EVOLVE platform, a trispecific T cell engager with integrated CD2 costimulation, for the treatment of solid and hematologic tumors

Abstract

Ten CD3 T cell engagers (TCEs) have received regulatory approval for the treatment of hematologic and solid tumors. However, limited costimulatory signaling essential for sustained T cell effector activity may limit CD3 TCE clinical efficacy and response duration. The CD2 receptor is an attractive costimulation target owing to its association with T cell receptor signaling and favorable expression profile. We show that CD2 costimulation is superior in maintaining T cell viability and effector function relative to other pathways in in vitro chronic stimulation assays. The extracellular domain of CD58, the predominant CD2 ligand, is functional as an antibody fusion, improving bispecific potency. We observe that higher CD3 affinity molecules have the potential for superagonism in the context of an integrated CD2 agonist. Evaluation of TCEs with integrated CD2 costimulation and attenuated CD3 binding identified optimal CD3 affinity agonists that avoid target-independent T cell activation and demonstrated an increased therapeutic index relative to nonattenuated CD3 agonists. This platform shows increased tumor-killing efficacy as compared to CD3 affinity-matched bispecifics for known tumor targets such as HER2, CD20, B7-H4, and UL16-binding protein 2 (ULBP2). We demonstrate that ULBP2-targeted trispecifics with integrated CD2 costimulation and optimized CD3 affinity are superior to higher-affinity CD3 molecules in in vivo mouse efficacy studies. This integrated CD2 costimulation platform, which we termed EVOLVE, represents a next-generation TCE platform to increase T cell effector function in the tumor microenvironment and has the potential to address unmet patient needs by improving the depth and durability of clinical antitumor T cell responses.

Keywords: T cell engager; costimulation; immune–oncology; immunotherapy.

Fig. 1.

CD2 costimulation is superior to other costimulatory receptor–ligand pairs. (A) The gene names, synonyms, transmembrane types, and extracellular domain graphical depiction of the six ligands used in the costimulation screen (Left). The schematic of the costimulation screen with five rounds of pan T cells being added to a plate containing the extracellular domain of the ligand and OKT3 with the readouts listed (Right). Created with BioRender.com. (B) T cell viability and CD8 T cell expansion for the screen shown in A. T cell expansion was normalized to 100% for ease of data representation. (C) Further, T cell activation by CD25, and T cell effector function by intracellular IFNγ and Granzyme B. This experiment was repeated with other ratios of anti-CD3 to Fc-fusion protein and the data shown are representative of all ratios tested. Statistical tests between ligands compared to the CD3 only condition were carried out across the row using Friedman’s test with Dunn’s multiple comparison correction. (D) CD8+ tumor infiltrating lymphocytes were identified from seven publicly accessible single cell RNA-Seq datasets of bladder, head and neck, and non–small cell lung cancer (Materials and Methods). (E) The percentage of CD8+ T cells or further identified Naïve-Like (N-L), Effector Memory (EM), Precursor Exhausted (PEx), or Terminally Exhausted (TEx), that express CD2, CD28, or 4-1BB were quantified. Each point is a different single-cell RNA-Seq dataset for n = 7 total. Statistical tests between subsets were carried out using one-way ANOVA with Holm–Sidak multiple comparison correction. (F) TDCC assays with activated T cells with an E:T ratio of 5:1 and a 48-h readout (except for EGFR/HCT116 which had a 5-d readout). BCMA bispecific (black) and with integrated CD2 costimulation (red) were assayed in TDCC with BCMA-low myeloma U266-B1 cells. CD19 bispecific (black; based on TNB-486) and with integrated CD2 costimulation were assayed in TDCC with CD19-high Burkitt’s lymphoma Raji cells. EGFR bispecific (black) and with integrated CD2 costimulation (red) were assayed in TDCC with EGFR-positive colon carcinoma HCT116 cells. PSMA bispecific (black; based on TNB-585) and with integrated CD2 costimulation (red) were assayed in TDCC with engineered PSMA-low prostate LNCaP cells. Cartoons for the bispecific formats (black) and with integrated CD2 agonist (red) are shown next to each curve. Each dataset is representative of 2 to 3 similar experiments.

Fig. 2.

Detuning CD3 affinity to optimize a trispecific molecule with integrated CD2 costimulation. (A) ULBP2-targeting bispecific molecules with 10 different CD3 affinities were tested for CD3 binding by ELISA. These data are representative of multiple experiments with subsets of molecules. (B) T cell activation by ULBP2-targeting bispecific molecules with 10 different CD3 affinities was assayed with HCT116 tumor cells and Jurkat-NFAT reporter cells. These data are representative of multiple experiments with subsets of molecules. (C) Schematic of a trispecific molecule with an antitumor-associated antigen (TAA) arm, an affinity-tuned anti-CD3 arm, effector null mutations, knobs-in-holes led heavy chain (HC) heterodimerization, and the CD2 agonist. Light chain pairing mutations are shown as electrostatic interactions. (D) Jurkat-NFAT reporter binding and activation assays for HER2-targeting bispecific and trispecific molecules. The bispecific molecules are shown in lighter colors and dotted lines while the trispecific molecules with integrated CD2 are shown as matching darker colors for each CD3 affinity (CD3, CD3.2, CD3.6, and CD3.8) and solid lines. NFAT-activation was assayed with HER2+ HCC1954 tumor cells or without (−tumor cells) like in B. (E) Jurkat-NFAT reporter binding and activation assays for ULBP2-targeting bispecific and trispecific molecules. The bispecific molecules are shown in lighter colors and dotted lines while the trispecific molecules with integrated CD2 are shown as matching darker colors for each CD3 affinity (CD3, CD3.2, CD3.6, and CD3.8) and solid lines. NFAT-activation was assayed with ULBP2+ COR-L105 tumor cells or without (−tumor cells) like in D and B. These data are representative of multiple experiments with subsets of molecules. (F) PBMCs were incubated with biologics and positive control (SEB) and negative control (IgG1) for 24 h and then cytokines were measured. All biologics are at approximately 300 nM while SEB is at 1 μg/mL. Each point (different symbol) corresponds to a different human PBMC donor for a n = 3. Statistical tests between biologics were carried out using Friedman’s test. These data are representative of multiple experiments with subsets of molecules.

Fig. 3.

CD3.6 affinity is the optimal CD3 affinity for a trispecific molecule with integrated CD2 costimulation. (A) Tumor killing data from T cell–dependenT cell cytotoxicity assays of 3 d with either HER2-targeting (Left) or ULBP2-targeting (Right) trispecific molecules incubated with HCC1954 or COR-L105 tumor cells (E:T ratio of 15:1 PBMCs to tumor cells), respectively. Statistical significance was found between the groups of TAA × CD3 × CD2 and TAA × CD3.2 × CD2 and TAA × CD3.6 × CD2 and TAA × CD3.8 × CD2 molecules for both targets. These data are an average of three different human PBMC donors and representative of three or more similar experiments each with multiple donors. (B) EC90 for each molecule in A. was pinpointed and then the cytokine release for IL-2, IL-6, IFNγ, and TNFα was calculated. Each set of molecules, either HER2-targeting (Left) or ULBP2-targeting (Right) was normalized to TAA × CD3 × CD2. Statistical analysis was done using ANOVA with Holm–Šídák’s multiple comparisons test. (C) HER2-targeting molecules were assessed in target-dependent and target-independent T cell activation (CD25) or cytokine release (TNFα, IL-6). (D) ULBP2-targeting molecules were assessed in target-dependent and target-independent T cell activation (CD25) or cytokine release (TNFα, IL-6). Fold changes between target-dependent and target-independent T cell activation based on EC50s from the curve fitting are marked on each graph. If the curve does not have an EC50, the highest point is used as a minimum for the fold change and is depicted using “>.” Each point is an average of three different human PBMC donors. The data are a compilation of data in Figs. 2F and 3A and SI Appendix, Fig. S4 A and B.

Fig. 4.

Optimal CD3.6 affinity as a platform for a tumor targeting trispecific molecule with integrated CD2 costimulation and its in vivo validation. (A) ULBP2 × CD3.6 and ULBP2 × CD3.6 × CD2 were evaluated in an in vitro TDCC assay with COR-L105 tumor cells and human PBMCs at an E:T ratio of 15:1 and a 5-d readout. The data are an average of eight different human PBMC donors and are representative of at least three similar experiments. (B) B7-H4 × CD3.6 and B7-H4 × CD3.6 × CD2 were evaluated in a TDCC assay with MDA-MB-468 tumor cells and human PBMCs at an E:T ratio of 15:1 and a 4-d readout. The data are an average of three different human PBMC donors and are representative of at least three similar experiments. (C) CD20 × CD3.6 and CD20 × CD3.6 × CD2 were evaluated in a TDCC assay with Raji tumor cells and human PBMCs at an E:T ratio of 15:1 and a 3-d readout. The data are an average of duplicates of one human PBMC donor and are representative of at least three similar experiments across multiple human PBMC donors. (D) COR-L105 tumor cells were coengrafted with expanded T cells into NSG mice and dosed twice a week for 3 wk with ULBP2-targeting bi- and tri-specific molecules with five different CD3 affinities. Tumor volume over the course of the experiment is shown and the number of tumor free mice (TFM) are noted for each group. Statistical P-values are shown for each group on day 43; also see SI Appendix, Fig. S10B for full statistical analysis. (E) COR-L105 tumor cells were coengrafted with expanded T cells into NSG mice and dosed twice a week for 3 wk with ULBP2 × CD3.6 × CD2 at three different concentrations. Tumor volume over the course of the experiment is shown with the legend containing statistical P-values for area under the curve through day 45 using a t test to Vehicle (PBS) condition.

Fig. 5.

A schematic of the mechanism of action of an EVOLVE platform trispecific TCE with integrated CD2 costimulation. Target-dependent activation of the molecule relies on tumor antigen targeting (red), while Fc mutations eliminate ADCC, antibody-dependenT cellular phagocytosis (ADCP), and complement-dependent cytotoxicity (CDC) activity of the molecule (purple). The molecule provides the T cell with signal 1 by affinity-tuned CD3 engagement (blue) and signal 2 by integrated CD2 costimulation (green) all to decrease T cell exhaustion and limit CRS risk. Finally, with both signal 1 and signal 2, the T cell is activated and produces cytokines (signal 3; orange) that amplify the immune response of nearby T cells. Created with BioRender.com.