recruiting; **completed, no available information; ***active, not recruiting.

Minimizing On-target, Off-tumor Toxicity of CAR-T Cells

Leukemia antigens shared with myeloid progenitors may lead to myeloablation, adding toxicity to the CAR-T side effects, such as CRS and neurotoxicity. To increase the CAR-Ts therapeutic window, different approaches have been developed: (i) Through specific gene or single-base epitope editing, healthy myeloid cells may become resistant to CAR-Ts while preserving the protein function. This approach has been validated for CD33 (68), which is now being tested in a phase I clinical trial (NCT05945849), whereas CD45 (69), cKIT, CD123, and FLT3 (70) provided a preclinical rationale to combine autologous gene-edited HSPCs transplantation followed by CAR-Ts. (ii) Logic gating, making antitumor response conditional on combinatorial recognition of two ligands. In the NOT-gated approach, antileukemia CAR is coexpressed with an inhibitory CAR specific for an antigen restricted to healthy cells. NOT-gated CAR-T cells are thus programmed to kill only those target cells that have the leukemia marker and lack healthy cell markers (71). IF-BETTER CAR-T relies on a receptor design that combines a sensitivity-tuned CAR with a targeted chimeric costimulatory receptor. Therefore, IF-BETTER CAR-T can selectively recognize different antigen densities between AML blasts and healthy myeloid cells allowing specific killing. This approach has been proved successful in preclinical models where IF BETTER ADGRE2 and CLEC12A CAR-T recognize only ADGRE2^high/MedCLEC12A^pos LSCs, thus limiting off-target cytolysis on ADGRE2^low CLEC12A^neg HSPCs (72). (iii) Tuning of CAR-T activities and safety switches are novel approaches to reduce off-target killing. These strategies rely on specific CAR design that allows for cytotoxicity modulation or CAR-T suicide, using specific mAbs, binders, or drugs, to tune CAR-Ts activity based on disease burden and side effects (71). In the approach described by Cartellieri and colleagues, a universal CAR-T (uniCAR) was combined with a specific CD123-targeting module with a short half-life, enabling a rapid switch-off of the uniCAR system (73).

Reduced Autologous T-cell Fitness and Manufacturing Time

The collection of adequate numbers of autologous T cells is often impractical in patients with AML because of the suppressed normal hemolympoiesis. In addition, the long CAR-T manufacturing time misaligns with the aggressivity of relapsing AML. Therefore, genetically modified off the shelf allogeneic CAR-Ts from healthy donors may represent a viable solution, which, however, poses challenges in CAR-Ts manufacturing to decrease the risk of GVHD. Garner and colleagues, by using Caribou’s CRISPR hybrid RNA-DNA guides in combination with Cas12a, made five edits in the manufacture of allogeneic anti-CLL-1 CAR-Ts to enhance their persistence and potentially achieve durable therapeutic responses (74).

Immunosuppressive Leukemic TME

The AML microenvironment is characterized by T-cell exhaustion, increase of regulatory T cells, and production of soluble factors that inhibit T-cell proliferation (48), all of which likely dampen CAR T-cell effectiveness. Therefore, approaches combining CAR-T with mAbs or ICIs could potentially increase CAR-T persistence and efficacy within an immunosuppressive microenvironment (75).

In conclusion, CAR-Ts are challenging but promising platforms for achieving clinical response in R/R AML. More effort must be made to define the CAR-T clinical setting and timing, and their interplay with haplo- or allo-HSCT to improve disease control while preserving adequate immune function (70, 76).

NK-Cell Therapies

NK cells are innate lymphocytes that recognize virally infected or transformed cells in an MHC-unrestricted manner through specific activating and inhibitory receptors. Depending on the type and extent of receptor activation, NK cells can deliver cytotoxic or protective signals. Normal cells are spared because of the binding and activation of killer cell immunoglobulin-like receptors on NK cells and MHC-I molecules on the target. Cells that downregulate MHC-I, as cancer cells frequently do, are attacked by NK cells, provided that they also express activating receptors. The property of NK cells to recognize their targets in an MHC-unrestricted manner candidates them as universal immunotherapy effectors in the allogeneic context, bypassing some of the main limits of autologous CAR-Ts.

NK cells have been used to treat several cancer types in preclinical and clinical settings, showing potential therapeutic efficacy. In AML, NK cells were used to prevent relapse after chemotherapy and/or allo-HSCT (NCT01898793, NCT01853358), showing safe infusion and reduced risk of AML relapse and GVHD (77). However, these studies also raise concerns about the clinical use of NK cells in AML. First, they showed limited in vivo survival (∼2 weeks), although high numbers are needed to ensure a favorable balance between effector and target cells. Novel cell culture protocols have been developed to expand NKs ex vivo, though the procedure remains complex. Secondly, AML blasts can suppress NK function, by inducing highly immunosuppressive microenvironments or by deregulating the expression of inhibitory NK cell ligands. New methodologies to harness the antileukemic potential of NK cells are emerging, such as combined treatments with cytokine, CAR-modified NK cells, or therapeutic targeting of TME inhibitory receptors (78, 79).

Adoptive Transfer of Naked NK Cells

The first approach to enhance NK cell immunosurveillance was the infusion of NK cells taken from patients’ peripheral blood and expanded/activated ex vivo using interleukin 2 (IL2) or IFNγ (autologous NK cells). Despite no significant clinical responses, this approach stimulated interest in allogeneic NK cell transfer (80). The potential benefits of NK cells were first described in the T-cell–depleted haplo-HSCT setting, where the anti-AML effect of allogeneic NKs was attributed to the alloreactivity between killer cell immunoglobulin-like receptors on donor NKs and HLA-I on recipient cells. A recent phase I clinical trial with 13 patients with myeloid malignancies showed a ∼70% CR rate with high-dose chemotherapy followed by repeated infusions of haploidentical NK cells. Another study reported CR in 5/19 patients with R/R AML infused with IL2-activated haploidentical NK cells (81). IL2 also increases Tregs, thus antagonizing the cytotoxic potential of NK cells. To avoid this hazard, haploidentical NKs were given together with IL15, which induced potent proliferation of both NK and CD8⁺ cells and 32% CR in a phase I/II trial on R/R AML (NCT01385423). Thus, NK cell therapy is emerging as a potentially powerful therapeutic tool for AML, though challenges persist (autologous vs. allogeneic source, optimal in vitro expansion and infused quantity, proper timing of NK cell administration).

CAR-NK Cells

Several preclinical studies have demonstrated that NK cells can be transduced with CARs. Administration of anti-CD123 CAR-NK cells to human AML xenografts showed high antileukemic activity with negligible toxicity, as compared with anti-CD123 CAR-Ts (82). Currently, four clinical studies are ongoing to evaluate the efficacy of anti-CD33 CAR-NK cells in AML (NCT05215015, NCT04623944, NCT05247957, NCT05008575). Table 4 reports the pros, cons, and open issues of adoptive TCR-T versus CAR-T versus CAR-NK cells-based therapies.

Table 4.

Pros, cons and open issues of adoptive cell-based therapies.

	TCR-T cells	CAR-T cells	Naked NK cells	CAR-NK cells
Pros	• Potent cytotoxic activity • High specificity • Reduced risk of GVHD and CRS	• Potent cytotoxic activity • Long-lasting response • High specificity and affinity • Generation of a strong anticancer immunity • High response rate in R/R AML or in patients ineligible for allo-HSCT	• Innate ability to recognize cancer cells in an MHC-unrestricted manner • Reduced risk of GVHD and cytokine storm • May be used as an “off-the-shelf” product	• Increased specificity and affinity compared with naked NK cells • Potent cytotoxic activity • Reduced risk of GVHD and cytokine storm • Potential for “off-the-shelf” use, easily accessible
Cons	• MHC-restricted killing • High manufacturing cost and time • Immune suppression from AML cells	• Risk of CRS • Risk of neurotoxicity • High manufacturing cost and time • Lack of AML-specific antigens (myeloablation) • Immune suppression from AML cells	• May require high doses to achieve efficacy • Ex vivo expansion procedures are complex • AML can inhibit NK cells function	• May require ex vivo expansion • Short half-life • Lack of AML-specific surface markers • Limited clinical data in AML
Open issues	• Feasibility and safety of repeated TCR-T administration • Improve accrual in clinical trial • Generation of “off the shelf” TCR-T • Evaluation of the potential of combinatorial therapies	• Safety improvement and toxicity reduction • Generation of CAR-T cells from healthy donors to overcome long manufacturing time • Development of strategies for CAR-T-cell delivery • Overcome resistance mechanisms (antigens downregulation, T-cell exhaustion) • Evaluation of the potential of combinatorial therapies • Possible combinations with allo-HSCT to reconstitute the hematopoietic system • Generation of an artificial AML-specific antigen by genetic editing of allo-T cells	• Development of strategies to enhance efficacy (adoptive transfer of cytokine-induced therapies) and persistence • Feasibility and safety of repeated NK cell administrations • Enhancement of NK cells production • Therapeutic targeting of TME inhibitory receptors	• Optimization of engineering and manufacturing processes • Identification of optimal combinatorial therapies • Evaluation of long-term safety and efficacy in clinical trials

Abbreviations: CRS, cytokine release syndrome; GVHD, graft versus host disease.

Monoclonal Antibodies, Antibody-Drug-Conjugates, and Bispecific Antibodies

Unconjugated mAbs bind specific antigens on target tumor cells and activate innate immune effector functions and complement through their engagement with Fc receptors (antibody-dependent cell-mediated cytotoxicity, antibody-dependent cell-mediated phagocytosis or complement-dependent cytotoxicity). To enhance antitumor activity, antibody–drug conjugates (ADC) have been generated by connecting mAbs to cytotoxic agents through a linker. Upon target-antigen binding, ADCs are internalized and release the conjugated cytotoxic drug, minimizing systemic toxicity.

CD33. GO, the only FDA-approved mAb in AML, is a humanized anti-CD33 mAb conjugated to a highly toxic natural product calicheamicin, which exerts its toxic effect by double-strand DNA cleavage. After being used in elderly patients with relapsed CD33⁺ AML ineligible for chemotherapy, GO was withdrawn in 2010 because of hematologic and hepatic toxicities in younger patients with newly diagnosed AML, in the absence of documented efficacy (83). Because CD33 is rapidly reexpressed after antibody binding, many studies have used lower and serial doses of GO to maintain efficacy while minimizing toxicity. Better efficacy of fractionated GO was shown in a phase III trial (ALFA-0701) involving 271 patients with de novo AML and in the phase II MyloFrance-1 trial, showing a 33% remission in relapsed AML (84). Best efficacy in terms of OS, however, was obtained in favorable risk AML, with almost absent effects in adverse risk AML. The use of fractionated GO was then reapproved by the FDA alone or with chemotherapy for de novo AML and as a single agent for CD33⁺ R/R AML (84). A recent randomized trial confirmed its efficacy when added to first-line intensive chemotherapy in adults with NPM1-mutated AML (85). Overall, a meta-analysis of several clinical studies showed that adding GO to standard chemotherapy improves relapse-free and OS in the favorable and intermediate cytogenetic risk groups (86). Other anti-CD33 immunoconjugates are under clinical evaluation: (i) Vadastuximab talirine (SGN-CD33A), conjugated to the thiol-containing maytansinoid derivative AVE9633, which was tested in a phase I trial in combination with HMAs in older patients with untreated AML (NCT02785900), showing higher remission rates (∼70%) than HMAs, but also higher toxicity (87); and (ii) Lintuzumab, an anti-CD33 mAb linked to a particle-emitting bismuth-213 (²¹³Bi) or actinium-225 (²²⁵Ac), which showed objective remission after low-dose cytarabine in ∼28% untreated patients with AML (NCT02575963). The association of lintuzumab ²²⁵Ac with salvage chemotherapy regimen in 15 R/R patients with AML resulted in 67% CR (88).

CD123. The tagraxofusp immunotoxin, which targets the IL3 receptor CD123, consists of human IL3 fused to a truncated diphtheria toxin, was approved by the FDA and the European Medicines Agency for the treatment of adult and pediatric patients with blastic plasmacytoid dendritic cell (DC) neoplasm, which overexpresses CD123 in most cases (89). CD123 is a potential target also for AML. Talacotuzumab (CSL362) is a fully humanized anti-CD123 mAb with enhanced NK cells-mediated cytotoxicity, which showed encouraging results in two phase I studies (NCT01632852 and NCT01272145) on R/R and high-risk AML (90). However, a phase II/III trial comparing decitabine alone or in combination with Talacotuzumab in elderly with de novo AML was terminated because of lack of efficacy (NCT024721145). Preclinical models showed strong antileukemic activity for anti-CD123 ADCs linked to a DNA-alkylating cytotoxic payload (indolinobenzodiazepine pseudodimer), such as IMGN632, which was then tested in a phase I/II trial (NCT03386513) in 66 R/R AML adult patients yelding ∼15% CR and ∼55% partial remission (91), and now under investigation as monotherapy or with azacytidine and/or venetoclax (NCT04086264) in patients with MRD after frontline treatment (92). Moreover, a pivotal registration frontline study of IMGN632 in patients with blastic plasmacytoid DC neoplasm (CADENZA) confirmed its safer clinical profile, as compared with tagraxofusp (low-grade peripheral edema, reversible veno-occlusive disease in <5%, absence of capillary leak syndrome and treatment-related deaths; ref. 93).

CD70. In preclinical studies, disruption of the CD27-CD70 interaction with an anti-CD70 mAb inhibits AML growth in vitro and in vivo, prolonging animal survival. Cusatuzumab (ARGX-110), an anti-CD70 mAb with afucosylation of the Fc region, followed by azacytidine, induced 83% CR in untreated old patients with AML (NCT03030612; ref. 94). Notably, azacitidine or venetoclax induce CD70 expression in LSCs in preclinical models and are currently tested with cusatuzumab in newly diagnosed AML (NCT03030612) or previously untreated de novo or secondary AML (NCT04150887).

CD45. CD45 is a tyrosine phosphatase expressed on AML blasts, which dephosphorylates and downregulates Src family kinases in normal HSPCs, modulates lymphocyte maturation, activation, and proliferation, and promotes motility and retention of progenitor cells. Iomab-B, an anti-CD45 mAb conjugated to ¹³¹I, showed good tolerability when used in combination with the reduced-intensity conditioning (RIC) regimen in patients with AML undergoing allo-HSCT (NCT00008177), and good preliminary results in phase III clinical trial with R/R AML in the elderly (NCT02665065). A novel ⁹⁰Y-BC8-immunoconjugated anti-CD45 also showed good tolerability in a phase I trial (NCT01300572) in combination with RIC in R/R AML ineligible for allo-HSCT (95).

In conclusion, different mAbs have shown promising clinical activity when used in combination with chemotherapy or HMAs and/or novel agents. Identifying molecular predictors of treatment response will help refine patient selection and optimize the usage of these novel approaches.

Bispecific T-cell Engaging mAbs

BiTEs are bi-specific mAbs that recognize T cells (CD3) and blasts. By forming a direct physical bridge, BiTEs facilitate the creation of cytolytic synapses, inducing apoptosis of tumor cells. Notably, T-cell engagement and activation by BiTEs is independent of the presence of MHC or costimulatory receptors and represents an off the shelf strategy. However, the identification of suitable target antigens and the presence of host active T cells remain major challenges. BITEs against different AML antigens (e.g., CD33, CD123, CLL-1, and FLT3) are currently in early clinical trials with preliminary evidence of safety and some clinical efficacy (96). AMG330 is a CD3*CD33 human BiTE that induced prolonged survival in murine AML and cytotoxicity in patient-derived R/R AML blasts and is now under clinical evaluation in combination with pembrolizumab in R/R AML (NCT04478695). Expression of activating/inhibitory checkpoint molecules on AML cells seems critical for T-cell activation following CD3*CD33 treatment. AMG673 is a new CD3*CD33 BiTE fused to the N-terminus of a single-chain Fc region, to increase its half-life, which also showed a good safety profile and some efficacy in a phase I trial (NCT03224819; ref. 96). Promising results have been reported in both preclinical and early clinical studies with both the investigational dual affinity re-targeting antibody (CD123*CD3; MGD006 or flotetuzumab and the CD123-directed bispecific antibody vibecotamab; ref. 97). Again, preexisting active T cells seem crucial for their efficacy. Finally, T-cell engagers are being tested with ICIs, showing disease eradication in the absence of adverse events in AML xenografts.

Bispecific and Trispecific Killer-Cell Engagers

Bispecific killer-cell engagers (BiKEs) and trispecific killer-cell engagers (TriKEs) target a tumor antigen and CD16 on the surface of NK cells (BiKEs), alone or in combination with IL15 to induce NK cell expansion (TriKEs; ref. 98). CD16*CD33, CD16*NKG2DL, or CD16*CD33*IL15 showed in vitro efficacy against primary AML lines by inducing NK activation/proliferation and cytokines release (99). Because TriKES displayed greater NK cytotoxicity towards AML blasts, a phase I/II trial was recently initiated on CD33⁺ R/R AML (NCT03214666; ref. 100). Expectedly, BiKEs and TriKEs require the presence of adequate quantities of active circulating NK cells and may be potentiated by the infusion of ex-vivo expanded NK cells. Novel constructs targeting the most suitable AML antigens are under development, such as the CD123 TriKE, shown to possess antitumor activity and modest toxicity in preclinical models.

Vaccines

Therapeutic cancer vaccines hold immense potential to unleash the body’s natural defenses against cancer cells, and the unique ability to generate lasting immune memory, potentially providing long-term protection. However, despite decades of intense investigation predating ICIs, CAR-T cells, and bispecifics, cancer vaccine therapies are yet to translate this potential into clinical successes. As with the abovementioned immunotherapy modalities, the choice of the antigen is critical for vaccine efficacy and should factor in its expression, capacity to be processed by antigen-presenting cells (APC), and its ability to elicit a strong cytotoxic T-cell response. Moreover, optimal adjuvants and administration routes have to be considered.

Despite many efforts, peptide-based vaccination in AML proved to be the least effective, probably because of its inability to effectively stimulate memory CD4⁺ T cells and induce durable immune responses. The most investigated antigen was WT1, which is highly expressed by most AML subtypes, involved in AML maintenance/growth, and has been extensively tested as an immunological target in preclinical models. A few clinical trials were performed with WT1 vaccines on limited numbers of patients with AML/MDS, showing safety and feasibility, but with few patients obtaining long-term remissions. Although T-cell responses were documented in most patients, the main problem was the induction of tolerance upon repeated administrations.

DC-based vaccination is an alternative antigen-delivery system. Monocytes (Mo-DC) or leukemia cells themselves (DCleu) have been used for this scope. With either, tumor antigens are exposed to T cells by professional APCs to generate a strong immune response. Mo-DC can be generated ex vivo from autologous or allogenic CD14⁺ monocytes and then “instructed” to present specific TAAs by gene transfer, such as WT1, PRAME, or hTERT, or, alternatively, multiple TAAs by direct loading of apoptotic cells or total AML-blasts lysates (101). Several trials have been conducted with DC vaccines in AML, usually in CR after chemotherapy or HSCT, which showed feasibility, safety, and good correlations between clinical and specific T-cell responses. In a WT1-DC vaccine trial (NCT00965224), 13/30 patients with AML in long remission showed circulating anti-WT1 CD8⁺ T cells. A DCleu vaccine, either alone or combined with autologous NKs, showed good clinical responses in postremission patients (102), yet modest results were then obtained in R/R AML (103). In general, despite some preliminary encouraging clinical data, Mo-DC vaccination has not been extensively employed in AML, probably due to the intrinsic difficulty of generating in vivo proficient APCs. Ongoing research aims at optimizing in vitro APC differentiation and activation, or DC administration with better adjuvants or in combination with drugs known to reduce PD1 expression and MDSCs activity (104).

Whole-cell vaccines using irradiated AML cells have been used as alternative antigen-delivery systems, which, however, failed to show a potent antileukemic immune response in most patients. Blasts were engineered to co-express T-cell- or DC-activating molecules (CD80 or GM-CSF) in preclinical models, prompting several phase I-II clinical trials that showed feasibility and tolerability but poor clinical efficacy, probably due to the concomitant activation of immunosuppressive MDSC by GM-CSF (105). Emerging evidence suggests that the so-called immunogenic cell death (ICD) exerts a cancer vaccine-like effect and could be exploited to increase the immunostimulatory capacities of whole cancer cell-based vaccines. ICD can be triggered by chemotherapy, radiation, or targeted therapies and is characterized by the release of tumor antigens to stimulate an immune response, or the exposure/secretion of molecules (e.g., CRT, high-mobility group box 1 and ATP) that work as eat me signals and potentiate the uptake of blasts by DCs, promoting a robust anticancer immunity. ICD has been investigated, both in vitro and in vivo, to boost the immunogenicity of leukemic blasts (revised in refs. 105, 106).

In conclusion, current attempts to develop cancer vaccines in AMLs have not yielded the desired results. Research on therapeutic cancer vaccines is however extremely active and focused on key areas, including identification of ideal immunogenic neoantigens, optimization of delivery platforms, and bypassing of the immunosuppressive TME, the latter appearing as a critical limit to the success of cancer vaccination strategies.

Concluding Remarks

Although immune-based therapies are changing the natural history of several solid tumors and some hematological diseases, their impact on AML is still modest, except for allo-HSCT, the most powerful immunotherapy approach for high-risk AML. Despite the plethora of novel immunotherapy approaches available, the anti-CD33 ADC GO remains the only immune-based approved drug in AML. Others are expected to be approved shortly, probably in combination with other treatments and/or in specific clinical settings (e.g., CR maintenance). AML is highly heterogeneous, and individual patients may respond differently to treatments and at different disease stages. Different ongoing studies in AML aim to adapt each of the available immunotherapy approaches (ICIs, CAR-T/NK cells, mAbs, and vaccines) and to investigate the effects of combining immune-stimulating strategies with the blockade of immune-suppressive mechanisms. Most notably, preliminary preclinical and clinical evidence predict a successful future for the combination of immunotherapy with other treatment strategies, including chemotherapy, venetoclax, or demethylating agents. The final goal is to provide mechanism-based and personalized therapies able to sustain remissions, eventually substituting allo-HSCT and chemotherapy.

Authors’ Disclosures

C. Restelli reports grants from AIRC during the conduct of the study. M. Ruella reports personal fees from NanoString, Bristol Myers Squibb, GlaxoSmithKline, Bayer, and Scaylite; grants from Beckman Coulter and Oxford Nanoimaging; grants and other support from viTToria Biotherapeutics; grants and personal fees from AbClon; and grants from CURIOX outside the submitted work; in addition, M. Ruella has a patent for Treatment of Cancer using anti-CD19 chimeric antigen receptor issued, a patent for Anti-CD123 chimeric antigen receptor (CAR) for use in cancer treatment issued, and a patent for Treatment of Cancer using humanized anti-BCMA chimeric antigen receptor issued. No disclosures were reported by the other authors.