Peptide-scFv antigen recognition domains effectively confer CAR T cell multiantigen specificity

Abstract

The emergence of immune escape is a significant roadblock to developing effective chimeric antigen receptor (CAR) T cell therapies against hematological malignancies, including acute myeloid leukemia (AML). Here, we demonstrate feasibility of targeting two antigens simultaneously by combining a GRP78-specific peptide antigen recognition domain with a CD123-specific scFv to generate a peptide-scFv bispecific antigen recognition domain (78.123). To achieve this, we test linkers with varying length and flexibility and perform immunophenotypic and functional characterization. We demonstrate that bispecific CAR T cells successfully recognize and kill tumor cells that express GRP78, CD123, or both antigens and have improved antitumor activity compared to their monospecific counterparts when both antigens are expressed. Protein structure prediction suggests that linker length and compactness influence the functionality of the generated bispecific CARs. Thus, we present a bispecific CAR design strategy to prevent immune escape in AML that can be extended to other peptide-scFv combinations.

Keywords: AML; B7H3; CAR T cell therapy; CD123; GRP78; bispecific CAR; chimeric antigen receptor; immune escape; leukemia; structure prediction.

Graphical abstract

Figure 1

Bispecific 78.123 CAR T cells can be successfully manufactured and recognize antigen-positive targets in vitro (A) Schematic of mono- and bispecific CAR constructs. (B) Graph depicting length and rigidity of chosen linkers. (C) Graphical depiction of workflow in generating bispecific CAR T cells. (D and E) T cells were analyzed by flow cytometry for CAR expression using (D) anti-CD19 antibody (GRP78 95.5% ± 1.8%, B2M 92.3% ± 1.4%, G4S3 96.7% ± 1.9%, GPcPcPc 91.0% ± 3.6%, mtIgG4 89.3% ± 2.9%, n = 4 biological replicates, mean ± SD) and (E) recombinant CD123 protein (CD123 74.48% ± 4.5%, B2M 37.8% ± 36%, (G4S)3 87.5% ± 7.8%, GPcPcPc 83.6% ± 9.2%, mtIgG4 41.0% ± 32%). Evaluation of IFN-γ secretion by ELISA. (F–I) Non-transduced (NT), control CAR, and mono- and bispecific CAR T cells were cocultured with RPMI8402 (F) KG1a (G), MOLM13 (H), or recombinant protein (I, 1 μg/mL) at a 2:1 E/T ratio. Supernatants were harvested after 24 h and analyzed for IFN-γ by ELISA assay (n = 4 biological replicates,mean ± SD, one-way ANOVA, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). (J–M) Evaluation of antitumor activity CAR T cells by flow-based and luciferase-based cytotoxicity assay. Monospecific (GRP78, CD123), bispecific (B2M, (G4S)3, GPcPcPc, mtIgG4), or control effector T cells (NT, HER2 CAR) were cocultured in the presence of target cells of different antigen densities at five different E/T ratios. (J) RPMI8402: 1:1 p < 0.05 HER2 vs. CD123, HER2 vs. B2M, HER2 vs. mIgG4, p < 0.01 B2M vs. G4S3, G4S2 vs. mIgG4; 1:2 p < 0.05 CD123 vs. B2M; 1:4 p < 0.05 HER2 vs. CD123, HER2 vs. B2M, B2M vs. mIgG4, p < 0.01 CD123 vs. (G4S)3. (K) MOLM13 and (L) KG1a at 1:1 HER2 vs. CARs, p < 0.01 all conditions. (M) KG1a KO, HER2 vs. GRP78, G4S3, B2M, mIgG4, p < 0.01; HER2 vs. CD123 and HER2 vs. GPcPcPc, data not significant. Two-way ANOVA,mean ± SD, Tukey’s multiple comparisons.

Figure 2

78.123 CAR T cells maintain antitumor activity upon serial stimulation (A and B) Effector T cells and (A) KG1a or (B) KG1a KO cells were cocultured at a 1:1 E/T ratio. Fresh target cells were added every 72 h if a luciferase-based cytotoxicity assay demonstrated greater than 50% killing (n = 3 biological replicates). Number of stimulations are presented for each donor (KG1a A, KG1a KO B, n = 4 biological replicates,mean ± SD,one-way ANOVA, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). (C and D) Multiplex analysis of cytokine production by CAR T cells from stimulation 1 or 3 from (C) KG1a or (D) KG1a KO cells at a 1:1 E/T ratio.

Figure 3

Bispecific CAR T cells demonstrate calcium flux in the presence of antigen, suggestive of an intact immune synapse (A) Representative single-cell calcium flux analysis of monospecific and bispecific CAR T cells interacting with MOLM13 AML cells over 1 h. MOLM13 cells are labeled in blue, T cells are labeled in red, and calcium flux is indicated by green-yellow. Scale bar, 10 μm. (B and C) (B) Calcium flux quantification curves and (C) area under the curve (AUC) starting at 1 min post coculture to 10 min for the total peak area of three separate donors (biological replicates, and within each replicate the following N is technical replicates) with NT (n = 28), HER2 (n = 39), GRP78 (N = 27), CD123 (n = 27), (G4S)3 (n = 17), B2M (n = 24), GPcPcPc (n = 23), and mtIgG4 (n = 22). Two-way ANOVA (mean ± SD, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001) was used to determine statistical significance with uncorrected Fisher’s LSD test.

Figure 4

Bispecific CAR T cells have potent anti-AML activity and persist in vivo NSG mice were injected intravenously via tail vein with MOLM13.GFP.ffluc cells, After 7 days, mice were injected with CAR T cells. Mice were monitored using IVIS imaging to track bioluminescence (total flux photons/s). (A) Bioluminescence and (B) survival of mice injected with MOLM13.GFP.ffluc cells (p < 0.0001, log-rank Mantel-Cox, CD123 vs. G4S3, p < 0.001). (C–F) NSG mice were injected intravenously via tail vein with KG1a.GFP.ffluc cells or KG1a KO.GFP.ffluc cells. On day 7, mice received a single T cell dose. Mice were monitored using an in vivo imaging system to track (C) bioluminescence (total flux photons/s, AUC analysis of mono- vs. bispecific, data not significant) and (D) survival of mice injected with KG1a.GFP.ffluc (p < 0.01, log-rank Mantel-Cox). (E) Bioluminescence (total flux photons/s) and (F) survival (p < 0.01, log-rank Mantel-Cox) of mice injected with KG1a KO.GFP.ffluc.

Figure 5

Structural predictions for extracellular portions of mono- and bispecific CAR T cells via AlphaFold2 (A and B) Structural predictions of (A) GRP78 peptide and (B) CD123 CAR. Predicted aligned errors (PAEs) are shown on the left and structures on the right for each construct in their respective panels. (C–F) Structural predictions of 78.123 bispecific CARs with different linkers such as (C) (G4S)3, (D) B2M, (E) mIgG4, and (F) GPcPcPc are shown. Similar to (A) and (B), PAE scores are shown on the left and structures on the right in respective panels. As indicated by the PAE scores, also reflected in the structures, interactions between heavy and light chains of CD123 are compromised in the presence of (D) B2M and (E) mIgG4 linkers. (G) Summary of in vitro and in vivo results of 78.123 bispecific CAR T cells.

Figure 6

Bispecific CAR T cells induce cytotoxic response in vitro and in vivo to primary samples and patient-derived xenografts (A–C) Analysis of a panel of ten pediatric bone marrow AM samples collected at diagnosis except for patient 3 collected at relapse. (A) Genetic driver information and (B) cell-surface GRP78 and CD123 antigen expression as determined by flow cytometry. (C) IFN-γ ELISAs were performed from supernatant collected from coculture assays between effector T cells (n = 3 donors, biological replicates) and primary AML samples at a 1:1 E/T ratio after 24 h (mean ± SD, one-way ANOVA, Tukey’s multiple comparisons test to HER2 control CAR, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). (D) Patient-derived xenograft with UBTF-TD54 duplication expressing YFP.ffluc were injected into NSG-SGM3 mice. On day 7, mice received a single 3 × 10⁶ T cell dose. Bioluminescence (total flux photons/s) was measured (n = 5 per group, day 47, p > 0.01 UBTF-TD54 vs. GRP78, UBTF-TD54 vs. CD123, UBTF-TD54 vs. (G4S)3).

Figure 7

Characterization of 78.B7H3 bispecific CAR T cells (A) Schematic of mono- and bispecific 78.B7H3 CAR constructs. (B and C) Transduced T cells analyzed by flow cytometry for transduction efficiency by (Fab′)₂ (B, NT 11.1% ± 12.9%, HER2 73.1% ± 29.5%, GRP78 19.6% ± 13.8%, B7H3 CD8 66.3% ± 13.2%, B7H3 CD28 51.8% ± 12.0%, 78.B7H3 CD28 78.86% ± 9.3%, 78.B7H3 CD8 16.7% ± 12.8%, mean ± SD, one-way ANOVA, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, n = 3) and CD19 (C, NT 4.1% ± 2.3%, HER2 5.6% ± 3.6%, GRP78 95.7% ± 2.0%, B7H3 CD8 5.5% ± 3.5%, B7H3 CD28 10.0% ± 4.2%, 78.B7H3 CD28 89.6% ± 4.2%, 78.B7H3 CD8 4.2% ± 3.0%, mean ± SD, one-way ANOVA, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, n = 4–7). (D) Cell viability measured via trypan blue exclusion (day 2: p < 0.05 NT vs. HER2, NT vs. B7H3 CD8, p < 0.01 NT vs. GRP78, NT vs. 78.B7H3 CD8; day 7: p < 0.05 NT vs. B7H3 CD8, NT vs. B7H3 CD28, HER2 vs. GRP78, HER2 vs. B7H3 CD8, p < 0.01 NT vs. GRP78, NT vs. 78.B7H3 CD8, HER2 vs. 78.B7H3 CD8, GRP78 vs. 78.B7H3 CD28; day 10: p < 0.05 NT vs. B7H3 CD28, B7H3 CD8 vs. B7H3 CD8, 78.B7H3 CD28 vs. B7H3 CD28, p < 0.01 GRP78 vs. B7H3 CD8, p < 0.001 NT vs. GRP78, HER2 vs. GRP78, GRP78 vs. 78.B7H3 CD28, GRP78 vs. 78.B7H3 CD8; mean ± SD, two-way ANOVA mixed modeling). (E) T cell expansion measured up until 10 days post transduction (day 2: p < 0.05 NT vs. GRP78; day 5: p < 0.05 NT vs. HER2, NT vs. B7H3 CD8, NT vs. 78.B7H3 CD28, NT vs. 78.B7H3 CD8, HER2 vs. GRP78, HER2 vs. B7H3 CD8, HER2 vs. B7H3 CD28, p < 0.01 NT vs. B7H3 CD28, p < 0.001 NT vs. GRP78; day 7: p < 0.05 HER2 vs. 78.B7H3 CD28, HER2 vs. 78.B7H3 CD8, 78.B7H3 CD28 vs. 78.B7H3 CD8; day 10: p < 0.05 NT vs. GRP78, p < 0.01 NT vs. 78.B7H3 CD8, HER2 vs. GRP78, HER2 vs. 78.B7H3 CD8, p < 0.001 GRP78 vs. 78.B7H3 CD28, 78.B7H3 CD28 vs. 78.B7H3 CD8;mean ± SD, two-way ANOVA mixed modeling). (F) Immunophenotype of CAR T cells on days 6–8 post transduction. CD4: p < 0.05 78.B7H3 CD28 vs. 78.B7H3 CD8, p < 0.01 B7H3 CD8 vs. 78.B7H3 CD28; CD8: p < 0.05 GRP78 vs. 78.B7H3 CD28, B7H3 CD8 vs. 78.B7H3 CD8, p < 0.01 NT vs. HER2, p < 0.001 HER2 vs. 78.B7H3 CD8; [EM: CCR7⁻, CD45RO⁺; CM: CCR7⁺, CD45RO⁺; naive-like: CCR7⁺CD45RO⁻; EMRA: CCR7⁻, CD45RO⁻] CD8 EM p < 0.05 GRP78 vs. B7H3 CD8; CD8 CM p < 0.05 GRP78 vs. B7H3 CD28; CD8 N p < 0.05 HER2 vs. GRP78; mean ± SD, two-way ANOVA mixed modeling. (G and H) Cytotoxicity assay of CAR T cells single antigen against target cells THP-1 (G, GRP78⁺B7H3⁺) and KG1a (H, GRP78⁺B7H3⁻) at seven different E/T ratios (THP-1 1:1 p < 0.05 HER2 vs. GRP78, p < 0.01 HER2 vs. 78.B7H3 CD28, p < 0.001 HER2 vs. B7H3 monospecific CARs; KG1a 1:1 p < 0.05 HER2 vs. GRP78, GRP78 vs. B7H3 CD28, p < 0.01 GRP78 vs. B7H3 CD8, B7H3 CD8 vs. 78.B7H3 CD28, 78.B7H3 CD28 vs. 78.B7H3 CD8, p < 0.0001 HER2 vs. 78.B7H3 CD28, GRP78 vs. 78.B7H3 CD8, B7H3 CD28 vs. 78.B7H3 CD28; mean ± SD, two-way ANOVA, Tukey’s multiple comparisons).