Immune-based therapies in acute leukemia

Abstract

Treatment resistance remains a leading cause of acute leukemia -related deaths. Thus, there is an unmet need to develop novel approaches to improve outcome. New immune-based therapies with chimeric antigen receptor (CAR) T cells, bi-specific T cell engagers (BiTEs) and immune checkpoint blockers (ICB) emerged as effective treatment options for chemo-resistant B-cell acute lymphoblastic leukemia (B-ALL) and acute myeloid leukemia (AML). However, many patients show resistance to these immune-based approaches. This review describes critical lessons learned from immune-based approaches targeting high-risk B-ALL and AML, such as the leukemia-intrinsic (e.g. target antigen loss, tumor heterogeneity) and extrinsic (e.g. immunosuppressive microenvironment) mechanisms driving treatment resistance, and discusses alternate approaches aimed at enhancing the effectiveness of these immune-based treatment regimens.

PATHOGENESIS AND THERAPEUTIC TARGETING OF B-ALL AND AML

Acute leukemia is a leading cause of cancer-related death in children, and an outstanding clinical urgency in adults [1, 2]. This clonal hematopoietic neoplasm is characterized by the proliferation and accumulation of hematopoietic progenitor cells in the bone marrow that impede normal blood production, and that subsequently egress from the bone marrow to obstruct the function of major organs. The success of standard chemotherapy in the treatment of acute leukemia is highly dependent on the transformed hematopoietic lineage presenting at diagnosis as well as on a variety of leukemia-intrinsic genomic alterations with biological and clinical relevance. Therapies harnessing a patient’s own immune system emerged as a novel and effective means of treating cancer. In this review, we discuss the use of chimeric antigen receptor (CAR) T cells, bi-specific T cell engagers (BiTEs) and immune checkpoint blockade (ICB) in the treatment of acute leukemia. Immune-based approaches rely on host or engineered T cell function for tumor killing, and clearly, resistance to these approaches is the result of both leukemia-intrinsic and extrinsic mechanisms impacting the function of effector T cells effector. We highlight the lessons learned from application of immune therapies to B cell progenitor acute lymphoblastic leukemia (B-ALL) and acute myeloid leukemia (AML), followed by a discussion of alternative strategies circumventing resistance to emerging immunotherapies.

ALL is the most common form of cancer in children, accounting for approximately 30% of all pediatric cancers. Of these, approximately 85% of cases are B-ALL [1]. T-ALL represents an important minority of cases (15%) but because of current limitations of T cell-directed therapy (e.g. fratricide, whereby CAR-T cells target antigen present on healthy host T cells) this review will focus on B-ALL where significant advances have been made. B-ALL results from the transformation of B cell progenitors within the bone marrow [1, 3-5]. Diagnosis of B cell malignancies capitalizes on the immunophenotypic similarities shared between healthy and transformed B cell progenitors, such as the expression of B lineage transmembrane complex subunits, such as CD19, that normally act to regulate BCR activation signaling [6, 7]. In addition, B-ALL blasts express high levels of numerous B cell differentiation surface antigens, including cell surface ectoenzyme, CD10, immunoglobulin gene super-family member, CD22, and the activated-glycosylated phosphoprotein, CD20. [8]. Advances in conventional therapy over the past four decades led to a startling improvement in survival (>90% five-year survival), however, resistance and subsequent relapse associated with standard chemotherapy still remains a leading cause of pediatric cancer-related deaths [1]. Using genome-wide approaches, numerous studies characterized the underlying biological pathways that underpin the emergence and relapse of B-ALL. A Darwinian model of clonal evolution driving leukemia-intrinsic treatment resistance proposed based on the discovery of relapse-specific genetic and epigenetic alterations associated with drug resistance, including IKZF1, CREPPB and NT5C2 mutations [9-14]. While considerable success has been achieved targeting Ph+ (BCR/ABL) and Ph-like ALL there have been limited advances in therapies targeting genetic events responsible for relapse [15-17].

Acute Myeloid Leukemia (AML) also originates in the bone marrow, and while less common than ALL in children (accounting for 20% of pediatric leukemias [18]), it is one of the most common forms of acute leukemia in adults (approximately 80% of leukemias in this group [19]). Transformed myeloid progenitors share many immunophenotypic and functional characteristics of hematopoietic stem and progenitor cells (HSPC). This includes markers such as the Siglec family member CD33, which is expressed on common myeloid progenitors and their downstream progeny [20]. Genetic lesions driving AML can be found throughout multiple hematopoietic lineages and may promote pre-leukemic conditions, such as clonal hematopoiesis, consistent with the notion of genetic mutations affecting early hematopoietic progenitors (e.g. TET2, DNMT3A mutations) [21]. The genetic lesions driving AML are relatively distinct between childhood and adult cases [18, 22]. Furthermore, unlike B-ALL, AML may emerge from multiple hematopoietic stem cell and progenitor compartments with varying levels of epigenetic and genetic clonal heterogeneity [22, 23]. Despite considerable efforts, overall survival for AML remains poor especially in adults (<50% five-year overall rate), which can largely be attributed to limited advances in treatment regimens that, for the last few decades, have relied on the use of two non-targeted cytotoxic drugs - cytarabine and anthracyclines, such as daunorubicin. Gene discovery studies identified targetable genetic and epigenetic lesions that contribute to AML pathogenesis including internal duplication events in the FLT3 locus (FLT3-ITD mutations) that enhance the sensitivity of AML blasts to FLT3 inhibitors [24]. In addition, the use of CD33 directed immunotherapy, IDH inhibitors and BH3 mimetics have also shown promise[25]. While such progress is welcome, strategies overcoming AML heterogeneity are in their infancy.

IMMUNE-BASED THERAPIES IN B-ALL AND AML.

Immune-based approaches to target relapse/refractory cases of acute leukemia are emerging based on promising clinical results. Both CAR-T cells and BiTEs such as blinatumomab, were approved by the U.S. Food and Drug Administration (FDAⁱ) for use in multiple B cell malignancies, with early clinical trials demonstrating excellent responses in pediatric and adult relapse/refractory B-ALL [26-30]. In addition, ICB approaches are in trials for AML [31], based on positive responses seen in treating solid tumors [32]. These approaches share the requirement of efficient T cell mediated killing of target cells, despite vast differences in their mechanisms of action (Box 1).

Box 1: common immunotherapeutic approaches.

Immunotherapy is an attractive strategy for treatment of many types of tumor. Here we will give an overview of the general mechanisms of action of common immunotherapeutic strategies.

CAR-T cells:

CAR-T cells are generated from patient-derived autologous T cells, that are transduced with lentiviral CAR constructs allowing for antigen-specific engineered T cell receptor (TCR) function. Following ex vivo expansion, CAR-T cells are infused back into the patient to allow engineered T cells to eliminate antigen-expressing target cells. Prior to CAR infusion, patients are pre-conditioned with lymphodepleting agents, such as cyclophosphamide [57]. The main challenges in CAR-T therapy are identifying a suitable tumor antigen, with minimal expression on healthy cells, as well as production of sufficient numbers of T cells from patients, often following extensive chemotherapy.

BiTE / BiKE:

BiTE are bi-specific monoclonal antibodies, that simultaneously target T cells and tumor cells, thus forming a direct physical link between them [139]. This leads to killing of the tumor cell by the T cell, independent of MHC-I presence or interaction with co-stimulatory receptors [139]. BiTE are an attractive treatment because they represent an “off the shelf’ strategy, unlike CAR-T that have to be uniquely tailored for each patient. Another attractive possibility is targeting of NK cells in the tumor microenvironment, using bi-specific killer cell engagers (BiKE) or tri-specific killer cell engagers (TriKE). These are bi-specific antibodies targeting a tumor cell antigen and CD16, expressed on NK cells (BiKE) [140], or antibodies targeting a tumor antigen and CD16 combined with IL-15, inducing NK cell expansion (TriKE) [141]. Linkage of tumor and NK cells induces tumor cell killing by the NK cells.

Immune checkpoint blockade:

Tumors can induce an immunosuppressive microenvironment, preventing their clearance by the immune system. A key feature of this is T cell exhaustion, mediated by expression of immune checkpoint molecules on infiltrating T cells and on tumor cells. An important axis for T cell exhaustion in the tumor microenvironment is the PD-1 / PD-L1 interaction: PD-1, expressed on T cells, interacts with its ligand PD-L1, expressed on tumor cells. This interaction induces T cell exhaustion, characterized by poor effector function. T cell exhaustion is triggered by prolonged exposure to antigen stimulation, and is also characteristic of chronic infections [142]. In recent years, the emergence of immune checkpoint blockade has revolutionized the treatment of many solid tumors. Treatment with antibodies blocking the PD-1 / PD-L1 interaction leads to re-invigoration of T cells in the tumor microenvironment and is remarkably effective in solid tumors [143].

CAR-T cells

CAR-T cell therapy has been highly successful in the treatment of relapse/refractory B-ALL. Early trials in B-ALL therapy utilized “second-generation” CAR constructs, such as FDA-approved ‘Tisagenlecleucel’, whereby autologous T cells are transduced with lentiviral vectors expressing a CAR comprised of single-chain fragment (scFv) from a murine monoclonal antibody specific for CD19, a CD3-zeta domain for T-cell activation, and either a 4-1BB (CD137) or CD28 domain for TCR co-stimulation [33-37]. Upon patient infusion, CAR-T cells engage antigen-expressing target cells in an MHC-independent manner to elicit a cytotoxic response. This approach is effective especially since many patients have been highly pretreated with chemotherapy. In spite of dramatic results seen with CAR-T cell therapy, a significant subset of patients develop resistance.

The success of CD19-CAR-T treatment in B-ALL led to increased interest in identifying suitable targets for CAR-T therapy in AML. An ideal CAR-T target candidate should be widely expressed on malignant cells, to ensure eradication of the disease, and be rarely expressed on healthy cells, to avoid treatment-induced toxicity [38]. To date no ideal target has been identified in AML. The most prominent candidates for CAR-T therapy in AML are CD33 [39], CD123 [40] and CLL1 [41]. These and other potential targets have all been extensively reviewed elsewhere [42], and we will not delve into their individual mechanisms. However, pre-clinical models for these targets showed widespread effects on non-AML cells. Therefore, prolonged CAR-T treatment aimed at these targets will likely prove to be challenging, as discussed later in this review.

BiTEs

Blinatumomab in B-ALL links host CD3⁺ T cells with CD19⁺ B cells via a bispecific T cell-engaging single-chain antibody construct, thus activating T cells to promote cytolysis of proximal CD19⁺ cells [43]. In AML, current strategies are similar to CAR-T strategies, with emphasis on targets that are expressed on the majority of AML blasts, such as CD33 and CD123 [44], which are currently in clinical trials. Additional strategies target cell-surface proteins that are often overexpressed or mutated in AML blasts, such as FLT3 [45]. However, these strategies also face the challenge of overcoming the lack of a specific marker for AML blasts, which may lead to significant adverse effects.

Immune Checkpoint Blockade

One of the key determinants for successful ICB is the presence of tumor neo-antigens, which is associated with a high mutational load. However, compared to highly-mutated solid tumors, B-ALL and AML have a very low mutational load, with the exception of cases harboring mutations affecting DNA mismatch repair genes [46-48], rendering these diseases unattractive candidates for ICB. In AML, malignant blasts express PD-L1 [49, 50], yet ICB has yielded only modest results in clinical trials. Hypomethylating agents (HMA) are a well-established treatment strategy for AML, targeting the dysregulated epigenetic machinery in AML blasts [51]. However, treatment with HMA leads to upregulation of several checkpoint molecules in AML patients [52]. A clinical trial combining ICB with the hypomethylating agent Azacitidine has shown remarkable efficacy, with an overall response rate of 58% in HMA-naïve patients [53]. This provides an important advance in treatment of elderly patients, who are not candidates for HCT. Notably, profiling of AML patients relapsing after allogeneic HCT revealed that T cell exhaustion is a significant contributor to failure of the graft vs leukemia response and to relapse [54, 55], indicating that ICB can also be an attractive strategy for treatment of these patients.

LESSONS LEARNED FROM IMMUNE-BASED THERAPIES

Optimizing the design of therapeutic agents

Clinical use of immune-based therapies relies heavily on the ability of the agent to effectively engage its target to elicit a strong and enduring immune response. CAR-T cell design has become a major area of research with the ultimate goal of maximizing tumor cytotoxicity upon antigen engagement, whilst allowing the formation of long-lived memory CAR-T cells. These requirements are interconnected, as highlighted by numerous studies exploring the effectiveness of different CAR-T signaling complex configurations. Early CAR designs utilized highly active TCR signaling components, such as the CD28 co-stimulatory domain [33, 56]. Despite robust early responses to antigen, T cell persistence and a lack of T cell memory formation prevented enduring responses to CAR-T therapy, leading to CD19⁺ B-ALL relapses. Subsequent trials utilized a 4-1BB signaling co-stimulatory domain coupled to CD3ζ signaling domain and single-chain antibody recognition domain recognizing CD19 [34, 57], allowing greater memory T cell formation whilst retaining anti-tumor function [58]. However, it should be noted that CD19⁺ B-ALL relapses following CAR-T therapy also resulted from insufficient CAR-T expansion upon infusion [36] (Figure 1).

Figure 1. Intrinsic and extrinsic regulators of B-ALL-targeting CAR-T and BiTE therapy.

CAR-T and BiTE therapy targeting CD19⁺ B-ALL rely on effective T cell function to exert antileukemic function. Multiple extrinsic regulatory mechanisms can potentially impact antileukemic T cell function, including hematopoietic (e.g. MDSCs, Tregs), non-hematopoietic (e.g. vascular and perivascular niche) and biochemical (e.g. hypoxia) factors. CAR-T and BiTE therapy escape may be driven by intrinsic mechanisms, such as the existence of CD19- B-ALL blasts prior to therapy, leukemic lineage switching and target antigen reduction. In addition, extrinsic mechanisms such as T cell exhaustion, insufficient CAR-T expansion, antigen masking by CAR-transduced B-ALL blasts, antigen loss via trogocytosis, and a protective leukemic niche may all promote resistance to immune-based therapies.

T cells are isolated from the blood of leukemia patients, and the quality of harvested cells presents a key obstacle for manufacturing. Minimizing culture duration throughout the transduction process, achieving complete depletion of circulating B-ALL blasts, isolating high quality naive T cells from patients, and generating the ideal composition of T cell subsets following ex-vivo culture (naive versus memory) all present major manufacturing challenges when attempting to maximize in-vivo CAR-T cell persistence [59-61]. For example, CD4⁺ and CD8⁺ T cells display divergent responses to CAR signaling, ultimately impacting the cytotoxic and memory capacity of infused CAR-T cells [62, 63]. CD28-based CD8 CAR-T cells generate rapid and potent cytotoxic responses with poor T cell persistence, whereas CD28-based CD4 CAR-T cells retain some cytotoxic function, but possess a higher capacity to form persistent memory T cells [64]. Endogenous TCR signaling also deleteriously effects CD8 CAR-T function, however, introduction of CAR sequences directly into the endogenous TCR locus may allow for appropriate tonic T cell signaling whilst minimizing the chance of random mutagenesis caused by lentiviral integration [65]. Although BiTE therapies, such as blinatumomab, avoid the challenges associated with cellular engineering, these low molecular weight therapeutics are limited by the need for constant infusion [66]. Blinatumomab’s short half-life of 2 hours and efficient renal clearance has led to generation of more stable agents, such as AFM11, which allow weekly infusion [67].

Immune therapies toxicities

Whilst ICB therapy is associated with many reported toxicities [68], adverse side-effects associated with CAR-T therapy and BiTE therapy have been extensively reported with frequent events of cytokine release syndrome (CRS) capable of driving a life-threatening multiple organ dysfunction syndrome, neurotoxicity, and the expected B cell aplasia [43, 69-72]. CRS-associated neurotoxicity, including headaches, seizures and rare cases of acute cerebral edema, occurred in multiple CAR-T B-ALL clinical trials, despite rarely occurring in solid tumor CAR-T trials (reviewed in [73]). The risk of CRS is decreased when treatment is initiated at low tumor burden and effectively managed using anti-inflammatory agents, such as IL-6 receptor antagonizing antibody, tocilizumab, and corticosteroid treatment [70]. B cell aplasia, induced by CD19-CAR-T therapy, results in loss of antibody production, known as hypogammaglobulinemia, rendering patients susceptible to infection, however, this condition can be addressed through immunoglobulin replacement therapy [74]. Limiting T cell activity is a major concern in CAR-T therapy-induced CRS, with efforts aimed at limiting potential off-target activity or excessive immune therapies toxicities [75]. One alternative is the use of inducible “suicide genes” in the constructs, which will enable control of CAR-T expansion and survival in the patients [76, 77].

In AML, expression of CAR-T targets on non-AML hematopoietic cells leads to severe toxicities due to bone marrow failure [42]. Furthermore, some potential targets are expressed on other, non-malignant tissues. CD33, the most prominent candidate for AML CAR-T therapy, is expressed on Kupffer cells in the liver and targeting CD33 can induce hepatic toxicity [78]. Due to the toxicity of CAR-T therapy, strategies utilizing transiently active CAR T cells [79] may be used in AML therapy as a means of inducing disease remission, hence providing a bridge to hematopoietic stem cell transplant. In this scenario, CAR-T cells will be used for a short period of time as a means to eliminate malignant cells, and the transplant will mitigate the damage induced by CAR-T killing of healthy hematopoietic cells. In addition, CRISPR/Cas9 mediated CD33-deficient autologous hematopoietic progenitors as a rescue stem cell source are potential strategies to facilitate entry of CAR T therapy into the clinic [80].

Tumor heterogeneity

Even with the most potent immune-based therapy, the stability and abundance of target antigen expression on the surface of leukemic blasts is critical for CAR-T and BiTE success. Therefore, a greater understanding of heterogeneity present within an individual patient’s leukemic blast population may inform the use of antigen-specific treatment agents (Figure 1). For example, a study of MLL-rearranged B-ALL patients receiving CD19 CAR-T showed successful elimination of CD19⁺ cells, but two patient relapse cases were reported with CAR-T-resistant MLL-rearranged AML arising from two distinct mechanisms [81]. First, one patient relapsed with AML blasts harboring the specific IGH rearrangements detected at B-ALL diagnosis, suggesting de-differentiation of CD19⁺ B-ALL blasts to the CD19⁻, CAR T-resistant myeloid lineage. Conversely, the second case of AML relapse emerged from pre-existing transformed myeloid clone unrelated to the diagnosis B-ALL population. Lineage-switching has also been reported in pre-clinical models of CAR-T therapy, and strongly suggests that more extensive screening may be necessary to circumvent such issues [82].

Antigen Escape

Loss or reduction of antigen expression remains one of the main mechanisms of CAR-T relapse, and to a lesser extent blinatumomab-resistance [83]. In B-ALL, CD19, and more recently CD20 [84] and CD22 [85], have been recognized as ideal CAR-T targets based on their restricted B-lineage expression. Although these approaches result in B cell aplasia, there is evidence of sustained pre-existing humoral immunity in CAR-treated patients, whilst immunoglobulin infusion provides further immune protection [74]. CD19, however, is believed to be dispensable for B-ALL blast survival [86]. Therefore, CAR-T treatment may select for blasts downregulating surface CD19 expression without consequence to leukemic fitness. In the case of CD19-directed therapies, leukemia-intrinsic reduction of target antigen expression results from CD19 locus-specific genomic deletions or the selection for expression of treatment-resistant mRNA splice isoforms [87, 88]. Rare mechanisms of CD19 down-regulation in human CAR-T patients include CD19 retention in the Golgi apparatus [89] and CD19 antigen masking via CAR-transduced B-ALL blasts [60]. In addition, CD19 antigen transfer to interacting CAR-T cells, through a process known as ‘trogocytosis’, occurs in B-ALL xenograft studies [90]. In summary, lineage switch and antigen reduction reveal that immune-based approaches are subject to the same evolutionary pressures observed in patients who relapse from conventional therapy emphasizing that tumor heterogeneity remains the primary obstacle for therapy of cancer (Figure 1).

Immune microenvironment

Immune-based treatments rely on both efficient T cell signaling and a permissive extrinsic milieu, suggesting the tumor microenvironment may also represent a barrier to tumor cell killing. Cases of leukemic relapse following CAR-T therapy in which CD19 antigen expression is sustained are believed to result from CAR-T cell exhaustion or lack of expansion, however, few studies have attempted to map the composition of the immune microenvironment throughout the course of immune therapy that may cause this phenomenon. Activation of monocytes and macrophages during CAR-T therapy of murine B-ALL drives CRS via the excessive secretion of IL-1 and IL-6 [91, 92], consistent with human studies of blinatumomab CRS [93]. Further, regulatory T cell (Treg) abundance is linked to blinatumomab efficacy in the treatment of high-risk B-ALL patients [94]. Depletion of myeloid-derived suppressor cells (MDSCs) significantly enhanced intra-tumoral CAR-T migration and tumor killing in neuroblastoma xenografts [95], however, the impact of MDSCs in B-ALL and AML pathogenesis remains controversial. Therefore, studies will be required to shed light on the contribution of immunosuppressive immune cell types, such as Treg and MDSCs, in inhibiting immune treatment (Figure 1). Notably, the immune microenvironment may be significantly different in pediatric and adult patients, leading to different responses.

Non-hematopoietic microenvironment

Leukemia cells dramatically alter the vascular and stromal niche providing a refuge from chemotherapy-induced killing. Under healthy conditions, bone marrow-resident HSCs are believed to localize to sinusoidal vessels associated with low oxygen concentrations resulting in the stabilization of Hypoxia Inducing Factors (HIFs), such as HIF-1α [96, 97]. Whilst hypoxic conditions favor HSC quiescence and maintenance of stem cell function, induction of HIFs drives B-ALL and AML chemo-resistance [98, 99]. As a consequence of these re-modeling events, T cells are hindered in their ability to home, adhere and extravasate into these leukemic reservoirs. Solid tumor studies show significant re-modeling of tumor-associated vascular endothelium that hinder T cell homing to the tumor site, thus, blocking the effects of T cell immunotherapy [100]. AML progression causes dramatic increases in vascular permeability and decreased blood flow, resulting in the formation of a hypoxic leukemic niche [101]. In addition, pre-clinical B-ALL models show disease progression coinciding with an increasingly hypoxic bone marrow niche [102], and a significant decrease in the motility of B-ALL-associated T cells in vivo [103]. In B-ALL, fibrosis is predictive of treatment outcome [104], but the role of matrix remodeling in immune cell function and in leukemia resistance still needs to be determined. Collectively, these finding may have important implications in the use of CAR-T and BiTE therapy, as small reservoirs of leukemia cells may be inaccessible to T cell-mediated killing due to the defects in T cell homing and motility within these aberrantly vascularized and hypoxic areas (Figure 1).

ADDRESSING OBSTACLES TO EFFECTIVE IMMUNE-BASED THERAPY

Patient screening at single cell resolution

Single cell studies have provided unprecedented resolution when analyzing the heterogeneity that exists within the hematopoietic stem and progenitor cell compartment. It is now possible to identify specific sub-clonal mutations that may help define the cellular origin of disease. As such, using the example of CD19-directed therapies, screening for the existence of CD19⁻ leukemic blasts prior to CAR-T therapy using flow cytometry and, potentially, single cell genotyping and transcriptomic approaches may assist in identifying patients at risk of failing treatment. For example, single cell analysis of HSCs from chronic myeloid leukemia patients allowed the identification of quiescent BCR-ABL1⁺ stem cells that failed to respond to tyrosine-kinase inhibition therapy, as well as identification of cell extrinsic signaling events affecting non-leukemic BCR-ABL1⁻ stem cell function [105]. More recently, van Galen and colleagues utilized single cell genotyping to track the presence of AML-associated FLT3-ITD mutations throughout the hematopoietic compartment, and found significant inter-patient variation in predicted transformed hematopoietic progenitor subpopulations [23]. In the case of B-ALL, multiple studies suggest transformation occurs at committed B cell progenitor stages, however, there are many examples of B-ALL-associated genetic lesions present within multipotent hematopoietic progenitor compartment [3-5]. Although this approach may be difficult to implement in the clinic in the near future, high-resolution single-cell approaches may enable identification of patients at risk of treatment resistance, and highlight the potential for targeting both myeloid and lymphoid leukemic blasts simultaneously in these patients to mitigate disease relapse.

Genetic profiling of patient HLA subtype has also revealed significant impact on response to ICB [106], and may also determine response to other types of immunotherapy. In AML, HLA subtype is an important factor in determining activation of NK cells following IL-2 treatment, and patients carrying the HLA-B-21M subtype harbor better-educated NK-cells with a higher capacity to degranulate when stimulated with AML blasts [107]. It is likely that as more high-resolution studies are performed, other determinants of response to therapy will emerge.

Identifying and sustaining the target antigen

Most treatment strategies rely on sustained antigen expression, maximizing the ability of effector cells to exert anti-leukemic effects. Little is known about leukemic cell-intrinsic regulators of antigen expression throughout the course of CD19 CAR-T therapy. During normal B cell development, for example, B cell lineage master-regulator, PAX5, directly activates the expression of CD19 during early B cell commitment [108, 109]. In classical Hodgkin Lymphoma, reduced CD19 surface expression by hypermethylation of the CD19 promoter was readily reversible by the DNA demethylating agent 5-aza-deoxycytidine [110]. In human B-ALL cell lines, targeting of CD22 with an immunotoxin resulted in treatment resistance caused by selection for primitive progenitor CD22⁻ B-ALL blasts [111]. Here, combination of immunotoxin and 5-aza-deoxycytidine treatment mitigated the incidence of CD22 downregulation, however, the specific effect on CD22 locus methylation remains unclear. Therefore, multiple mechanisms including epigenetic mechanisms, such as DNA methylation, can regulate the surface abundance of immunotherapy targets. A greater understanding of how extrinsic signaling events that occur throughout CAR-T therapy re-wire the expression program of leukemic blasts to drive antigen downregulation may shed light on novel interventions promoting the sustained expression of leukemia-specific target antigens.

Unlike B-ALL, AML blasts share many antigens with HSPC that are critical for effective hematopoiesis. Therefore, targeted strategies such as CAR-T or BiTEs may be used as a pathway to clinical remission prior to hematopoietic stem cell transplantation (HSCT), and rely on conditional ablation of CAR-T cells prior to HSCT. Notably, multilineage engraftment following transplantation of CD33-deficient HSCs in mice and rhesus macaques, rendered healthy HSCs resistant to CD33 CAR-T killing, however, whether this genome editing-based approach is safe for human applications remains to be seen [80].

Disease heterogeneity may support the use of a CAR-T with two or more independently-activated CARs [112]. In many respects this is the most obvious solution replicating decades of experience with multi-agent conventional chemotherapy. Pre-clinical work has identified several suitable targets for combinatorial targeting [112, 113]. These include combination of CD33 and CLL1, which are currently being evaluated independently for AML, as well as combinations with novel targets. Analysis of surface protein expression in AML has identified novel target combinations such as ADGRE2⁺CD33⁺, CCR1⁺CLL1⁺, CD70⁺CD33⁺ and LILRB2⁺CCL1⁺ [112] (Figure 2). Early clinical trials are underway assessing the function of tandem CD19/CD20 CAR-T constructs in B-ALL therapy, hence, minimizing the chances of target cells reducing the expression of both target antigens [84, 114]. Another alternative for AML may be the use of logic-gated synthetic Notch receptors as a means of discriminating antigens expressed on tumor cells and healthy tissue [115]. Here, engineered T cell require the engagement of two distinct antigens for T cell activation, which has proven to be successful in targeting cancers expressing tumor-specific neo-antigens and/or surface molecules absent in healthy adult tissue [116].

Figure 2. Immune-based approaches for the treatment of AML.

AML blasts express cell surface molecules that can be targeted by CAR-T cells or BiTE. AML cell surface molecules can be targeted individually, by a single-specificity CAR-T cell, or in combination, by dual-specificity CAR-T cells. BiTE can be used to connect CD3⁺ T cells to AML surface molecules to induce killing of AML blasts. The bone marrow vascular niche secretes CXCL12, which binds CXCR4 on AML blasts and mediates homing of leukemic cells to the bone marrow. This axis can be blocked by CXCR4 inhibitors. CSF1R⁺ cells in the AML microenvironment support AML growth, and can be targeted by CSF1R inhibitors.

Disarming a hostile microenvironment

The tumor microenvironment plays an important role in mediating immunosuppression, and can also be considered as a target for enhancing the efficacy of immunotherapy [117]. While there is greater understanding of the microenvironment in solid tumors, there is significant evidence for a role for the bone marrow microenvironment in supporting hematological malignancies [118]. The bone marrow microenvironment consists of mesenchymal cells, vascular cells and hematopoietic cells, which can shelter and nurture the malignant cells. The bone marrow niche has an established role in maintaining hematopoietic stem cells [119], and there is mounting evidence for a role of the bone marrow niche in maintaining leukemic cells. Further, recent single-cell studies have shown the vast heterogeneity that exists across the hematopoietic and non-hematopoietic bone marrow microenvironment in both healthy and malignant conditions [120, 121]. Therefore, comprehensive mapping of the leukemic bone marrow microenvironment may enable targeting of the bone marrow niche as an effective strategy in enhancing responses to treatment in leukemia.

One well-characterized example of an interaction between the bone marrow niche and leukemic cells is the CXCL12-CXCR4 axis [122, 123] (Figure 2). Bone marrow stromal cells secrete CXCL12, the ligand for CXCR4, which is highly expressed by leukemic cells. Blocking this interaction prevents homing of leukemic cells to the bone marrow and makes them more susceptible to therapy. Preclinical and clinical studies targeting the CXCR4:CXCL12 axis have revealed biological activity and safety of this approach. Ongoing phase III trials will be needed to establish efficacy. CXCR4 antagonists are currently being investigated in clinical trials for AMLⁱⁱ [124]. Their combination with immunotherapy has been tested in pre-clinical cancer models of ovarian cancer [125] and may also be efficient for acute leukemias.

Combined approaches disarming extrinsic suppressive factors affecting T cell function are beginning to be explored in both pre-clinical and clinical settings. “Armored” CAR-T cells allow targeting of tumor cells, whilst simultaneously manipulating the tumor microenvironment to favor efficient killing [126]. Transmembrane protein, CD40L, is normally upregulated on activated T cells to enhance antigen presentation from proximal antigen presenting cells (APCs) [127]. Accordingly, CD19 CAR-T cells constitutively expressing CD40-L promoted the recruitment and function of APCs, enhanced the pro-inflammatory milieu and facilitated immune effector cell expansion in a pre-clinical model of lymphoma. In addition, CD19 CAR-T cells constitutively expressing IL-18 also enhanced responses to CAR-T through similar pro-inflammatory mechanisms, and a possible increase in formation of both endogenous and CAR memory T cells [128].

CAR-T cells capable of secreting PD-1-blocking single-chain variable fragments (scFv) have shown excellent pre-clinical utility, as this approach allows for localized paracrine and autocrine checkpoint inhibition to enable efficient target killing [129]. This sophisticated approach complements clinical trials in which PD-1 inhibitors, Pembrolizumab or Nivolumab, will be combined with CD19 CAR-T cell therapy in the hope of circumventing CAR-T cell exhaustion during treatment of relapsed/refractory cases of either B cell Non-Hodgkin lymphoma or B-ALL [130, 131].

Controlling vascular endothelial integrity and hypoxia would be another potential means of unleashing T cell effector function within largely inaccessible leukemic niches. Interestingly, targeted inhibition of the hypoxia-induced NOX4-NOS3 axis partially reverses vascular leakiness induced by AML xenografts [101]. It is conceivable that restoration of vascular integrity may increase local oxygen levels as well as the motility of leukemia-targeting T cells into previously poorly-perfused regions.

Pre-clinical platforms for testing complex interactions

Pre-clinical systems are an invaluable tool for understanding the underlying mechanisms governing the success of immunotherapies. To date, the majority of pre-clinical efforts have been dedicated to optimizing the performance of CAR-T cells. Expanding knowledge of non-hematopoietic participants in CAR-T responses, such as the vascular endothelium and perivascular niche, may provide important insights into CAR-T cell effector function within the chemo-resistant leukemic niche, and help identify novel targets for therapy. One potential opportunity, successfully applied in solid tumors [132], is the use of ex-vivo organotypic biomimetic platforms aimed to recapitulate a humanized leukemic niche. These systems allow monitoring of immune and non-hematopoietic cell types infused into ex-vivo devices in response to manipulation of extrinsic factors, such as hypoxia, media composition and extracellular matrix integrity. Using such an approach, it may be possible to mimic different immune treatment conditions using primary human cells.

Alternate strategies for immunotherapy

While the strategies described above take advantage of T cells in the leukemic microenvironment, targeting innate sensing pathways can also be an efficient therapeutic strategy in AML. AML originates from myeloid cells, that express high levels of different pathogen recognition receptors (PRR). PRR engagement results in activation of a strong inflammatory response, with the goal of eliminating pathogens [133]. This can potentially be co-opted to eliminate tumor cells.

In murine models of AML, activation of the cytoplasmic DNA sensor STING leads to production of inflammatory cytokines IFN-β, TNF and IL-6, maturation of antigen presenting cells, and activation of tumor specific T cells. This, in turn, prolonged survival in a murine AML model and led to generation of long-term memory against AML-specific antigens [134]. STING agonists are currently in clinical trials for other malignanciesⁱⁱⁱ.

Inflammasome activation results in production of IL-1β, a key cytokine required for maturation of hematopoietic stem cells [135]. Therefore, triggering inflammasome activation may induce differentiation of leukemic stem cells. Indeed, in a mouse model of AML, activation of the inflammasome by RIPK3-mediated necroptosis led to elimination of leukemia-initiating cells and to differentiation of transformed cells. RIPK3 levels were found to be reduced in samples from AML patients, suggesting that suppression of this pathway is necessary for AML development [136]. These reports suggest that targeting of the innate immune sensing machinery can be an efficient strategy to eliminate malignant hematopoietic cells. However, broad activation of these pathways can lead to significant toxicity, and should be implemented with care.

The myeloid origin of AML can also play a role in disease development. In mice, deficiencies in antigen presentation by dendritic cells to CD4+ T cells can lead to myeloproliferative disease, suggesting that functional DC-T cell interaction is crucial for maintaining myeloid cells in check. Naive CD4+ T cells produce high levels of Flt3 ligand, suggesting a potential mechanism for myeloid expansion in this model [137]. However, the role of antigen presentation in human AML development remains unclear.

Another attractive immunotherapeutic strategy is targeting of other immune cells in the tumor microenvironment. One notable example in AML is the M-CSF receptor, encoded by CSF1R, expressed on tumor-supporting cells in the bone marrow (Figure 2). CSF1R-expressing cells secrete HGF and other cytokines required for AML growth. While the identity of the CSF1R-expressing cells remains unclear, targeting of these cells in pre-clinical models, via CSF1R inhibition, led to disease regression [138], and CSF1R blockade is currently being evaluated in clinical trials^iv.

CONCLUDING REMARKS

While there has been substantial progress in the treatment of acute leukemia, particularly in children, and promising new approaches offer optimism, far too many patients succumb to the disease. Harnessing the immune system against leukemic cells can be a powerful treatment strategy for these patients. Immune-based therapies, such as CAR-T cell and BiTE therapy, have proven to be effective means of targeting chemo-resistant acute leukemia, however, efforts to improve drug design and manufacturing will be paramount to achieving optimal T cell effector and memory function. Future studies aimed at understanding the mechanisms of resistance – both intrinsic and extrinsic may lead to novel combination approaches that effectively permit effector T cell function whilst maintaining the targeted antigen expression on the surface of leukemic blasts (see ‘Outstanding Questions’). In addition, technological advances, leading to improved characterization of disease heterogeneity and composition of both the hematopoietic and non-hematopoietic microenvironment, offer precision-based approaches to optimizing immune-based therapies.

OUTSTANDING QUESTIONS.

• What is the optimal drug design of CAR and BiTE therapies to enhance both T cell effector function and the ability to form robust T cell memory?

• What quality control measures should be put into place for engineered T cell manufacturing to ensure robust T cell expansion and enduring CAR activity following infusion?

• What are the key intrinsic regulators of target antigen expression, and how is the function of these regulators impacted throughout the course of immune-based therapy?

• Can we augment the activity of antigen regulators to provide higher and prolonged antigen availability throughout the course of treatment?

• Prior to immune therapy, can we identify leukemic sub-clones lacking antigen expression and, as such, formulate alternate treatment strategies to mitigate patient relapse?

• How does leukemia re-model its microenvironment to foster blast survival?

• Which components of the leukemic microenvironment – hematopoietic or non-hematopoietic – effect the function of effector T cells throughout immune therapy?

• What other components of the leukemic microenvironment can be targeted for therapy?

• Can we identify combination approaches that allow for optimal effector T cell function, whilst overcoming critical intrinsic and extrinsic mediators of antigen escape?

HIGHLIGHTS.

• Early CAR-T and bi-specific T cell engagers (BiTEs) are effective immune-based approaches to eliminate chemo-resistant leukemia, however, many patients will not respond to these novel agents.

• Improved CAR-T and BiTE design and manufacturing practices will be paramount to generating optimal effector T cell function and the generation of a persistent memory T cell pool.

• The effectiveness of immune therapy relies on both leukemia-intrinsic and leukemia-extrinsic factors.

• Understanding the heterogeneity that exists within the leukemic blast population and surrounding leukemic microenvironment may allow preempting of potential therapy relapse.

• Potential combination approaches should minimize the chance of antigen escape whilst overcoming key leukemia-intrinsic and environmental factors that negatively impact T cell effector function.

Glossary

B-cell aplasia

Absence of B-cells following CAR-T killing of CD-19 expressing cells

Blast

An undifferentiated hematopoietic cell, rarely found in the blood of healthy individuals and found in high percentages in leukemia patients

Clonal evolution

Accumulation of mutations and epigenetic changes in individual cells, affecting their proliferative capacity

Clonal hematopoiesis

An aging-related phenomenon where genetically distinct subpopulations of hematopoietic cells arise from hematopoietic stem cells or early progenitors

Co-stimulation

Additional signals required for activation of T or B cells, which are provided through interaction of molecules like CD28, present on T or B cells, with molecules on antigen presenting cells

Graft vs leukemia

Elimination of malignant leukemic cell by donor T cells after hematopoietic stem cell transplantation

Hematopoietic stem and progenitor cells

Immature cells that can develop into all types of blood cells, including white blood cells, red blood cells and platelets

High-risk acute leukemia

Disease that is more resistant to treatment and more likely to relapse

Mutational load

The number of deleterious mutations occurring in a genome

Myeloid derived suppressor cells (MDSC)

Immature myeloid cells that are prevalent in chronic inflammation or in tumors, and play a role in suppressing immune responses

Pre-clinical models

Cell and animal models for evaluation of the efficacy of potential therapies before their examination in clinical trials

Second-generation CAR-T

Autologous T cells transduced with lentiviral vectors expressing a CAR specific for a tumor antigen, CD3ζ domain for T-cell activation and 4-1BB or CD28 domains for TCR co-stimulation

Single-chain antibody

A fusion protein of the variable regions of immunoglobulin heavy and light chains, connected by a flexible linker peptide. The fusion protein retains the specificity of the original immunoglobulin

Tumor neo-antigens

Peptides that arise in tumor cells due to tumor-specific mutations, that can be recognized by the host immune system

Xenograft

Transplantation of patient-derived cells into an animal recipient to model disease in a pre-clinical setting