Harnessing Autologous Immune Effector Mechanisms in Acute Myeloid Leukemia: 2023 Update of Trials and Tribulations

Abstract

Numerous recent advances have been made in therapeutic approaches toward acute myeloid leukemia (AML). Since 2017, we have seen eleven novel Food & Drug Administration (FDA)-approved medications for AML, all of which extend beyond the classical cytarabine-based cytostatic chemotherapy. In the recent two decades, the role of immune surveillance in AML has been intensively investigated. The power of one’s own innate and adaptive immunity has been harnessed pharmacologically toward the goal of clearance of AML cells. Specifically, pre-clinical studies have shown great promise for antibodies that disinhibit T cells and macrophages by blocking checkpoint receptors within the immunologic synapse, thereby resulting in the elimination of AML cells. Anti-CD33 CAR-T therapies and anti-CD3/CD123 bispecific antibodies have also exhibited encouraging results in pre-clinical and early clinical studies. However, despite these translational efforts, we currently have no immune-based therapies for AML on the market, with the exception of gemtuzumab ozogamicin. In this focused review, we discuss molecular target validation and the most relevant clinical updates for immune-based experimental therapeutics including anti-CD47 monoclonal antibodies, CAR-T therapies, and bispecific T cell engagers. We highlight barriers to the clinical translation of these therapies in AML, and we propose solutions to optimize the manufacturing and delivery of the most novel immune-based therapies in the pipeline.

I. INTRODUCTION

Although numerous advances have been made in acute myeloid leukemia (AML) therapeutics, morbidity and mortality rates remain high. Food and Drug Administration (FDA) approvals of immunotherapy for solid tumors in the last decade have sparked interest in studying immunotherapeutics in hematologic malignancies. Allogeneic stem cell transplant (allo-HCT) elicits graft-versus-leukemia effect, and numerous recent clinical trials have explored agents that engage one’s own immune system to combat AML. While these agents span a broad range of mechanisms, one of the most pivotal of these exploits the principle that inhibitory receptors or “checkpoints” on T cells restrict their activation, and that the administration of antagonistic antibodies can bolster tumor-specific responses. Other “checkpoint” receptor/ligand pairs exist within the innate system as well, notably on macrophages, and can similarly be blocked to enhance anti-tumor immunity.

Thus far, clinical trials data disappointingly have not translated into any recent FDA approvals to date.¹ The experimental pipeline of immunotherapies in AML includes macrophage checkpoint inhibitors, T cell checkpoint inhibitors, chimeric antigen receptor (CAR)-T cells, bispecific T cell engagers (BiTEs), bispecific killer cell engagers (BiKEs), trispecific killer cell engagers (TriKEs), CAR-NK cells, and dual-affinity re-targeting (DART) molecules. There are many ongoing challenges to translational efforts related to immunotherapies in AML, and there is an unmet need for new insights and solutions regarding how current investigational agents can be optimized in the upcoming months and years to more safely and effectively harness autologous anti-cancer immunity.

II. MACROPHAGE CHECKPOINT BLOCKADE

II.1. Molecular Target Validation

The macrophage inhibitory checkpoint CD47/SIRPα was one of the first and most promising targets to be explored in AML. Through binding of its cognate receptor signal-regulatory protein alpha (SIRPα) on the surface of macrophages and dendritic cells, CD47 elicits a co-inhibitory signal that suppresses phagocytosis. CD47 thus serves as a negative checkpoint or “don’t eat me” signal to phagocytes.² Upregulation of CD47 by a number of solid and hematologic cancers allows for malignant cells to evade immune clearance.³ The first demonstration of target validation of CD47 was done by Stanford University in 2009, when it was found that CD47 is constitutively upregulated by leukemic stem cells (LSCs) in AML, and that greater surface expression of CD47 by these cells is associated with poorer prognosis.^4,5

II.2. Pre-clinical Data

In the late 2000s, a growing body of evidence suggested that CD47 upregulation contributes to disease pathogenesis.^4,5 For example, in xenograft experiments, a higher surface CD47 correlates with the ability of human leukemia cell lines to engraft and elicit fulminant disease. Jaiswal et al. showed that the patient-derived AML line MOLM-13, which expresses low CD47, was unable to cause disease in immunodeficient mice that succumbed to higher CD47-expressing AML cell lines.⁴ Engraftment by a pathogenic AML cell line that expresses relatively high levels of CD47 was prevented by pre-treatment with a CD47 blocking antibody prior to challenge in immunodeficient mice.⁵ Clinical correlative studies suggested that greater CD47 expression was predictive of worse overall survival in multiple independent cohorts (Figure 1).⁵

Fig. 1.

Summary of the novel experimental therapeutics targeting the interface of AML cells and immune effector cells.

A number of high-impact laboratory experiments in the 2000s demonstrated therapeutic benefit when CD47 was inhibited. Majeti et al. show that in human AML xenografted mice, daily intraperitoneal injections with the anti-CD47 antibody B6H12.2 resulted in near complete elimination of leukemia from peripheral blood.⁵ Both in vitro and in vivo administration of a CD47 blocking antibody facilitated phagocytosis of AML cells by murine macrophages. In vivo efficacy in leukemic mice has since been recapitulated with multiple humanized CD47 blocking agents.^6,7 Chao et al. demonstrated in vitro that CD47 blockade did not elicit phagocytosis of a number of healthy human cells from a variety of tissue types including peripheral blood, bone marrow, and fetal bladder by human macrophages.⁸

This selectivity for leukemic cells is believed to reflect an important underlying immunologic principle: in order for a phagocyte to engulf a cell, it requires not only the absence of an inhibitory signal but also the presence of a pro-phagocytic one. In many human cancers including AML, this pro-phagocytic signal is believed to be cell surface calreticulin. Binding of calreticulin was shown to be necessary for phagocytosis of leukemic cells elicited by anti-CD47 antibody in vitro.⁸ Thus, most healthy self-renewing cells which express negligible levels of surface calreticulin are likely to be spared by therapy. Ultimately, the therapeutic efficacy and relative selectivity observed in these preclinical experiments catapulted CD47 blockade into the clinic.

II.3. Translational Efforts

The most extensively studied anti-CD47 agent is Hu5F9-G4 (magrolimab), a humanized IgG4 monoclonal antibody (Table 1). Despite its preclinical efficacy as an individual agent in mice, testing of magrolimab as a monotherapy in a Phase I dose escalation study provided insufficient promise for continued development as an individual agent. In this Phase I study, no objective responses were observed, but 73% of patients experienced stable disease.^9,10 It has been suggested that combination approaches to CD47 blockade, especially with therapies that upregulate pro-phagocytic signals on leukemic cells, might be more efficacious. One such approach is combination with azacitidine, a hypomethylating agent (HMA) that has been shown to upregulate calreticulin on leukemic cells and bolster anti-CD47 mAb mediated phagocytosis in vitro.¹¹ Recent results from a Phase 1b trial of magrolimab plus azacitidine showed an overall response rate (ORR) of 64% in AML and 91% in myelodysplastic syndrome (MDS). The rates of complete response (CR) were 40% in AML and 42% in MDS. In a subset of 12 patients with TP53-mutant AML, the rate of CR plus CRi was 75%.¹² The recent update from 2022 studying a larger cohort of 72 patients with TP53-mutant AML showed a lower CR plus CRi rate of 41.6%.¹³

TABLE 1.

Active and Enrolling Clinical Trials Investigating Checkpoint Inhibitors as of September 2023

Intervention	Name of Study	Target	Disease	Design Phase	Enrollment	Identifier	Primary Objective	Country
Pembrolizumab	BLAST MRD AML-2: Blockade of PD-1 Added to Standard Therapy to Target Measurable Residual Disease in AML 2- A Randomized PH 2 Study of Anti-PD-1 Pembrolizumab in Combination With Azacitidine and Venetoclax as Frontline Therapy in Unfit Patients With AML	PD-1	AML AML Arising From Previous MDS Secondary AML Therapy-Related AML	II	76	NCT04284787	MRD(−) CR or MRD(−) CR with incomplete count recovery	USA
Pembrolizumab	BLAST MRD AML-1: Blockade of PD-1 Added to Standard Therapy to Target Measurable Residual Disease in AML 1- A Randomized PH 2 Study of Anti-PD-1 Pembrolizumab in Combination With Intensive Chemotherapy as Frontline Therapy in Patients With AML	PD-1	AML AML Arising From Previous MDS Secondary AML Therapy-Related AML	II	124	NCT04214249	MRD(−) CR and rate of MRD(−) CR at end of therapy and end of consolidation	USA
Pembrolizumab	MRD-guided Treatment With Pembrolizumab and Azacitidine in NPM1mut AML Patients With an Imminent Hematological Relapse	PD-1	AML	II	28	NCT03769532	Proportion of event-free patients	Germany
Nivolimab and Relatimab	An Open-Label PH II Study of Relatlimab (BMS-986016) With Nivolumab (BMS-936558) in Combination With 5-Azacytidine for the Treatment of Patients With Refractory/Relapsed AML and Newly Diagnosed Older AML Patients	PD-1 and LAG-3	AML	II	30	NCT04913922	Maximum tolerated dose (MTD) and dose-limiting toxicities (DLT)	Germany
Nivolimab and Ipilimumab	An Open-Label PH II Study of Nivolumab (BMS-936558) in Combination With 5-Azacytidine (Vidaza) or Nivolumab With Ipilimumab in Combination With 5-Azacytidine for the Treatment of Patients With Refractory/ Relapsed AML and Newly Diagnosed Older AML (>65 Years) Patients	PD-1 and CTLA-4	Acute Bilineal Leukemia Acute Biphenotypic Leukemia AML Arising From Previous MDS CMML MDS Recurrent AML Refractory AML Secondary AML Therapy-Related AML	II	182	NCT02397720	MTD, DLT and ORR	USA
Magrolimab	A PH 3, Randomized, Double-Blind, Placebo-Controlled Study Evaluating the Safety and Efficacy of Magrolimab Versus Placebo in Combination With Venetoclax and Azacitidine in Newly Diagnosed, Previously Untreated Patients With AML Who Are Ineligible for Intensive Chemotherapy	CD47	AML	III	432	NCT05079230	CR and OS	USA
Sabatolimab	A PH Ib/II, Open Label Study of Sabatolimab as a Treatment for Patients With AML and Presence of Measurable Residual Disease After Allogeneic Stem Cell Transplantation	TIM-3	AML	I and II	59	NCT04623216	DLT, investigator-assessed absence of relapse	France and Germany
Sabatolimab	A Study of MBG453 in Combination With Azacitidine and Venetoclax in AML Patients Unfit for Chemotherapy (STIMULUS-AML1)	TIM-3	AML	II	86	NCT04150029	DLT; CR	USA
LYT-200	A Phase 1 Open-label, Multi-center Study of the Safety, Pharmacokinetics (PK), and Anti-tumor Activity of LYT- 200 only in Patients with Relapsed/Refractory Acute Myeloid Leukemia (AML), or with Relapsed/refractory, High-risk Myelodysplastic Syndrome (MDS)	Galectin-9	AML MDS	I	50	NCT05829226	Safety Tolerability	USA

Abbreviations : AML, acute myeloid leukemia; CR, complete remission; DLT, dose-limiting toxicity; MDS, myelodysplastic syndrome; MRD, measurable residual disease; OS, overall survival; PH, phase

III. T CELL CHECKPOINT BLOCKADE

III.1. Molecular Target Validation

T cell checkpoint inhibition in AML was suggested based on the historical efficacy of the graft-versus-leukemia effect of allo-HCT, in which alloreactive T cells have continuous efficacy after transplant. Indeed, T cell-depleted allo-HCTs have been associated with increased risk of disease relapse in AML.¹⁴ The major targetable T cell checkpoints include PD-1 (CD279), CTLA-4 (CD152), and TIM-3 (CD366). Inhibitory signaling occurs when PD-1 on T cells binds to PD-L1 on target cells, CTLA-4 on T cells binds to CD80 (B7–1) and CD86 (B7–2) on target cells, and TIM-3 on T cells binds to galectin-9 on target cells.¹⁵

III.2. Pre-clinical Data

Studies in murine models of AML have established the precedence for T cell checkpoint blockade. The most well-studied checkpoint pathway to date in this context is PD-1/PD-L1. In both C1498 and DA1–3b models of murine AML, PD-1 inhibition results in prolonged survival.¹⁶ Furthermore, C1498 cancer cells upregulate PD-L1 after a period of growth in C57BL/6 mice, and CD8(+) T cells in organs of high C1498 leukemic burden upregulate PD-1.¹⁷ PD-L1 blockade with a monoclonal antibody in these mice increases the proliferation and cytotoxicity of T cells in the leukemic microenvironment, extends survival, and enhances the efficacy of adoptively transferred polyclonal AML-specific cytotoxic T lymphocytes (CTLs).¹⁷

While the efficacy noted above is in part due to the disinhibition of effector T cells, PD-1/PD-L1 blockade likely also alters the expansion and function of regulatory T cells (T_regs), which also express PD-1. T_regs are highly prevalent in the peripheral blood and bone marrow of AML patients, increased in number relative to healthy controls, and inhibit anti-cancer immune responses.¹⁸ Interaction of PD-L1 on AML cells and PD-1 on T_regs drives T_reg expansion.¹⁹ Furthermore, PD-L1 blockade in mice hinders the ability of endogenous T_regs to suppress adoptively transferred CTLs and attenuates T_reg numbers while reducing tumor burden.¹⁹ These data suggest that PD-L1 blockade might be beneficial in AML by diminishing T_reg number or function and allowing for the expansion of AML-specific effector cell populations.

III.3. Translational Efforts

PD-1 and CTLA-4 inhibitors showed limited efficacy as monotherapies in preliminary trials of hematologic malignancies and high-risk MDS, respectively.^20,21 Therefore, combinations of checkpoint inhibitors with other therapies are the focus of most current investigations in AML. Checkpoint molecules are upregulated after treatment with HMAs, and this correlates with inferior clinical outcomes.²²

In the setting of newly diagnosed AML, trials of PD-L1 inhibitors in combination with HMAs have thus far suggested low clinical efficacy with significant risks of toxicity. A Phase II trial of the PD-L1 inhibitor durvalumab combined with azacitidine versus azacitidine alone found that the addition of durvalumab did not improve clinical efficacy. The ORR (CR + CR with incomplete count recovery) was comparable in both arms: 31.3% vs 35.4%, respectively.²³ Despite the lack of additional efficacy, toxicity was higher in the durvalumab group. A single-arm, open-label Phase I study of another anti-PD-L1 agent, avelumab, with decitabine as first-line treatment enrolled 7 patients and similarly concluded no clinical benefit, with only 1 patient achieving CR.²⁴ Limited results of PD-1 inhibition in this setting may be more encouraging. In a Phase 2 study of pembrolizumab and standard course azacitidine in 17 newly diagnosed patients, 8 (47%) achieved CR/CRi.²⁵ In the relapsed/refractory (R/R) setting, early studies of nivolumab in combination with azacytidine were performed in a Phase II trial. The trial showed an encouraging ORR of 33% and CR rate of 22%.²⁶ Pembrolizumab was studied in combination with decitabine in R/R AML in the single-arm open-label PD-AML trial (NCT02996474), where 6 of 10 patients achieved stable disease or better.²⁷

The CTLA-4 antibody ipilimumab has also had limited success in clinical trials. In a Phase I/Ib study, low-dose ipilimumab did not result in CR, while high-dose ipilimumab resulted in 23% CR rate, with 42% CR in AML specifically.²⁸ The study was partly limited by graft-versus-host disease. A similar study in 29 patients with various hematologic malignancies, only 2 of whom had AML, treatment with low-dose ipilimumab after HCT resulted in few objective responses, and no responses were seen in the 2 patients with AML.²⁹ The use of checkpoint inhibitors in the post-HCT setting (compared to the front-line setting) may be limited by the inherent defect in patients’ T cells and immune cells. Additional T cell checkpoint inhibitor trials are underway as of September 2023 (Table 1).

Sabatolimab is an anti-TIM-3 antibody that is being studied in combination with azacitidine and venetoclax in the STIMULUS-AML1 trial. This is currently under active investigation in MDS in the randomized Phase II multi-center STIMULUS-MDS1 trial, but efficacy data is currently not mature for AML. Preliminary data in the Phase 1B setting showed ORR of 40% with median duration of response of 12.6 months.³⁰

IV. CAR-T THERAPIES

IV.1. Molecular Target Validation

Seminal studies in the 2000s have paved the way for CAR-T therapy in hematologic malignancies. CAR-T cells are ex vivo genetically modified autologous T cells containing a chimeric receptor (variable region of an antibody and co-stimulatory domains) designed to target a specific tumor antigen.³¹ Target validation for B cell leukemias and lymphomas involved CD19, which led to translational success. Myeloid surface antigens including CD33, FLT3, and CD123 have been proposed as targets of CAR constructs. However, no current FDA-approved CAR-T cells are available for AML as of September 2023.

IV.2. Pre-clinical Data

Extensive pre-clinical studies have investigated a variety of different CAR-T cell targets in AML. These targets include CD33, FLT3, CD123, LILRB4, CLL-1, B7-H3, NKG2D, and CD116/CD131.^31–34 CAR-T cells designed to target these myeloid antigens have shown variable efficacy in single-arm studies. The original studies showed that EBV-specific cytotoxic T cells can be genetically modified with an anti-CD33 CAR, and these major histocompatibility complex (MHC)-unrestricted cells can exert anti-leukemic activity. CD33-directed CAR-T cells have shown efficacy in murine models: they can prevent the development of AML in vivo and can also delay the progression of disease that is already established.³⁵ However, early studies on anti-CD33 CAR-T cells showed that these CARs also target normal CD34(+) hematopoietic progenitors.³⁶ CAR-T cells against CD123, or IL-3Rα, were thus developed, as CD123 was thought to be a more specific marker for cancerous cells compared to CD33.³⁷ Additional myeloid antigen-targeted CARs were generated in the subsequent years. For example, compound CAR-T cells were studied in murine models: this double-prong (CD33 and CD123) has been shown to target LSCs.³⁸

IV.3. Translational Efforts

These pre-clinical data have led to many clinical trials in various stages, though the data is sparse and case-based at best, with limited to no large-scale systematic studies. Anti-CD33 CAR-T cells can lead to decreased AML cell burden, but T cell exhaustion can occur.³⁹ Preliminary data from NCT04351022 which investigated anti-CD38 CAR-T cells showed that 4 out of 6 patients achieved some form of complete remission lasting a median of only 191 days at best.⁴⁰ Flow cytometry revealed depletion of CD38(+) blasts after the anti-CD38 CAR-T cell infusion.⁴⁰ Four patients received anti-LeY CAR-T cells with 3 showing some degree of response to the infusion and 1 having morphologic/immunophenotypic remission for 23 months.⁴¹ The CAR-T cells persisted for up to 10 months and no grade 3 or 4 toxicities were observed.⁴¹ A Phase 1 trial investigating anti-NKG2D CAR-T cells in 7 AML patients did not show any objective response when a single injection of the lowest cell dose was used.⁴² There was evidence of mild improvement in AML and disease stability with higher doses or multiple doses. The side effect profile was favorable with no dose limiting toxicities and no grade 3 or higher adverse event attributable to the CAR-T cell infusion.⁴² These limited studies merit larger scale efforts, especially in the Phase 3 setting. Dual-antigen targeting is also being studied.

V. ANTI-CD33 ANTIBODIES

V.1. Molecular Target Validation

CD33 is found on multipotent hematopoietic progenitors, myelomonocytic precursors and further differentiated members of the myeloid lineage including monocytes and neutrophils.⁴³ While CD33 is expressed by mature leukemic blasts, it gained particular interest as a therapeutic target based on the notion that some AML cells seem to be derived predominantly from a CD33(+) progenitor. Seminal studies were conducted at the Fred Hutchinson Cancer Center: targeting of CD33 was pursued in the hopes eradicating the pool of stem cells that seed the cancer over time and drive relapse.⁴³

V.2. Pre-clinical Data

Monoclonal antibody (mAb) targeting of CD33 was first attempted to ascertain whether there was any intrinsic cytotoxicity while also tagging leukemic cells for phagocytosis. Crosslinking of CD33 with a mAb on human AML cells in vitro was shown to halt cell proliferation and induce apoptotic-like cell death.⁴⁴ In vitro targeting of CD33 can indeed fix complement and elicit antibody-dependent cellular cytotoxicity (ADCC), when utilizing a mAb with the appropriate Fc. Ultimately, the success of simple CD33 targeting was initially limited by its immunogenicity, whereby patients would develop human anti-murine antibody responses.

While unconjugated antibodies had proven largely ineffective, new strategies emerged to exploit CD33, most notably in the way of antibody-mediated delivery of toxins. Now, clinical oncology is replete with antibodies conjugated to cytotoxic compounds, or payloads. Upon binding to CD33, these toxin-antibody complexes can be internalized, where they accumulate in the target cell resulting in cell death.⁴³ The first-in-class agent was gemtuzumab ozogamicin (GO) – a CD33 mAb connected to calicheamicin. While GO is highly stable in circulation, the calicheamicin derivative is readily released when exposed to acid within cells to exert its cytotoxic effects.⁴³ Preclinically, GO has shown selective cytotoxicity against CD33(+) AML cells in vitro, which led to translational efforts.

V.3. Translational Efforts

Clinically, GO was originally investigated for the treatment of acute promyelocytic leukemia but was more recently approved by the FDA in 2017 for broader use in newly diagnosed and R/R CD33(+) AML.⁴⁵ ALFA-0701 was a seminal trial that led to the approval of GO in newly diagnosed AML along with 7+3 induction chemo.⁴⁵ Induction chemotherapy with cytarabine and anthracycline (“7+3”) in combination with GO improved OS compared to the 7+3 alone; the 2-year OS was 53.2% vs 41.9%, respectively. Since then, GO has been approved in multiple settings, including the consolidation, maintenance, and R/R settings. The EORTC-GIMEMA AML-19 trial led to the approval of GO monotherapy in the newly diagnosed setting.⁴⁶ This trial showed improved OS with GO monotherapy compared to best supportive care (4.9 months vs. 3.6 months). Other ongoing approaches are employing CD33 as a CAR-T or CAR-NK target, or as an antigen targeted by bispecific antibodies.

VI. BI-SPECIFIC T CELL ENGAGERS (BiTEs)

VI.1. Molecular Target Validation

The introduction of the first-in-class BiTE, known as blinatumomab, sparked interest in the use of BiTEs in other hematologic malignancies. BiTEs consist of two single chain variable fragments specific for CD3 and a tumor antigen, respectively. One of the main benefits of BiTEs is that they are MHC-independent, as MHC is often downregulated in malignancies. BiTEs established their presence for the treatment of hematologic malignancies in the TOWER study of blinatumomab, which targets CD3 and CD19 in B cell ALL.⁴⁷

VI.2. Pre-clinical Data

Pre-clinical data on BiTEs began in the early 2010s, when bispecific constructs were generated against CD3 on T cells and CD33 on myeloid cells in primary samples of AML and in AML cell lines. AMG 330 (the CD3/CD33 BiTE) was shown to lyse the AML cell lines KG-1 and U937, and AMG 330 could expand T cells also.⁴⁸ Murine studies showed that AMG 330 could induce infiltration of T cells into HL60 tumors and also inhibit tumor growth.⁴⁸ This BiTE showed high activity in CD33(+) cells in a dose-dependent manner, which correlated with CD33 expression.⁴⁹ AMG 330 can expand CD3(+)/CD45RA(−)/CCR7(+) memory T cells and cause CD33(+)cell lysis at low effector:target ratios.⁵⁰

VI.3. Translational Efforts

The leading agent in clinical trials had been AMG 330, which showed promising in vivo efficacy regarding depletion of CD33(+) blasts and CD33(+) myeloid-derived suppressor cells. A Phase 1 trial (NCT02520427) using AMG 330 showed preliminary results of 4 of 35 patients achieving some form of CR.⁵¹ However, 24 of 35 patients needed to discontinue treatment due to disease progression and 15 patients had treatment-related serious adverse events, including 11 with cytokine release syndrome. This trial (NCT04478695) has been unfortunately terminated (Table 2). Another Phase 1 trial (NCT03144245) using AMV564 (which bivalently targets CD33/CD3) had preliminary results of 13–91% reduction in bone marrow blasts in 12 of 18 patients and no dose-limiting toxicity.⁵²

TABLE 2.

Active and Enrolling Clinical Trials Investigating BiTEs and CAR-T Therapies as of September 2023

Intervention	Name of Study	Target	Disease	Design Phase	Enrollment	Identifier	Primary Objective	Country
CLN-049	A PH 1, Open-label, Preliminary Pharmacokinetics and Safety Study of CLN-049 (An Fms-like Tyrosine Kinase 3 [FLT3] x Cluster of Differentiation 3 [CD3] Bispecific T Cell Engager) in Patients With Relapsed/Refractory AML	FLT3 x CD3	Relapsed/Refractory AML	I	9	NCT05143996	Preliminary Pharmacokinetics and Safety	USA
APVO436	PH 1B Open-Label, Dose-Escalation and Dose-Expansion Study of APVO436 in Patients With Relapsed or Refractory AML or High-Grade MDS	CD123 x CD3	AML and MDS	I	136	NCT03647800	Dose escalation, tolerability and safety	USA
anti-CD38 CAR-T	Pilot Study of the Efficacy and Safety of CD38 Targeted Chimeric Antigen Receptor Engineered T-Cells in the Treatment of CD38 Positive Relapsed or Refractory AML	CD38 CART	AML	I and II	20	NCT04351022	Adverse Events	China
anti-FLT3 CAR-T	Pilot Study of the Safety and Efficacy of Anti-FLT3 Chimeric Antigen Receptor Engineered T-Cells in the Treatment of Relapsed or Refractory AML	FLT3	AML	I and II	5	NCT05023707	Adverse Events	China
anti-CD33 CAR-T	Open-Label, Nonrandomized, Single-Arm PH I/II Study to Evaluate the Safety and Tolerability of Functionally Enhanced CD33 CAR T Cells in Subjects With Relapsed or Refractory AML	CD33	AML	I and II	25	NCT04835519	Adverse Events	USA
anti-CD19 CAR-T	Pilot Study of the Efficacy and Safety of CD19 Targeted Chimeric Antigen Receptor Engineered T Cell in the Treatment of Relapsed or Refractory CD19 Positive AML	CD19	AML	I and II	15	NCT03896854	Adverse Events	China

Abbreviations: AML, acute myeloid leukemia; CAR-T, chimeric antigen receptor-T; MDS, myelodysplastic syndrome; PH, phase

CD123 is another cell surface marker overexpressed in hematologic malignancies including AML. In vitro, two CD3/CD123 BiTE antibodies targeting different epitopes on CD123 both showed killing of AML blasts with one being 10 times more potent than the other.⁵³ This led to multiple CD3/CD123 BiTE Phase 1 trials, though the trial data is not mature. There is one ongoing Phase 1 trial (NCT03038230) using MCLA-117, a BiTE targeting CD3/CLEC12A. CLEC12A is a myeloid differentiation antigen expressed in 90–95% of AML regardless of subtype and is not classically expressed on normal hematopoietic stem cells.⁵⁴ MCLA-117 showed in vitro and ex vivo lysis of 23–98% of AML blasts.⁵⁴ Another target is a CD3/FLT3 BiTE (Table 2). FLT3 is another common marker in AML with limited expression in normal tissue. A CD3/FLT3 BiTE showed in vitro, in vivo, and ex vivo elimination of FLT3-expressing cells.⁵⁵

VII. BIKEs, TRIKEs, CAR-NK CELLS, AND DARTs

Other immune-based therapies in AML that are taking root in clinical soil include BiKEs, TriKEs, CAR-NK cells, and DART molecules.⁵⁶ BiKEs and TriKEs harness the power of NK-dependent ADCC against cancer cells and have 2 or 3 variable portions of antibodies. The value of NK cells is evident in that these tumor-targeting cells are not MHC-restricted and do not require priming before killing tumor cells. BiKEs and TriKEs can directly crosslink CD16 (FcγRIII) on NK cells to drive ADCC and are far less likely to elicit graft-versus-host disease or cytokine release-related toxicities. One disadvantage of these agents is the complex methodology required for synthesis of these agents, which involves generation of the relevant variable fragments and a linker between the respective binding sites of the BiKE. CAR-NK cells have been engineered against CD33; these cells have shown efficacy against OCI-AML2 and primary AML cells. They can prevent engraftment of AML cells.⁵⁶ However, there is no Phase 3 clinical trial data available. Finally, DARTs are also making their way into the clinic. These agents utilize a protein design different from BiTEs to simultaneously target 2 antigens. Pre-clinical studies for CD3xCD123 DART in AML have been reported, but clinical trials have not led to FDA approval thus far.⁵⁷

VIII. CHALLENGES TO TRANSLATION OF IMMUNOTHERAPEUTICS

Thus far, the recent 10 years have witnessed overall suboptimal clinical trial results for immune-based therapies in AML. A variety of reasons may underlie these suboptimal translational efforts. We now have insight into these failures and foresight into how to improve the chance of successful clinical translation.

The first challenge is the inherent disease biology of AML, namely the risk for clonal evolution over time leading to changes in myeloid antigen expression (i.e. loss or immune escape). Antigen loss of immune-based targets, such as CD33 or CD123, may render the respective antibodies or CAR-T cells ineffective. The literature for antigen loss is more robust for CD19 in B cell lymphoma, in which CD19 loss can render the autologous CAR inert, and a similar mechanism may be present in AML. Another cell-intrinsic challenge is that the tumor mutational burden in AML is relatively low compared to solid tumors: a lower repertoire of neoantigens on the AML surface proteome may limit antigen recognition of tumor-specific T cells. This is particularly true of therapeutic antibodies that harness the power of adaptive immunity, like anti-PD-1 or anti-CTLA-4 mAbs. Such agents are highly efficacious in melanoma, renal cell carcinoma, and lung cancer, but not in AML due to limited surface neoantigens.

Next, pharmacodynamic problems are inherent to some immune-based therapies based on mechanism of action. These include significant adverse effect profiles related to off-target effects. Although numerous therapies have undergone molecular target validation, the molecular targets are sometimes non-specific and can also be present on the healthy hematopoietic stem/progenitor population. For example, CD47 is found on erythroid progenitors, and magrolimab can lead to prolonged or profound anemia. Since CD47 is co-expressed with pro-phagocytic stimuli on the surface of aging erythrocytes, these cells are particularly susceptible to clearance once CD47 is artificially blocked. CAR-T cells can lead to myelosuppression since the CARs may target the specified antigen on normal cells (i.e. hypogammaglobulinemia).

Thirdly, the suboptimal innate and adaptive immunologic status of patients with hematologic malignancies poses a major hurdle to the efficacy of immune-based therapies. The efficacy of most immune-based therapies relies on intact immune effector mechanisms in peripheral blood, as most of these pharmacologic agents harness the power of existing immune cells. For example, if the T cell frequency or macrophage frequency is low in a patient, pembrolizumab or magrolimab will not work well. Ideally, an optimal effector:target ratio would exist prior to administration of immune-based therapies, so heavily pre-treated patients may not benefit. Furthermore, CAR-T cell therapy requires a sufficient number of existing autologous T cells for successful manufacturing of the CAR-T cell product, which poses a problem in heavily pre-treated patients.

Finally, suboptimal clinical trial design may be responsible for lack of FDA approvals, especially since these therapies are novel and we are unsure about the optimal setting where these therapies are most effective. Clinical trial endpoints have a great degree of variability depending on the study population and the disease setting. This is especially true in AML which is a heterogenous disease with potential for multiple immune escape mechanisms. Some unanswered questions include:

• Should immunotherapeutics be employed at the time of frank relapse, at the time of measurable residual disease positivity but morphologic remission, as conditioning therapy prior to allo-HCT, and/or as maintenance therapy only?

• Should immunotherapeutics be employed in the first-line setting for maximal efficacy, since this is the time when the immune milieu (the patient’s catalogue of all circulating immune cells) is intact and most functional prior to any chemotherapy having been given?

• Should immunotherapeutics be used in combination with other AML-directed therapies to allow for synergistic cell killing (i.e. increased target antigen expression with azacitidine), or as monotherapy to ensure that antigen loss does not occur with use of a partner drug? If we extrapolate from the malignant B cell literature, the R/R setting appears to be optimal study setting for CAR-T cells, but the MRD(+) setting or R/R setting might be optimal for BiTEs.

IX. PROPOSED SOLUTIONS FOR THE FUTURE PIPELINE

Learning from previous wins and failures of immunotherapies, we recognize that optimization of manufacturing and administration of immunotherapies is warranted.

One solution is pharmacokinetic adjustments and dosing strategies. For example, magrolimab can be administered in a “priming and maintenance” regimen to help prevent anemia: a “priming” dose of the drug is given which clears aged red blood cells, prompts reticulocytosis, and stimulates a process termed erythrocyte pruning, allowing erythrocytes to downregulate CD47 from their surface.⁵⁸ This results in a pool of younger cells with lower surface expression of CD47 overall, which are consequently less susceptible to drug-induced clearance.⁵⁸ Therapy is resumed 1 week later with a “maintenance” therapeutic dose up to 30 times greater, with lower risk of recurring or worsening anemia.⁵⁸ This dosing approach has reportedly decreased the incidence of severe anemias in patients treated with magrolimab compared to the first Phase I trial conducted with this agent as a monotherapy, during which 89% of patients experienced treatment-related anemia.⁹

In the case of CD47 blockade-induced anemia, another solution is to employ anti-CD47 agents that have lower binding affinity for erythrocytes. TTI-621 is a recombinant, human SIRPα-Fc fusion protein that was designed with this purpose.⁵⁹ TTI-622, ALX-148, and IMM01 are also SIRPα-Fc fusion proteins which bind CD47 and are under study. Since CD47 blockade with the previously mentioned agents both blocks CD47 on erythrocytes and opsonizes them with a potentially pro-phagocytic Fc, an alternative approach to reducing anemic risk would be to block SIRPα. This is currently being explored with a mAb against SIRPα, CC-95251.

There are also a handful of solutions to the CAR-T challenges of limited durability, limited efficacy, and high toxicity. One potential solution to this problem of low durability of CAR-T cell responses is the use of a higher infusion dose of CAR-T cells or more frequent infusions. Durability of anti-CD33 CAR-T cells can be enhanced using the costimulatory molecule 4–1BB; these cells have increased survival in vitro compared to anti-CD33 CAR-T cells with the costimulatory molecule CD28.³⁹ As for CAR-T efficacy, combination therapies using CAR-T cells can also be considered, as some medications may have synergy with CAR-T cells. Decitabine was shown to improve anti-CD123 CAR-T cell effectiveness in vitro and in vivo (You et al., 2020). Similarly, anti-CLL-1 CAR-T cells were more effective when PD-1 was silenced: this silencing of PD-1 was achieved in the engineered CAR-T cells via addition of the PD-1 silencing shRNA sequence into this lentiviral vector.⁶⁰

Regarding the cytopenias related to CAR-T cells, the work-around for this is having a termination mechanism in place after the CAR-T cell infusion. The key will be to limit toxicity (cytopenias) while retaining efficacy (anti-leukemic effect). One option that has been explored in murine models involves the use of transiently active mRNAs encoding CARs. RNA modification might prevent cytopenias, which is a major toxicity with AML-directed CAR-T cells.⁶¹ Furthermore, rapidly switchable platforms for CAR-T cells may confer CAR-T cells with “on” and “off” signals so that normal hematopoietic cells can be spared and the duration of cytopenias can be limited.⁶² Biodegradable CAR-T products may also help prevent the problem of prolonged cytopenias, as has been shown for anti-CD123 CAR-T cells.⁶³ These engineered products can be created by electroporation of anti-CD123 CAR mRNA.⁶³ Other solutions to spare toxicities after CAR-T infusion include T cell ablation with alemtuzumab, and T cell ablation with rituximab using CD20-co-expressing CAR-T cells.⁶⁴ It is worth translating these concepts into human studies, though technical design will be complex.

In summary, our field will benefit from a further or deeper understanding of the immunologic niche such that immune-based therapies can come to market more rapidly. Pre-clinical rationale has already been established for multiple agents, but translation into clinical daily routine is lagging. Novel investigational therapies that target unique antigens are also of high interest. LYT-200, for example, is a humanized anti-galectin-9 antibody which has recently shown promising results in murine models, with potentially better efficacy compared to TIM-3 antibodies.⁶⁵ This data was presented at ASH 2022, and clinical trials are actively recruiting. A druggable immunologic niche may be the holy grail of the next generation of AML therapies.⁶⁶ We hope to gain further understanding of the unique escape mechanisms of AML cells in the near future. One of the leading agents in the pipeline previously appeared to be magrolimab, but the ENHANCE study was recently closed due to futility. Conclusions regarding future utility will rely heavily on the results of ongoing Phase 3 trials, which may come to fruition in the near future.

HIGHLIGHTS.

• Pre-clinical studies have shown great promise for checkpoint inhibition in AML.

• CAR-T therapies and BiTEs have shown variable success in AML.

• BiKEs, TriKEs, CAR-NK cells, and DART molecules are actively under investigation.

• Technical challenges must first be overcome for successful translation.