The CD137 receptor plays a key role in mediating immune response by promoting T cell proliferation, survival, and memory. Effective agonism of CD137 has the potential to reinvigorate potent antitumor immunity either alone or in combination with other immune-checkpoint therapies. In this study, we describe the discovery and characterization of a unique CD137 agonist, 7A5, a fully human IgG1 Fc effector-null monoclonal antibody. The biological properties of 7A5 were investigated through in vitro and in vivo studies. 7A5 binds CD137, and the binding epitope overlaps with the CD137L binding site based on structure. 7A5 engages CD137 receptor and activates NF-κB cell signaling independent of cross-linking or Fc effector function. In addition, T cell activation measured by cytokine IFNγ production is induced by 7A5 in peripheral blood mononuclear cell costimulation assay. Human tumor xenograft mouse models reconstituted with human immune cells were used to determine antitumor activity in vivo. Monotherapy with 7A5 inhibits tumor growth, and this activity is enhanced in combination with a PD-L1 antagonist antibody. Furthermore, the intratumoral immune gene expression signature in response to 7A5 is highly suggestive of enhanced T cell infiltration and activation. Taken together, these results demonstrate 7A5 is a differentiated CD137 agonist antibody with biological properties that warrant its further development as a cancer immunotherapy.
This article is featured in Highlights of This Issue, p. 973
Approaches to boost or restore effective antitumor T cell immune responses have included the targeting of immune coreceptors (also known as immune checkpoints) found on lymphocytes and that normally function to modulate T cell activation and function (1). Immune-checkpoint inhibitors (e.g., anti–PD-1 and anti–PD-L1) have demonstrated long-lasting responses in a subset of patients across many types of cancers (2). However, the majority of patients either do not respond or progress after an initial response to therapy (3). Mechanisms of resistance to immune therapy include T cell exhaustion, dysregulation of antigen presentation, or alterations in IFNγ signaling. Intense efforts are focused toward identifying new checkpoints or combination therapies in order to extend the impact of immunotherapy across a greater number of cancer patients. Initial strategies have targeted coinhibitory receptors such as CTLA-4 and PD-1/PD-L1 and have led to an enhanced therapeutic benefit across a number of tumor types (4). However, success with targeting costimulatory receptors, specifically with agonist antibodies, has proven to be more elusive. Clinical trials testing T cell agonist molecules, such as CD134 (OX-40, TNFRSF4), CD278 (ICOS), CD357 (GITR, TNFRSF18), and CD137 (4-1BB, TNFRSF9), have observed only modest clinical benefit (5–7), potentially reflective of the challenge in designing an optimal agonist that safely and effectively activates T cells.
CD137 is a member of the TNF receptor superfamily (TNFRSF; ref. 6) and its expression on T cells is tightly regulated in response to antigen stimulation. Productive engagement of CD137 by CD137L leads to rapid receptor activation and signaling through NF-κB and mitogen-activated protein kinase (ERK, JNK, and p38) pathways (8–12), and delayed activation of AKT signaling (13) that promotes T cell expansion with increased effector function, survival, and immunologic memory. CD137 is also expressed on other cell types, including dendritic cells (DC), regulatory T cells (Treg), natural killer (NK) cells, B cells, and activated endothelial cells (14, 15); and CD137 engagement on these cells has been shown to lead to enhanced suppression of Tregs, activation of antigen-presenting DCs, and cytotoxicity of NK cells (15).
CD137 ligand (CD137L, 4-1BBL, TNFSF9) is found expressed on antigen-presenting cells such as DCs and B cells, and in some nonimmune cells (cardiac myocytes, neurons); CD137 agonism has been shown to lead to metabolic changes that enhance T cell function and survival in the tumor microenvironment (16). The potential for engagement of the CD137 axis to enhance T cell function has also been illustrated through observations of enhanced T cell survival, persistence, and metabolic fitness for T cells engineered to express chimeric antigen receptors that contain CD137 cytoplasmic domain (17, 18). Preclinical mouse studies have further demonstrated the key immunomodulatory role of CD137 to enhance T cell anticancer immunity (19).
The collective preclinical data on the biology and functions of CD137 have precipitated an interest to develop effective therapeutics that target this costimulatory immune axis to boost antitumor immunity in cancer patients (20). To date, two anti-human CD137 agonist antibodies have been investigated in the clinic. Urelumab (Bristol-Myers Squibb) is a fully human IgG4 potent agonist antibody that demonstrated signs of activity in clinical trials, but its development was compromised due to liver toxicity in patients (20, 21). Lowering doses of urelumab mitigated the liver toxicity, but also dampened its clinical activity (22, 23). Utomilumab (Pfizer) is a fully humanized IgG2 weak CD137 agonist antibody that lacked any dose-limiting toxicities but had only modest activity when combined with PD-1 blockade (24–26). The crystal structures of the urelumab:CD137 and utomilumab:CD137 complexes have recently been reported and suggest potential mechanisms that may explain the observed differences of these agents in the clinic (27).
In this report, we describe the discovery and characterization of 7A5, a differentiated human CD137 agonist monoclonal antibody. 7A5 is derived from a human Fab phage display library, is engineered to an Fc effector-null phenotype, shows activity independently of Fcγ receptor engagement, and was functionally selected to productively and effectively engage the CD137 receptor. 7A5 binds human CD137 and blocks binding of CD137L, and the structural data reveal a binding epitope that overlaps with CD137L. 7A5 specifically binds to activated primary T cells and productively engages CD137 signaling, and leads to T cell activation in cell-based assays. In established xenograft immune-humanized mouse models, 7A5 treatment results in peripheral T cell activation and expansion, and antitumor efficacy accompanied by an intratumoral immune gene expression signature of T cell infiltration and activation. The combination of 7A5 with an antagonist PD-L1 antibody further enhances tumor growth inhibition. The preclinical characterization and activity of 7A5 support clinical development of this molecule as a differentiated agonist with the potential to enhance antitumor T cell immunity.
Materials and Methods
CD137 antibody 7A5 was derived from a human Fab phage display library (Eli Lilly and Company) panned against recombinant human extracellular CD137 protein (R&D Systems). The DNA sequence of 7A5 (WO 2019/027754) was cloned into expression vector with an IgG1 Fc effector-null backbone (28). Two light-chain DNA sequences were used that differ in coding of an N-terminal alanine. The heavy- and light-chain amino acid sequences are depicted in Supplementary Fig. S1. 7A5 Fab fragment DNA sequence was also subcloned into expression vector. CD137 reference antibodies BMS20H4.9 (US7288638) and MOR-7480.1 (US20130078240) were cloned into human IgG4 and IgG2 backbones, respectively. Antibodies were expressed in HEK-293 or CHO cells (ATCC) and purified by protein A chromatography (POROS A; Life Technologies), whereas Fab was purified using CaptureSelect IgG-CH1 Affinity Matrix (Thermo Fisher).
Recombinant protein expression
Human CD137 extracellular domain (ECD) protein was generated as described in Supplementary Materials and Methods.
Human CD137-Fc protein (R&D Systems) was immobilized onto 96-well plate (50 ng/well) overnight at 4°C and blocked with 0.2% BSA in PBS/0.05% Tween-20 (PBS/T) for 1 to 2 hours at room temperature. 7A5 or control human IgG1 was added and incubated for 1 to 2 hours at room temperature. Plates were washed, and HRP-conjugated goat anti-human IgG, F(ab')2 fragment-specific antibody (Jackson ImmunoResearch Laboratories) was added and incubated for 45 minutes at room temperature. After washing, TMB peroxidase substrate was added and absorbance measured at 650 nm using a microplate reader. Binding EC50 was determined with GraphPad Prism software (GraphPad Software).
The binding ELISA with human CD137-FLAG was run similarly; 50 ng/well was immobilized onto 96-well plate overnight at 4°C, blocked (5% BSA in PBS/T), and 7A5 or human CD137L (R&D Systems) added and incubated for 1 hour at room temperature. After washing, HRP-conjugated mouse anti-His antibody (Sigma) for CD137L binding or HRP-conjugated goat anti-Fab antibody (Jackson ImmunoResearch Laboratories) for 7A5 binding was added and incubated at room temperature followed by TMB peroxidase substrate. Absorbance readings were plotted, and EC50 values were calculated with GraphPad Prism software.
Human CD137-FLAG was immobilized onto 96-well plate as described above. CD137L was titrated and mixed with 7A5, added to wells, and incubated for 1 hour at room temperature. After washing, HRP-conjugated goat anti-Fab antibody was added, and TMB peroxidase substrate was used for signal detection. The percentage of antibody remaining bound to CD137 was plotted, and IC50 values were calculated using GraphPad Prism software.
The binding kinetics of 7A5 to human and cynomolgus monkey CD137-Fc proteins (R&D Systems, Eli Lilly and Company) was measured by surface plasmon resonance on a Biacore T200 instrument (GE Healthcare BioSciences) as described in Supplementary Materials and Methods.
Human CD137 expressing stable cells were generated in human HEK-293 cells (ATCC) by transfecting plasmid DNA pLVX-EF1a-IRES-Puro/hCD137 using Lipofectamine 2000 (Invitrogen) and selected in 0.5 μg/mL puromycin media (DMEM/10% FBS/1% glutamax). No additional cell line authentication was performed. CD137-293 cells plated into 96-well plate at 1 × 105 cells/well were stained with CD137 antibodies, Fab or control human IgG1 for 1 hour at 4°C. Secondary antibody R-phycoerythrin-conjugated goat anti-human IgG, F(ab')2 fragment-specific antibody (Jackson ImmunoResearch Laboratories), was added, and cells were incubated at 4°C for 30 minutes. Live/dead cell staining was performed with Far-Red LIVE/DEAD Fixable Dead Cell Stain (Life Technologies), processed on the IntelliCyt HTFC Screening System (IntelliCyt) and analyzed using FlowJo software (Tree Star).
Human peripheral blood mononuclear cells (PBMCs; NY Blood Center) stimulated for 5 days with Dynabeads Human T-Activator CD3/CD28 beads (Life Technologies) were plated into 96-well plate (1–2 × 105 cell/well). Commercial antibodies were from BD Biosciences (anti-CD3 339186, anti-CD4 560650, and anti-CD8 641409). CD137 antibodies and control IgG1 were added to cells and incubated for 1 hour at 4°C. Secondary antibody staining was with R-phycoerythrin–conjugated goat anti-human IgG, F(ab')2 fragment–specific antibody for 1 hour at 4°C. Live/dead cell staining was performed with Zombie Red Fixable Viability kit (BioLegend), acquisitions were performed on Fortessa X-20 (BD Biosciences), and analyzed with FlowJo software.
NF-κB reporter assay
Human CD137-293 cells were either transiently or stably transfected with pGL4.32[luc2P/NF-κB-RE/Hygro] plasmid DNA containing 5 copies of NF-κB response element (Promega; E8491) using Lipofectamine 2000. No additional cell line authentication was performed. Transiently expressed NF-κB luciferase reporter cells were incubated overnight at 37°C, then treated with 7A5 antibody or Fab and control IgG1 at 20 μg/mL for 5.5 hours at 37°C and processed using the ONE-Glo Luciferase Assay System (Promega). Luminescence signal was quantified using a plate reader, and data were analyzed with GraphPad Prism software. Stable NF-κB luciferase reporter cells were selected with 100 μg/mL hygromycin and 0.5 μg/mL puromycin containing media. Cells were plated in a 384-well plate using the Thermo MultiDrop Combi Reagent Dispenser (Thermo Fisher Scientific) and cultured overnight at 37°C. CD137L, CD137 antibodies, and control human IgG1 dilutions in PBS were prepared with Hamilton Star (Hamilton Company) or manually, 10-point 3-fold thru-plate dilution starting at 1 μmol/L (except 7A5 at 7.184 μmol/L), transferred and incubated with antibodies for 5.5 hours at 37°C. Cells were processed, luminescence signal was measured as described above, and data were analyzed with GraphPad Prism.
Western blot for TRAF2 phosphorylation
Human CD137-293 cells were treated with 7A5 or control human IgG1 as indicated at 37°C. Cell lysates were prepared in NP-40 lysis buffer (Life Technologies) with HALT protease and phosphatase inhibitors (Thermo Scientific), analyzed on 4% to 12% Bis-Tris NuPage gel, and protein was transferred onto nitrocellulose membrane with iBlot system (Life Technologies). The blot was incubated with primary antibodies: phospho-TRAF2 or TRAF2 (Cell Signaling Technology; 13908, 4724) or glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Life Technologies; AM4300). Secondary antibody–HRP conjugates were used and enhanced chemiluminescence (ECL) signal detected with ImageQuant LAS 4000 digital imaging system (GE Healthcare BioSciences).
PBMC costimulation assay
Human PBMCs were isolated from whole blood or leukopak obtained from NY Blood Center, AllCells, or Biological Specialty using Ficoll density gradient centrifugation (Ficoll-Paque PLUS; GE Healthcare) and grown in RPMI (Life Technologies) with Indicate as 10% FCS (HyClone Laboratories). PBMCs were plated (1.5 × 105 cells/well) onto CD3 antibody (BD Biosciences, clone HIT3) coated 96-well plate (2 ng/well). CD28 antibody (BioLegend, 302933) was added to the plate (0.4 μg/mL) followed by test antibodies and incubated for 96 hours at 37°C. Human IFNγ levels were detected using the DuoSet ELISA Kit (R&D Systems, DY285), and data were analyzed with GraphPad Prism software.
Treg suppression assay
CD4+ T cells were isolated from human whole blood (NY Blood Center) using the RosetteSep Human CD4+ T cell Enrichment kit (STEMCELL Technologies) and Ficoll density gradient centrifugation. Mononuclear cells were collected and treated with ACK Lysis Buffer (GIBCO). Human CD25+ Treg and CD25− T effector cells were purified with the Miltenyi Biotec CD25+ Microbeads and LS columns. T effector cells were labeled with CFSE (Celltrace CFSE proliferation kit, Thermo Fisher) at 2 μmol/L final and combined with Tregs at 4:1 ratio (4 × 104 cells: 1 × 104 cells) or cultured alone (5 × 104 cells) in triplicate onto 96-well round-bottom plates. Miltenyi Treg Suppression Inspector Beads were added at 1 bead per cell. CD137 antibodies or control IgG1 were added to the cells and incubated for 5 days at 37°C. Cells were analyzed for CFSE dilution by flow cytometry on Fortessa X-20.
Antibody-dependent cellular cytotoxicity (ADCC) effector function assay
CD137-positive or -negative HEK-293 cells and FcγRIIIa-positive Jurkat cells containing NFAT luciferase reporter (Eli Lilly and Company) were utilized to measure 7A5 effector function as described in Supplementary Materials and Methods.
In vivo efficacy studies
All animal research methods and studies were approved by the Institutional Animal Care and Use Committee and performed in accordance with current regulations and standards of the United States Department of Agriculture and the National Institute of Health. All experiments with engrafted human PBMCs or expanded T cells were performed in female NOD/SCID Gamma (NSG) mice at 7 weeks age (Jackson Laboratories). Mice were housed in a 12-hour light/dark cycle facility under pathogen-free conditions in microisolator cages with standard laboratory chow and water ad libitum. Body weight and tumor volume were measured twice weekly. Tumor volume was calculated using the formula [tumor volume (mm3) = π/6 × length × width2] and plotted as geometric means ± standard error of the mean (SEM). The %T/C was calculated as described in Supplementary Materials and Methods. Statistical analysis of tumor volume data was performed with a two-way repeated-measures analysis of variance by time and treatment using the MIXED procedures in SAS software (Version 9.2). A Bliss independence analysis was performed to determine if combination treatment tested was additive or greater than additive or less than additive as compared with either single agent.
Established tumor models
HCC827 human NSCLC tumor cells (ATCC) were injected subcutaneously into the right flank of female NSG mice (10 × 106 cells). When tumors reached approximately 250 mm3 to 400 mm3 in size, mice were randomized into groups, and human expanded T cells (3 × 106 cells) were injected intravenously (i.v.). As control, tumor cells alone were implanted with no T cells. Treatment groups included human IgG1, 7A5, anti–PD-L1 LY3300054 (28), and the combination of 7A5 + anti–PD-L1 antibodies, all dosed intraperitoneal (i.p.) as indicated. Tumor volumes were measured twice per week using electronic calipers.
Preventative tumor models
Freshly isolated human PBMCs were combined with NCI-H292 human NSCLC tumor cells (ATCC) at a 1:4 ratio of PBMCs to tumor cells (2.5 × 106 PBMC and 10 × 106 NCI-H292 cells). Each mouse was implanted subcutaneously on the right flank with 0.2 mL of the solution on day 0. One control group receiving tumor cells alone was included in each study. Mice were randomly assigned to treatment groups that included human IgG1, 7A5, anti–PD-L1, and combinations of 7A5 + anti–PD-L1 antibodies, all dosed i.p. as indicated. Tumor volumes were measured twice per week using electronic calipers.
The mechanism of action in vivo study is described in Supplementary Materials and Methods. Snap-frozen tumors were lysed with the MagMAX-96 Total RNA isolation kit (Life Technologies) and homogenized with TissueLyser (Qiagen). Total RNA was isolated using MagMAX Express-96 Deep Well Magnetic Particle Processor (Life Technologies). Sample total RNA (500 ng) was analyzed in duplicate for gene expression by a custom designed QuantiGene 2.0 plex assay (Affymetrix) and analysis performed on FLEXMAP 3D Luminex instrument (Thermo Fisher). Mean fluorescence intensity (MFI) data were converted to relative gene expression (normalized adjusted net MFI) using a quality control analysis script (Eli Lilly and Company; ref. 29).
Discovery and binding properties of CD137 monoclonal antibody 7A5
CD137 is a TNFRSF-superfamily member that forms a trimeric complex with its ligand. We speculated that anti-CD137 antibodies with lower binding potency and the ability to function as monovalent, effectorless antibody fragments might produce a differentiated approach to developing receptor agonists. To identify anti-CD137 antibodies that functioned in the absence of receptor cross-linking, we screened a human Fab phage display library for Fab capable of binding to CD137 using recombinant human extracellular CD137 protein as antigen, and then evaluated this collection for the ability to bind human CD137 on cells. We identified 73 Fabs and further characterized these antibodies before selecting 7A5 based on the in vitro biophysical and functional activity in the absence of cross-linking. 7A5 was cloned into human IgG1 effector-null backbone to minimize the potential for ADCC and complement-dependent cytotoxicity (CDC)-mediated targeting of activated CD137-expressing lymphocytes.
In ELISA assays, 7A5 binds immobilized human CD137 in a dose-dependent manner with an EC50 of 0.017 nmol/L (Fig. 1A). 7A5 also demonstrated binding to cynomolgus monkey CD137 protein in ELISA (EC50 of 0.011 nmol/L) but not to mouse (Supplementary Fig. S2). Biacore analysis revealed similar binding affinity to human and cynomolgus monkey CD137, with KD value of 5.36 nmol/L (kon = 1.33 × 106 M−1 s−1, koff = 7.13 × 10−3 s−1) and 6.12 nmol/L (kon = 3.59 × 105 M−1 s−1, koff = 2.19 × 10−3 s−1), respectively.
Flow cytometry analysis was performed to examine the ability of 7A5 to bind surface CD137. Using human CD137-293 cells, 7A5 IgG1 and Fab were shown to bind surface CD137 with 7A5 Fab binding at lower levels (Fig. 1B); no specific binding of 7A5 was observed on 293 cells. We next evaluated the ability of 7A5 IgG1 and Fab to trigger CD137 signaling using 293 cells expressing CD137 and an NF-κB–driven luciferase reporter transgene; both 7A5 IgG1 and Fab activated luciferase expression, with 15-fold and 3-fold increase relative to control IgG1 (Fig. 1C).
To evaluate the lack of Fc effector function of 7A5, we applied a panel of Fcγ receptor binding assays to assess FcγRI, FcγRIIa(H), FcγRIIb, FcγRIIIa(F), and FcγRIIIa(V) activities. No specific binding to Fcγ receptors was observed for the effectorless 7A5, whereas the effector competent version of 7A5 and rituximab control showed binding to Fcγ receptors (Supplementary Table S1). In ADCC assays, using CD137-293 cells and FcγRIIIa-positive Jurkat cells cultured with 7A5 showed no NFAT luciferase reporter activation relative to controls (Fig. 1D; Supplementary Table S2). Similarly, no 7A5-mediated binding to human complement C1q was observed by ELISA or cell lysis in CDC assays (Supplementary Tables S1 and S2).
In vitro functional activity of 7A5
Engagement of CD137 by CD137L triggers receptor signaling cascade that is composed of NF-κB and other pathways that lead to T cell activation and effector function responses (8, 14). To evaluate the functional properties of 7A5, we performed a series of in vitro studies using CD137-293 cells and primary activated human T cells. We also evaluated the functional activity of BMS20H4.9, the parent antibody of the potent CD137 agonist urelumab. 7A5 binding to CD137-293 cells was considerably lower in intensity compared with BMS20H4.9 (Fig. 2A). To evaluate the ability of 7A5 to functionally signal through CD137, we used CD137-293 cells expressing an NF-κB–driven luciferase reporter transgene and evaluated 7A5, BMS20H4.9, MOR-7480.1 (the parent antibody of utomilumab), as well as CD137L and an irrelevant IgG1 control (Fig. 2B). Although BMS20H4.9 and MOR-7480.1 demonstrated potent and very weak agonist activity in these assays (AC50 = 2.39 nmol/L and >100 nmol/L), 7A5 showed agonist activity relatively similar to CD137L (AC50 = 24.30 nmol/L and 29.30 nmol/L).
CD137 and members of the TNFRSF recruit to their cytoplasmic domain TNFR-associated factor adaptor proteins to form intracellular signaling complexes that contribute to the signalosome (9, 30). TRAF2, an E3 ubiquitin ligase that mediates K63 ubiquitination, links CD137 to NF-κB and other signaling pathways. To evaluate whether 7A5 binding to CD137 resulted in TRAF2 recruitment, we measured TRAF2 phosphorylation at Ser11 in CD137-293 cells. Treatment with 7A5 increased the phospho-TRAF2 protein in a dose-dependent manner with no changes in the amount of TRAF2 or GAPDH protein across treatments (Fig. 2C). We also examined the CD137 receptor internalizing properties of 7A5 in CD137-293 cells; 7A5 did not induce internalization of human CD137 under the assay conditions (Supplementary Fig. S3).
To evaluate the ability of 7A5 to bind primary human T cells, we utilized PBMC cultures activated by CD3/CD28 antibody–coated beads. Both 7A5 and BMS20H4.9 bound to CD8 and CD4 T cells, with BMS20H4.9 showing binding saturation throughout the titration range and 7A5 showing a dose-dependent increase, with CD8 T cell binding being more robust (Fig. 2D). Activation of T cells was evaluated in a PBMC costimulation assay using suboptimal plate-bound CD3 and CD28 and measuring enhancement of cytokine IFNγ production. Both 7A5 and BMS20H4.9 enhanced T cell activation dose dependently as measured by IFNγ production relative to control human IgG1, with BMS20H4.9 demonstrating more robust activation (Fig. 2E). To assess the impact of 7A5 on Treg function, we evaluated its effect on Treg-suppressive activity; 7A5 treatment inhibited Treg suppression of T effector proliferation in a dose-dependent manner, whereas BMS20H4.9 inhibited Treg suppression at all the tested doses (Fig. 2F). Finally, stimulation of T cells with 7A5 in a different assay increased the frequency of CD8-positive memory T cells in vitro (Supplementary Fig. S4).
The 7A5 epitope overlaps the CD137L binding site on CD137
To understand the molecular mechanism by which 7A5 interacts with human CD137 we utilized high resolution X-ray crystallography structural data in addition to ELISA based binding and blocking of the native complex. Figure 3A shows the structure of the 7A5 variable region (shades of red cartoon) in complex with CD137 (wheat surface). The interacting epitope of 7A5 overlaps with the CD137L binding site (blue) as extrapolated from the recently reported structure of human CD137L:CD137 complex (PDB:6BWV; ref. 31). Given these observations, we examined the ligand blocking properties of 7A5 to gain additional mechanistic insights using solid surface phase assays. Because ligand binding to CD137 is of higher affinity compared with that of 7A5 (Fig. 3B, inset), we titrated ligand concentration in the blocking competition ELISA with 7A5. Figure 3B shows 7A5 blocked formation of the CD137L–CD137 complex at IC50 = 0.40 nmol/L. These findings of ligand blocking function support the internal crystallography data in Fig. 3A.
Biological activity of 7A5 in humanized murine models in vivo
To evaluate the ability of 7A5 to enhance T cell antitumor activity, we performed studies in immune-efficient NSG animals reconstituted with human immune cells and implanted with human tumors. These studies included a preventive or coimplantation model and a therapeutic model with established tumors, each reconstituted with allogeneic human PBMCs and T cells, respectively. These models are based on the ability to modulate alloreactivity of engrafted human immune cells against the tumor and measure peripheral and intratumoral immune cell modulation. PD-L1 therapy has shown antitumor efficacy in these models when using PD-L1–positive tumor cell lines as xenografts (28). We therefore proceeded to examine the antitumor activity of 7A5 in an established HCC827 human NSCLC tumor xenograft mouse model reconstituted with allogeneic human expanded T cells. There was a dose-dependent response to 7A5 therapy with significant tumor growth inhibition at 3, 10, and 30 mg/kg doses (T/C% = 41, P = 0.005; T/C% = 13, P < 0.001; T/C% = 22, P < 0.001) but not at 1 mg/kg (T/C% = 66, P = 0.141) relative to control human IgG1. Figure 4A shows the tumor volumes for 7A5 in the 10 and 1 mg/kg dose arms and positive control anti–PD-L1 antibody. There were no observed body weight changes (Supplementary Fig. S5A).
The therapeutic potential of 7A5 in combination with an immune-checkpoint inhibitor, anti–PD-L1 antibody LY3300054 (28), was evaluated in the established HCC827 tumor model. As shown in Fig. 4B, 7A5 monotherapy treatment demonstrated enhanced alloreactivity resulting in potent antitumor activity relative to control human IgG1 (T/C% = 13, P < 0.001) and was similar to that of anti–PD-L1 antibody monotherapy (T/C% = 8, P < 0.001). The combination therapy of 7A5 and anti–PD-L1 antibodies demonstrated additional efficacy, resulting in tumor regression (−45%, P < 0.001) that was synergistic as determined by the Bliss independence method. Treatments were well tolerated with no body weight changes observed (Supplementary Fig. S5B).
We also explored the antitumor efficacy of 7A5 in the preventative NCI-H292 human NSCLC tumor xenograft mouse model that coimplanted human PBMCs and NCI-H292 tumor cells. As shown in Fig. 4C, compared with control human IgG1 therapy, 7A5 monotherapy resulted in tumor growth inhibition at the 1 mg/kg dose (T/C% = 39%, P = 0.017) but was attenuated at the higher dosage. We observed 7A5 efficacy at the higher dose in a separate study (Supplementary Fig. S6), which suggests different donor PBMCs may provide varying levels of immune cell engraftment and response in this model. Anti–PD-L1 monotherapy treatment at 0.25 mg/kg had a similar antitumor effect (T/C% = 31%, P = 0.003). The combination of 7A5 and anti–PD-L1 showed a trend of an enhanced antitumor effect relative to 7A5 or anti–PD-L1 alone (T/C% = 13%: P < 0.001 vs. 7A5, P = 0.052 vs. anti–PD-L1). Body weight changes were not observed in response to the treatments (Supplementary Fig. S5C). In some in vivo models, CD3 immunohistochemistry of tumors from animals treated with 7A5 indicates enhanced T cells as shown in Supplementary Fig. S7 for the L55 established tumor model, and further examination will be informative in the HCC827 and NCI-H292 models.
Mechanistic analysis of 7A5 in vivo: intratumor immune gene expression
To explore in more detail the mechanism of action of 7A5 in the in vivo studies, we evaluated intratumor pharmacodynamic changes for engrafted human immune cells during 7A5 treatment. Tumor samples from the NCI-H292 tumor model coimplanted with human PBMCs (Supplementary Fig. S8) were assessed for immune-related gene expression changes in response to 7A5, anti–PD-L1 antibody LY3300054, or human IgG1 using a custom-designed QuantiGene Plex RNA panel. Figure 5A shows 7A5 treatment induced significant changes to the immune gene signature in tumors (fold difference ≥2-fold, P ≤ 0.05) compared with control human IgG1. There was predominantly upregulation of multiple immune-related genes, including transcripts associated with T cell infiltration, activation, and effector function (e.g., CD3E, CD8B, IFNG, GZMB, PRF1, TIGIT, LAG3, and TBX21 [TBET]). Genes indicative of other immune cell type infiltration and function also increased following 7A5 treatment, such as myeloid (ITGAM [CD11B], ITGAX [CD11C]), B cell (CD19), and NK cell (NCAM1)-associated transcripts. Changes to the immune gene signature were also observed in response to anti–PD-L1 antibody, as reported previously (28). Notably, the gene expression profile following 7A5 treatment overlaps but is distinct from that of anti–PD-L1 (Fig. 5B); there was a greater number of immune genes upregulated in tumors by agonist 7A5 relative to the response of anti–PD-L1. Only two genes from the panel, TGFB2 and IL8, were downregulated by 7A5 and anti–PD-L1 antibody, respectively.
In this report, we describe 7A5, a fully human antibody that functions as an agonist to CD137. 7A5 binds an epitope on CD137 that overlaps with the binding site of CD137L and has been engineered on an IgG1 backbone to lack detectable Fc receptor binding and effector function (ADCC) to avoid Fc-mediated receptor clustering. In cell-based in vitro assays, 7A5 triggers CD137 receptor and pathway activation, and enhances T cell activation in response to suboptimal CD3/CD28 antibody stimulation. In vivo, 7A5 mediates potent antitumor activity in humanized tumor xenograft mouse models reconstituted with human immune cells (PBMCs or T cells), and this activity is further enhanced by combination with anti–PD-L1 therapy. High-content molecular analysis of tumors from 7A5-treated animals demonstrates that 7A5 therapy induces an immune gene signature consistent with T cell activation and tumor infiltration, properties that likely contribute to its antitumor mechanism.
Although immunotherapy approaches such as checkpoint inhibition, cell therapy, and redirection have provided significant benefit to subsets of cancer patients, there remains a critical need to develop additional strategies that enhance T cell function and expand the breadth of patients who benefit from these therapies. Because T cell costimulatory receptors play key roles in the activation, expansion, and effector functions of T cells, approaches to boost T cell antitumor activity through costimulatory receptor agonism have been developed. Clinical programs that have evaluated agonist antibodies that target CD28, OX-40, GITR, ICOS, and CD137 have demonstrated the potential and challenges for targeting T cell agonism (5–7).
Productive engagement of costimulatory T cell receptors likely requires the application of agonists with optimal, rather than most potent, functional properties to prevent T cell overactivation. The clinical experience with CD137 is instructive in this regard, with urelumab (BMS-663513), a very potent agonist of CD137, demonstrating toxicities at pharmacodynamically acceptable doses, whereas utomilumab (PF-05082566), a weak agonist of CD137, demonstrated limited toxicity, but lacked monotherapy activity in early trials (20). An additional complexity with targeting TNFRSF members such as CD137 is that productive receptor engagement typically requires oligomerization and receptor clustering (32), and bivalent antibodies may not effectively enable such native or ligand-like activity; recent literature indicates that TNFRSF stimulation can be strongly influenced by antibody properties such as binding epitope, isotype, Fc effector function, and cross-linking requirements (33–35).
We hypothesized that an optimally active CD137 agonist antibody would mimic the activity of CD137L in terms of binding and activation of T cells, and that CD137 binding antibodies could be identified that productively agonized CD137 without the requirement for oligomerization and receptor clustering. Our data show that 7A5 antibody mimics the activity of CD137L in in vitro assays and functionally engages the CD137 receptor as an IgG1 Fc effector-null antibody agonist without the need for effector function or antibody cross-linking. Structural analysis of 7A5 binding to CD137 shows epitope overlap with CD137L and a binding mode unique from that of the recently described urelumab and utomilumab epitopes (27). Urelumab is a ligand nonblocker and its epitope is nonoverlapping with the ligand, whereas the utomilumab epitope is adjacent to the ligand and reported to sterically hinder ligand interaction. We demonstrate 7A5 completely blocks the binding of ligand to CD137 and the binding epitope is embedded within the ligand binding site, further supporting its distinction.
Binding of 7A5 triggered sufficient CD137 receptor axis activation to productively engage T cells in vitro. Furthermore, this activity translated into significant in vivo antitumor efficacy across several immune-humanized mouse models. Insights into the potential mechanism of in vivo efficacy were obtained by examining the tumor immune gene signature in response to 7A5 therapy. We observed significant upregulation within tumor tissues of multiple genes functioning in T cell activation and infiltration. In addition, we observed increased intratumoral expression of genes corresponding to innate immune cell types such as myeloid, B cell, and NK cell. These findings are concordant with the immunostimulatory roles that CD137 has on T cells and other immune cell types (15) and suggest that 7A5 activation of the CD137 axis has the potential to promote broader immune changes to the tumor environment. Furthermore, these data provide a starting foundation for biomarker consideration and evaluation in the clinic. Finally, combination of 7A5 with anti–PD-L1 therapy resulted in enhancement of the antitumor effect relative to each of the monotherapies, as predicted for the combination of effective T cell checkpoint and costimulation-targeting agents. Preliminary observations in a less responsive PD-L1 model (L55 tumor xenograft; Supplementary Fig. S9) also demonstrate a 7A5 and PD-L1 combination benefit that warrants further study. Combination findings were also observed with an anti–PD-1 antibody in humanized mouse models (Supplementary Fig. S6), which supports targeting both PD-1 and CD137 pathways and exploring potential combination with other checkpoint agents including anti–CTLA-4 antibody.
In summary, we describe the identification and characterization of a human CD137 monoclonal antibody with differentiating agonist functional properties in vitro that translate to in vivo antitumor efficacy and immunologic response. The preclinical characterization and activity of 7A5 support clinical development of this molecule as a differentiated agonist with the potential to enhance antitumor T cell immunity in the clinic.
Disclosure of Potential Conflicts of Interest
H. Kotanides is a senior research advisor at Eli Lilly and Company and has ownership interest (including patents) in the same. R.M. Sattler is a principal research associate at Eli Lilly and Company. J. Li is a research associate at Eli Lilly. J.N. Haidar is a research advisor at Eli Lilly and Company. C. Burns is a senior principal research associate at Eli Lilly & Company. A. Forest is a senior principal research associate at Eli Lilly. X. Chen is a principal research associate scientist at Eli Lilly. M. Topper is a principal research associate at Eli Lilly and Company and has ownership interest (including patents) in the same. L. Boucher is a senior research associate at Eli Lilly and Company. R.D. Novosiadly has ownership interest (including patents) in Eli Lilly and BMS shares. D.L. Ludwig is a former employee VP Research Biologics Technology at Eli Lilly and Company. G.D. Plowman is VP oncology research at Eli Lilly. M. Kalos is CSO at Eli Lilly and Company. No potential conflicts of interest were disclosed by the other authors.
Conception and design: H. Kotanides, J.N. Haidar, D.L. Ludwig, G.D. Plowman, M. Kalos
Development of methodology: H. Kotanides, R.M. Sattler, C. Carpenito, J. Shen, J.N. Haidar, C. Burns, L. Shen, A. Forest, X. Chen, M. Topper, Y. Zhang, D. Burtrum, R.D. Novosiadly, M. Kalos
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.M. Sattler, M.B. Lebron, C. Carpenito, J. Li, D. Surguladze, J.N. Haidar, C. Burns, L. Shen, I. Inigo, A.L. Pennello, A. Forest, X. Chen, D. Chin, A. Sonyi, M. Topper, L. Boucher, Y. Zhang
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Kotanides, R.M. Sattler, C. Carpenito, J. Li, D. Surguladze, J.N. Haidar, C. Burns, L. Shen, I. Inigo, A.L. Pennello, A. Forest, X. Chen, A. Sonyi, M. Topper, L. Boucher, P. Sharma, R.D. Novosiadly, D.L. Ludwig, M. Kalos
Writing, review, and/or revision of the manuscript: H. Kotanides, R.M. Sattler, M.B. Lebron, J. Shen, J.N. Haidar, C. Burns, L. Shen, A.L. Pennello, A. Forest, X. Chen, A. Sonyi, M. Topper, D.L. Ludwig, G.D. Plowman, M. Kalos
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.N. Haidar, A. Forest, M. Topper, R.D. Novosiadly
Study supervision: H. Kotanides, C. Carpenito, M. Kalos
Other (identification of 7A5): J. Shen
We thank Yiwen Li, Krishnadatt Persaud, Sagit Hindi-Jacoel, Timothy D. Bailey, Timothy R Mack, Christopher Mark Moxham, and Gerald E. Hall for technical assistance and helpful discussions; Thompson N. Doman and Jason R. Manro for computational and statistical analysis; Stephen Antonysamy, Margaret C. Kearins, and J. Michael Sauder for coordinates of the 7A5–Fab complex with CD137; Carl June (University of Pennsylvania) for providing the L55 cell line. This research was supported by Eli Lilly and Company.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.