Vectors containing cDNA encoding mouse Ox40l (Tnfsf4), mouse mannose-binding lectin 1 (Mbl1), mouse 4-1bbl (Tnfsf9), mouse Cd70, mouse Gitrl (Tnfsf18), or mouse dihydrofolate reductase (Dhfr) were previously described (23–25). The MBL-OX40L gene ( Supplementary Figure S1A ) in the pCAGGS plasmid (GenBank accession number LT727518.1) was also previously described (23). The MBL-OX40L protein consists of an N-terminal PA-peptide tag (GVAMPGAEDDVV), a collagen-like domain of Mbl1 (¹⁸Ser-¹²⁶Gly), and a C-terminal extracellular THD domain of Ox40l (⁵¹Ser-¹⁹⁸Leu). The Fc-MBL-OX40L gene ( Supplementary Figure S1B ) was constructed by inserting the MBL-OX40L gene into the 3’ end of the human IgG1 Fc gene using In-Fusion cloning (638947, In-Fusion Snap Assembly Master Mix, Takara Bio, Shiga, Japan). The Fc-MBL-OX40L protein is composed of an N-terminal Fc (hinge, CH₂, and CH₃ domains) and a C-terminal MBL-OX40L. The gene for Fc-scOX40L ( Supplementary Figure S1C ) was previously reported (24). The genes for scOX40L-Fc ( Supplementary Figure S1D ) and MBL-scOX40L ( Supplementary Figure S1E ) were constructed using In-Fusion cloning. The genes for Fc-sc4-1BBL, Fc-scCD70, and Fc-scGITRL were previously reported (24). The pCAGGS plasmid encoding mouse Ox40-Fc was also previously described (23).

To produce recombinant proteins, the expression vector was transfected into DG44-CHO cells (Dhfr negative, CRL-9096, ATCC, Manassas, VA, U.S.A.), and a stable DG44-CHO cell clone was established. In brief, DG44-CHO cells were cultured in IMDM (098-06465, FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) containing 0.1 mM hypoxanthine, 0.016 mM thymidine, 0.002 mM methotrexate (MTX), 50 µM 2-mercaptoethanol (2-ME), 2 mM L-Alanyl-L-Glutamine, 10% FCS, 100 U/mL penicillin, and 100 µg/mL streptomycin. The cells were then transfected with a pcDNA3.1 vector encoding both Dhfr gene and neomycin resistance gene, and concurrently with the pCAGGS vector encoding MBL-OX40L, Fc-MBL-OX40L, Fc-scOX40L, scOX40L-Fc, or MBL-scOX40L, at a 1:10 ratio. The transfected DG44-CHO cells were cultured in complete IMDM supplemented with 0.5 mg/mL G418, without hypoxanthine, thymidine, and MTX. To induce a high degree of gene amplification, the MTX concentration was increased stepwise from 50 to 800 nM, and individual clones were isolated from heterogeneous cell pools by limiting dilution. To select cell clones with the highest productivity, the expression level of recombinant protein in the culture supernatant was analyzed by using a sandwich enzyme-linked immunosorbent assay (ELISA). Selected DG44-CHO cell clones were cultured either in IMDM containing 2% FCS or BalanCD™ CHO GROWTH A/BalanCD™ CHO FEED4 (554-34287, 530-94421, FUJIFILM).

For the purification of Fc-MBL-OX40L, Fc-scOX40L, and scOX40L-Fc, the culture supernatant containing the respective recombinant OX40L proteins was applied to a HiTrap™ rProtein A FF column (17507901, Cytiva, Tokyo, Japan). After rinsing the column with phosphate-buffered saline (PBS), the bound Fc protein was eluted with elution buffer (80 mM glycine-HCl, 0.8 M arginine, pH 3.0), and the protein fractions were immediately neutralized with 1 M Tris-HCl (pH 9.0). MBL-scOX40L, possessing a His-tag, was purified using a HisTrap™ excel column (17371205, Cytiva). The column was washed with wash buffer (20 mM sodium phosphate, 0.5 M NaCl, 30 mM imidazole, pH 7.4) and eluted with elution buffer (20 mM sodium phosphate, 0.5 M NaCl, 500 mM imidazole, 0.8 M L-arginine, pH 7.4). MBL-OX40L and MBL-scOX40L containing a PA tag were absorbed onto Anti PA tag Antibody Beads (012-25841, FUJIFILM). The beads were washed with PBS and eluted with elution buffer (0.1 M glycine-HCl, 1 M arginine, 2 M MgCl₂·6H₂O, pH 3.0). The eluate was dialyzed against PBS, concentrated using an Amicon Ultra-4 (UFC801008, Merck MilliporeSigma, Burlington, Massachusetts, U.S.A.), and filtered with Millex-GP (0.22 µm, SLGPR33RS, Merck). The total protein concentration was determined using a bicinchoninic acid (BCA) assay (297-73101, FUJIFILM).

To detect the binding between OX40 and OX40L, purified OX40L proteins were added to ELISA plates (439454, Thermo Fisher, Waltham, MA, U.S.A.) pre-coated with 0.4 µg/mL of OX40-Fc (23, 26). OX40L protein was detected with anti-PA tag IgG (NZ-1, 012-25863, FUJIFILM) and peroxidase-conjugated AffiniPure Goat Anti-Rat IgG (H+L) (112-035-167, Jackson ImmunoResearch, West Grove, PA, U.S.A.).

The following reagents were used for the measurement of cytokines from T cells: anti-IL-2 (JES6-1A12, 503701, BioLegend, San Diego, CA, U.S.A.) capture antibody, biotin-anti-IL-2 (JES6-5H4, 503803, BioLegend) detection antibody, anti-IFN-γ (R4-6A2, 505702, BioLegend) capture antibody, biotin-anti-IFN-γ (XMG1.2, 505804, BioLegend) detection antibody, anti-IL-4 (11B11, 504101, BioLegend) capture antibody, biotin-anti-IL-4 (BVD6-24G2, 504201, BioLegend) detection antibody, anti-IL-17A (TC11-18H10.1, 506901, BioLegend) capture antibody, biotin-anti-IL-17A (TC11-8H4, 507001, BioLegend) detection antibody, HRP streptavidin (405210, BioLegend), and 3,3’,5,5’-tetramethylbenzidine (TMB, 421101, BioLegend). The absorbance was measured at 450 nm or 620 nm using a FilterMax F5 (Molecular Devices, San Jose, CA, U.S.A.).

An OX40⁺ T cell hybridoma was previously established and used to evaluate OX40 activity in T cells (23–25, 27, 28). To test the binding activity of OX40L proteins to cell-associated OX40 proteins, the T cell hybridoma was first incubated with the respective OX40L proteins, followed by incubation with anti-PA tag antibody, and then with goat anti-rat IgG2a-FITC (MRG2a-83, 407506, BioLegend).

To perform the activation-induced marker (AIM) assay, a valuable method to detect antigen-specific CD4⁺ T cells (29, 30), draining lymph node (dLN) cells were initially incubated with IgG from rat serum (14131, Merck) to block nonspecific binding. The following antibodies were used for cell-surface staining: PE/Cyanine7 anti-mouse CD4 (GK1.5, 100421, BioLegend), Alexa Fluor^® 488 anti-mouse CD25 (PC61, 102018, BioLegend), and APC anti-mouse CD134 (OX-40) (OX-86, 119413, BioLegend). For Foxp3 staining, cells were fixed and permeabilized with the eBioscience™ Foxp3/Transcription Factor Staining Buffer Set (00-5523-00, Thermo Fisher), then stained with PE Anti-Mo/Ra FOXP3, eBioscience™ (FJK-16s, 12-5773-08, Thermo Fisher). Data were acquired on a FACS Canto II or a Celesta (BD Biosciences, Franklin Lakes, NJ, U.S.A.) and analyzed with BD FACS Diva software version 9 or Flowing software version 2 (Turku Bioscience, Turku, Finland, http://flowingsoftware.btk.fi/).

Recombinant OX40L proteins were mixed with 4× Laemmli sample buffer (240 mM Tris-HCl, pH 6.8, 40% glycerol, 8% sodium dodecyl sulfate (SDS), 0.1% bromophenol blue) with or without 4% 2-ME and incubated at 60-100°C for 1 to 5 minutes. Reducing or non-reducing samples were separated by SDS-polyacrylamide gel electrophoresis (PAGE), transferred onto polyvinylidene difluoride (PVDF) membranes (034-25663, FUJIFILM), and analyzed by immunoblotting with anti-PA tag antibody and peroxidase-conjugated AffiniPure Goat Anti-Rat IgG (H+L).

To detect the expression levels of IκBα and β-actin proteins, the OX40⁺ T cell hybridoma was lysed for 30 minutes in ice-cold radioimmunoprecipitation assay (RIPA) buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM EDTA, 1% Nonidet P-40 (NP-40), 50 mM NaF, 1 mM Na₃VO₄, 1% sodium deoxycholate, and 0.1% SDS) with a protease inhibitor mixture (P8340, Merck). Insoluble debris was removed by centrifugation at 15,000×g for 10 minutes. The clear lysate was reduced with 4× Laemmli sample buffer for 5 minitues at 100°C. Samples were separated by SDS-PAGE, transferred onto PVDF membranes, and analyzed by immunoblotting with anti-IκBα (L35A5, 4814, Cell Signaling Technology, Danvers, MA, U.S.A.), anti-β-actin (C4, MAB1501, Merck) or anti-β-tubulin (TUB2.1, T4026, Merk), followed by goat anti-mouse IgG (H+L chain) pAb-HRP (330, MEDICAL & BIOLOGICAL LABORATORIES CO., LTD., Tokyo, Japan). The reaction was visualized with a chemiluminescence detection system using Immobilon Classico Western HRP substrate (WBLUC0500, Merck) and LAS-4000 mini (FUJIFILM).

C57BL/6 mice (Japan SLC, Shizuoka, Japan) were bred under specific pathogen-free conditions. Animal experimental protocols were approved by the Animal Care and Use Committee of the University of Toyama (Approved numbers: A2024PHA-02 and A2024PHA-03) and conducted in accordance with the Institutional Animal Experiment Handling Rules of the University of Toyama. CD4⁺ and CD8⁺ T cells were purified from the spleens of C57BL/6 mice using CD4 (L3T4) microbeads (130-117-043, Miltenyi Biotec, Bergish Gladbach, Germany) and CD8 (Ly-2) microbeads (130-117-044), respectively. T cells were cultured in RPMI medium with 10% FCS, 2 mM L-Alanyl-L-Glutamine, 100 U/mL penicillin, 100 µg/mL streptomycin, and 50 µM 2-ME. T cells were plated in 200 µL volumes in 96-well flat-bottom culture plates precoated with 10 µg/mL of anti-CD3ε antibody (low endotoxin, azide-free, 145-2C11, 100340, Biolegend) in triplicate with increasing concentrations of respective OX40L proteins. Culture supernatants were harvested at 72 hours for cytokine ELISA.

Groups of C57BL/6 mice from both male and female mice aged between 5 and 15 weeks were immunized subcutaneously at the footpad with 100 µg of 2,4,6-trinitrophenyl-conjugated keyhole limpet hemocyanin (TNP-KLH) emulsified in complete Freund’s adjuvant (CFA) (F5881, Merck) on day 0. Recombinant proteins in PBS were injected intraperitoneally on days 1 and 3. On day 7, pooled dLN cells were plated in 200 µL volumes in 96-well plates in triplicate in the presence or absence of KLH. Cells were harvested at 18 hours for the AIM assay. Culture supernatants were harvested at 72 hours for cytokine ELISA. Cell viability on day 3 was assessed with a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium (MTT) (M2128, Merck) assay.

Statistical significance was evaluated using either the Tukey-Kramer or Dunnett’s multiple comparison test in JMP Pro version 17.0.0 (SAS Institute, Cary, NC, U.S.A.). A p-value of less than 0.05 was considered statistically significant.

Highly clustered trimeric OX40L molecules, expressed by antigen-presenting cells, exhibit potent agonistic functions and are likely the natural source of costimulatory activity for OX40 on T cells during physiologically relevant immune responses. Identifying a substitute for naturally expressed OX40L on the cell surface remains a significant challenge. We hypothesized that highly multimerized OX40L trimers could serve as potent agonists for OX40. To investigate how the structure of the OX40L protein influences its agonistic activity toward OX40, we prepared five different soluble OX40L-fusion proteins, as illustrated in Figure 1 . The DNA nucleotide sequences corresponding to these OX40L proteins are provided in Supplementary Figure S1 . The structure of MBL-OX40L protein consists of a PA peptide tag (PA), a collagen-like domain derived from mannose-binding lectin (MBL), and the THD of OX40L (OX40L), from the N-terminus to the C-terminus ( Figure 1A ) (23). The attachment of the Fc domain of IgG1 (Fc) to MBL induces the formation of oligomerized trimeric structures of CD40L, which efficiently stimulates CD40 on B cells (31). Consequently, we prepared Fc-MBL-OX40L, comprising Fc, PA, MBL, and OX40L ( Figure 1B ). The structure of Fc-scOX40L is composed of Fc, PA, OX40L, a GGGSGGG peptide linker (Linker), OX40L, Linker, OX40L, and a histidine peptide tag (His₆) ( Figure 1C ) (24). The structure of scOX40L-Fc contains OX40L, Linker, OX40L, Linker, OX40L, Fc, PA, and His₆ ( Figure 1D ). The structure of MBL-scOX40L includes PA, MBL, OX40, Linker, OX40L, Linker, OX40L, and His₆ ( Figure 1E ). The structural moieties PA, His, and Fc were employed for the detection and purification of the protein. The moieties MBL and Fc were utilized to facilitate the formation of a trimer and a dimer, respectively. The tandem repeats of three OX40Ls, scOX40L, were designed to organize the formation of an active OX40L trimer.

MBL-OX40L primarily displayed dimer, trimer, and hexamer structures based on molecular weight as observed in immunoblotting and SDS-PAGE, suggesting the intrinsic role of the collagenous trimerization domain of mannose-binding lectin in forming multimerized structures (32–34). In contrast, it appeared as a monomer after boiling and reduction in Laemmli’s buffer ( Figure 2A ; Supplementary Figure S2A ) (23). Fc-MBL-OX40L showed dimer, hexamer, and higher-order multimer structures ( Figure 2B ; Supplementary Figure S2B ). Fc-scOX40L and scOX40L-Fc exhibited a dimer structure ( Figures 2C, D ; Supplementary Figures S2C, D ). MBL-scOX40L formed dimer, trimer, and multimeric structures ( Figure 2E ; Supplementary Figure S2E ). These electrophoretic analyses suggest that MBL-OX40L, Fc-MBL-OX40L, and MBL-scOX40L containing MBL form multimeric conformations, whereas Fc-scOX40L and scOX40L-Fc form dimeric structures.

Next, to confirm the binding activity of these OX40L proteins to OX40, each OX40L protein was serially diluted and added to the wells of a microtiter ELISA plate pre-adsorbed with OX40-Fc. Our previous studies demonstrated that MBL-OX40L and Fc-scOX40L specifically interact with OX40 (23, 24). Consistent with previous results, all five OX40L proteins and the control anti-OX40 monoclonal antibody (OX86) bound to OX40-Fc in a dose-dependent manner ( Figures 3A-F ) and exhibited different binding activities ( Figure 3G ). The OX40L moieties of these proteins did not cross-react with 4-1BB, CD27, or GITR (23, 24). Additionally, these five OX40L proteins interacted differentially with the cell surface OX40 expressed by the T cell hybridoma ( Figures 4A-G ; Supplementary Figures S3A-E ). These results demonstrate that the OX40L-fusion proteins used in this study specifically bind to OX40.

OX40 activates the classical NF-κB pathway (27, 35). To evaluate the functional activity of OX40L-fusion proteins in terms of NF-κB activation, the T cell hybridoma used in the experiments shown in Figure 4 ; Supplementary Figure S3 was stimulated with the respective OX40L-fusion proteins. As demonstrated in Figure 5 ; Supplementary Figure S4 , all five OX40L-fusion proteins induced the degradation of IκBα, a hallmark of classical NF-κB pathway activation. The anti-OX40 agonistic antibody (OX86) did not efficiently promote the degradation of IκBα ( Supplementary Figure S5 ), indicating that OX40L proteins possess an advantage in activating the NF-κB pathway.

Signals through OX40 promote TCR/CD3 signals and enhance cytokine production from T cells. To evaluate the costimulatory activity of OX40L-fusion proteins, murine splenic CD4⁺ and CD8⁺ T cells were cultured with plate-immobilized anti-CD3 agonistic antibody and increasing concentrations of the respective soluble OX40L-fusion proteins. Since OX40 is highly upregulated on activated T cells, microwells were coated with a relatively higher concentration of anti-CD3 antibody (10 µg/mL). Lower concentrations of anti-CD3 did not exhibit productive cytokine responses (24). All five OX40L-fusion proteins and the control anti-OX40 agonistic antibody (OX86) significantly increased the production of IL-2 and IFN-γ from CD4⁺ and CD8⁺ T cells ( Figures 6A-F ). Collectively, these results demonstrate that all OX40L-fusion proteins have significant costimulatory activity on T cells in vitro.

To demonstrate the agonistic function of OX40L-fusion proteins in vivo, mice were immunized with TNP-KLH and CFA on day 0, and then injected with Fc-MBL-OX40L (B), Fc-scOX40L (C), or scOX40L-Fc (D) on days 1 and 3 to analyze antigen-specific CD4⁺ T cell responses. We selected these OX40L-fusion proteins because the IgG1 Fc domain moiety may enhance the overall in vivo stability of the OX40L-fusion proteins, thereby providing potent agonistic activity. On day 7, dLN cells were cultured with or without KLH. Anti-OX40 agonistic antibody (OX86) and phosphate-buffered saline (PBS) were administered as positive and negative controls, respectively.

Interestingly, Fc-scOX40L (C) elicited stronger antigen-specific T cell responses than Fc-MBL-OX40L (B) and scOX40L-Fc (D), as assessed by antigen recall responses of dLN cells, in terms of IFN-γ, IL-17, and IL-4 production and proliferation ( Figure 7 ; Supplementary Figure S6 ). Both scOX40L-Fc (D) and anti-OX40 agonistic antibody (OX86) also triggered significantly higher T cell responses compared to the None (PBS) control group, although these responses were comparatively weaker ( Figure 7 ). Unexpectedly, Fc-MBL-OX40L (B) failed to effectively stimulate in vivo T cell responses, with levels comparable to those observed in the None (PBS) control group ( Figure 7 ; Supplementary Figure S6 ). These findings indicate that oligomerization of OX40L with MBL and IgG1 Fc domains, as well as its stabilization through fusion with IgG1 Fc domain, may not be sufficient to produce a more effective OX40 agonist.

Detection of antigen-responding CD4⁺ T cells is crucial for understanding the outcome of T-cell immunity regulated by the OX40-OX40L system. The AIM assay is used to identify antigen-specific T cells based on the upregulation of activation markers after antigen recall (29, 30). We analyzed Foxp3^-CD4⁺ and Foxp3⁺CD4⁺ populations in dLN cells from TNP-KLH-immunized mice in terms of the upregulation of the activation markers OX40 and CD25 following stimulation with KLH for 18 hours in vitro ( Figure 8 ). We found that Fc-scOX40L (C) showed a significantly increased frequency of OX40⁺CD25⁺Foxp3^-CD4⁺ cells compared to the control (PBS) or the anti-OX40 agonistic antibody (OX86) ( Figure 8 ; Supplementary Figure S6 ). Furthermore, Fc-scOX40L (C) exhibited a significantly enhanced frequency of OX40⁺CD25⁺Foxp3⁺CD4⁺ cells ( Figure 8 ; Supplementary Figure S6 ). OX40L proteins other than Fc-scOX40L (C) also induced the upregulation of AIM markers, but the levels were marginal (data not shown). These results indicate that an OX40L-fusion protein with an optimized structure may work as a potent agonist for OX40 and that one of the OX40L proteins, Fc-scOX40L (C), allows the expansion of antigen-specific CD4⁺ T cells in the primary phase of the immune response.

The results demonstrated in Figures 7 , 8 ; Supplementary Figure S6 indicate that a molecule with a structure like Fc-scOX40L (C) would function as a better agonist. To test this hypothesis, we performed similar experiments using molecules in which the OX40L part of Fc-scOX40L (C) was replaced with other TNF ligands (TNFLs), i.e., Fc-sc4-1BBL, Fc-scCD70, and Fc-scGITRL, as we previously reported (24). The TNFL protein moiety was required to induce both proliferative and cytokine responses, and the Fc-only protein could not induce effective T cell responses ( Supplementary Figure S7A ). These Fc-scTNFL proteins significantly enhanced the number of cell divisions in the presence of an anti-CD3 antibody, as determined by the dilution of carboxyfluorescein diacetate succinimidyl ester (CFSE) ( Supplementary Figure S7B ). Mice were immunized with TNP-KLH and CFA, followed by the administration of the respective Fc-scTNFL proteins. Draining lymph node cells from TNP-KLH-immunized mice injected with the respective Fc-scTNFL proteins exhibited higher recall cytokine responses than those treated with the control (PBS) ( Figure 9 ). Characteristically, Fc-scOX40L promoted the induction of IL-4-producing cells ( Figures 7 , 9 ). In the AIM assay, upon antigen recall, all Fc-scTNFL groups induced the upregulation of the activation markers OX40 and CD25 in both Foxp3^-CD4⁺ and Foxp3⁺CD4⁺ populations ( Figure 10 ). These results suggest that an Fc molecule with scTNFL at the C-terminus can efficiently agonize antigen-primed CD4⁺ T cells in vivo.

In summary, our results demonstrate that a potent agonist, capable of enhancing T cell responses in vivo, can be developed by incorporating an active single-chain TNF trimer structure into a suitable protein scaffold, such as the Fc domain of an IgG.