Structure-guided engineering of immunotherapies targeting TRBC1 and TRBC2 in T cell malignancies

Abstract

Peripheral T cell lymphomas are typically aggressive with a poor prognosis. Unlike other hematologic malignancies, the lack of target antigens to discriminate healthy from malignant cells limits the efficacy of immunotherapeutic approaches. The T cell receptor expresses one of two highly homologous chains [T cell receptor β-chain constant (TRBC) domains 1 and 2] in a mutually exclusive manner, making it a promising target. Here we demonstrate specificity redirection by rational design using structure-guided computational biology to generate a TRBC2-specific antibody (KFN), complementing the antibody previously described by our laboratory with unique TRBC1 specificity (Jovi-1) in targeting broader spectrum of T cell malignancies clonally expressing either of the two chains. This permits generation of paired reagents (chimeric antigen receptor-T cells) specific for TRBC1 and TRBC2, with preclinical evidence to support their efficacy in T cell malignancies.

Fig. 1. Structural modeling and interaction between TCR-targeting antibodies and TRBC1/2.

a Homology between human TRBC1 and TRBC2. Numbering according to constant region of proteins (TRBC1, UniProtKB - P01850 and TRBC2, UniProtKB - A0A5B9) Corresponding positions of amino acids from the β chain of the A6 TCR used in this study are also shown. b Superimposition of HuJovi-1 Fab-TCR complex on TCR CD3 complex structure showing how specificity for TRBC is mediated in the context of the CD3 sheath. Right: Close-up of the interface between TCR complex and HuJovi-1 Fab. The Fab heavy and light chains are shown in green and gray, respectively. Despite proximity to CD3ε (yellow), neither the heavy nor the light chains form appropriate shape complementarity to interact with the CD3ε subunit of the TCR complex. c Molecular interface of the interaction between HuJovi-1 and TRBC1 (PDB ID 7AMP). Three key amino acids (Thr28, Tyr32, and Tyr98) within HuJovi-1 drive the specificity for TRBC1 (all antibody aa positions are referring to Kabat numbering scheme). Thr28 of HuJovi-1 mediates contact with Lys119, while Tyr32 mediates contact with Asn119. Tyr98 lies across the binding pocket and forms interactions with both Asn119 and Lys120. d Interaction between HuJovi-1 and TRBC2 (PDB ID 7AMQ). Inversion of Lys and Asn at positions 119 and 120 of the TCR β-chain removes the interacting partners of Thr28 and Ty32 in the antibody.

Fig. 2. CDR mutagenesis of HuJovi-1 for TRBC2 specificity.

a Molecular model of HuJovi-1 (Lys28 mutant; Mut1) in the context of TRBC2, suggesting that this residue can form a hydrogen bond with Lys119 in TRBC2. b ELISA binding of Thr28 mutant (pink) and HuJovi-1 (blue) to TRBC1 (triangles) and TRBC2 (squares). c Molecular model of HuJovi-1 (Lys28/Phe32; Mut2) in the context of TRBC2. d ELISA binding of Lys28/Phe32 Mut2 (green) and HuJovi-1 (blue) to TRBC1 (triangles) and TRBC2 (squares). e Molecular model of HuJovi-1 (Lys28/Phe32/Asn96; Mut3 – KFN) (PDB ID 7AMS). f ELISA binding of KFN (orange) and HuJovi-1 (blue) to TRBC1 (triangles) and TRBC2 (squares). d–f n = 3 technical replicates. Data shown as mean ± SD with non-linear fit, sigmoidal 4PL model. g Generation and interrogation of a phage display library based on HuJovi-1 structural analysis. Top: Amino acid distribution plot of residues in HuJovi-1 CDRs that were randomized to generate the library. Bottom: Frequency distribution of amino acids of 189 unique TRBC2-specific clones that were obtained from the library after selection. h Affinity ELISA showing the 17 highest affinity TRBC2 binding clones obtained after selection. (all antibody aa positions are referring to Kabat numbering scheme). Source data are provided as a Source Data file.

Fig. 3. Biophysical characterization of KFN antibody compared with HuJovi-1.

a Interface of KFN-TRBC2 from solved crystal structure (PDB ID 7AMS). The Fab heavy chain is colored in green while the TCR β-chain is in pink. The electron density is shown as a chicken wire net. Panels show the region around the Lys28 residue (left), Phe32 residue (center), and the Asn96 residue (right). b Left: Dot plot of a representative staining of healthy donor PBMC with HuJovi-1 and KFN IgG antibodies. Right: Distribution of TRBC1⁺ and TRBC2⁺ PBMC in four healthy donors (n = 4 biologically independent samples), as determined by HuJovi-1/KFN IgG staining. Data shown as mean ± SD. c Staining of TCR⁺ cell lines with HuJovi-1 (left) and KFN (right). d Staining of HPB-ALL TRBC1, HPB-ALL TRBC2, or HPB-ALL TCR KO cell lines with a titration of HuJovi-1 (blue) and KFN (orange) IgG. n = 4 technical replicates. Data shown as mean ± SD with non-linear fit, sigmoidal 4PL model. e Differential scanning calorimetry demonstrating similar stabilities of HuJovi-1 (blue) and KFN (orange) scFv-Fc antibodies. n = 2 technical replicates. Data shown as mean ± SD. f Kinetic profile of KFN scFv-Fc on recombinant TRBC1 (blue) and TRBC2 (orange). Source data are provided as a Source Data file.

Fig. 4. Functional characterization of HuJovi-1 and KFN CAR.

Schematic of anti-TRBC1 (a) and anti-TRBC2 (b) CAR architectures. Flow cytometry-based killing of Jurkat TRBC1, Jurkat TRBC2, or Jurkat TCR KO cells by HuJovi-1 (c) and KFN (d) CAR-T cells at 1:8 E:T ratio, 72 h; donor n = 9 (n = 8 for Jurkat TRBC1 in (c) biologically independent samples. *p < 0.05, ***p < 0.001, ****p < 0.0001 by two-way ANOVA and Dunnett’s test for multiple comparisons versus aCD19 CAR. Data shown as mean ± SD. IFN-γ and IL-2 secretion by HuJovi-1 (e) and KFN (f) –based CARs against Jurkat TRBC1, TRBC2 and TCR KO. Donor n = 6 (n = 9 for Jurkat TRBC1) biologically independent samples for TRBC1 CAR, and n = 3 (n = 6 for Jurkat TRBC2) biologically independent samples for TRBC2 CAR. *p < 0.05, **p < 0.01, ****p < 0.0001 by two-way ANOVA and Dunnett’s test for multiple comparisons versus aCD19 CAR. Data shown as mean ± SD. Sorted healthy donor T cells (TRBC1⁺ and TRBC2⁺) forward (left) and reverse (right) killing for HuJovi-1 (g) and KFN (h) –based CARs at 4:1, 1:1 and 1:4 E:T ratios, 72 h. Donor n = 4 biologically independent samples, *p < 0.05, **p < 0.01, two-way ANOVA and Sidak’s post-test for multiple comparisons TRBC1 versus TRBC2. Data shown as mean ± SD. i FACS-based killing of primary T-PLL tumor samples (TRBC1 n = 3, TRBC2 n = 3) at 4:1, 1:1 and 1:4 E:T ratios, 72 h. PBMC donor n = 8 (T-PLL TRBC1⁺) and n = 7 (T-PLL TRBC2⁺) biologically independent samples for HuJovi-1 CAR and n = 8 (T-PLL TRBC1⁺) and n = 11 (T-PLL TRBC2⁺) biologically independent samples for KFN CAR. **p > 0.01, ***p < 0.001, ****P < 0.0001, two-way ANOVA and Sidak’s post-test for multiple comparisons versus aCD19 CAR. T-PLL = T cell prolymphocytic leukemia. Data shown as min to max box and whiskers. Source data and exact p values are provided as a Source Data file.

Fig. 5. In vivo characterization of HuJovi-1 CAR and KFN CAR.

a Schematic of NSG model with 2.5e6 HPB-ALL cells/animal (n = 6/group). b Median total flux radiance of HPB-ALL TRBC2 tumor burden (from a). Two-way ANOVA with Dunnett’s post test against KFN CAR, ***p < 0.001, ****p < 0.0001. Data shown as median with 95% CI. c Kaplan-Meier survival curve (cut-off 1e9 p/s/cm²/sr) (from a). Log-rank test ***p < 0.001. d (from left) HPB-ALL count in CAR treated mice; % of TRBC2 expressing cells in residual HPB-ALL population; human T cell count (CD3⁺) in treated mice; % RQR8⁺ (CAR) T cells in CD3⁺ population. One-way ANOVA with Dunnett’s post-test. *P < 0.05, **P < 0.01, ****P < 0.0001. Data shown as mean ± SD. e Schematic of NSG model with 5e6 Jurkat TRBC1 or 1e6 Jurkat TRBC2 cells/animal (n = 6/group). f (left) Median total radiance of Jurkat TRBC1 tumor burden. Two-way ANOVA with Dunnett’s post-test against HuJovi-1 CAR, **p < 0.01. Data shown as median with 95% CI. (right) Kaplan-Meier survival curve (cut-off 1e9 p/s/cm²/sr). # 1 mouse from aCD19 CAR at d34 and remaining mice from the non-transduced group at d55, sacrificed due to xenogeneic graft versus host disease. Log-rank test **p < 0.01. g (left) Median total radiance of Jurkat TRBC2 tumor burden. Two-way ANOVA with Dunnett’s post-test against KFN CAR, ****p < 0.0001. Data shown as median with 95% CI. (right) Kaplan-Meier survival curve (cut-off 1e9 p/s/cm²/sr). Log-rank test ***p < 0.001. Source data and exact p values are provided as a Source Data file.