Tumor suppression of novel anti-PD-1 antibodies mediated through CD28 costimulatory pathway

Abstract

Classical antagonistic antibodies (Abs) targeting PD-1, such as pembrolizumab and nivolumab, act through blockade of the PD-1-PDL-1 interaction. Here, we have identified novel antagonistic anti-PD-1 Abs not blocking the PD-1-PDL-1 interaction. The nonblocking Abs recognize epitopes on PD-1 located on the opposing face of the PDL-1 interaction and overlap with a newly identified evolutionarily conserved patch. These nonblocking Abs act predominantly through the CD28 coreceptor. Importantly, a combination of blocking and nonblocking Abs synergize in the functional recovery of antigen-specific exhausted CD8 T cells. Interestingly, nonblocking anti-PD-1 Abs have equivalent antitumor activity compared with blocker Abs in two mouse tumor models, and combination therapy using both classes of Abs enhanced tumor suppression in the mouse immunogenic tumor model. The identification of the novel nonblocker anti-PD-1 Abs and their synergy with classical blocker Abs may be instrumental in potentiating immunotherapy strategies and antitumor activity.

Figure 1.

Anti–PD-1 Abs significantly enhance proliferation of Ag-specific exhausted CD8 T cells. (A) Recovery of proliferation in HIV-specific CD8 T cells following stimulation with an HIV-derived peptide. Results from a representative experiment are shown and expressed as the percentage of CFSE-low CD8 T cells. 8–10 replicates were performed for each experimental condition. (B) Cumulative results from multiple CFSE experiments (n = 2–6) are shown for the 10 anti–PD-1 Abs identified with antagonistic properties similar to pembrolizumab. Results have been generated assessing the proliferation of CD8 T cells specific to three HIV-derived peptides (FLGKIWPSYK restricted by A*0201 and RLRPGGKKK or RMRGAHTNDVK restricted by A*0301) in patients B08 and B09. For comparative purposes across multiple assays and the different anti–PD-1 Abs, the level of proliferation in the pembrolizumab-treated samples was set as a 100% reference. Pembrolizumab was used as a positive control, while untreated (Neg) or mouse IgG1 isotype control Ab was used as a negative control in each experiment. Graphs show the mean ± SD. ****, P < 0.0001 for all anti–PD-1 Abs relative to the IgG1 control (unpaired t test with Welch’s correction).

Figure 2.

Epitope mapping and structural studies of the prioritized anti–PD-1 Ab clones. (A) Ability of anti–PD-1 Abs to block the PD-1–PDL-1 interaction in a Luminex biochemical assay. PD-1–coated beads were incubated in the presence or absence of a competitor anti–PD-1 Ab, and then beads were stained with different concentrations of biotin-labeled PDL-1 protein. Data (n = 2) are mean ± SD. MFI, mean fluorescence intensity. (B) Potency of anti–PD-1 Abs in blocking the PD-1–PDL-1 interaction in a Luminex biochemical assay. PD-1–coated beads were incubated with a fixed concentration of PDL-1 equivalent to the half-maximal inhibitory concentration value for the PD-1–PDL-1 interaction in this assay. The PD-1–PDL-1 complex, bound at 50% in equilibrium, was then treated with increasing concentrations of anti–PD-1 Ab to determine if they were capable of completely disrupting the PD-1–PDL-1 interaction with pembrolizumab used as a positive blocking Ab control. (C) Epitope mapping by site directed mutagenesis of PD-1. Defined epitopes were identified for anti–PD-1 Abs that were either blocking or nonblocking of the PD-1–PDL-1 interaction using HeLa cells transfected with expression vectors encoding PD-1 with substitutions at solvent accessible residues. Amino acid substitutions in PD-1 are indicated in blue lettering above each histogram. Representative data are shown for n = 3 experiments. (D) Ab competitive binding studies for cell-surface PD-1. Jurkat PD-1 cells were incubated with excess of 137F2, 135C12, or 136B4 mouse Abs and then stained with a minimal concentration of the indicated humanized anti–PD-1 Abs (n = 3). (E) hPD-1 and NB01a Fab (humanized version of the mouse 135C12 Ab) complexes were purified by size-exclusion chromatography and crystallized. Crystals diffracted to 2.2 Å resolution, and the structure was solved by molecular replacement. The structure reveals that the binding site of NB01a is adjacent to residues in purple involved in the PD-1 interaction with either PDL-1 or PDL-2. The CC′ loop (residues 70–74) of hPDL-1 is disordered and indicated as a dashed line. Loops connecting β strands BC (57–63), C′D (84–92), and FG (127–133) were also disordered. Strands are named following the canonical designation. Cα superpositioning of the hPD-1 present in the NB01a Fab and hPDL-1 (PDB accession no. 4ZQK) complexes show that NB01a Fab and PDL-1 bind distinct nonoverlapping sites on PD-1. (F) Mapping of variable residues between human PD-1 and monkey, dog, horse, mouse, and rat PD-1 revealed an evolutionarily conserved patch (P1) on the opposite face of PD-1 from the PDL-1 or PDL-2 interaction site (yellow, orange, and light green colored residues). The P1 patch overlaps with the binding epitopes for the 135C12/NB01 (orange residues) and 136B4 (light green residues) antagonistic Abs that are nonblocking of PD-1–PDL-1. Residues N49, N58, N74, and N116 that are predicted N-linked glycosylation sites are show in brown on the PD-1 model to be excluded from the P1 patch, 135C12/NB01, and 136B4 Ab binding epitopes. (G) Cα superpositioning of hPDL-1 coordinates of the NB01a complex with pembrolizumab (PDB accession no. 5GGS) and the nivolumab (PDB accession no. 5GGR) confirms that NB01a Fab binding to PD-1 does not interfere with the binding of either pembrolizumab or nivolumab anti–PD-1 Abs. hPDL-1 is shown as a ribbon diagram in E and F, with the hPDL-1 binding surface (PDB accession no. 4ZQK) colored in purple in G.

Figure 3.

Blocking and nonblocking anti–PD-1 Ab combinations synergize in recovering both the proliferation and functional activity of exhausted Ag-specific CD8 T cells. (A) Cumulative results (three to six experiments) of the recovery of the proliferation of HIV-specific CD8 T cells after treatment with single and/or the combination of blocking and/or nonblocking anti–PD-1 Abs. Results are expressed as the percentage of CFSE-low CD8 T cells, and 8–10 replicates were performed for each experimental condition. (B) Recovery of T cell functionality evaluated by measuring cytokine levels of IFNγ, IL-2, TNFα, and IL-10 in the cell medium following an Ag-specific CD8 T cell stimulation. Cumulative results are shown using the PBLs from six to eight different viremic HIV-positive donors. Untreated samples (Neg) were used as a negative control in each experiment. Data represent mean ± SD. *, P < 0.036; **, P < 0.0079; ***, P < 0.0009; ****, P < 0.0001 (unpaired t test with Welch’s correction).

Figure 4.

Nonblocking anti–PD-1 Abs restore Ca²⁺ flux and AKT signaling to exhausted T cells. (A) Phospho-flow signaling studies performed with PD-1–expressing functionally exhausted T cells. Intracellular staining of phosphoproteins important to T cell signaling showed increased phosphorylation upon stimulation with anti-CD3/CD28 Abs. FACS histogram profiles for memory CD4 T cells show that a PDL-1 Fc fusion protein suppressed phosphorylation of AKT pT308, AKT pS473, and PDK1 pS242. Preincubation of cells with NB01b nonblocking or pembrolizumab blocking anti–PD-1 Abs significantly relieved PDL-1–mediated suppression of these phosphosignaling proteins at 5 and 15 min after stimulation. Data are presented as mean ± SEM for five to eight individual experiments. ns, not significant; *, P < 0.05; **, P < 0.0064; ****, P < 0.0001. (B) Exhausted T cells had reduced Ca²⁺ mobilization when stimulated with anti-CD3/CD28 in the presence of PDL-1 Fc fusion protein. (C) This suppression was restored with NB01b, pembrolizumab, or nivolumab. Synergistic increase in Ca²⁺ release was observed in exhausted T cells stimulated with anti-CD3/CD28 + PDL-1 when coincubated with NB01b and either pembrolizumab (B) or nivolumab (C). Representative data are shown from three independent experiments, and clinical Ab preparation of pembrolizumab was used for these studies. FDR, false discovery rate; MFI, mean fluorescence intensity.

Figure 5.

Nonblocking anti–PD-1 Abs act primarily through the CD28 costimulatory receptor–associated pathway. (A) IPs for two representative experiments are shown where the PD-1 receptor and associated protein complex were immunoprecipitated with NB01b, pembrolizumab, or a combination of NB01b and pembrolizumab in nonstimulated and anti-CD3/CD28 Ab–stimulated Jurkat PD-1 cells. In stimulated cells, pembrolizumab coprecipitated high levels of PD-1, CD28, SHP-2, PI3K, and the phosphorylated form of Src (p-Src) with the PD-1 complex. (B) NB01b pulled down equivalent levels of PD-1, SHP-2, and p-Src, but significantly reduced levels of CD28 and PI3K were observed in three or four separate experiments. AKT was only weakly pulled down in the PD-1 complex (A; right blot); however, there was a trend toward higher levels of AKT being immunoprecipitated with pembrolizumab compared with NB01b. (C) Control IPs performed by preincubating cells with either biotinylated pembrolizumab or NB01b before T cell stimulation shows that PD-1 pulled down with streptavidin (S) does not constitutively form a complex with CD28. Similarly, IPs performed with anti-CD3/CD28 coated beads show a strong pull-down of CD28 without detectable levels of associated PD-1. (D) The nonblocking anti–PD-1 Ab NB01b showed weak to no activation of the NFAT promoter in a Jurkat PD-1 luciferase reporter cell line when stimulated with 293T cells coexpressing a TCR activator and the PDL-1 receptor. Pretreatment of Jurkat PD-1 cells with the blocking anti–PD-1 Ab pembrolizumab strongly promoted NFAT activation. In contrast, both NB01b and pembrolizumab relieved PD-1–mediated suppression of NF-κB activation. Graphs show the mean ± SD and are representative of two independent experiments. *, P < 0.045; ***, P < 0.001; ****, P < 0.0001 (unpaired t test with Welch’s correction). RLU, relative luminescence units.

Figure 6.

Enhancement of tumor clearance by the combination of blocking and nonblocking anti–PD-1 Abs in the PD-1 HuGEMM in vivo MC38 tumor model. (A) Experimental scheme. (B–E) Mice successfully engrafted with the MC38 tumor cell line were treated twice weekly with PBS control (B), NB01b (C), pembrolizumab or nivolumab (D), or a combination of NB01b with either pembrolizumab or nivolumab (E), with tumor volumes measured in parallel. A collective analysis of three separate studies with n = 9–10 mice per arm per study is presented in A–D. P values determined by pairwise comparison using a mixed-effect linear model showed that all Ab arms of the study had reduced tumor growth compared with the vehicle control arm. (F) Suppression of tumor volume in mice engrafted with PDL-1–high MC38 cell line with either single or the combination of blocking and nonblocking anti–PD-1 Abs. Blocking (pembrolizumab or nivolumab) and nonblocking (NB01b) anti–PD-1 Ab monotherapies exerted equivalent suppression of tumor growth and mean percent tumor inhibition. Modeling of the cubic root transformed tumor volume in three separate studies as a function of time demonstrated a statistically significant reduction in tumor volume over time for the anti–PD-1 combination therapy relative to anti–PD-1 monotherapies. (G) Combination anti–PD-1 Ab therapy significantly enhanced the proportion of mice that controlled tumor growth (9 out of 19 mice) relative to Ab monotherapies (13 out of 59 mice). Survival analysis for a study performed using NB01b and nivolumab, with statistical differences determined using the log-rank test. Graphs show the mean ± SEM, unless otherwise indicated. *, P = 0.0223; **, P < 0.008; ***, P = 0.0005; ****, P < 0.0001.

Figure 7.

Blocking and nonblocking anti–PD-1 Abs suppress tumor growth and prolong survival of mice implanted with B16F10 cells in the PD-1 HuGEMM in vivo tumor model. (A) Experimental scheme. (B–E) Mice successfully engrafted with the poorly immunogenic B16F10 tumor cell line were treated twice weekly with PBS control (B), NB01b (C), pembrolizumab (D), or a combination of NB01b and pembrolizumab (E), with tumor volumes measured in parallel in the treatment (n = 10 mice) and control (vehicle; n = 10 mice) groups. P values determined by pairwise comparison using a mixed-effect linear model showed that blocking (pembrolizumab), nonblocking (NB01b), and combination anti–PD-1 Ab therapies had reduced tumor volume growth relative to the vehicle control. (F) Survival analysis with statistical difference determined using the log-rank test. Graphs show the mean ± SEM. **, P = 0.008; ****, P < 0.0001.