Checkpoint blockade therapies targeting PD-1/PD-L1 and CTLA-4 are clinically successful but also evoke adverse events due to systemic T-cell activation. We engineered a bispecific, mAb targeting CD28 homolog (CD28H), a newly identified B7 family receptor that is constitutively expressed on T and natural killer (NK) cells, with a PD-L1 antibody to potentiate tumor-specific immune responses. The bispecific antibody led to T-cell costimulation, induced NK-cell cytotoxicity of PD-L1–expressing tumor cells, and activated tissue-resident memory CD8+ T cells. Mechanistically, the CD28H agonistic arm of the bispecific antibody reduced PD-L1/PD-1–induced SHP2 phosphorylation while simultaneously augmenting T-cell receptor signaling by activating the MAPK and AKT pathways. This bispecific approach could be used to target multiple immune cells, including CD8+ T cells, tissue-resident memory T cells, and NK cells, in a tumor-specific manner that may lead to induction of durable, therapeutic antitumor responses.
Immune checkpoint inhibitor (ICI) treatment targeting CTLA-4 (1), PD-1 (2), and PD-L1 are breakthrough therapies that result in durable, long-term responses in patients with cancer (3). Despite the phenomenal clinical success of ICI therapies, responses are limited to a patient subset, and a significant number of patients have disease that is resistant to single-agent therapy. ICI therapy is also accompanied by the risk of fatal immune-related adverse events due to systemic targeting of T cells. ICI combination therapies have been quickly adopted to overcome resistance and have been shown to achieve greater overall response rates due to different modalities of action of CTLA-4 and PD-1 (4, 5). However, this success is also accompanied by higher incidents of immune-related adverse events (4). This varied response to ICI therapies highlights the need to target alternative pathways with novel targeting platforms to mediate antitumor immunity. One approach is to target T-cell costimulators as a monotherapy or in combination with ICIs to elicit antitumor responses (5).
Targeted combination therapy with a bispecific antibody is an advantageous strategy to overcome systemic responses by pairing tumor-associated antigens in the context of T-cell engagement (6). The majority of bispecific antibodies in clinical trials are bispecific T-cell engagers consisting of two variable-domain antibody fragments, one of which typically targets CD3 coupled to tumor-associated antigen–specific targeting arms (7, 8). Here, we describe the generation, functional effects, and mechanism of action of a tetravalent bispecific antibody targeting PD-L1 in combination with an agonistic antibody to the B7 family receptor, CD28 homolog (CD28H; a.k.a TMIGD2/IGPR-1).
CD28H is a costimulatory receptor that shares close homology with CD28 and has a broad expression on adaptive and innate immune cells and mediates cell adhesion and barrier functions on endothelial cells (9–11). CD28H is expressed on a majority of peripheral-tissue T cells, natural killer (NK) cells, plasmacytoid dendritic cells (pDC), and innate lymphoid cells (ILC; refs. 9, 12) and functions as a costimulatory receptor on T cells (9). In peripheral tissues, as well as in certain tumors, CD28H expression has been observed on tissue-resident memory (TRM) CD8+ T cells (13–15). A recent report indicates that CD28H activation on NK cells synergizes with activating NK receptors NKp46 and 2B4 to induce targeted killing through antibody-dependent cellular cytotoxicity (ADCC; ref. 16). B7H7 (HHLA2), the cognate ligand for CD28H, mediates T-cell costimulation (9) and NK cell–mediated lysis of B7H7-expressing target cells (16). B7H7 is expressed primarily on macrophages, and normal epithelial cells in some tissues, and is also known to be overexpressed on many tumors (9, 17, 18).
Here, we showed that a bispecific CD28H/PD-L1 antibody (referred to here as “BiS-Ab”) increased activation of primary human T cells in an antigen-specific manner when cross-linked with PD-L1 expressed on tumor targets. While activation of the CD28H pathway promoted the formation of CD8+ TRM cells, NK cells stimulated with BiS-Ab displayed significantly higher cytotoxicity against PD-L1–expressing tumor cells than did anti-CD28H or anti–PD-L1 alone. BiS-Ab–induced activation of T cells in the context of PD-L1 required an intact CD28H signaling domain and phosphorylation of intracellular tyrosine motifs. Signaling studies indicated that anti-CD28H induced SHP2 dephosphorylation in the presence of ongoing PD-L1–PD-1 engagement in T cells. Our results indicated that BiS-Ab activated antigen-specific T-cell responses in the context of PD-L1 cross-linking, potentiated NK cell–directed killing of PD-L1–expressing tumor cells and induced the generation of CD8+ TRM cells. This multipronged immune activation indicates that this bispecific strategy has therapeutic potential to activate tumor-infiltrating lymphocyte (TIL) responses.
Materials and Methods
Patient and healthy donor samples and approvals
All patient samples [dissociated tissue and tumor cells (DTC) and peripheral blood mononuclear cells PBMC] were sourced from Conversant Bio (now, Discovery Life Sciences). The DTCs (2–3 vials of 1–5e6/mL, >60% viability) were generated by a proprietary mechanical and enzymatic digestion and were obtained as frozen single cells and stored in liquid nitrogen until use: 15 colorectal carcinomas (adenocarcinoma), 13 lung cancers (non–small cell, adenocarcinoma, squamous, and large cell carcinoma), 1 pancreatic cancer, 4 kidney cancers (clear cell and spindle cell carcinoma), 3 ovarian cancers, and 4 breast cancers were obtained from Discovery Life Sciences. All human samples (PBMCs, NK cells, T cells) used in this study were obtained from AstraZeneca employee volunteers who provided written informed consent in accordance with Institutional Review Baord–approved AstraZeneca protocol 2010-001 (CRRI 1003006) and the AstraZeneca Gaithersburg Research Blood Program (Pro00004341). All the cells were isolated from fresh blood using MACS negative selection kits (Miltenyi Biotec) by the Astrazeneca cell services facility. The human tonsil samples were procured from National Disease Research Interchange (Philadelphia, PA).
Isolation of human T cells and NK cells
All human cells were isolated from fresh peripheral blood of healthy donors (AstraZeneca Gaithersburg Research Blood Program) by negative selection using the EasySep Human Enrichment Kits (catalog no. 19051, total T cells, catalog no. 19055 for total T cells) and a Robosep machine (StemCell Technologies, StemCell Robosep S model). The purity was approximately 98% for both NK and T cells as confirmed by flow cytometry.
The human glioblastoma cell line U-87 MG (ATCC HTB-14) and the mouse mastocytoma cell line P815 (ATCC TIB-64) were both maintained in DMEM (Gibco), supplemented with 10% FCS (Gibco), penicillin (100 U/mL; Gibco), and streptomycin (100 μg/mL; Gibco). The NFAT/PD-1 Jurkat cells line (Clone 3LB9) was obtained from Promega and grown in RPMI1640 (ATCC) with 10% FBS (Gibco), hygromycin B (100 μg/mL, Invitrogen), and G418 (0.5 mg/mL, Thermo Fisher Scientific) for a maximum of five passages. The human CD28H gene (TMIGD2) was cloned into the lentivirus vector pCDH-puro (Systems BioSciences) and used to generate lentivirus. Jurkat PD-1 NFAT reporter cell lines were grown in RPMI1640 + 10% FBS + 0.5 mg/mL G418, transduced with human CD28H gene containing lentivirus described above at 1 × 108 PFU/mL, and selected with puromycin (2 μg/μL, Thermo Fisher Scientific) for a week. Jurkat PD-1 NFAT reporter cells were also engineered to express a signaling-deficient version of CD28H. The intracellular signaling domain–mutated CD28H (CD28H-ΔICD) was generated by removing the intracellular portion of CD28H and fusing it with an SNKL tag encoding FLAG and Rho-1D4 tags; specifically, CD28H residues M1-W171 were fused to SNKL. The CD28H TYRTM was generated by mutating all the intracellular domain (ICD) tyrosines to phenylalanines, specifically at residues Y192F, Y197F, and Y222F. The signaling-deficient CD28H was cloned into pCDH-puro, which was used to generate lentivirus and transduce cells as described for native CD28H above. The CHO-PD-L1-OKT3 cells were graciously provided by Amplimmune and grown in MEM + 10% FBS + 0.5 mg/mL G418. The CHO-OKT3-B7H7 was generated with full-length B7H7 cloned into the lentiviral vector pCDH-hygro (Systems BioSciences). The resulting plasmid was transfected into 293 cells to generate lentivirus containing B7H7. CHO-OKT3 cells (grown in MEM + 10% FBS) were transduced with Lenti-B7H7-hygro and selected with hygromycin B (400 μg/mL). The CHO-OKT3-B7H7 cell line was grown in MEM + 10% FBS medium. CHO-PDL1-Her2 cell line was derived with full-length human Her2 cloned into the lentiviral vector pCDH-puro (Systems BioSciences). The resulting plasmid was transfected into 293 cells to generate lentivirus containing Her2. The CHO-PD-L1-OKT3 cell line (grown in Ham's F12 + 10% FBS + 0.5 mg/mL G418) was transduced with Lenti Her2-His-puro and selected with puromycin (10 μg/mL). The CD32 L cells-B7H7 was generated with full-length B7H7 cloned into the lentiviral vector pCDH-hygro (Systems BioSciences). The resulting plasmid was transfected into 293 cells to generate lentivirus containing B7H7. CD32 L cells (grown in RPMI1640 + 10% FBS) were transduced with Lenti-B7H7-hygro and selected with hygromycin B (400 μg/mL).
Bispecific antibody generation
The variable-domain heavy-chain (VH) and light-chain (VL) sequences of the CD28H-specific hamster antibody, clone 4–5 (9), were humanized by identifying the closest matching VH and VL human antibody framework regions from the IMGT database (www.imgt.org) at the amino acid level and grafting the hamster complementarity-determining regions onto these frameworks to generate the humanized anti-CD28H, AMP945 FGL. The VH and VL sequences of the two parent antibodies of AMP945 FGL (anti-CD28H), durvalumab (anti–PD-L1; DrugBank accession number: DB11714) and trastuzumab (anti–Her-2; DrugBank accession number: DB00072), were then used to generate a series of bivalent bispecific antibody constructs based on the BiS2Ab and BiS3Ab formats described previously (19). For the panel of constructs we designed, IgG backbone and the appended single-chain variable fragment (scFv). The scFvs were further stabilized by the addition of an interchain disulfide bond between VH amino acid position 44 and VL amino acid position 101. Initially, all bispecific antibody constructs were generated with the human IgG4P fragment crystallizable gamma (Fcγ) domain, which facilitates Fcγ receptor clustering by the bispecific antibody. This domain also contains a serine-to-proline substitution in the hinge region at position 228 [according to EU numbering of the full-length IgG4 VH (IgG4P; 20)] to prevent inter-Fab arm exchange. For ADCC studies with primary human NK cells (described below), the human IgG1 isotype was used. For the CD28H–Her2 bispecific studies (described below), human IgG4P was used with matched isotype monospecific antibodies and isotype control. All genetic constructs encoding CD28H–PD-L1 and CD28H-Her2 bispecific antibodies were synthesized by GeneArt (Thermo Fisher Scientific) and cloned into the pOE mammalian expression vector, using standard restriction enzyme cloning and ligation. After the sequences of the bispecific antibody constructs were confirmed by Sanger sequencing, recombinant bispecific antibodies were expressed by transient transfection of Chinese hamster ovary (CHO) cells and protein purification, as described previously (21). The bispecific antibody panel was assessed for expression yield after purification by spectrometry at 280 nm (Perkin Elmer Envision) and for protein monomer, aggregate, and fragmentation after purification by spectrometry at 280 nm (Perkin Elmer Envision) and for protein monomer, aggregate, and fragmentation after purification by analytic size exclusion chromatography as described previously (19). The best CD28H–PD-L1 bispecific antibody which retained equivalent binding of both antigens to their parental IgG, PDH70024, was a BiS3Ab format, with the AMP945 FGL (anti-CD28H) as the Fab and durvalumab (anti–PD-L1) as an scFv appended to the C-terminus of the human IgG4P Fc domain. The CD28H-Her2 bispecific antibody, NKB0034, was a BiS2Ab format with trastuzumab (anti-Her2) as the Fab and AMP945 FGL (anti-CD28H) as an scFv appended to the N-terminus of the trastuzumab VH domain.
Simultaneous antigen-binding assays were performed by biolayer interferometry on the Octet RED384 (FortéBio) at 25°C in assay buffer containing PBS (14190-169, Thermo Fisher Scientific), 0.1% [volume for volume (v/v)] BSA (A9576-50ML, Sigma), and 0.01% (v/v) Tween-20 (P9416; MilliPore Sigma), pH 7.4, using tilted-bottom, black 384-well plates (18-5076; FortéBio). Assays were set up with High Precision Streptavidin (SAX) Biosensors (18-5117; FortéBio) according to the manufacturer's instructions. The first antigen, biotinylated recombinant human (rh)CD28H-Fc (8316-TR-050, R&D Systems), was loaded onto biosensors at 5 μg/mL for 1 minute for a wavelength shift of approximately 0.7 nm, followed by 1-minute dissociation in a fresh well containing buffer only. The biosensor was then moved to a well containing the bispecific antibody (PDH70024 or NKB0034) at a concentration of 5 μg/mL, and the wavelength shift was measured over 5 minutes. After a 5-minute dissociation period, the biosensors were moved to wells containing the second antigen, rhB7-H1, at 5 μg/mL (R&D Systems) or in-house–generated recombinant histidine-tagged Her-2 ECD (AstraZeneca, described below) for a further 5 minutes, and the wavelength shift was measured. Finally, the biosensor was moved to a fresh buffer well for a 5-minute dissociation period. Data were analyzed with Octet data analysis software, version 9.0 (FortéBio). Recombinant Her2 extracellular domain (ECD): the ECD of Her2 was PCR-amplified and cloned into a mammalian expression vector. A 6× his tag was included at the C-terminus for protein purification purposes. The Her2-ECD-his containing expression vector was transfected into the Expi293 (Thermo Fisher Scientifc, catalog no. A14527) using 293fectin (Thermo Fisher Scientific, catalog no. 12347019). Supernatant was harvested 7 days post transfection (total volume 2.4 L). A 5 mL HisTrap HP column was used for the first step of purification, and the HiLoad 26/60 Superdex 200 PG column for the polishing step. Fractions containing monomeric protein were pooled and concentrated and the final Her2 ECD protein concentration was determined by BCA assay.
NFAT reporter assays
The functional potency of the CD28H and PD-L1 scFV in BiS-Ab was determined in a PD-1/PD-L1 blockade assay (Promega, catalog no. J1255) according to the manufacturer's protocol. Briefly, target CHO cells stably expressing PD-L1 and OKT3 were cocultured with PD-1–expressing NFAT reporter parental Jurkat T cells or Jurkat T cells expressing CD28H (wild-type, ΔICD, or TYRTM) in a 1:1 ratio (4 × 104 cells each) for 6 hours with dose dilutions (serial 1:3 dilutions of Bis CD28H/PD-L1, CD28H, PD-L1, and isotype with a starting concentration of 0.1 nmol/L; 1:4 dilutions of Bis CD28H/PD-L1, CD28H, PD-L1, CD28H+PD-L1, and isotype with a starting concentration of 100 nmol/L) of soluble test antibody. After 6 hours, NFAT luciferase was detected with the Bright-Glo Luciferase Assay System (Promega), and luminescence was read with an Envision Plate reader (Perkin Elmer). Curve fitting was performed on the analyzed data using a log(agonist) versus response variable slope model [Y = Bottom + (Top-Bottom)/(1+10⁁((logEC50-X)*HillSlope))] with Prism software (version 8; GraphPad).
Preparation of stimulation plate
On the day before stimulation, untreated 48-well plates (Nunc;150787) were coated withanti-CD3 (1 μg/mL; BioLegend; 317326) along with NIP (5 μg/mL; AstraZeneca) anti-CD28H (5 μg/mL; AstraZeneca), anti-PD-L1 (5 μg/mL; AstraZeneca), combined anti-CD28H and anti-PD-L1, BiS-Ab (5 μg/mL; AstraZeneca), B7H7-Fc (R&D Systems; 8084-B7) with/without anti-B7H7 (AstraZeneca, Clone 20C5) or anti-CD28 (2.5 μg/mL; eBioscience; 16-0288-85) in the absence or presence of PD-L1-Fc (5 μg/mL; R&D Systems; 156-B7), as indicated, in 200 μL PBS overnight. Plates were washed with PBS three times and air dried on the bench before use.
Cell culture and stimulation for intracellular signaling
Total CD4+ T cells were isolated with negative selection kits (Miltenyi Biotec; 130-104-454), according to the manufacturer's protocol, from healthy donor cells and Jurkat cell lines (CD28H wild-type, CD28H-ΔICD, or CD28H TYRTM). Cells were cultured in RPMI medium containing 2 mmol/L l-glutamine (Gibco; 11875-093) and 10% FBS (Gibco; 10082-147) at 37°C. For stimulation, cells were stimulated by seeding at 4 × 105 in 200 mL of medium in 48-well plates for the indicated time period. B7H7-Fc (R&D, 8084-B7-050) was used at 1 μg/mL, and PD-L1–Fc (R&D, 156-B7-01M) was used at 5 μg/mL. Anti-B7H7 antibody (clone 20C5, Amplimmune) was used at 5 μg/mL.
Cells were washed with PBS twice and resuspended with lysis buffer (20 mmol/L Tris Cl, pH 7.5; 1 mmol/L EDTA (Invitrogen/Thermo Fisher Scientific; 15575-038); 150 mmol/L NaCl; 0.5% CHAPS (Sigma-Aldrich; C3023-25G); 10% glycerol (Invitrogen/Thermo Fisher Scientific; 15514-011)) on ice. Lysates (total 15 μg of proteins) were mixed with loading dye buffer (6X Laemmli buffer, Alfa Aesar; J61337). NuPAGE 4%–12% Bis-Tris gels (Thermo Fisher Scientific) were run to separate proteins. Proteins on gel were transferred to nitrocellulose membranes (iBlot 2 Transfer Stacks from Invitrogen/Thermo Fisher Scientific), and then blocked with blocking buffer [3% BSA in Tris-buffered saline with polysorbate 20 (TBST)] for 1 hour at room temperature. Membranes were incubated with probing solution (appropriate primary antibody dilution in blocking buffer, indicated below) overnight at 4°C. Membranes were washed three times with TBST for 5 minutes each, followed by secondary antibody incubation for 1 hour at room temperature. Membranes were washed three times with TBST for 5 minutes each. SuperSignal West Pico or Dura Chemiluminescent substrate kits (Thermo Fisher Scientific) were used to visualize the signal with X-ray film (CL-X Posure Film from Thermo Fisher Scientific; 34091) or ImageQuant LAS 4000 mini (GE Healthcare).
Antibodies for immunoblot
Anti-phospho-ERK1/2 (catalog no. 9101), anti-ERK1/2 (catalog no. 4695), anti-phospho-S6K1 (catalog no. 9205), anti-S6K1 (catalog no. 2708), anti-phospho-AKT (S473; catalog no. 4060), anti-phospho-AKT (T308; catalog no. 13038), and anti-AKT (catalog no. 4691) were purchased from Cell Signaling Technology. Anti-phospho-SHP2 (PA5-17186) was purchased from Invitrogen/Thermo Fisher Scientific. Anti-SHP2 (sc-280) was purchased from Santa Cruz Biotechnology. Anti-CD28H and anti-PD-L1 were generated by AstraZeneca. All primary antibodies were diluted 1:1,000. Secondary anti-rabbit-IgG linked with horseradish peroxidase (HRP; catalog no. 7074) and secondary anti-mouse-IgG linked with HRP (catalog no. 7076) were purchased from Cell Signaling Technology and diluted 1:10,000.
In vitro T-cell costimulation assays
Antibodies were coated on 96-well plates overnight at 4°C and washed three times with PBS. Carboxyfluorescein succinimidyl ester (CFSE)-labeled (1 nm; Invitrogen) healthy donor human total T cells or CD4+ or CD8+ T cells were seeded at 100,000 cells per well for either 48 or 72 hours. Antibody clones and sources were as follows: human CD3 (1 μg/mL, OKT3; BioLegend), human CD28 (1 μg/mL, clone 28.2, NA/LE; BD Biosciences), anti-CD28H (33.33 nmol/L, 945 AMP FGL; AstraZeneca), anti–PD-L1 (33.33 nmol/L, clone 4736; AstraZeneca), and bispecific CD28H/PD-L1 (33.33 nmol/L; AstraZeneca). For the majority of the assays described here, the Bis3 format of the CD28H/PD-L1 in the IgG1TM format was utilized; use of other formats for some assays are specifically mentioned in the “Results” section. Cytokines were detected in 25 μL of undiluted supernatants of 2- to 3-day costimulation cultures using the Meso Scale Diagnostic human TH1/TH2 10-plex, or the human IFNγ Tissue Culture Kit single-plex assay. The MSD assay plate was read using a MESO Sector 600 plate reader.
T cell–PD-L1 suppression assays
Human recombinant PD-L1–Fc chimeric protein or a control IgG–Fc (20 nmol/L; R&D) was coated overnight at 4°C on 96-well plates with OKT3 (1 μg/mL; BioLegend). Two-day preactivated (CD3/28 T-cell activation/expansion kit; Miltenyi Biotec) human peripheral total T cells were seeded with soluble antibodies (BiS CD28H/PD-L1, CD28H, PD-L1, isotype) and assessed for proliferation (Ki67, clone B56; BD Biosciences) and intracellular effector cytokines (granzyme B, clone GB11; BioLegend) by FACS analysis (protocol described below). IFNγ (Meso Scale Diagnostic, IFNγ Tissue Culture Kit single-plex assay) was measured in 25 μL of undiluted culture supernatant after 2–3 days of restimulation.
T-cell cytomegalovirus antigen recall assays
Peripheral blood mononuclear cells (0.5 × 106 PBMCs; isolated via Ficoll gradient of fresh blood from donors; the AstraZeneca Gaithersburg Research Blood Program) derived from cytomegalovirus (CMV)-positive donors (three donors) were restimulated with CMV peptide (1 μg/mL final of pepTivator pp65; Miltenyi Biotec) in the presence of increasing dose dilutions (10–100 nmol/L) of soluble treatment antibodies (BiS CD28H/PD-L1, CD28H, PD-L1, isotype). After 4 days, supernatants were assessed for IFNγ production (IFNγ Tissue Culture Kit single plex assay, Meso Scale Diagnostic) as described above.
In vitro induction of TRM cells
Isolated total T cells were (0.3 × 106) CFSE-labeled (as described above) and were seeded on 96-well plates precoated overnight with OKT3 (1 μg/mL) in RPMI1640 medium (Gibco) supplemented with 10% FCS (Gibco), penicillin (100 U/mL; Gibco), and streptomycin (100 μg/mL; Gibco). Following a published protocol (22), in vitro TRM cells were induced by addition of IL15 (50 ng/mL; PeproTech) for 3 days, followed by the addition of TGFβ (50 ng/mL; PeproTech) for a subsequent 3 days. As controls for TRM induction, OKT3-stimulated T cells were also treated with IL2 only (100 IU/mL; PeproTech) or IL15 only for 6 days. Flow cytometry analysis of surface markers (CD4, CD8, CD69, and CD103) and proliferation (CFSE dilution) was performed on day 6 (protocols below). For treatments, plates were coated with OKT3 and a specific treatment antibody (anti-CD28H, anti–PD-L1, or BiS-Ab; all at 1 μg/mL) simultaneously and tested for in vitro induction of TRM cells. Day 6 cells were also tested for surface (CXCR3, CD14, CD39, CD28H, PD-1) and intracellular (BLIMP-1, EOMES) TRM markers via flow cytometry as described below.
NK cytotoxicity assays
U87MG (PD-L1+) or P815 target cells were labeled with europium, using DELFIA EuTDA Cytotoxicity Reagents (PerkinElmer) according to the manufacturer's protocol, and were incubated with the indicated antibodies at a fixed concentration (66 mmol/L) for 15 minutes before isolated NK cells (from healthy donor peripheral blood) were added at an effector-to-target cell ratio of 20:1. Cells were incubated for 4 hours before release of europium in the supernatants was detected with DELFIA Eu-Solution (PerkinElmer). Anti-NKG2D (MAB139; R&D) was used as positive control for the redirected killing assay. For these set of assays, the Fc-enabled IgG1 format of monoclonal antibodies and CD28H/PD-L1 BiS-Ab was utilized.
All the dissociated tumor cells (generated by mechanical and enzymatic digestion) were purchased from Conversant Bio (now, Discovery Life Sciences) as cryopreserved samples and stored in liquid nitrogen. For analysis, samples were quickly thawed, washed twice with X-VIVO 15 media, assessed for cell viability on a TC20 cell counter (Bio-Rad), and a minimum of 3 × 105 cells were used for the flow analysis as described below. For functional assays, 1 × 105 dissociated tumor cells were seeded on 96-well plates precoated overnight with OKT3 (1 μg/mL) and soluble treatment antibodies (20 nmol/L). After 5 days, supernatants were collected for cytokine analysis as described above. Dissociated tumor cells were cultured in X-VIVO 15 serum-free medium (Lonza) supplemented with 5% normal human serum (Sigma), penicillin (100 U/mL; Gibco), streptomycin (100 μg/mL; Gibco), and IL2 (10–50 IU/mL; PeproTech).
Cells were washed with PBS, blocked for 15 minutes on ice (Human Trustain FcX; BioLegend), and stained with a viability dye for 15 minutes (Live/Dead Fixable Blue Dead Cell Stain Kit; Thermo Fisher Scientific). Cells were then stained with surface antibodies in FACS buffer (PBS + 2% FCS + 0.2 mmol/L EDTA) for 30 minutes on ice. For intracellular cytokine staining, cells were fixed with fixation buffer for 15 minutes at room temperature and stained with antibodies in permeabilization buffer for 30 minutes on ice (eBioscience Foxp3 Transcription Factor Fixation/Permeabilization kit; Thermo Fisher Scientific). Samples were acquired on FACSymphony or LSRFortessa flow cytometer (BD Biosciences), using BD FACSDiva software (BD Biosciences); data were analyzed using FlowJo software (version 10; FlowJo). Antibodies used in flow studies are listed in Supplementary Table S1.
All data were analyzed using GraphPad Prism Software 8 and are represented as mean ± SD (error bars). Statistical significance was analyzed by Student t test, Mann–Whitney test, or one-way ANOVA, followed by Sidak multiple comparison, as indicated in the figure legends.
The CD28H/B7H7 pathway activates primary T cells via MAPK and mTOR signaling
CD28H is a T-cell costimulatory receptor, and an agonistic antibody targeting CD28H that has previously been shown to induce proliferation of T cells (9). A fully humanized anti-CD28H (AMP945 FGL) antibody (generated in-house, see Materials and Methods) activated primary human CD4+ and CD8+ T cells when stimulated with suboptimal anti-human CD3 (OKT3; Supplementary Fig. S1A). Immunoblot analysis showed that anti-CD28H and anti-CD3 activated the pAKT signaling in primary human T cells (Supplementary Fig. S1B). To determine the downstream signaling pathways that were attributable to CD28H-induced proliferation of T cells, we assessed MAPK and mTOR pathways, both of which play a role in T-cell proliferation, cytokine production, and T-cell metabolism (23). Upon anti-CD28H cross-linking in the presence of anti-CD3, we observed the activation of pERK within 10 minutes and increased activation in 30 minutes. This latent and sustained activation of pERK is distinct from the early ERK activation that occurs within minutes of CD3 cross-linking (ref. 24; Supplementary Fig. S1B). Cross-linking of CD28H also activated the mTOR pathway, as observed by phosphorylation of ribosomal S6-P70K, which occurs downstream of the mTORC1 complex (Supplementary Fig. S1B). B7H7 is the cognate ligand of CD28H and activates T cells in the presence of anti-CD3 (9). Similar to anti-CD28H, B7H7-Fc induced the activation of pERK and pS6-P70K in primary T cells in the presence of OKT3, which was reversed in the presence of an anti-B7H7 blocking antibody (Supplementary Fig. S1C).
To further confirm the agonistic activity of the CD28H antibody in a more physiologic setting, we overexpressed CD28H in Jurkat T cells with a T-cell receptor (TCR)-responsive NFAT luciferase reporter. Soluble anti-CD28H mediated increased luciferase production over that of OKT3, but only in the context of coculturing CD28H-expressing Jurkat cells with CHO cells expressing surface-Fc cross-linking CD64 (Supplementary Fig. S1D and S1E).
BiS-ab is a bispecific antibody comprising agonistic anti-CD28H and anti–PD-L1
In normal donor peripheral blood, CD28H expression is higher on naïve T cells than on effector memory T cells (9). In human tumors, where the majority of T cells are effector memory cells, 30%–50% of tumor-infiltrating T cells still express CD28H (12). To cross-link CD28H on these cells, we designed a bispecific antibody with an anti-CD28H as the T-cell engager paired with an anti–PD-L1. TILs derived from non–small cell lung carcinoma (NSCLC) had higher coexpression of PD-1 and CD28H in the CD8+ T-cell pool than in CD4+ T cells, whereas peripheral T cells were single expressers of either PD-1 or CD28H (Fig. 1A). Multiple formats and orientations of BiS-Ab were investigated to identify the antibody format that retained activity comparable with that of the parent immunoglobulins and that also had favorable characteristics for development (i.e., high expression, low propensity to aggregate and/or fragment upon production). The final bispecific antibody format was constructed with anti-CD28H specificity as the Fab, and anti–PD-L1 specificity was engineered as an scFv appended to the C-terminus of the Fc domain (IgG4) in the so-called BiSAb3 format as described previously (ref. 19; Fig. 1B). Biolayer interferometry using recombinant CD28H as the immobilized antigen and a bispecific antibody with isotype and CD28H antibody as the soluble analyte were conducted to compare the CD28H binding affinity of BiS-Ab with that of the parent antibody (Fig. 1C). Results indicated that both arms of BiS-Ab bound to their respective antigens with similar affinity as their parent antibodies (Fig. 1D). Both arms of BiS-Ab bound simultaneously to soluble forms of their specific target antigens (recombinant CD28H and PD–L1–Fc; Fig. 1D).
We hypothesized that once both BiS-Ab arms engaged their respective ligands, PD-L1 would allow the Fab domain to cross-link CD28H and induce a proliferative response in tumor-infiltrating T cells. At the same time, blocking PD-L1 by the anti–PD-L1 scFv would suppress the inhibitory signaling emanating from PD-L1–expressing tumors on PD-1+ TILs. To test the functionality of the anti–PD-L1 arm, we compared BiS-Ab with an anti–PD-L1 for the ability to block the PD-1/PD-L1 interaction using a blockade bioassay (25). In a two-cell assay system using PD-1–expressing Jurkat T cells with a TCR-responsive NFAT reporter cassette, we observed similar potency and half-maximal effective concentrations (EC50) of the anti–PD-L1 scFv in the bispecific construct and monoclonal antibody format (BiS-Ab, EC50 = 0.00028 nmol/L; anti–PD-L1, EC50 = 0.0001 nmol/L; Fig. 1E).
BiS-Ab–induced CD28H agonism in the context of PD-1/PD-L1 blockade
The stimulatory properties of BiS-Ab, along with anti-CD28H, anti–PD-L1, or a combination of the parental antibodies were evaluated for activation of primary human CD4+ and CD8+ T cells. BiS-Ab induced high activation, as assessed by CFSE dilution (Fig. 2A; Supplementary Fig. S2A) and cytokine production (Supplementary Fig. S2B and S2C) in both CD4+ and CD8+ T cells. This effect was not due to a synergistic effect of the two antigenic arms because combining the CD28H and PD-L1 antibodies individually did not provide any additive activation over the single arms alone (Fig. 2A). The single-arm CD28H antibody induced greater activation in CD8+ T cells than in CD4+ T cells (Supplementary Fig. S2A).
To demonstrate the cross-linking effect of PD-L1 from BiS-Ab and to assess the effects of soluble treatment antibodies, we stimulated human T cells with either a control IgG–Fc or a PD-L1–Fc with anti-CD3 in a PD-L1 suppression assay modified from a previous study (26). Ex vivo–purified human donor T cells were preactivated for 24 hours with anti-CD3/28 beads to upregulate surface PD-1 to maximally facilitate PD-L1–mediated repression and invoke an effector phenotype (Supplementary Fig. S2D). We also tested for surface CD28H after activation and confirmed that no receptor downmodulation due to TCR stimulation occurred (Supplementary Fig. S2D). Activated T cells were then stimulated with either plate-coated control IgG–Fc/anti-CD3 or PD-L1–Fc/anti-CD3 in the presence of the soluble BiS-Ab or single-arm CD28H or PD-L1 antibodies. Under these conditions, only BiS-Ab increased T-cell proliferation (as measured by Ki67) and induced granzyme B production and IFNγ secretion (Fig. 2B and C). This response was specific to PD-L1–mediated T-cell suppression, as there was no observable difference in Ki67, granzyme B, or cytokine response by any of the soluble antibodies under the control conditions (Supplementary Fig. S2E and S2F). These results demonstrate that primary T-cell activation through CD28H agonism occurred specifically in the context of PD-L1–mediated cross-linking by BiS-Ab.
To validate the efficacy of BiS-Ab in a more physiologic setting, we assessed its ability to modulate CMV antigen–specific T-cell responses in human PBMCs with known reactivity to the CMV peptide. HLA-A02–positive donor PBMCs were cultured in the presence of CMV peptide and fixed concentrations of soluble antibody for 72 hours. BiS-Ab induced a significantly higher IFNγ response than the combination of the anti-CD28H and anti–PD-L1 (Fig. 2D). These data indicated that the CD28H pathway was stimulated through cross-linking PD-L1 by BiS-Ab to induce TCR-dependent activation of antigen-specific human CD4+ and CD8+ T cells.
T-cell activation by BiS-Ab requires intracellular CD28H receptor signaling
Results of a previous study with CD28-CD28H chimeric receptors indicates that phosphorylation of two key tyrosine residues are required for mediating CD28H-associated downstream signaling (9). We engineered PD-1 NFAT reporter Jurkat T cells (25) to overexpress either wild-type CD28H, an ICD-deleted CD28H (ΔICD), or a triple tyrosine-to-phenylalanine (TYRTM)-mutated ICD CD28H receptor to assess whether the ICD was required for BiS-Ab–mediated signaling. These cell lines expressed similar levels of surface CD3, CD28H, and PD-1 (Fig. 3A). CD28H-overexpressing reporter Jurkat T cells were cocultured with a CHO-OKT3 cell line that overexpressed B7H7, the ligand for CD28H. The surface expression of the ligand in the target CHO cells was confirmed by FACS (Supplementary Fig. S3A). B7H7 increased NFAT activation specifically in the CD28H Jurkat T cells, but not in the parental Jurkat T cells (Supplementary Fig. S3B). Disruption of B7H7-CD28H interactions with a B7H7 ligand–blocking antibody prevented NFAT induction (Supplementary Fig. S3B). This finding was confirmed with a mouse fibroblast cell line (L cells) expressing surface cross-linked OKT3 with or without B7H7 overexpression. In this system, IL2 production was observed only in L cells expressing B7H7- and CD28H-expressing Jurkat cells (Supplementary Fig. S3C).
Next, we set up cocultures of the various Jurkat T cells overexpressing CD28H and CHO cells expressing anti-CD3 and PD-L1 in the presence of soluble antibodies, combinations, or BiS-Ab at various concentrations. PD-L1/PD-1–mediated suppression of NFAT (luciferase) in the Jurkat T cells was reduced by anti–PD-L1 alone or in combination with anti-CD28H, but not by isotype or CD28H antibody alone (Fig. 3B). BiS-Ab, however, induced 3- to 4-fold higher (7.5-fold over isotype) activation of NFAT than that induced by anti–PD-L1 alone (2.5- to 3-fold over isotype), suggesting a unique modality of T-cell activation. This synergistic T-cell activation by BiS-Ab was lost in CD28H-ΔICD–expressing Jurkat cells, suggesting that an intact CD28H signaling domain was required for BiS-Ab activity (Fig. 3C). The loss of synergistic NFAT activation was specific only to CD28H signaling, as activation mediated by anti–PD-L1 remained in the CD28H-ΔICD Jurkat cells (Fig. 3C). Similarly, BiS-Ab–mediated synergistic NFAT activation was lost with the TYRTM CD28H–expressing Jurkat cells (Supplementary Fig. S3D). These results indicated that signaling through the tyrosine residues of CD28H ICD was crucial for BiS-Ab–mediated NFAT activation in the context of PD-L1/PD-1–mediated T-cell inhibition.
We also engineered BiS-Ab with isotype controls, replacing each antigen-specific binding domain with isotype control (non-binding) antibodies. These isotype controls (bispecific isotype/PD-L1 or bispecific CD28H/isotype) failed to elicit the NFAT activation seen with the anti-CD28H/PD-L1 BiS-Ab, indicating that the enhanced TCR response seen with BiS-Ab was due to simultaneous blockade of the PD-1/PD-L1 pathway in the context of CD28H agonism (Fig. 3D).
To test the ability of the CD28H to costimulate in a bispecific format with other tumor-associated antigens as binding partners, we constructed a bispecific antibody targeting CD28H and Her2 in the BiS2 and IgG4P format and tested its ability to activate CD28H-expressing Jurkat T cells when stimulated with CHO-OKT3 cells expressing surface Her2. Results showed statistically significant NFAT activation of CD28H Jurkat cells only by the bispecific CD28H/Her2 antibody, but not by the antigen-monospecific antibodies or the combination of anti-CD28H with anti-Her2 (Fig. 3E). Furthermore, this activation was specific to the Jurkat T cells expressing wild-type CD28H and was not observed in either the parental Jurkat T cells or the CD28H ΔICD Jurkat T cells (Fig. 3E). This demonstrates the feasibility of anti-CD28H activation of T cells in the bispecific antibody format via other tumor-specific antigens (Fig. 3E).
Decreased PD-L1/PD-1–induced SHP2 phosphorylation by anti-CD28H
To explore the mechanism of action by which BiS-Ab induced T-cell stimulation, we tested for MAPK activation in primary T cells with single antibodies, as well as BiS-Ab in the context of cross-linked TCRs. We found that BiS-Ab phosphorylated MAPKs (pERK and pAKT) in primary T cells in a manner similar to the MAPK activation that was observed with anti-CD28H (Supplementary Fig. S4A). However, BiS-Ab had a more prolonged activation of the mTOR dependent pP70-S6K compared with anti-CD28H alone, with activation comparable to pP70-S6K activation with CD28 agonism (Supplementary Fig. S4A). In Jurkat T cells, anti-CD28H or B7H7-Fc induced high pERK activation, similar to that observed in primary T cells. This activation required a signaling-competent CD28H, as observed by the lack of pERK activation in Jurkat cells overexpressing CD28H TYRTM (Fig. 4A). In contrast, cross-linked PD-L1–Fc in the context of CD3 inhibited phosphorylation of ERK in in a dose-dependent manner (Supplementary Fig. S4B). Simultaneous T-cell activation with anti-CD28H or B7H7-Fc in the presence of PD-L1–Fc, however, allowed for sustained pERK activation, despite the PD-L1/PD-1–mediated repression of TCR signaling (Fig. 4B; Supplementary Fig. S4C). PD-L1/PD-1 binding results in phosphorylation of SHP1 and SHP2 (27), which directly represses TCR (28) and CD28 signaling (29, 30). PD-L1–Fc stimulation of CD28H Jurkat cells induced phosphorylation of SHP2 in as little as 30 minutes after anti-CD3 stimulation (Fig. 4C). However, stimulation of Jurkat cells with PD-L1–Fc in the presence of anti-CD28H, B7H7–Fc, or BiS-Ab reversed SHP2 phosphorylation, indicating a direct decreased repression of PD-L1/PD-1 signaling by the CD28H/B7H7 pathway (Fig. 4C). These data indicated that the CD28H pathway was able to directly or indirectly activate the MAPK pathway (pAKT and pERK) downstream of the TCR, in addition to downmodulating PD-L1 inhibitory responses. These divergent mechanisms coalesced to constitute a positive feedback loop and elicit T-cell costimulation, despite ongoing PD-1/PD-L1–mediated suppression.
BiS-Ab–induced NK cell–redirected killing
Among innate cells, CD28H is known to be expressed on NK cells, pDCs, and ILCs (9, 12). Peripheral circulating NK cells constitutively express CD28H, and one study has indicated that its surface expression is retained in the majority of tumor-infiltrating NK cells (12). CD28H expression, however, is not homogeneous within NK cells. Hence, we characterized the surface expression of activation and inhibitor receptors within CD28H-positive and -negative cells. Among all the human NK-cell markers, only CD25 was differentially expressed between CD28H-positive and -negative NK cells (Fig. 5A; Supplementary Fig. S5A). Activation with anti-CD28H upregulated surface CD137 on isolated human blood NK cells (Supplementary Fig. S5B). To determine whether CD28H engagement on NK cells induced killing of target tumor cells, freshly isolated human NK cells were cocultured with P815 tumor cells expressing high levels of Fc receptors. Tumor cells were preincubated with a fixed concentration of anti-CD28H or anti-NKG2D. Cytotoxicity and IFNγ production by NK cells toward P815 targets were induced only by anti-CD28H and anti-NKG2D and not by isotype or anti–PD-L1 (Fig. 5B; Supplementary Fig. S5C). To investigate the ability of BiS-Ab to mediate cytotoxicity against PD-L1–expressing target cells, the glioblastoma cell line U87-MG was cultured with peripheral human NK cells in the presence of isotype, single-antigen antibodies, or BiS-Ab. Anti–PD-L1 induced a high NK-cell ADCC, due to engagement of the Fc portion of the antibody with CD16 on NK cells (Fig. 5C). Similar NK-cell cytotoxicity was also seen with BiS-Ab and anti–PD-L1 containing effector-enabled IgG1 in the Fc region. Anti-CD28H alone or in the bispecific format with the isotype arm resulted in low cytotoxicity. CD28H/PD-L1 bispecific IgG1, however, elicited significantly more cytotoxicity than anti–PD-L1 alone (Fig. 5C). CD28H/PD-L1 BiS-Ab–mediated cytotoxicity was also accompanied by IFNγ production by NK cells (Supplementary Fig. S5D). These results indicated that the agonistic CD28H pathway contributed to the increased tumor cell cytotoxicity seen with BiS-Ab by boosting the anti–PD-L1–mediated NK-cell ADCC.
Induction of TRM cells by CD28H
On peripheral T cells, CD28H is expressed predominantly on naïve subsets, with lower expression on memory and effector memory T cells (9). In peripheral tissues, such as lung and intestine, high expression of CD28H is found on TRM cells, a subset of non-circulating CD8+ T cells that express both early activation marker CD69 and the αE integrin, CD103 (13). CD28H is enriched on TRM cells in TILs derived from pancreatic cancer and associates with a less differentiated phenotype (13). To determine whether the CD28H pathway costimulates TRM cells or induces in vitro differentiation of TRM cells (22), total T cells derived from peripheral blood were stimulated with anti-CD3 with or without IL2 only, IL15 only for 6 days, or IL15 for 3 days, followed by TGFβ for an additional 3 days (TRM; ref. 22). As reported previously (22), CD69+CD103+ TRM cells differentiated only in presence of of IL15 and TGFβ (Supplementary Fig. S6A). TRM cells were more abundant within the CD8+ T-cell pool compared with the CD4+ T-cell compartment (Supplementary Fig. S6B). Surface markers and intracellular transcription factors indicated differential expression of some, but not all, of the reported TRM markers, including CD14, CD39, and PD-1 (Supplementary Fig. S6C). The in vitro differentiation did not affect surface CD28H expression on TRM or non-TRM cell populations (Supplementary Fig. S6C).
Next, we tested the effects of anti-CD28H or isotype antibody under TRM cytokine-skewing conditions. Anti-CD28H enhanced the differentiation of CD69+CD103+ TRM cells in the presence of IL15 and TGFβ (Fig. 6A). This effect of CD28H on TRM cells was specific to CD8+ T cells, but not CD4+ T cells (Fig. 6B) and was statistically significant compared with the isotype antibody only under IL15 and TGFβ differentiation conditions. To extend these findings to BiS-Ab, we tested its ability to induce TRM in CD4+ and CD8+ T cells in the context of anti-CD3. Our results showed a significant induction of the TRM phenotype by BiS-Ab compared with isotype and anti-CD28H (Fig. 6C). BiS-Ab, similar to anti-CD28H, specifically increased TRM differentiation in CD8+ T cells, but not in CD4+ T cells (Fig. 6D).
Costimulation of tumor-derived TILs by BiS-Ab
To understand the physiologic significance of BiS-Ab costimulation in a tumor setting, we utilized TILs from different cancer indications. FACS revealed that lung and kidney tumors had the highest infiltration of CD45+ cells (Supplementary Fig. S7A) and that CD8+ T-cell frequency was higher than that of NK cells within the CD45+ pool (Supplementary Fig. S7B and S7C). The TRM fraction was prevalent within CD8+ T cells, and the frequencies were higher than in CD4+ T cells (Supplementary Fig. S7D). TILs derived from colorectal and ovarian tumors expressed CD28H predominantly on CD8+ T cells (40%), whereas lower expression of CD28H was seen on CD8+ TILs from other cancers (20%; Fig. 7A). CD28H had high expression in the CD69+CD103+CD8+ TRM fraction than in non-TRM CD8+ T cells (Fig. 7B). CD28H expression on tumor-infiltrating NK cells was as high (50%–60% in lung and kidney tumors) or higher (85% in colorectal and ovarian tumors) than on peripheral blood NK cells (Fig. 7C).
To assess the functional significance of targeting CD28H against tumors, in vitro T-cell activation assays were performed with TILs from various tumor indications. Dissociated tumor samples were stimulated with a fixed concentration of soluble isotype, anti–CD28H, anti–PD-L1, or BiS-Ab in the presence of anti-CD3 for 72 hours in the presence of low-dose IL2 to assess TIL activation and effector function. Increased IFNγ production in the supernatant of TILs was found only in the presence of BiS-Ab (Fig. 7D). These results indicated that BiS-Ab mediated a tumor-specific response. The resulting increase in IFNγ indicated that tumor-specific PD-L1 expression allowed cross-linking of BiS-Ab, leading to CD28H-mediated cell activation.
Despite the success of ICI therapies, patients whose disease is nonresponsive represent a significant population with unmet need. Because most of these therapies aim at boosting T-cell responses within tumors with either monotherapy or dual checkpoint inhibitor therapy, a combinatorial bispecific approach is promising. BiS-Ab described here abrogated PD-1/PD-L1–mediated suppression and delivered a costimulatory signal through a B7 family receptor with a “brakes-off/pedals-on” approach. The CD28H agonistic arm of BiS-Ab targeted a B7 receptor pathway to mediate T-cell costimulation (9). Although this receptor/ligand pathway has been characterized in T and NK cells, the signaling downstream of CD28H has not yet been demonstrated in primary T cells (9, 12). Our results showed that CD28H activated MAPK and pAKT, leading to downstream mTOR signaling in a manner similar to the CD28 pathway.
The biological role of B7H7, the ligand for CD28H, is contentious because it is highly expressed on a variety of tumor tissues, and many studies indicate an inhibitory role on T-cell activation (17, 18, 31, 32). Here, we showed a costimulatory role for B7H7, using CD28H-overexpressing Jurkat T cells, as well as primary human T cells, through activation and signaling analysis. B7H7-mediated phosphorylation of ERK/AKT, as well as mTOR activation, correlates with the activation kinetics observed for agonistic anti-CD28H, indicating a stimulatory role for this ligand-receptor pair on T cells. It is likely that the B7H7 ligand is promiscuous (33), and indeed, a publication has identified a KIR-like inhibitory receptor expressed on a subset of activated T cells that could account for the observed T-cell inhibition mediated by B7H7 (34). BiS-Ab described here retained the biological functions of CD28H-dependent T-cell costimulation but also reduced the suppressive effects of the anti–PD-L1 arm. This bispecific construct mediated increased costimulation and effector cytokine production in human T cells than the combination of the single antibodies, indicating a unique mechanism of action. In assay conditions designed to phenocopy TILs, in which T cells were activated to be effector-like and express PD-1, BiS-Ab induced T-cell activation only when it encountered PD-L1 in the context of TCRs. This suggests that BiS-Ab could drive T-cell activation in a tumor-specific manner, with minimal or no systemic activation. The possibility that BiS-Ab could encounter and bind CD28H-high–expressing naïve T and NK cells in the periphery cannot be discounted. However, without the cross-linking effect of anti–PD-L1, the CD28H antibodies would not induce T-cell activation.
Among innate cells, NK cells have a role in immune surveillance and are present both peripherally and within human tumors. We showed here that anti-CD28H activated NK cells and induced direct killing of tumor cells. A recent study shows that CD28H agonism in the presence of other NK agonists, including 2B4 and NKG2D, induces synergistic NK activation in the form of CD107a degranulation and effector cytokine secretion (16). This study also showed that interactions between NK cell–expressed CD28H and its ligand B7H7 on tumor cells induced high cytotoxicity that could be reversed by CD28H-blocking antibodies (16). Our results showed that BiS-Ab induced killing of target cells expressing PD-L1 via ADCC, which was much higher than the cytotoxicity induced by anti–PD-L1 alone. These findings suggest that BiS-Ab has the potential to agonize both adaptive T cells and innate NK cells within the tumor microenvironment to induce antitumor responses. Supporting this hypothesis are reports correlating improved overall survival with higher expression of B7H7 in patients with pancreatic ductal adenocarcinoma (PDAC) and gastric cancers (35–37). In light of the findings presented here and those of others, it is possible to predict that the increased overall survival that correlated with the higher B7H7 expression seen in PDAC and gastric tumors is due to a combination of T-cell activation through CD28H and/or direct lysis of tumor cells bearing B7H7 by NK cells within the tumor microenvironment, leading to antitumor immunity.
Within peripheral tissues and TILs, CD8+ TRM cells, another T-cell subset, are reported to retain high surface CD28H expression (13). The same report also shows that T cells exposed to the TRM-skewing cytokines IL15 and TGFβ express and retain high levels of CD28H (13). Our findings showed that CD28H, as well as BiS-Ab, increased in vitro skewing of TRM cells in human peripheral blood T cells, suggesting an important physiologic role for CD28H/B7H7 signaling in the maintenance of TRM cells. The observation that B7H7 has high expression in the epithelium of colon, lung, pancreas, and breast, among other tissues, than in innate immune cells (9, 18) suggests its physiologic role in T-cell stimulation in peripheral tissues. CD8+ TRM cells have been characterized as tissue-specific, non-migratory CD8+ memory cells that provide essential immunity against virus infections and local inflammation (38). Studies have highlighted a critical role for tissue-resident cells in antitumor immunity, increased overall survival, and favorable patient outcomes (39). In mouse models of melanoma, TRM cells play a fundamental role in maintaining cancer-immune equilibrium (15). Markers associated with TRM cells, including CD39 and CD103, distinguish tumor-reactive T cells from bystander T cells in primary human TILs (14, 40). Our results indicated that CD28H signaling contributed to increased numbers of CD39+, CD103+, and CD69+ TRM cells in vitro. This suggests an undiscovered role for CD28H receptor in TRM differentiation, which, unlike in mice, has not been very well characterized in humans. TRM cells in human TILs upregulate many markers of T-cell dysfunction, including PD-1, Tim3, and Lag-3, making them prime targets for checkpoint inhibitor therapy (41, 42). In this context, our results showing that BiS-Ab increased induction of CD8+ TRM cells imply that this targeting strategy might promote favorable antitumor responses. The absence of a mouse ortholog for CD28H or B7H7 makes it challenging to confirm these findings in tumor models in vivo. However, our results showing that BiS-Ab elicited higher cytokine responses in human primary TILs, but not T cells derived from peripheral blood, support the hypothesis that BiS-Ab induces activation of CD4+ and CD8+ T cells, TRM cells, and NK cells within the tumor microenvironment in a tumor-specific manner.
The mechanism of action of BiS-Ab in eliciting greater T-cell stimulation than the combination of the monospecific antibodies is unique and required phosphorylation of tyrosine residues in the intracellular signaling domain of CD28H. Once cross-linked, CD28H was able to reverse PD-L1–mediated suppression of ERK phosphorylation. Our data indicated that CD28H agonism alone was able to reverse SHP2 phosphorylation during active PD-L1/PD-1–induced T-cell inhibition. This held true for the ligand B7H7, which, when co-expressed with PD-L1 on the target cell surface, still activated PD-1–expressing T cells via CD28H (34). Deciphering the CD28H downstream signaling that leads to SHP dephosphorylation is beyond the scope of this work. However, our results indicated an active signaling requirement for tyrosine phosphorylation in CD28H-ΔICD. The ability of the CD28H/B7H7 pathway to short-circuit inhibitory signals is not unique to T cells and has been reported for NK cells (16). Other work indicates that PD-1 inhibition specifically targets CD28 costimulatory signals, and CD28 surface expression on T cells has proven to be a useful biomarker for successful anti–PD-1 therapy in tumors. Dysfunctional T cells that do not express CD28 or do not encounter B7.1 or B7.2, even when they express CD28, are not reinvigorated by anti–PD-1 therapy (29, 30). Published data indicate that CD28H expression correlates with exhausted T cells within NSCLC TILs (41). Because tumors do not express B7 ligands, but do express B7H7 and PD-L1, it is possible that in TILs, the CD28H/B7H7 pathway may activate and rescue dysfunctional T cells that are CD28-negative and have high expression of PD-1.
In summary, this body of work described a signaling mechanism invoked by the B7 family CD28H/B7H7 pathway in activating T cells, NK cells, and CD8+ TRM cells. The BiS-Ab targeting CD28H and PD-L1 described here could augment TIL-specific antitumor responses in populations of patients whose disease does not respond to ICI therapies.
M. Ramaswamy is an employee of AstraZeneca and has stock ownership and/or stock options or interests in the company. D.C. Jones is an employee of AstraZeneca and has stock ownership and/or stock options or interests in the company. H. Ghadially is an employee of AstraZeneca and has stock ownership and/or stock options or interests in the company. J.M. Riggs is an employee of AstraZeneca and has stock ownership and/or stock options or interests in the company. G.K. Bhat was a full-time employee of AstraZeneca and held stocks in the company during this project. R. Herbst is an employee of AstraZeneca and has stock ownership and/or stock options or interests in the company. D.J. Schofield is an employee of AstraZeneca and has stock ownership and/or stock options or interests in the company. G. Carlesso is an employee of AstraZeneca and has stock ownership and/or stock options or interests in the company. No disclosures were reported by the other authors.
M. Ramaswamy: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. T. Kim: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. D.C. Jones: Resources, data curation, formal analysis, validation, investigation, visualization, methodology, project administration. H. Ghadially: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration. T.I. Mahmoud: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. A. Garcia: Conceptualization, resources, data curation, formal analysis, validation, investigation, methodology. G. Browne: Resources, data curation, formal analysis, investigation, visualization, methodology, writing–original draft. Z. Zenonos: Resources, formal analysis, investigation, visualization, methodology. Y. Puplampu-Dove: Resources, data curation, formal analysis, validation, investigation, visualization, methodology. J.M. Riggs: Resources, formal analysis, investigation, visualization, methodology, project administration. G.K. Bhat: Investigation. R. Herbst: Conceptualization, resources, supervision, project administration, writing–review and editing. D.J. Schofield: Conceptualization, resources, formal analysis, supervision, methodology, writing–original draft, project administration, writing–review and editing. G. Carlesso: Conceptualization, resources, supervision, writing–review and editing.
We thank Shino Hanabuchi for her intellectual support and critical feedback on the article and Lorraine Clarke, Susan Wilson, Deepali Malhotra, and the AstraZeneca Blood and Flow Cytometry core teams for their technical input and support.
This study was funded by AstraZeneca.
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.