Systematic Engineering of TROP2-Targeted CAR T-Cell Therapy Overcomes Resistance Pathways in Solid Tumors

Abstract

Antibody-based therapies have revolutionized cancer treatment but have several limitations. These include downregulation of the target antigen, mutation of the target epitope, and, in the case of antibody-drug conjugates (ADC), resistance to the chemotherapy warhead. As TROP2-targeted therapy with ADCs yields responses in TROP2+ solid tumors, but the responses lack the durability observed with other immunotherapy-based approaches, we developed TROP2-targeting chimeric antigen receptor (CAR) T cells as an alternative. The TROP2-directed CAR T cells showed high potency against multiple solid tumor models. Moreover, TROP2-directed CAR T-cell therapy preserved high potency in models of ADC resistance and could be further engineered to prevent cell therapy resistance. This was achieved by leveraging fully human single-domain (VH-only) binder discovery to rationally engineer dual epitope binding-based (biparatopic) CARs. This work highlights the potency of CAR T-cell therapies and how rational engineering leveraging dual-VH targeting domains can overcome resistance pathways to current therapies. In future work, the CAR engineering approaches presented here can serve as a platform to be partnered with other strategies to address the suppressive tumor microenvironment.

Figure 1.

TROP2-targeted CARs are highly effective in vitro and in vivo against TROP2⁺ solid tumors. A, In vitro cytotoxicity of TROP2 CAR T cells against PC9 (EGFR exon 19 del NSCLC) and HCC70 (TNBC) compared with irrelevantly targeted negative controls. CAR T cells were incubated at indicated E:T ratios with targets for 24 hours. Data are representative of two independent experiments from unique donors with technical triplicates, with error bars showing SD. B, TROP2 CARs maintain cytotoxic potential against PC9 with the osimertinib-acquired resistant EGFR C797S mutation. Cell line targets were incubated with either osimertinib at increasing concentrations or CARs. For viability, osimertinib was normalized to vehicle, in which tumor viability for CARs was normalized to the irrelevantly targeted control CAR at indicated E:T. Measurement was performed as in (A). Data are representative of two independent experiments with unique donors and technical triplicates, with error bars showing SD. C, TROP2 CARs maintain activity against 3D PC9 spheroids. Red object area was plotted over time with technical triplicates, with error bars showing SD. D, In vivo activity of TROP2 CARs in NSCLC xenograft. PC9 was injected subcutaneously in NSG mice, and after tumors reached ∼100 mm³, irrelevantly targeted control or TROP2 CAR T were administered at indicated doses on day 0, as indicated by the arrow. In the TROP2 CAR 0.3E6 group, one of five mice and, in the 1.3E6 dose group, two of five mice, were euthanized at days 51 and 78 due to the general health of the mice (possibly GVHD) and not tumor growth. Data are representative of two different experiments with different donors. E, Orthotopic model of PC9. Mice were injected with PC9 ffLuc⁺ and, after engraftment confirmed, given control irrelevantly targeted control CARs or TROP2 CARs at indicated doses. Tumor was measured subsequently by bioluminescence imaging. Data are representative of two different experiments with different donors. F, Activity of TROP2 CARs in an ex vivo organoid model utilizing a microfluidic device against patient-derived lung adenocarcinoma DFCI243 (EGFR exon 19 del/T790M). Cell markers are indicated in the figure. G, Percentage of live/dead cell analysis of (F) performed by measuring the total cell area of each dye with technical triplicates, with error bars showing SD. H, In vivo activity of TROP2 CARs PDX DFCI243 (EGFR exon 19 del/T790M) and (EGFR L858R/MET amp), as well as DFCI642 (TROP2⁻EGFRm NSCLC transformed to SCLC). CAR cells (3 to 6 × 10⁵) were injected at day = 0, as indicated by the arrow, and tumor volumes were measured. Each line represents an independent mouse tumor volume measurement, with magenta for the irrelevant control and green for TROP2. P values are reported as follows compared with control: ns, P > 0.05; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001. PI, propidium iodide.

Figure 2.

Models of acquired resistance to TROP2 ADC can be overcome with TROP2 CARs. A, PC9 with KO of TROP2 were reconstituted with either WT TROP2 or the T256R mutation, which has been previously described to have developed in acquired resistance to TROP2 ADC. Mean fluorescence intensity after TROP2 staining by flow cytometry on viable (DAPI-negative) cells is shown. B, In vitro viability of constructs described in (A) after generation of spheroids and treatment in an ultralow attachment 96-well plate with datopotamab deruxtecan. Cells were incubated with ADC for 6 days, followed by viability assessment by luciferase assay. Viability was normalized to vehicle (0 µg/mL) for each construct. Data are representative of two different experiments with technical triplicates, with error bars showing SD. C, In vitro viability of either WT TROP2 or T256R TROP2 with TROP2 CARs at a 1:1 E:T ratio. Assay after 24 hours of incubation. Data are representative of two different experiments with different donors and technical triplicates, with error bars showing SD. D, In vitro viability of either WT TOP1 or E418K TOP1 with TROP2 CARs after 24 hours of incubation, with technical triplicates, with error bars showing SD. E, Model of resistance to TROP2 ADC by TROP2 T256R mutation or TOP1 mutation and how TROP2 CARs are able to maintain cytolytic activity, created with Biorender. P values are reported as follows compared with control: *, P ≤ 0.05; **, P ≤ 0.01; ****, P ≤ 0.0001. dsDNA, double-stranded DNA; GzmB, granzyme B.

Figure 3.

VH-only binder discovery, CAR engineering, and screening results in the identification of CARs with high efficacy and capable of binding unique epitopes. A, In vivo activity of selected constructs from (S4) with an orthotopic model of PC9. Mice were injected at day (d) = −14 with PC9 ffLuc⁺ and, after engraftment was confirmed, given irrelevantly targeted control CAR or TROP2 CAR at 0.3E6 CAR⁺/mouse, as indicated by the arrow. Tumor was measured subsequently by bioluminescence imaging. Data are representative of two different experiments with different donors. B, Survival of mice from (A) with statistics shown using log-rank (Mantel–Cox) of hRS7 to VH681. C, In silico protein–protein docking prediction of sacituzumab, modeled from the FASTA primary sequence using antibody homology modeling (MOE), in complex with TROP2. The inset shows putative key residues and their side chains participating in protein–protein contacts, namely van der Waals interactions (pink halos), at the epitope–partatope interface. D, In vitro viability of TROP2 (hRS7-based) CARs against WT, KO, or murine Q237-252 substitution of TROP2. Performed at an E:T ratio of 2:1 and cocultured for 24 hours. Data are representative of two different experiments with different donors, with technical triplicates, with error bars showing SD. E, TROP2 VH binds to distinct epitopes compared with scFv-based CAR constructs. mTROP2 is murine TROP2, whereas mCPD and mCRD are murine CRD and CPD, with the remaining domains derived from the human TROP2 amino acid sequence. Heatmap demonstrates that selected TROP2 VH constructs have activity against a unique domain compared with scFv-based (hRS7 and dato) CAR constructs. Viability is normalized to an irrelevantly targeted control CAR for each construct. Data are representative of two different experiments with different donors. F, In silico protein–protein docking prediction of TROP2 VH375 and VH681, modeled from the FASTA primary sequence using antibody homology modeling (MOE), in complex with TROP2. The inset shows putative key residues and their side chains participating in protein–protein contacts, namely van der Waals interactions (pink halos), at the epitope–partatope interface. P values are reported as follows compared with control: *, P ≤ 0.05; ****, P ≤ 0.0001. TY domain, thyroglobulin type-1 domain.

Figure 4.

Rational epitope binding–based design of biparatopic CARs leveraging single-domain VH-only binders overcomes models of resistance to single epitope–targeted approaches. A, Schematic of biparatopic TROP2 VH-based CARs in a second-generation CAR vector with a linker between anti-CRD VH and anti-CPD VH, as well as a hinge/transmembrane domain (H/TM), 4-1BB costimulatory domain, and CD3ζ signaling domain. B, In vitro cytotoxicity of TROP2 scFv, VH, and biparatopic VH CAR against PC9 with various substitutions of TROP2 domains with murine domains as indicated. Performed at a 2:1 E:T ratio and cocultured for 24 hours before viability was assessed via luciferase assay and normalized to an irrelevant BCMA control. Data are representative of two different experiments with different donors. C, In silico protein–protein docking prediction of biparatopic TROP2 VH681_375, modeled from the FASTA primary sequence using antibody homology modeling (MOE), in complex with TROP2. D, In vitro activity in a live-cell imaging assay of TROP2 CARs against PC9. PC9 TROP2 KO was stably transduced with murine CPD and mCherry, and live-cell imaging was performed on Incucyte with counting of red objects. The cytotoxicity index is the red cell object count relative to time (t) = 0 when CARs were applied at an E:T ratio of 0.25:1. Data are representative of two different experiments with different donors, with technical triplicates with error shading showing SEM. E, Similar to (D), with PC9 TROP2 murine CRD clone transduced with GFP, with counting of green objects. Data are representative of two different experiments with different donors, with technical triplicates with error shading showing SEM. F, PC9 harboring murine Q237-252 (CPD) or murine CRD tumors were injected subcutaneously into separate flanks of NSG mice, and after tumors reached ∼100 mm³, CD19 irrelevant control or TROP2 CARs were administered at a one-time dose of 2.5E6 CAR⁺ cells via the tail vein, as indicated by the arrow. Tumors measured by caliper measurement of the mQ237-252 tumor on the left flank, with individual (dim) and mean (solid) measurements shown for groups. Data are representative of two different experiments with different donors. G, Tumors were measured by caliper measurement of the mCRD tumor on the right flank, with individual (dim) and mean (solid) measurements shown for groups. Data are representative of two different experiments with different donors. H, Survival of mice from F to G with biparatopic TROP2 CAR VH681_375 compared with benchmark hRS7 CAR (60 vs. 36.5 days, P = 0.001). Survival statistics were reported by log-rank (Mantel–Cox) test. P values are reported as follows compared with control: **, P ≤ 0.01; ****, P ≤ 0.0001.