TRBC1-targeting antibody-drug conjugates for the treatment of T cell cancers

Abstract

Antibody and chimeric antigen receptor (CAR) T cell-mediated targeted therapies have improved survival in patients with solid and haematologic malignancies^1-9. Adults with T cell leukaemias and lymphomas, collectively called T cell cancers, have short survival^10,11 and lack such targeted therapies. Thus, T cell cancers particularly warrant the development of CAR T cells and antibodies to improve patient outcomes. Preclinical studies showed that targeting T cell receptor β-chain constant region 1 (TRBC1) can kill cancerous T cells while preserving sufficient healthy T cells to maintain immunity¹², making TRBC1 an attractive target to treat T cell cancers. However, the first-in-human clinical trial of anti-TRBC1 CAR T cells reported a low response rate and unexplained loss of anti-TRBC1 CAR T cells^13,14. Here we demonstrate that CAR T cells are lost due to killing by the patient's normal T cells, reducing their efficacy. To circumvent this issue, we developed an antibody-drug conjugate that could kill TRBC1⁺ cancer cells in vitro and cure human T cell cancers in mouse models. The anti-TRBC1 antibody-drug conjugate may provide an optimal format for TRBC1 targeting and produce superior responses in patients with T cell cancers.

Extended Data Fig. 1 |. Generation and testing of anti-TRBC1 CAR T cells.

a, b, Illustration depicting the anti-TRBC1 CAR construct c, CAR T cells were stained with anti-mouse scFv-PE antibody or anti-NGFR-APC antibody followed by analysis using flow cytometry. Control T cells indicate staining in unedited T cells. d, The aggregate data of the experiment is shown in Fig. 1a. CAR T cells were incubated with cancer cells (SUP-T1 or H9 or Jurkat cells) for 48 h. The percentage of surviving cancer cells is shown in the bar graphs. Bar graphs represent mean ± standard error of mean using three technical replicates. Number of biological replicates, n = 2. e, The aggregate data of the experiment is shown in Fig. 1c. CAR T cells were incubated with normal T cells. Flow cytometry was used to assess NGFR and TRBC1 expression after 48 h. The percentage of surviving normal TRBC1⁺ or TRBC2⁺ T cells and CAR T cells are shown in the bar graphs. Bar graphs represent mean ± standard error of mean using three technical replicates. Number of biological replicates, n = 3. f, g, Anti-TRBC1 CAR T cells (2.5 × 10⁴) were incubated with 2.5 × 10⁴ (1:1) or 5 × 10⁴ (1:2) or 12.5 × 10⁴ (1:5) Jurkat cells, in the presence or absence of normal T cells. After 48 h, flow cytometry was used to assess GFP and NGFR expression. Numbers beside flow plots show the percentage of surviving cells in each condition (f) and aggregate data from three technical replicates are shown in (g). Bar graphs represent mean ± standard error of mean. Number of biological replicates, n = 3. In (d, e, g) p values obtained by one-way ANOVA with Šidák’s multiple comparison test. The diagrams in a and b were created using BioRender.

Extended Data Fig. 2 |. Synthesis and characterization of anti-TRBC1 ADCs.

a, Schematic of the anti-TRBC1 antibody conjugation to SG3249. b, Hydrophobic interaction chromatography of anti-TRBC1 antibody, SG3249 and anti-TRBC1-SG3249 ADC. Number of repeated experiments, n = 4. c, The deconvoluted mass spectra of the anti-TRBC1 antibody and the anti-TRBC1-SG3249 ADC. HC = heavy chain and LC = light chain. HC+1 and HC+2 indicate heavy chains conjugated with one or two molecules of SG3249. LC+1 indicates light chain conjugated with one molecule of SG3249. Number of repeated experiments, n = 2. d, Size exclusion chromatography (SEC) of anti-TRBC1-SG3249 and anti-TRBC1 antibody. SEC standards include: A, thyroglobulin (MW 670 kDa); B, gamma globulin (MW 158 kDa); C, ovalbumin (MW 44 kDa); D, myoglobin (MW 17 kDa); E, vitamin B12 (MW 1.35 kDa). e, Schematic representation of the anti-TRBC1 antibody conjugation to MC-VC-MMAE. f, Hydrophobic interaction chromatography analysis of anti-TRBC1 antibody, MC-VC-MMAE and anti-TRBC1-MMAE ADC. Number of repeated experiments, N = 2.

Extended Data Fig. 3 |. Anti-TRBC1-SG3249 stability and binding epitope.

a, Anti-TRBC1-SG3249 ADC was incubated in human serum at 7 μg/mL concentration for 0, 3, and 7 days at 37 °C. After incubation with human serum, anti-TRBC1-ADC (at 100 ng/mL) was added to TRBC1⁺ cells (Jurkat and H9) and TRBC1⁻ cells (HPB-ALL and SUP-T1). After 5 days, cancer cell viability was assessed using luminescence. Bar graphs represent mean ± standard error of mean using three technical replicates. Number of biological replicates n = 3. b, TRBC1 and TRBC2 amino acid sequence alignment. The distinct amino acid residues are highlighted in red. The anti-TRBC1 antibody binding epitope at position 3,4 are shown inside a box. c, Jurkat, Jurkat TCR-KO, Jurkat TRBC2⁺, HPBALL or HPB-ALL TRBC1⁺ cancer cell lines were stained with anti-TRBC1-PE antibody. The histograms of the anti-TRBC1-PE stain of the indicated cell lines are shown on the left. The nucleotide and amino acid sequences of the anti-TRBC1-antibody binding epitope are shown on the right.

Extended Data Fig. 4 |. Anti-TRBC1-SG3249 kills TRBC1⁺ cells in vivo.

a, b, Flow cytometry to assess Jurkat cells (CD3+, GFP+, top right quadrant) on day 205, from bone marrow of mice treated with anti-TRBC1-SG3249 in Fig. 5a. For positive control, 5 mice were injected with 1 × 10⁶ Jurkat cells followed flow cytometry from bone marrow on day 21 “Jurkat injected NSG mice no ADC”. Data from 5 mice in each group shown in (b). c, TapeStation gel image of PCR product from mouse cells (black arrowhead shows amplified mouse β2 microglobulin sequence, 343 base pairs) and Jurkat cells (grey arrowhead shows amplified IVISbrite Red-F-luc-GFP sequence in Jurkat cells, 199 base pairs). Negative control “neg control” indicates sample with no input DNA, positive control “pos control” indicates sample with input DNA from mouse EMT6 cells and Red-F-luc-GFP transduced Jurkat cells. Data from 5 mice in each group shown in (c). d, Weights of 5 mice in each group as means ± S.E.M from Fig. 5a. e, Representative sections of hematoxylin and eosin-stained skin and liver from the 5 mice treated with anti-TRBC1-SG3249 ADC as in Fig. 5a. Untreated mice used as control (NSG mice, no ADC). Number of biological replicates, n = 5. f, Bioluminescence imaging of H9-injected NSG mice used in experiment Fig. 5g. g, Weights of 5 mice in each group as means ± S.E.M from Fig. 5g. h, Timeline of in vivo experiment using NSG mice injected with Jurkat cells and normal human T cells. On day 12, mice received mIgG2a-SG3249 or anti-TRBC1-SG3249 ADC. i, j, Flow cytometry on day 20 to assess Jurkat cells (CD3+, GFP+, top right quadrant), normal human T cells (CD3+, GFP-, top left quadrant), and aggregate data from 3 mice shown in (f). In (b) p value by two-tailed Mann-Whitney test. In (j) p values by one-way ANOVA with Šidák’s multiple comparison test. The diagram in h was created using BioRender.

Extended Data Fig. 5 |. Anti-TRBC1-SG3249 kills TRBC1+ patient-derived xenografts.

a, Jurkat, Jurkat TCR-KO, and T cell cancer PDX samples were stained with anti-TCRαβ-APC and anti-TRBC1-PE antibodies. b, Timeline of in vivo experiment using NSG mice injected with PDX cells. On day 25, mice were intravenously injected with either mIgG2a-SG3249 or anti-TRBC1-SG3249 ADC. c, d, Flow cytometry on day 21 and day 33 to assess circulating PDX cells (gated on the top right CD3+, and TRBC1+ cells), and aggregate data from 4 mice are shown in (d). In (d) p value obtained by one-way ANOVA with Šidák’s multiple comparison test. The diagram in b was created using BioRender.

Extended Data Fig. 6 |. Chimeric anti-TRBC1-SG3249 ADC has comparable cytotoxicity.

a, Schematic representation of the mouse IgG2a anti-TRBC1 antibody and chimeric IgG1 anti-TRBC1 antibody. The mouse IgG2a heavy chain (HC) and kappa light chain (LC) constant regions were replaced with human IgG1 HC and human kappa LC constant regions. b, Hydrophobic interaction chromatography analysis of chimeric anti-TRBC1 antibody and chimeric anti-TRBC1-SG3249 ADC. Number of repeated experiments n = 2. c, H9 or Jurkat cells were incubated with indicated concentrations of anti-TRBC1-SG3249 ADC or chimeric anti-TRBC1-SG3249 ADC for 5 days. The H9 and Jurkat cells expressed luciferase and luminescence was used to assess cell viability. Data plotted as means ± standard error of mean from three technical replicates. The calculated IC₅₀ for anti-TRBC1-SG3249 and chimeric anti-TRBC1-SG3249 ADC is shown at the right of the graphs. Number of repeated experiments n = 3. The diagram in a was created using BioRender.

Fig. 1 |. Anti-TRBC1 CAR T cell activity is limited by normal T cell-mediated killing of the CAR T cells.

a, NGFR-expressing CAR T cells (anti-TRBC1 or anti-CD19) were incubated with GFP-expressing cancer cells (SUP-T1, H9, Jurkat) for 48 h. GFP (cancer cell) and NGFR (CAR T cells) expression was assessed using flow cytometry. The percentage of surviving cells is shown. n = 3 biological replicates. b, Illustration of anti-TRBC1 CAR T cells killing TRBC1⁺ cancer cells. Anti-TRBC1 scFv on anti-TRBC1 CAR T cells binds to TRBC1 on cancerous T cells, leading to CAR T cell activation and the killing of TRBC1⁺ cancer cells. c, CAR T cells were incubated with normal T cells for 48 h. n = 3 biological replicates. d, Illustration of bidirectional killing of anti-TRBC1 CAR T cells and TRBC1⁺ normal T cells. Anti-TRBC1-CAR T cell binding to TRBC1 on normal T cells leads to the killing of TRBC1⁺ normal T cells. TRBC1⁺ binding also leads to the activation of TRBC1⁺ normal T cells, leading to reciprocal killing of anti-TRBC1 CAR T cells by normal T cells. e, CAR T cells were incubated with normal T cells and H9 or Jurkat cells for 48 h. n = 3 biological replicates. f, Illustration of the killing events that occur when anti-TRBC1 CAR T cells are incubated with normal T cells and TRBC1⁺ T cell cancers. Bidirectional killing leads to the depletion of both anti-TRBC1 CAR T cells and TRBC1⁺ normal T cells. The incomplete killing of TRBC1⁺ T cell cancers leads to cancer progression. g, CAR T cells were incubated with SUP-T1, H9 or Jurkat cells in the presence or absence of normal T cells. Cancer cells express GFP and were quantified using live-cell imaging. The arrows indicate the timepoints of the addition of Jurkat cells. Data are mean ± s.e.m. of three technical replicates. n = 2 biological replicates. The diagrams in b, d and f were created using BioRender.

Fig. 2 |. Anti-TRBC1 antibody binding to the TRBC1⁺ TCR leads to antibody internalization into lysosomes.

a, Illustration of the internalization of anti-TRBC1 antibodies into T cells. Anti-TRBC1–pHrodo antibody binds to the TRBC1⁺ TCR leading to its endocytosis. The anti-TRBC1–pHrodo antibody emits red fluorescence after lysosome endosome fusion. b, Anti-TRBC1–pHrodo antibody was added to Jurkat or Jurkat TCR-KO cells and analysed using live-cell imaging. c, Quantification of fluorescence as red calibrated unit (RCU) over time after the addition of anti-TRCB1–pHrodo antibodies to the indicated cell lines. Data are mean ± s.e.m. of three technical replicates. For b and c, n = 3 biological replicates. d,e, H9 (d) and Jurkat (e) cells were incubated with anti-TRBC1 antibodies, and then analysed using confocal microscopy. LAMP1 antibodies and DAPI staining mark lysosomes and nuclei, respectively. Anti-mouse IgG2a–Alexa568 was used to detect the location of anti-TRBC1 antibodies. Scale bars, 200 μm for b; 10 μm for d,e. n = 2 biological replicates. The diagram in a was created using BioRender.

Fig. 3 |. The performance of anti-TRBC1–SG3249 ADC in vitro.

a, Jurkat cells or Jurkat TCR-KO cells were treated with anti-TRBC1 antibodies, anti-TRBC1–SG3249 ADC or SG3199 free drug at the indicated concentration. Cell growth was quantified as green calibrated unit (GCU) using live-cell imaging. Data are mean ± s.e.m. of three technical replicates. n = 3 biological replicates. b, T cell cancer cell lines were incubated with the indicated concentration of anti-TRBC1–SG3249 ADC for 5 days. The cancer cell viability was measured using the luciferase assay. The mIgG2a–SG3249 ADC was used as a negative control. The IC₅₀ for anti-TRBC1–SG3249 is indicated in the graphs. Data are mean ± s.e.m. of three technical replicates. c, H9 and Jurkat cells were treated with anti-TRBC1 antibodies or a sublethal dose of the anti-TRBC1–SG3249 ADC, followed by pH2Ax staining. The cells expressed cytosolic GFP, and DAPI staining marks the nuclei. Cells were imaged using confocal microscopy. Scale bars, 10 μm. n = 2 biological replicates. DIC, differential interference contrast.

Fig. 4 |. Anti-TRBC1–SG3249 ADC kills cancer cells in the presence of normal T cells.

a,b, T cell cancers were cultured with normal T cells in the presence of anti-TRBC1–SG3249 (25 ng ml⁻¹) or mIgG2a–SG3249 (as a negative control ADC). After 5 days, GFP and TRBC1 expression (a) and viability (b) were assessed using flow cytometry. The numbers on the plots indicate cells counted by flow cytometry in a representative experiment (a), with data from three human T cell donors (n = 3 biological replicates) shown in b. Data are mean ± s.e.m. c, The effect of anti-TRBC1–SG3249 ADC added to co-culture of T cell cancers and normal T cells. Anti-TRBC1 ADCs kill TRBC1-expressing cancer cells and normal T cells, while TRBC2⁺ normal T cells survive. d,e, Normal T cells were incubated with anti-TRBC1–SG3249 ADC or mIgG2a–SG3249 ADC. After 5 days, TRBC1⁺ and TRBC2⁺ cells (d) and viability (e) were assessed using flow cytometry. The numbers on the plots indicate cells counted by flow cytometry (d), with data from three human T cell donors (n = 3 biological replicates) shown in e. Data are mean ± s.e.m. f, Normal T cells from three donors were sorted to obtain TRBC1⁺ and TRBC2⁺ populations. TRBC1⁺ or TRBC2⁺ normal T cells were incubated with anti-TRBC1–SG3249 ADC for 5 days and then analysed using the MTS viability assay. n = 2 biological replicates. g,h, CD3⁺ lymphocytes were isolated from two patients with TRBC1-expressing CTCL and incubated with anti-TRBC1–SG3249 ADC or with mIgG2a–SG3249 ADC for 5 days. Flow cytometry analysis (g) and quantification (h) of the number of tumour cells (patient 1, TRBC1⁺ and C26⁻ cells; and patient 2, TRBC1⁺CD7⁻ cells) is shown. The numbers on the plots indicate the number of cells counted by flow cytometry (g), with data from three technical replicates shown in h. For b, e and h, P values were calculated using one-way analysis of variance (ANOVA) with Šidák’s multiple-comparison test. The diagram in c was created using BioRender.

Fig. 5 |. ADC activity in vivo.

a, The timeline of the in vivo experiment in NSG mice injected with Jurkat cells. b,c, NSG mice were intravenously (i.v.) injected with Jurkat cells expressing luciferase and GFP. On day 8, mice were intravenously injected with one of three different ADCs: mIgG2a–SG3249, anti-TRBC1–MMAE or anti-TRBC1–SG3249. Bioluminescence imaging was performed on the indicated days (b) and aggregate data are shown (c). d,e, Flow cytometry analysis of circulating Jurkat cells (CD3⁺GFP⁺, top right quadrant) on day 21 (d), and aggregate data from five mice (e). f, Kaplan–Meier survival curves of Jurkat-bearing NSG mice after various treatments, with five mice in each group. The median survival was as follows: 27 days (mIgG2a–SG3249), 38 days (anti-TRBC1–MMAE) and end-point not reached (undefined; anti-TRBC1–SG3249). Statistical analysis was performed using the log-rank Mantle–Cox test; P = 0.0034. g, The timeline of the in vivo experiment using NSG mice injected with H9 cells. h, NSG mice were intravenously injected with H9 cells expressing luciferase and GFP. On day 8, mice were intravenously injected with either mIgG2a–SG3249 or anti-TRBC1–SG3249 ADC. Bioluminescence imaging was performed on the indicated days. i,j, Flow cytometry analysis of circulating H9 cells (CD3⁺GFP⁺, top right quadrant) on day 20 (i), and aggregate data from 5 mice (j). k, Kaplan–Meier survival curves of H9-bearing NSG mice after various treatments, with five mice in each group. The median survival was as follows: 24 days (mIgG2a–SG3249) and not reached (undefined; anti-TRBC1–SG3249). Statistical analysis was performed using the log-rank Mantle–Cox test; P = 0.002. P values were calculated using one-way ANOVA with Šidák’s multiple-comparison test (e) and two-tailed Mann–Whitney tests (j). The diagrams in a and g were created using BioRender.