Evaluation of multi-antigen targeting ADCC strategies in pediatric BCP-ALL

Abstract

Background

Blinatumomab, inotuzumab or autologous anti-CD19 chimeric antigen receptor (CAR)-T cells have revolutionized the treatment of relapsed or refractory B-cell precursor acute lymphoblastic leukemia (BCP-ALL). However, tumor escape through antigenic modulation accounts for almost 40% of subsequent relapses. Multi-antigen targeting strategies should be developed, and it is urgent to identify new targets.

Methods

We investigated the extensive immunophenotyping of 13 BCP-ALL from pediatric patients by using the BioLegend Human Cell Surface Marker Screening Kit. Then, to assess whether targeting each antigen with monoclonal antibodies could lead to leukemic cell lysis, long-term antibody-dependent cellular cytotoxicity (ADCC) assays were performed using murine monoclonal antibodies and human T cells armed with murine CD16.

Results

13 highly expressed antigens were selected. With the antibodies tested here, the most significant lysis was observed by targeting CD24 and CD156c. The double targeting of CD24-CD123 appeared to be even more effective. Triple targeting was associated with a reduction in ADCC activity.

Conclusion

CD24 therefore emerged as an effective target in BCP-ALL, and the combination of CD24 and CD123 as a potential effective double-targeting strategy. The combination of different recognition modalities (eg, a CAR and CD16) should be tested to determine whether it provides synergistic cytotoxic activity in triple targeting.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Anti-CD19 chimeric antigen receptor (CAR)-T cells revolutionized the therapeutic approach in relapsed and refractory B-cell precursor acute lymphoblastic leukemia (BCP-ALL) may be cured by anti-CD19 CAR-T. Unfortunately, 50% of patients relapse after this targeted therapy. Tumor escape by antigen modulation accounts for 40% of relapses. Multi-antigen targeted therapy may avoid such escape mechanism. Identification of efficient combination of targets on BCP-ALL is needed. The CD16+ T cells appear as a simple and scalable multi-antigen targeting strategy when combined with different antibodies.

WHAT THIS STUDY ADDS

• Our study provides an in-depth exploration of surface antigen expression in pediatric BCP-ALL. An efficient antibody-dependent cellular cytotoxicity (ADCC)-mediated lysis of BCP-ALL is obtained by targeting CD24 with antibodies combined with CD16-T cells. The combination CD24-CD123 appeared as a potential effective double-targeting strategy in ADCC.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The identification of new potential antigenic targets in BCP-ALL may offer new therapeutic strategies in refractory diseases, notably in case of relapses after targeted therapies. The use of CD16-T cells also appeared as a promising strategy for scalable multi-antigen targeting.

Introduction

The current first-line approach for B-cell precursor acute lymphoblastic leukemia is still based on chemotherapy, and leads to an overall survival greater than 90%.¹ However, 10–15% of patients relapse after a first line of treatment, and prognosis worsens with the number of relapses.¹ In the last decade, the anti-CD19 chimeric antigen receptor (CAR) T-cell therapy (tisagenlecleucel) has revolutionized therapeutic approaches to relapse/refractory B-cell precursor acute lymphoblastic leukemia (BCP-ALL), as the early response rate rose to 80% in the pivotal pediatric study.² Nevertheless, after 10 years of “real life” experience, relapses occurred, and the 3-year event-free survival of patients included in the pivotal study fell to 44.4%.³

The main driver of post CAR T-cell relapse of BCP-ALL is the loss of CD19 expression (41% of cases), secondary to alternative splicing, mutation, or deletion. A high disease burden at infusion, a prior poor response to blinatumomab and a 4-1BB co-stimulatory domain in the CAR construct are associated with a higher risk of CD19-negative BCP-ALL relapse, which is associated with a very poor prognosis.^{4 5}

Targeting several antigens simultaneously or sequentially may provide greater tumor coverage and potentially circumvent antigen escape. Dual or multi-antigen targeting strategies were therefore explored in preclinical studies, by using either tandem or dual CAR T-cell constructions.⁶ In a clinical setting, some studies reported encouraging early results for dual targeting (CD19/CD22) in BCP-ALL, unfortunately followed by a significant rate of relapses.^{7 8}

An alternative strategy is based on universal immune receptors (UIR). These UIRs provide an adaptable specificity, through the use of an extracellular adaptor moiety (ie, antibodies, single-chain variable fragments, small molecules) targeting an antigen, and bridging a receptor on T cells.79,11 We reported one of the first UIR T cells, based on the principle of antibody-dependent cellular cytotoxicity (ADCC) using genetically modified T cells expressing CD16. First developed to increase the ADCC activity in patients receiving monoclonal antibodies, the CD16-T lymphocytes could allow flexible multi-antigen targeting strategies when combined with several monoclonal antibodies.812,15 Adoptive transfer of CD16-T lymphocytes coupled with the injection of two or three different monoclonal antibodies or polyclonal antibodies could be evaluated as a simple and scalable multi-antigen targeting strategy. To this end, it is necessary to directly identify antigens whose targeting by an antibody effectively induces ADCC.

In this study, we first performed extensive immunophenotyping of BCP-ALL in pediatric patients and then selected 13 highly expressed antigens. We determined whether the selected antigens enabled lysis of leukemic cells by ADCC. ADCC assays were performed, using murine antibodies widely available and the cellular tool human T cells expressing the murine CD16 (mCD16 T cells) previously described.¹⁶

Materials and methods

Cells

Leukemic samples

Residual bone marrow aspiration samples collected from leukemic pediatric patients in the Nantes hospital pediatric hematology unit were used for experiments. All samples contained more than 60% of blasts. The National Cancer Institute (NCI) criteria and the genetic subtypes of the BCP-ALL used in the experiment have been detailed in the online supplemental table S1. Mononuclear cells were isolated by using density gradient centrifugation on a FICOLL-Paque solution (Eurobio, Les Ullis, France). Cells were then used fresh or after thawing. Patients and their guardians/parents provided an informed signed consent form for the biocollection. Investigations were therefore conducted in accordance with the principles of the Declaration of Helsinki.

B-cell line

An Epstein-Barr virus lymphoblastoid cell line (EBV-LCL) derived from donor peripheral blood mononuclear cells infected with B95-8 marmoset-derived EBV in the presence of 1 µg/mL of cyclosporin-A (sandimmun) was used as a control for phenotyping and ADCC assays. A B-cell lineage Namalwa (ATCC CRL-1432, B lymphocyte isolated from a patient with Burkitt’s lymphoma) was also used as control for ADCC assays. EBV-LCL and Namalwa cell line were cultured in RPMI 1640 Medium (Gibco) medium supplemented with 10% fetal bovine serum (Biosera) and 1 GlutaMAX-I (1X) (Gibco).

ADCC effector T cells: anti-CMV-mCD16 T cells

A cytomegalovirus (CMV)-specific CD8⁺ T-cell population, genetically modified to express the murine Fc$\gamma$RIII (CD16) linked to the human Fc$\epsilon$RI$\gamma$ chain was obtained as previously described.^{8 17} Briefly, anti-HLA-A2 CMVpp65-T cells from an HLA-A2 CMV-seropositive healthy donor (HD) were selected and amplified using recombinant pMHC HLA-A2 CMVpp65^N9V tetramer. The selected anti-CMV-CD8⁺ T-cell population was stimulated (day 0) and transduced (day 2) with retroviral vector expressing the murine CD16-human $\gamma$ receptor. On day 14, mCD16-T cells were immunoselected, amplified, then frozen on day 16 after the stimulation. The anti-CMV-mCD16-T cells were thawed and kept overnight or several days at 37°C in X-VIVO 15 medium (Lonza) supplemented with 4% human serum and interleukin-2 (300 UI/mL) before ADCC assays. The ADCC activity of these anti-CMV-mCD16-T cells (mCD16-T cells) was controlled against EBV-LCL preincubated with anti-CD20 murine antibodies (mAb) (murine IgG1, AT80 clone).

Antibodies

Monoclonal mAb specific for human selected antigens CD10, CD19, CD22, CD24, CD47, CD74, CD123, CD135, CD137, CD156c, CD205, CD229 and CD268 were provided by BioLegend (BioLegend Europe B.V., Amsterdam, the Netherlands). The specific anti-human CD20 and anti-CD71 mAb were purchased from Santa Cruz Biotechnology (Heidelberg, Germany) and ABD Serotec Bio-Rad (Marne La Coquette, France), respectively. For ADCC assays, mAbs were used at a final concentration of 1 µg/mL. The complete list of mAbs is provided in online supplemental table S2.

Immunophenotyping

Extensive immunophenotyping of leukemic cells was performed using the Human Cell Surface Marker Screening Kit from BioLegend (BioLegend Europe B.V., Amsterdam, the Netherlands; LEGENDScreen), containing 371 phycoerythrin (PE)-conjugated, lyophilized and pre-titrated antibodies to cell surface markers as well as 10 Ig isotype controls in 96-well plates. Leukemic samples were first incubated with an anti-CD19-APC coupled antibody (BioLegend, clone HIB19) in order to select the leukemic CD19⁺ blasts, and then1,318 19 × 10⁵ cells per well were plated in the 96-well plates. Staining was performed at 4°C in the dark for 20–30 min, followed by washing with Cell Staining Buffer before fixation and analysis. For data analysis, at least 10,000 CD19⁺ cells were acquired per sample, using the CANTO II cytometer (Cytocell—Flow cytometry and FACS (Fluorescence Activated Cell Sorting) core facility SFR Bonamy, BioCore, Inserm UMS 016, CNRS UAR 3556, Nantes, France). Data were processed using FlowJo V.10.8.1 software (BD LifeSciences). For each PE-mAb, relative fluorescence intensity (RFI in log) was calculated as the median fluorescence intensity of the targeted-antigen coupled antibody/median fluorescence intensity of the unspecific control isotype.

⁵¹Cr release cytotoxicity assays

Target cells were previously labeled with 75 µCi ⁵¹Cr for 1 hour at 37°C then washed four times with Roswell Park Memorial Institute medium (RPMI), (Fetal Calf Serum) FCS 10%. Target cells previously incubated with mAb and effector mCD16-T cells were then plated at the indicated effector-to-target ratio (E:T ratio) in flat-bottom 96-well plates. After a 4 hours incubation at 37°C, 25 µL of supernatant was removed from each well, mixed with 100 µL scintillation fluid (Ultima Gold XR) (PerkinElmer Health Sciences, Groninger, the Netherlands), and released ⁵¹Cr activity was counted in a scintillation counter (MicroBeta JET) (PerkinElmer Health Sciences, Groninger, the Netherlands). Each test was performed in triplicate. Results are expressed as the percentage of lysis, which is calculated according to the following equation: (experimental release − spontaneous release) / (maximal release − spontaneous release) × 100.

Long-term ADCC killing assays

The analysis of leukemic cell lysis was performed by using long-term ADCC assays over 24 hours, as this better reflects the response observed in vivo.²⁰ Medullary BCP-ALL samples were thawed and kept overnight at +4°C. Target cells (80,000 cells/well) were sensitized for 15 min with murine mAb at a final concentration of 1 µg/mL, and plated with ADCC effector T cells, in X-VIVO 15 media (Lonza, Basel, Switzerland) at an E:T ratio of 3:1 in a flat-bottom, 96-well plate, then incubated at 37°C. At different time points, that is, 0 hours (H0) and 24 hours (H24) after co-culture, cell suspension was collected. Cells were then washed once with phosphate-buffered saline (PBS)-EDTA (0.02%) and then with PBS, before being labeled with a Fixable Viability Stain 780 (BD Biosciences, Le Pont de Claix, France) over 10 min at room temperature (RT). Cells were then washed in PBS and labeled with a PE-coupled anti-CD22 antibody (Dako product provided by Technologies Agilent France, Les Ulis, France) and an FITC (fluorescein 5-isothiocyanate)-coupled anti-CD3 antibody (Beckman Coulter France SAS, Roissy, France) for 15 min at RT, in order to gate residual viable leukemic cells and effector mCD16-T cells, respectively. After two more washes with PBS 0.1% human albumin, cells were fixed and were acquired on BD FACSCanto II (Cytocell—Flow Cytometry and FACS core facility, SFR Bonamy, BioCore, Inserm UMS 016, CNRS UAR 3556, Nantes, France). For data analysis, a first logical gate was based on FCS (forward scatter)/SSC (side scatter), then singlets were isolated. Second, viable (VS780^low) cells were selected, and the two last gates on CD22⁺ and CD3⁺ cells were performed to identify leukemic cells and T cells, respectively. Representative plots with gating strategy and controls are shown in online supplemental figure S1. A total of 10,000 events in the viable CD3⁺ gate were acquired. For each condition, the total number of viable leukemic cells (VS780^low/CD22⁺) at H0 was reported as 100%. At H24, the percentage of residual viable leukemic cells was calculated as follows: = (count of viable CD22⁺ cells at H24 / count of viable CD22⁺ cells at H0) × 100.

Statistical analysis

The correlation between the level of expression of each antigen tested (RFI) and specific ADCC-mediated lysis was determined by using a simple linear regression test (GraphPad Prism V.9.5.1, GraphPad software, Boston, USA).

Results

Extensive immunophenotyping and antigenic target selection on BCP-ALL

Extensive immunophenotyping was performed for 13 BCP-ALL (11 de novo and 2 relapses) of several standard, intermediate or high risk subtypes (online supplemental table S1). The 361 antigens tested were separated into 15 categories (see online supplemental table S3). Unexpressed antigens (n=129) were first excluded. Second, the selection of antigens of interest was carried out among the antigens belonging to B-cell markers (n=7/36) and/or those for which a targeted therapy has been developed (n=10/47). The 13 antigens that meet the three main following criteria: B-cell specificity, and/or high expression levels and/or therapeutic targets that are already used in patients were selected. CD20, which is expressed by several BCP-ALL, and highly by the EBV-LCL, and CD137, not expressed on BCP-ALL cells, were used as positive and negative controls, respectively. The level of expression of these 15 antigens on the 13 BCP-ALL samples is presented in figure 1. The mean RFI of the antigens selected ranged from 2.6 for the CD268 (BAFF-R) to 478 for the CD10. As reported in the literature, our data confirm higher expression of CD19 than CD22 on BCP-ALL.²¹ Finally, the 13 antigens highly expressed by the majority of samples were selected (mean of RFI ranging from 2.6 to 478). Note that antigens shared by B and T cells (CD52 and CD27) were not selected. The selected antigens had very different molecular weights and functions (figure 2A). The levels of expression of each antigen in healthy tissues were collected in a database (https://www.proteinatlas.org/; https://www.uniprot.org/)²² and are represented in figure 2B.

Figure 1. Expression levels of the 15 selected antigens on primary BCP-ALL sample. The extensive immunophenotyping was performed with PE-coupled antibodies provided in the Human Cell Surface Marker Screening Kit (BioLegend Europe B.V., Amsterdam, the Netherlands). RFI in log was calculated as the MFI with the PE-specific antigen antibody divided by the MFI of the isotype control mAb. Results are presented as a median and 95% CI. Each point symbolizes one ALL sample (n=13 donors, each sample has been tested one time). ALL, acute lymphoblastic leukemia; BCP, B-cell precursor; EBV-LCL, Epstein-Barr Virus lymphoblastoid cell line; mAb, murine antibodies; MFI, median fluorescence intensity; PE, phycoerythrin; RFI, relative fluorescence intensity.

Figure 2. Details of selected antigens. (A) Table with the name and main features of 15 selected antigens. (B) Heat map of levels of expression of each antigen in healthy tissues.^{29 30}.

Monotargeting: variability of ADCC efficiency depending on the targeted antigen

Survival of primary B-ALL cells ex vivo is often difficult and represents a limitation for the development of novel and targeted therapy. Our study was performed on primary BCP-ALL samples from patients, which have been frozen and thawed before use. The viability of BCP-ALL samples alone in culture was therefore monitored in all assays and assessed by flow cytometry at the following time points: 0 hours, 4 hours (n=6/8) and 24 hours of culture. These analyses were performed by using a fixable viability stain (VS780) and a PE-coupled anti-CD22 antibody. As presented in figure 3A, the viability of ALL samples was between 70% (n=1) and 100% at H0, and remained above 50% at 24 hours, except for ALL 7 and 8 (30 and 10%, respectively). However, for these two ALL-B samples, the decrease of the percentage of CD22+/VS780^low cells may be linked to the downmodulation of CD22 expression on the majority of cells (online supplemental figure S2).

Figure 3. Monotargeting long-term ADCC assays. Leukemic cells and mCD16-T cells were co-incubated at the E:T ratio 3:1 in X-VIVO 15 medium in the absence of or with 1 µg/mL of murine purified antibodies and incubated at 37°C for 24 hours. At 0 hours (H0), 4 hours (H4) and 24 hours (H24) of co-culture, residual viable cells were analyzed by MFC. (A) Percentage of residual viable leukemic cells (VS780^low/CD22+) at H0, H4 and H24 for eight BCP-ALL alone (eight donors, one test for ALL-1, 2, 7, and 8; two technical repeats for ALL-5 and 6, 3 replicates for ALL-3 and four replicates for ALL-4). (B) Percentage of residual viable leukemic cells of four BCP-ALL cells after 24 hours, alone, or in co-culture with mCD16 T cells in the absence or presence of the different murine mAbs (four donors, one replicate for each sample). (C) Correlation between the level of CD24 expression by four BCP-ALL cells and residual viable CD22+ cells after 24 hours of co-culture in long-term ADCC assays with an anti-CD24 mAb (four donors and one replicate for each sample). A simple linear regression test was used (GraphPad Prism V.9.5.1, GraphPad software). ADCC, antibody-dependent cellular cytotoxicity; ALL, acute lymphoblastic leukemia; BCP, B-cell precursor; E:T, effector-to-target ratio; mAb, murine antibodies; MFC: Multiparametric Flow Cytometry; RFI, relative fluorescence intensity.

It is important to note that all ADCC assays were performed in an allogeneic situation as the mCD16-T-cell population was derived from one HD and BCP-ALL samples were derived from patients. In order to reduce allogeneic recognition, in the long-term ADCC assays, we used mCD16-T cells with a restricted T-cell receptor (TCR) specificity (HLA-A2-CMVpp65^N9V T cells). However, even with these mCD16-T cells, five out of the eight BCP-ALL tested were lysed, among which three were strongly lysed (ALL-4, 7 and 8) (figures3B 4). Therefore, for all analyses, mAb-independent lysis has been taken into account and is shown in all graphs (no mAb).

Figure 4. Multi-antigen targeting ADCC assays. Primary BCP-ALL leukemic cells and mCD16-T cells were co-cultured at the effector/target ratio of 3:1 in X-VIVO 15 medium in the absence of or with 1 µg/mL of each murine purified antibody and incubated at 37°C for 24 hours. At 0 hours (H0) and 24 hours (H24) of co-culture, residual viable cells were analyzed by MFC. (A) Percentage of residual viable leukemic cells targeted with anti-CD22, anti-CD24, anti-CD268 mAbs (three donors, two replicates for ALL-4 and 5, one for ALL-6). (B) Percentage of residual viable leukemic cells targeted with anti-CD24, anti-CD47, anti-CD123 (four donors, two replicates for ALL-3, one replicate for ALL-4, 7 and 8). ADCC, antibody-dependent cellular cytotoxicity; ALL, acute lymphoblastic leukemia; BCP, B-cell precursor; mAb, murine antibodies, MFC: Multiparametric flow cytometry.

The capacity of 15 mAb, specific to the previously selected antigens, to induce ADCC against four primary BCP-ALL when combined with mCD16-T cells was evaluated. Since isotype antibody influences ADCC activity, when possible, we selected murine IgG1. Thus, out of 15 mAbs tested, 13 are IgG1 (87%), 1 is IgG2a (anti-CD24), and 1 is IgG2b (anti-CD22) (online supplemental table S2). We had previously shown that the mCD16 T lymphocytes were able to mediate ADCC activity in the presence of mouse IgG1, IgG2a, IgG2b and IgG3 anti-human CD20 mAbs, which recognize closely related epitopes in the large extracellular loop of CD20. In addition, our previous ADCC assays performed with mCD16 T cells and murine antibodies against B-lymphocyte antigens using a range of antibody concentrations between 10⁻⁵ and 10 µg/mL revealed bell-shaped curves with, in most cases, a peak of ADCC activity at 1 µg/mL.¹⁶ Positive ADCC controls were performed using a -EBV (Epstein Barr Virus) cell line as target cells, pre-incubated with a murine IgG1 anti-CD19 and anti-CD20 and co-cultured with mCD16-T cells (online supplemental figure S3).

The percentages of residual viable leukemic cells after 24 hours in the absence or presence of mAb (1 µg/mL) and co-cultured with mCD16-T cells are presented in figure 3B. The percentage of residual viable leukemic cells measured from 7% to 100% was dependent on both the target antigen and the BCP-ALL tested. For most of the samples, a high ADCC activity was observed with anti-CD22, CD24, CD47 and CD156c (ADAM10) mAbs (figure 3B). Interestingly, of these four antigens, we observed only for CD24 a correlation between its expression level on BCP-ALL and ADCC lysis (p=0.14) (figure 3C).

Multi-antigen targeting combination

The CD24, which is associated with a high ADCC in all BCP-ALL samples, was selected to be tested in a multitargeting approach. Moreover, the following antigens for which ADCC was observed for all but one sample were also selected to try to assess additive or synergistic effect when targeted in combinations: CD22, CD47, CD123, CD268.

24-hour ADCC assays were performed using two combinations of three antibodies: CD22-CD24-CD268 (figure 4A) and CD24-CD47-CD123 (figure 4B). mAbs combinations were tested against five BCP-ALL samples which express the antigens selected in the combination: ALL-4, which was most sensitive to ADCC lysis in single-target assays, and four additional BCP-ALL samples expressing varying levels of selected antigens. For the anti-CD22-CD24-CD268 antibodies combination, none of the four antibody combinations tested in three BCP-ALL were able to induce higher ADCC lysis than CD24 monotargeting. The addition of anti-CD22 and/or anti-CD268 mAb to anti-CD24 mAb led to a decrease in the leukemic cell lysis as compared with anti-CD24 targeting alone (figure 4A). This decrease in ADCC activity did not seem to be related to the lack of mAb binding, since when revealing with a secondary anti-mouse coupled antibody, the binding of antibodies on target cells continued (data not shown).

For the combination of anti-CD24-CD47-CD123 antibodies tested against four BCP-ALL, the combination of the anti-CD47 mAb with anti-CD24 mAb decreased the anti-CD24 ADCC activity. Only the simultaneous targeting of CD123 and CD24 did not lead to a decrease in anti-CD24 ADCC-related lysis, and in one case (ALL-4), this combination increased the ADCC slightly, suggesting an additive effect (figure 4B).

ADCC cytotoxicity assays on Namalwa cell line

To counteract the variability linked to the use of primary BCP-ALL samples, we also performed ADCC assays with the Namalwa cell line, a B lymphocyte isolated from a patient with Burkitt’s lymphoma. The expression of CD19, CD22 and CD24 was evaluated. After incubation with the first antibody (concentrations range from 10⁻⁵ µg/mL to 10 µg/mL), cells were washed before incubation with the R-PE goat anti-mouse mAb used at saturating concentration. As it is shown in figure 5A (left panel), at 1 µg/mL, the maximum RFI reached a plateau for CD19 and CD22. Note that in contrast to primary BCP-ALL, CD19 and CD22 are expressed at the same level on Namalwa cell line. The RFI for CD24 is higher than those of CD19 and CD22 and is equivalent to that observed on primary BCP-ALL (218 ie, 209). In Chrome-51 ADCC assays in 4 hours with anti-CD24 mAb (clone ML-5), 70% lysis is observed at 1 µg/mL, 53% at 0.1 µg/mL, and no lysis is observed at 0.01 µg/mL (figure 5B). We also performed long-term ADCC killing assays with anti-CD24 and anti-CD19 antibodies at 1 µg/mL or 0.1 µg/mL (figure 5C). Note that, unlike primary BCP-ALL, at 24 hours, the percentage of viable Namalwa cell line increases due to its proliferation. The results confirmed that an efficient ADCC-mediated lysis is obtained with the 1 µg/mL of anti-CD24 Ab (n=3). Consistent with Chrome-51 ADCC assays, no ADCC-mediated lysis is observed at 0.1 µg/mL (n=3) (figure 5C). In addition, dual-targeting ADCC assays with anti-CD24 either at 1 µg/mL or at 0.1 µg/mL and anti-CD19 at 1 µg/mL revealed the decrease of ADCC compare to anti-CD24 alone. The combination CD24-CD123 could not have been tested as the Namalwa cell line is negative for CD123 expression (data not shown).

Figure 5. Target antigen evaluation for ADCC against Namalwa cell line. (A) Namalwa staining with the murine mAbs against CD19 (HIB19, IgG1), CD22 (S-HCL-1, IgG2b) and CD24 (ML-5, IgG2a) and then washed before incubation with the R-PE goat anti-mouse mAb used at saturating concentration. (B) ADCC activity of mCD16-T cells toward Namalwa cell line preincubating with murine anti-CD24 mAbs. Tests were performed in ⁵¹Cr-release assay. Results are expressed as per cent of specific lysis (mean of triplicate, at effector-to-target ratio=30:1). (C) Mono and dual-targeting long-term ADCC killing assays. Results are expressed as the per cent of residual viable cells at H0, H4 and H24 hours after co-culture with mCD16-T cells and with or without mAbs. ADCC, antibody-dependent cellular cytotoxicity; mAb, murine antibodies; PE, phycoerythrin.

Discussion

First, the results presented here provide an in-depth exploration of surface antigen expression in pediatric BCP-ALL. The expression profile of different subtypes of BCP-ALL (KMT2A rearranged, ETV6::RUNX1 or hyperdiploid karyotype) appeared to be in accordance with the previous description of such BCP-ALL.²³ However, because of the small number of BCP-ALL samples used, it was difficult to draw conclusions about the relative antigen expression and/or ADCC sensitivity of each subtype. Multiparametric flow cytometry analysis is an essential tool both in the initial diagnosis work-up, and to assess response to the treatment of BCP-ALL. In routine procedures, the expression of 20–25 markers is generally used to evaluate BCP-ALL at diagnosis.²⁴ Leukemia-associated aberrant immunophenotype (LAIP) marker expression may sometimes be difficult to interpret, and recently, Boris et al reported a better understanding of seven “LAIP markers” landscape in BCP-ALL: CD19, CD21, CD66c, CD58, CD81, CD123, and NG2.²⁵ By providing a broad analysis of the expression profile of 361 antigens by BCP-ALL, this study starts to address this issue.

Second, we selected 13 membrane antigens which can be targeted by murine monoclonal antibodies and our innovative human cytotoxic T-lymphocytes expressing the mouse FcγRIII receptor (mCD16), to directly determine which of them were best suited to induce ADCC-dependent lysis. The level of ADCC-mediated lysis was very variable according to the antibody used and the BCP-ALL tested. As previously reported, we confirmed that antibody opsonization of the target cell is necessary, but not sufficient, to induce ADCC.²⁵ Among the 13 mAbs tested, those targeting CD24 (clone ML5) and CD156c (ADAM10, Clone SHM14) led to effective ADCC-mediated cytotoxicity and, only for CD24, the lysis seemed to be correlated to its level of expression on BCP-ALL, as already described for EpCAM (epithelial cell adhesion molecule, CD326).²⁶

Numerous parameters can influence the efficiency of ADCC, and their relative importance is still unclear. First, ADCC requires a minimal threshold number of antibodies, depending on the affinity and avidity of the antibody, which are linked to the interaction between the bivalent Fab and the antigen. The distance between the epitope and the tumor cell surface also influences ADCC potency²⁷ as does the angle at which the mAbs bind to their epitopes. Indeed, a membrane-proximal epitope is associated with more efficient ADCC potency²⁷ and variations in the orientation of the mAbs, when they bind to the antigen, lead to a 75-fold difference in ADCC potency in the HIV-1 immune response.²⁸ Moreover, the composition of the Fc fragment also impacts the ADCC and may be optimized.²⁹ The evaluation of all these parameters for one target antigen is difficult and requires extensive studies involving the generation of a library of antibodies recognizing different epitopes on the same antigen with different affinities. Our study represents a first step in the search for membrane proteins of interest for ADCC targeting, and identified CD24 and CD156c as potential targets for BCP-ALL. These preliminary results need to be refined by further analysis involving different IgG subclasses, Fc-engineered antibodies, or antibodies targeting different epitopes on the same antigen.

CD24 is a broadly expressed glycosylphosphatidylinositol-anchored protein of 20–70 kD, but its level of expression and glycosylation is highly variable depending on the cell subtype.³⁰ Similarly, like two of the most successful targets in oncology, CD20 and CD52, CD24 also has a small extracellular domain, making it an excellent candidate for the engagement of the Fc-mediated ADCC mechanism.²⁷ CD24 is expressed in several cancer cells, interacts with several tyrosine kinases (Src, STAT3, MAPK), and promotes cell proliferation.³¹ Interestingly, CD24 interacts with Siglec10 in humans, and acts as a “do not eat me” signal, decreasing the host immune response to the tumor.³⁰ In 1988, a preliminary report described remission after treatment with a combination of anti-CD24 and anti-CD21 B-cell antibodies in two patients with oligoclonal B-cell lymphoproliferative syndrome which developed after bone marrow transplantation.³² Later, a multicenter, prospective study analyzed the use of the same antibodies (specific for CD21 (BL13 mouse IgG1) and CD24 (ALB9 mouse-IgG1) in 58 patients presenting with B-lymphoproliferative disorder after bone marrow or organ transplantation, yielding some complete remissions (n=36). Although 27 of the 58 patients treated developed transient neutropenia, among which three had favorable infections, the tolerance was good.³³

The CD156c (ADAM10) is a transmembrane protein (a disintegrin and metalloprotease) involved in cell adhesion and proteolysis.³⁴ The main function of CD156c is the shedding of the extracellular domain of transmembrane proteins such as epidermal growth factor, betacellulin, Notch, CD30 or class I chain-related proteins A and B (MHC class I chain-related protein A and B; MIC-A and B).³⁵ Dysregulation of ectodomain shedding of these molecules is associated with autoimmune and cardiovascular diseases, neurodegeneration, infection, inflammation and cancer.^{36 37} An overexpression of ADAM10 has already been described in leukemia, lymphoma and also in colon, prostate and ovary cancer.³⁶ Moreover, the sheddase activity of ADAM10, particularly with MIC-A and MIC-B, was therefore associated with an impaired immune response against cancer cells, but also the decreased efficacy of monoclonal therapeutic antibodies.38,40 INCB7839, which is a sheddase inhibitor (ADAM10 and ADAM17), showed antitumoral efficacy in vitro when combined with a HER2 inhibitor in breast cancer cell lines.⁴¹ In Australia, Janes and colleagues also developed an antibody directed against a substrate-binding pocket in ADAM10 (8C7) which inhibits tumor growth in mouse models of colorectal cancers and glioma.⁴² Subsequently, this team developed a fully human anti-ADAM10 antibody, specific to the same epitope IH5, which is able to inhibit tumor growth in preclinical models of colorectal cancer, in combination with irinotecan.⁴³ Today, to our knowledge, INCB7839 is the only ADAM10-targeting molecule used in clinical studies, since it is used in a phase I trial involving children treated for high-grade glioma (ClinicalTrial.gov).

The third objective of our study was to evaluate cocktails of antibodies targeting two or more cell surface receptors to enhance ADCC activity and BCP-ALL lysis. Indeed, targeting multiple cell surface antigens may increase both antibody occupancy and antigen clustering through crosslinking and Fc–Fc interactions, resulting in enhanced avidity interactions. However, none of the six combinations of two mAbs and two combinations of three mAbs tested (anti-CD22-CD24-CD268 and anti-CD24-CD47-CD123) showed any synergistic or additive ADCC activities. Additive and/or synergistic tumor cell lysis has already been described in multi-antigen targeting strategies, mainly through complement-dependent cytotoxicity (CDC). Recently, Oostindie et al demonstrated that an IgG antibody mixture of modified anti-CD20 and anti-CD37 antibodies could enhance hexamerization and CDC against tumor B cells, but ADCC was not evaluated in parallel.⁴⁴ Jacobsen and colleagues reported a mixture of six antibodies which, in synergistic pairs, target each of the HER family members EGFR, HER2 and HER3. This mixture, known as Pan-HER (Sym013, Symphogen), was shown to induce potent growth inhibition in a range of cancer cell lines and xenograft models. Pan-HER was also capable of overcoming resistance in HER family-expressing tumors due to an increase in ligand production. Pan-HER induced a level of ADCC in vitro similar to that of cetuximab or trastuzumab, whereas the level of CDC was clearly enhanced.⁴⁵ However, enhanced ADCC activity was described with an mAb cocktail targeting several epitopes of HER2 on human breast cancer cell line BT474 and in xenograft models, as compared with trastuzumab alone.⁴⁶ Although the reasons for this increased ADCC were not explored, this may be attributable to a greater Fc density at the cell surface, or alternatively, a broader range of Fc orientations available for FcγRIIIa engagement.

Since we reported one of the first chimeric Fc-receptor cells, based on the principle of ADCC using genetically modified T cells expressing CD16,⁸ other teams have developed CD16⁺ T cells.^{12 13} However, none of them report the use of these CD16⁺ T cells in ADCC assays with antibodies mixtures to evaluate multi-antigen targeting strategies. The reason may be that, like in our study, they did not observe any additive effects.

Interestingly, although very few oncology studies report the ADCC activity of antibody cocktails, such data are emerging in the field of infectious diseases, in particular for anti-influenza antibodies.³⁸

The single-agent use of monoclonal antibodies for therapy is in many ways contrary to antibody function as established in natural biology. In traditional drug development, single-agent safety and activity is usually required for a drug to be approved, and drug combinations are typically only sought following first approval, often using empirical approaches. However, physiologically, a combinatorial, polyclonal response is the starting point for the immune response to an antigen. An appreciation of this paradox has steadily grown, and antibody-based therapeutics have evolved from canonical monoclonal antibodies to more complex antibody architectures and formats in an effort to enhance functional activity.

Our study has some limitations. First, it was carried out on a small number of primary BCP-ALL samples. Second, the use of primary clinical BCP-ALL samples is associated with extremely variable results, due to both variability in antigen expression and cell viability in vitro. The continuation of our work could be simplified by the use of BCP-ALL cell lines, as the preliminary assays that we have performed with the Namalwa cell line confirmed our first results. In addition, the use of another positive control (different from the anti-CD20 ADCC) involving either blinatumomab-induced lysis or anti-CD19 CAR-T cells, when available, should reinforce the validity of our results. Third, despite the use of mCD16 T lymphocytes with an HLA/peptide-restricted TCR specificity (ie, anti-HLA-A2 CMVpp65), a variable and unpredictable alloreactivity was still observed. The observation of this allogeneic recognition is consistent with data from Melenhorst et al who reported that memory viral antigen-specific T cells can cross-react with specific mismatched HLA-peptide complexes, even in the absence of CMV or EBV peptides.³⁹ Some studies also report alloreactivity-dependent lysis with CAR T cells at variable levels (from 5% to 50% of the cytotoxic activity).12,1447 In general, this lysis is greater when the cytotoxic assays are performed over 24 hours or longer.¹² In order to improve our mCD16-T cell tools, we planned to insert the mCD16 transgene in the TRAC locus by CRISPR-Cas9 to disrupt the TCR expression. Finally, the research presented is a very first exploratory step to identify new potential targets in pediatric BCP-ALL. Regarding the level of expression and the ADCC-mediated lysis obtained with murine monoclonal IgG1/2a antibodies, the CD24 appeared to be a potential candidate. However, these preliminary results need to be refined. Indeed, in this screening step, no tests with different IgG subclasses, nor engineering of antibodies, nor targeting of different epitopes on a same antigen (CD24) have been performed. Further research, involving several anti-CD24 IgG subclasses and polyclonal antibodies are currently ongoing. Similarly, antigens for which no ADCC activity was observed here may be tested using other specific antibodies (targeting more favorable epitopes, or more efficient) in future research.

As a conclusion, our study reports the broad phenotyping of BCP-ALL and the selection of antigens of interest for ADCC targeting by original assays using murine antibodies and mCD16-T lymphocytes as effector cells. Among the 13 antigens selected, CD24 and CD156c have emerged as ADCC-effective targets. Our study is one of the few to report the analysis of ADCC activity with so many antibody combinations. Our results showed that the increase in avidity by multi-antigen targeting rarely led to an increase in ADCC, or even to inhibitory effects. However, such synergies have already been demonstrated with natural polyclonal antiviral responses, suggesting that synergistic combinations are possible. Therefore, analyses of ADCC activity obtained with cocktails of antibodies, or preferentially with polyclonal antibodies or Fc-optimized monoclonal antibodies, against an antigen, should continue to be evaluated as a tumor cell multi-antigen targeting strategy. These studies are necessary for the clinical development of multi-antigen targeting strategies against leukemic cells, based on ADCC activity, using mAbs and cell adoptive transfers with natural killer cells or FcR⁺-T lymphocytes. Another multi-antigen targeting strategy that we are evaluating is to engineer T-lymphocytes with two recognition systems: CD16 (ADCC) and CAR.

Data availability statement

Data are available upon reasonable request.