Dual targeting OX40 and IL-2 receptor enhances antitumor activity through tumor-infiltrating Treg depletion and CD8+ T-cell proliferation

Abstract

Background

The strong regulatory T cell (Treg) inhibitory activity and dysfunctional cytotoxic T lymphocytes (CTLs) represent major barriers to effective antitumor immunity, particularly in late-stage cancer. Multiple anti-OX40 (aOX40) agonistic antibodies have been developed but exhibit limited antitumor efficacy. Interleukin-2 (IL-2) effectivity expands CTLs but has severe side effects.

Methods

We construct an aOX40-mIL2-Fc bispecific antibody through Fab physical blocking and attenuated IL-2 with Rβ reducing N88D mutation. We also produced aOX40-Fc and IL-2/aOX40-Fc as a comparison using the 293F expression system. Single-cell and flow cytometry were used to analyze the change of T-cell subsets in the tumor microenvironment (TME). Mouse tumor models were used to assess the antitumor efficacy of aOX40-mIL2-Fc by tumor growth and survival, and toxicity by body weight loss, inflammatory cytokine production, and natural killer (NK) cell proliferation in the blood. The tumor-bearing mice were randomly assigned, and the average size was similar among various groups.

Results

aOX40-mIL2-Fc bispecific antibody-cytokine exhibited a synergistic therapeutic effect with limited toxicity, outperforming IL-2-Fc or aOX40 alone treatment, and conferring resistance to tumor rechallenge. On cellular mechanisms, aOX40-mIL2-Fc treatment showed great Treg depletion and increased both stem-like and effector functional terminal CD8⁺ T cells in the TME, while avoiding NK cells expansion in the periphery. Furthermore, this bispecific antibody remarkably improved the anti-programmed death-ligand 1 (PD-L1) therapeutic effect.

Conclusions

Our study unveils a novel approach to IL-2 design that addresses several critical shortcomings of existing strategies and elucidates the cellular mechanisms underlying aOX40-mIL2-Fc therapy. Meanwhile, combining aOX40-mIL2-Fc with PD-L1 blockade represents a strategic approach to enhance tumor control and overcome resistance to immune checkpoint blockade therapies synergistically.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Anti-OX40 (aOX40) antibody depletes regulatory T cell (Treg) among various mice tumor models, but insufficiently expands CD8⁺ T cells. Interleukin (IL)-2 increases cytotoxic T lymphocyte expansion but also enhances Treg quantity and has severe toxicity at high doses.

WHAT THIS STUDY ADDS

• Due to the discrepancy in expression levels of OX40 receptors on tumorous T-cell subsets, we designed a novel aOX40-mIL2-Fc bispecific antibody that effectively achieves Treg depletion inside the tumor microenvironment and tumor-infiltrating lymphocytes rejuvenation for optimized tumor control, while avoiding IL-2-induced side effects in the periphery.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our finding provides a novel strategy for IL-2-based immunotherapy, addressing the problems that IL-2 induces Treg proliferation and toxicity. Our data highlight the potential of aOX40-mIL2-Fc as a new therapeutic strategy to combine with immune checkpoint inhibitors, chemotherapy, radiotherapy, or other immunotherapies.

Introduction

Following T-cell receptor stimulation, the tumor necrosis factor receptor family member OX40 is transiently expressed on CD4+ T and CD8+ T cells, playing critical roles in cell activation, effector function, and long-term survival.¹ OX40 ligand (OX40L), expressed on activated antigen-presenting cells, facilitates signaling transduction to activate antigen-specific T cells.² However, Tregs greatly upregulate OX40 expression inside the tumor microenvironment (TME).^{3 4} Despite several being under clinical investigation, no anti-OX40 antibodies have been approved, most of which do not achieve clear antitumor effects.^{5 6} The Fc region of antibodies plays a critical role in engaging immune effector mechanisms. Both human IgG1 (human immunoglobulin G subclass 1) and mouse IgG2a (mouse immunoglobulin G subclass 2a) exhibit a preferential affinity for activating Fcγ receptors, thereby promoting the clearance of target cells via antibody-dependent cellular cytotoxicity and antibody-dependent cellular phagocytosis. Recent studies have shown that the hIgG1 backbone of aOX40 antibodies is essential for effectively depleting Treg cells within the TME through Fc-Fcγ receptor interactions.^{7 8} However, despite effective regulatory T cell (Treg) depletion, these antibodies have demonstrated limited capacity to enhance the expansion of tumor-infiltrating lymphocytes (TILs), which restricts their overall therapeutic efficacy.^{9 10} We hypothesize that Treg depletion alone is insufficient to rejuvenate TILs. Therefore, the addition of exogenous T cell-stimulatory cytokines may be necessary to expand and activate TILs for effective tumor control.

Interleukin-2 (IL-2), a pleiotropic cytokine that can proliferate and activate cytotoxic T cells or natural killer (NK) cells, has been approved for treating metastatic melanoma and renal cell cancer.¹¹ IL-2 preferentially binds to the high-affinity IL-2Rαβγ receptor complex, then peripheral or intratumoral Treg expressed trimer IL-2 receptor complex could absorb a large proportion of IL-2, potentially impairing the therapeutic efficacy.^{12 13} High-dose IL-2 expands CD8⁺ T and NK cells for tumor control but causes severe peripheral toxicity.^{14 15} To address these limitations, no-alpha IL-2 muteins and anti-programmed cell death protein-1 (PD-1) fusion proteins have been developed to selectively target and activate PD-1⁺ tumor-specific T cells, reducing CD25 binding to prevent Treg activation.16,18 Despite these advances, such approaches may still induce peripheral NK cell expansion and associated toxicity. Other IL-2 muteins, designed to reduce binding to CD122 or CD132, demonstrate significantly lower peripheral toxicity by limiting NK activation and proliferation.19,21 However, these CD25-biased IL-2 muteins may still bind Tregs and could weaken antitumor activity.^{22 23} To improve IL-2-based cancer therapy, the innovative design of IL-2 needs to have a distinct binding affinity to these target cells, aiming to expand tumor-specific CD8⁺ T cells and decrease Tregs within the TME while preventing peripheral NK and endothelial cell activation.

We found that IL-2 signaling is essential during anti-OX40 (aOX40) antibody-mediated tumor control. To achieve efficient Treg depletion while enhancing TIL functionality within the TME, we engineered a novel anti-OX40 bispecific fusion protein with attenuated IL-2, using a combination of Fab physical blocking and the Rβ binding reducing IL-2 N88D mutation to alter the IL-2’s binding pattern. This design is intended to minimize NK cell proliferation in the peripheral circulation, thereby decreasing cytokine consumption and systemic toxicity, while effectively depleting Treg within the TME and rejuvenating TIL for enhanced tumor control.

Methods

Mice

BALB/c and C57BL/6 mice, aged 6–8 weeks, were purchased from Vital River Laboratories (Beijing, China). C57BL Rag1 KO mice were purchased from the model animal research center of Nanjing University. All mice were maintained under specific pathogen-free conditions in the Institute of Biophysics and Tsinghua University animal facilities. All studies were approved by the Animal Care and Use Committee of the Institute of Biophysics and Tsinghua University.

Cell lines and reagents

Invitrogen FreeStyle 293-F cells (R79007) were cultured in SMM 293-TII medium (M293TII, Sino Biological) or EX-CELL 293 serum-free medium (14571C, Sigma-Aldrich). MC38 and CT26 cell lines were purchased from the American Type Culture Collection. MC38-OVA cells were selected from single-cell clones transduced with a lentivirus expressing OVA. Both MC38 and MC38-OVA cells were cultured and maintained in Dulbecco’s Modified Eagle’s Medium supplemented with 10% heat-inactivated fetal bovine serum, 2 mM l-glutamine, 0.1 mM minimum essential medium non-essential amino acids, 100 U/mL penicillin, and 100 mg/mL streptomycin. Purified mouse splenocytes and CTLL2 cells were maintained in complete Roswell Park Memorial Institute (RPMI) 1640 medium, supplemented with 10% heat-inactivated fetal bovine serum. All cell lines were cultured at 37°C in a 5% CO₂ atmosphere and routinely tested for Mycoplasma contamination using a Mycoplasma detection kit (R&D Systems). Antibodies, including anti-CD8 antibody (TIB210), FcγRII/III blocking antibody (2.4G2), anti-NK1.1 antibody (PK136), anti-interferon (IFN)-γ antibody (XMG1.2), anti-IL-2 antibody (JES6-1A12), and anti-programmed death-ligand 1 (PD-L1) antibody (10F. 9G2), were purchased from Bio X Cell (USA). FTY720 was purchased from Sigma-Aldrich.

Production of bispecific and monoclonal antibodies

Using the heterodimeric Fc variant KiHss-AkKh technology, the wild-type human IL-2 or IL-2 muteins were fused with the knob variant Fc region, while the aOX40 Fab was fused with the hole variant Fc region. The aOX40-mIL2-Fc antibody was generated by transient co-transfection of three plasmid constructs into FreeStyle 293-F cells. The supernatant containing bispecific antibodies was purified using protein A affinity chromatography according to the manufacturer’s protocol.

Tumor growth and treatment

A total of 5×10⁵ MC38, 7×10⁵ MC38OVA, and 4×10⁵ CT26 cells in 100 µL phosphate-buffered saline (PBS) were inoculated subcutaneously into the right dorsal flanks of 6–8 weeks old mice. Tumor-bearing mice were randomly assigned to treatment groups when tumors reached approximately 100–150 mm³. For the depletion of NK1.1 and CD8⁺ T cells, 400 µg or 200 µg antibodies were injected intraperitoneally 1 day before the initial treatment, followed by biweekly administration for 2 weeks. To neutralize IL-2, 200 µg of mouse IL-2 blocking antibody was administered on days 12, 15, and 18. FTY720 exhibits agonistic activity at sphingosine 1-phosphate (S1P) receptors, thereby inhibiting S1P/S1P1-dependent lymphocyte egress from secondary lymphoid tissues and the thymus. FTY720 was intraperitoneally administered at a dose of 25 µg 1 day before treatment initiation, followed by 10 µg every other day for 2 weeks. Anti-IFN-γ was intraperitoneally injected (500 µg per mouse) 1 day before aOX40-mIL2-Fc treatment. For anti-PD-L1 and anti-cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) treatment, 200 µg of anti-PD-L1 or 100 µg of anti-CTLA-4 antibody were administered intraperitoneally. Tumor volumes were measured using the formula: volume=(length×width×height)/2.

Flow cytometry analysis

To evaluate the infiltration of immune cells and the ratio of CD8⁺ T/Treg, tumor-bearing mice were sacrificed and tumors were harvested 2 days after a single or second treatment. Single-cell suspensions from the spleen, tumor, or in vitro co-cultured cells were incubated with anti-FcγIII/II receptor (clone 2.4G2) for 15 min at 4°C to block non-specific binding before staining with the conjugated antibodies. LIVE/DEAD fixable yellow dye (Thermo Fisher Scientific) was employed to exclude dead cells. For intracellular staining of IFN-γ and Foxp3, samples were fixed, permeabilized, and subsequently stained with anti-mouse IFN-γ or anti-mouse Foxp3 antibodies. All staining procedures were conducted in the dark at 4°C. Cytokine levels in the supernatants from mice serum were measured using the BD CBA Mouse Th1/Th2/Th17 Kit. Data were collected on FACS Fortessa flow cytometer (BD) and analyzed using FlowJo (TreeStar) software.

Ex vivo binding assay

To evaluate the binding affinity of fusion protein, single-cell suspension from C57BL6 spleen (1×10e6) or MC38 tumor was incubated with anti-CD16/32 (anti-FcγIII/II receptor, clone 2.4G2) for 30 min at 4°C. IL-2, aOX40 or various forms of IL-2/aOX40 antibodies were added for 15 min on ice, followed by two washes and staining with anti-human IgG Fcγ-PE for 20 min on ice. After additional washes, samples were analyzed.

CTLL-2 proliferation assay

To assess IL-2 biological activity, various fusion proteins were co-cultured with CTLL-2 cells. Purified proteins, including IL-2/aOX40-Fc, aOX40-IL2-Fc, or aOX40-mIL2-Fc, were serially diluted in 100 µL medium and added to a 96-well plate. Each well received 3×10³ of CTLL-2 cells in 100 µL medium and was incubated for 72 hours at 37°C in a 5% CO₂. Following incubation, 20 µL of CCK-8 reagent (Cell Counting Kit-8) was added, and the plate was incubated for an additional 3–4 hours at 37°C in 5% CO₂. Absorbance was read at 450 nm.

CD8⁺ T-cell binding and expansion in vitro

Splenocytes from C57BL6 mice were cultured with aCD3 and aCD28 (1 µg/mL) and subsequently stimulated with or without IL-2. OX40 and CD25 receptor expression on CD8⁺ T cells was detected after IL-2 or aOX40-Fc stimulation. For binding affinity assessment, splenocytes were incubated with aOX40 or IL-2/aOX40-Fc antibodies for 15 min on ice, followed by two washes and staining with anti-human IgG Fcγ-PE for 20 min on ice, then analyzed for CD8⁺ T cells binding throughflow cytometry. For proliferation analysis, CD8⁺ T cells were sorted from the cultural splenocytes using a negative CD8⁺ T Isolation Kit (STEMCELL Technologies) following the manufacturer’s instructions, and co-cultured with aOX40 or IL-2/aOX40 fusion proteins for 72 hours, followed by evaluation with the CCK-8 assay.

Toxicity

C57BL/6 mice bearing MC38 tumors were treated with varying doses of fusion proteins, with body weight monitored daily. Serum samples were collected 12 hours after the second antibody injection for cytokine analysis, measuring serum IFN-γ levels via CBA. Peripheral blood was collected 48 hours after the second injection to assess NK cell numbers using flow cytometry.

Single-cell RNA sequencing and analysis

C57bL/6 mice (n=4 per group) were inoculated with MC38 cells and treated with PBS or 15 µg of aOX40-mIL2-Fc by intraperitoneal injection on days 13 and 16. The tumor tissues were collected and digested on day 19. CD45⁺ cells were sorted throughflow cytometry for single-cell sequencing. Single-cell RNA sequencing (scRNA-seq) libraries were constructed using the Chromium Controller system and supporting reagents (10x Genomics), following the manufacturer’s protocols. Sequencing was performed on an Illumina NovaSeq 6000 system (Annoroad Gene Technology, Beijing, China), with paired-end sequencing and a read length of 150 bp. After getting raw sequencing data, sample demultiplexing, barcode processing, read alignment, and unique molecular identifier (UMI) counting were performed using the 10x Cell Ranger analysis pipeline. Further quality control, feature selection, dimensionality reduction, unsupervised clustering, and differential expression analysis were carried out with the Seurat R package V.5.0.1.

The Cancer Genome Atlas database and KM plotter database analysis

The DiffExp module was used to analyze the differential expression of the OX40 gene between tumor and adjacent normal tissues across all The Cancer Genome Atlas (TCGA) tumors. Distributions of gene expression levels are displayed using box plots, and with statistical significance of differential expression is evaluated using the Wilcoxon test. The correlation module draws the expression scatterplots between OX40 and Foxp3 in a specific cancer type, together with the Spearman’s rho value and estimated statistical significance (TIMER; https://cistrome.shinyapps.io/timer/). The Kaplan Meier (KM) plotter website is an online survival analysis tool that can be used to evaluate the impact of individual genes on cancer prognosis. The Kaplan-Meier curve was generated by comparing patients with high versus low Foxp3 expression levels in advanced colorectal adenocarcinoma (KM plotter; https://www.kmplot.com/analysis/).

Statistical analysis

Data analyses were performed with GraphPad Prism statistical software (V.8.0) and presented as mean±SEM. P value was determined using a two-way analysis of variance for tumor growth, followed by multiple comparisons test as indicated in the figure captions. Log-rank tests for survival and unpaired two-tailed t-tests were used for other analyses. A p value of <0.05 was considered statistically significant.

Results

Anti-OX40 function depends on IL-2 signaling

We first investigate the expression of the OX40 across multiple human cancers using TCGA database. We found that OX40 expression was elevated in patients’ tumor tissue relative to adjacent normal tissue, suggesting a potential role in tumor progression (online supplemental figure 1A). OX40 expression correlated positively with Foxp3 in several epithelial tumors, indicating its preferential expression on tumor-infiltrating Treg cells (online supplemental figure 1B). Further analysis of single-cell RNA sequencing data from patients with clinical colorectal cancer confirmed that OX40 was predominantly expressed on Tregs, with minimal expression on other CD4⁺ or CD8⁺ T cells (online supplemental figure 1C,D).²⁴ Notably, higher Treg infiltration was associated with reduced patient survival (online supplemental figure 1E). These observations led us to hypothesize that aOX40 antibody may induce Treg depletion to relieve immunosuppression and enhance CD8⁺ T cell-mediated tumor control.

aOX40 antibodies have been explored in various preclinical mouse models but show limited therapeutic efficacy. To further enhance the therapeutic potential of aOX40, we investigated its underlying antitumor mechanisms using wild-type hIgG1 (aOX40-Fc) and Fc-silent mutant aOX40 (aOX40-mutFc) antibodies. In established MC38 colon cancer tumor-bearing mice, aOX40-Fc treatment demonstrated superior therapeutic effects over the Fc-silent aOX40-mutFc, highlighting the critical role of Fc-FcγR interaction mediated immune effects (figure 1A). Analysis of cell subsets in the TME revealed a reduction in Treg ratios and quantities following wildtype-Fc treatment but not the mutant-Fc, while CD8⁺ T-cell ratios remained largely unaffected in both groups (figure 1B,C). It has been reported that OX40-mediated differentiation of T cells to effector function requires IL-2 receptor signaling.25,27 We further found that blockade of IL-2 signaling with a neutralizing antibody impressively abrogated most of the antitumor effect of aOX40-Fc (figure 1D). These suggest that both Fc-dependent Treg depletion and IL-2-mediated activation of effector T cells are essential for OX40 antibody-mediated tumor control, and Treg depletion might allow more endogenous IL-2 to activate TILs.

Figure 1. IL-2 signaling is essential for aOX40-mediated therapeutic effect. (A) C57BL/6 mice (n=6 per group) were inoculated with 5×10⁵ MC38 cells and then treated with phosphate-buffered saline, 15 µg of aOX40 wild-type Fc or mutant Fc fusion protein by intraperitoneal injection on days 12, 15, and 18. Tumor size was measured every 3 days. (B and C) Mice were treated as in A. The tumor tissues were collected 2 days after the first treatment. The percentages and the absolute number of Treg cells among CD4⁺ T cells or CD8⁺ T cells were detected by flow cytometry. (D) Mice (n=5 per group) were injected subcutaneously with MC38 cells and intraperitoneally treated with 15 µg of aOX40 wild-type Fc or/and 200 µg of mouse IL-2 neutralization antibody on days 12, 15, and 18. Representative results from two independent experiments are shown as mean ±SEM. Two-way analysis of variance tests were applied to analyze the tumor growth data. Unpaired two-tailed t-tests were used to analyze the other data. ns (not significant), **p<0.01, and ****p<0.0001. aOX40, anti-OX40; IL, interleukin; Treg, regulatory T cell.

aOX40 fusion with IL-2 improves CD8⁺ T-cell proliferation in vitro

The insufficient CD8⁺ T-cell expansion during aOX40 treatment suggests that endogenous IL-2 is limited within tumor tissue. We hypothesized that replenishing exogenous IL-2 may enhance CD8⁺ T-cell proliferation and further increase aOX40’s antitumor efficacy. To explore this hypothesis, we first examined the expression profiles of CD25 and OX40 across immune subsets within the tumor. Both receptors were highly expressed on Treg cells, while CD8⁺ T cells expressed markedly lower levels (online supplemental figure 2A,B), suggesting that native IL-2 is preferentially sequestered by Tregs. To determine whether OX40 stimulation could enhance IL-2 responsiveness in CD8⁺ T cells, we co-cultured splenocytes with aOX40 antibody or IL-2. aOX40 engagement induced upregulation of CD25 on CD8⁺ T cells, and IL-2 stimulation reciprocally increased OX40 expression (figure 2A), indicating a positive feedback loop that could enhance dual receptor engagement. Based on these findings, we engineered an IL-2/anti-OX40-Fc (IL-2/aOX40-Fc) bispecific fusion protein (figure 2B). Binding affinity studies revealed that IL-2/aOX40-Fc exhibited superior affinity with activated CD8⁺ T cells compared with aOX40 alone (online supplemental figure 2C). Notably, co-culture with IL-2/aOX40-Fc robustly induced CD8⁺ T-cell expansion in vitro, whereas aOX40 monotherapy had minimal proliferative effect (online supplemental figure 2D). These results demonstrate that IL-2 is the primary driver of CD8⁺ T-cell proliferation and that its fusion to aOX40 can improve the functional engagement and expansion of effector T cells in vitro.

Figure 2. IL-2/aOX40 wild-type Fc depletes Treg cells while mutant Fc enhances CD8⁺ T proliferation. (A) Splenocytes were stimulated with aOX40 antibody or IL-2 for 72 hours, and the mean fluorescence intensity (MFI) of CD25 or OX40 on CD8⁺ T cells was analyzed using flow cytometry. (B) Schematic structure of IL-2/aOX40 bispecific antibody. IL-2/aOX40 comprises an Fab to OX40 and IL-2, fused to human IgG1(wild-type or mutant Fc). (C) Mice were subcutaneously inoculated with MC38 cells and treated with PBS, 25 µg of IL-2/aOX40 wildtype-Fc or mutant-Fc, or equal molar IL-2 or aOX40 by intraperitoneal injection. Tumor tissues were collected 2 days after treatment and percentages of Treg cells were detected. (D) MC38 tumor tissues were collected 2 days after the second treatment and an absolute quantity of CD8⁺ T cells was detected after 25 µg IL-2/aOX40-Fc treatment. For IL-2/aOX40-mutFc, MC38-OVA bearing mice treated with PBS, 25 µg of IL-2/aOX40-mutFc, or equal molar IL-2 and aOX40 mixture. (E) MC38 bearing mice (n=5 per group) were treated with PBS, 25 µg of IL-2/aOX40-Fc, or equal molar IL-2 and aOX40 mixture on days 14, 17, and 20. (F) Mice were inoculated with MC38 cells and treated with different doses of IL-2/aOX40-Fc by intraperitoneal injection on days 13, 16, and 19. Mouse body weight was measured daily. (G) C57BL/6 mice (n=5 per group) were inoculated with MC38-OVA cells and treated with PBS, 15 µg of IL-2/aOX40 wild-type Fc, or mutant Fc on days 15, 18, and 21. (H) MC38-OVA tumor-bearing mice (n=5 per group) were treated with PBS, 15 µg of IL-2/aOX40-mutFc on days 16, 19, and 22. 100 µg of anti-CTLA-4 was administered 1 day before each IL-2/aOX40-mutFc treatment. Data indicate means±SEM and are repeated two or three independent times. Statistical analysis for (A, C, and D) was performed using unpaired two-tailed t-tests. For tumor growth curves, a two-way analysis of variance was applied. ns (not significant), **p<0.01, ***p<0.001, and ****p<0.0001. aOX40, anti-OX40; CTLA-4, cytotoxic T-lymphocyte associated protein 4; IL, interleukin; PBS, phosphate-buffered saline; Treg, regulatory T cell.

IL-2/aOX40 cannot achieve Treg depletion and CD8⁺ T expansion concurrently

To assess the in vivo activity of the IL-2/aOX40-Fc bispecific fusion protein, we first evaluated its binding distribution among tumor-infiltrating T-cell subsets. On incubation of dissociated tumor tissue cells with graded concentrations of proteins, we observed that IL-2/aOX40-Fc had a greater binding affinity for tumorous Tregs relative to conventional CD4⁺ T or CD8⁺ T cells (online supplemental figure 3A). Notably, IL-2/aOX40-Fc exhibited the highest binding affinity to tumor-infiltrating Tregs compared with those from draining lymph nodes (dLNs) or spleen (online supplemental figure 3B). The IL-2/aOX40-Fc bispecific antibody even displayed higher Treg binding compared with the aOX40 antibody, suggesting this fusion protein might further increase Treg depletion efficiency (online supplemental figure 3C). In MC38 tumor-bearing mice, IL-2/aOX40-Fc treatment induced substantial Treg depletion and elevated CD8⁺ T/Treg ratio more effectively than IL-2, aOX40, or their combination (figure 2C and online supplemental figure 3D). Importantly, Treg frequencies in the dLNs and spleen remained unchanged, avoiding disruption to the peripheral immune equilibrium (online supplemental figure 3E). Unexpectedly, despite its enhanced Treg depletion, IL-2/aOX40-Fc failed to increase the total number of CD8⁺ T cells in the tumor (figure 2D). To explore the reason why IL-2/aOX40-Fc lost the capacity of CD8⁺ T proliferation in vivo, we generated an Fc-silent version of the fusion protein (IL-2/aOX40-mutFc). Compared with the wild-type Fc version, IL-2/aOX40-mutFc exhibited reduced Treg depletion but significantly enhanced CD8⁺ T-cell expansion in the tumor (figure 2C,D). These findings demonstrate a functional trade-off: the wild-type Fc enables Treg depletion but limits CD8⁺ T-cell expansion, whereas the Fc-silent format supports CD8⁺ T-cell proliferation but compromises Treg clearance.

To evaluate antitumor efficacy, we compared IL-2/aOX40-Fc to monotherapies and combination treatments. The bispecific fusion protein showed superior tumor control in MC38-bearing mice (figure 2E), but high-dose treatment induced acute body weight loss and early mortality, indicative of dose-dependent systemic toxicity (figure 2F). Despite compromised Treg depletion ability, IL-2/aOX40-mutFc showed comparable tumor inhibition as IL-2/aOX40-Fc, suggesting the potential antitumor ability of CD8⁺ T-cell expansion (figure 2G). To investigate whether achieving both Treg cell depletion and CD8⁺ T-cell expansion could further increase the antitumor effect, we performed combination therapy with anti-CTLA-4 and IL-2/aOX40-mutFc in MC38-OVA tumor-bearing mice and revealed that the combined treatment significantly improved therapeutic effects compared with a single treatment, indicating a synergistic function (figure 2H).

Finally, we investigated the immune cell populations required for therapeutic efficacy. Depletion of CD8⁺ T cells, but not NK cells, abrogated the therapeutic effects of IL-2/aOX40-Fc, underscoring the critical role of cytotoxic T cells (online supplemental figure 4A,B). Administration of FTY720 to block lymphocyte egress from lymphoid organs during IL-2/aOX40-Fc treatment did not impair antitumor efficacy, indicating that pre-existing intratumoral CD8⁺ T cells are sufficient to mediate tumor control (online supplemental figure 4C). Similarly, IL-2/aOX40-mutFc efficacy was also dependent on CD8⁺ T cells (online supplemental figure 4D). Even though the IL-2/aOX40-mutFc improved CD8⁺ T quantity in TME and exhibited superior therapeutic efficacy in the CT26 model, it also displayed the highest levels of serum IFN-γ, inducing potential toxicity (online supplemental figure 4E,F). Collectively, these findings indicate that both formats rely on intratumoral CD8⁺ T cells for efficacy, but achieving optimal therapeutic benefit requires a strategy that balances Treg depletion with effector T-cell expansion while minimizing systemic toxicity.

“2+1” form of IL-2/aOX40 with attenuated IL-2 reduces toxicity

Given the essential role of CD8⁺ T cells in mediating tumor control, we sought to design an IL-2/aOX40 fusion protein that could expand intratumoral CD8⁺ T cells while minimizing peripheral toxicity. To preserve the ability to deplete Tregs, we retained the wild-type Fc region, which maintains interaction with activating Fcγ receptors. To limit IL-2 activity outside the tumor, we inserted the IL-2 moiety into the hinge region of anti-OX40 heavy chain, generating both homodimeric and heterodimeric formats of aOX40-IL2-Fc (figure 3A and online supplemental figure 5A). This configuration was hypothesized to impose steric hindrance on IL-2, reducing its accessibility to the high-affinity IL-2 receptor (CD25/CD122/CD132) and thereby preventing peripheral toxicity and potential activated CD8⁺ T cells depletion. On evaluation in vivo, the homodimeric aOX40-IL2-Fc failed to deplete intratumoral Tregs or alter the Treg/CD4⁺ T-cell ratio (online supplemental figure 5B). The heterodimer maintained Treg depletion, and this “2+1” form of aOX40-IL2-Fc showed improved antitumor control compared with the original IL-2/aOX40-Fc protein across various tumor models (online supplemental figure 5C,D). To assess whether steric hindrance design reduced peripheral toxicity, we measured serum inflammatory cytokines following treatment. Mice receiving aOX40-IL2-Fc exhibited substantially lower circulating IFN-γ levels than those treated with IL-2/aOX40-Fc (figure 3B). However, despite improved safety at moderate doses, high-dose (>100 µg) aOX40-IL2-Fc still induced significant body weight loss, indicating that physical hindrance alone was insufficient to fully abrogate off-tumor effects (figure 3C).

Figure 3. Fab physical blocking and N88D mutation attenuate IL-2-induced toxicity. (A) Schematic structure of aOX40-IL2-Fc. An Fab of aOX40 was added to the N terminal of IL-2 compared with IL-2/aOX40-Fc. (B) Mice were treated with 15 µg of aOX40-IL2-Fc or equal molar IL-2/aOX40-Fc. Serum was collected 12 hours after the second injection and IFN-γ levels were measured by CBA. (C) C57BL/6 mice were inoculated with MC38 cells and treated with different doses of aOX40-IL2-Fc by intraperitoneal injection on days 13, 16, and 19. Mouse body weight was measured daily. (D) Schematic structure of aOX40-mIL2-Fc. Attenuated IL-2 containing N88D mutation physically blocks through Fab of aOX40. (E) CTLL-2 cells co-cultured with serial dilutions of IL-2/aOX40-Fc, aOX40-IL2-Fc, or aOX40-mIL2-Fc for 3 days, cell proliferation was detected by Cell Counting Kit-8. (F) MC38 tumor-bearing mice (n=6 per group) were treated with phosphate-buffered saline, 15 µg of aOX40-mIL2-Fc, or equal molar IL-2/aOX40-Fc on days 12 and 15. Periphery blood was collected at day 17 and natural killer cells were measured by flow cytometry. (G) MC38 tumor-bearing mice were treated with different concentrations of aOX40-mIL2-Fc on days 13, 16, and 19. Body weight changes were measured post-treatment. Data are presented as mean±SEM and are repeated two independent times. For tumor growth curves, a two-way analysis of variance was applied. Unpaired two-tailed t-tests were used to analyze the other data. **p<0.01, ****p<0.0001. aOX40, anti-OX40; IFN, interferon; IL, interleukin.

IL-2 could activate the receptor on peripheral NK and endothelial cells, which contribute to “on-target, off-tumor” toxicity. To further reduce peripheral toxicity, the N88D mutation (attenuating the CD122 binding) was introduced into “2+1” form of aOX40-IL2-Fc to construct aOX40-mIL2-Fc²⁸ (figure 3D). Using CTLL-2 cells to quantify IL-2 bioactivity, aOX40-IL2-Fc induced 10-fold less cell proliferation compared with wild-type IL-2, while aOX40-mIL2-Fc fusion protein showed a 1,000-fold reduction in IL-2 bioactivity (figure 3E). In vivo, treatment with IL-2/aOX40-Fc expanded circulating NK cell populations, whereas aOX40-mIL2-Fc did not, suggesting minimal peripheral IL-2 activity (figure 3F). Body weight was monitored throughout treatment, impressively, administrating up to 200 µg of aOX40-mutIL2-Fc into tumor-bearing mice did not significantly induce body-weight loss (figure 3G). All these data demonstrate that the “2+1” form of aOX40-mIL-2-Fc with attenuated IL-2 (N88D) variant and physical constraint effectively decreases IL-2-induced NK cell proliferation in the periphery, thereby reducing toxicity.

aOX40-mIL2-Fc remodels a distinct T-cell immunophenotypic signature in tumor

To explore the cellular mechanism following aOX40-mIL2-Fc therapy, we isolated intratumoral CD45⁺ cells and analyzed the transcriptional signatures at the single-cell level (figure 4A). As T-cell functions are critical for therapeutic effect, we separated the CD3⁺ T cells into four subsets and found that the proportion of CD8⁺ T cells was significantly enhanced following therapy (figure 4B,C). Meanwhile, the Treg cell population sharply decreased after the fusion protein treatment, and the cell subset ratios in CD3⁺ T were also changed (figure 4D). These data indicated that aOX40-mIL2-Fc reshaped the TME by depleting Tregs and expanding CD8⁺ T cells. We further analyzed the CD8⁺ T-cell phenotypes in different therapy groups, finding that the majority of the CD8⁺ T cells in tumors without therapy exhibited a terminal exhausted phenotype (figure 4E,F). In contrast, after aOX40-mIL2-Fc treatment, the CD8⁺ T cells not only contained the terminal phenotype but also included a stem-like subset with high expression of CXCR3 and Tcf7, increasing the stem-like/terminal phenotype ratio (figure 4G). Exhausted T cells were further separated into three clusters, the CD8⁺ T cells after fusion protein treatment enriched in cluster 0 and expressed genes encoding co-inhibitory cell surface receptors (PD-1 and T cell immunoglobulin and mucin domain containing protein 3 (Tim3)), while also achieving higher transcript levels of genes encoding granzyme B, perforin, and IFN-γ compared with the control group, indicating that aOX40-mIL2-Fc enhanced the tumor-killing function of terminal CD8⁺ T cells (figure 4H,I). It was reported that stem-like cells maintain polyfunctionality, persist long-term, and can differentiate into terminal TILs.²⁹ Thus, expanding the population of stem-like CD8⁺ T cells and enhancing effector molecule expression on exhausted CD8⁺ T post aOX40-mIL2-Fc therapy could improve the antitumor effect.

Figure 4. aOX40-mIL2-Fc altered T-cell ratio and increased cytotoxic activity of CD8⁺ T. (A) The experimental design scheme of scRNA-seq. (B) UMAP plot of 3,513 T cells showing four cell subsets (1,846 cells from the control group and 1,667 cells from aOX40-mIL2-Fc treated group). (C) Bubble plot showing signature genes of T-cell subsets. (D) Cell compositions and the ratio of T-cell subsets from control and aOX40-mIL2-Fc treated group. (E) UMAP plot of 1,868 CD8⁺ T cells showing two cell subsets (814 cells from the control group and 1,054 cells from the aOX40-mIL2-Fc treated group). (F) Bubble plot showing signature genes of CD8⁺ T-cell subsets. (G) Proportions of stem-like and exhausted T cells within CD8⁺ T cells from control and aOX40-mIL2-Fc treated group. (H) UMAP plot of CD8⁺_ex cells showing three subsets (720 cells from the control group and 526 cells from aOX40-mIL2-Fc treated group). (I) Feature plot exhibiting the distribution of expression levels of Pdcd1, Havcr2, Gzmb, Prf1, and Ifng in CD8_ex cells. aOX40, anti-OX40; FACS, Fluorescence-activated cell sorting; IL, interleukin; Treg, regulatory T cell; i.p.,intraperitoneal; s.c.,subcutaneous; UMAP, Uniform Manifold Approximation and Projection.

Synergy of intratumoral Tregs depletion and CD8⁺ T expansion improves antitumor function

After demonstrating that aOX40-mIL2-Fc diminished binding to the IL-2 receptor and reduced peripheral toxicity, we sought to determine its impact on tumor therapeutic efficacy. In mouse models of MC38 and CT26 tumors, aOX40-mIL2-Fc showed comparable therapeutic efficacy to aOX40-IL2-Fc, while exhibiting a better effect than the original IL-2/aOX40-Fc (figure 5A,B). The antitumor effect was abrogated in Rag1−/− mice, which lack adaptive immune cells, confirming the necessity of adaptive immunity for the therapeutic efficacy of aOX40-mIL2-Fc (figure 5C). We further explored the cellular and molecular mechanisms underlying the enhanced antitumor effects. Since the attenuated IL-2 reduced NK proliferation in the bloodstream, we wonder whether decreased binding affinity also influenced the Treg depletion efficiency. In vitro co-culture experiment with splenocytes revealed that although the N88D mutation reduced binding to the CD122 receptor, the overall binding avidity of aOX40-mIL2-Fc on Treg cells remained unchanged compared with IL-2/aOX40-Fc (figure 5D). Further analysis of in vivo mechanisms following fusion protein treatments proved that aOX40-mIL2-Fc effectively depleted Tregs in tumors 48 hours after the initial treatment (figure 5E). As cytotoxic T lymphocytes (CTLs) are essential for controlling tumor growth, we further examined the CD8⁺ T cells within the TME among different groups. While the original IL-2/aOX40-Fc fusion protein cannot expand CD8⁺ T cells in the TME, aOX40-mIL2-Fc treatment significantly increased the absolute number of total CD8⁺ T cells in MC38 tumor-bearing mice (figure 5F).

Figure 5. aOX40-mIL2-Fc mediated Tregs depletion and IFN-γ⁺CD8⁺ T expansion improve antitumor function. (A) C57BL/6 mice (n=6 per group) were inoculated with MC38 cells and treated with PBS, 15 µg of aOX40-mIL2-Fc, aOX40-IL-2-Fc or equal molar IL-2/aOX40-Fc by intraperitoneal injection on days 13, 16, and 19. (B) CT26 tumor-bearing mice (n=5 per group) treated with PBS, 21.5 µg of aOX40-mIL2-Fc or equal molar IL-2/aOX40-Fc and aOX40-IL2-Fc on days 9, 12, and 15. (C) Rag1−/− (n=5 per group) mice bearing MC38 tumors were treated with PBS or 15 µg of aOX40-mIL2-Fc at days 10, 13, and 16. (D) Binding avidity of bispecific proteins to Treg. Splenocytes were incubated with serial dilutions of fusion proteins and mean fluorescence intensity on Treg cells was detected by flow cytometry. (E–G) Mice (n=6 per group) were treated as in A. The tumor tissues were collected 2 days after the first treatment. The percentages and the absolute number of Treg cells among CD4⁺ T cells were detected (E). Flow cytometry analysis was performed with dissociated tumor samples for total CD8⁺ T cells and IFN-γ⁺CD8⁺ T cells on day 18 (F and G). (H) C57BL/6 mice bearing MC38 tumors (n=6 per group) were treated with PBS, or 15 µg of aOX40-mIL2-Fc on days 13, 16, and 19. To block IFN-γ, mice were intraperitoneally injected with 500 µg of neutralization antibody 1 day before aOX40-mIL2-Fc treatment. Data are presented as mean±SEM and are repeated two or three independent times. For tumor growth curves, a two-way analysis of variance was applied. Unpaired two-tailed t-tests were used to analyze the other data. ns (not significant), ***p<0.001, ****p<0.0001. aOX40, anti-OX40; IFN, interferon; IL, interleukin; MFI, mean fluorescence intensity; PBS, phosphate-buffered saline; Treg, regulatory T cell.

IFN-γ is one of the effector cytokines produced by tumor-specific CTLs and is associated with favorable clinical outcomes.^{30 31} We then assessed the effect of aOX40-mIL2-Fc on IFN-γ production in tumor tissue via flow cytometry. MC38-bearing mice treated with aOX40-mIL2-Fc induced a marked increase in the number of IFN-γ⁺CD8⁺ T cells compared with the control or IL-2/aOX40-Fc treatment (figure 5G). To confirm the role of IFN-γ in the antitumor activity of aOX40-mIL2-Fc, we administered an IFN-γ neutralizing antibody during therapy. This neutralization largely abolished the antitumor activity, indicating that IFN-γ was essential for fusion protein-induced therapeutic effects (figure 5H). Taken together, these findings highlighted a synergistic mechanism involving Treg depletion and CTL activation for enhancing antitumor function.

Recent studies showed many constructions to conjugate no-α IL-2 mutant (abolished CD25 binding) with antibody against TIL’s high expressing PD-1 molecule to specifically expand tumor-specific CD8⁺ T cells. Limited studies emphasize the necessity of CD25 expression on CD8⁺ T cells for effective antiviral or antitumor responses.^{22 32 33} To determine whether CD25 receptor binding is essential, we constructed aOX40-IL-2-Fc harboring F42A or F42A/N88D double mutations. Treatment on MC38 bearing mice revealed that these fusion proteins neither depleted Tregs nor altered Treg/CD4⁺ T ratio, leading to diminished antitumor efficacy compared with the aOX40-IL2-Fc, even though they increased the ratio of CD8⁺ T cells in TME (online supplemental figure 6A–C). F42A/N88D double mutations even induced Treg expansion in the spleen (online supplemental figure 6D,E). These data suggested that CD25 binding is essential for fusion protein-mediated Treg depletion and the F42A mutational IL-2 largely abolished therapeutic effects. Therefore, Treg depletion and CTL expansion are both crucial for the aOX40-mIL2-Fc antitumor effect.

To further analyze whether the combination of OX40 target and IL-2 achieves a unique therapeutic advantage over other known bispecific or cytokine-fusion strategies, we also constructed anti-programmed cell death protein-1 (aPD-1)-IL-2 bispecific antibodies, which have been extensively studied.^{16 17} The PD-1 molecule is highly expressed on intratumoral CD8⁺ T cells; therefore, aPD-1 antibody has been used to deliver IL-2 molecules biased to CD8⁺ T cells. As aPD-1-IL-2 primarily targets CD8⁺ T cells, we introduced a silent Fc to the fusion protein to avoid CD8+ T-cell depletion. Consistent with previous reports, aPD-1-mIL2-mutFc fusion protein with IL-2 (N88D) variant showed significantly reduced peripheral toxicity than that with wild-type IL-2 (online supplemental figure 7A,B).²² Similar as aOX40-mIL2-Fc, the aPD-1-mIL2-mutFc treatment increased the absolute quantity of intratumoral CD8⁺ T cells (online supplemental figure 7C). However, the aPD-1-mIL2-mutFc therapy did not cause a reduction of the intratumoral Treg/CD4⁺ T-cell ratio, while the absolute number of Treg cells was higher than the untreated group (online supplemental figure 7D). We also examined the proportion of Treg cells in the spleen and found that the Treg/CD4⁺ T-cell ratio had increased, indicating remarkable expansion of Treg cells in the periphery (online supplemental figure 7E). Indeed, the aOX40-mIL2-Fc antibody proved more effective in controlling tumors than aPD-1-mIL2-mutFc (online supplemental figure 7F). These results demonstrated that combining other targets, such as PD-1 with IL-2, can increase the quantity of CD8⁺ T cells, but it may not deplete the tumor-infiltrating Tregs. Thus, the aOX40-mIL2-Fc antibody shows the unique therapeutic advantages. The combination of the OX40 target and IL-2 achieves Treg depletion and CD8⁺ T-cell expansion concurrently, exhibiting a synergistic therapeutic effect (online supplemental figure 8).

aOX40-mIL2-Fc induces systemic immunity and overcomes immune checkpoint blockade resistance

Our previous results showed that activation of pre-existing tumor-infiltrating T cells is sufficient for tumor control and CD8⁺ T was expanded after aOX40-mutIL2-Fc treatment. To test whether increased CTLs from TME can have a systemic therapy impact, we employed a bilateral tumor model. An MC38-established tumor on the right flank was treated intratumorally, while a second tumor was inoculated on the left flank at the initiation of therapy. Following treatment with aOX40-mIL2-Fc, the local tumor was effectively eradicated compared with the control group, and notably, the growth of the distal tumors was also inhibited (figure 6A). This suggests that aOX40-mIL2-Fc-mediated activation of TILs is not only sufficient for local-tumor control but also for distal-tumor control. To examine immune memory induced by fusion protein treatment, mice that achieved tumor eradication after therapy were rechallenged by subcutaneous injection of MC38 cells on the left flank. Remarkably, none of these rechallenged mice developed tumors, whereas tumors progressively increased in the control mice (figure 6B). These findings suggest that mice achieved primary tumor clearance through intratumoral aOX40-mIL2-Fc therapy generated a systemic and protective memory immunity response for distal tumor control.

Figure 6. aOX40-mIL2-Fc eradicates distal tumor and overcomes anti-PD-L1 resistance. (A) C57BL/6 mice (n=7 or 12 per group) were inoculated with 5×10⁵ MC38 cells on the right flank of mice as in situ tumors and then inoculated with 1×10⁶ MC38 cells on the left flank on day 13 as distal tumors. Mice were intratumorally treated with 5 µg of aOX40-mIL2-Fc on days 13, 16, and 19 in situ tumors. Tumor growth was measured every 3 days. (B) Wild-type C57BL/6 mice or mice with tumor clearance by aOX40-mIL2-Fc ((n=5 or 8 per group) were injected subcutaneously with 3×10⁶ of MC38 on the left flank. (C) Tumor growth curves of C57BL/6 mice (n=6 per group) inoculated with MC38 cells and then treated with phosphate-buffered saline, 15 µg of aOX40-mIL2-Fc, 200 ug of anti-PD-L1, or both on days 17, 20, and 23. Tumor size was measured every 3 days, and the mouse survival curve was shown. Data are presented as mean±SEM and are repeated two independent times. For tumor growth curves, a two-way analysis of variance was applied. Statistical analysis for the survival curve data was performed using log-rank tests. ****p<0.0001. aOX40, anti-OX40; aPD-L1,anti-programmed death-ligand 1; IL, interleukin; PD-L1, programmed death-ligand 1.

Immune checkpoint blockades, such as anti-PD-1/PD-L1 have demonstrated therapeutic efficacy in clinics across a broad range of tumors, but only a minority of patients responded effectively. Multiple inhibitory mechanisms during TME may mediate the resistance to PD-1/PD-L1 blockade therapy, such as high infiltration of Tregs or insufficient proliferation of CD8⁺ T due to lack of IL-2 in advanced tumors.^{34 35} To test whether aOX40-mIL2-Fc can overcome anti-PD-L1 resistance, anti-PD-L1, and aOX40-mIL2-Fc were compared. Anti-PD-L1 therapy resulted in only partial tumor growth control, and the tumor relapsed in a few weeks. Impressively, the combination of aOX40-mIL2-Fc with anti-PD-L1 effectively inhibited tumor growth, prolonged survival, and led to tumor eradication completely (figure 6C). These results demonstrated that aOX40-mIL2-Fc can synergize with anti-PD-L1 checkpoint blockade for enhanced antitumor therapeutic efficacy.

Discussion

OX40 is transiently expressed on activated CD8⁺ T cells but continuously highly expressed in tumorous Tregs. Most antibodies targeting OX40 are based on human IgG1 or mouse IgG2a, which effectively deplete Tregs in TME.^{36 37} While human IgG1 antibodies have shown efficacy in murine tumor models, their therapeutic effects in clinical trials have been limited.^{5 38 39} Meanwhile, the proliferation and activation of CD8⁺ T cells within the TME strongly correlate with superior clinical outcomes across various cancers.^{40 41} Our findings demonstrate that Treg depletion alone is insufficient for aOX40 treatment and aOX40 function depends on IL-2 signaling, emphasizing the importance of IL-2 stimulation and the proliferation of CD8⁺ T cells. Meanwhile, IL-2 or aOX40 treatment induces upregulation of each other’s receptor on CD8⁺ T cells in vitro, which promoted us to construct IL-2 and aOX40 fusion protein. After carefully designing and efforts of screening, we finally constructed aOX40 antibody fused with an IL-2 Rβ attenuated IL-2 mutant (aOX40-mIL2-Fc), which shows several distinct characteristics: (1) aOX40 fused with IL-2 showed increased Treg binding ability in vitro and depleted intratumoral Treg cells more effectively than single aOX40 treatment in vivo. (2) aOX40-mIL2-Fc with attenuated IL-2 through N88D mutation and physical blocking showed much reduced peripheral NK proliferation and systemic toxicity. (3) Impressively, aOX40-mIL2-Fc expands tumor-infiltrating CD8⁺ T cells with high IFN-γ effector secretion, which contributes to improved tumor control. (4) aOX40-mIL2-Fc treatment induced systemic and durable protective memory immunity, and overcame PD-1/PD-L1 blockade therapy resistance.

IL-2 is a potent stimulator of cytotoxic T cells and has been approved for treating metastatic melanoma and renal carcinoma. High doses of IL-2 can activate CTLs but cause severe systemic side effects, while low doses preferentially bind to Treg cells in the TME, limiting their antitumor efficacy.⁴² Approaches like PEGylation, specific mutations, or antibody blocking can attenuate IL-2’s affinity for CD25 and decrease the Treg binding.^{43 44} Our design used a novel strategy by choosing OX40, which is highly expressed on Tregs in TME. aOX40-mIL2-Fc fusion protein achieves high binding avidity with Tregs and depletes intratumoral Treg cells effectively. Unexpectedly, we found that IL-2 molecular numbers on the aOX40-IL2-Fc antibody are critical for Treg depletion efficacy. Early researchers chose two IL-2 molecules fused to the C terminal of the antibody,^{45 46} while anti-PD-1-laIL-2 or anti-CD8-laIL-2 only contains one CD25 binding-abolished IL-2 molecule.^{17 47} In our comparative study of homodimer versus heterodimer aOX40-IL2-Fc constructs, we found that only the heterodimer could effectively deplete Tregs, while the Treg ratio was unchanged after homodimer therapy, emphasizing that only “2+1” form containing a single IL-2 molecule paired with dual aOX40 Fabs can reduce the Treg population in the tumors.

Compared with aOX40 antibodies, the IL-2/aOX40-Fc fusion protein promotes the proliferation of activated CD8⁺ T cells in vitro. In vivo, IL-2/aOX40-Fc depletes Tregs efficiently, but the expansion of CD8⁺ T cells was observed only after IL-2/aOX40-mutFc treatment. These data indicate that activated CD8⁺ T cells upregulate the expression of CD25 and OX40 receptors, which increases the risk of CD8⁺ T-cell depletion via wild-type-Fc. Due to the CD25 receptor expression level on activated CD8⁺ T being higher than OX40, we employed Fab physical blocking and N88D mutation to switch IL-2 dominated binding to aOX40 antibody-mediated targeting. Attenuated aOX40-mIL2-Fc achieved CTL proliferation and activated cell function in TME compared with IL-2/aOX40-Fc. Although the N88D mutation of IL-2 diminished its binding to CD122, due to OX40’s high expression on tumorous Treg, it preserved the overall binding avidity on Tregs by adding another aOX40 Fab on the N-terminal of attenuated IL-2, ensuring effective Treg cells depletion. Taken together, our design of aOX40-mIL2-Fc fusion protein with attenuated IL-2 uses the differential expression levels of OX40 receptors on tumorous T-cell subsets, enabling targeted depletion of Tregs while promoting CD8⁺ T-cell activation and proliferation.

Most designs aim to mitigate Treg binding feature IL-2 modifications that diminish CD25 interaction.^{22 32} However, recent studies have emphasized that CD25 expression on CD8⁺ T cells is vital for eliciting robust antitumor responses.^{22 33} Our exploration of various aOX40-IL2-Fc fusion proteins with F42A or F42A/N88D mutations revealed that loss of CD25 binding significantly impairs Treg depletion and overall antitumor efficacy, underscoring the critical role of CD25 in our therapeutic strategy. Various approaches have been attempted to improve the antitumor efficacy of IL-2 while reducing toxicity.⁴⁸ Meanwhile, localized expression of IL-2 may inadvertently enter circulation, activating NK cells systemically.⁴⁹ Our approach circumvents these challenges by enabling systemic delivery of the aOX40-mIL2-Fc fusion protein while reducing IL-2 induced NK cell binding and proliferation in the periphery through Fab blocking and the CD122 binding reduced N88D mutation.

Although immune checkpoint inhibitors are widely used in the clinic, their efficacy is often limited.^{34 35} While releasing brakes on dysfunctional CTLs can improve TIL function, this might not be adequate to enhance the therapeutic effect without additional stimulation. The aOX40-mIL2-Fc fusion protein promotes the activation and expansion of tumor-specific CD8⁺ T cells, while concurrently depleting Tregs, thereby increasing the CD8⁺ T/Treg ratio. Thus, combining aOX40-mIL2-Fc with PD-L1 blockade represents a strategic approach to synergistically enhance tumor control and overcome resistance to aPD-L1 therapies.

Tumor-resident CD4⁺ Treg cells play important roles in orchestrating an immune suppressive microenvironment. Treg cell infiltration is widely reported in cancers and has been shown to correlate with impaired CD8⁺ T-cell response and poor prognosis.^{50 51} Notably, the accumulation of Treg is observed in non-responsive patients with various cancer types following PD-1 blockade and combination therapies, suggesting the critical requirement of Treg depletion as well as CD8⁺ T-cell activation for optimal tumor control.52,54 Here, our aOX40-mIL2-Fc fusion protein simultaneously achieves tumor-infiltrating Treg cells depletion and CD8⁺ T cells expansion, exhibiting a potential synergistic therapeutic effect for the cancer types with Treg-enriching and insufficient CTLs infiltrating, such as colorectal cancer, lung cancer, and hepatocellular carcinoma. In conclusion, our study presents an innovative aOX40-mIL2-Fc bispecific antibody that demonstrates a potent antitumor effect while remaining safe in the periphery. Our strategy not only addresses significant limitations of conventional IL-2 and aOX40 therapies but also facilitates tumor clearance and induction of long-term systemic memory. Overall, our study elucidates the cellular mechanisms underlying aOX40-mIL2-Fc therapy, holding high translational potential for antitumor immunotherapy.

Data availability statement

Data are available upon reasonable request.