Systematically optimized BCMA/CS1 bispecific CAR-T cells robustly control heterogeneous multiple myeloma

Abstract

Chimeric antigen receptor (CAR)-T cell therapy has shown remarkable clinical efficacy against B-cell malignancies, yet marked vulnerability to antigen escape and tumor relapse exists. Here we report the rational design and optimization of bispecific CAR-T cells with robust activity against heterogeneous multiple myeloma (MM) that is resistant to conventional CAR-T cell therapy targeting B-cell maturation antigen (BCMA). We demonstrate that BCMA/CS1 bispecific CAR-T cells exhibit superior CAR expression and function compared to T cells that co-express individual BCMA and CS1 CARs. Combination therapy with anti-PD-1 antibody further accelerates the rate of initial tumor clearance in vivo, while CAR-T cell treatment alone achieves durable tumor-free survival even upon tumor re-challenge. Taken together, the BCMA/CS1 bispecific CAR presents a promising treatment approach to prevent antigen escape in CAR-T cell therapy against MM, and the vertically integrated optimization process can be used to develop robust cell-based therapy against novel disease targets.

Conflict of interest statement

Y.Y.C. and E.Z. declare competing financial interest in the form of a patent application whose value may be affected by the publication of this work. The other authors declare no competing interests.

Figures

Fig. 1. High-throughput generation and screening of BCMA/CS1 OR-gate CARs.

a A vertically integrated optimization process for CAR-T cell therapy development. b Schematic of a single-chain bispecific (OR-gate) CAR, which contains two ligand-binding domains connected in tandem. A panel of CAR variants was constructed from two CS1-binding scFvs and three BCMA-specific binding domains. Both murine and humanized versions of BCMA-binding scFvs (c11D5.3 and J22.9-xi) were evaluated. The CS1-binding huLuc63 and Luc90 scFvs were fixed at the membrane-distal and membrane-proximal positions, respectively, based on binding-epitope analysis. c Methodology for producing and screening bispecific CARs. CAR-T cells were generated in a 14-day cycle. Starting on day 9 post-stimulation, CAR-T cells were characterized for antitumor function in various assays, which could last up to 2 weeks.

Fig. 2. OR-gate CAR panel exhibits range of efficacy against BCMA+ and CS1+ targets.

a Cell-lysis activity of single-input and bispecific CD8+ CAR-T cells against K562 targets engineered to express either BCMA or CS1. Cells were seeded at an effector-to-target (E:T) ratio of 2:1, where effector-cell seeding was based on CAR+ T-cell count. The fraction of viable K562 cells left after a 20-h coincubation was quantified by fluorescence imaging of target cells using IncuCyte. All bispecific CARs in this panel contained a short extracellular spacer. b Proliferation of single-input and bispecific BCMA/CS1 CD8 + CAR-T cells upon antigen stimulation. CAR-T cells were stained with CellTrace Violet (CTV) dye. CTV median fluorescence intensity (MFI) was quantified by flow cytometry after a 5-day coincubation with parental (BCMA−/CS1−), BCMA+, or CS1+  K562 target cells at a 2:1 E:T ratio. CARs containing huLuc63 paired with dAPRIL, J22.9-xi, and huJ22.9-xi were subsequently eliminated from the panel based on poor cytotoxicity and/or T-cell proliferation. c Cytotoxicity of reduced bispecific CAR-T-cell panel upon repeated antigen challenge. CD8+ CAR-T cells were coincubated with K562 target cells at a 1:1 E:T ratio and rechallenged every 2 days with fresh target cells. Viable target-cell count was quantified by flow cytometry 2 days after each target-cell addition. ‘C#’ denotes the challenge number and ‘D#’ denotes the number of days post challenge. d Characterization of the remaining K562 target-cell populations after four challenges from c reveals differing antigen preference among the panel of bispecific CARs. Values shown are the means of technical triplicate samples, with error bars indicating +1 standard deviation (SD). P-values were calculated by unpaired two-tailed Student’s t-test; n.s. not statistically significant (p > 0.05); *p < 0.05; **p < 0.01, with Bonferroni correction for multiple comparisons applied. P-values in c were calculated for the final time point for each construct, relative to the top-performing CAR, c11D5.3-Luc90. Source data are provided as a Source Data File. P-values in a and b can be referenced in the Source Data File.

Fig. 3. OR-gate CAR-T-cells outperform T cells co-expressing two separate CARs.

a Single-input and bispecific CARs were tagged with an N-terminal FLAG tag. In DualCAR constructs, the CS1 CAR was N-terminally tagged with a FLAG tag while the BCMA CAR was N-terminally tagged with a HA tag. b CAR surface expression levels were quantified by surface antibody staining of FLAG and HA tags followed by flow cytometry. Each single-input and single-chain bispecific CAR was tagged with an N-terminal FLAG. The first CAR in each DualCAR construct was tagged with the FLAG while the second CAR was tagged with HA. See Supplementary Fig. 4b for construct schematics. c Transduction efficiency as measured by % CAR+ (FLAG+, HA+, or FLAG+HA+) Data shown represent the mean value from three different donors, with error bars indicating +1 SD. d Median fluorescence intensity (MFI) of FLAG antibody staining for T cells expressing either OR-gate or DualCAR constructs. e NM CAR-T-cell proliferation following a 5-day coincubation with MM.1 S target cells. Values shown are the means of technical triplicate samples from the same donor, with error bars indicating +1 SD. Data shown are representative of results from two independent experiments performed with cells from two different healthy donors. P-values for all panels were calculated by unpaired two-tailed Student’s t-test; n.s. not statistically significant (p > 0.05); *p < 0.05; **p < 0.01, with Bonferroni correction for multiple comparisons applied. P-values in c were calculated for the OR-gate CAR constructs, relative to their corresponding DualCAR constructs. Source data are provided as a Source Data File. P-values in e can be referenced in the Source Data File.

Fig. 4. BCMA/CS1 OR-gate CAR-T cells prevent antigen escape in vivo.

a Evaluation of single-input and bispecific CAR-T cells in vivo. Mice were engrafted with a mixture of 1.5 × 106 MM.1 S cells containing a 1:1:1 ratio of BCMA+/CS1−, BCMA−/CS1+, and BCMA+/CS1+ cells. Tumor-bearing animals were treated with 1.5 × 106 EGFRt- or CAR-expressing T cells on day 5 (5 days after tumor injection) and day 13. Six mice were included in each initial treatment group but only five mice in the huLuc63-c11D5.3 group were redosed due to limited T-cell availability. Tumor progression was monitored by bioluminescence imaging. b Survival of mice shown in a. Statistical difference (depicted) between survival of huLuc63-c11D5.3-treatment group compared with other treatment groups was determined using log-rank analysis, applying a Chi-square distribution with one degree of freedom; n.s. not statistically significant (p > 0.05); *p < 0.05; **p < 0.01. P-values for the different treatment groups are as follows: Untreated, p = 0.024; EGFRt, p = 0.024; Luc90, p = 0.011; huLuc63, p = 0.015; c11D5.3, p = 0.008; huc11D5.3-Luc90, p = 0.114. c BCMA/CS1 antigen expression on tumors harvested from mice treated with CS1 single-input Luc90 Short, huLuc63 Long, or BCMA single-input c11D5.3 Long CAR-T cells. Values shown are the means of n = 4 technical replicates for cells stained prior to injection. Thirteen to 18 tumor samples were collected from different locations in the mice (six mice per test group) at the time of sacrifice, with error bars indicating +1 standard deviation (SD), and each sample corresponding to 100 or more events as detected by flow cytometry. d Antigen expression pattern on tumor cells recovered at the time of animal sacrifice. Each data point d corresponds to an individual tumor sample recovered that included more than 100 tumor cells as detected by flow cytometry; each mouse generally contained multiple tumors at the time of sacrifice. Source data are provided as a Source Data File.

Fig. 5. Combination therapy with anti-PD-1 increases initial antitumor efficacy but not durability of response in vivo.

a Mice were engrafted with 1.5 × 106 WT MM.1 S cells. Tumor-bearing animals were treated with 1.5 × 106 EGFRt- or CAR-expressing T cells on day 8 (8 days after tumor injection) and day 16. Tumor progression was monitored by bioluminescence imaging and shown for individual animals in each test group (n = 6). On day 133, animals that had been tumor-free for at least 7 weeks (3 each in the huLuc63-c11D5.3 Short groups with and without anti-PD-1) were rechallenged with 1.5 × 106 WT MM.1 S cells. Black arrows indicate time of cell injections; red arrows indicate instances of CAR-T cells eradicating palpable tumor nodules. The endpoint for each animal is marked with an “X.” b CD38 and PD-1 expression on T cells persisting in mice at time of animal sacrifice. Each data point in b corresponds to an individual tumor mass, tissue (e.g., brain and spleen), or cardiac blood sample recovered. For PD-1 expression: 23 samples were collected from six mice in the group not treated with anti-PD-1, and 26 samples were collected from six mice in the group treated with anti-PD-1. For CD38 expression: seven samples were collected from two mice in the group not treated with anti-PD-1, and nine samples were collected from three mice in the group treated with anti-PD-1. Only samples that contained at least 10 human CD45+ cells as detected by flow cytometry were included for analysis. P-values were calculated by unpaired two-tailed Student’s t-test. Source data are provided as a Source Data File.