WHAT IS ALREADY KNOWN ON THIS TOPIC

CD28 bispecific antibodies are an emerging class of immunotherapies with the potential to treat a wide range of cancers. However, little is known about the optimal design of CD28 engagers, including the ideal potency of these molecules for maximizing clinical success.

WHAT THIS STUDY ADDS

We describe the preclinical development of a CD28 × Nectin-4 bispecific antibody (bsAb) optimized for the treatment of advanced urothelial cancer (UC) that elicits costimulation of tumor-specific T cells at the site of the tumor. We detail our approach to empirically determine the optimal potency of CD28 engagement from a panel of molecules containing CD28 binding domains with different affinities. This is the first example of a CD28 bsAb targeting Nectin-4 and represents a new therapeutic strategy for UC.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICYIntroduction

Immune checkpoint therapy (ICT) has revolutionized cancer treatment; however, the limited response rates to single-agent ICT highlight the urgent need for additional therapeutic strategies.1 In hematological malignancies, bispecific T-cell engaging antibodies (TCEs) have emerged as a promising therapeutic modality, evidenced by several recent Food and Drug Administration (FDA) approvals.2 In contrast, TCEs have faced significant challenges in solid tumors, with limited efficacy or toxicity often implicated in discontinuation of clinical use.3 4 A key barrier to TCE efficacy in solid tumors is the complex and immunosuppressive tumor microenvironment which hinders both the antitumor activity and durability of these therapies.5

Recent reports suggest that an important characteristic of the suppressive solid tumor microenvironment is insufficient T-cell costimulation stemming from a lack of antigen-presenting cells expressing the CD28 ligands CD80 and CD86 as well as cytotoxic T lymphocyte-associated antigen 4 (CTLA-4)-mediated sequestration of these proteins.5–7 CD28 is a membrane-bound, homodimeric T-cell costimulatory receptor expressed by most T cells, and is integral to T-cell activation, proliferation, and the formation of a memory response.8 Interaction with its ligands on antigen-presenting cells provides a necessary signal (termed Signal 2) that enhances T-cell receptor (TCR) signaling (termed Signal 1) and is required for full T-cell activation. Insufficient costimulation causes T cells to enter a state of anergy, characterized by hypo-responsiveness to cognate peptide major histocompatibility complex (MHC) resulting in reduced effector function and immunosuppression.9 10 In the context of cancer therapy, augmenting CD28 signaling has the potential to amplify endogenous T-cell responses against tumor cells.11 However, a key challenge lies in achieving selective activation of tumor-specific T cells to prevent systemic immune activation and associated toxicities, such as those seen in the phase I clinical trial of the CD28 super-agonist antibody TGN1412.12 13

A bispecific antibody (bsAb) approach that integrates CD28 engagement with the binding of a tumor-associated antigen (TAA) has the potential to direct costimulation specifically to the site of the tumor. This strategy has the benefit of enhancing the function of antitumor T cells while minimizing systemic immune activation. Several clinical trials are currently investigating this approach;11 however, key questions remain, such as the optimal potency of CD28 activation in bsAb format and the potential for clinical development as monotherapies. Insights from TCE development have shown that optimizing potency through affinity tuning of anti-CD3 antibodies can improve therapeutic windows.14 15 We hypothesize that similar tuning will be necessary to optimize the therapeutic window for CD28 bsAbs depending on the TAA, indication and potential combination partners. Therefore, there is a need for novel CD28-engaging antibodies with a range of potencies to explore optimal engagement strategies across various tumor targets and indications.

Nectin-4 is a cell adhesion molecule overexpressed in multiple malignancies, including urothelial, breast, skin, lung, and pancreatic cancers, with limited expression in healthy adult tissues.16–18 Moreover, Nectin-4 has been implicated in promoting tumor cell proliferation, migration, and survival, making it a compelling target for therapeutic intervention.18 In urothelial cancer (UC), a Nectin-4-directed antibody drug conjugate (ADC) is approved for use in locally advanced and metastatic disease (la/mUC) while additional recently approved therapies include several checkpoint inhibitors as well as FGFR inhibitors for a subset of patients.19 Despite these advancements, options for la/mUC patients who become refractory to first-line therapies remain limited, with 5-year survival rates below 10% for those with metastatic disease.20–23 Thus, there is an urgent need for novel therapeutic approaches and new modalities to improve patient outcomes.

In this study, we describe the identification and optimization of a CD28 × Nectin-4 clinical development candidate by evaluating a panel of bsAbs with a range of potencies. These molecules are designed to enhance immune-mediated tumor eradication while minimizing off-target effects by activating T cells only in the setting of both TCR/peptide-MHC engagement and Nectin-4-positive tumor cells. A detailed analysis of the binding properties, in vitro functional activity, and therapeutic potential of these molecules in preclinical models allowed us to select a clinical development candidate, termed RNDO-564, that is designed to support robust antitumor activity while preserving a wide therapeutic window. Our results suggest a promising new immune-activating therapeutic modality for the treatment of Nectin-4-positive cancers that is differentiated from the current standard of care and, to our knowledge, is the first time Nectin-4-targeted CD28 activation will be investigated as a treatment for UC.

ResultsDiscovery of anti-CD28 antibodies

We used a next-generation sequencing (NGS)-based discovery approach to identify 764 unique monoclonal antibodies (mAbs) from antibody repertoires of humanized rodents immunized with CD2824 (figure 1A). Based on high-throughput cell binding analysis to CD28-positive and -negative cell lines, we selected 27 mAbs for further characterization. Additional studies confirmed that 25 of the 27 candidates showed specific binding to human CD28 protein and a 30-fold range of binding strength to CD28-positive Jurkat cells, while none of the mAbs bound to mouse CD28 (figure 1B). We then evaluated the costimulatory activity of each CD28 mAb in combination with the anti-CD3 antibody OKT3 by measuring interleukin (IL)-2 production from human CD4+ T cells incubated with each antibody co-treatment. All but one of our CD28 candidate mAbs enhanced IL-2 production above that of OKT3 stimulation alone, although generally not to the extent of the strong positive control CD28 mAb TGN1412. Sequence analysis of the CD28 mAbs revealed six clonotypes in the set of 27 antibodies based on CDR3 (complementarity determining region 3) similarity. To screen additional mAb candidates that could have more desirable features, such as varying affinities or better developability properties, we identified 269 additional members of these clonotypes from the full antibody repertoire data set. After another round of screening similar to that described for the initial set of 27 mAbs, we selected five CD28 mAbs (online supplemental figure S1), based on their specific CD28 binding, lack of sequence liabilities and range of target affinity (double to triple-digit nM affinities to recombinant human CD28 protein) and converted them into bsAb format.

Figure 1

Discovery of a diverse panel of CD28 agonist antibodies. (A) Our CD28 antibody discovery approach consisted of four key steps: immunization of humanized rodents, antibody repertoire analysis and high-throughput screening of monoclonal antibodies (mAbs), detailed binding and functional analysis of a subset of antibodies and conversion of selected candidates to bsAb format. (B) Summary of step 3 of our CD28 antibody discovery approach. CD28-positive Jurkat cell binding dose response results for 27 anti-CD28 mAbs reported as EC50 values (column 2). Binding signal as measured by ELISA for each antibody tested at a concentration of 1.6 μg/mL, 0.16 μg/mL and 0.016 μg/mL to human CD28 (columns 3–5) and at 1.6 µg/mL to off-target proteins (columns 6–9). Human CD4+ T-cell costimulation as measured by IL-2 production (column 10). IL-2 was measured in supernatants from 96-well plates containing 1×105 T cells in a 200 μL volume incubated overnight in the presence of 1 μg/mL plate-coated anti-CD3 antibody OKT3 and 16 µg/mL of each anti-CD28 mAb. bsAb, bispecific antibody; CTLA-4, cytotoxic T lymphocyte-associated antigen 4; IL, interleukin; mu, mouse.

Design and specificity of CD28 × Nectin-4 bsAbs

Following an antibody discovery strategy similar to that described for CD28, we identified the Nectin-4 binding antibody, Nectin-4_001, based on its high target affinity (124 pM and 328 pM to human and cynomolgus Nectin-4, respectively, figure 2B, online supplemental figure S2) and paired it with our panel of five CD28 antibodies for conversion to bsAb format (figure 2A). To prevent Fc receptor and complement binding of the bsAbs, we used a hinge-stabilized, silenced IgG4 Fc, with knobs-into-holes mutations to drive efficient formation of heterodimers.25 26 We observed varying binding strength to Jurkat cells for the set of five CD28 × Nectin-4 bsAbs, as expected given the range of CD28 protein-binding affinities of these molecules (figure 2C). Binding to the Nectin-4 positive cell line T-47D was highly similar because the five bsAbs contain an identical Nectin-4-binding domain. None of the bsAbs showed binding to a target cell line negative for both CD28 and Nectin-4 (UM-UC-3).

Figure 2

CD28 × Nectin-4 bispecific antibody panel. (A) Schematic of bispecific antibody format. A human IgG4 Fc scaffold was used, including silencing mutations and knob-into-hole mutations to facilitate heterodimer formation. VH (variable heavy chain), VL (variable light chain), CH1/2/3 (heavy chain constant domain 1/2/3), CL (light chain constant domain). (B) Surface plasmon resonance analysis of human Nectin-4 binding by the bsAb CD28_075 × Nectin-4_001. The reported values are averages of triplicate measurements at two loading densities and a representative binding trace is shown. (C) Analysis of cell binding by flow cytometry of the panel of five CD28 × Nectin-4 bsAbs. Binding results to CD28-positive Jurkat cells, Nectin-4-positive T-47D cells and a CD28- and Nectin-4-negative cell line, UM-UC-3 are shown. The mean and SD of triplicate measurements are shown. bsAb, bispecific antibody.

Biophysical properties of CD28 × Nectin-4 bsAbs

The panel of CD28 × Nectin-4 bsAbs displayed robust transient expression yields and purity above 90% (table 1). We further polished the bsAbs to greater than 95% purity and analyzed the melting and aggregation temperatures (Tm and Tagg), finding values within the 90% threshold of approved mAbs.27 We then assessed stability after four cycles of freeze-thaw and after thermal stress by incubation at 40°C for 1 month. All antibodies showed excellent stability with ≤2.0% change in monomer content.

Table 1

Biophysical properties of CD28 × Nectin-4 bispecific antibodies. All antibodies show >90% purity and yield between 24 and 42 milligrams (mgs) from a 200 mL transient transfection after a 1-step protein A purification. Melting (Tm) and aggregation (Tagg) temperatures were measured on 2-step purified proteins on the Uncle. Long-term thermal stability was tested by incubating samples for 4 weeks at 40°C. Stability was assessed by SEC-HPLC analysis of percent monomer content, decrease in percent monomer from the initial time point (T0) to 4 weeks (D28) at 40°C is reported. Stability after four freeze-thaw cycles (FT4) was analyzed by percent change in monomer compared with T0 by SEC-HPLC.

CD28 × Nectin-4 bsAbs exhibit target-dependent T-cell costimulation

To determine whether the CD28 × Nectin-4 bsAbs can provide T-cell costimulation, we developed a co-culture assay using primary human T cells and human tumor cell lines ranging in Nectin-4 expression (online supplemental figure S3). TCR stimulus (Signal 1) in our in vitro assays was provided by a CD3 bispecific antibody (CD3 bsAb) targeting the TAA 5T4 (expressed by all tumor cell lines that we used). We first assessed tumor cell cytotoxicity in the presence of a subefficacious concentration of CD3 bsAb combined with a titration of each CD28 × Nectin-4 bsAb across the panel of Nectin-4-positive and -negative cell lines (figure 3A). All the CD28 × Nectin-4 bsAbs showed dose-dependent enhanced cytotoxicity of the Nectin-4-positive cell lines. We next measured production of IL-2 following a similar assay design and observed increased IL-2 in the presence of Nectin-4-positive cells (figure 3B). Importantly, we did not detect cytotoxicity or cytokine production in the absence of Signal 1 (online supplemental figure S4A,B). We also assessed whether T-cell proliferation was enhanced by the CD28 × Nectin-4 bsAbs and found dose-dependent augmented proliferation of both CD4+ and CD8+ T cells only in the presence of Signal 1 (figure 3C, online supplemental figure S4C). CD28 antibodies that non-specifically activate T cells in the absence of Signal 1 pose a safety risk due to uncontrolled immune activation and are termed super-agonists.13 We assessed the potential for super-agonist activity in our CD28 × Nectin-4 bsAb panel by using an FDA-recommended assay where human peripheral blood mononuclear cells (PBMCs) are incubated with immobilized antibodies and tested for activation by cytokine release.28 29 We compared our CD28 × Nectin-4 bsAbs as well as the CD28 parental mAbs to the quintessential CD28 super-agonist TGN1412 by measuring production of interferon (IFN)-γ. We did not observe super-agonist activity from the five CD28 × Nectin-4 bsAbs or parental CD28 mAbs, whereas the positive control TGN1412 showed robust IFN-γ production (figure 3D).

Figure 3

CD28 × Nectin-4 bsAbs exhibit target-dependent T-cell costimulation. (A) T cells were incubated with target cells at a 5:1 E:T ratio in a 200 μL assay volume. CD28 × Nectin-4 bsAbs were titrated with a limiting concentration of CD3 bsAb. To assess tumor cell cytotoxicity, the concentration of the CD3 bsAb was adjusted for each cell line to correspond to the EC10 concentrations as follows: T-47D at 5.5 pM, RT4, RT-112 and UM-UC-3 at 27.6 pM. Assays were incubated for three days, after which time cell viability was assessed. The mean and SD of triplicate values are shown. (B) In an IL-2 release assay, CD28 bsAbs were titrated with a fixed CD3 bsAb concentration of 103 pM, a level of Signal 1 optimal for analysis of cytokine release. Assay supernatants were collected after 24 hours of incubation and analyzed for production of IL-2. The mean and SD of duplicate values are shown. (C) T cells labeled with CellTrace Violet (CTV) were incubated with T-47D target cells at a 2:1 E:T ratio. Proliferation of CD8+ and CD4+ T cells by CTV dilution was measured after five days of incubation by flow cytometry. The mean and SD of triplicate measurements are plotted for panels A, B and C. (D) Assessment of the potential for super-agonist activity was measured by IFN-γ production in a solid-phase cytokine release assay containing 1×105 human peripheral blood mononuclear cells (PBMCs) incubated with immobilized CD28 × Nectin-4 bsAbs or the parental CD28 monoclonal antibodies in a 200 μL assay volume. Triplicate measurements of two representative PBMC donors are plotted. bsAb, bispecific antibody; E:T, effector to target; IFN, interferon; IL, interleukin.

CD28 costimulation supports T-cell cytotoxic activity at a low E:T ratio

We next assessed the costimulatory activity of our CD28 × Nectin-4 molecules at a low effector to target (E:T) ratio of 1:4, as tumor-specific T cells are likely present at low numbers within the tumor microenvironment.9 30 For comparison, we also tested a 5:1 E:T ratio, where the CD3 bsAb induced high levels of tumor cell cytotoxicity alone, with further potency enhancements after the addition of each of the CD28 × Nectin-4 bsAbs (figure 4A, online supplemental figure S5A). However, the CD3 bsAb alone at a 1:4 E:T ratio had no activity (figure 4B, online supplemental figure S5B). The addition of CD28 × Nectin-4 bsAbs restored activity to varying degrees correlating to the bsAb potency, suggesting the potential for activity in a setting with limited tumor-specific T cells.

Figure 4

CD28 costimulation supports T-cell redirected cytotoxicity at low E:T ratios and differentially impacts cytokine production. 1×104 RT4 bladder cancer cells were incubated for three days with human T cells from one donor at either (A) a 5:1 or (B) 1:4 effector to target cell ratio and a titration of CD3 bsAb with or without 4 µg/mL of the indicated CD28 × Nectin-4 bsAb followed by assessment of tumor cell cytotoxicity. The mean and SD of triplicate values are shown. (C) Assessment of cytokine production was conducted by co-culturing 1×104 RT4 bladder cancer cells with freshly isolated human peripheral blood mononuclear cells (donor #9021) at a 5:1 effector-to-target (E:T) ratio and a titration of CD3 bsAb, with or without 4 µg/mL of each CD28 × Nectin-4 bsAb in a 200 µL volume. Assay supernatants were collected after 48 hours to assess production of the cytokines IL-2, IFN-γ, IL-6, IL-4, IL-10, IL-1β and TNF-α. Fold enhancement of production of each cytokine by the addition of each CD28 × Nectin-4 bsAb was calculated relative to the CD3 bsAb treatment alone by dividing the area under the curve for each co-treatment condition with that of the CD3 bsAb treatment alone (GraphPad Prism V.10). Data from a representative donor is shown. A dashed horizontal line indicates a fold change of 1. bsAb, bispecific antibody; IFN, interferon; IL, interleukin; TNF, tumor necrosis factor.

CD28 × Nectin-4 costimulation differentially impacts cytokine production

We next investigated the impact of CD28 × Nectin-4 costimulation on production of cytokines associated with T-cell effector function and cytokine release syndrome (CRS)31 (online supplemental figure S6). To measure the change in cytokine production from the addition of each CD28 × Nectin-4 bsAb, we calculated the fold enhancement of each co-treatment relative to the CD3 bsAb titration alone. IL-2 was the most enhanced cytokine in each co-treatment, with smaller increases in IFN-γ, IL-4, IL-1β and tumor necrosis factor (TNF)-α (figure 4C, online supplemental figure S7). Importantly, we did not detect substantial increases in the CRS-associated cytokine IL-6.

Clinical development candidate selection

Our in vitro assessments demonstrated a range of activity in the panel of CD28 × Nectin-4 bsAbs. To identify an optimal clinical candidate, we considered both potency and maximum activity. CD28_076 × Nectin-4_001 and CD28_077 × Nectin-4_001 had reduced maximum activity in some assays such as cytotoxicity at low E:T ratios (figure 4B, online supplemental figure S5B), suggesting a potential risk for insufficient clinical activity. In contrast, the three other bsAbs in the panel had similar maximum activity but a range of potencies (figure 3). CD28_075 x Nectin-4_001 was the least potent of these three molecules, suggesting it may maintain efficacy while benefiting from a wider therapeutic window. Thus, we selected it for clinical development, renaming it to RNDO-564.

RNDO-564 enhances T-cell function in settings with mixed Nectin-4 positive and negative tumor cells

Recent studies report heterogeneous Nectin-4 expression in both primary and metastatic tumors in patients with la/mUC.32 33 We assessed Nectin-4 expression on dissociated tumor cells (DTCs) from patients with UC (n=7), and found that the Nectin-4-positive fraction ranged from 9% to 51% (figure 5A). Thus, we were interested in determining the activity of RNDO-564 in a setting with mixed Nectin-4-positive and -negative tumor cells. We generated a Nectin-4-negative MCF-7 cell line (MCF-7KO), then combined these cells and the Nectin-4-positive parental MCF-7 cells (MCF-7P) at various ratios in a cytotoxicity assay containing RNDO-564 and a subefficacious concentration of CD3 bsAb. We assessed the viability of the MCF7P and MCF7KO populations, observing enhanced cytotoxicity against both MCF7P and MCF7KO target cells at each ratio tested (figure 5B). As the fraction of MCF7KO cells increased, maximum levels of cytotoxicity decreased, perhaps due to the reduction in target cells providing the costimulatory signal necessary to enhance T-cell function. However, activity was greatly elevated over that observed under the control condition containing 100% MCF7KO cells. We then assessed co-cultures of allogeneic T cells with bladder DTCs treated with RNDO-564 alone or in combination with CD3 bsAb (figure 5C). We observed increased T-cell activation and cytokine production from the combination treatment as well as smaller increases from each individual treatment. T-cell activation correlated to the fraction of Nectin-4-positive DTCs (online supplemental figure S8), in agreement with the pattern of cytotoxicity enhancement seen in our MCF7 model system. The small increases from RNDO-564 monotherapy treatment may indicate alloreactivity between the T-cell donor and DTCs or activation of tumor-specific T cells within the DTC population. Taken together, these experiments demonstrate beneficial activity by RNDO-564 in a mixed Nectin-4-positive and -negative tumor cell setting with either tumor cell lines or UC patient DTCs.

Figure 5

RNDO-564 enhances T-cell function in settings with mixed Nectin-4-positive and -negative target cells. (A) Nectin-4 expression was analyzed on seven patients with bladder cancer dissociated tumor cell samples (DTCs) by flow cytometry. Tumor cells were identified by EpCAM expression. (B) Parental MCF7 (MCF7P) and Nectin-4 knock-out MCF7 (MCF7KO) cells were labeled with CellTrace Violet and CFSE (carboxyfluorescein diacetate succinimidyl ester), to distinguish the two populations and mixed at varying ratios of MCF7P : MCF7KO as indicated by the X-axis. Human T cells were added at a 1:1 E:T ratio, along with CD3 bsAb (5.5 pM) and RNDO-564 (1 μg/mL) and incubated for three days. Target cell cytotoxicity was assessed using flow cytometry by calculating the number of live labeled target cells per well compared with untreated controls. The mean and SD of triplicate values are shown. (C) Four of the DTC samples from panel A were co-cultured with allogeneic T cells at a 10:1 E:T ratio along with 64 μg/mL RNDO-564 with or without 4 ng/mL of CD3 bsAb. Assay supernatants were analyzed for cytokine production after 48 hours of incubation and T-cell activation was assessed by flow cytometry analysis of CD25 expression after 72 hours of incubation. bsAb, bispecific antibody; E:T, effector to target; IL, interleukin.

CD28 costimulation enhances the function of chronically stimulated T cells

Chronically stimulated T cells with reduced effector activity may be found in the tumor environment,5 thus we were interested in determining if T-cell costimulation could reverse the loss of cytotoxic function associated with this state. To test this in vitro, we repeatedly activated T cells with beads conjugated to CD3 antibodies.34 After serially activating the T cells, we profiled their exhaustion marker expression by flow cytometry, finding increases in PD-1, CD39, TIM-3, LAG3, CTLA-4 and TOX (online supplemental figure S9). We then assessed their function in a cytotoxicity assay with either CD3 bsAb or co-treatment with CD3 bsAb and RNDO-564 (figure 6A). Serially stimulated T cells treated with CD3 bsAb alone showed reduced tumor cell cytotoxicity, while the addition of RNDO-564 completely restored cytotoxic function. For comparison, we tested T-cell function prior to serial stimulation, observing complete cytotoxicity of tumor cells with or without costimulation.

Figure 6

CD28 costimulation enhances the function of chronically stimulated T cells. (A) Human T cells were activated with anti-CD3 beads at a 1:1 ratio every two days. After the fourth stimulation, T cells were collected for use in a cytotoxicity assay with either CD3 bsAb (1 µg/mL) or a combination of CD3 bsAb and RNDO-564 (4 µg/mL) at a 2:1 E:T ratio with T-47D target cells. Target cell survival was measured using xCELLigence cell index values after six days of incubation. The mean of triplicate values is plotted with SD indicated by error bars. (B) 3×103 T47D cells were co-cultured with human T cells at 1:1 E:T ratio along with the indicated treatments of CD3 bsAb (1 µg/mL), RNDO-564 (4 µg/mL), IL-2 (10 U/mL) and anti-IL-2 antibody (4 µg/mL) in xCELLigence PET 96-well plates. T cells were re-stimulated by the addition of 3×103 T-47D target cells every 48–72 hours (on days 0, 3, 5 and 7) and percent target cell survival was measured using xCELLigence cell index values. Percent survival in each condition was calculated relative to the average of untreated control wells. Each condition was tested in four to eight replicates and the mean and SD of all replicates are plotted. bsAb, bispecific antibody; E:T, effector to target; IL, interleukin.

Given that CD28 costimulation supports high levels of IL-2 production, we next investigated the impact of this cytokine on tumor cell cytotoxicity in a serial stimulation assay where T cells were repeatedly challenged with tumor cells in the presence of CD3 bsAb (figure 6B). We monitored tumor cell viability to assess T-cell cytotoxic function when treated with either CD3 bsAb and RNDO-564 (dual treatment) or CD3 bsAb + RNDO-564 + an IL-2 antagonist antibody (dual treatment + αIL-2). For comparison, we also tested treatment with CD3 bsAb alone and CD3 bsAb + IL-2 with and without the addition of IL-2 antagonism. Interestingly, dual treatment + αIL-2 initially showed similar levels of tumor cell cytotoxicity as dual treatment; however, by the third challenge tumor cell cytotoxicity plateaued and tumor growth control was lost. We also observed limited tumor cell cytotoxicity in the presence of monotherapy treatment with CD3 bsAb alone. In contrast, treatment with CD3 bsAb + IL-2 showed tumor cell cytotoxicity equivalent to the dual treatment condition. These results highlight the important role of IL-2 in supporting sustained T-cell function and indicate that CD28 costimulation enhances T-cell function through additional mechanisms independent of IL-2. CD28 costimulation is important for the formation of a T-cell memory response8 and so we investigated the memory phenotype following treatment with a dose titration of RNDO-564 or a non-targeting CD28 bsAb negative control in the presence of Signal 1. RNDO-564 preferentially supported a central memory phenotype in CD8+ and CD4+ T cells (online supplemental figure S10).

RNDO-564 is effective as a single agent and in combination with ICT in tumor-bearing syngeneic mouse models

To evaluate the in vivo activity of RNDO-564, transgenic mice expressing human CD28 extracellular domain were subcutaneously implanted with a mouse tumor cell line (MC38) expressing human Nectin-4 (figure 7A, online supplemental figure S11). MC38 cells express retroviral antigens which are recognized as foreign by mouse T cells, thus providing Signal 1.35 Therefore, we hypothesized that T-cell costimulation with RNDO-564 would elicit monotherapy activity by enhancing the endogenous immune response. After tumors reached an average volume of 100 mm3, twice weekly treatments were initiated at 10, 1, and 0.1 mg/kg. We observed significant tumor regression in all treatment groups, with complete responses in all members of the 10 mg/kg group by day 45 (figure 7B). We next tested the combination of RNDO-564 with an anti-mouse PD-1 (programmed cell death protein 1) antibody, both dosed at 1 mg/kg, selected based on their moderate activity as monotherapy treatments. All animals in the combination group showed stable disease or complete response, whereas in both monotherapy groups 25% of animals did not respond to treatment (figure 7C). These results provide evidence for complementary activity between PD-1 inhibition and CD28 agonism consistent with literature reports.36–38

Figure 7

RNDO-564 has single-agent activity in an immune-primed syngeneic tumor model. (A) Study schematic. (B and C) Immuno-competent mice expressing human CD28 in place of mouse CD28 were subcutaneously implanted with MC38-hNectin-4 tumor cells. When the tumors reached 100 mm3 the animals were randomized to treatment groups (eight/group) and RNDO-564 (0.1, 1 or 10 mg/kg) or an isotype control (10 mg/kg) was injected intraperitoneally, two times a week for 39 days. (C) Additional groups with combination treatments were tested. Anti-PD-1 antibody at 1 mg/kg was tested alone or as a combination with 1 mg/kg of RNDO-564 and compared with a vehicle control. Tumor volumes were monitored over time and depicted as average±SEM. P values were calculated by two-way analysis of variance using Dunnett’s correction and shown for day 35 for each treatment group relative to isotype control (A) or vehicle (B) (ns – not significant, *p<0.05, **p<0.01, ***p<0.001 ****p<0.0001). Individual animal curves are shown below each figure. aPD-1, anti-programmed cell death protein 1; ICT, immune checkpoint therapy.

Pharmacokinetics and tolerability of intravenous RNDO-564 in non-human primates

As RNDO-564 cross-reacts with cynomolgus monkey CD28 and Nectin-4, this species represents a relevant model for pharmacokinetics (PK) and tolerability. Cynomolgus monkeys were injected with an intravenous infusion of either 1 or 10 mg/kg RNDO-564 followed by observation and intravenous sampling. The mean half-life was 6.2 days, consistent with expectations of a standard IgG (online supplemental figure S12A).39 Analysis of immune cell composition and production of T-cell activation markers did not reveal any RNDO-564 dependent significant changes in any immune cell subset or increase in T-cell activation (data not shown). We also measured serum levels of the cytokines IL-2, IL-4, IL-6, IL-8, IL-10, IFN-γ and TNF-α, observing no significant changes in cytokines across any of the individuals (online supplemental figure S12B). These results suggest RNDO-564 is well-tolerated and exhibits normal PK.

Extended developability assessment of RNDO-564

The favorable developability characteristics observed during assessment of our bsAb panel (table 1) translated to desirable manufacturability properties for RNDO-564. A CHO stable cell line expressing RNDO-564 produces 5.2 g/L of desired product in a bioreactor at ≥95% purity (by size exclusion high performance liquid chromatography (SEC-HPLC)). A standard mAb purification process with affinity capture, low pH viral inactivation, intermediate anion exchange, and a final mixed mode polishing step can be successfully applied to deliver pure bsAb with high yields. RNDO-564 shows negligible change in purity under stress conditions, including elevated temperature, oxidative stress, low pH and freeze-thaw. For example, after incubation at 40°C for 4 weeks RNDO-564 formulated at 50 mg/mL showed a ≤1% change in monomer content by SEC-HPLC (online supplemental table S1).

Discussion

CD28 agonist therapeutics were largely abandoned following the acute toxicity observed in the 2006 TeGenero phase one trial evaluating the experimental mAb TGN1412. However, subsequent studies of TGN1412 established preclinical assays that reduce the target risk by identifying CD28 super-agonists early in discovery. Furthermore, advances in bsAb technology enable tumor-antigen binding-dependent CD28 costimulation, further minimizing the risk of systemic immune activation. These developments result