WHAT IS ALREADY KNOWN ON THIS TOPIC

Combining programmed cell death protein-1 (PD-1)/programmed death-ligand 1 (PD-L1) blockade with 4-1BB agonism has shown promise in reinvigorating antitumor T-cell responses in preclinical models. The novel combination therapy of acasunlimab (DuoBody-PD-L1×4-1BB), an investigational PD-L1 and 4-1BB-targeting bispecific antibody, with pembrolizumab has recently demonstrated promising clinical activity in checkpoint inhibitor (CPI)-relapsed/refractory metastatic non-small cell lung cancer (NSCLC). This study investigated the mechanisms of antitumor immunity conferred by this combination therapy.

WHAT THIS STUDY ADDS

This study offers key mechanistic insights into the complementary immune modulatory effects of acasunlimab-mediated conditional T-cell costimulation through 4-1BB combined with full PD-1 blockade. The combination amplifies the depth and duration of antitumor immune responses by promoting autocrine interleukin 2–CD25 signaling and inducing stem-like CD8+ tumor-infiltrating T cells with enhanced effector functions.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

These findings provide a robust mechanistic rationale for combining targeted 4-1BB costimulation in tumor and lymphoid tissues with full PD-1 blockade to amplify adaptive antitumor immunity, establishing the preclinical foundation for the clinical development of acasunlimab in combination with pembrolizumab. This combination approach is currently being evaluated in a pivotal Phase 3 study (ABBIL1TY NSCLC-06; NCT06635824) for patients with metastatic NSCLC who have failed prior CPI therapy.

Background

Programmed cell death protein-1 (PD-1)/programmed death-ligand 1 (PD-L1) checkpoint inhibitors (CPIs) have transformed the treatment paradigm and prognosis for several advanced solid tumor indications.1 However, clinical efficacy to CPI-containing therapy is observed only in a subset of patients, with the majority progressing after an initial response. This underscores the need for better patient selection strategies and combination therapies to improve patient outcomes. The combination of immune CPIs targeting complementary pathways has demonstrated improved antitumor activity compared with single-agent CPIs in the clinic.2 To unleash the full potential of adaptive immunity, it may be beneficial to combine blockade of inhibitory signals with simultaneous activation of costimulatory receptors to promote T-cell proliferation, survival, and effector functions, and to reinvigorate T cells that became exhausted and dysfunctional due to sustained antigen stimulation in the tumor microenvironment (TME).3–7 One such costimulatory receptor is 4-1BB that can induce expansion of T-cell clones, enhance memory differentiation, and improve functionality and survival of dysfunctional CD8+ T cells in the TME.8–10 4-1BB is a clinically validated target, with agonistic 4-1BB monospecific antibodies demonstrating promising initial biological activity, however with a limited therapeutic window due to dose-limiting hepatotoxicity.9 11 Therefore, next-generation 4-1BB agonists aim to broaden the therapeutic window by restricting 4-1BB activation to tumor and lymphoid tissues in a manner strictly dependent on cotargeting complementary immune checkpoint pathways.9

The potential for concurrent targeting of 4-1BB and PD-1 is supported by their coexpression on antigen-specific CD8+ T cells during different states of activation, including activated, memory and (pre)exhausted CD8+ T cells in the TME.5 12–16 Preclinical studies have shown that the combination of PD-1/PD-L1 (PD-(L)1)-blocking agents and 4-1BB agonists can reinvigorate the function of exhausted CD8+ tumor-infiltrating lymphocytes (TILs) and exert synergistic antitumor effects associated with potent tumor antigen-specific T-cell responses.14 17–20

PD-L1 and 4-1BB targeting bispecific antibodies, which combine conditional 4-1BB agonist activity with PD-(L)1 checkpoint blockade, have shown promise in enhancing antitumor activity, while allowing for an improved safety profile over earlier generation 4-1BB monoclonal antibodies.21–27 Acasunlimab (GEN1046/BNT311/DuoBody-PD-L1×4-1BB) is a PD-L1×4-1BB bispecific antibody that is in clinical development for the treatment of patients with solid tumors (NCT03917381, NCT04937153, NCT05117242, and NCT06635824). Due to its innovative design using the clinically validated DuoBody platform for the generation of bispecific antibodies in combination with an inert fragment crystallizable (Fc) region, the 4-1BB agonist activity of acasunlimab is strictly dependent on binding to PD-L1+ tumor cells and/or PD-L1+ immune cells, thus predominantly limiting the immune response to the site of the tumor and secondary lymphoid organs. At the same time, the PD-L1-specific arm of acasunlimab functions as a classical immune CPI by blocking the PD-1/PD-L1 axis irrespective of 4-1BB binding. Acasunlimab was shown to exert superior T-cell activation and T-cell-mediated cytotoxicity compared with clinically approved PD-L1 antibodies in vitro, and to promote expansion of tumor-reactive TILs ex vivo.25 In preclinical mouse tumor models, acasunlimab exerted potent antitumor activity, which was associated with enhanced intratumoral CD8+ T-cell infiltration.25 In the clinic, single-agent acasunlimab has shown a manageable safety profile and early clinical activity in heavily pretreated patients with advanced solid tumors, including patients who had progressed on prior CPI-containing therapy.25

Pharmacokinetic (PK)/pharmacodynamic modeling using clinical data from the dose escalation phase of the first-in-human study indicated that doses of acasunlimab achieving optimal 4-1BB activation result in partial PD-1/PD-L1 axis blockade.28 This suggests that blockade of the PD-1 pathway could be enhanced by combining acasunlimab with an anti-PD-1-blocking antibody, allowing for concurrent optimum 4-1BB stimulation and complete inhibition of the PD-1 pathway by fully disrupting its interactions with both PD-L1 and PD-L2.

In this study, we provide preclinical evidence that combining acasunlimab-induced conditional 4-1BB costimulation with complete blockade of the PD-1 pathway leads to improved depth and duration of the antitumor immune response through distinct and complementary immune modulatory effects.

MethodsHuman primary leukocytes

Details on the preparation of human primary lymphocytes and dendritic cells (DCs) used in the in vitro T-cell assays are described in the online supplemental materials and methods. Briefly, CD14+ monocytes were obtained from human healthy donor peripheral blood mononuclear cells (PBMCs) and differentiated into immature DCs (iDCs) using granulocyte macrophage-colony stimulating factor (BioLegend, 766106) and interleukin (IL)-4 (BioLegend, 766206), and then lipopolysaccharide-matured to mature DCs (mDCs). For the antigen-specific T-cell assays, PBMC-derived CD8+ T cells were electroporated to express a human Claudin-6 (CLDN6)-specific mouse T-cell receptor (TCR) and human PD-1, and iDCs were electroporated to express human CLDN6.

In vitro studiesMixed lymphocyte reaction with functional T cells

In the mixed lymphocyte reaction (MLR) assay (online supplemental figure 1A), purified CD8+ T cells were thawed and resuspended at 1×106 cells/mL in Roswell Park Memorial Institute (RPMI) 1640 complete medium (Thermo Fisher Scientific, A1049101) supplemented with 10% fetal bovine serum (FBS; Gibco, 16140071) and 10 ng/mL IL-2 (BioLegend, 589106) at 37°C O/N. CD8+ T cells were then co-cultured with allogeneic mDCs in 96-well plates (1:10 mDC:T-cell ratio) and incubated with a dose range of acasunlimab, including clinically relevant concentrations,25 28 either alone or combined with pembrolizumab, or with control antibodies in MLR assay medium at 37°C for 5 days. Single-agent pembrolizumab samples were tested with an antibody dose range, whereas in combination with the acasunlimab dose range, 1 µg/mL pembrolizumab was tested, which was previously determined as the maximal-effective concentration in this assay. Supernatants were harvested for cytokine analysis.

MLR with dysfunctional T cells

Dysfunctional T cells (Tdys) were prepared by stimulating 1×106 cells/mL purified healthy donor CD3+ T cells twice with anti-CD3/CD28 Dynabeads (Gibco, 11 161D) (1:1 bead-to-cell ratio) in MLR assay medium supplemented with 5% FBS and 10 ng/mL IL-2 at 37°C for a total of 120 hours (after an initial 72 hours, the beads were removed and fresh anti-CD3/CD28 beads were added for an additional 48 hours). The dysfunctional phenotype of the T cells was confirmed as described in the online supplemental materials and methods (online supplemental figure 2A–D). Dysfunctional CD3+ T cells were rested for 24 hours in RPMI 1640 medium supplemented with 10% FBS and 10 ng/mL IL-2 before co-culturing with allogeneic mDCs in 96-well plates (1:4 or 1:10 DC:T-cell ratio) and incubating with a dose range of acasunlimab, either alone or combined with 0.8 µg/mL pembrolizumab (previously determined as the maximal-effective concentration in this assay), or with control antibodies in MLR assay medium at 37°C for 5 days. Supernatants were harvested for cytokine analysis.

Antigen-specific T-cell proliferation assay

Electroporated human CD8+ T cells expressing the CLDN6-TCR and human PD-1 were labeled with carboxyfluorescein succinimidyl ester (CFSE) using the Vybrant CFDA SE Cell Tracer Kit (Life Technologies, V12883). 75,000 CFSE-labeled T cells were co-cultured with 7,500 autologous iDCs electroporated to express human CLDN6 (1:10 DC:T-cell ratio) in 96-well plates and incubated with a dose range of acasunlimab, either alone or combined with 0.8 µg/mL pembrolizumab (previously determined as the maximal-effective concentration in this assay), or with control antibodies in T-cell proliferation assay medium at 37°C for 4 days. CD8+ T-cell proliferation was calculated from CFSE dilution measured by flow cytometry. Generation peaks were automatically fitted using the proliferation modeling tool in FlowJo and used to calculate expansion index values. Supernatants were harvested for cytokine analysis (online supplemental figure 1B).

Antigen-specific T-cell-mediated cytotoxicity assay

MDA-MB-231_hCLDN6 cells, which endogenously express human PD-L1, were seeded (1.5×104 cells/well) in xCELLigence E-plates (Agilent, 05232368001) 1 day before adding electroporated CD8+ T cells expressing a CLDN6-TCR and human PD-1 at an effector to target cell (E:T) ratio of 3:1. The co-cultures of tumor cells and CD8+ T cells were incubated with acasunlimab fixed concentrations of 0.003 µg/mL (ie, predetermined EC20) or 0.0015 µg/mL (ie, <EC20), either alone or with 0.8 µg/mL pembrolizumab (predetermined maximal-effective concentration), or with control antibodies. Cells in the E-plates were cultured in the xCELLigence real-time cell analysis instrument (ACEA Biosciences) for 140–170 hours without disturbance, with impedance measurements at 2-hour intervals. Impedance measurement data were normalized to the time of co-culture start for each treatment condition and data were expressed relative to T cell-tumor cell co-cultures without antibody (set to 100%). Pooled real-time cell analysis over the assay period was performed using GraphPad Prism’s area under the curve (AUC) function (online supplemental figure 1C). Expression of CD107a and GZMB by the CD8+ T cells was also assessed as described in the online supplemental materials and methods.

In vivo studiesPK analysis of anti-mPD-L1×m4-1BB and mouse mPBPK/RO modeling

MC38 tumor-bearing mice were treated with 5 or 20 mg/kg anti-mouse PD-L1 × anti-mouse 4-1BB (anti-mPD-L1×m4-1BB) antibody twice weekly for three weeks (Q2W×3) as single agent or in combination with 10 mg/kg anti-mouse PD-1 (anti-mPD-1) antibody clone RMP1-14 (three mice per group) and blood samples were collected for PK analysis as described in the online supplemental materials and methods. A minimal physiologically-based PK model with receptor occupancy (mPBPK/RO) was developed by supplementing a published mPBPK model with a tumor compartment.29 Tumor distribution of the antibody was taken as previously described.30 T cells could interact with nearby tumor cells via both mPD-L1 on the tumor cells engaging mPD-1 on the T-cell membrane and via anti-mPD-L1×m4-1BB engagement of m4-1BB on the T cells and mPD-L1 on the tumor cells. Cells in the model were assumed to be static; no growth or death was considered, only the formation of complexes on their membranes. PK parameters (clearance and distribution rates) were estimated from data of anti-mPD-L1×m4-1BB (online supplemental figure 3) and anti-mPD-1.31 A typical tumor volume of 50 mm3 was used to initialize the model. T-cell infiltration and 4-1BB expression at baseline were taken from immunohistochemistry (IHC) analysis of phosphate-buffered saline (PBS)-treated control animals on day 7 under the assumption that the relative proportion would not change in these control animals between the start of treatment and day 7.

In vivo efficacy study in C57BL/6 mice implanted with B16F10, MB49, MC38 and Pan02 tumors

C57BL/6 mice were purchased from Vital River Laboratories Research Models and Services (Beijing, China) or GemPharmatech. Mice were subcutaneously (SC) injected in the right flank with B16F10 (2×105 cells/mouse), MB49 (1×106 cells/mouse), MC38 (1×106 cells/mouse), or Pan02 (3×106 cells/mouse) in 100 µL PBS. At randomization of the animals into treatment groups, the mean tumor volumes reached 39 mm3 for B16F10, 59 mm3 for MB49, 60 mm3 for MC38, and 90 mm3 for Pan02. Tumor-bearing mice were treated by repeated intraperitoneal (IP) injections twice weekly for three weeks (2QW×3) of 10 mg/kg anti-mPD-1 antibody clone RMP1-14, 5 mg/kg anti-mPD-L1×m4-1BB, the combination thereof, or with PBS (10 mice per group). Animals implanted with MC38 tumor cells that achieved complete tumor regression (CR) after treatment were rechallenged by SC injection of MC38 tumor cells on day 138 after the initial start of treatment. As a control group for the rechallenge analysis, a second cohort of treatment-naïve animals was inoculated with MC38 tumor cells.

In vivo efficacy study in hPD-1/hPD-L1/h4-1BB tKI mice implanted with MC38-hPD-L1 and MC38 tumors

C57BL/6-Pdcd1tm1(PDCD1)Bcgen Cd274tm1(CD274)Bcgen Tnfrsf9tm1(TNFRSF9)Bcgen/Bcgen mice engineered to express extracellular domains of human PD-1, PD-L1 and 4-1BB in the mouse PD-1, PD-L1 and 4-1BB gene loci (hPD-1/hPD-L1/h4-1BB) triple knock-in (tKI) mice were purchased from Beijing Biocytogen (130569). In this model, chimeric acasunlimab (chi-acasunlimab) was tested, which is a chimeric antibody containing the VH/VL sequences of acasunlimab in a mouse IgG2a backbone with Fc-inertness mutations. The experimental set-up was based on previous experience25 and dose optimization studies (data not shown). Mice were SC injected in the right flank with 1×106 MC38-hPD-L1 or wild type (WT) MC38 cells in 100 µL PBS. At randomization of the animals into treatment groups, the mean tumor volumes reached 128 mm3 for MC38-hPD-L1 and 61 mm3 for MC38. Tumor-bearing mice were treated 2QW×3 by IP injections with chi-acasunlimab (5 mg/kg or 10 mg/kg), pembrolizumab (10 mg/kg), the combination thereof, or matched isotype control antibodies mIgG2a-ctrl and IgG4 (eight mice per group). In the MC38-hPD-L1 tumor model, animals with CR after treatment were rechallenged by SC injection of 1×106 MC38-hPD-L1 tumor cells in the left flank on day 143 after the initial start of treatment. As a control group for the rechallenge analysis, a second cohort of treatment-naïve animals was inoculated with MC38-hPD-L1 tumor cells.

In vivo MoA studies using plasma and tissues from MC38 tumor-bearing C57BL/6 mice

For the in vivo mechanism of action (MoA) studies, C57BL/6 mice were injected with MC38 cells and treated as described for the efficacy study with the only difference that a 2QW dosing frequency was used. On day 7 and day 14 after the start of treatment, mice were euthanized and tumor tissue and/or tumor-draining lymph nodes (tdLNs) were collected for IHC, flow cytometry, TCR sequencing, and RNA sequencing as described in the online supplemental materials and methods.

Statistical analysesIn vitro studies

For the MLR and antigen-specific T-cell proliferation assays, adjusted p values for single-agent activity were calculated using one-way analysis of variance with Dunnett’s multiple comparisons test and Highest Single Agent (HSA) synergy analysis was performed using the SynergyFinder R package.32 For HSA synergy analysis, data were processed for each donor or donor pair separately and the measured value of each sample was normalized by subtracting the control values (no treatment control wells) and expressed as a percentage of the maximal value in the assay. Synergy was defined as the excess over the maximum single-agent response based on the HSA model stating that the expected combination effect equals the higher effect of individual drugs. HSA scores>10 are suggestive of synergy. Statistical analyses of the in vitro cytotoxicity studies were performed using a non-parametric, paired Friedman test with Dunn’s multiple comparisons test on percentages of CD107a+/GZMB+ cells or the AUC of normalized real-time analysis data.

In vivo studies

For the in vivo studies in mouse models, tumor growth rate was determined by performing a simple linear regression of the log-transformed tumor volumes. Kaplan-Meier curves of progression-free survival (PFS), defined as mice with tumor volumes<1000 mm3 for B16F10 and MB49, and <500 mm3 for MC38 and Pan02, were analyzed using Mantel-Cox analyses in SPSS. Therapeutic synergy in the in vivo efficacy studies was defined as an antitumor effect in which the combination of agents demonstrated significant superiority (p<0.05) relative to the activity shown by each agent alone.33 For IHC, flow cytometry, and cytokine readouts, differences between treatment groups were analyzed using Mann-Whitney or Kruskal-Wallis analysis in GraphPad Prism. Only significant differences (p<0.05) between treatment groups are indicated in the presented figures. Differences were classified as a trend when 0.05≤p<0.1. All reported p values are two-sided.

TCRseq data analysis

For TCRseq data analysis, clonal diversity was calculated using the generalized diversity index (Hill numbers) over a range of diversity orders (q) to generate a smooth curve with the alphaDiversity function from the alakazam R package in the Immcantation framework. Pairwise statistical significance between groups was assessed by constructing a bootstrap delta distribution across groups with 100 realizations. Clonal abundance distribution between treatment groups and tissue compartments (tumor and tdLNs) was estimated using the estimateAbundance function from the Alakazam R package with 100 bootstrap realizations to derive 95% CIs.

RNAseq data analysis

For RNAseq data analysis, differential gene expression between treatment groups was calculated in R V.4.4.1 with a Welch’s modified t-test. Differentially expressed transcripts with a p value<0.01 and fold change>2 were retained for further analysis using Ingenuity Pathway Analysis (IPA) software (Ingenuity Systems) as detailed in the online supplemental materials and methods.

ResultsCombination of acasunlimab and pembrolizumab potentiates T-cell proliferation, cytokine secretion and cytotoxic activity in vitro

To investigate whether additional PD-1 blockade can enhance the single-agent effects of acasunlimab on T-cell activation and cytotoxicity, the combination of acasunlimab with pembrolizumab was studied in vitro using MLR and antigen-specific T-cell assays (online supplemental figure 1).

Co-cultures of mDCs with purified allogeneic CD8+ T cells were treated with test and control antibodies across a range of concentrations (online supplemental figure 1A). Acasunlimab exerted single-agent activity indicative of immune activation in all three allogeneic donor pairs, showing a significant concentration-dependent increase in IL-2 and interferon gamma (IFNγ) release (figure 1A, online supplemental figure 4; online supplemental data file 1). The effects of pembrolizumab were significant for IFNγ and negligible for IL-2 secretion. Acasunlimab in combination with pembrolizumab strongly potentiated IL-2 and, to a lesser extent, IFNγ secretion relative to each antibody alone, which was predicted to be synergistic in all tested donor pairs as defined by the HSA synergy model (HSA synergy scores>10; figure 1B).

Figure 1

Potentiation of cytokine secretion by combining acasunlimab and pembrolizumab in MLR assays. MLR assay (online supplemental figure 1A) in co-cultures of mDC with purified CD8+ T cells (A, B) or co-cultures of mDC with dysfunctional CD3+ T cells (C, D) incubated with acasunlimab, pembrolizumab or the combination thereof (acasunlimab + 1 µg/mL pembrolizumab). (A) Mean IL-2 or IFNγ secretion in purified CD8+ T cell MLRs from one representative donor pair (donor pair 1) is shown with in (B) the corresponding HSA scores of all donor pairs tested (n=3). (C) Mean IFNγ secretion from one representative donor pair (donor pair 3) in MLRs with Tdys with in (D) the respective HSA synergy scores of all donor pairs tested (n=4). Scores >10 are considered predictive of synergy (B, D). Error bars denote SD of duplicate wells. HSA, Highest Single Agent; IFNγ, interferon gamma; IL-2, interleukin 2; mDC, mature dendritic cell; MLR, mixed lymphocyte reaction; PD-L1, programmed death-ligand 1; Tdys, dysfunctional T cells.

Since 4-1BB and PD-1 are coexpressed on exhausted antigen-experienced T cells,5 12–15 we evaluated the ability of concurrent 4-1BB agonism and PD-1/PD-L1 blockade to reinvigorate Tdys. An in vitro model for Tdys was generated by repeated CD3/CD28 stimulation of CD3+ T cells (online supplemental figure 1A) resulting in a T-cell phenotype resembling that of exhausted T cells (online supplemental figure 2A–C). These cells exhibited strongly reduced IFNγ secretion compared with functional T cells when co-cultured with mDCs (online supplemental figure 2D; >90% average decrease). In the MLR assay with Tdys, acasunlimab increased secretion of IFNγ in all donors tested (figure 1C) and IL-2 in three out of four donors (online supplemental figure 2E,F; online supplemental data file 1). Pembrolizumab also induced secretion of IFNγ by Tdys but had a limited effect on IL-2 secretion. The combination of acasunlimab with pembrolizumab potentiated IFNγ secretion compared with either single agent, and was predicted to be synergistic in three out of four donor pairs at clinically relevant acasunlimab concentrations (0.1 to 10 µg/mL)25 28 (figure 1D). The combination also enhanced IL-2 secretion, with a predicted synergistic effect in two out of four donors (online supplemental figure 2G). Together, these results indicate that the combination of acasunlimab and an anti-PD-1 antibody can potentiate cytokine secretion by functional T cells and reinvigorate cytokine release by Tdys in vitro.

Next, we evaluated the effect of acasunlimab and pembrolizumab on T-cell proliferation and cytokine secretion in an antigen-specific setting, where human CD8+ T cells, engineered to express a CLDN6-specific TCR and high levels of PD-1, were co-cultured with autologous iDCs expressing the cognate antigen CLDN6 (online supplemental figure 1B). Acasunlimab significantly enhanced T-cell proliferation in a concentration-dependent manner (figure 2A; Online supplemental figure 5A; online supplemental data file 1). The effect of acasunlimab on T-cell proliferation was further enhanced when combined with pembrolizumab. The combination was predicted to be synergistic across the broad range of tested acasunlimab concentrations (figure 2B) with most prominent effects at low to intermediate concentrations (≤0.4 µg/mL) in all four donors (online supplemental figure 5B). Additionally, the acasunlimab-induced increase in IFNγ and IL-2 secretion (online supplemental data file 1) was further enhanced when combined with pembrolizumab (figure 2C) with synergistic effects predicted in all three donors tested (online supplemental figure 5C,D).

Figure 2

Potentiation of antigen-specific T-cell proliferation, cytokine secretion and cytotoxicity by combining acasunlimab and pembrolizumab in vitro. (A–C) CFSE-labeled human CLDN6-TCR+ PD-1+ CD8+ T cells incubated with CLDN6+ iDCs in the presence of acasunlimab in combination with pembrolizumab (0.8 µg/mL) (online supplemental figure1B). (A) T-cell proliferation as determined by flow cytometry analysis of CFSE label dilution, expressed as mean expansion index±SD of duplicate wells of one representative donor (donor 3). (B) HSA synergy scores for the expansion index of all donors (n=4) with scores>10 considered as predictive of synergy. (C) Mean IFNγ and IL-2 secretion±SD of duplicate wells of one representative donor. (D) CLDN6-TCR+PD-1+ CD8+ T cells were co-cultured with previously seeded MDA-MB-231_hCLDN6 tumor cells (hCLDN6+ PD-L1+) in the presence of 0.0033 or 0.0015 µg/mL acasunlimab, 0.8 µg/mL pembrolizumab or the combination thereof (online supplemental figure 1C). Cytotoxicity of the CD8+ T cells toward MDA-MB-231_CLDN6 cells was monitored by electrical impedance measurement over 140-170 hours using the xCELLigence real-time cell analysis. Data were normalized to the time point of co-culture start and expressed relative to co-cultures incubated without antibodies, which was set to 100%. AUC (total area) analysis of normalized data is shown as mean±SD (n=6–7). Friedman test with Dunn’s multiple comparison test was used to compare AUC between treatment groups (*p<0.05, **p<0.01). AUC, area under the curve; CFSE, carboxyfluorescein succinimidyl ester; hCLDN6, human Claudin-6; iDCs, immature dendritic cells; IFNγ, interferon gamma; IL-2, interleukin 2; PD-1, programmed cell death protein-1; TCR, T-cell receptor.

In an in vitro antigen-specific cytotoxicity assay, CLDN6-TCR+/PD-1+ CD8+ T cells were co-cultured with PD-L1+ MDA-MB-231 tumor cells expressing CLDN6 (online supplemental figure 1C). Acasunlimab significantly increased CD8+ T-cell mediated tumor-cell killing (p<0.05 for 0.0033 µg/mL, that is, EC20 concentration acasunlimab, vs control; figure 2D, online supplemental figure 6A). This effect was further enhanced when acasunlimab was combined with pembrolizumab, which had no significant effect as a single agent (p<0.05 combination vs 0.015 µg/mL acasunlimab; p<0.01 combination vs pembrolizumab). Additionally, the combination further potentiated the proportion of CD107a+GZMB+ CD8+ T cells relative to each single agent (online supplemental figure 6B).

Together, these results show that combining acasunlimab with an anti-PD-1 antibody potentiates the proliferation, cytokine secretion and cytotoxic activity of antigen-specific CD8+ T cells.

Combination of anti-mPD-L1×m4-1BB and anti-mPD-1 potentiates antitumor activity in vivo across multiple syngeneic mouse tumor models with distinct tumor biology

As acasunlimab and pembrolizumab are not mouse-cross-reactive, in vivo antitumor activity of the combination was evaluated using an Fc-inert mouse IgG2a surrogate bispecific antibody targeting mouse PD-L1 and mouse 4-1BB (anti-mPD-L1×m4-1BB)24 and anti-mPD-1 antibody clone RMP1-14 in syngeneic mouse models.

An mPBPK/RO was developed to predict the amount of trimers formed by anti-mPD-L1×m4-1BB when simultaneously bound to 4-1BB and PD-L1 relative to the number of disrupted PD-1/PD-L1 complexes at different doses of anti-mPD-L1×m4-1BB in presence or absence of anti-mPD-1 (figure 3A). Optimal trimer formation, and consequently optimal 4-1BB activation, was predicted to occur in the dose range of 1–5 mg/kg anti-mPD-L1×m4-1BB. However, the disruption of PD-1/PD-L1 complexes was incomplete (27–87%) at these doses (online supplemental table 1), similar to the bell-shaped response previously described for acasunlimab.28 Increasing the dose of anti-mPD-L1×m4-1BB further enhanced the disruption of PD-1/PD-L1 complexes at the expense of trimer formation, resulting in reduced 4-1BB stimulation. In contrast, adding 10 mg/kg of anti-mPD-1 was predicted to disrupt more than 99% of the PD-1/PD-L1 complexes per T cell without impacting trimer formation at any anti-mPD-L1×m4-1BB dose level (figure 3A; online supplemental table 2). According to the model, 5 mg/kg anti-mPD-L1×m4-1BB falls within the peak range of trimer formation while also achieving substantial disruption of PD-1/PD-L1 complexes (87%; online supplemental table 1). Therefore, this dose was selected as the preferred dose in follow-up in vivo studies.

Figure 3

Therapeutic efficacy and memory response of anti-mPD-L1×m4-1BB and anti-mPD-1 combination in MC38 tumor-bearing mice. (A) mPBPK/RO model predictions of 4-1BB and PD-L1 engagement. Simulations were performed for anti-mPD-L1×m4-1BB monotherapy (0.05–20 mg/kg) and combination with anti-mPD-1 (10 mg/kg). Shown is the amount of 4-1BB/anti-mPD-L1×m4-1BB/PD-L1-crosslinked trimers formed versus the number of PD-1/PD-L1 complexes disrupted. (B) 1×106 MC38 tumor cells were injected SC in the right flank of C57BL/6 mice. After tumor establishment (60 mm3 average tumor volume), mice were randomized and treated with PBS, anti-mPD-L1×m4-1BB (5 mg/kg), anti-mPD-1 (10 mg/kg) or the combination thereof at the indicated time points (n=10 per group). (C) Tumor growth of individual mice in each group. CR: number of animals with a complete response. (D) PFS, defined as the percentage of mice with tumor volume smaller than 500 mm3, is shown as a Kaplan–Meier curve. Mantel-Cox analysis was used to compare PFS between treatment groups (*p<0.05, **p<0.01, ***p<0.001). (E) Mice with a CR after combination treatment (shown in C) were rechallenged by SC injection of 1×106 MC38 cells in the left flank 138 days after start treatment. As a control group, a second cohort of naïve mice was inoculated with 1×106 tumor cells. Tumor growth of individual mice in each group is shown. Results are representative of three independent experiments. i.p., intraperitoneal; mPBPK/RO, minimal physiologically-based pharmacokinetic model with receptor occupancy; PBS, phosphate-buffered saline; PD-1, programmed cell death protein-1; PD-L1, programmed death-ligand 1; PFS, progression-free survival; SC, subcutaneous; TV, tumor volume.

In MC38 tumor-bearing C57BL/6 mice, single-agent treatment with 5 mg/kg anti-mPD-L1×m4-1BB or 10 mg/kg anti-mPD-1 delayed tumor outgrowth (figure 3B,C), as evidenced by a lower tumor growth rate (p<0.0001 vs control; online supplemental figure 7) and increased PFS relative to the control group (p<0.001 and p=0.012, respectively; figure 3D). The antitumor activity was potentiated when anti-mPD-L1×m4-1BB was combined with anti-mPD-1 resulting in a lower tumor growth rate (p<0.005; online supplemental figure 7) and enhanced PFS (p≤0.01; figure 3D) relative to each single agent. Notably, durable CR (no palpable tumor left) was observed in 7/10 mice treated with the combination, but in none of the mice treated with the single agent (figure 3C), indicating therapeutic synergy. After rechallenging these seven tumor-free mice with MC38 tumor cells on day 138 after the initial start of treatment, tumor outgrowth was suppressed with no tumors observed in 6/7 mice during the entire follow-up period of 63 days (figure 3E). Together, these data are consistent with a protective immune memory response in mice treated with the combination of anti-mPD-L1×m4-1BB and anti-mPD-1.

Antitumor activity of anti-mPD-L1×m4-1BB in combination with anti-mPD-1 was confirmed in three additional syngeneic models selected to represent TME diversity. In the MB49 urothelial carcinoma model that is characterized by high immune infiltration and known to be responsive to CPIs,34 tumor growth was delayed and PFS increased by anti-mPD-L1×m4-1BB and anti-mPD-1 as single agents (p<0.0001 vs control; online supplemental figure 8). The antitumor activity was further enhanced when anti-mPD-L1×m4-1BB was combined with anti-mPD-1, resulting in a reduced tumor growth rate and extended PFS compared with each single agent (p<0.05 vs anti-mPD-L1×m4-1BB, p<0.01 vs anti-mPD-1). In the immunologically cold CPI-resistant B16F10 melanoma model,34 35 modest single-agent activity was only observed for anti-mPD-L1×m4-1BB, resulting in reduced tumor growth rate (p<0.05 vs control; online supplemental figure 9). Combining anti-mPD-L1×m4-1BB with anti-mPD-1 tended to further enhance tumor growth inhibition (p<0.01 vs control). While anti-mPD-L1×m4-1BB showed a trend of increased PFS (p=0.06 vs control), PFS was significantly extended by the combination (p<0.001 vs control). Finally, in the highly immunosuppressive Pan02 pancreatic ductal adenocarcinoma model that is resistant to anti-PD-(L)1 therapy,35 36 only the combination of anti-mPD-L1×m4-1BB and anti-mPD-1 reduced tumor growth rate (p<0.05 vs control) whereas the single-agent treatments did not (online supplemental figure 10). PFS was increased by anti-mPD-L1×m4-1BB single agent (p<0.05 vs control), which tended to be further extended when anti-mPD-L1×m4-1BB was combined with anti-mPD-1 (p<0.001 vs control). These findings from diverse syngeneic tumor models support potentiation of antitumor activity by the combination of anti-mPD-L1×m4-1BB and anti-mPD-1, with more pronounced effects in tumor models with pre-existing immune infiltration.

Combination of chi-acasunlimab and pembrolizumab potentiates antitumor activity in vivo in a mouse model expressing the human targets h4-1BB, hPD-L1 and hPD-1

To evaluate the in vivo effect of the combination in a model closer to the human setting, hPD-1/hPD-L1/h4-1BB tKI mice expressing the human targets were treated with chi-acasunlimab and pembrolizumab. Chi-acasunlimab was shown to block PD-1/PD-L1 and induce 4-1BB agonist activity in cell-based reporter assays and to activate T cells in human PBMC cultures in vitro to the same level as acasunlimab (online supplemental figure 11). Treating MC38-hPD-L1 tumor-bearing tKI mice with 5 mg/kg chi-acasunlimab and/or 10 mg/kg pembrolizumab was well tolerated, as no body weight loss (online supplemental figure 12A) nor clinical observations were reported. Tumor outgrowth was delayed relative to isotype control in mice treated with chi-acasunlimab or pembrolizumab, with CR in 5/8 and 2/8 mice, respectively, which was also reflected by increased PFS (p<0.001 vs control) and decreased tumor growth rate (figure 4A–C, online supplemental figure 12B). When chi-acasunlimab was combined with pembrolizumab, tumor outgrowth inhibition resulting in CR was increased to 7/8 mice and PFS was further extended (p=0.006 vs pembrolizumab, p=0.218 vs chi-acasunlimab). All animals with a CR (pembrolizumab: n=2; chi-acasunlimab: n=5; combination: n=7) were protected from tumor outgrowth on rechallenge with MC38-hPD-L1 tumor cells on day 143 after the initial start of treatment (figure 4D), consistent with a protective immune memory response. Notably, the percentage of intratumoral cells expressing PD-L2 increased on day 5 after starting chi-acasunlimab-containing treatment (either alone or in combination with pembrolizumab) compared with the control and pembrolizumab-only groups (online supplemental figure 12C). Together, these data further support the rationale for combining acasunlimab with an anti-PD-1 agent to achieve full PD-1 blockade and inhibit the compensatory PD-1/PD-L2 pathway.

Figure 4

Therapeutic efficacy and memory response of chi-acasunlimab and pembrolizumab combination in MC38-hPD-L1 tumor-bearing hPD-1/hPD-L1/h4-1BB tKI mice. (A) 1×106 MC38-hPD-L1 tumor cells were injected SC in the right flank of hPD-1/hPD-L1/h4-1BB tKI mice. After tumor establishment (128 mm3 average tumor volume), mice were randomized and treated with isotype control antibodies (5 mg/kg mIgG2a + 10 mg/kg IgG4), chi-acasunlimab (5 mg/kg), pembrolizumab (10 mg/kg) or the combination thereof at the indicated time points (n=8 per group). (B) Tumor growth of individual mice in each group. CR: number of animals with a complete response. (C) PFS, defined as the percentage of mice with tumor volume smaller than 500 mm3, is shown as a Kaplan–Meier curve. Mantel-Cox analysis was used to compare PFS between treatment groups, with *p<0.05, **p<0.01, ***p<0.001. (D) Mice with a CR after treatment (shown in B) were rechallenged by SC injection of 1×106 MC38-hPD-L1 cells in the left flank 143 days after start treatment. As a control group, a second cohort of naïve hPD-1/hPD-L1/h4-1BB tKI mice was inoculated with 1×106 tumor cells. Tumor growth of individual mice in each group is shown. PFS, progression-free survival; SC, subcutaneous; tKI, triple knock-in; TV, tumor volume.

Next, the role of human PD-L1 expression on tumor cells in the antitumor activity of the combination of chi-acasunlimab with pembrolizumab was assessed using the hPD-1/hPD-L1/h4-1BB tKI mice implanted with MC38 WT tumor cells. Although antitumor activity was reduced compared with the MC38-hPD-L1-bearing tKI model, delayed outgrowth of MC38 WT tumors was observed in tKI mice treated with either 10 mg/kg chi-acasunlimab or 10 mg/kg pembrolizumab, resulting in increased PFS (p<0.05 vs control), 1 CR in each group, and decreased tumor growth rate for chi-acasunlimab (p<0.05 vs control) (online supplemental figure 13). The combination of chi-acasunlimab with pembrolizumab tended to further delay tumor outgrowth, evidenced by a more significant increase in PFS and decrease in tumor growth rate compared with the isotype control (p<0.001), and reduced tumor growth rate compared with chi-acasunlimab alone (p<0.01). However, PFS in the combination group was not significantly improved compared with the single-agent treatments. Since chi-acasunlimab and pembrolizumab bind only to host cells expressing the human targets in the tKI mice, bu