Cancer immune therapy with PD-1-dependent CD137 co-stimulation provides localized tumour killing without systemic toxicity

Abstract

Expression of the cell surface receptor CD137 has been shown to enhance anti-cancer T cell function via engagement with its natural ligand 4-1BBL. CD137 ligation with engineered ligands has emerged as a cancer immunotherapy strategy, yet clinical development of agonists has been hindered by either toxicity or limited efficacy. Here we show that a CD137/PD-1 bispecific antibody, IBI319, is able to overcome these limitations by coupling CD137 activation to PD-1-crosslinking. In CT26 and MC38 syngeneic mouse tumour models, IBI319 restricts T cell co-stimulation to PD-1-rich microenvironments, such as tumours and tumour-draining lymph nodes, hence systemic (liver) toxicity arising from generalised T cell activation is reduced. Besides limiting systemic T cell co-stimulation, the anti-PD-1 arm of IBI319 also exhibits checkpoint blockade functions, with an overall result of T and NK cell infiltration into tumours. Toxicology profiling in non-human primates shows that IBI319 is a well-tolerated molecule with IgG-like pharmacokinetic properties, thus a suitable candidate for further clinical development.

Fig. 1. IBI319 activates CD137 in a PD-1 dependent manner.

a Structure diagram of IBI319 on a human IgG1 backbone. IBI319 consists of a Fab region binding to PD-1 and CD137 and an Fc fragment with null effector function. b The dissociation constants (K_Ds) of IBI319, αPD-1, and αCD137 for human PD-1 and human CD137, respectively, determined via SPR (3 independent experiments; mean ± standard deviation (SD); n.a. not analysed). c CD137 ligand-blocking properties of IBI319, αCD137, in-house-generated urelumab and utomilumab via BLI. d PD-1 and CD137 expression on CD4 + and CD8 + T cells from a representative healthy donor before (black) and after activation by anti-CD3/CD28 antibody-coated beads (red) measured by flow cytometry (FACS). e Binding affinities of IBI319 and control antibodies for activated CD4 + T cells and CD8 + T cells measured by FACS. f Binding affinities of IBI319 and control antibodies for Jurkat-CD137 cells co-cultured with CHO-S or CHO-S-PD-1 cells determined by FACS and calculated from the CTV + CFSE + population. e, f two technical replicates of one representative experiment from three independent experiments; mean and SD.

Fig. 2. IBI319 blocks the PD-1/PD-L1 interaction and selectively enhances CD137 activity in a PD-1-dependent manner.

a PD-1/PD-L1 blocking activities of different antibodies presented as NFAT-mediated luciferase activity. Fold induction was calculated as the mean relative light units (RLU) of tested antibodies/mean RLU of the no antibody control. b CD137 agonist activities of different antibodies presented as NFκB-mediated luciferase activity. Jurkat-CD137-NFκB-Luc (Jurkat-CD137) cells were co-cultured with two times the number of CHO-S-PD-1 cells (left), without CHO-S-PD-1 cells (middle), and with different amounts of CHO-S-PD-1 cells (right). Fold induction was calculated as the mean RLU of tested antibodies/mean RLU of the no antibody control. c Three dimensional (3D) and 2 dimensional (2D) cell culture systems to test the cis- and trans-cell crosslinking of IBI319. (Left) The 3D and 2D cell culture models when treated with IBI319. (Right) The relative luciferase signals of the Jurkat-CD137-NFκB-Luc2P-PD-1 alone and Jurkat-CD137 + CHO-S-PD-1 groups were calculated with the equation: (mean RLU of tested antibodies in 3D culture/mean RLU of the control in 3D culture)/(mean RLU of tested antibodies in 2D culture/mean RLU of the no antibody control in 2D culture). d Relative IFN-γ production in an MLR assay. Antibodies were 4-fold serially diluted from 200 nM. e Relative IFN-γ production level in a T cell co-stimulation assay. Primary T cells were treated with 1 μg/mL CD3/CD28 activator, and antibodies were serially diluted 10-fold from 400 nM in the presence of CHO-S or CHO-S-PD-1 cells. f IFN-γ production level in a cytomegalovirus (CMV) T cell memory recall assay. PBMCs from a donor previously infected with CMV were co-cultured with a CMV peptide pool and antibodies at 50 nM for 4 days (data for three donors in Supplementary Fig. 2e). a and b show two technical replicates of one representative experiment from three independent experiments. mean and SD. c, d, e, f show individual values of three independent experiments and the mean ± SD. Statistics were performed in the indicated groups: two-way ANOVA with Tukey’s multiple comparisons test (b) and two-sided paired t-test (c, d, e, f); p values are indicated in the graphs.

Fig. 3. IBI319 shows strong antitumour efficacy while avoiding hepatotoxicity in mouse models.

a (Left) Tumour volumes of CT26-hPD-L1 tumours in BALB/c-hPD-1/hPD-L1/hCD137 mice treated with the indicated antibodies (20 mg/kg, i.p.) on days 11, 18 and 25 post-tumour implantation (n = 8 per group). The mice in the hIgG1 and αCD137 groups were euthanized on day 27 post-tumour implantation due to ethical issues. (Right) Spaghetti plots show the individual tumour volumes in each group. The numbers in the plots indicate tumour-free mice/all. b Tumour volumes of MC38 tumours in C57BL/6-hPD-1/hCD137 mice treated with the indicated antibodies on days 9, 16 and 23 (n = 7 per group). c and d Single-cell RNA sequencing of CD45 + cells in MC38 tumours after antibody treatments. c A thumbnail heatmap showing the expression of signature genes in each cell cluster (detailed heatmap in Supplementary Fig. 4b). d The percentages of different cell clusters in the αPD-1, IBI319 and αPD-1 + αCD137 groups. e Tumour volumes of MC38 tumours in C57BL/6-hPD-1/hCD137 mice treated with the indicated antibodies (3 + 3 mg/kg for the αPD-1/Hel + αCD137/Hel-Fc group and 3 mg/kg for all other groups) on days 9, 16 and 23 (n = 6 per group for the IgG1 and αPD-1/Hel + αCD137/Hel-Fc groups; n = 5 per group for all other groups). Hel: hen egg white lysozyme. f Livers from the mice in b were stained for CD45 and F4/80 via IHC. (Left) Representative images. (Right) Quantification of whole pathological digital slides via HALO software (n = 3 per group, mean and SD); images of all quantified groups can be found in Supplementary Fig. 5b. a, b and e Tumour volumes are presented as the mean, and the error bar represents the SEM. Statistics were performed in the indicated groups: two-way ANOVA with Tukey’s multiple comparisons test (a and b) and two-sided, un-paired t test (f); p values are indicated in the graphs.

Fig. 4. IBI319 is a tolerated molecule with an IgG-like pharmacokinetic (PK) profile suitable for further clinical development.

a Serum concentrations and receptor occupancies of IBI319 in cynomolgus monkeys following a single intravenous administration of 0.1 mg/kg, 1 mg/kg, or 10 mg/kg. n = 6 per group (three males and three females). conc. concentration, RO receptor occupancy, mean and SD. b Time-concentration curves of IBI319 following the first (day 1) and fourth (day 22) intravenous administrations of 1 mg/kg, 10 mg/kg, or 50 mg/kg IBI319 into cynomolgus monkeys. n = 10 per group (five males and five females); mean and SD.

Fig. 5. Working model of IBI319.

IBI319 activates CD137 in a manner restricted to PD-1-rich microenvironments, such as tumours and tumour-draining lymph nodes, limiting potential liver toxicity caused by the αCD137 and αPD-1 combination.