Introduction
Antibody blockade of PD-1/PD-L1 interactions has resulted in regulatory approvals in at least 14 different cancer indications, demonstrating the broad utility of this approach [
1,
2]. However, most patients do not achieve long-term responses, and additional interventions are required. Combination with targeted and chemotherapies has improved outcomes to PD-1 blockade as evidenced by the recent combination approvals in renal cell carcinoma [
3] and breast cancer [
4].
A significant factor in the resistance to PD-1 blockade therapy is the lack of tumor-specific T cell responses, which may be attributed to low levels of neoantigens, inefficient antigen processing and presentation and poor T cell priming [
5,
6]. Therefore, combination with agents that enhance these processes is rational for improving outcomes to PD-1 blockade.
The TNFR superfamily member CD27 is a T cell costimulatory molecule with an important and non-redundant role in their activation, proliferation and survival [
7‐
10]. This point is best illustrated in individuals with deficiency of either CD27 or its ligand CD70 that results in sometimes fatal EBV-driven lymphoproliferation, hypogammaglobulinemia and lymphoma development [
11‐
15]. Among the costimulatory molecules 4-1BB (CD137), OX40 (CD134) and ICOS (CD278), CD27 is uniquely expressed at high levels on naïve T cells making it especially well-suited to help prime and promote new T cell responses [
7,
16‐
18].
The potent activity of CD27 agonist antibodies in T cell activation and antitumor activity has been well documented in mouse models and human primary cultures [
19,
20]. The antitumor activity of the anti-human CD27 agonist mAb, varlilumab, was established using human CD27 transgenic (huCD27-Tg) mice and found to be mediated through a combination of T cell costimulation and reduction/inhibition of Treg cells [
21]. Importantly, a strong synergy in antitumor activity was observed when combining CD27 agonist mAbs with PD-1/PD-L1 blockade. Transcriptome analysis of the expanding CD8 T cells from these studies demonstrated complementary profiles of proliferation and cytotoxicity for CD27 agonism and PD-1/PD-L1 blockade, respectively, which converged in the combination treatment to yield a particularly notable enhanced expression of genes associated with T cell activation [
22].
Varlilumab has been in clinical trials both as monotherapy and in combination with PD-1 blockade. As monotherapy in patients with advanced cancers, varlilumab was well tolerated at all dose levels, induced pharmacological changes consistent with the preclinical data (T cell activation and Treg depletion) and demonstrated clinical activity with 3 patients having long-term clinical benefit without any additional therapy [
23]. Varlilumab in combination with nivolumab in advanced solid tumor patients was generally well tolerated through the highest dose of varlilumab (10 mg/kg) and nivolumab (3 mg/kg) administered every 2 weeks [
24]. Clinical benefit was observed in patients with low probability to respond to nivolumab monotherapy. In particular, we observed durable responses in ovarian cancer patients with PD-L1 low or negative tumors and glioblastoma multiform patients with unmethylated gene promoter for O-6-methylguanine-DNA methyltransferase [
25].
The strong rationale for combining CD27 costimulation and PD-1 blockade, together with the supportive preclinical and clinical data motivated us to engineer these activities into a single molecule. To this end, we developed novel CD27 and PD-L1 human antibodies using human Ig transgenic mice and expressed them as a whole IgG genetically fused to a single-chain variable fragment (scFv) yielding a tetravalent bispecific antibody (BsAb) [
26] referred to as CDX-527. Here we report the generation, characterization and functional activities of CDX-527, and of a surrogate BsAb that is cross-reactive with mouse PD-L1 and used to demonstrate in vivo activity.
Materials and methods
Development and characterization of bispecific antibodies
Antibodies to CD27 and PD-L1 were generated by immunization of H2L2 human Ig transgenic mice (Harbour Antibodies BV) with recombinant human CD27 or PD-L1. Splenocytes were used for hybridoma preparation by standard polyethylene glycol fusion techniques. The variable heavy and light chain regions of selected antibodies were cloned into a human IgG1κ expression vector, expressed in ExpiCHO cells (Invitrogen), and further characterized. The PD-L1 antagonist mAb avelumab (AVE) was similarly prepared using the published sequence (WHO Drug Information, Vol. 29, No. 2, p. 203, 2015).
For the BsAb, an expression vector encoded the full-length anti-PD-L1 mAb 9H9 IgG1κ heavy and light chains and the scFv of the anti-CD27 2B3 mAb genetically linked in VL–VH orientation to the C-terminus of the 9H9 mAb heavy chain. Cysteine residues were introduced, one in 2B3 VL and one in 2B3 VH, to stabilize the scFv domains. An analogous vector was prepared in which the 9H9 VH and VL were replaced by the avelumab VH and VL. These vectors were transfected into HD-BIOP3 cells (Horizon Discovery) and proteins were purified by protein A and size-exclusion chromatography. All purified mAbs and BsAbs contained < 0.5 endotoxin units/mg.
Binding and blocking assays
Affinity determination using bio-layer interferometry
The mAbs or BsAbs were captured on anti-human Fc capture biosensors (ForteBio). Binding was determined by exposing the loaded biosensor to human PD-L1-HIS (R&D Systems) or human CD27 (generated in-house). Affinity measurements were determined using twofold serial dilutions of analyte ranging from 50 to 0.195 nM. The association and dissociation curves were fitted to a 1:1 binding model using the data analysis software according to the manufacturer’s guidelines.
ELISA assays
Extracellular domains of human and cynomolgus CD27 were generated and purified from transient transfections using Protein L. The fusion protein of human PD-L1 and mouse Ig Fc (msFc) domain was purified from transient transfections using Protein A. The following fusion proteins were purchased: mouse PD-L1 with mouse Ig Fc domain (ACROBioSystems), and mouse, cynomolgus macaque and rat PD-L1 with human Ig Fc (R&D Systems). For ELISA, plates coated with recombinant protein were exposed to samples and binding was detected using an HRP-labeled goat-anti-human IgG (Fc-specific) antibody and developed with 3,3′,5,5′-tetramethylbenzidine substrate. For BsAb binding, wells were coated with CD27 protein. BsAb dilutions were allowed to bind before adding human or mouse PD-L1-msFc which was detected with an HRP-labeled goat anti-mouse IgG (Fc specific) antibody.
Flow cytometry
HEK293 cells transfected with human CD27 or PD-L1 (Crown Bioscience) were incubated with mAbs for 20 min, and the bound antibodies were detected with a phycoerythrin (PE)-labeled goat anti-human IgG Fc-specific probe (Jackson ImmunoResearch). Cell-associated fluorescence was determined by analysis using a FACSCanto II™ instrument (BD Biosciences). To assess the effect of mAbs or BsAbs on ligand binding, CD27 expressing Ramos cells (ATCC) or HEK293-PD-L1 cells were briefly pre-incubated with the mAbs, BsAbs or controls, followed by the addition of 0.5 µg/ml human CD70-biotin (US Biological) or 0.5 µg/ml PD-1-biotin (R&D Systems), respectively. Binding of biotinylated ligands was detected with streptavidin PE (SA–PE) and analyzed on a FACSCanto II™ instrument.
CD27 agonist assays
NFκB reporter assay
A stable cell line was developed from HEK293 NFκB-luciferase reporter cell line (Signosis, Inc.) transfected with human CD27. Cells were incubated with mAbs or BsAb for 6 h at 37 °C, 6% CO2. Luciferase was detected with the Luciferase Assay System (Promega).
Human T cell costimulation
Ninety-six well tissue culture plates were prepared by adding 1 µg/ml anti-CD3 mAb (OKT3- eBioscience), and/or 2 µg/ml recombinant human PD-L1 and coated overnight at 4 °C. After washing the wells with PBS, 100,000 CD3+ cells isolated by magnetic bead separation from peripheral blood mononuclear cells (PBMC) were added to each well in media. The plates were incubated for 3 days at 37 °C and 5% CO2, and supernatants were harvested and analyzed for IL-2 or IFN-γ production by ELISA (R&D Systems).
Cell-based PD-1 signaling assay
The effect of the BsAb on blockade of PD-1 signaling was performed per manufacturer’s instructions with a cell-based method in which blocking PD-1 signaling allows T cell receptor (TCR) activation and induces luminescence via the NFAT pathway (Promega). Luminescence was detected by the addition of Bio-Glo reagent and quantitated on a PerkinElmer Victor X luminometer.
Mixed lymphocyte reaction (MLR)
Human PBMCs were isolated from buffy coats using Ficoll separation, and CD4+ cells were further isolated using magnetic bead separation (Miltenyi). Monocyte-derived dendritic cells (DC) were generated from PMBCs by adhering to plastic and then cultured for 7 days in RPMI medium containing 10% FBS, 10 ng/ml IL-4 plus 100 ng/ml GM-CSF (R&D Systems). Cells were harvested and confirmed to be 80% DCs by expression of CD11c. The CD4+ cells and DCs from allogeneic donors were co-incubated at a 10:1 ratio in the presence of mAb or BsAb for 3 days. Supernatants were harvested and analyzed for IL-2 or IFN-γ production by ELISA (R&D Systems).
Antibody-dependent cellular cytotoxicity (ADCC)
ADCC activity was evaluated using a commercially available ADCC Reporter Bioassay Kit (Promega). This assay indirectly measures ADCC through quantitation of the FcγRIIIa receptor activity. Target cells included tumor cell lines Ramos (CD27+) and MDA-MB-231 (PD-L1+) (ATCC) or HEK293 cell lines expressing either CD27 or PD-L1.
Enhancement of vaccine-specific T cell responses
The huCD27-Tg mice [
19] received intraperitoneal (i.p.) administrations of 0.05 mg of BsAb or mAbs and 5 mg of ovalbumin (OVA) (Sigma-Aldrich). After 1 week, spleen cells were harvested and ELISPOT analysis was performed with and without 2 µg/ml SIINFEKL peptide (GenScript) incubated overnight in IFN-γ Ab-coated 96-well filtration plates (Sigma-Aldrich). Spots were developed by using an IFN-γ antibody set and a 3-amino-9-ethylcarbazole substrate (BD Biosciences) and counted by ZellNet Consulting, Inc.
Mouse tumor models
For the syngeneic lymphoma model, huCD27-Tg Balb/C mice were injected intravenously (i.v.) with BCL1 cells (1 × 106) on day 0, followed by i.p. injection of mAbs or BsAbs (0.2 mg) on days 5 or 7. Mice were observed daily for survival. Xenograft tumor studies were performed using the human B cell lymphoma cell line Raji (ATCC) with 0.5 × 106 cells implanted subcutaneously (s.c.) on day 0 into the flanks of SCID mice cells followed by i.p. administration of mAbs or BsAbs (0.1 mg) on days 5, 8, 12, 15, 19, and 22. Tumors were measured by caliper twice per week, and mice were euthanized according to pre-defined endpoint criteria.
Pilot non-human primate study
Three male cynomolgus monkeys (Citoxlab, Stilwell KS) received a single 7 mg/kg slow bolus i.v. injection (2–3 min) of CDX-527, via a cephalic catheter. Animals were followed for 21 days. Evaluations included clinical signs, body temperature, clinical pathology parameters (hematology, coagulation, clinical chemistry and urinalysis), and toxicokinetic parameters. Body weights were recorded once prior to BsAb administration and weekly thereafter. This was designed as a survival study with no planned necropsy.
Quantitation of CDX-527 concentration and anti-drug antibodies (ADA) was performed using a Mesoscale Discovery platform (MSD). For pharmacokinetics (PK), the plates were coated with a human CD27-Fc. The bound CDX-527 was detected by adding human PD-L1-msFc and a ruthenium-labeled SULFO-Tag F(ab’)2 (Mesoscale diagnostics) goat anti-mouse IgG (Fc-specific) followed by tripropylamine. The ADA assay used streptavidin-coated plates to capture biotinylated CDX-527. Serum samples were then added, and reactive antibodies were detected with ruthenium-conjugated CDX-527 and tripropylamine.
Statistical analysis
Statistical significance was evaluated using two-way ANOVA or paired Student’s t-test as appropriate. For tumor survival studies, the Mantel-Cox test was used.
Discussion
We developed a BsAb, designated CDX-527, based on the strong scientific rationale for coupling CD27 costimulation with PD-1 blockade, which included preclinical data demonstrating complementary and synergistic effects in tumor models [
22] and clinical data supporting safety and efficacy at the same dose and regimen for each antibody [
24,
25]. To generate the BsAb, we initially developed and characterized novel fully human antibodies to CD27 (mAb 2B3) and PD-L1 (mAb 9H9) that were primarily selected for CD27 agonist activity and inhibition of PD-1 signaling, respectively. The 9H9 antibody was also selected for its potent inhibition of CD80 binding to PD-L1, which may reduce the recently revealed cis-interactions between CD80 and PD-L1 on antigen presenting cells (APC) thereby resulting in more available CD80 for costimulation through CD28 [
28]. CDX-527 was designed as a tetravalent IgG1-scFv construct that is produced and purified by standard technologies applied for mAbs. As expected, CDX-527 retained the high-affinity binding to PD-L1 and to CD27, similar to the parental mAbs, and could engage each target simultaneously as demonstrated using a bifunctional ELISA.
Efficient antibody-mediated CD27 signaling requires receptor clustering optimally achieved by crosslinking the CD27 mAb, e.g., via FcγR binding [
29,
30]. Using a CD27-signaling reporter cell line that expresses low levels of PD-L1 and no FcR, we demonstrated that CDX-527 indeed was a more potent agonist than the parental CD27 mAb 2B3. The CD27 agonist activity of CDX-527 was further enhanced through FcR cross-linking, demonstrating that both PD-L1 and Fc receptor cross-linking can contribute to CD27 agonism with the BsAb.
The enhanced CD27 agonist activity of CDX-527 relative to mAb targeting was further demonstrated using in vitro T cell activation assays measured primarily by IL-2 production. Direct CD27 costimulation with the BsAb was achieved when human T cells were cultured on plates precoated with the anti-CD3 mAb OKT3 and PD-L1, but not achieved with the parental CD27 or PD-L1 mAbs even when combined. CDX-527 is efficiently cross-linked through binding to PD-L1, whereas the parental CD27 mAb is not cross-linked, and PD-1 blockade using the parental PD-L1 mAb, in the absence of T cell activation, does not have a measurable effect. As observed previously for CD27 costimulation [
20,
29], CDX-527 required both TCR stimulation and cross-linking to result in costimulation of T cells.
In MLR assays, addition of CDX-527 led to significantly greater T cell activation than the individual or combined CD27 and PD-L1 mAbs. Despite the presence of FcR on the APCs, the CD27 mAb did not enhance T cell activation either alone or in combination with the PD-L1 mAb. Presumably, in this setting the FcR-mediated cross-linking of the 2B3 mAb is suboptimal for good agonist activity. In contrast, PD-1 blockade with the PD-L1 mAb did enhance T cell activation in the MLR studies, as has been reported with other PD-(L)1 inhibitors [
31‐
33]. The greater activity of CDX-527 relative to the PD-L1 mAbs is most likely due to more efficient cross-linking of the BsAb that can be mediated through both FcR and PD-L1 interactions.
To study in vivo activity of the BsAb, we leveraged huCD27 transgenic mice [
19] and the murine cross-reactivity of the PD-L1 mAb, avelumab. The surrogate BsAb, AVEx2B3, demonstrated potent activity in vivo. As previously reported, CD27 agonist mAbs significantly boost the number of antigen-specific CD8 effector cells in an FcR-dependent manner when combined with a vaccine [
19] and the response is further increased in the context of PD-1/PD-L1 blockade [
22]. AVEx2B3 was more effective in boosting the CD8 effector T cell response to vaccine than either the CD27 or PD-L1 mAb, although its activity did not exceed the strong effect seen with the combination of the parental mAbs at the doses tested.
The superiority of the BsAb over the combination of CD27 and PD-L1 mAbs was clearly evident in a disseminated BCL1 lymphoma model. BCL1 expresses PD-L1 but not CD27 which would allow costimulation of T cells in the huCD27-Tg mice, enhancement of CD27 costimulation by crosslinking through PD-L1 in addition to FcR, and direct tumor killing through PD-L1. Other syngeneic tumor models in huCD27-Tg mice have shown that reductions in CD27-expressing Treg cells using varlilumab can also contribute to anti-tumor activity [
21]. We have previously shown that the efficacy of anti-CD27 therapy in the BCL1 model correlated with the level of CD27 agonism more strongly than Treg reductions [
21], suggesting that the additional cross-linking through PD-L1 enhanced CD27 signaling and antitumor activity of the BsAb.
In addition to the immune modulating activities of CD27 agonism and PD-1 blockade, the BsAb may have direct antitumor activity due to its capacity to bind FcR. We observed similarly potent ADCC with CDX-527 and the 9H9 mAb of PD-L1 expressing tumor cells, a mechanism that has been associated with antitumor activity for PD-L1 mAbs that can engage activating FcγRs [
34,
35]. Although the in vitro ADCC activity of CDX-527 was less potent than the 2B3 mAb, AVEx2B3 and 2B3 both similarly improved the survival of immunodeficient mice transplanted with the human Burkitt’s lymphoma line Raji that has high CD27 expression. The direct antitumor activity against Raji cells likely involves innate ADCC effector cells such as NK or macrophages [
36] but is unlikely to involve complement meditated cytotoxicity based on our in vitro data.
The inclusion of the intact human IgG1 Fc domain in the BsAb design was important for achieving mAb-like PK through FcRn binding. This was demonstrated in a small study performed in macaques that estimated the elimination half-life (t
1/2) for CDX-527 to be 5.31 days. In these animals, no obvious toxicity was observed, although the strong ADA response makes longer-term evaluations challenging in this species. Immunogenicity in non-human primates, however, is not considered predictive of immunogenicity in humans [
37]. Collectively, these data support CDX-527 as a promising approach to enhance the activity of PD-1 blockade, and activities to support evaluation of CDX-527 in cancer patients are underway. In addition to expanding the clinical benefits of PD-1 blockade, future combinations of CDX-527 with in situ vaccination strategies that induce immunogenic cell death [
38], including certain chemotherapies [
39] and radiation treatment [
40,
41], may further improve outcomes.
Acknowledgements
The authors would like to thank Shannon Renn-Bingham and Ioannis Papayannopoulos for their analytical expertise, Cate Swarce for protein purification, and April R. Baronas, Mallary Rocheleau, Kathleen Borrelli, Lindsay Dolan, Kshiti R. Patel and Amelia Fields for preclinical studies.
Compliance with ethical standards
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.