Since the early studies in the eighteenth century on the development of the first vaccines, researchers have attempted to eliminate tumors by harnessing the immune system. One strategy, adoptive cell therapy, uses T cells that can recognize tumor antigens through tumor-specific receptors. Since the 1980s, chimeric antigen receptor T cells (CAR T cells) have revolutionized the treatment algorithms of patients with lymphoma, acute lymphoblastic leukemia, and multiple myeloma [1]. They have shown dramatic success in the clinic, improving survival and quality of life for patients that would otherwise reach the end of care with conventional therapies. Currently, six CAR T cell products are approved by the United States Food and Drug Administration (FDA), and approvals are expanding to Europe and many other countries around the world.

There are distinct considerations in acute myeloid leukemia (AML), the most common acute leukemia in adults. AML is an aggressive blood cancer characterized by a collection of immature cells of myeloid lineage that exhibit partial or complete arrest of maturation [2]. The heterogeneity and intrinsic variability of the tumor make patient responses hard to predict, and around 75% of patients ultimately relapse. Treatment resistance (10–40%) and relapse remain the major consequences during disease follow-up, highlighting the urgent need for novel therapeutic approaches [3]. Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is the only curative option, but many patients are not suitable candidates [4].

Many in vitro and in vivo studies have shown that CAR T cells against surface proteins, such as CD33, CD123, CLL-1, CD13, CD7, NKG2D ligand, CD38, CD70, and TIM3, effectively eradicate AML cells [5]. However, the clinical trials are limited, with not very promising response rates accompanied by high ‘on-target off-tumor’ toxicity due to the frequent expression on healthy hematopoietic stem cells or progenitors, as well as other tissues. 

Normal hematopoietic stem cells give rise to mature cells of the myeloid, lymphoid, and erythroid/megakaryocyte lineages. Single-cell RNA sequencing (scRNA-seq) analyses have shown that normal hematopoietic stem cell (HSC) commitment proceeds through a series of increasingly lineage-committed progenitor states ^[6][13]. AML consists of leukemia stem cells (LSCs) and differentiated cells. LSCs sustain the disease and display self-renewal, quiescence, and therapy resistance. Differentiated AML cells that lack stem cell characteristics affect tumor biology through pathologic effects on the tumor microenvironment ^[7][14].

AML is a highly heterogeneous disease between patients due to the presence of specific chromosomal abnormalities, gene mutations, or gene fusions. Epigenetic modifications such as histone modifications, DNA methylation, and post-transcriptional regulation of mRNAs by noncoding RNAs have additional roles in the pathogenic heterogeneity ^[8][22].Furthermore, a crucial problem in patients with relapse or treatment resistance is intratumoral heterogeneity, formed with different subclones of leukemia cells, with distinct genetic and epigenetic features coexisting within a single patient ^[9][10][23,24]. Highly heterogeneous LSCs have variable drug sensitivity. Some LSCs acquire quiescence, which plays a major role in drug resistance profiles ^[11][12][25,26]. Recent advances in genomic, transcriptomic, epigenomic, and proteomic data have helped to elucidate the biological differences between pre-treatment AML cells and their equivalents in relapse. 

CAR T cells can bind to cell surface molecules without requiring any antigen processing or HLA expression ^[13][30]. The choice of which surface antigen to target is the critical step in manufacturing. The most important feature of an ideal target is the unique and high expression profile on tumor cells, above the detection and activation threshold for CAR T cells, as well as tolerable or no expression on healthy tissues to prevent toxicity. CD19 is now a widely accepted B lineage target of lymphoma and leukemia that is expressed in all tumor cells and absent in normal HSCs as well as all normal tissues ^[14][31]. Despite many discoveries related to the immunopathology of AML, a single AML-specific target has still remained elusive.

LSCs reside in a specialized niche that promotes their survival and chemoresistance, through which they can alter their microenvironment ^[15][41]. LSCs secrete pro-angiogenic VEGF and interleukins to stimulate angiogenesis to provide additional nutrients, oxygen, and growth factors and to promote proliferation ^[16][17][42,43]. AML blasts support a low-arginine microenvironment ^[18][44]. AML LSCs induce the expression of a growth-arresting protein, GAS6, in BM stromal cells ^[19][45]. AML cells promote the expression of immunomodulatory factors that impair cytotoxic T lymphocyte (CTL) activation in the tumor microenvironment ^[20][46]; these factors include programmed death receptor (PD-1), transforming growth factor β (TGF β), arginase II, prostaglandin E2 (PGE2), cytotoxic T lymphocyte-associated protein 4 (CTLA-4), lymphocyte activation gene 3 (LAG3), and T cell immunoglobulin and mucin-containing-3 (TIM3) on T cells ^[21][47]. Furthermore, leukemia cells modulate the NK cell receptor repertoire that inhibits NK cell activity ^[22][48]. The interactions are reciprocal: the niche cells also foster LSC growth. Assessment by scRNA-seq, the cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq), or single-cell assay for transposase-accessible chromatin (ATAC) sequencing has paved the way for the study of cellular heterogeneity and intercellular hierarchies and for the obtaining of insights into the cellular ecosystem of malignant and normal cells.

3. Promising Strategies to Overcome Challenges

3. Promising Strategies to Overcome Challenges

3.1. Safer Targets with Less ‘On-Target Off-Tumor’ Effect

3.1. Safer targets with less ‘on-target off-tumor’ effect:

Some potential targets appear more reliable in terms of off-tumor toxicity, especially on normal hematopoietic cells, but none of them have proven to be an ideal target in AML. C-type lectin like molecule-1 (CLL-1) is a type II transmembrane glycoprotein overexpressed in over 90% of AML patients on AML blasts, LSCs and differentiated myeloid cells absent in normal CD34+CD38- hematopoietic stem cells.  CD70, a ligand for CD27 identified as a type II transmembrane glycoprotein, was reported to be expressed on AML bulk cells and leukemic stem cells but not on normal hematopoietic stem cells. It showed promising anti-tumor effect without toxicity on healthy HSCs ^[23][56].  Efforts to discover efficient targets for immunotherapeutic strategies accelerated following technical progress in proteomic and transcriptomic assays. These assays help to understand cellular behavior on the protein level instead of immunophenotyping malignant cells. Hoffman et al described mass-spectrometry based phenotyping of HL60 and NB4 cell lines ^[24][63]. Perna et al performed surface-specific proteomic and transcriptomic studies in AML patients and normal tissues to indicate a potential therapeutic target. However, none of the surface proteins showed a similar expression profile to CD19. These studies suggested a combinatorial targeting strategy, which is discussed further in other sections ^[25][29]. Kohnke and colleagues aimed to discover de novo targets using cell-surface capture technology to detect the surfaceome, a set of proteins expressed on the surface of primary AML patient samples, including surface receptor, transporters, adhesion molecules among others. They identified three promising targets: CD148, ITGA4 and Integrin beta-7. Among these, Integrin beta-7 was the most favorable due to low or absent expression in healthy hematopoietic tissues ^[26][64]. In a recent scRNA-sec approach, two antigen targets-CSF1R and CD86 revealed potential targets for CAR T-cell therapy with broad expression on AML blasts accompanied by minimal toxicities toward relevant healthy cells and tissues ^[27][65].

3.2. Limiting the ‘On Target-Off Tumor’ Effect

3.2. Limiting the ‘on target-off tumor’ effect:

Unacceptably severe or prolonged toxicities, especially cytopenias and infections, can occur more frequently than in clinical trials with CD19 CAR T cells in lymphomas due to older and therapy-resistant patient populations. It is crucial to maintain anti-cancer immune surveillance and clinical efficacy while avoiding toxicity. Safety or suicide genes are widely used to alleviate toxicity from CAR T cells ^[28][29][30][54,66,67]. Alternatively, administration of cross-reactive CAR T cells can be a bridge to allogeneic stem cell transplants. Tasian et al showed three different approaches (anti-CD123 messenger RNA electroporated CAR T cells, administration of alemtuzumab, and administration of rituximab to CD20-coexpressing CART123) to eliminate CD123 CAR T cells without effecting the antitumor activity in murine models ^[31][68].

Many cell engineering approaches have attempted to improve safety in designing CARs, such as logic gating of T cell recognition and SnyNotch receptors ^[32][69]. Mutation in the anti-CD123 CAR antigen binding domain reduced the antigen binding affinity, as reported by Archangeli et al ^[33][70]. Mild adverse events were demonstrated in the interim analysis of a phase 1 trial of rapidly switchable universal CAR-T (UniCAR) targeting CD123 ^[34][35][71,72]. Benmebarek and colleagues generated a controllable CAR platform-synthetic agonist receptor (SAR) T cells only activated in the presence of their CD33 or CD123 scFv construct in vitro and in AML xenograft models ^[36][73]. Dimerizing Agent Regulated ImmunoReceptor Complex (DARIC) is a split receptor design that modulates CAR T cell activation by rapamycin to dimerize units. This approach aims to aid hematopoietic recovery and mitigate toxicity. Cooper and colleagues developed a lentiviral DARIC construct that targets a C2 splice isoform with the membrane proximal domain of CD33, and a Phase I study using this strategy is now open for enrollment ^[37][74]. ResWearchers have recently demonstrated the potential benefits of DOX-inducible CAR-T therapy, allowing the control of CAR-T using an external trigger. It is an effective and sensitive way to turn CAR-T activity ON or OFF in order to prevent unwanted side effects and reduce prolonged toxicities ^[38][75]. Potentially, the most feasible approach is to knock out the targeted antigen in normal marrow cells. CD33 deletion in primary HSPCs maintained their full function in terms of engraftment and differentiation, and it efficiently reduced off-tumor targeting while preserving on-tumor efficacy ^[39][40][76,77].

3.3. Combinatorial Antigen Targeting for Heterogeneity

3.3. Combinatorial Antigen Targeting for Heterogeneity:

Combinations of CARs against different AML targets might be a promising solution due to the lack of a leukemia-specific target antigen ^[32][69]. Previously, promising results were reported with dual or tri-specific CAR T cells against B-cell malignancies and solid tumor models to overcome the heterogeneity and antigen escape ^[41][42][78,79]. Indeed, the phenomenon of ‘antigen escape’ as a reason of failure in CAR T cells for AML has not been clearly demonstrated in previous pre-clinical reports ^[43][44][80,81]. Nonetheless, regardless of antigen expression levels, dual-targeting CAR T cells were associated with increased T-cell activation and proliferation. This effect might be due to increased interaction with the target cell, favoring immune synapse formation and subsequent T cell recruitment ^[45][46][82,83]. How to combine suitable pairs of antigens to enhance therapeutic efficacy without increasing off-tumor toxicity are still pieces of the puzzle yet to be solved. Perna et al suggested four possible combinatorial pairings (CD33+ADGRE, CLEC12A+CCR1, CD33+CD70 and LILRB+CLEC12A) to target AML with an algorithm integrating proteomics and transcriptomics ^[25][29].

The combinatorial antigen targeting strategy has already been applied in AML. Bicistronic CD123 and CD33 CAR T cells showed significant anti-tumor activity in artificially created cell lines (CD33+CD123-, CD33-CD123+) and in vivo ^[47][84]. In a phase I trial, CLL-1 and CD33 bicistronic CAR T cells reported remarkable results ^[48][85]. Similarly, 10 of 11 pediatric R/R AML patients infused with CLL-1 or CLL-1-CD33 dual CAR T cells had a response (5 reached CR/MRD-) without dose-limiting toxicities ^[49][86]. Atilla et al reported that dual targeting with either a CD33 CAR or a CD123 CAR and a CLL-1 CAR increased anti-tumor activity most profoundly when the target antigen expression on the tumor cells was low. Since the expression of each target antigen is highly variable, reswearchers chose to modify T cells with two separate vectors targeting CLL-1 and CD33/CD123 to get a mixed product rather than a bicistronic or tandem CAR design in which the molar ratio of each target is fixed [80]. A universal CAR T cell platform (on/off switching mechanism) successfully targeted CD33 and CD123 positive AML blasts in vitro and in vivo ^[50][87]. Haubner et al presented a novel combinatorial ADGRE-2 targeting CAR and CLEC12A-targeting chimeric costimulatory receptor (CCR) (IF-BETTER gate) that triggered high anti-leukemic activity in vitro and in vivo while sparing vital normal hematopoietic cells ^[51][88].

3.4. Neoantigens

3.4. Neoantigens:

Neoantigens are limited to malignant clones that arise from somatic mutation ^[52][53][89,90]. Because recurrent gene alterations can be shared by AML patients, neoantigens trigger potent anti-leukemic responses ^[54][91]. Distinct from other cancers, AML presents with low mutational burden, so recognizing neoantigens arising from mutations is rare ^[55][56][92,93]. Neoantigens from driver gene mutations appear to be ideal targets for immunotherapy since immune evasion is unlikely ^[57][94].

3.5. T Cell Receptor (TCR) T Cells for Treatment of AML

3.5. T cell receptor (TCR) T cells for treatment of AML:

T cell receptor (TCR) engineered T cells act through their modified TCRs and tumor-associated antigens (TAAs) presented by human leukocyte antigen (HLA) molecules on the surfaces of leukemic cells. The target protein can be expressed intracellularly or on the cell surface. TCR T cells have less stringent antigen requirements for T cell activation than CAR T cells ^[58][100]. TCR-T cell immunotherapy in AML has still barriers that need to be addressed. The major drawbacks are that TAAs might be expressed by non-malignant cells causing on-target, off-tumor toxicities, dose-related toxicity, limited persistence, and chance of immune escape ^[59][60][101,102]. Dose optimization of TCR-T cells, combining the treatment with exogenous cytokines (e.g., IL-21, IL-7 and IL-15), or adding genetically engineered signaling during cell expansion and demethylating agents such as decitabine might overcome the disadvantages of TCR-T cell application ^[61][103]. One other limitation of TCR transfer is the mispairing of endogenous and exogenous TCR components that impair the function; this limitation might be prevented by swapping the constant regions of mouse and human TCRs or codon-optimized cysteine modified TCRs in which TCR-α and β are linked by a T2A sequence ^[62][63][64][104-106]. Another approach uses TCR-like CAR T cells that contain scFv and CAR signaling mechanisms that recognize peptides in the context of MHC class I molecules ^[65][107].

3.6. CAR NK Cells

3.6. CAR NK cells:

NK cells are lymphoid cells involved in the innate immune response; they are programmed to kill virus-infected and malignant cells without causing significant graft vs host disease, CRS or neurotoxicity ^[66][115]. AML has been an attractive target for NK cell therapy as an allogeneic product ^[67][116]. Despite several manipulations for longer persistence of NK cells, the response to NK cell infusions varies without long-term remissions ^[67][116]. NK cells differentiated into cytokine induced memory-like NK cells following stimulation with IL-12, IL-15 and IL-18 and showed a distinct transcriptional and surface proteomic profile as well as enhanced functionality ^[68][117].  Successful application of CD33-targeted CAR-modified NK cells by transduction of blood-derived primary NK cells showed promising cytotoxicity with unimpeded proliferation in vitro and in vivo without observable side effects ^[69][122]. 

3.7. Manipulations in Manufacturing

3.7. Manipulations in Manufacturing:

AML is a highly aggressive disease affecting older populations. In a relapsed refractory setting, patients receive many lines of immunosuppressive therapies prior to apheresis, which affects T cell function, and the timeline of manufacturing raises serious concerns in practice. Optimal CAR construct design will preserve the naïve and central memory phenotype as well as the persistence of T cells. It has been shown that naïve and early memory T-cells have been enriched by decitabine administered with CD123 CAR T-cells ^[70][126].

Administration of off-the shelf ready-to-use products (allogeneic CAR T cells) generated from healthy donors will provide a valuable solution, since the CAR-T cells product are pre-manufactured without need for customized manufacturing for a specific patient.  DNA transposon systems are sophisticated systems for stable genetic modification that can deliver large genetic cargos and can be used to reduce cost ^[71][131]. Clinical-grade CAR-T cell products using Sleeping Beauty and piggyBac for multiple myeloma and leukemia are under investigation ^[72][73][74][132-134]. 

3.8. Strategies to Overcome the Negative Effects of Microenvironment

3.8. Strategies to Overcome the Negative Effects of Microenvironment:

There are several approaches described to modulate immunosuppressive microenvironment. Immune evasion such as upregulating immune checkpoint proteins has proven to be a way which can dampen the antitumor response and limit efficacy of CAR-T cell therapy. Although immune checkpoint blockade in AML has not proven beneficial, ^[75][76][77][136-138] there may be still additional effects in combining CAR T cells and immune checkpoint blockage (PD-1, CTLA-4, etc) that will improve T cell persistence and anti-tumor efficacy ^[78][139]. 

3.9. Allogeneic Hematopoietic Stem Cell Transplantation with CAR T Cells

3.9. Allogeneic Hematopoietic Stem Cell Transplantation with CAR T cells:

When and how to combine allo-HSCT with adoptive immunotherapy in AML is still debated. The mechanisms of resistance to T cell-mediated antitumor effects after allo-HSCT are well-defined in sophisticated murine models of allo-HSCT ^[79][145]. Combining a novel myeloablative irradiation-based conditioning regimen with regulatory and conventional T cell immunotherapy in haploidentical transplantation was shown to eradicate AML ^[80][146]. Published studies on how and when to combine CAR T cells in the setting of allo-HSCT showed conflicting results. 

4. Summary and Conclusions

4. Summary and Conclusion:

The tremendous advances in understanding the molecular and cellular mechanisms of AML have made it possible to manipulate the immune system and BM niches. Treating AML with CAR T cells is still in an immature stage. Experience with allogeneic stem cell transplantation, which is the most effective immune cellular therapy for AML, is guiding other directed therapies. One of the major challenges in developing CAR-T cell therapy for AML is the lack of a suitable antigen that is expressed uniquely on AML cells. Identifying and isolating target antigens that are homogeneously and stably expressed in all leukemic blasts and leukemic stem cells with limited on-target off-tumor toxicity, investigating complex interactions in the AML microenvironment, and seeking a suitable cell source all improve the fine-tuning of CARs.

Sophisticated methods for ex vivo manufacturing are now changing the in vivo dynamics and the character of the final product (Figure). In AML, personalization should be taken a step further in directed cellular therapies with platforms that will standardize the optimal CAR design for the target antigen or antigens in line with patient-specific immunophenotyping findings, the selection of a compatible carrier cell, and the cellular subtype.