A CD38-directed, single-chain T-cell engager targets leukemia stem cells through IFN-γ–induced CD38 expression

Despite therapeutics recently approved by the US Food and Drug Administration, only 30% of adult patients with acute myeloid leukemia (AML) are alive at 5 years from diagnosis.^1,2 Both initial treatment refractoriness and postremission relapse remain major challenges.^3,4 Allogeneic hematopoietic stem cell (HSC) transplantation is the only curative approach for most patients with AML, but treatment-related morbidity and mortality limit its use to clinically fit and younger patients.^5,6 As such, a significant number of patients with AML are in dire need for more effective and less toxic therapies.

In AML, leukemia stem cells (LSCs) are primitive clonal cells capable of self-renewal and disease initiation and maintenance.⁷ These cells are highly treatment resistant,⁸ and therefore are likely responsible for disease refractoriness and relapse. Compared with their “bulk blast” progeny, AML LSCs are characterized by quiescence, distinct metabolism, and phenotypic plasticity.⁹ They are, however, difficult to identify because they can reside in immunophenotypic-diverse leukemic cell subpopulations¹⁰ and share common antigens with both bulk blasts and normal HSCs.¹¹ In human patients with AML, LSCs are frequently enriched in the CD34^posCD38^neg fraction, whereas more differentiated AML progenitors and bulk AML blasts reside in the CD34^posCD38^pos fraction.⁹ 

Bispecific T-cell engager antibodies that redirect T cells to killing of cancer cells by simultaneously binding a tumor associated antigen (TAA) on cancer cells and the CD3 receptor on T-cells have shown robust clinical activity in hematological cancers.¹² Several AML-associated TAAs have been targeted by T-cell engagers, including CD33, CD123, CD13, and C-type lectin-like molecule-1 (CLL-1).¹³ Although these targets are ubiquitously expressed on AML blasts and LSCs, they are also highly expressed by normal HSCs,¹⁴ leading to off-target activity and reducing their clinical efficacy.

CD38 is an ectoenzyme transmembrane glycoprotein involved in the metabolism of extracellular nicotinamide adenine dinucleotide.¹⁵ This antigen has successfully been exploited as a therapeutic target in multiple myeloma (MM) and T-cell acute lymphoblastic leukemia.^16,17 However, because LSCs are mainly CD38^neg, the therapeutic targeting of CD38 is of secondary interest in AML.¹⁸ We show, however, when targeted by a novel compact single-chain T-cell engager, CD38 is a very promising therapeutic target in AML. Specifically, CD38 expression is induced by interferon gamma (IFN-γ), which converts CD34^posCD38^neg into CD34^posCD38^pos blasts and marks the AML cell population for elimination using a CD38-directed immunotherapeutic. To exploit this feed-forward mechanism, we developed a novel single-domain, CD38-targeting T-cell engager (hereafter called CD38 biologic nested inside chains [CD38-BIONIC]). We demonstrate that, once engaged with CD38^pos AML cells, CD38-BIONIC (BN-CD38) induces T cells to release IFN-γ and convert bystander CD38^neg cells (including LSCs) into CD38^pos blasts. Consequently, BN-CD38 eliminates both CD38^pos and CD38^neg leukemic blasts and offers an attractive therapeutic approach for AML.

Samples from patients with AML were obtained from the City of Hope Hematopoietic Tissue Biorepository according to institutional review board–approved protocols and after written informed consent from all participants before inclusion in the study.

Single-cell RNA sequencing (scRNAseq) was conducted in primary AML bone marrow (BM) mononuclear cells (MNCs) treated either with vehicle or IFN-γ for 5 hours at 10 ng/mL. Bulk RNAseq was conducted in CD34^pos blasts purified from patients with AML, and exposed overnight to either vehicle or 10 ng/mL IFN-γ. Library preparation, sequencing methods, and data analyses are detailed in supplemental Materials, available on the Blood website.

Supplemental Materials detail the conceptualization, generation, and characterization of the single-chain T-cell engager called CD38-CD3 BIONIC (BN-CD38) and the control that is CD38-mutated BN-CD38 (BN-CD38Mut).

Primary peripheral blood (PB) or BM MNCs from patients with AML were treated with 1.0 ng/mL of BN-CD38 or controls that included BN-CD38Mut, CD38 nanobody, or human immunoglobulin G (IgG) for 5 days and stained with a customized 36–surface marker panel according to Standard BioTools standard cell surface staining protocol.

AML cell lines (THP1 and U937) xenografts and patient-derived xenograft (PDX) models were generated and handled under approved City of Hope institutional animal care and use committee protocols. Once engraftment was confirmed, the mice were injected with either BIONICs or control IgG (2.5 mg/kg per mouse) along with healthy donor–derived pan T cells (3 million T cells), as described in supplemental Materials and supplemental Figures.

CD38 expression is induced in response to T helper type 1 cell–derived IFN-γ.¹⁹ To gain insights in CD38 mechanisms of regulation in AML cells, we exposed primary BM AML blasts to IFN-γ. We observed a significant dose-dependent increase in surface CD38 in both CD34^pos and CD34^neg AML blasts (Figure 1A-C). In colony forming unit assays, IFN-γ treatment significantly decreased the clonogenic activities of AML cells (Figure 1D; P = .002), independent of their cytogenetic and molecular features (supplemental Table 1). To understand how IFN-γ induces CD38 expression, we treated total BM MNCs from a patient with relapsed AML with IFN-γ or vehicle and carried out scRNAseq. We observed significant transcriptional changes involving >1000 distinct genes in IFN-γ–treated cells compared with in the vehicle-treated AML cells. Among the most upregulated genes in the IFN-γ–treated cells (>2.5 log₂ fold change; P = 0; false discovery rate [FDR] = 0), we found CD38 and 2 previously reported transcriptional regulators of CD38 expression, interferon regulatory factor 1 (IRF-1) and signal transducer and activator of transcription 1^20-23 (Figure 1E; supplemental Table 2).

Total RNAseq from isolated CD34^pos AML blasts of patients with relapsed AML (n = 3) confirmed that upregulation of CD38 and IRF-1 but not of signal transducer and activator of transcription 1 (>2.5 log₂ fold change FDR of <0.25; P = .0007) in the IFN-γ–treated cells (Figure 1F-G; supplemental Figure 1A; supplemental Table 3). These results were further confirmed by quantitative reverse transcription polymerase chain reaction and flow cytometry assays (Figure 1H; supplemental Figure 1B-D). Gene set enrichment analysis showed that IFN-γ treatment also induced a significant (FDR = 0) upregulation of genes involved in apoptosis, DNA repair, reactive oxygen species regulation, and inflammation (ie, tumor necrosis factor α), among others (supplemental Figure 1E), and significant suppression of genes involved in the pro-oncogenic Wnt β catenin pathways (supplemental Figure 1F) and cell survival (eg, CD109,²⁴ KDM6B,²⁵ and MYCN²⁶; supplemental Table 3).

To investigate the role of IRF-1 in IFN-γ–induced CD38 transcription, we analyzed the CD38 promoter using the JASPAR 2022 database.²⁷ We found 2 partially overlapping and highly conserved IRF-1 consensus sequences in the CD38 promoter starting at −6 and −27 nucleotides from the transcriptional starting site (supplemental Figure 1G). To verify these experimentally, we developed a luciferase-based reporter system to measure CD38 promoter (WT-CD38) activation. We transfected the CD38 luciferase-based reporter construct into CD38-negative HEK-293 cells along with a vector expressing IRF-1. Ectopic expression of IRF-1 significantly increased CD38 promoter luciferase activity (P < .0001) compared with cells transfected with the CD38 reporter vector and a vector-null control (Ctrl; Figure 1J). To assess the specificity of the IRF-1 transcription reporter assay, we also generated a second CD38 promoter reporter system lacking the IRF-1 consensus sequences only (−6 to −27, Mut-CD38; Figure 1I) and transfected with this construct in CD38-negative (HEK-293) and CD38-positive THP-1 cells. The ectopic IRF-1 expression, in contrast, failed to induce any luciferase activity (Figure 1J-K). In addition, we knocked down IRF-1 expression with small interfering RNAs (siRNA) in AML cells treated with IFN-γ and observed a significant downregulation of both CD38 messenger RNA and surface protein compared with cells transfected with nontargeting siRNA (Ctrl-siRNA; Figure 1L; supplemental Figure 1H-I).

CD4^pos T helper type 1 and CD8^pos cytotoxic T cells release IFN-γ once engaged on their target(s).²⁸ Thus, we hypothesized that a CD38-directed T-cell engager would not only kill CD34^posCD38^pos AML blasts but also convert CD34^posCD38^neg LSCs into CD34^posCD38^pos blasts via the secreted IFN-γ, thereby expanding the targetable AML cell population also to the CD38^neg blasts (Figure 2A). To this end, we hypothesized that a compact T-cell engager, mimicking the distance between the T-cell receptor (TCR) and the major histocompatibility complex (MHC) while limiting steric hindrance, would produce an optimal immunological synapse.^29,30 To achieve a compact structure while simultaneously addressing challenges in generating immune engagers (eg, light chain pairing), we created a single-chain construct, inserting a CD38 nanobody (CD38-NB) between the light and heavy chains of an anti-CD3 Fab using 2 short peptide linkers. The net result is a compact, single-chain molecule we have named a CD38-CD3 BIONIC (BN-CD38; Figure 2B). As a point of comparison, superposition of the cryoelectron microscope structure of the TCR (PDB:6JXR) and the MHC structure (PDB:4GRL) indicates the distance between T-cell membrane and antigen presenting cells is ∼140 Å (Figure 2C), which is consistent with previously reported values.³¹ Simple modeling indicates that BN-CD38 closely matches the TCR-MHC distance (Figure 2D).

Surface plasmon resonance and analytical cytometry studies show that BN-CD38 retained binding affinity to both CD38 and CD3 antigens (Figure 2E). Moreover, the CD38-NB alone and BN-CD38 also show comparable binding kinetics to CD38 (supplemental Figure 2A). We also generated a BN-CD38Mut as a negative control. Replacing serine with histidine at residue 57 (Ser57His) within the anti-CD38 nanobody (based on nanobody numbering), located within the CD38-CD38 nanobody interface (supplemental Figure 2B) abrogated the binding to CD38. Binding to T-cell surface CD3 remained unaffected (supplemental Figure 2B-C).

To test the activity of BN-CD38, we cocultured T cells (hereafter called “effector”) with CD38^pos (hereafter called “target”) AML cells (ie, THP1, MV-4;11, HL-60, U937, NOMO1, and SKM1) and MM cells (ie, MM.1S) at an effector-to-target (E:T) ratio of 1:1 in the presence of BN-CD38. We observed significant killing activity (P < .0001), which correlated with the levels of CD38 expression (supplemental Figure 2D-E). No significant killing was observed in CD38^neg cells, that is, the AML SKM1 cells or CD38-knockout HL-60 AML cells, in which the CD38 was silenced via CRISPR (supplemental Figure 2D-E). Killing activity was also not observed in the presence of BN-CD38Mut or nonspecific human IgG, instead of BN-CD38 (supplemental Figure 2F). The 50% inhibitory concentration (IC₅₀) for BN-CD38 was calculated to be within the 0.01 to 0.1 ng/mL range, whereas the IC₅₀ of BN-CD38Mut was at least 100-fold higher (Figure 2F; supplemental Figure 2G). Significant increases in expression of early (CD69) and late (CD25) T-cell activation markers was observed both on CD4^pos and CD8^pos T cells cocultured with CD38^pos THP1 cells (P < .0001) in the presence of BN-CD38, and not in those cocultures with control treatments (human IgG or BN-CD38Mut; Figure 2G-H; supplemental Figure 2H-I). Furthermore, a significant increase in secretion levels of IFN-γ, tumor necrosis factor α, and interleukin-2, among others, by T cells was observed in cocultures of T cells and CD38^pos THP1 cells treated with BN-CD38 (supplemental Figure 3A-B). No cytokine secretion was detected in the absence of CD38^pos THP1 cells (supplemental Figure 3C) or in T cells cocultured with CD38^neg target cells (ie, CD38-knockout HL-60 cells; supplemental Figure 3D). Of particular interest, levels of BN-CD38 that induced engagement and activation for CD4^pos and CD8^pos T cells were comparable (supplemental Figure 3E-F).³² 

To assess whether BN-CD38 could effectively engage blasts and activate autologous T cells, we cocultured total BM or PB blasts and autologous T cells from the same patients with AML in the presence of 1 ng/mL of BN-CD38 or controls (ie, IgG, CD38-NB, or BN-CD38Mut). Compared with controls, exposure to BN-CD38 resulted in a significant decrease of both CD38^pos and CD38^neg blasts and their clonogenic activity (P < .01; Figure 2I-L; supplemental Table 1). Consistently, we observed upregulation of early (CD69) and late (CD25) T-cell activation markers (P < .0001; supplemental Figure 2J-K) and, after 5 days from the initial treatment, a significant expansion of patient-derived CD4^pos and CD8^pos T cells (P < .01; supplemental Figure 2L). To further investigate whether BN-CD38 induces specific killing of the CD38^neg LSC–enriched population as well, we purified these cells from the total AML MNCs. Specifically, as shown in the graphical experimental design in Figure 2M, we first treated the CD38^pos enriched blast population with their autologous effector cells and either BN-CD38, BN-CD38Mut, or control IgG for 24 hours. We then collected the CD38^pos supernatant (SN) from each experimental group and used it to treat the CD38^neg population cocultured with autologous T cells (E:T, 1:1), for 24 hours. We observed significant apoptotic induction of the CD38^neg blasts treated with BN-CD38–enriched SN, compared with levels in blasts treated with BN-CD38Mut SN (51.4% ± 7.4% vs 2.1% ± 2.6%; P = .0001; n = 4 normalized for the respective Ctrl IgG SN), as assessed by annexin-V/propidium iodide staining (Figure 2N-O).

To investigate how exposure to BN-CD38 modifies distinct cellular components of the BM from patients with AML, we assembled a 36-metal mass cytometry (CyTOF) panel enabling us to discriminate AML and immune cell subpopulations (supplemental Tables 4 and 5).

Using supervised gating on vi_SNE (FlowSOM), we showed that BN-CD38 but not IgG, BN-CD38Mut, or CD38-NB significantly reduced the viability of both CD34^posCD38^pos and CD34^posCD38^neg AML blasts (Figure 3A-C; supplemental Figure 4A). BN-CD38 treatment was also associated with a significant expansion of CD8^pos effector memory cells (P < .01), terminally differentiated effector memory CD45RA^pos cells (P < .05; Figure 3D), central memory regulatory T cells (P < .0001), and natural killer (NK) T cells (P < .01; Figure 3E-F). Consistent with these results, IFN-γ induction was also observed in the SNs of total AML cells treated with BN-CD38 (P = .04; supplemental Figure 4B). B cells were not significantly affected by exposure to BN-CD38, whereas NK cell expansion was observed in 5 of 6 primary samples with detectable NK cells (supplemental Figure 4C). Data-driven self-stratifying clustering CITRUS (cluster identification, characterization, and regression) analysis also shows that BN-CD38 significantly induces leukemic cytoreduction, while expanding T cells and phagocytic CD14^posCD16^neg classical monocytes (Figure 3G-I).³³ Furthermore, BN-CD38 induced a significant increase in CD38 messenger RNA and protein expression (P < .0001) in primary AML samples, which was effectively reverted when 2 μg/mL anti-human IFN-γ neutralizing antibody was added to the culture (Figure 3J-K). We then treated PB MNCs from patients with AML for 5 days with 1 ng/mL BN-CD38 and 2 μg/mL anti-human IFN-γ neutralizing antibody; we confirmed that blocking IFN-γ significantly reversed the BN-CD38 killing activity for CD34^posCD38^neg cells (P < .05; Figure 3L-M), without affecting T-cell expansion. Because IFN-γ can impair T-cell killing activity through the induction of programmed cell death protein 1,³⁴ as also observed in the T cells of AML MNCs treated ex vivo with BN-CD38 compared with the relative controls (P = .02; n = 7; supplemental Figure 4D-E), we investigated whether those T cells maintained their killing capabilities once re-exposed to the tumor cells. Despite programmed cell death protein 1 upregulation, BN-CD38–treated T cells eliminated autologous AML cells and THP1 cells upon rechallenge, to a level that was comparable with that of the first exposure (supplemental Figure 4F-G).

To study the antileukemic activity of BN-CD38 in vivo, we IV injected 1 × 10⁶ luciferase-positive CD38^pos THP1 cells into nonobese diabetic/severe combined immunodeficiency (NOD-SCID) IL2Rgnull (NSG) mice (Figure 4A). On day 19 after injection, mice were randomized into 2 groups and treated with either 2.5 mg/kg BN-CD38 (n = 8) or human IgG (n = 10), and 3 × 10⁶ human T cells were given IV weekly. The mice receiving BN-CD38 plus T cells were treated for a total of 4 weeks, whereas control mice were treated for 2 weeks, because the control animals had all died by day 32. A significant decrease in leukemic burden was observed in the BN-CD38–treated mice compared with human IgG–treated controls as early as 1 week after treatment initiation (day 25; P = .003; Figure 4B-C). Although lymphoproliferation and graft-versus-host disease (GVHD) were expected side-effects associated with allogeneic T-cell infusion, 38% of BN-CD38–treated mice remained AML- and GVHD-free and remained this way for 120 days after the last treatment until they were humanely euthanized and no signs of disease were found, whereas the median survival for control mice was 29 days (Figure 4D). The antileukemic activity of BN-CD38 was also tested in mice xenografted with U937 cells, which have lower CD38 expression (CD38^dim) and are less susceptible to BN-CD38 T-cell killing compared with THP-1 cells (supplemental Figure 5A-B). Mice treated with BN-CD38 and human T cells and not those treated with BN-CD38Mut or Ctrl-IgG under the same experimental setting, showed near-complete disease eradication (supplemental Figure 5C-F).

Next, we prepared 3 PDXs using AML MNCs from 2 patients with refractory disease with complex karyotypes (patient identifiers: AML-4 and AML-15) and 1 patient with relapsed disease with a normal karyotype in which the CD34^posCD38^neg fraction in the total blast was >20% (AML-22; supplemental Table 1). One million AML MNCs from each patient were injected into 3 independent cohorts of irradiated NSG mice. To assess LSC eradication upon BN-CD38 treatment, we first performed a multiple transplantation study using the PDX-1 model (AML-15). Specifically, as shown in supplemental Figure 5G, on day 20, upon confirmation of successful engraftment (>1% circulating human AML blasts in the PB), PDX-1 engrafted mice were randomized into 3 groups to receive BN-CD38, BN-CD38Mut, or IgG (2.5 mg/kg per week) respectively, together with healthy donor human T cells (3 × 10⁶ cells per week). After 3 treatments, mice were humanely euthanized, and leukemia burden was assessed. We observed significant decreases in AML cell engraftment in BN-CD38–treated animals (supplemental Figure 5H). We then transplanted BM MNCs of each donor into secondary NSG recipients. After 10 days, the secondary recipient mice were given the same treatment, and at the end of treatment, we observed that BN-CD38–treated mice had near disease eradication, both in the spleen and the BM, compared with mice treated with BN-CD38Mut or IgG (P < .0001; supplemental Figure 5I-K). To assess the long-term effect of BN-CD38–induced T-cell killing on AML cells, we performed a tertiary transplantation and followed-up for survival. Mice transplanted with BM MNCs from BN-CD38Mut– or IgG-treated mice had significantly lower survival (median survival, 28 days) than those transplanted with the BM MNCs from BN-CD38–treated mice (P = .002), which were followed-up to 256 days (Figure 4E); no signs of disease were observed before euthanasia. To assess the BN-CD38–directed antileukemic effect, we used a different PDX model (PDX-2, AML-22) with limited GVHD complications and conducted multiple treatments without multiple transplantation. Specifically, mice after engraftment were treated for 4 weeks with BN-CD38 or relative controls in combination with 3 × 10⁶ T cells, and when T-cell engraftment was assessed, the same mice were treated only with BN-CD38 or relative controls once a week for 2 subsequent weeks (Figure 4F). Almost 70% of the BN-CD38–treated mice remained disease free, compared with the relative control groups, which showed a median survival of 12 weeks (P = .0007), even when the BN-CD38–treated mice at week 20 were rechallenged with autologous AML cells (Figure 4F). Although the mice did not show signs of distress by week 29 (201 days), they were euthanized, and flow cytometry analysis showed absence of leukemia engraftment, and human T cells were still detected (supplemental Figure 5L). Because we observed ex vivo that BN-CD38 can redirect patient-derived AML T cells against autologous AML blasts, we investigated whether similar effects occurred in vivo. For this purpose, we used the complex karyotype TP53-mutated PDX-3 (AML-4; Figure 4G). Using a lower number of autologous T cells (1 × 10⁶ cells) once a week for 3 subsequent weeks because of sample limitation, we observed T-cell expansion and significant reduction in BM and spleen tumor engraftment only in the mice treated with BN-CD38 (BN-CD38 vs BN-CD38Mut: P = .002, BM; P < .0001, spleen; Figure 4H-I), further supporting the therapeutic potential of BN-CD38 to induce T-cell activation in autologous conditions.

A concern with therapeutically targeting CD38 is CD38 antigen expression on the surface of normal HSC precursor cells. Thus, depletion of normal CD38^pos HSC precursor cells could result in long-term hematological toxicity. To this end, our data indicate that in colony forming unit, flow cytometry, and mass cytometry assays, BN-CD38 and/or IFN-γ did not perturb the clonogenic potential or the total number of the uncommitted CD34^posCD38^low/neg HSCs obtained from healthy donors (Figure 5A-D; supplemental Figure 6A-D). Moreover, BN-CD38 and/or IFN-γ induced variable, not statistically significant reduction of the committed CD38^pos normal hematopoietic progenitors (Figure 5E; supplemental Figure 6B-D). A trend toward upregulation of CD4^posCD69^pos and CD8^posCD69^pos T cells was observed (Figure 5F) but at a level significantly lower than that observed with AML blasts (supplemental Figure 6E) or CD38^pos cell lines (supplemental Figure 6F). Of note, we did not observe a fratricidal effect of BN-CD38 on mature BM or PB noncancer cells expressing CD38 (T, B, and NK cells, and monocytes) compared with each internal control (Figure 5G-I; supplemental Figure 6G-H). In contrast to AML blasts, no significant CD38 upregulation was observed in normal BM MNCs exposed to BN-CD38 for 48 hours (Figure 5J-K; supplemental Figure 6I). To further address this subject, we performed a killing assay in which we cocultured the total CD34^pos normal population, which contains both HSCs (∼7%) and CD38⁺ hematopoietic progenitors (∼92%; supplemental Figure 6J-K), together with CD38^posCD34^neg AML cells (1:10 ratio) and T cells (E:T ratio 1:1) and assessed the BN-CD38 T-cell killing activity of both total healthy CD34^pos and AML cells at different concentrations. We did not observe dose-dependent killing of normal HSCs as instead found in the cocultured AML cells (P = .01; Figure 5L). To further investigate the impact of BN-CD38 on normal hematopoietic cells, we treated human CD34^pos HSC–engrafted NSG mice (hu-CD34), which produce multilineage human immune cells, with BN-CD38 or controls, weekly for 3 consecutive weeks. Under these conditions, we did not observe differences in BM engraftment or in lineage development between BN-CD38–treated or control-treated mice (supplemental Figure 7A-C). We also did not observe, among the different treatment groups, differences in human HSCs engraftment or hematopoietic lineage development when we used a second CD34^pos humanized animal model in which BN-CD38 was administrated together with 3 × 10⁶ autologous T cells (Figure 5M-O; supplemental Figure 7D-F) obtained from the same donor from which the CD34^pos cells used for engraftment were isolated.

Unlike MM or T-cell acute lymphoblastic leukemia, the clinical relevance of CD38 as a target in AML remains to be clearly demonstrated. Of note, CD38 expression in AML cell lines and patients with primary AML can be induced by all-trans retinoic acid.^35,36 Consistent with previous studies,^17,18,32 we show here that IFN-γ induces CD38 upregulation in AML blasts,³⁷ through the activation of IRF-1, a transcription factor that plays a pivotal role in IFN-γ intracellular signaling, as demonstrated by promoter analysis and functional experiments.^23,38 IRF-1 is a tumor suppressor gene^39-41 located in the long arm of chromosome 5. Interestingly, we observed IRF-1 upregulation in all AML cell lines and primary samples, including those carrying monoallelic 5q deletion, a chromosomal aberration found in ∼15% of patients with AML and myelodysplastic syndromes.⁴² 

Consistent with recent results demonstrating that mouse immune cells produce high levels of IFN-γ that block AML growth,⁴³ we show the anticlonogenic activity of IFN-γ on AML LSCs but not on normal HSCs. Gene expression profiling highlighted that primary AML blasts exposed to IFN-γ upregulate genes involved in apoptosis, DNA repair, and reactive oxygen species, and downregulate genes involved in Wnt signaling that supports LSCs clonogenicity and AML survival.^44-46 

Considering these results, we reasoned that we could target not only CD38^pos but also CD38^neg AML blasts with immunotherapeutics that induce IFN-γ degranulation from T cells, and thus we designed and produced the BN-CD38 chain T-cell engager. We demonstrated that not only did the BN-CD38 directly kill CD38^pos AML blasts but also activated engaged T cells to produce IFN-γ. We further showed that this approach successfully converted CD38^neg into CD38^pos blasts, thereby expanding the BN-CD38–targetable AML cell population.

Le Dieu et al have characterized an aberrant T-cell activation signature in patients with AML⁴⁷ and demonstrated that although T cells isolated from patients with AML are phenotypically effector cytotoxic T lymphocytes, they exhibit impairment in effectively forming immune synapses.^47,48 We demonstrate in vitro, ex vivo, and in vivo that BN-CD38 produced an effective synapse between T cells and CD38^pos AML cells. Importantly, the BN-CD38–engaged T cells efficiently degranulated and released measurable levels of IFN-γ, which proceeded to induce high CD38 expression on both CD38^pos and CD38^neg AML blasts. Using CyTOF immunophenotyping assays, we demonstrated that BN-CD38 significantly reduced AML blasts concurrently with expansion of autologous CD4^pos and CD8^pos T cells and lack of a fratricide effect toward other CD38^pos mature cells (ie, NK cells, NK T cells, or B cells). Interestingly, in the BM samples obtained from patients with AML, BN-CD38 not only spares CD38^pos noncancer immune subsets but expanded NK T cells, NK cells, and CD14^posCD16^neg classical monocytes, which have high capacity for phagocytosis, migration, and innate immune responses.³³ Moreover, in healthy donor BM samples, BN-CD38 did not compromise engraftment and normal hematopoiesis in 2 independent models of CD34^pos humanized animals, supporting lack of hematological toxicity that could prevent the broad use of this molecule. Although the mechanisms of CD38 regulation in normal HSCs upon IFN-γ treatment will need further investigation, we suggest that this effect may be because of a different epigenetic and transcriptional signature of HSCs compared with LSCs, as previously reported.^49,50 

In AML, CD33, CD123, CLEC12A, and FLT3 represent suitable targets for T-cell engagers⁵¹ but, to date, the clinical benefits have been limited. The limitation to successfully use these engagers ranges from the immunophenotypic heterogeneity of AML cells, to the overlap of TAAs with those on normal HSCs, to the limited immune activity of T cells present in patients with AML.⁵² BN-CD38 exhibits a potent cytotoxicity profile with an IC₅₀ ranging from 0.01 to 0.1 ng/mL after overnight incubation at an E:T ratio of 1:1, compared with other T-cell engagers, which have an IC₅₀ range of 0.01 to 68 ng/mL at an E:T of 5:1 or 10:1 after 48 hours of AML cell lines and effectors coculture.^53-56 We also show that BN-CD38 exhibits potent cytotoxicity in in vivo models without the need of using preactivated T cells as reported with many other engagers including those targeting CD33, FLT3, and BCMA.^53,56-58 

CD38-directed chimeric antigen receptor T cells have also been developed and tested as a potential AML therapy.^59-62 In preclinical experiments, CD38-directed chimeric antigen receptor T cells mediated killing against CD38^pos blasts, but the ability of these cells to eliminate LSCs was not fully characterized.⁵⁹ In contrast, we show that BN-CD38 can target LSCs, through mechanisms that are IFN-γ dependent, further supporting that CD38-BN can induce the release of effective levels of IFN-γ even in the context of the AML immune suppressive environment concomitantly blocking LSCs clonogenicity and enhancing their direct targeting, an observation that, to the best of our knowledge, has not been reported before.

Although BN-CD38 increased the survival of mice xenografted with different AML cell lines and primary leukemia cells, it is reasonable to assume that patients with AML with a higher frequency of CD38⁻ LSCs may require a longer treatment schedule to achieve remission, a question that we are planning to address in the clinical setting.

Taken together, here we report a novel strategy to specifically target LSCs, and in doing so, we demonstrate that the biologic activities of BN-CD38 overcome the AML tumor suppressive environment and potentiate the efficacy of CD38-directed T-cell therapy, providing the scientific rationale for rapidly translating BN-CD38 into the clinic as an effective therapeutic approach for AML.

The authors are grateful to Tinisha McDonald, Kelly Synold, and Elena Pulkinen for technical support in this work. The authors also thank Hyeran Choi for assisting with fluorescence-activated cell sorting analysis of BIONICs constructs. The authors thank Yuriy Shostak and Christoph Pittius for the scientific and administrative support in this work.

Research reported in this publication was, in part, supported by the generous contributions of the Nason-Hollingsworth Project for Multiple Myeloma (A.K., F.P.). This research was also supported by the City of Hope Preclinical Drug Development Venture Program Award (F.P., J.C.W., G.M.), and partially supported by the National Institutes of Health (NIH), National Cancer Institute (NCI) grants R01-CA238429 (F.P.) and R50-CA252135 (E.C.). Research reported in this publication included work performed at the City of Hope X-ray, Liquid Tissue Bank, Analytical Cytometry, and Integrative Genomics, and Small Imaging cores supported by the NIH, NCI under award number P30CA033572.

The content is solely the responsibility of the authors and does not represent the official views of the National Cancer Institute or the National Institutes of Health.

Contribution: M. Murtadha, M.P., L.L., B.Z., G.M., J.C.W., and F.P. designed the research; M. Murtadha, M.P., Y.Z., E.C., O.N., A.A.D., M.S., H.V., T.T., M. Moloudizargari, A.G., K.L., C.O., L.N., and A.D. performed research; L.G. provided support with selection of the patient samples for the PDX models; M. Murtadha, M.P., and J.F.S. analyzed data; A.P. and A.C. provided biostatistical support; A.P. provided bioinformatics support; M. Murtadha, M.P., J.F.S., G.M., J.C.W., and F.P. wrote the manuscript; A.K. scientifically edited the manuscript; and all authors reviewed and approved of the manuscript.

Conflict-of-interest disclosure: M. Murtadha, M.P., G.M., J.C.W., and F.P. have pending patents covering certain aspects of this technology in anticipation of commercialization. The remaining authors declare no competing financial interests.

Correspondence: Flavia Pichiorri, Department of Hematology and Hematopoietic Cell Transplantation, Judy and Bernard Briskin Center for Multiple Myeloma Research, Department of Hematologic Malignancies Translational Science, Beckman Research Institute, City of Hope, 1500 Duarte Rd, Duarte, CA 91010; email: fpichiorri@coh.org; John C. Williams, Department of Cancer Biology and Molecular Medicine, Beckman Research Institute, City of Hope, 1500 Duarte Rd, Duarte, CA 91010; email: jcwilliams@coh.org; and Guido Marcucci, Department of Hematologic Malignancies Translational Science, Beckman Research Institute, 1500 Duarte Rd, Duarte, CA 91010; email: gmarcucci@coh.org.