Anti-CD40 Antibodies Fused to CD40 Ligand Have Superagonist Properties

CD40 is a potent activating TNFR superfamily member expressed on APCs and B cells (1). Dendritic cells (DCs) respond to infections by internalizing Ags and activating pathogen-associated molecular patterns and present foreign Ags on their MHC molecules to Ag-specific T cells, initiating a cycle of DC maturation via CD40 ligand (CD40L) expressed on the activated cognate T cell, which then directs cellular and humoral Ag-specific T and B cell responses to the pathogen (1). The intense focus on the development of strong CD40 agonists for therapeutic applications reflect this central position of CD40L–CD40 interaction in the development of T and B cell adaptive immune responses. In particular, CD40 agonists have high potential as adjuvants to stimulate Ag-based vaccines for protective and therapeutic cellular immunity through various mechanisms, administered either alone or in combination with Ags or other immune modulators (2–14).

Enhancing the biological activity of anti-CD40 agonistic Abs or CD40L compounds has thus become important for improving potential in vivo efficacy. Approaches toward this end include primary screening to identify CD40-reactive mAbs that have the most potent agonist activity, e.g., inducing cytokine secretion or CD86 surface expression on human DCs (12). Maximizing the agonist efficacy and utility of clinical candidate anti-CD40 Abs typically involves affinity maturation of the H and L chain variable regions (9), enhancing cross-linking of the C region with FcR (15), and screening and, in the case of CD40L agonists, making multimeric forms to enhance receptor cross-linking (4, 16). Screening can also include identifying potent agonists without any need for Fc interaction (17, 18), which may be problematic for human platelet activation if FcγRIIA interaction is maintained (15). Agonistic anti-CD40 Abs can either bind to sites that overlap the CD40L-interacting region or may interact with a site distinct from its ligand-binding region (6, 15, 18, 19). Agonists that do not interfere with CD40L action permit possible synergy with CD40L naturally expressed on activated T cells, but it is not clear if this distinction has clinical relevance (12).

In this study, we explore soluble CD40L cooperation with certain agonistic anti-human CD40 (anti-hCD40) vehicles for enhancing CD40 activation efficacy of anti-CD40 Abs. We further demonstrate via both in vitro and in vivo assays that directly fusing CD40L to anti-CD40 H or L chain C termini confers superagonist activity to anti-CD40 Abs, thus providing a new mode of increasing their adjuvant activity.

Generation and screening strategies for making in-house mAbs were as described (20, 21), except that anti–DEC-205 (22) fused to human CD40 ectodomain via the H chain C termini was used as immunogen. GenBank sequences HQ738666.1, HQ738667.1, KP684035, KP684036, KP684037, KP684038, KP701435, and KP701436 describe the human IgG4 chimeric forms of, respectively, the in-house derived 12E12, 11B6, 24A3, and 12B4 anti-hCD40 H and L chains. Unless otherwise indicated, humanized V region sequences for the anti-hCD40 12E12 and 11B6 fused to human IgG4 H and κ L chain C regions were used. Methods for expression vectors and protein production via transient or stable Chinese hamster ovary (CHO subline S [CHO-S]) cells transfection and quality assurance including CD40 binding specificity were as described (23–25). All anti-CD40 mAbs, unless otherwise indicated, were configured on identical human IgG4 H chain and human κ-chain backbones. Mouse anti-hCD40 S2C6, a previously described CD40 agonistic Ab (3), was from Mabtech. For size-exclusion chromatography (SEC) analysis or purification, proteins were loaded onto a Sephadex 200 10/300 GL prepacked gel filtration column (GE Healthcare) equilibrated in 90 mM Na₂SO₄, 10 mM NaH₂PO₄, 1 mM Na₂EDTA, and 0.02% NaN₃ (pH 8.0) buffer and run at 0.5 ml/min at 25°C on a Akta Purifier instrument (GE Healthcare). Storage buffer was PBS with 125 mM hydroxypropyl β-cyclodextrin (Cavitron W7 HP5).

Frozen human PBMCs resuspended in RPMI 1640 medium (Life Technologies), 50 U/ml penicillin/streptomycin (Life Technologies, reference [ref] 15140-122), 100× HEPES Buffer (Life Technologies, ref 15630-080), 100× nonessential amino acids (Life Technologies, ref 11140-050), 100× sodium pyruvate (Life Technologies, ref 11360-070), 100× GlutaMAX (Life Technologies, ref 35050-061), and 1000× 2-ME (Life Technologies, ref 21985-023) (complete RPMI) + 10% FBS with 50 U/ml benzonase nuclease (MilliporeSigma, catalog [cat] 70746) were washed 2× in PBS, suspended at 1E7/ml in PBS, and then stained with 1.25 μM CellTrace CFSE Cell Proliferation Kit (Invitrogen) at room temperature (RT) for 7 min in the dark. Then, 10 vol of cold 100% FBS were added, and cells were kept at 4°C for 5 min, followed by 2× washes with PBS. Cells (1E6/ml) in complete RPMI 1640 medium with 10% FBS with human IL-4 (10 ng/ml) (R&D Systems, 204-IL/CF) and human IL-21 (5 ng/ml) were added to each culture well, along with various (10-fold dilutions) amounts of anti-hCD40 mAbs. After 5 d of culture, the cells were stained as described below for surface and LIVE/DEAD markers followed by flow cytometry analysis gating on singlets, live cells, and CFSE⁻/CD19⁺ cells. In some experiments, a constant suboptimal (e.g., ∼5% of maximal response, 100 ng/ml, or 6 nM) amount of monomeric soluble CD40L (sCD40L, R&D Systems, 6245-CL-050) was added to test cooperation between the agonistic Abs and CD40L.

To make human myeloid-derived DCs (MDDC), 1E6 human blood monocytes/ml were cultured in six-well plates (2 ml per well) in complete RPMI 1640 medium + 10% FBS + 10 ng/ml human IL-4 + 100 ng/ml human GM-CSF (Sanofi, NDC 0024-5843-01). Half of the medium was changed at day 2 and also at day 4, maintaining the same concentration of IL-4 and GM-CSF. Cells were harvested at day 5 without scraping but with gentle washing and plated in a 96-well V-bottom plate in 200 μl at 1E5 cells per well. Typically, 1E6 DCs were derived from 2E6 monocytes. Then, 10 ng/ml IL-4 and 100 ng/ml GM-CSF and different concentrations (10-fold dilutions) of the anti-hCD40 mAbs or anti-hCD40 IgG4 fusion proteins were added, and in some conditions, a constant suboptimal amount of sCD40L was added (1 μg/ml or 60 nM or 100 ng/ml or 6 nM). After 24 or 48 h, supernatants were analyzed for secreted cytokines, and in some cases, the cells were stained for cell surface activation markers, as described below, gating on singlets, live cells, CD3⁻/CD19⁻/CD56⁻/CD14^high−/CD66b⁻ (when the panel allowed it), and CD11c⁺/HLA-DR⁺ to quantify the surface activation markers HLA-DR, CD80, CD83, and CD86. Cytokine assays for DC secretion and mouse plasma samples were done by Luminex multiplex assays using kits (MilliporeSigma) or in-house reagents (21) with analysis by Bio-Plex software (Bio-Rad Laboratories).

CHO cells stably transfected with a pCET 1019 HS-puro-SceI vector (MilliporeSigma) carrying a human CD40 ectodomain cDNA insert (National Center for Biotechnology Information ref sequence: NP_001241.1 residues 1–193) fused to human FAS transmembrane and intracellular domains (sequence identifier: XP_011538069.1 residues 189–350) were grown in CD CHO/M5 media (Life Technologies) with puromycin selection to establish a bulk stably transfected cell population. 10K cells in a final volume of 100 μl culture medium with the indicated concentration of anti-hCD40 Ab were incubated at 37°C in 5% CO₂. MTT was added after 4 d (American Type Culture Collection cat 30-1010K) at 10 μl per well and incubated at 37°C for 4 h and then developed with 150 μl isopropanol containing 0.04 N HCl with mixing to solubilize the reduced MTT. Absorbance at 570–650 nm was determined in a Paradigm instrument (Molecular Devices). Because only live cells can reduce MTT, a loss of absorbance is indicative of cell death. In control experiments, nontransfected CHO cells were not affected by any of the tested reagents.

A titration series of anti-hCD40 mouse or human IgG4 mAbs was added to a constant amount (1 μg/ml) of human CD40 ectodomain human Fc or mouse Fc fusion protein at 2 μg/ml, and the mixture was incubated for 1 h at 4°C and then added to an equal volume of 2E5 human CD40L stably transfected L cells (26). The cells were incubated for 1 h at 4°C, washed, then incubated with mouse-absorbed goat anti-mouse or anti-human IgG-RPE reagent (Agilent Technologies, PJR33), then washed again, and analyzed on a FACSArray Bioanalyzer and associated software (BD Biosciences), scoring the percentage of cells brighter than the gate set for cells probed only with the detecting reagent (defined as percentage of maximum binding).

Surface plasmon resonance (SPR) assay binding measurements were performed on a SensíQ Pioneer instrument (SensíQ Technologies, Oklahoma City, OK). Protein A (100 μg/ml in 10 mM NaAc [pH 4.5]) (Thermo Fisher Scientific, cat 21193) was immobilized using amine coupling chemistry on COOH2 or COOH5 sensor chips at 25°C, following the manufacturer’s recommended protocols. Running buffer was 10 mM HEPES, 3.4 mM EDTA, 0.005% Tween 20, and 8.8 g/L NaCl (pH 7.5). Subsequently, channel 1 was used to inject anti-hCD40 mAbs at a concentration of 125 nM; channels 1–2 were used to inject a dilution series of cohesin–human CD40 ectodomain (GenBank AAO43990.1 residues 22–193) or human OX40 (GenBank AAB33944.1 residues 26–211) protein (25, 12.5, 6.25, 3.125, 1.6, and 0.8 nM for anti-hCD40 12E12 and anti-hCD40 12E12–CD40L and 400, 200, 100, 50, 25, and 12.5 nM for anti-hCD40 11B6 and anti-hCD40 11B6–CD40L) at 25 μL/minute for 2 min. Cohesin–human OX40 ectodomain protein (400 nM injected for 6 min plus 5 min dissociation time at 25 μl/min) over a protein A coated the surface with bound anti-OX40 IgG1 or human OX40L linked to the L chain C termini. Control human IgG1 (hIgG1) and IgG1-Flex v1-OX40L proteins gave no or low (<6 resonance units [RU]) binding to OX40 in this assay. Surfaces were regenerated through injection of 20 mM NaOH for 1 min (25 μl). The binding data were analyzed with Qdat software (SensíQ Technologies). Qdat analysis allows setting of binding models that vary binding site numbers (n = 1–3) and compute a goodness of fit measured as the averaged residual SD. When this value is <1% of the maximum response of the curve set, this indicates an excellent fit. We found that analysis of the data using a one-site binding model was acceptable (3.5 ± 1.3% of residual SD relative to the maximum response of the curve set), but using two-site (1.4 ± 1.2%) or three-site (1.5 ± 0.5%) binding models improved the fit. RU curves shown are values normalized compared with control channels.

ExpiCHO-S cells (Thermo Fisher Scientific) stably expressing a human CD40-mCherry fusion protein (i.e., stably transfected with a pCET 1019 HS-puro-SceI vector [MilliporeSigma]) carrying a human CD40 cDNA insert (NM_001250.6 residues 31 to 864 fused to mCherry ANF29837.1 residues 330–562) were used as a model to study CD40 cluster formation. The cells were incubated in CD CHO/M5 media (Life Technologies) at a concentration of 1E6 cells/ml in a six-well plate with rounded 25-mm diameter cover slips (Electron Microscopy Sciences) at 37°C in the presence of 10 nM anti-hCD40 Ab. After 1 h, the cover slips were gently washed 2× with PBS and then resuspended in 1% paraformaldehyde in PBS (Thermo Fisher Scientific) for 10 min at RT. Two more washes in PBS followed, and the cover slips were then mounted on super frosted microscope slides (Thermo Fisher Scientific) using ProLong Gold antifade mountant with DAPI (Invitrogen). Alternatively, MDDC were cultured at a concentration of 1E6 cells/ml in a six-well plate on rounded 25-mm diameter cover slips at 37°C in the presence of 100 nM anti-hCD40 mAb fused at the H chain C terminus to a Flex V1 Doc Var1 module (23, 24) in stable noncovalent association with a cohesin-eGFP fusion protein (LDITH6 residues fused to a cohesin domain from cellulosomal scaffolding protein [Hungateiclostridium thermocellum] WP_065674352.1_residues_1044–1213 with a f1 flexible linker, AVY25163.1 residues 580–608 to ABF13214.1, eGFP residues 123–361 followed by a EPEA sequence used for C-tag affinity matrix CaptureSelect [Thermo Fisher Scientific 191307005]). After 6 h, the cover slips were processed as previously described. The slides were left overnight at RT in the dark. The day after, the slides were imaged by Leica TCS SP5 Confocal Microscopy and subsequently analyzed with Fiji-ImageJ software (University of Wisconsin-Madison). The images were taken with a confocal instrument maintaining the same setting between different treatments/slides and analyzed with Fiji software for quantification of clusters over beam area transects. Twenty-one cells were analyzed for the anti-hCD40 11B6-CD40L MDDC clustering experiment performing spatial image correlation spectroscopy, using between two and seven 64 × 64 squares per image, depending on the dimension of the cell, and seventeen cells for anti-hCD40 12E12 treatment were analyzed performing spatial image correlation spectroscopy, using between two and five 64 × 64 squares each, depending on the dimension of the cell. In the human CD40 CHO clustering experiment, 23 human CD40-mCherry CHO cells treated for 1 h with 10 nM anti-hCD40 12E12, 21 human CD40-mCherry cells treated for 1 h with anti-hCD40 11B6-CD40L, and 23 human CD40-mCherry untreated cells were analyzed performing spatial image correlation spectroscopy, analyzing between two to four 64 × 64 squares per each cell, depending on the dimension of the cell. The analysis protocol uses the software to quantitate the fluorescent intensity of labeled receptors as a function of the beam area of the confocal microscope. This provides a quantitative measure of the state of target molecule aggregation (clustering) on the cell surface, thus higher values indicate more uniform receptor distribution across the transect, whereas lower values indicate receptor clusters (27). Images were modified by adjusting brightness and contrast to highlight the differences between treatments, whereas calculations to determine differences in the cluster formation were done on raw images.

CHO cells stably transfected with a pCET 1019 HS-puro-SceI vector carrying a human CD40 cDNA insert (NM_001250.6 residues 31–864, C928) were grown in 50% CD CHO and 50% CHO M5 media (Life Technologies) with 2× GlutaMax (Thermo Fisher Scientific) and Primosin (1 ml per 500 ml, InvivoGen) with puromycin (10 μg/ml, AG Scientific, P-1033-SOL) selection to establish a bulk stably transfected cell population. Cells were dispensed in culture media with 1% BSA (250K in 50 μl) in V-bottom 96-well plates, and 100 nM of each test mAb fused at the H chain C terminus to a Flex V1 Doc Var1 module (23, 24) in stable noncovalent association with a cohesin-mCherry fusion protein (C3808, LDITH6 residues fused to a cohesin domain from cellulosomal-scaffolding protein [H. thermocellum] WP_065674352.1 residues 1044–1213 with a f1 flexible linker AVY25163.1 residues 580–608 to mCherry ANF29837.1 residues 330–562, preceded by codons encoding ML and followed by a EPEA sequence used for C-tag affinity matrix), with purification of the encoded secreted protein. The tested Abs saturate CD40 binding sites on these cells at 100 nM (data not shown). At 30-min intervals, the labeled Ab complex was added to cells, kept at 37°C in a cell culture incubator, and at the last (zero) time point, an equal volume of ice-cold PBS was added to all time points with centrifugation at 1600 rpm for 6 min with liquid removal by flicking. Then, 110 μl of cold PBS was added to one time course row (for total binding analysis), and 100 μl of ice-cold 0.1 M glycine and 0.1 M NaCl (pH 2.5) was added to a parallel time course row (i.e., acid stripping treatment to selectively remove cell surface–bound mAb). After 1 min, 10 μl of 1 M Tris HCl (pH 9) was added to the acid treatment row to neutralize the acid, and a further 100 μl cold PBS was added to all rows followed by centrifugation at 1600 rpm for 6 min with liquid removal by flicking. Note that mCherry fluorescence is not compromised by the acid treatment in this time frame (data not shown). After a final wash in PBS, cells were resuspended in 100 μl of PBS, and 75 μl was dispensed into Black Fluor Micro 2 plates (Thermo Fisher Scientific) for reading fluorescence at Ex 570_Em 625 nM in a SpectraMax Paradigm instrument (Molecular Devices).

CHO-S cells stably transfected with a plasmid construct expressing human OX40 ectodomain (GenBank AAB33944.1 residues 26–211) fused to FAS transmembrane and intracellular domains (GenPept XP_011538069.1 residues 187–350) were incubated at 4°C for 30 min with a titration series of anti-OX40 hIgG1, anti-OX40 hIgG1–OX40L, control IgG1, and control hIgG1. The cells were washed in PBS and probed with anti-human IgG-RPE reagent at 4°C for 30 min, washed with PBS, and then 500,000 cells per point were analyzed by flow cytometry on a FACSArray Bioanalyzer (BD Biosciences), scoring the percentage of cells brighter than the gate set for cells probed only with the detecting reagent (defined as percentage of P1 of Parent).

For this study, 10 μg of anti-hCD40 CP-870,893 (CP) Ab [CP is a Pfizer Ab tested in various clinical trials (8)], 10 μg anti-hCD40 11B6 Ab, or the molar equivalent of anti-hCD40 11B6-CD40L (11B6-CD40L) (12.25 μg) were injected into the i.p. cavity of six female and seven male human CD40 transgenic mice ranging in age from 7–24 wk. Human CD40 transgenic mice have the human CD40 region BAC inserted into a wild type C57BL/6 background (Taconic Biosciences) (28). Five unimmunized male C57BL/6 mice transgenic for the human CLEC-10A region were used as controls for non-CD40–activated cell and serum cytokine status. Mice were housed in microisolator cages and fed irradiated food and normal water. The Baylor Scott and White Research Institute and South Dallas Veteran’s Administration Institutional Care and Use Committees approved all mouse protocols. After 4 h, the mice were euthanized, and blood and spleens were collected. For cytokine analyses, 100 μl of blood was transferred in a 1.5 ml microfuge tube and allowed to clot at RT for 30 min and then centrifuged at 14,000 g at 4°C for 30 min. After centrifugation, serum was collected and frozen at −80°C. The rest of the blood (∼300 μl) collected in 1/10 volume 0.5 M Na₂EDTA was used for PBMC isolation. Following a 30 min RT rest period, whole blood was diluted in 1 ml of PBS and layered above 2 ml of Ficoll Paque PLUS (GE Healthcare). After centrifugation for 25 min at 670 g at 4°C with an acceleration of 1 and a deceleration equal to 0, PBMCs were collected into a 15-ml tube and washed in staining media (1× PBS with 2% FBS and 2 mM EDTA). Additional PBS washes were performed prior to staining for FACS. Single cell suspensions of spleen were prepared as described. Organs were harvested and placed in media (complete RPMI with 10% FCS) in 15-ml tubes. Using a 5-ml syringe plunger, samples were crushed against 70-μM filters into a 50-ml tube filled with PBS + 10% FCS + 1 mM EDTA. Cells were spun down and then washed in PBS. RBCs were then lysed adding 1 ml ammonium–chloride–potassium (Life Technologies, A10492-01) at RT for 2 min. The reaction was blocked with 30 ml ice-cold PBS + 2 mM EDTA. Cells were spun down, washed twice in PBS + 2 mM EDTA, then resuspended in PBS, and filtered. Cells were then stained for surface markers as described below and analyzed by flow cytometry with FlowJo software. After gating on lymphocytes, cells were gated sequentially on singlets, live cells, and IgD⁺/B220⁺ to identify the B cells, whereas gating on CD11c⁺/MHC class II⁺ cells in the IgD⁻/B220⁻ population was used to identify the DCs.

Blood from two different donors was collected in acid–citrate–dextrose tubes and processed by Ficoll Paque separation in SepMate (STEMCELL Technologies, Cambridge, MA), using three acid–citrate–dextrose tubes for each SepMate, according to the manufacturer’s protocol. Recovery of viable cells was in the range 1.1E8 of 98.5% and 1.13E8 of 99.3%. The PBMCs were processed with the EasySep CD4⁺ T cell Negative Isolation Kit (STEMCELL Technologies, 17952). Cells were first stimulated with PHA/IL-2/anti-CD3/CD28 beads. PHA stock (Remel-purified PHA, Thermo Fisher Scientific, R30852801) was in PBS and 2% FCS at 1 mg/ml and was used at 2 μg/ml. Anti-CD3/CD28 beads for polyclonal stimulation (Life Technologies Dynabeads Human T activator, ref 11161D, 40,000 bead per μl) were used at five T cells to one bead. IL-2 (pharmaceutical-grade Proleukin) at 1E7 U/ml was used at 20 U/ml. The isolated CD4⁺ T cells from both donors were individually stimulated at 1E6/ml in 50-ml conical tubes with 10% FCS and complete RPMI and kept at 37°C at 7% CO₂ for 48 h. A small subset was tested by flow cytometry at 24 and 48 h to validate increased expression of OX40 (data not shown). After 48 h, anti-CD3/CD28 beads were removed using a magnet (STEMCELL Technologies), and cells were labeled with CFSE to enable monitoring of the proliferation status using the Cell Trace Labeling Kit (Thermo Fisher Scientific, 34554) reconstituted with 18 μl DMSO to make a 5 mM stock. Cells were washed twice in PBS warmed to 37°C following the bead depletion. Cells were then resuspended in 1.25 μM CFSE in the warm PBS at a concentration of 1E6/ml at RT for 10 min, then quenched with 10 volumes of ice-cold 10% FCS complete RPMI, and kept at 4°C for an additional 5 min. Cells were washed twice in media or 2% FCS with PBS and then resuspended in media and added at 100,000 cells per well (50 μl) into V-bottom culture plates for stimulation. Test proteins were added as a dilution series from 2 μM to 0.05 nM. Cell controls without test vehicles were included, and some wells were also stimulated with anti-CD3/CD28 (100,000 T cells to 40,000 beads, ratio of 2.5 T cells to one bead) as positive signal controls. Cells were then cultured an additional 5 d, and culture supernatants were analyzed by Luminex for IL-5, IL-13, IFN-γ, and TNF-α cytokine levels (MilliporeSigma). The remaining cells were analyzed by flow cytometry (FACSCanto II, BD Biosciences) for CFSE proliferation after staining with LIVE/DEAD Fixable Aqua Dead Cell Stain (Thermo Fisher Scientific) and anti-CD4 mAb using FlowJo software (Ashland, OR), gating on live cells and CD4⁺.

Human cells were transferred to a V-bottom plate, washed twice in PBS, and incubated for 20 min at 4°C with LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Thermo Fisher Scientific, cat L34965) at a 1:50 dilution in a volume of 50 μl. Cells were washed twice with PBS and incubated for 30 min on ice with the mix of Abs in a volume of 50 μl PBS + 2% FCS + 1 mM EDTA. Finally, cells were washed and resuspended in 200 μl BD stabilizing fixative (BD Biosciences, cat 338036) diluted 1:3. All analysis plots were pregated on live (using LIVE/DEAD stain) and singlet events. Cells were analyzed with a FACSCanto II, FACSArray Bioanalyzer, or an LSR Fortessa (BD Biosciences). Data were analyzed with FlowJo software. The following Abs were used for analysis of human DC activation. One panel included human CD80–PE, clone L307.4, ref 340294 (Becton Dickinson [BD]); human CD83–allophycocyanin, clone HB15e, ref 551073 (BD); human CD86–FITC, clone 2331 (FUN-1), ref 555657 (BD); human CD11c–PE–cyanine 7, clone 3.9, ref 25011642, (eBioscience); human HLA–DR-V450, clone G46-6, ref 561359 (BD); and human CD40-PE–cyanine 5, clone 5C3, ref 555590 (BD). A different panel was composed by human CD14–FITC, clone M5E2, ref 301804 (BioLegend); human CD66b–FITC, clone G10F5, ref 305104 (BioLegend); human CD56–FITC, clone NCAM, ref 340410 (BD); human CD19–FITC, clone 4G7, ref 347543 (BD); human CD3–FITC, clone HIT3A, ref 555339 (BD); human CD11c–allophycocyanin–cyanine 7, clone Bu15, ref 337218 (BioLegend); human HLA–DR-BV711, clone G46-6, ref 563696 (BD); human CD83–BV421, clone HB15e, ref 562630 (BD); and human CD86–PE–cyanine 5, clone IT2.2, ref 555666 (BD). Human CD19–allophycocyanin, clone HIB19, ref 555415 (BD) and human CD3–PerCP, clone SK7, ref 347344 (BD) were used in the human B cell proliferation panel. The following panel was used for the in vivo mouse experiments: mB220–Alexa Fluor 488, clone RA3-6B2, ref 103225 (BioLegend); mouse CD69–PE, clone H1.2F3, ref 104508 (BioLegend); mouse CD11c–PE-Cy7, clone N418, ref 117318 (BioLegend); mouse MHC class II–Alexa Fluor 700, clone M5/114.15.2, ref 107622 (BioLegend); mouse IgD-BV711, clone 11/26c.2a, ref 405731 (BioLegend); mouse CD86-BV605, clone GL-1, ref 105037 (BioLegend); and Fc block, ref 70-0161-U500 (Tonbo Biosciences). Human CD4–PE-Cy7, clone SK3, ref 557852 (BD) was used in the OX40 agonistic activity assay.

Data are presented as means (±SEM). Statistical significance was determined by Student t test with or without Welch correction. A p value <0.05 was considered statistically significant. GraphPad Prism software was used for statistical calculations.

CD40 expressed on B cells, when engaged by CD40L expressed on Ag-activated CD4⁺ helper T cells secreting cytokines IL-4 and IL-21, drives the proliferation of B cells, events that are typically confined to germinal centers of lymphoid organs (1). We tested a matched panel of anti-hCD40 Abs, formatted as human IgG4 (hIgG4) and human κ L chain, for their efficiency in driving proliferation of human peripheral B cells in the presence of IL-4 and IL-21 (Fig. 1A). These Abs covered a ≥1000-fold range of agonist efficacies with rank order CP > 12E12 ≥ 12B4 > 11B6 > 24A3 [CP is a Pfizer Ab (8), and the others were developed in-house]. Repeating this assay in the presence of a fixed, suboptimal concentration of sCD40L had no effect on the dose response of the 12B4 and 12E12 Abs, slightly increased the potency of the CP Ab, but synergized with the 11B6 and 24A3 Abs to greatly (≥100-fold) increase their efficacy (Fig. 1A, Supplemental Table I, upper panel). The synergistic cooperation of these two weak CD40 agonist signals suggests that the interaction of these specific mAbs with CD40 potentiates productive interaction with CD40L, or vice versa. Cooperative effects between substances that activate distinct receptors on a single cell is a well-known phenomenon, e.g., as observed between agonistic anti-CD40 mAbs and FcRII (15), but in this case, our two activating agents are acting on the same receptor type. (Fig. 2)

Sentinel DCs, when exposed to foreign Ags and pathogen-derived danger signals, process and present Ag peptides in their MHC to cognate Ag-specific T cells (29). CD40 expressed on DCs interacts with CD40L expressed on adjacent Ag-activated T cells, and this event is critical for initiating immunity, partly via increasing expression of cell surface DC activation molecules (e.g., CD86 and HLA) and invoking secretion of inflammatory cytokines by the DCs (30). Thus, we assayed the panel of anti-hCD40 hIgG4 mAbs for their ability to initiate cytokine secretion on matured MDDC. This same Ab panel elicited cytokine production over a similar >1000-fold range of efficacies as observed with the B cell proliferation assay, with a rank order CP > 12B4 > 12E12 ≥ 11B6 ≥ 24A3 (Fig. 1B, 1C, Supplemental Table I, lower panel). These results were consistent with the increased expression of surface activation markers analyzed through flow cytometry. When assayed with a suboptimal dose of sCD40L, there was no effect on the dose response of the 12B4 and 12E12 Abs. However, low-dose sCD40L increased the potency of the CP Ab by ≈10-fold, whereas strong synergy for DC activation (>100-fold) was again observed between sCD40L and the 11B6 and 24A3 Abs (Fig. 1B-C). For the 11B6 and 24A3 Abs, the effect of sCD40L was dose dependent, but even at the low 6 nM sCD40L dose, striking augmentation in cytokine production was observed. This effect was consistent between two low sCD40L doses tested (6 and 60 nM). Note there was also an increase in efficacy (i.e., maximal cytokine production achieved) between Ab alone and Ab with sCD40L, unlike the responses in B cell proliferation in which only potency was increased (Fig. 1A), indicating a greater potential for cooperation between these two agonist types on DCs.

The panel of anti-hCD40 mAbs we studied bind with relatively high affinity to CD40 as determined by SPR analysis, with rank order for on-rate of 11B6 > 12B4 > 12E12 > 24A3 > CP and rank order for off-rate 12B4 > 11B6 > 24A3 > 12E12 > CP (Table I). Previous studies with agonistic anti-CD40 Abs have shown no apparent correlation between these kinetic parameters and activation potential (31), and the SPR data for the mAbs within this study are in accordance with this conclusion.

One possible mechanism for the synergy between sCD40L and anti-hCD40 mAb activation of CD40 could be via their access to separate sites on the CD40 ectodomain. It is known that the agonistic anti-CD40 mAb CDX-1140 interacts with CD40 at a site distinct from CD40L, and analogous synergy between sCD40L and this mAb was observed (18). Thus, we tested our panel of anti-hCD40 IgG4 Abs for their ability to prevent the binding of CD40 to CD40L expressed on the surface of L cells. The 12B4 and 12E12 Abs in stoichiometric amounts prevented CD40 binding to cell surface CD40L, whereas >20-fold higher levels of the CP Ab were required to even partially block CD40 binding to CD40L, but the 11B6 and 24A3 Abs and the previously described agonistic S2C6 Ab (3) had a minimal effect on CD40L binding to CD40 even at the highest mAb doses (Fig. 2). These data were consistent for these mAbs either as the original mouse Abs or reformatted as chimeric hIgG4. Based on their affinity constants (Table I), all of these Abs would have fully occupied CD40 binding sites at the key discriminating concentration of 1 μg/ml in which 12E12 and 12B4 mAbs fully block CD40L binding. These data show that the 12B4 and 12E12 mAbs bind to sites on CD40 that are absolutely required for CD40–CD40L interaction, whereas the CP, 11B6, and 24A3 mAbs bind to CD40 sites with minimal interference to CD40–CD40L interaction. Thus, synergy between sCD40L and anti-hCD40 mAb for B cell and DC activation is associated with simultaneous access of both these agonists to distinct parts of CD40L.

Synergistic cooperation between sCD40L and agonistic anti-CD40 mAbs may be a valuable property in vivo, e.g., via allowing the CD40L on activated T cells access to CD40 on DCs already occupied by the mAb. Alternatively, agonistic anti-CD40 mAb and sCD40L could be delivered simultaneously in vivo for possible therapeutic benefit via the enhanced CD40 activation observed in vitro. A trimeric form of sCD40L has shown efficacy in preclinical studies (5) and may become available in the future to enable clinical validation of combining sCD40L with synergizing agonistic mAbs. In this study, we explored the concept of physically associating sCD40L with agonistic mAb by direct fusion with an obvious potential benefit of establishing a highly active agonist as a single agent. For this purpose, the entire ectodomain of human CD40L was fused to the L chain C termini of the anti-hCD40 mAbs via a glycosylation-rich flexible linker sequence [called Flex V1 or ASQTPTNTISVTPTNNSTPTNNSNPKPNPAS (24)]. These anti-hCD40–CD40L mAbs were efficiently expressed in 293 and CHO cells as homogenous-secreted products (Supplemental Fig. 1).

We selected two non-CD40L–blocking anti-hCD40 IgG4 mAbs 11B6 and CP and the CD40L-blocking 12B4 mAb to fuse to CD40L and tested them relative to the nonfused mAbs for their efficacy in eliciting B cell proliferation. All of these CD40L-fused mAbs were highly potent in this assay, matching the efficacy of the highly potent CP mAb coadministered with suboptimal sCD40L (Fig. 3A). The CD40L adduct greatly increased the potency of the parent mAbs (>1000-fold for 11B6 and >100-fold for CP and 12B4), and the increase for the 11B6 mAb was >10-fold more robust that when coadministered with sCD40L but was less for the intrinsically more potent 12B4 and CP mAbs (Fig. 3A). CD40L fused to the control IgG4 mAb was also active in inducing B cell proliferation, but ≥100-fold less than the 11B6, 12B4, and CP CD40L fusion proteins (Fig. 3A). The enhanced activity of hIgG4-CD40L relative to sCD40L is likely due to the divalent nature of this construct. Note that although the 12B4 mAb fails to synergize with sCD40L, CD40L fusion enhances 12B4 potency.

A full panel of anti-hCD40 IgG4 mAbs fused to CD40L was then developed and tested relative to the nonfused mAbs for their efficacy in eliciting DC activation. Remarkably, directly linking CD40L to all the mAbs except 12E12 dramatically increased their efficacy (i.e., the maximal response) compared with the synergy observed with adding unlinked sCD40L to the 11B6, CP, and 24A3 mAbs (Fig. 3B, 3C). Furthermore, consistent with the B cell proliferation assay (Fig. 3A), mAb 12B4, which competes directly with CD40L for CD40 occupancy, also benefited greatly from CD40L fusion (Fig. 3B, 3C). However, the CD40L fusion to the CD40L-blocking 12E12 mAb increased efficacy at the 10 nM dose, but this effect was lost at the higher 100 nM dose. The control hIgG4-CD40L fusion was more active than hIgG4 + sCD40L, as would be expected from its dimer configuration. In an independent experiment with a similar panel of anti-hCD40 and control IgG4 mAbs, with and without added or directly linked CD40L, similar effects on enhanced expression of DC activation markers CD83, CD86, and HLA-DR were observed (Fig. 3D–F), and the corresponding cytokine secretion data are presented in Supplemental Fig. 2, together with similar effects on DC activation markers in a third independent experiment.

FAS (CD95) belongs to the TNFR family and contains an intracellular “death domain” that can trigger apoptosis in response to its physiologic ligand, FASL (32). We constructed a fusion protein expressing human CD40 ectodomain residues 21–193 fused to human FAS residues 187–350 and established stably transfected CHO cells expressing the CD40 ectodomain linked to the FAS transmembrane and intracellular domains. CD40 agonists elicit killing of these cells as determined by loss of mitochondrial reduction of the tetrazolium salt MTT (33). FAS and CD40 are in the same TNFR family, and mechanisms of receptor activation (external to the cell) are similar, but the intracellular signaling pathways are different (i.e., apoptosis versus selected cytokine and cell surface marker activation). This fusion construct provides a convenient surrogate assay format for analysis of CD40 activation based on transfected CHO cells. In this assay, anti-hCD40 11B6 IgG4 and anti-hCD40 12E12 IgG4 show similar efficacy (as determined by maximal decrease in MTT reduction) and similar potency (EC₅₀ ≈ 2.5 and 1 nM, respectively), but anti-hCD40 11B6-CD40L IgG4 has increased efficacy (i.e., greater maximal decrease in MTT reduction) and significantly increased potency (EC₅₀ ≈ 2.5 pM) (Fig. 4). Thus, CD40L fusion to the partial agonist anti-hCD40 11B6 mAb can greatly increase potency and efficacy on three distinct CD40-bearing cell types.

We used SPR analysis to probe the impact upon the CD40 binding kinetics of anti-hCD40 11B6 and anti-hCD40 12E12 mAbs of CD40L fused to their L chain C termini by immobilizing them onto a protein A/G surface and flowing soluble human CD40 ectodomain over them in the liquid phase. The CD40L adduct on the anti-hCD40 12E12 mAb did not significantly alter the Ab on- or off-rates compared with the parental anti-hCD40 12E12 mAb (Fig. 5, upper panels). This was expected because the anti-hCD40 12E12 mAb competes for the CD40L binding site on CD40, and a human IgG4 control mAb with CD40L fused in a similar manner to the L chain showed no detectable binding to CD40 in this format. In contrast, the CD40L adduct on the anti-hCD40 11B6 mAb clearly stabilized the Ab off-rate compared with the parental anti-hCD40 11B6 mAb (Fig. 5, lower panels). Specifically, whereas the on-rate was marginally impacted, the off-rate decreased by ∼15-fold, indicating cooperativity between the anti-hCD40 mAb and CD40L in binding to CD40. This is likely due to the multivalent nature of the anti-hCD40 mAb–CD40L fusion.

To investigate if the location and type of linking of the CD40L fusion was important for increased binding affinity and activation efficacy, we tested a series of anti-hCD40 11B6 IgG4 Abs with and without the Flex V1 linker joining the human CD40L ectodomain to either the L chain or H chain C termini. SPR analysis showed that compared with the parental humanized anti-hCD40 11B6 IgG4 mAb, all forms with fused CD40L had characteristic slight (≤2-fold) reductions in on-rate and significant (>6-fold) decreases in the off-rate (Fig. 6A).

All the isoforms tested elicited similar highly potent activities for activating cytokine secretion in human MDDC cultures (Fig. 6B). Also, in the CD40-FAS CHO killing assay, all the 11B6-CD40L fusion variants had very high potency and efficacy compared with control CD40 agonist proteins (Fig. 6C). Thus, for the 11B6 mAb, the position and mode of attachment via direct fusion or through a flexible linker sequence had little impact on the increased affinity and biological activity associated with the fusion to CD40L. The 11B6-CD40L fusion variants had >10-fold higher efficacy than the highly active megaCD40 trimer (Fig. 6B).

We compared the rate and extent of CD40-mediated internalization of anti-hCD40 11B6-CD40L versus anti-hCD40 12E12. These two Abs are well matched for their binding to CD40 based on SPR analysis (KD [equilibrium dissociation constant] of 12 and 28 nM, respectively; (Fig. 5), but they differ dramatically in potency of CD40 activation (Fig. 3). Because ligand engagement leads to the formation of cross-linked CD40 clusters on the cell membrane, an event followed by CD40 internalization and downstream signaling (34), we compared the ability of anti-hCD40 11B6-CD40L versus anti-hCD40 12E12 to cluster and internalize cell surface CD40.

We first compared the ability of these two anti-hCD40 Abs to induce the formation of cross-linked CD40 clusters on the cell membrane. We used MDDC as a model to visualize cluster formation through confocal microscopy. Treating the cells for 6 h at 37°C with 100 nM anti-hCD40 11B6-CD40L induced stronger CD40 cluster formation compared with the same treatment with anti-hCD40 12E12 (Fig. 7A, 7C). We confirmed this using CHO cells expressing a human CD40-mCherry fusion protein, treated with 10 nM of anti-hCD40 12E12 or 10 nM anti-hCD40 11B6-CD40L for 1 h or left untreated (Fig. 7B, 7C). These data are concordant with the increased signaling potency of anti-hCD40 11B6-CD40L mAb because CD40 clustering is likely the initial trigger for CD40 activation.

We then used CHO cells expressing human CD40 as a model and assayed anti-CD40–mediated binding and internalization of a noncovalently attached mCherry module. Binding of both Abs was rapid and reached saturation within ∼30 min at 37 and 0°C (Fig. 8A, 8B). Treating the cells in cold isotonic acid buffer (pH 2.5) for 1 min removed ∼75% of the cell-associated anti-hCD40 12E12 label when binding was performed at 0°C (Fig. 8A). The detectable binding at the zero time point likely reflects binding and internalization occurring during the initial ∼6 min centrifugation and washing step. When binding was performed at 37°C, there was a trend (residual label 29 ± 3% versus 26 ± 3%, NS) to greater label retention at 37°C, perhaps reflecting slight internalization. In contrast, when anti-hCD40 11B6-CD40L binding was performed at 0°C, the acid stripping removed only ∼60% of the cell-associated label versus the ∼75% observed with the anti-hCD40 12E12 mAb, and this difference between the two mAbs is significant (p < 0.0001, Fig. 8B), indicating anti-hCD40 11B6-CD40L internalization, as defined by acid resistance, is readily detectable, even to a low extent, at 0°C. Internalization of the anti-hCD40 11B6-CD40L label was much greater at 37°C, with acid-resistant label increasing from 50–90% over the 4.5-h time course (Fig. 8B). Thus, anti-hCD40 11B6-CD40L internalizes to a much greater extent than anti-hCD40 12E12, with significant acid-resistant signal detectable even at 0°C. This property is a sole result of the CD40L adduct because the binding and internalization properties of the anti-hCD40 11B6 mAb without fusion to CD40L were very similar to the anti-hCD40 12E12 mAb (Fig. 8C).

In the simplest molecular state, anti-hCD40 11B6-CD40L is a prototypical IgG Ab that is a typical Ab dimer of H and L chain pairs connected by covalent inter–H chain disulfide bonds in the H chain hinge region and noncovalent interactions in the H and L chain constant domain regions (35) with CD40L monomer units linked to the H or L chain C termini. Native CD40L forms a self-assembled trimer similar to that found for other members of the TNF family (36), and this raises the possibility that the superagonist activity of anti-hCD40 11B6-CD40L is driven by intramultimeric CD40L–CD40L associations, resulting in higher order molecular structures, i.e., aggregates. Furthermore, recombinant Abs have well-recognized propensities to self-aggregate, driven by intrinsic properties defined by their sequences, as well as extrinsic factors such as pH, temperature, excipients, and concentration [reviewed in (37)]. These facts raise the possibility that the high potency of anti-hCD40 11B6-CD40L results from such higher order aggregates that drive e.g., enhanced CD40 clustering.

To address this issue, we analyzed by native SEC protein A–purified anti-hCD40 and anti-hCD40 11B6-CD40L. The anti-hCD40 humanized 11B6 mAb we analyzed was largely (∼78%) not aggregated [i.e., prototypical IgG 2H + 2L chain composition, with an SEC estimated mass ∼200 kDa, with lesser amounts of high molecular mass species aggregates of ∼520 kDa and 670 kDa (Supplemental Fig. 3A)]. In two different preparations of humanized anti-hCD40 11B6-CD40L with CD40L attached to the L chain C terminus via a flexible linker sequence, SEC analysis revealed the presence of ∼45% (Fig. 9A) and ∼49% nonaggregated forms (Fig. 9B) (SEC estimated mass ∼295 kDa) with the residual material being higher molecular mass species aggregates of ≥670 kDa (Fig. 9A, 9B, Supplemental Fig. 3B–C). We separated the nonaggregated form from higher m.w. species by preparative SEC and characterized the fractions pooled according to three m.w. categories (highly aggregated, aggregated, and nonaggregated) by SDS-PAGE (Fig. 9A, 9B, Supplemental Fig. 1) and analyzed their potency for DC activation (Fig. 9C–F).

Compared with the unfractionated starting anti-hCD40 11B6-CD40L material, all of the fractions analyzed representing the highly aggregated, aggregated, and nonaggregated Ab induced similar highly potent DC activation as measured by cytokine secretion and cell surface marker activation, and this result was observed in both independent fractionation experiments (Fig. 9C–F). Thus, the observed partial aggregation of the anti-hCD40 11B6-CD40L does not have an impact on the high potency for DC activation. To confirm the stability of the collected fractionated material used in the DCs activation assay, the nonaggregated form from a freshly thawed vial of anti-hCD40 11B6-CD40L was separated according to its m.w., dialyzed in storage buffer, filtered, and immediately run on SEC. To mimic the conditions of the material used in the DCs activation assay, the nonaggregated fraction was left at 4°C for 1 wk. Subsequently, it was run again on SEC to investigate its stability. The results show that the purified nonaggregated fraction is perfectly stable at these conditions for the analyzed period of time (Supplemental Fig. 3D).

Anti-hCD40 CP mAb adjuvant infusion in cancer patients undergoing chemotherapy triggers immune activation detected by increased plasma inflammatory cytokines (i.e., cytokine release syndrome), increased B cell expression of costimulatory molecules, and transient depletion of B cells (7). In these patients, dose-limiting toxicity was determined to be 0.2 mg/kg, although 0.3 mg/kg was the limit determined in patients receiving this anti-CD40 agonist alone (38).

To compare the biological activity of anti-hCD40 11B6-CD40L, anti-hCD40 11B6, and the clinically relevant anti-hCD40 CP in vivo, we tested their short-term (4 h) effects on immune cells in human CD40 transgenic C57BL/6 mice. Administration of these agents to nontransgenic C57BL/6 mice served as ref controls because these anti-hCD40 Abs have no reactivity with mouse CD40. In blood and spleen, we measured CD86 and CD69 levels on B cells and DCs as a measure of activation, the percentage of B cells and DCs relative to total lymphocytes as a measure of cell depletion, as well as measuring blood cytokine levels. By all measures anti-hCD40 11B6 (10 μg administered i.p. or ≈ 0.4 mg/kg) gave no detectable effects compared with controls, reflecting the weak agonist activities of this Ab (Fig. 1, 2). However, compared with anti-hCD40 11B6 and negative controls, an equimolar dose of anti-hCD40 11B6-CD40L mAb (12.25 μg administered i.p.) increased expression of CD86 levels on blood B cells and splenic DCs, increased CD69 levels on both splenic and blood B cells, depleted B cells from the blood, and elicited or increased IL-6, MCP-1, and IP-10 levels in the blood (Fig. 10A, 10B). These in vivo data recapitulate the change from weak agonist properties of anti-hCD40 11B6 mAb to the strong agonist efficacies observed in vitro with anti-hCD40 11B6-CD40L mAb (Figs. 2, 3, 4, 5, and (8). Compared with the anti-hCD40 11B6 mAb and negative controls, an equimolar dose of the strong agonist anti-hCD40 CP mAb (10 μg administered i.p.) also elicited increased expression of CD86 levels on blood B cells, increased CD69 levels on both splenic and blood B cells, and elicited or increased IL-6, MCP-1, and IP-10 levels in the blood (Fig. 10A, 10B). However, expression of CD86 on splenic DCs and blood B cells was not significantly different from responses to the anti-hCD40 11B6 mAb and negative controls (Fig. 10A). The overall high variation in responses, likely compounded by sex and age differences in the tested mice, masked the potential significance of differences measured directly between the anti-hCD40 11B6-CD40L and anti-hCD40 CP mAbs, but these in vivo data clearly trend toward higher potency for the measured parameters elicited by anti-hCD40 11B6-CD40L mAb.

OX40 (also known as TNFRSF4 or CD134) is transiently expressed by CD4⁺ and CD8⁺ T cells following TCR stimulation and interacts with OX40L expressed on activated DCs to enhance proliferation and expression of effector molecules and cytokines by these T cells [reviewed in (39)]. Thus, both OX40L-Fc fusion protein and agonistic anti-OX40 Abs are under clinical study as adjuvants for boosting immunity against e.g., cancer (40, 41). Agonistic anti-OX40 Abs are typically configured as IgG1 to facilitate cross-linking of the Fc receptors to help cluster surface OX40, enabling activation of downstream signaling pathways (42–44). However, soluble versions of OX40L trimer can act independently of Fc cross-linking (45).

To test if linking natural ligand to an agonistic Ab is a general strategy for enhancing efficacy and potency, we linked human OX40L to the L chain C terminus of an agonistic anti-OX40 IgG1 Ab in clinical development (humanized OX40 mAb24 from patent US 9738723). (Fig. 11A shows that both the humanized anti-OX40 Ab24 IgG1 and anti-OX40 Ab24-OX40L fusion IgG1 proteins bound similarly to CHO cells expressing the human OX40 ectodomain. The parental anti-OX40 mAb has a relatively low off-rate; however, SPR analysis revealed a ∼2-fold enhancement in off-rate associated with the addition of OX40L to the anti-OX40 mAb L chain (Fig. 11B).

Anti-OX40 hIgG1, anti-OX40–OX40L hIgG1, and hIgG1-OX40L proteins were tested for their ability to evoke proliferation and cytokine production of primed human CD4⁺ T cells in vitro. In the absence of any Fc cross-linking agent, picomolar levels of anti-OX40–OX40L were sufficient to drive both proliferation and cytokine production in this assay. In contrast, nanomolar amounts of hIgG1–OX40L were required to drive proliferation and cytokine production, whereas anti-OX40 was inactive in this assay. These data (Fig. 11C) show that fusing OX40L to agonistic OX40 Ab profoundly changes the characteristics of the separate entities: 1) the anti-OX40–OX40L IgG1 fusion increases potency of the OX40L IgG1 by ≥2 logs; 2) the efficacy (maximum response) is increased; and 3) the activity is independent of Fc cross-linking.

Agonistic anti-CD40 mAbs are under intense study as adjuvants for vaccines or checkpoint inhibitor therapies, and this has led to a multitude of approaches to increase their potencies. An intriguing aspect of agonistic anti-CD40 mAbs is that some do not block CD40L action and can cooperate with sCD40L. To further explore this phenomenon, we tested a panel of anti-hCD40 mAbs with a range of in vitro activation potencies for cooperation with soluble CD40L. As a high potency benchmark control, the panel included the clinical candidate agonistic Ab CP that does not rely on FcγR binding for its functional activity (46). Given the known effects of FcγR binding on CD40 activation on some anti-CD40 mAbs (15), we configured all tested Abs onto an identical disulfide bond stabilized and FcR null human IgG4 framework (47). Our results confirmed that some anti-hCD40 mAbs can synergize with soluble CD40L, and this correlated with their inability to block CD40L binding to CD40. Interestingly, the extent of synergy was markedly greater on DCs compared with B cells, which were overall markedly more sensitive to the CD40 agonists (compare Supplemental Table I upper versus lower panels). Variation of cell type potency is a common effect and reflects both differences in receptor number and efficiency or complexity of associated intracellular signaling pathways (48).

The potential therapeutic benefit of synergy between agonistic anti-hCD40 mAb and sCD40L could be exploited by coadministering the mAb and sCD40L; however, this has not been tested in vivo and also presents significant logistic and regulatory hurdles for clinical application of two separate products. Thus, we explored directly fusing CD40L to the anti-hCD40 mAbs to create single entity agonists, and this resulted in surprising increases in potency and cell type–specific efficacy compared with coadministration of separate anti-hCD40 mAb and sCD40L. In particular, maximal cytokine production (i.e., efficacy) by MDDC was severalfold higher with anti-hCD40–CD40L constructs compared with anti-hCD40 mAb agonists, even in synergy with sCD40L. The potency of these constructs was also greatly superior to megaCD40L, a high activity trimer form of sCD40L (4).

Given the extensive data showing dependency of anti-CD40 mAb potency via cross-linking to FcR (15, 46, 49, 50) or increased potency of engineered multivalent sCD40L (4, 16, 51), an obvious mechanistic explanation for the high potency and efficacy of the tetravalent anti-hCD40–CD40L constructs is via increased cross-linking of cell surface CD40. The observed stabilization of soluble CD40 off-rate when bound to immobilized anti-hCD40 11B6 mAb most likely reflects increased affinity via spatially increased concentration of CD40 binding through both anti-hCD40 mAb V regions and proximal-linked CD40L. However, the high clustering of cell surface CD40 via anti-hCD40–CD40L mAb compared with anti-hCD40 mAb is direct evidence for an enhanced cross-linking mechanism of action. Ligand-induced cell surface receptor clustering into lipid rafts is a well-recognized initial step in CD40 activation (52–54). Such clustering initiates recruitment of intracellular signaling modules and triggering of internalizing endocytic mechanisms designed to either dissociate the ligand–receptor complex and recycle the receptor to the cell surface or to divert the ligand–receptor complex into degradation pathways (1, 34). The increase in cell surface CD40 clustering via anti-hCD40 11B6-CD40L mAb compared with anti-hCD40 12E12 mAb correlates with the considerable increase in resistance to acid stripping of the bound anti-hCD40 11B6-CD40L mAb into cells expressing CD40, further supporting a superagonist mechanism based on tetravalent-induced CD40 clustering, triggering robust internalization, and signaling.

The in vitro superagonist properties of anti-hCD40 11B6-CD40L mAb were recapitulated in vivo. One day after a single dose of anti-hCD40 11B6 mAb administration to human CD40 transgenic mice, there was barely detectable induction of CD40 activation markers CD86 and CD69 on B cells and DCs from blood and spleen; however, a molar equivalent single dose of anti-hCD40 11B6-CD40L mAb resulted in high levels of expression of these markers. Also, as observed in human studies (7), the strong CD40 activation via anti-hCD40 11B6-CD40L resulted in B cell depletion. The in vivo study did not fully recapitulate observed differences in vitro between efficacy of the anti-hCD40 11B6-CD40L and anti-CD40 CP mAbs, although the data favored higher efficacy of the anti-hCD40 11B6-CD40L mAb. This could be further explored via additional dose ranging and time course studies. The potential benefit of higher agonist activity of anti-CD40–CD40L mAbs for human adjuvant therapies will require additional studies related to drug manufacture cost and defining dose-limiting toxicity levels but could also depend on advantages in the longevity of DC stimulation in vivo. Such studies are beyond the scope of this work.

The superagonist anti-hCD40–CD40L Abs we have engineered may have application as adjuvants with likely dose-sparing advantages for use in similar settings as current clinical applications of highly agonistic anti-hCD40 Abs. However, our main purpose for this work was to engineer highly DC-activating anti-hCD40 Ab–based vehicles for Ag delivery as vaccines with potential for adjuvant-free administration. Our current CD40-targeting vaccine platform elicits powerful Ab and T cell responses but requires coadministration of the TLR3 agonist poly IC for maximal efficacy (25, 55). This may result from loss of agonist potency by the anti-hCD40 12E12 vehicle mAb when fused to Ags (24), which can be potentially overcome by using e.g., our newly developed superagonist anti-hCD40 11B6-CD40L vehicles. Obviously, intrinsic adjuvant activity via anti-hCD40–CD40L–Ag vaccines is an attractive area to explore for simplifying the formulation to one component in excipient to solubilize or stabilize. In follow-up unpublished work, we have explored this notion and confirmed that anti-hCD40 11B6-CD40L–Ag fusions retain high agonist activity, and this fundamentally changes in vitro and in vivo immune responses relative to currently used anti-hCD40 12E12–Ag fusions.

Importantly, using antireceptor mAb–ligand fusion to generate pharmacologically useful superagonists may be a general concept. Anti-OX40 agonist Abs are an emerging novel class of adjuvants finding clinical utility in e.g., cancer treatment, sometimes in combination with checkpoint inhibitors (41, 56). Thus, we selected a clinical candidate anti-OX40 mAb agonist and fused OX40L to this mAb. This improved in vitro potency compared with the OX40L IgG1 by ≥2 logs, the anti-OX40 mAb activity became independent of Fc cross-linking, and the efficacy (maximum response) increased. This suggests that the strategy of fusing natural ligand to agonistic mAb for enhancing efficacy and potency may be generally applicable beyond the CD40–CD40L receptor–ligand system. It is important to note that this strategy combines the typical very high affinity of therapeutic candidate antireceptor mAbs with the usually lower affinity of the soluble natural ligand to generate a multivalent mAb with increased affinity and cross-linking capacity. Others have described mAbs fused to ligands for targeting the ligand to the surface of specific cell types [e.g., cancer cells (57)], but our concept retains the native cell type receptor–ligand specificity with great benefit to agonist efficacy and potency.

We thank Yong-Jun Liu, the previous Baylor Institute for Immunology Research Director, for initiating the development of anti-CD40 Abs with enhanced agonist potency. We thank the reviewers for thorough and insightful comments and suggestions for improving this study and manuscript.

V.C., S.Z., M.M., Y.L., and G.Z. are named inventors on patent applications based on this work held by Inserm Transfert. The other authors have no financial conflicts of interest.