Glycoengineering-based anti-PD-1-iRGD peptide conjugate boosts antitumor efficacy through T cell engagement

Summary

Despite the important breakthroughs of immune checkpoint inhibitors in recent years, the objective response rates remain limited. Here, we synthesize programmed cell death protein-1 (PD-1) antibody-iRGD cyclic peptide conjugate (αPD-1-(iRGD)₂) through glycoengineering methods. In addition to enhancing tissue penetration, αPD-1-(iRGD)₂ simultaneously engages tumor cells and PD-1⁺ T cells via dual targeting, thus mediating tumor-specific T cell activation and proliferation with mild effects on non-specific T cells. In multiple syngeneic mouse models, αPD-1-(iRGD)₂ effectively reduces tumor growth with satisfactory biosafety. Moreover, results of flow cytometry and single-cell RNA-seq reveal that αPD-1-(iRGD)₂ remodels the tumor microenvironment and expands a population of “better effector” CD8⁺ tumor infiltrating T cells expressing stem- and memory-associated genes, including Tcf7, Il7r, Lef1, and Bach2. Conclusively, αPD-1-(iRGD)₂ is a promising antibody conjugate therapeutic beyond antibody-drug conjugate for cancer immunotherapy.

Graphical abstract

Highlights

• •αPD-1-(iRGD)₂ effectively penetrates and distributes in tumor tissue
• •αPD-1-(iRGD)₂ engages tumor cells with tumor-specific T cells
• •αPD-1-(iRGD)₂ remodels TME and promotes the expansion of tumor-specific TILs

Pan et al. design a glycoengineering-based antibody-peptide conjugate, αPD-1-(iRGD)₂, which penetrates deeply into a tumor, engages T cells and tumor cells, activates tumor-specific T cells, and expands a population of “better effector” CD8⁺ T cells with enhanced therapeutic potential for cancer immunotherapy.

Introduction

Cancer immunotherapy, represented by immune checkpoint inhibitors (ICIs), has brought significant survival benefits for patients with advanced cancer in recent years.^1,2,3 Although indications of ICIs have expanded over time, the ORR remains merely 10%–30% among several types of solid tumors, which prompts researchers to develop novel ICIs with better antitumor efficacy.^4,5,6 Bispecific T cell engagers (BiTEs) are emerging immunotherapy with promising cancer treatment potential.⁷ Due to their affinity to both T cells and tumor cells, BiTEs exhibit engagement effects and can mediate effective tumor control at lower doses compared to nature antibodies.⁸ However, BiTEs traditionally contain one or more CD3 binding domains to realize T cell engaging, which bypasses T cell recepter (TCR) and peptide-major histocompatibility complex (pMHC) interaction and may induce unselective T cell activation and severe side effects.⁹ Considering that PD-1⁺ T cells have been reported to be mainly tumor specific, we attempted to design a form of BiTE based on PD-1 antibodies to transfer the general activation of T cells induced by traditional BiTEs into a tumor-specific mode.¹⁰

However, analogous to chimeric antigen receptor T cell immunotherapy, the success of BiTEs in hematologic neoplasms has yet to be obtained in solid tumors. A key obstacle is the penetration difficulty caused by the dense microenvironment of solid tumors.⁸ Meanwhile, PD-1/programmed cell death protein-1 legend 1 (PD-L1) pathway inhibitors currently prescribed in clinic all belong to antibodies with a relative molecular mass of around 150 kDa.¹¹ The huge scale impedes the distribution of PD-1 antibodies and PD-1 antibody-derived BiTEs in tumor bulks. Recently, a positron emission tomography imaging study with zirconium-89-labeled pembrolizumab demonstrated that it required approximately 5–7 days for antibodies to reach peak concentrations in the tumor region.^12,13 Meanwhile, multidimensional angiography demonstrated that natural antibody penetration into the tumor is limited to the perivascular area, about 35% of the total tumor.¹⁴ Slow and uneven local tumor distribution limits the efficacy of PD-1 monoclonal antibodies. The results of a small-sample clinical trial suggested that the local maximum standardized uptake value of PD-1 antibody was positively correlated with the patients’ clinical response, progression-free survival, and overall survival.¹² However, there is no correlation between clinical responses and PD-1/PD-L1 immunohistochemical staining scores, suggesting enhancing penetrability is a potential strategy to improve the efficacy of PD-1 blockade and PD-1 antibody-derived BiTEs. For the convenience of immunology research in syngeneic mouse tumor models, we selected the antibody CS1003 developed by CStone Pharmaceuticals Co. (Suzhou) as the target for subsequent development. CS1003 is a humanized antibody with similar affinities for both human and murine PD-1 proteins (Kd 6.13 × 10⁻⁹ M for hPD-1, 3.99 × 10⁻⁹ M for mPD-1).¹⁵ Its binding site differs from approved antibodies such as pembrolizumab and nivolumab. In comparison to the clinically used pembrolizumab, CS1003 exhibits similar affinity and pharmacokinetic characteristics. In the MC38-hPD-L1 murine colon cancer tumor model, under equivalent dosage conditions, CS1003 and nivolumab demonstrate comparable antitumor effects.

Internalizing RGD (iRGD, c(CRGDKGPDC)), a promising tool in targeted drug delivery, operates through binding to αv integrins on tumor vasculature, facilitating extravasation into the tumor tissue.¹⁶ Subsequently, proteolytic cleavage exposes a CendR motif (CRGDK), allowing iRGD to interact with neuropilin-1 (NRP-1) on both tumor and endothelial cells. This interaction activates a transcytosis process, promoting efficient penetration deep into the tumor.¹⁷ Thus, iRGD’s multi-step mechanism, involving vascular binding and proteolytic activation underscores its potential in improving the targeted delivery of therapeutic agents to solid tumors. iRGD has been applied to deliver agents, including chemotherapeutics, single-chain mAbs, and nanoparticles, toward tumor bulks through interacting with integrins and neuropilin-1 (NRP1).^18,19 Our team has reported that T cell membrane modification with iRGD significantly strengthened T cell penetration and antitumor efficacy.^20,21 Theoretically, the attachment of iRGD can also facilitate the tumor penetration of αPD-1. However, iRGD modification of a native double-chain monoclonal antibody (mAb) hasn’t been reported. Traditionally, iRGD modification was conducted through plasmid transfection and subsequent eukaryotic or prokaryotic expression. The prokaryotic system cannot express double-chain mAbs, while the eukaryotic system has difficulty in constructing the cyclic structure of iRGD in full conversion (partial cyclization often observed).^22,23 Hence, we employed a nongenetic glycoengineering method, LacNAc-conju platform, to connect the conserved glycosylation site of the Fc domain with the αPD-1. LacNAc-conju platform is a one-step site-specific native IgG modification technique previously constructed by our team,^24,25 which allows for site-specific attachment of iRGD peptide with a peptide antibody ratio of close to 2.

In this study, we generated programmed cell death protein-1 antibody-iRGD cyclic peptide conjugate (αPD-1-(iRGD)₂) through glycoengineering and bio-orthogonal reaction. Similar to BiTEs, αPD-1-(iRGD)₂ engages T cells and tumor cells to activate tumor-specific T cells and promote tumor elimination. Systemic administration of αPD-1-(iRGD)₂ elicited tumor microenvironment (TME) modulation and impressive tumor growth control in various mouse tumor models. Moreover, αPD-1-(iRGD)₂ activated and expanded a subset of tumor infiltrating T cells (TILs) expressing stem and memory-associated genes (Tcf7, Il7r, Lef1, and Bach2), which may be induced by engaging tumor-specific T cells with cancer cells.

Results

Synthesis and characterization of αPD-1-(iRGD)₂

An anti-PD-1 antibody, CS1003, was selected for its cross-affinity to both human and mouse PD-1. Although the whole synthesis procedure of αPD-1-(iRGD)₂ was accomplished in a one-pot reaction, we would split the reaction into several parts and describe them separately in a logical order (Figure 1A). Firstly, the glycan structure on the N297 residue of the antibody CS1003 was modified by removing native sugars using EndoS (an endoglycosidase for specific hydrolysis at N297) and trimming α-(1,6)-fucose with AlfC (an α-fucosidase). Then, bovine β-1,4-galactosyltransferase 1 (Y289L) (β4GalT1) was used to add a galactose molecule to the GlcNAc residue, creating a monoantennary disaccharide structure, LacNAc, on N297 of CS1003. To meet the one-pot procedure of this platform, we introduced the GDP-fucose group to the N₃-iRGD peptide via click chemistry (Figures 1B and S1A). Electrospray ionization-mass spectroscopy (ESI-MS) manifested the mass of core substrate GDP-FAmP4Prop as 845.1989 [M-H⁺], which was within the allowable range of error (Figure S1B). The donor substrate of the FT enzyme, GDP-Fuc-iRGD, was purified through Prep-HPLC, and ESI-MS indicated its mass to be 936.7961 [M-2H⁺]/2 (Figure S1C). Then, using the FT enzyme and GDP-Fuc-iRGD, we nearly fucosylated all of the LacNAc units on CS1003 (1.88 iRGD per antibody) (Figure 1C). ESI-MS confirmed the relative molecular weight of αPD-1-(iRGD)₂ was 148833.00 Da. With glycoengineering, the cyclic structure of iRGD was preserved, and the complexity of plasmid transfection was spared (Figure S1D).

Figure 1.

Synthesis and characterization of αPD-1-(iRGD)₂

(A) The pattern diagram of the structure and chemical synthesis procedures of αPD-1-(iRGD)₂.

(B) Production of core substrate GDP-fucose-iRGD.

(C) ESI-MS characterization of αPD-1-(iRGD)₂.

(D and E) The binding affinity of αPD-1-(iRGD)₂ and unmodified antibody toward human (D) and murine (E) PD-1 protein by ELISA.

(F) Mean fluorescence intensity (MFI) fold change of PD-1-Jurkat cells after being incubated with αPD-1 or αPD-1-(iRGD)₂ for 1 h. Data represent mean ± SEM; for (D)–(F), n = 3.

(G–K) Flow cytometric analysis of changes in relative averaged fluorescence intensities of cancer cell lines (N87, HGC27, MFC, and B16) and normal cell line (293T) cocultured with 10 μg/mL αPD-1-(iRGD)₂ or 10 μg/mL αPD-1 before coculturing with anti-human IgG-PE (Abcam, #ab7005). The concentration of free iRGD (Genescript Biotech Corporation) was 100 μg/mL, and αNRP1 (Biolegend, #354502 and #145201) was 15 μg/mL. Data represent mean ± SEM; for (G)–(K), n = 3. For (G)–(J), one-way ANOVA test and Tukey’s multiple comparisons test. n.s., not significant; ∗p < 0.5; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

After the synthesis procedure, characterization assays were conducted to ensure the stability and bio-activity of PD-1 mAb. Generally, enzyme-linked immunosorbent assay (ELISA) showed no difference in binding affinity toward both human and murine PD-1 protein between αPD-1-(iRGD)₂ and the unmodified αPD-1 (Figures 1D and 1E). The calculated dissociation constant (Kd) of αPD-1-(iRGD)₂ to human PD-1 was 1.918 × 10⁻⁸ nM, and murine PD-1 was 2.65 × 10⁻⁸ nM, while Kd of αPD-1 to human PD-1 was 1.931 × 10⁻⁸ nM, and murine PD-1 was 1.867 × 10⁻⁸ nM. We transfected hPD-1 into Jurkat cells (PD-1-Jurkat) in order to mimic PD-1⁺ T cells (Figure S1E). Within the cell level, we found that effective concentration 50 (EC₅₀) of αPD-1-(iRGD)₂ to PD-1-Jurkat was about 0.04 μg/mL, while that of αPD-1 was 0.06 μg/mL (Figure 1F). Pharmacokinetics analysis demonstrated that αPD-1-(iRGD)₂ shared similar half-life (more than 7 days) and serum stability to unmodified αPD-1 (Figures S1F–S1H). Expression of iRGD receptors (integrin αvβ5 and NRP-1) was identified next. Both human and murine cancer cell lines (NCIN87, HGC27, MFC, and B16F10) exhibited higher expression of integrin αvβ5 and NRP-1, while normal cell lines (293T, HBE, and GES-1) rarely expressed iRGD receptors (Figures S1I–S1l). αPD-1-(iRGD)₂ attained stronger affinity to HGC27, NCI-N87, MFC, and B16F10 cancer cell lines than unmodified αPD-1 (Figures 1G–1J and S1M). Competitive inhibitors (free iRGD and αNRP-1 and anti-integrin αvβ5) at least partially abrogated the affinity of αPD-1-(iRGD)₂ toward cancer cell lines. Combination of αNRP-1 and anti-integrin αvβ5 exhibited complete inhibition, indicating that both integrin αvβ5 and NRP-1 contributed to the cell binding of αPD-1-(iRGD)₂ (Figures S1N and S1O). Meanwhile, both αPD-1 and αPD-1-(iRGD)₂ exhibited little binding to HBE, 293T, and GES-1 normal cell lines, implying attachment of iRGD would not increase the accumulation of αPD-1 in normal tissue (Figures 1K and S1P–S1Q).

For the convenience of subsequent in vivo distribution assay and fluorescence imaging, we attached extra fluorophores, Cy5, to αPD-1-(iRGD)₂ (designated as αPD-1-(iRGD)₂-Cy5) and αPD-1 (designated as αPD-1-Cy5) also through glycoengineering methods. ESI-MS showed the Cy5-antibody-ratio of αPD-1-Cy5 was about 1.8, and that of αPD-1-(iRGD)₂-Cy5 was about 1.75 (Figures S1R–S1T).

αPD-1-(iRGD)₂ engages PD-1⁺ T cells and tumor cells

Due to the dual targeting capacity of αPD-1-(iRGD)₂ to PD-1 and iRGD receptors, we first examined whether αPD-1-(iRGD)₂ might engage T cells and tumor cells (Figure 2A). Considering that naive T cells do not express PD-1, we applied CD3/CD28 beads, IL2, IL7, and IL15 to induce PD-1 expression of PBMC and OT-I cells. PD-1 expression of PBMC, OT-I, and PD-1-Jurkat were checked before subsequent assays (Figures S2A and S2B). Afterward, we conducted a conjugate formation assay at an effector target ratio (E:T) of 1:10 between OT-I cells and B16-OVA cells. Flow cytometry exhibited significantly increased cell engagement between OT-I cells and tumor cells in the αPD-1-(iRGD)₂ group (Figures 2B and 2C). The engagement of T cells and tumor cells was competitively inhibited by iRGD, anti-integrin αvβ5, αNRP-1, and αPD-1. Fluorescence imaging of 5,6- carboxyfluorescein diacetate, succinimidyl ester (CFSE)-labeled HGC27 cells, and CellTracker Deep Red Dye-labeled T cells cocultured with different agents visually displayed the engagement of cells mediated by αPD-1-(iRGD)₂ (Figure S2C). After that, we focused on whether αPD-1-(iRGD)₂ could mediate better tumor killing efficacy. In both monolayer culture system and multicellular spheroid (MCS) system, αPD-1-(iRGD)₂ effectively promoted the cytotoxicity of OT-I against B16-OVA cells in a concentration-dependent manner (Figures 2D and 2E). This enhanced cytotoxicity could be inhibited by excessive dissociative iRGD, αNRP-1, and αPD-1 (Figure 2F). We further tested the capacity of αPD-1-(iRGD)₂ to activate T cells. After coculturing, αPD-1-(iRGD)₂ significantly elevated the expression of activation markers (CD69 and CD25) and cytotoxicity markers (GZMB and IFNγ) on T cells (Figures 2G–2I). PD-L1 was intermediately or slightly expressed on B16-OVA cells and HGC27 cells, indicating that apart from PD-1 blockade, BiTE-like effect also contributed to the enhanced antitumor efficacy of αPD-1-(iRGD)₂ (Figures S2D and S2E). These interesting findings support the idea that αPD-1-(iRGD)₂ can be considered a format of BiTE that does not target CD3.

Figure 2.

αPD-1-(iRGD)₂ engages tumor cells and T cells

(A) Diagram of αPD-1-(iRGD)₂ engaging tumor cells and T cells.

(B) Percentage of CFSE (eBioscience, #65-0850-84) and Dye670 (eBioscience, #65-0840-85) double-positive cells in B16-OVA cells incubated with OT-I cells at an E:T ratio of 1:10.

(C) Representative flow cytometry results of (B). Data represent mean ± SEM; for (B), n = 3.

(D) Lysis percentage of monolayer B16-OVA cells coculturing with OT-I at an E:T rate of 1:10 under indicated antibody concentration. The concentration of free iRGD if added was 10 μg/mL. Two-way ANOVA test and Tukey’s multiple comparisons test.

(E) Lysis percentage of MCSs constructed with B16-OVA cells coculturing with OT-I at an E:T rate of 1:10 under indicated antibody concentration. Concentration of free iRGD if added was 10 μg/mL.

(F) Lysis percentage of B16-OVA cells coculturing with OT-I cells at an E:T rate of 10:1. Concentrations of αPD-1 and αPD-1-(iRGD)₂ were 10 μg/mL, iRGD was 100 μg/mL, and αNRP1 was 15 μg/mL. Specially in the αPD-1-(iRGD)₂+αPD-1 group, the concentration of αPD-1 was 50 μg/mL.

(G) Expression of GZMB in OT-I cells under coculture conditions in (F).

(H) Expression of IFNγ (Biolegend, #505810) in OT-I cells under coculture conditions in (F).

(I) Expression of CD25 (eBioscience, #12-0251-83) and CD69 (Biolegend, #104506) in OT-I cells under coculture conditions in (F). For (B) and (F)–(I), one-way ANOVA test and Tukey’s multiple comparisons test. For (D) and (E), two-way ANOVA test and Tukey’s multiple comparisons test. Data represent mean ± SEM; for (D)–(I), n = 3. n.s., not significant; ∗p < 0.5; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

αPD-1-(iRGD)₂ penetrates and preferentially distributes in tumor bulk

In order to analyze the distribution characteristics of αPD-1-(iRGD)₂, we first constructed MCSs with a human gastric cancer cell line, HGC27 (Figure 3A). αPD-1-(iRGD)₂-Cy5 infiltrated into the core of MCSs rapidly and robustly, compared with αPD-1 alone or with different doses of free iRGD (Figures 3B and 3C). The high-dose (100 μg/mL) free iRGD group failed to exhibit equivalent penetration as αPD-1-(iRGD)₂-Cy5.

Figure 3.

Targeted distribution and penetration of αPD-1-(iRGD)₂ within tumor bulk

(A) Diagram of penetrability assay of Cy5-labeled αPD-1-(iRGD)₂.

(B) Confocal microscopy images showing penetration of Cy5-labeled antibody in preconstructed MCSs. Magnification, ×50; scale bar, 200 μm.

(C) The relative fluorescence intensity of MCSs incubated with Cy5-labeled antibody or control agents under indicated periods.

(D) Average radiant efficiency at tumor location of mice inoculated with MFC cells. Cy5-labeled antibody (50 μg) was intraperitoneally injected.

(E) Bioluminescence images of Cy5-labeled antibody at intervals after intraperitoneal injection to mice inoculated with MFC cells.

(F) Ex vivo images of tumor, liver, heart, kidney, lung, and spleen collected 48 h after intraperitoneal injection.

(G) Average radiant efficiency of resected tumor bulk 48 h after fluorescent agents were administered.

(H) Immunofluorescence analysis of tumor sections at 48 h after last injection. αPD-1 was labeled with anti-human IgG-FITC (Abcam, #ab6854); CD8 T cells were stained with anti-mouse CD8-PE (Abcam, #ab25498); nucleus, blue. Magnification, ×20; scale bar, 200 μm. Data represent mean ± SEM; for (B)–(G), n = 3. For (C) and (E), two-way ANOVA test and Tukey’s multiple comparisons test. For (G), one-way ANOVA test and Tukey’s multiple comparisons test. n.s., not significant; ∗p < 0.5; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

Afterward, the in vivo distribution of αPD-1-(iRGD)₂ was examined in the murine gastric cancer model. A subcutaneous injection of 1 × 10⁶ MFC cells was performed at the left inguinal groove of 6- to 8-week-old 615 mice. Upon reaching a calculated tumor volume of 200 mm³, 50 μg of αPD-1-(iRGD)₂-Cy5 or αPD-1-Cy5 was administered intraperitoneally. Near-infrared imaging of the living body displayed the rapid and intensive distribution of αPD-1-(iRGD)₂-Cy5 in the tumors (Figures 3D and 3E). 48 h after injection, tumors and main organs were resected to evaluate the fluorescence intensity. The average ex vivo radiant efficacy of tumor bulk doubled compared with αPD-1-Cy5 (Figures 3F and 3G). No significant increase in organ absorption was observed, indicating αPD-1-(iRGD)₂-Cy5 mainly distributed in tumor bulk. Immunofluorescence of tumor sections also exhibited increased infiltration of αPD-1-(iRGD)₂ in the tumor (Figure 3H). These observations were consistent with the in vitro findings that αPD-1-(iRGD)₂ had potent affinity to tumor cells.

αPD-1-(iRGD)₂ effectively reduces tumor growth in murine tumor models

ICIs aim to mobilize the host’s immune system, rescuing T cells from exhausted status and reviving immune response against cancer cells.²⁶ Therefore, we selected immune-competent murine tumor models to evaluate the antitumor efficacy of αPD-1-(iRGD)₂. In the MFC subcutaneous tumor model, mice treated with αPD-1-(iRGD)₂ exhibited delayed tumor growth and remarkably lighter tumor weight at the endpoint (Figures 4A–4C and S3A). αPD-1 alone also inhibited the tumor growth to a certain degree, while αPD-1 combined with free iRGD showed little difference compared to the αPD-1 group. The antitumor efficacy of αPD-1-(iRGD)₂ was also tested in the B16F10 murine melanoma model and 4T1 orthoptic tumor model, which were considered as “cold tumor” due to poor immune infiltration. Impressively, αPD-1-(iRGD)₂ also postponed tumor growth and remarkably lightened tumor weight at the endpoint, which indicates that αPD-1-(iRGD)₂ might have broad-spectrum antitumor effects (Figures 4D–4F and S3B–S3F). For survival analysis, αPD-1-(iRGD)₂ significantly extended survival in the MFC tumor model (Figure 4G).

Figure 4.

Antitumor efficacy of αPD-1-(iRGD)₂

(A) Schematic of the treatment regimen in MFC mouse gastric tumor model. Briefly, mice were treated with 1 × 10⁶ MFC cells and injected intraperitoneally with PBS (100 μL), αPD-1 (5 mg/kg) alone, or with free iRGD (2.5 μg) and αPD-1-(iRGD)₂ (5 mg/kg) every 3 days.

(B and C) Tumor growth profile (B) and weight (C) of tumors collected at the endpoint of (A).

(D) Schematic of the treatment regimen in B16F10 mouse melanoma tumor model. Briefly, mice were treated with 1 × 10⁵ B16F10 cells and injected intraperitoneally with PBS (100 μL control), αPD-1(5 mg/kg) alone, or with free iRGD (2.5 μg) and αPD-1-(iRGD)₂ (5 mg/kg) every 3 days.

(E and F) Tumor growth profile (E) and weight (F) of tumors collected at the endpoint of (D).

(G) Survival plot of MFC mouse gastric cancer model. Briefly, mice were treated with 1 × 10⁶ MFC cells and injected intraperitoneally with PBS (100 μL control), αPD-1 (5 mg/kg) alone, or with free iRGD (4 μmol/kg) and αPD-1-(iRGD)₂ (5 mg/kg) every 3 days. A total of 4 injections were assigned to each mouse.

(H–P) Quantitative flow cytometry results indicating the abundance and characteristics of TILs in an MFC model. 4- to 6-week-old 615 mice were treated with 1 × 10⁶ MFC cells and injected intraperitoneally with PBS (100 μL), αPD-1 (5 mg/kg) alone, or with free iRGD(4 μmol/kg) and αPD-1-(iRGD)₂ (5 mg/kg) every 3 days. (H) CD3⁺ T cells, (I) CD4⁺ T cells, (J) CD8⁺ T cells, (K) PD-1⁺CD8⁺ T cells, (L) IFNγ⁺CD8⁺ T cells, (M) CD137⁺CD8⁺ T cells, (N) Ki67⁺CD8⁺ T cells, (O) CD137⁺CD4⁺ T cells, and (P) Ki67⁺CD4⁺ T cells. Flow cytometry plots here utilized the following agents: anti-mouse CD3-FITC (Biolegend, #100204), anti-mouse CD3e-PC7 (Biolegend, #100320), anti-mouse CD4-PE (Biolegend, #100408), anti-mouse CD8-APC (Biolegend, #100712), anti-mouse CD8-FITC (Biolegend, #100706), anti-mouse Ki67-PC7 (Biolegend, #151217), anti-mouse PD-1-PerCP5.5 (Biolegend, #124334), anti-mouse IFNγ-PerCP5.5 (Biolegend, #505822), and anti-mouse CD137-PE (Biolegend, #106106). Data represent mean ± SEM; for (A)–(F), n = 6; for (G), n = 8; for (H)–(P), n = 5. For (B) and (E), two-way ANOVA test and Tukey’s multiple comparisons test. For (C), (F), and (H)–(P), one-way ANOVA test and Tukey’s multiple comparisons test; n.s., not significant; ∗p < 0.5; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

Quantitative flow cytometry results revealed that in the MFC tumor model, αPD-1-(iRGD)₂ group had significantly more CD8⁺ T cells in the TME than other groups. No difference was observed in CD3⁺ T cells or CD4⁺ T cells (Figures 4H–4J). PD-1 expression on CD8⁺ T cells was similarly inhibited in αPD-1, αPD-1+iRGD, and αPD-1-(iRGD)₂ groups (Figure 4K). αPD-1-(iRGD)₂ also significantly elevated the expression of IFNγ in CD8⁺ T cells, while αPD-1 alone or with free iRGD exhibited no influence (Figure 4L). In the 4T1 tumor model, increases of CD8⁺ T cells, CD137⁺ T cells, GZMB⁺CD8⁺ T cells, and IFNγ⁺CD8⁺ T cells were observed in mice treated with αPD-1-(iRGD)₂ (Figures S3G–S3M). A similar trend was also observed in tumor-draining lymph nodes of the MFC tumor model (Figures S3N–S3Q).

During the treatment, weight loss and other side effects were not observed (Figure S4A). At the end of the experiment, no observable damage was recorded in the main organs as indicated by H&E staining (Figures S4B and S3C). These results demonstrated that αPD-1-(iRGD)₂ exhibited significant tumor suppression effects with few side effects in these mouse models.

αPD-1-(iRGD)₂ promotes tumor-specific T cell activation and proliferation

Flow cytometry analysis exhibited the increase of CD137⁺CD8⁺ T cells and Ki67⁺CD8⁺ T cells in mice treated with αPD-1-(iRGD)₂ (Figures 4M and 4N). However, increased expressions of CD137 and Ki67 were not observed in CD4⁺ T cells (Figures 4O and 4P). These results indicated that the antitumor effect of αPD-1-(iRGD)₂ relied on the expansion of tumor-specific CD8⁺ T cells in the TME. To further clarify the antitumor mechanism of αPD-1-(iRGD)₂, we conducted T cell migration inhibition and CD8⁺ T cell depletion assay. After the injection of fingolimod hydrochloride (FTY720, an S1P receptor agonist, MCE, #HY-12005) to block T cell trafficking to tumor tissue during αPD-1-(iRGD)₂ treatment, tumor suppression was not significantly affected (Figures 5A–5C and S4D). Moreover, depleting CD8⁺ T cell subsets with anti-CD8 (Bioxcell, # BE0117) confirmed that CD8⁺ T cells mainly contributed to the therapeutic effect (Figures 5D–5F and S4E). Here, we indicated the antitumor effect of αPD-1-(iRGD)₂ depended on pre-existing T cells in the tumor microenvironment instead of recruiting T cells from the peripheral.

Figure 5.

Antitumor efficacy of αPD-1-(iRGD)₂ depends on pre-existing intra-tumoral CD8⁺ T cells

(A) Schematic of the T cell migration inhibition regimen in MFC mouse gastric tumor model. Briefly, mice were treated with 1 × 10⁶ MFC cells and injected intraperitoneally with PBS (100 μL control), αPD-1-(iRGD)₂ (5 mg/kg), and FTY720 (MCE, #HY-12005, 20 μg) every 3 days.

(B) Flow cytometry analysis of T cell abundance in peripheral blood to validate the T cell migration inhibition of FTY720.

(C) Tumor growth profile of (A).

(D) Schematic of the CD8⁺ T cell depletion regimen in MFC mouse gastric tumor model. Briefly, mice were treated with 1 × 10⁶ MFC cells and injected intraperitoneally with PBS (100 μL control), αPD-1-(iRGD)₂ (5 mg/kg), and αCD8 (BioXCell, #BE0117, 200 μg) every 3 days.

(E) Flow cytometry analysis of CD8⁺ T cell abundance in tumor to validate CD8⁺ T cell depletion.

(F) Tumor growth profile of CD8⁺ T cell depletion assay.

(G) Schematic of the treatment regimen in MFC mouse gastric tumor model. Briefly, mice were treated with 1 × 10⁶ MFC cells and injected intraperitoneally with PBS (100 μL control), αPD-1 (0.1 mg/kg) alone, or with free iRGD (4 μmol/kg) and αPD-1-(iRGD)₂ (0.1 mg/kg).

(H) Weight of murine subcutaneous tumors resected at the end of (G).

(I) Tumor growth profile of (G).

(J) Flow cytometry results indicating the abundance of CD8⁺ T cell (Biolegend, #100702) from resected tumor bulk at the endpoint of (G).

(K) Flow cytometry quantification of PD-1 (Biolegend, #135206) expression on CD8⁺ T cell from resected tumor bulk at the endpoint of (G).

(L) Flow cytometry quantification of CD39 (Biolegend, #143806) expression on PD-1⁺CD8⁺ T cells from resected tumor bulk at the endpoint of (F).

(M and N) Flow cytometry results indicating the abundance of CD4⁺ T cell (Biolegend, #100408) (M) and Treg (Biolegend, #320014 and #100412; eBioscience, #12-0251-83) (N) from resected tumor bulk at the endpoint of (G).

(O) Flow cytometry analysis of propidium-iodide-positive tumor cells preloaded with GP33 (Genescript Biotech Corporation, 500 nM) or SIINFEKL peptide (Genescript Biotech Corporation, 500 nM) after simultaneously coculturing with OT-I cells. The ratio of unprimed, GP33-primed, SIINFEKL-primed B16F10 cells, and OT-I cells was 1:1:1:10. Concentration of αPD-1 and αPD-1-(iRGD)₂ was 10 μg/mL, and that of iRGD was 100 μg/mL

(P) Flow cytometry analysis of activation markers (CD25 and CD69) on OT-I cells and spleen T cells simultaneously cocultured with SIINFEKL peptide (500 nM) preloaded tumor cells. The ratio of OT-I cells, non-specific spleen T cells, and SIINFEKL-primed B16 cells was 5:5:1.

(Q) Histogram of the percentage of conjugated cells when 2 × 10⁵ Dye 670-stained NY-ESO-1_157-165-primed HLA-A∗0201-Raji cells were cocultured 1 h with mixed 2 × 10⁴ CFSE-stained 1G4-PD-1-Jurkat cells and 2 × 10⁴ Dye 450-stained PD-1-Jurkat cells. The concentration of αPD-1-(iRGD)₂ was 10 μg/mL, while blinatumomab (MCE, #HY-P9963) was 1 μg/mL

(R) Histogram of CD69 (Biolegend, #319102) expression on 1G4-PD-1-Jurkat and PD-1-Jurkat under the same condition in (P).

(S) Flow cytometry chart of (Q).

(T) Flow cytometry chart of (R). Data represent mean ± SEM. For (B) and (O)–(T), n = 3. For (C), (E), (F), and (H)–(N), n = 5. For (B), (E), (H), and (J)–(M), one-way ANOVA test and Tukey’s multiple comparisons test. For (C), (F), (I), and (O)–(R), two-way ANOVA test and Tukey’s multiple comparisons test. n.s., not significant; ∗p < 0.5; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

To further analyze the effect of αPD-1-(iRGD)₂ on intra-tumoral CD8⁺ T cells, we minimized the dose to 0.1 mg/kg and limited administration frequency, similar to other BiTEs (Figure 5G). Dose and frequency reduction didn’t diminish the antitumor efficacy of αPD-1-(iRGD)₂, while αPD-1 alone or αPD-1 with free iRGD group exhibited no tumor suppression (Figures 5H and 5I). Flow cytometry showed the increase of CD8⁺ T cells and CD39⁺CD8⁺ T cells in mice treated with αPD-1-(iRGD)₂, while comparable expression of PD-1 was detected in each group (Figures 5J–5L). CD4⁺ T cells and Treg abundance were not significantly modified as well (Figures 5M and 5N).

Interestingly, CD39⁺ and CD137⁺ CD8⁺ T cell increase was observed after the administration of αPD-1-(iRGD)₂. As CD39 and CD137 were both reported as tumor-specific markers for T cells, we further investigated the effect of αPD-1-(iRGD)₂ on tumor-specific T cells. We pretreated B16F10 cells with GP33 peptide (KAVYNFATC, Genscript) or OVA_257-264 peptide (SIINFEKL, Genscript) to simulate irrelevant or recognizable tumor cells of OT-I cells. After coculturing OT-I cells with mixed untreated, GP33-treated, or SIINFEKL-treated B16F10 cells for 12 h, αPD-1-(iRGD)₂ specifically promoted the cytotoxicity of OT-I cells to recognizable tumor cells (Figure 5O). When SIINFEKL-treated B16F10 cells were cocultured with 1:1 mixed OT-I cells and non-specific splenic T cells, αPD-1-(iRGD)₂ significantly increased the expression of T cell activation markers (CD25 and CD69) on OT-I cells other than non-specific splenic T cells (Figure 5P).

To demonstrate the differences in engaging and activating tumor-specific T cells between αPD-(iRGD)₂ and other BiTEs (e.g., blinatumomab, an anti-CD3/CD19 BiTE), we turned to NY-ESO-1_157-165 and its paired TCR, 1G4 α95:LY TCR, to simulated antigen-specific T cells.²⁷ Raji cells were transfected with HLA-A∗0201 to fit in the requirement of presenting the epitope of NY-ESO-1_157-165. PD-1-Jurkat cells were transfected with 1G4. The expressions of CD19 on HLA-A∗0201-Raji cells and CD3 on 1G4-PD-1-Jurkat cells were also identified with flow cytometry (Figures S4F–S4I). Dye 670-stained NY-ESO-1_157-165-primed HLA-A∗0201-Raji cells were cocultured 1 h with mixed 2 × 10⁴ CFSE-stained 1G4-PD-1-Jurkat cells and 2 × 10⁴ Dye 450-stained PD-1-Jurkat cells. αPD-(iRGD)₂ engaged comparable 1G4-PD-1-Jurkat cells as blinatumomab, while the amount of engaged PD-1-Jurkat cells was significantly less than blinatumomab (Figure 5Q). Meanwhile, αCD3 domain in blinatumomab caused uniformed high expression of CD69 on 1G4-PD-1-Jurkat and PD-1-Jurkat. Although the expression of CD69 induced by αPD-(iRGD)₂ was lower than blinatumomab, 1G4-PD-1-Jurkat expressed a significantly higher level of CD69 than PD-1-Jurkat (Figures 5R–5T). Therefore, we postulated that αPD-1-(iRGD)₂ engaged and activated tumor-specific CD8⁺ T cells to realize the impressive antitumor efficacy.

αPD-1-(iRGD)₂ remodels tumor immune microenvironment

We performed single-cell RNA sequencing of CD45⁺ tumor-infiltrating immune cells to further depict the immune landscape after the administration of αPD-1-(iRGD)₂. We separated CD45⁺ cells from different treatment groups with density gradient centrifugation and fluorescence-activated cell sorting. The CD45⁺ immune cells from six mice in the same treatment group were pooled together and barcoded with the same label. The results of single-cell RNA sequencing were exhibited via Uniform Manifold Approximation and Projection (UMAP) (Figures 6A–6C). UMAP displayed 27 clusters including 5 cell types. According to specific cell markers, we defined clusters 4, 7, 10, 11, 12, 15, 18, and 23 into T cells (Trac, Trbc, and Cd3d), clusters 6, 9, 19, and 22 as dendritic cells (DCs) (Cd74 and H2-Aa), clusters 0, 2, 6, 8, 14, 16, 21, and 24 as macrophages (Adgre1, Itgax, and Cd83), clusters 3, 5, and 20 as monocytes (Cd14), and clusters 12 and 26 as natural killer (NK) cells (Nkg7, Klrb1c, and Klrg1) (Figure S5).

Figure 6.

αPD-1-(iRGD)₂ remodels tumor microenvironment

(A) Schematics of the treatment regimen in MFC homograft models are shown. Briefly, mice bearing tumor burdens were treated with 1 × 10⁶ MFC cells and injected intraperitoneally with PBS (100 μL control), αPD-1 (5 mg/kg) alone, or with free iRGD (50 μg) and αPD-1-(iRGD)₂ (5 mg/kg) every 3 days.

(B) Two-dimensional (2D) UMAP visualization of CD45⁺ tumor-infiltrating immune cells colored according to subset (upper) and specific treatment group (lower).

(C) The ratio of each subset within the CD45⁺ tumor-infiltrating immune cells depicted in (B).

(D) Heatmap of average relative expression of selected genes in NK cells across treatment groups within CD45⁺ tumor-infiltrating immune cells.

(E and F) Heatmap of average relative expression of selected genes in CD4⁺ (E) and CD8⁺ (F) T cells across treatment groups within CD45⁺ tumor-infiltrating immune cells depicted in (B).

Consistent with our previous findings, enhancement of effector function was observed in CD8⁺ T cells, CD4⁺ T cells, and NK cells (Figures 6D–6F). In mice treated with αPD-1-(iRGD)₂, NK cells in TME expressed higher levels of effector genes (Nkg7, Gzma, Prf1, and Klrg1), co-stimulatory factors (Slamf7, Cd244a, and Klrk1), migration genes (Itgb2, Itga2, Icam1), and immune stimulatory cytokines and receptors (Ifng, Ifngr1, Il2ra, and Il2rb) (Figure 6D). In CD4⁺ T cells, mRNA levels of immune stimulating cytokines and receptors (Ifng, Il2ra, Il18r, etc.) and co-stimulatory factors (Cd27 and Cd28) were elevated, while those of Il4 and Il6 were downregulated (Figure 6E). For CD8⁺ T cells, the expression levels of several effector genes (Gzma, Ifng, Ifngr1, etc.), inflammatory cytokine receptors (Il2ra, Il7r, Il12rb1, etc.), and co-stimulatory factors (Cd28, Cd27, etc.) were upregulated in CD8⁺ TILs from mice that received αPD-1-(iRGD)₂ treatment (Figure 6F). One of the most notable changes in CD8⁺ T cells after αPD-1-(iRGD)₂ administration was the upregulation of several stem- and memory-associated genes (Lef1, Tcf7, Bach2, and Ikzf2).^28,29,30

Apart from T cells and NK cells, myeloid cells also participated in tumor suppression of αPD-1 directly or indirectly.^31,32,33 Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis was performed using differentially expressed gene signatures of DCs and macrophages. Although quantitative flow cytometry exhibited no change in abundance, the results of KEGG showed an enrichment of response to IFNγ, antigen processing, and presentation in DCs from mice treated with αPD-1-(iRGD)₂ (Figures S6A–S6D). Differentially expressed genes of macrophages were enriched in cellular response to lipopolysaccharide and biotic stimulus as well (Figures S6E–S6H). These observations indicated αPD-1-(iRGD)₂ simultaneously ameliorated the function of DCs and macrophages. For neutrophils, no change was observed in abundance as well (Figure S6I). In a 4T1 orthoptic tumor model, quantitative flow cytometry analysis of myeloid cells also showed little difference (Figures S6J–S6N). In conclusion, αPD-1-(iRGD)₂ profoundly modulates the immune suppressive TME thus facilitating the cytotoxicity of CD8⁺ T cells.

To gain further insights into how the differentiation program of CD8⁺ T cells was altered by αPD-1-(iRGD)₂, we reperformed UMAP analysis of predefined T cells (Figure 7A). UMAP displayed 10 clusters and was annotated into 6 subtypes according to cell markers, including better effector CD8⁺ T cells (cluster 4), effector CD8⁺ T cells (cluster 3), intermediate exhausted CD8⁺ T cells (clusters 0, 2, and 5), terminal exhausted CD8⁺ T cells (clusters 1 and 9), memory CD8⁺ T cells (cluster 6), and CD4⁺ T cells (clusters 7 and 8) (Figures 7B and 7C). Impressively, the abundance of a subset of CD8⁺ T cells (better effector CD8⁺ T cells), characterized by the expression of stem- and memory-associated genes (Tcf7, Il7r, Lef1, and Bach2) and intermediate expression of effector genes (Gzma, Gzmb, and Ifng), expanded in TILs of αPD-1-(iRGD)₂ treated mice (Figure 7D). In addition, this population of CD8⁺ TILs also expressed low levels of transcription for Pdcd1, Lag3, Havcr2 (TIM-3), and Entpd1 in line with a non-exhausted profile. According to previous work that PD-1-cis IL-2R agonism yields better effectors from stem-like CD8⁺ T cells, we identified this subset of CD8⁺ TILs as “better effector” T cells underlining their highly functional effector profile and lower degree of exhaustion.^34,35 The expansion of better effector CD8⁺ T cells implies that αPD-1-(iRGD)₂ reinforced the effector function of TILs while avoiding exhaustion and maintaining stem and memory characteristics of CD8⁺ TILs.

Figure 7.

αPD-1-(iRGD)₂ expands a population of better effector CD8⁺ TILs

(A) Two-dimensional (2D) UMAP visualization of CD3⁺ TILs colored according to subset (left) and specific treatment group (right).

(B) Ratio of each subset within the CD3⁺ tumor-infiltrating T cells depicted in (A).

(C) Heatmap of normalized expression of several representative genes within the clusters of CD8⁺ TILs depicted in (A).

(D) Normalized expression of several representative genes within the six clusters of CD3⁺ TILs.

Construction and evaluation of murine αPD-1-(iRGD)₂

Although CS1003 binds to both human and murine PD-1, its humanized Fc domain may cause an unpredictable influence on the antitumor efficacy of αPD-1-(iRGD)₂. To exclude the incompatible Fc domain and examine the suitability of LacNAc-conju to murine antibodies, we constructed αPD-1-(iRGD)₂ from murine αPD-1 (G4C2, TopAlliance). ESI-MS analysis confirmed that the product possessed an expected molecular weight, while ELISA results exhibited that αPD-1-(iRGD)₂ sustained its affinity to murine PD-1 (Figures S7A and S7B). After that, the tumor suppression function of αPD-1-(iRGD)₂ was evaluated in the murine gastric cancer tumor model. Murine αPD-1-(iRGD)₂ displayed superior antitumor efficacy compared with αPD-1 monotherapy (Figures S7C–S7E). During the regimen, no weight loss was observed (Figure S7F).

Discussion

BiTEs are an emerging and promising form of cancer immunotherapy for their engaging effect.³⁶ Nevertheless, current anti-CD3-based BiTEs have several drawbacks, including the recruitment of immune suppressive CD3⁺ T cell clusters, the risk of eliciting systemic cytokine storms, the upregulation of immune checkpoint molecules, the presence of an immunosuppressive TME, tumor antigen loss or escape, and suboptimal potency.³⁷ While anti-CD3-based BiTEs may cause universal CD3⁺ T cell mobilization since almost all T cells express CD3, PD-1 antibody-based BiTEs can specifically target and redirect tumor-specific CD8⁺ TILs, which have been documented to express higher levels of PD-1 and are temporarily functionally impaired.³⁸ Therefore, we chose αPD-1 as the modification object. In this work, we revealed that αPD-1-(iRGD)₂ mainly acted on PD-1⁺ TILs and avoided non-selective T cell activation and excessive TCR signaling. Considering the penetration-promoting effects of iRGD, we ruled out penetration-related factors and found that αPD-1-(iRGD)₂ could still significantly enhance T cell cytotoxicity. Besides, multiple competitive inhibitors, including free iRGD, αNRP1, and αPD-1, abrogated the engagement formation, tumor elimination, and T cell activation mediated by αPD-1-(iRGD)₂ to a certain degree. Even when we limited αPD-1-(iRGD)₂ doses to 0.1 mg/kg and reduced administration frequencies, αPD-1-(iRGD)₂ treatment could similarly amplify a tumor-specific CD8⁺ T cell subset and induce lasting tumor control without blocking PD-1 expression on CD8⁺ T cells, which further indicated that αPD-1-(iRGD)₂, upon penetrating the tumor core via iRGD, primarily engages tumor cells and T cells, rather than merely functioning as a PD-1 blockade.

Several research groups have also developed other forms of αPD-1-based BiTEs. The reported αPD-1-based BiTEs primarily target HER2, VEGF, and others.^39,40,41 These bispecific antibodies all utilize the IgG1 Fc domain, indicating that antibody-dependent cell-mediated cytotoxicity is also a crucial component of their antitumor effects. In contrast, our conjugate has weaker Fc effect minoring the possibility of tumor-specific T cell loss. Furthermore, unlike the other bispecific antibodies that employ antibody fragment design for tumor targeting, some of which exhibit strong affinity and can mediate tumor cytotoxicity without the need for TCR-MHC interaction, our conjugate demonstrated relatively weaker affinity for tumor cells. It still relies on the TCR-pMHC complex interaction to initiate subsequent cytotoxicity. This structural distinction forms the basis for the tumor-specific T cell activation characteristics exhibited by αPD-1-(iRGD)₂. Recently, αPD-1-αCLEC9A-based DC-T cell engager (BiCE) has also been reported, with Fc function null and aimed at potentiating cell circuits and communication pathways needed for effective antitumor immunity.⁴² Different from traditional BiTE, the focus of BiCE has been switched to interaction between immune cells, exhibiting more like immunotherapy. Based on different design ideas, αPD-1 derivatives have been an emerging direction of therapy exploration.

Unlike single-chain variable fragment (scFv)-based anti-CD3-iRGD generated by plasmid construction and expression, we performed a site-specific modification of both human and murine αPD-1 via a glycoengineering platform that preserved the natural characteristics of antibodies. It has several advantages as follows. Due to its larger molecular size and FcRn-mediated recycling processes, αPD-1-(iRGD)₂ with an Fc region might also have a longer half-life in circulation than scFvs.⁴³ Besides, αPD-1-(iRGD)₂ is more convenient to purify and displays increased solubility and stability. On the other hand, the current clinical success of BiTEs was restrained to hematology malignancies, as the fibrous and dense microenvironment of solid tumors largely constrained the clinical benefit.⁴⁴ To solve this problem, we conjugated iRGD peptide, which has been proved to boost the penetration of therapeutic agents into tumor bulk via successively binding to integrins and NRP-1.¹⁹ In an open-label, multicenter, phase I clinical study, iRGD peptide in combination with chemotherapeutics preliminarily exhibited superior antitumor efficacy in metastatic pancreatic ductal adenocarcinoma.⁴⁵ Our team previously anchored iRGD to the surface of T cells to facilitate the infiltration of T cells into MCSs and tumor bulk.²⁰ We also designed anti-CD3-iRGD, a bifunctional agent that possesses dual affinity to CD3 and iRGD receptors and provides an extra activation signal to further strengthen the antitumor effect.²¹ The pre-existing T cell landscape is another crucial factor that may dampen the response to BiTEs in cancer patients.⁴⁶ Our work demonstrated that different from previous designs, intra-tumoral specific CD8⁺ T cells were sufficient to support the antitumor effect of αPD-1-(iRGD)₂.

Additionally, the dose of free iRGD (2.5 μg) in the αPD-1+iRGD group was equivalent to the absolute scale of iRGD conjugated to αPD-1 in this study. However, the αPD-1 plus iRGD group did not exhibit improvement in antitumor efficacy over the αPD-1 group. Different from our observation, Sugahara et al. disclosed that Nab-paclitaxel (ABX) conjugated with iRGD peptide had comparable antitumor efficacy to ABX combined with free iRGD when iRGD was administered at the dose of 4 μmol/kg (about 7.58 μg for mouse) every other day.¹⁹ A possible explanation for our results may be that the half-life of free iRGD was limited to minutes, which requires frequent administration. Meanwhile, free iRGD and αPD-1 combination could not mediate the engagement of T cells and tumor cells. In our work, dose retrenchment did not impede the antitumor effect of αPD-1-(iRGD)₂, PD-1 expression on CD8 T cells was not restrained, and the tumor-specific T cell subset (CD39⁺PD-1⁺) exhibited significant expansion. Therefore, we speculate that αPD-1-(iRGD)₂ causes gathering between tumor cells and T cells to facilitate TCR-pMHC interaction and promotes tumor-specific T cell activation and proliferation. iRGD conjugation generates a BiTE-like antitumor mechanism, remarkably reduces the required dose of iRGD, and allows for longer injection intervals, thus improving treatment compliance and feasibility.

Last but not least, αPD-1-(iRGD)₂ reinforces CD8⁺ T cells and expands a population of “better effector” PD-1⁺TCF-1⁺CD8⁺ T cells that have emerged as important determinants of the immune response in chronic infections and cancer, and their abundance is critical to the success of cancer immunotherapies.^47,48 In our work, the abundance of a subset of T cells, expressing stem- and memory-associated genes, intermediate effector genes, and low levels of exhaustion markers, increased after the administration of αPD-1-(iRGD)₂. According to previous studies, we defined them as “better effectors.”³⁴ We suppose that the dual binding affinity of αPD-1-(iRGD)₂ maintains stem and memory characteristics of CD8⁺ TILs. The receptors of iRGD, including integrin αv and NRP1, have been reported to regulate antitumor CD8⁺ T cell immunity and response to PD-1 blockade.⁴⁹ Further exploration is required to clarify the modulation of integrin αV and NRP1 by iRGD.

The abundance of tumor-specific TILs along with the expression of iRGD receptors may be potential markers for the application of αPD1-(iRGD)₂. However, the application of αPD1-(iRGD)₂ in the B16F10 and 4T1 mouse tumor models maintained its antitumor efficacy, although B16F10 and 4T1 were defined as “cold tumors” with poor immune infiltration.^50,51 Meanwhile, various non-malignant conditions such as inflammation and wound healing cause upregulation of the same target molecules employed in tumor targeting.¹⁶ Further safety evaluation is also required before clinical translation. iRGD has been explored in a phase I clinical study as combination therapy with nab-paclitaxel and gemcitabine for the treatment of metastatic pancreatic ductal adenocarcinoma (NCT03517176).⁴⁵ The most common grade 3 or 4 events were neutropenia, anemia, leukopenia, and pulmonary embolism, which were similar to the nab-paclitaxel and gemcitabine group. Nevertheless, due to the limited number of patients, more extensive trials are imperative to further establish the safety profile of iRGD and its derivatives.

In the past decade, immunotherapy represented by PD-1/PD-L1 blockade has revolutionized the standard care for several types of tumors. However, underperforming drug uptake and poor immune infiltration largely limited the clinical benefit. In this work, we performed site-specific modification to both human and murine αPD-1 via our glycoengineering platform. We proved that αPD-1-(iRGD)₂ can enhance the therapeutic potential for cancer treatment by simultaneously engaging T cells and tumor cells, promoting tumor-specific T cell amplification, and expanding a population of “better effector” T cells. We believe that αPD-1-(iRGD)₂ represents a paradigm shift in the anti-PD-1 therapeutic approach, with the ability to remodel the TME via a distinct combinatorial mechanism, as well as setting up a clinical application scenario for the antibody conjugate.

Limitations of the study

The main limitation of this study is the imperfection of mechanism exploration for specific T cell expansion and antitumor efficacy confirmation in humanized tumor models that may be important for clinical translation. The contribution to antitumor effect induced by penetration enhancement and cell engagement needs to be further evaluated. Meanwhile, αPD-1-(iRGD)₂ should be benchmarked against other αPD-1-based bispecific antibodies and other immune cell engagers to provide more insights of antibody conjugate development.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
αPD-1-(iRGD)2 (CS1003)	This paper	N/A
αPD-1-(iRGD)2 (G4C2)	This paper	N/A
αPD-1-(iRGD)2-Cy5 (CS1003)	This paper	N/A
αPD-1-Cy5 (CS1003)	This paper	N/A
αPD-1 (CS1003)	(Li et al., 2021)15	N/A
αPD-1 (G4C2)	TopAlliance	N/A
Anti-human IgG-PE	Abcam	Cat# ab7005, RRID:AB_955489
Anti-human IgG-FITC	Abcam	Cat# ab6854, RRID:AB_955300
Anti-human CD3-APC	Biolegend	Cat# 300412, RRID:AB_314065
Anti-human CD3-PerCP	Biolegend	Cat# 981016, RRID:AB_2876777
Anti-human CD8-FITC	Biolegend	Cat# 980908, RRID:AB_2888883
Anti-human CD19-PE	Biolegend	Cat# 982402, RRID:AB_2616905
Anti-human HLA-A2-PE	Biolegend	Cat# 343305, RRID:AB_1877227
Anti-human PD-1-PE	Biolegend	Cat# 379209, RRID:AB_2922607
Anti-human CD69-PC7	Biolegend	Cat# 310912, RRID:AB_314847
Anti-human Integrin αvβ5-APC	Biolegend	Cat# 920011, RRID:AB_2894599
Anti-human NRP-1-PE	Biolegend	Cat# 354503, RRID:AB_11219194
Anti-human NRP-1	Biolegend	BioLegend Cat# 354502, RRID:AB_2564475
Anti-mouse GZMB-AF647	Biolegend	BioLegend Cat# 515406, RRID:AB_2294995
Anti-mouse IFNγ-APC	Biolegend	Cat# 505810, RRID:AB_315404
Anti-mouse IFNγ-PerCP5.5	Biolegend	Cat# 505822, RRID:AB_961361
Anti-mouse CD25-PE	eBioscience	Cat# 12-0251-83, RRID:AB_465608
Anti-mouse CD69-FITC	Biolegend	Cat# 104506, RRID:AB_313108
Anti-mouse CD3-FITC	Biolegend	Cat# 100204, RRID:AB_312660
Anti-mouse CD3e-PC7	Biolegend	Cat# 100320, RRID:AB_312685
Anti-mouse CD4-FITC	Biolegend	Cat# 100406, RRID:AB_312690
Anti-mouse CD4-PerCP5.5	Biolegend	Cat# 100434, RRID:AB_893330
Anti-mouse CD4-APC	Biolegend	Cat# 100412, RRID:AB_312696
Anti-mouse CD4-PE	Biolegend	Cat# 100408, RRID:AB_312692
Anti-mouse CD8-PE	Abcam	Cat# ab25498, RRID:AB_470588
Anti-mouse CD8-APC	Biolegend	Cat# 100712, RRID:AB_312751
Anti-mouse CD8-PerCP	Biolegend	Cat# 100734, RRID:AB_2075238
Anti-mouse CD8-FITC	Biolegend	Cat# 100706, RRID:AB_312745
Anti-mouse CD8	Bioxcell	Cat# BE0117, RRID:AB_10950145
Anti-mouse PD-1-PE	Biolegend	Cat# 135206, RRID:AB_1877232
Anti-mouse PD-1-FITC	Biolegend	Cat# 135213, RRID:AB_10689633
Anti-mouse CD137-PE	Biolegend	Cat# 106106, RRID:AB_2287565
Anti-mouse Ki67-PC7	Biolegend	Cat# 151217, RRID:AB_2910305
Anti-mouse CD39-PerCP	Biolegend	Cat# 143806, RRID:AB_2563393
Anti-mouse CD45-FITC	Biolegend	Cat# 103108, RRID:AB_312972
Anti-mouse Foxp3-AF647	Biolegend	Cat# 320014, RRID:AB_439750
Anti-mouse Integrin αvβ5	Biolegend	Cat# 153202, RRID:AB_2687270
Anti-mouse Integrin αvβ5-PE	Biolegend	Cat# 104106, RRID:AB_2129493
Anti-mouse NRP-1-PerCP5.5	Biolegend	Cat# 145208, RRID:AB_2562034
Anti-mouse PD-L1-PerCP5.5	Biolegend	Cat# 124334, RRID:AB_2629831
Anti-mouse CD11b-FITC	Biolegend	Cat# 101206, RRID:AB_312788
Anti-mouse CD11b-APC	Biolegend	Cat# 101212, RRID:AB_312795
Anti-mouse CD11c-FITC	Biolegend	Cat# 117306, RRID:AB_313775
Anti-mouse F4/80-FITC	Biolegend	Cat# 123108, RRID:AB_893502
Anti-mouse MHC II-APC	Biolegend	Cat# 107614, RRID:AB_313329
Anti-mouse CD83-PE	Biolegend	Cat# 121508, RRID:AB_572015
Anti-mouse CD103-PerCP	Biolegend	Cat# 121406, RRID:AB_1133989
Anti-mouse CD86-PerCP	Biolegend	Cat# 105028, RRID:AB_893420
Anti-mouse CD163-PE	Biolegend	Cat# 156704, RRID:AB_2860724
Anti-mouse Ly6G-PE	Biolegend	Cat# 127608, RRID:AB_1186104
Anti-mouse Ly6G-PC7	Biolegend	Cat# 127618, RRID:AB_1877262
Anti-mouse NRP-1	Biolegend	Cat# 145201, RRID:AB_2561840
Chemicals, peptides, and recombinant proteins
Recombinant Mouse PD-1 (HEK293, His)	MCE	HY-P73724
Recombinant human PD-1 (HEK293, His)	MCE	HY-P7396
Blinatumomab	MCE	HY-P9963
FTY720	MCE	HY-12005
Recombinant mouse IL7	Peprotech	#210-07
Recombinant mouse IL15	Peprotech	#210-15
Dynabeads for T cell activation/expansion ™ Mouse T activator CD3/CD28	Gibco	#11453D
iRGD, c(CRGDKGPDC)	Genescript Biotech Corporation	N/A
BCN-Cy5	Xi’an Ruixi Biological Technology Co., Ltd	N/A
N3-iRGD	Nanjing Yuan Peptide Biotech Ltd.	N/A
OVA257-264 peptide	Genscript Biotech Corporation	N/A
GP33 peptide	Genscript Biotech Corporation	N/A
NY-ESO-1157-165	Genscript Biotech Corporation	N/A
collagenase IV	Sigma-Aldrich	C4-BIOC
DNase I	Sigma-Aldrich	04536282001
CFSE	eBioscience	#65-0850-84
Cell Proliferation Dye eFluor™ 670	eBioscience	#65-0840-85
Cell Proliferation Dye eFluor™ 450	eBioscience	#65-0842-90
CellTracker™ Blue	Invitrogen	#C12881
Deep Red cell tracer	Invitrogen	#C37608
Zombie Aqua™ Fixable Viability Kit	Biolegend	#423101
Critical commercial assays
Cytofix/Cytoperm™ Fixation/Permeabilization Kit	BD Biosciences	#554714
Foxp3/transcription factor flow cytometry fixed membrane breaking buffer	EBioscience	#00-5523-00
GEXSCOPE Single Cell RNA-seq Kit	Singleron Biotechnologies	N/A
Deposited data
Mouse CD45+ cells single cell RNA-seq data	This paper	CNGB database: CNP0005642. DOI: 10.26036/CNP0005642
Experimental models: Cell lines
Cancer cell line: B16F10	Cell Bank of Shanghai	N/A
Cancer cell line: B16-OVA	Cell Bank of Shanghai	N/A
Cancer cell line: 4T1	Cell Bank of Shanghai	N/A
Cancer cell line: MFC	Cell Bank of Shanghai	N/A
Cancer cell line: HGC27	Cell Bank of Shanghai	N/A
Cancer cell line: Jurkat	Cell Bank of Shanghai	N/A
Cancer cell line: Raji	Cell Bank of Shanghai	N/A
HBE	Cell Bank of Shanghai	N/A
GES-1	Cell Bank of Shanghai	N/A
293T	Cell Bank of Shanghai	N/A
Experimental models: Organisms/strains
Mouse:C57BL/6J	Gem Pharmatech Co., Ltd.	RRID: IMSR_JAX:000664
Mouse: OT-I, C57BL/6J	Gem Pharmatech Co., Ltd.	N/A
Mouse: Balb/C	Gem Pharmatech Co., Ltd.	N/A
Mouse: 615	the Institute of Hematology, Chinese Academy of Medical Sciences	N/A
Software and algorithms
R 3.5.0	The R Foundation	http://www.r-project.org/
GraphPad Prism Version 7	Prism	N/A
FlowJo Software Version 10.6.2	FlowJo, LLC	N/A

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Jia Wei (jiawei99@nju.edu.cn).

Materials availability

Correspondence and requests for materials should be addressed to Jia Wei and Jie P. Li. All biological materials can be obtained from the corresponding authors following reasonable request.

Data and code availability

Raw and processed data from RNA-seg analysis of mouse tumor infiltrating CD45⁺ cells have been submitted to CNGB database (CNP0005642). This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and study participant details

Cell lines

All of the cells mentioned above were cultured in Roswell Park Memorial Institute (RPMI) 1640 (Corning) supplemented with 10% fetal bovine serum (FBS, Gibco), 1% penicillin/streptomycin solution 100X (Beyotime, penicillin 10 kU/ml, and streptomycin10 mg/ml) at 37°C and 5% CO₂. Cells were regularly tested for Mycoplasma with PCR method. The most recent cell line authentication was in September 2020 by short tandem repeat (STR) analysis.

Mice

All animals in our study were raised in the pathogen-free animal facilities at Nanjing University Medical School Affiliated Drum Tower Hospital (Nanjing, China). All animal experiments were approved by the Institutional Animal Care and Use Committee of Drum Tower Hospital (approval number: 2020AE01064). 615-line mice were purchased from the Institute of Hematology, Chinese Academy of Medical Sciences (Tianjin, China). C57BL/6 and OT-I mice were purchased from Gem Pharmatech Co., Ltd. (Nanjing, China). Animals of both sexes were used between the ages of 6–8 weeks at the beginning of the experiment, randomized, and assigned to experimental groups.

Method details

Synthesis of GDP-FAmP4PropP4-iRGD

To a solution of 500 μL GDP-FAm (100 mM) was added 1500 μL ddH₂O, 500 μL NaHCO₃ (200 mM), 1750 μL THF and 750 μL NHS-PEG₄-Prop (Confluore) (50 mM in tetrahydrofuran (THF)). The reaction was stirred at room temperature (r.t.) for 4 h and monitored by thin-layer chromatography (TLC). The solvent was removed under reduced pressure. The crude product was further purified through a Prep-HPLC system to give the desired product as a white solid (27.1 mg. 64%).

To a solution of 640 μL GDP-FAmP₄Prop (50 mM) in ddH₂O/DMSO (2048 μL/1920 μL), 64 μL CuSO₄ (50 mM in ddH₂O), 128 μL BTTP (50 mM in ddH₂O), 1280 μL N₃-iRGD (Nanjing Yuan Peptide Biotech Ltd.) (50 mM in DMSO), and 320 μL ascorbate (50 mM in ddH₂O) were added. The reaction was allowed for stirring at r.t. for 12 h and monitored by TLC. The crude product was further purified through a Prep-HPLC system to give the product a white solid (38.4 mg, 64%). HRMS (ESI-) calculated for C₆₃H₁₀₃N₂₃O₃₄P₂S₂ (M-2H⁺)/2 936.7905, found 936.7932.

HPLC system using a waters XBridge Prep C18 5um OBDTM column (19 × 150 mm). CH₃CN with 0.1% NH₃·H₂O (solvent A) and H₂O with 0.1% NH₃·H₂O (solvent B), were used as the mobile phase at a flow rate of 15 mL/min. The gradient was programmed as follows:100% B for 5 min, then a gradient to 10% A over 5 min, then a gradient to 95% A over 14 min, then 95% A for 2 min, then a gradient to 100% B over 2 min, then 100% B for 2 min.

Cloning, expression and purification of bovine β1,4-GalT1(Y289L), EndoS(Streptococcus pyogenes endoglycosidase S) and Alfc (Lactobacillus casei α-1,6-fucosidase)

The cloning, expression and purification of bovine β1,4-GalT1(Y289L), EndoS and AlfC were performed according to the reported procedure by Qasba, P. K et al., by Wang L. et al. and by Wu P. respectively.^52,53,54

Cloning, expression and purification of Helicobacter pylori α1,3 fucosyltransferase

Genes encoding the Helicobacter pylori α1,3 fucosyltransferase were synthesized and subcloned into a pET24b vector at NdeI and BamHI by Genscript. E. coli BL21(DE3) transformed with the plasmids were cultured at 37°C in LB with 50 μg/mL kanamycin until OD600 = 0.6–0.8. IPTG was added to a final concentration of 0.2mM and protein expression was induced for 16 h at 25°C. The cells were harvested by centrifugation and resuspended in lysis buffer (25 mM Tris pH 7.5, 500 mM NaCl, 20 mM imidazole, and 1 mM PMSF). Cells were lysed by sonication and the clarified supernatant was purified on Ni-NTA agarose (GE Health) following the manufacturer’s instructions. Fractions that were >90% purity, as judged by SDS-PAGE, were consolidated and dialyzed against Tris-buffered saline (25 mM Tris pH 7.5, 150 mM NaCl).

Cloning, expression and purification of αPD-1

The Fab sequence of αPD-1 antibody light chain and heavy chain were referenced to the patent (CN202210202373.1). The gene encoding the light chain and the heavy chain of αPD-1 were synthesized and cloned into a PPT5 vector respectively by Genescript. Then, FreeStyle 293F cells were grown to a density of ∼2.5×10⁶ cells/ml and transfected by direct addition of 0.37 μg/mL and 0.66 μg/mL of the light chain and heavy chain expression plasmid DNA, and 2.2 μg/mL polyethylenimine (linear 25 kDa PEI, Polysciences, Inc, Warrington, PA) to the suspension cultures. The cultures were diluted 1:1 with Freestyle 293 expression medium containing 4.4 mM valproic acid (2.2 mM final) 24 h after transfection, and protein production was continued for another 4–5 days at 37°C. After protein production, the antibodies were purified through the protein A agarose following the manufacturer’s instructions.

“One-pot” synthesis of αPD-1-(Galβ1,4) GlcNAc-FAmP4PropP4-iRGD

αPD-1 (8 mg/mL) was incubated with EndoS (0.05 mg/mL) and Alfc (1.5 mg/mL) in 50 mM Tris-HCl buffer (pH 7.5) at 30°C. After 24 h, the reaction mixtures were added with UDP-Gal (final concentration 5 mM), bovine β1,4-GalT1(Y289L) (final concentration 0.3 mg/mL), GDP-FAmP4PropP4-iRGD (final concentration 2 mM) and Hp1,3-FucT (final concentration 0.5 mg/mL), MgCl₂ (final concentration 10 mM) and MnCl₂ (final concentration 5 mM) followed by incubating at 30°C for 48 h. The modified antibody was purified with protein A resin to give the αPD-1-(Galβ1,4) GlcNAc-FAmP4PropP4-iRGD (calculated as 148833.00 Da, found as 148831.00 Da, MAR2) conjugates.

Synthesis of αPD-1-(Fucα1,6)-(GalNAzβ1,4)GlcNAc-FAm-iRGD

The αPD-1 antibody (15 mg/mL) was incubated with EndoS (0.1 mg/mL) in 50 mM Tris-HCl buffer (pH 7.5) at 37°C. After 1 h, UDP-GalNAz (final concentration 10 mM, Shanghai Bosen Biological Technology Co., Ltd., China), bovine β1,4-GalT1 (Y289L) (final concentration 0.5 mg/mL), GDP-FAm-iRGD (final concentration 5 mM), and Hp1,3-FucT (final concentration 1 mg/mL) were added to the reaction mixture along with MgCl₂ (final concentration 10 mM) and MnCl₂ (final concentration 5 mM). The incubation continued at 37°C for 24–48 h. Modified antibodies were purified using protein A resin to obtain the PD-1 antibody-iRGD conjugate, represented as αPD-1-(GalNAzβ1,4)GlcNAc-FAm-iRGD.

Synthesis of αPD-1-(Fucα1,6)-(GalNAzβ1,4BCN-Cy5)GlcNAc and αPD-1-(Fucα1,6)-(GalNAzβ1,4BCN-Cy5)GlcNAc-FAm-iRGD

αPD-1-(Fucα1,6)-(GalNAzβ1,4)GlcNAc (10 mg/mL) and αPD-1-(Fucα1,6)-(GalNAzβ1,4)GlcNAc-FAm-iRGD (10 mg/mL) were separately incubated with BCN-Cy5 (3–10 antibody equivalents) (Xi’an Ruixi Biological Technology Co., Ltd., China) in PBS buffer (pH 7.0) containing 10% DMF for 2–10 h. The products were purified through desalting columns to obtain αPD-1-(Fucα1,6)-(GalNAzβ1,4BCN-Cy5)GlcNAc (abbreviated as αPD-1-Cy5) and αPD-1-(Fucα1,6)-(GalNAzβ1,4BCN-Cy5)GlcNAc-FAm-iRGD (abbreviated as αPD-1-(iRGD)₂-Cy5).

Binding affinity assay

Recombinant PD-1 extracellular domains (PD-1, novoprotein) were diluted to a final concentration of 250 ng/mL with coating buffer and plated on 96-well plates (100 μL/well) at 4 °C overnight. After removing the coating solution, the plates were blocked with 3% (v/v) bovine serum albumin in PBS for 2 h at 37°C. After washing with PBST (PBS containing 0.03% tween 20) for 3 times, αPD-1 and αPD-1-(Galβ1,4)GlcNAc-FAmP4PropP4-iRGD were added to PBST (with 1% (v/v) bovine serum albumin in PBS) to a series of final concentrations (3000 ng/mL, 1000 ng/mL, 333.33 ng/mL, 111.11 ng/mL, 37.04 ng/mL, 12.35 ng/mL, 4.12 ng/mL, 1.37 ng/mL, 0.46 ng/mL, 0.15 ng/mL, 0.05 ng/mL, 0 ng/mL) and added to the plates respectively. After incubating for 1.5 h, the plates were washed 3 times with PBST, then horseradish peroxidase (HRP)-conjugated goat anti-human IgG antibody was added to each well and incubated for 1 h at 37°C. Finally, each well was washed with PBST 3 times, and then tetramethyl benzidine substrate was cotreated to produce color for visualization. The reaction in each well was stopped by adding 100 μL of 3 M HCl after 15 min of incubation. The absorbance was read at 450 nm on a Synergy LX plate reader.

Intact protein mass analysis

For LC-MS analysis, the purified proteins were analyzed on an Xevo G2-XS QTOF MS System (Waters Corporation) equipped with an electrospray ionization (ESI) source in conjunction with Waters Acuqity UPLC I-Class plus. Separation and desalting were carried out on a waters ACQUITY UPLC Protein BEH C4 Column (300 Å, 1.7 μm, 2.1 mm × 100 mm). Mobile phase A was 0.1% formic acid in water and mobile phase B was acetonitrile with 0.1% formic acid. A constant flow rate of 0.2 mL/min was used. Data were analyzed using Waters Unify software. Mass spectral deconvolution was performed using a Unify software (version 1.9.4, Waters Corporation).

Generation of stable cell lines expressing PD-1, 1G4 and HLA-A∗0201

HLA-A∗0201-Raji, Jurkat PD-1 and IG4 Jurkat PD-1 were derived from Raji, Jurkat and IG4 Jurkat, respectively, after lentiviral transduction. The PD-1 and HLA-A∗0201 genes were cloned into the FG12 cloning vector (14884, Addgene). Lentivirus containing supernatant was prepared by transfecting target plasmid, PMDLg/pRRE, pRSV-Rev and pMD2.G per 10cm dish of 70%–80% confluent HEK293T cells. Transfection was performed by incubating the plasmid mix in 1 mL Opti-MEM containing 0.05 mL polyethyleneimine for 15 min at room temperature before adding dropwise to the cells. The medium was replaced 8 h after transfection. The viral supernatant was collected after 48 h, briefly centrifuged at 1000× g and filtered through a 0.45 μm syringe. Target cells were plated in 24-well plates at 1.5 × 10⁵ cells per well and spin occulated with 0.5 mL lentiviral supernatant at 1000 × g for 90 min at room temperature, then incubated at 37°C. After 24 h, the cells were collected and harvested.

Stability studies in human serum

In an eppendorf tube, 200 μL human serum was mixed with 100 μL αPD-1 (1 mg/mL) or αPD-1-iRGD for each sample individually to give a final solution of 0.3 mg/mL ADC in human serum. Samples were incubated in human serum at 37°C for 3 and 7 days. Day 0 samples were processed directly. The antibody was purified using Protein A resin and confirmed by UPLC-TOF/MS analysis.

Pharmacokinetic studies

After intravenous administration into the tail vein of MFC mice, 40–50 μL each blood samples were collected from the eye veins (capillary blood collection) at 0h, 1h, 2h, 6h, 24h, 48h, 168h time points into anticoagulant tubes and stored on ice. After collection, the samples were immediately centrifuged to separate the plasma (3000–5000 rpm, 10 min, 4°C), and the antibody levels in the plasma were detected according to the following procedure: 1. PD-1 protein was diluted to 0.25 μg/mL in carbonate buffer and coated onto a 96-well enzyme-linked immunosorbent assay (ELISA) plate at 100 μL per well, then incubated overnight (16–24 h) at 2°C–8°C. The liquid in the wells was discarded and each well was washed with 200 μL of 0.3% PBST (PBS containing Tween 20) and then blotted dry. 2. Blocking solution (3% BSA-PBST) was added to each well at 200 μL per well and incubated at 30°C for 1.5–2 h. 3. The liquid in the wells was discarded and each well was washed three times with 200 μL of 0.3% PBST (500 mL PBS +150 μL Tween 20) and then blotted dry. 4. Mouse plasma samples were diluted in PBST +0.5% BSA to a final concentration range of 2–10 ng/mL and added to each well at 100 μL per well. The plate was incubated for 1 h at 30°C. 5. After three washes with PBST, each well was incubated with horseradish peroxidase (HRP)-conjugated goat anti-human IgG antibody (diluted 1:2000 in PBST +0.5% BSA) at 30°C for 1 h. 6. Finally, each well was washed three times with PBST and 100 μL of TMB substrate solution (TMB:H₂O₂ = 1:1) was added for color development. The reaction was stopped by adding 100 μL of 2 M HCl per well after 5 min of incubation. Absorbance was measured at 450 nm using a SynergyTM LX microplate reader.

OT-I cells acquisition and culture

Spleens of OT-I mice were resected and mechanically dissociated. Single cell suspensions were obtained through 40-μm nylon cell strainers (Biosharp). Splenocytes from OT-I mice were cultured at about 1 × 10⁶ cells/ml in 24 well-plates with RPMI (Thermo Fisher) along with 10% FBS (Thermo Fisher), 1% penicillin/streptomycin (Thermo Fisher), 2mM L-glutamin (Thermo Fisher), 50μM β-mecaptoethanol (Thermo Fisher), 10mM HEPES (Thermo Fisher) and IL-2 (100U/ml, Thermo Fisher) in the presence of OVA_257-264 peptide (0.1nM) at 37°C and 5% CO₂. 3 days later, activated cells were washed 3 times with RPMI 1640 and resuspended in T25 flasks at 1 × 10⁵ cell/ml in the presence of IL-2(100U/ml), IL-7, IL-15(10 ng/ml, Thermo Fisher).

Penetrability analysis of αPD-1-(iRGD)₂ in MCSs

MCSs were generated with HGC27 cells as previously described.²¹ Subcircular MCSs with a diameter of around 500μm were selected for this study. αPD-1-(iRGD)₂-Cy5 (10 μg/ml) or αPD-1(10 μg/ml) along or αPD-1(10 μg/ml) with free iRGD(10 μg/ml in L-iRGD, 100 μg/ml in H-iRGD) were cocultured with MCSs for 24 h at 37°C. In the cell penetrating assay, 1×10⁶ PD-1-Jurkat cells were labeled with CFSE (Abcam, Cambridge, UK). After labeling, PD-1-Jurkat cells along with αPD-1-(iRGD)₂ (10 μg/ml) or αPD-1(10 μg/ml) along or αPD-1(10 μg/ml) with free iRGD(100 μg/ml), were cocultured with MCSs for 24 h at 37°C. Then, the MCSs were washed with PBS and fixed with 4% paraformaldehyde and imaged using a ZEN 710 confocal microscope (Zeiss, Jena, Germany). Images were acquired close to the mid-height of the spheroids. Fluorescence intensity was calculated with Leica Application Suite X (LAS X).

Tumor-specific T cell activation assay

B16F10 cells were digested with pancreatic enzymes and made into single cell suspension at 1∗10⁷/mL with 100nM SIINFEKL peptide (Genscript Biotech Corporation) for 30min. After peptide loading, 2∗10⁴ B16F10 cells were cultured with 1∗10⁵ CellTracker Blue (Invitrogen, #C12881) labeled OT-I cells and 1∗10⁵ non-specific spleen cells from C57BL6J mice. 1 μg/well αPD-1 or αPD-1-(iRGD)₂ was added. After 4h incubation, cells were collected and stained with anti-mouse CD69-FITC (Biolegend, # 104506) and anti-mouse CD25-PE (ebioscience, # 12-0251-83) before flow cytometry analysis.

T cell-specific cytotoxicity assay

B16F10 cells were digested with pancreatic enzymes and made into single cell suspension at 1∗10⁷/mL with 100nM SIINFEKL peptide (Genscript Biotech Corporation) or GP33 peptide (Genscript Biotech Corporation) for 30min. After peptide loading, SIINFEKL-treated B16F10 cells were labeled with CFSE, GP33 treated B16F10 cells were labeled with Dye eFluor 670 (Thermo Fisher Scientific). 1∗10⁴ untreated B16F10 cells, 1∗10⁴ SIINFEKL treated B16F10 cells and 1∗10⁴ GP33 treated B16F10 cells were cultured with 2∗10⁵ CellTracker Blue (Invitrogen, #C12881) labeled OT-I cells. 1 μg/well αPD-1 or αPD-1-(iRGD)₂ was added. After 12h incubation, cells were collected and stained with 100 ng/ml Propidium iodide (PI, beyotime) before flow cytometry analysis.

In vivo real time near-infrared fluorescence imaging of Cy5 labeled αPD-1-(iRGD)₂

Near-infrared live body imaging was used to trace the distribution of αPD-1-(iRGD)₂-Cy5 or αPD-1. 50μg αPD-1-(iRGD)₂-Cy5, or αPD-1-Cy5 along or with 1μg free iRGD were injected i.p. into MFC tumor-bearing mice with an average tumor volume of 200mm³. After anesthesia, mice were scanned with CRi Maestro In Vivo Imaging System (Cambridge Research & Instrumentation, Massachusetts, USA) at different time points after injection. 48 h after the agents’ administration, mice were humanely sacrificed and tumor tissues and main organs were resected. The In vitro average radiant efficiency of tumor and main organ tissues was also scanned with CRi Maestro In Vivo Imaging System (Cambridge Research & Instrumentation, Massachusetts, USA).

Cytotoxicity assay

HLA-A∗2402⁺ PBMCs from healthy donors were collected via Ficoll density centrifugation. PBMCS were cultured in AIM-V medium (Gibco, USA) along with 10% FBS at 37°C and 5% CO₂. 2 h later, non-adherent T cells were collected and suspended in AIM-V medium (Gibco, USA) along with 10% FBS, 100IU interleukin-2 (IL2) (Peprotech, USA), 10 ng/ml IL7 (Peprotech, USA). and IL15 (PeproTech, USA). Then, 2×10⁵ amplified T cells along with αPD-1-(iRGD)₂ or αPD-1(1, 5, 10, 25, 50 μg/ml) along or with free iRGD (100 μg/ml) were co-cultured with 2×10⁴ CFSE labeled HGC27 cells for 24 h. In competitive inhibition assay, the concentration of αPD-1-(iRGD)₂ or αPD-1 was 10 μg/ml, free iRGD was 100 μg/ml, αNRP1 was 15 μg/mL, extra αPD-1 was 50 μg/mL. To induce exhaustion of T cells, amplified T cells were cultured in Dynabeads Human T-Expander CD3/CD28(Gibco), 500IU interleukin-2 (IL2) (Peprotech, USA), 10 ng/ml IL7 (Peprotech, USA). and IL15 (PeproTech, USA). After the incubation, 100 ng/ml Propidium Iodide (PI, beyotime) was added to cultural media. Tumor cell cytotoxicity was measured using flow cytometry. CFSE and PI double-positive cells were considered to be lysed tumor cells. The percentage of cytotoxic activity was calculated with the following equation:

$$\mathbf{\text{Lysis}}(\%)=\frac{{\text{CFSE}}^{+}{\text{PI}}^{+}\mathbf{\text{cells}}-\mathbf{\text{spontanious}}\mathbf{\text{lysis}}}{{\text{CFSE}}^{+}\mathbf{\text{cells}}-\mathbf{\text{spontanious}}\mathbf{\text{lysis}}}\times \mathbf{100}\%$$

T cell activation assay

2 × 10⁵ T cells or OT-I cells were cocultured with 2×10⁴ HGC27 cells or B16-OVA for 24h in 96 well ultra-low adsorption culture plate. The concentration of αPD-1-(iRGD)₂ or αPD-1 was 10 μg/ml, free iRGD was 100 μg/ml, αNRP1 was 15 μg/mL, extra αPD-1 was 50 μg/mL. After the incubation, the plate was centrifuged and the supernatant was discarded. 100 μL PBS was added into every well to form single cell suspensions. The staining and flow cytometry process were described in “Flow cytometry analysis” part.

Cell conjugates formation assay

1G4-PD-1-Jurkat cells or OT-I cells were labeled with 2μM CFSE (Abcam, Cambridge, UK). PD-1-Jurkat cells were stained with 10μM Dye 450 (eBioscience Cell Proliferation Dye eFluor 450). HGC27 cells and NY-ESO-1_157-165primed-HLA-A∗0201-Raji cells were labeled with Dye eFluor 670 (Thermo Fisher Scientific). In fluorescence imaging, PBMCs were labeled with CellTracker Deep Red Dye (Invitrogen). HGC27 cells were labeled with CFSE. After being washed 3 times with FACS buffer (DPBS containing 2% FBS), 2×10⁵ PD-1-Jurkat cells or OT-I cells or PBMCs were cocultured with 2×10⁴ HGC27 cells. αPD-1-(iRGD)₂ (10 μg/ml) or αPD-1(10 μg/ml) along or αPD-1(10 μg/ml) with free iRGD(100μg/ml) or Blinatumomab (1 μg/ml) was added into cultural media. After 1 h of incubation at 37°C, flow cytometry analysis was performed directly using BD Accuri C6 PLUS Flow Cytometry (BD Biosciences). All raw data were analyzed using FlowJo software (10.4, Tree Star). For fluorescence imaging, 1×10⁶ PBMC were cocultured with 1×10⁶ HGC27 cells with 10 μg/mL αPD-1-(iRGD)₂ or αPD-1 or indicated agents. If added, the concentration of free iRGD was 100 μg/ml, αNRP1 was 15 μg/mL, extra αPD-1 was 50 μg/mL. After 12h, fluorescence figures were taken with ZEN 710 confocal microscope (Zeiss, Jena, Germany).

Construction and treatment of mouse tumor models

For in vivo tumor suppression assay, 1 × 10⁶ MFC or 1 × 10⁵ B16F10 cells were subcutaneously inoculated into 6–8-week-old sex-matched 615-line mice or C57BL/6 mice (n = 6 or 5 per group). For 4T1 orthotopic model, 4 × 10⁵ 4T1 cells were injected at right lower breast fat pad. All mice were checked every day. The length of the long axis (a) and vertical axis (b) of the tumor nodule were measured with calipers. Tumor volume was calculated via the following formula:

$$\mathbf{\text{Tumor}}\mathbf{\text{volume}}\left({\mathit{m}\mathit{m}}^{\mathbf{3}}\right)=\frac{\mathit{a}\times {\mathit{b}}^{\mathbf{2}}}{\mathbf{2}}$$

Endpoints of animal experiments and timepoints of tissue collection (tumor and lymph nodes) were shown in figure annotations. In order to obtain single cell suspensions, tumor nodules were shredded and digested with 1 mg/ml collagenase IV (Sigma-Aldrich) in serum free RMPI-1640 for 2h at 37°C. After the incubation, cells were filtered via 40 μm nylon cell strainers (Biosharp), and washed twice with PBS. Then red blood cell lysis buffer (Biosharp) was applied for erythrocyte clearance.

After tumor inoculation, mice were randomized to treatment groups. In tumor suppression assays, mice were treated intraperitoneally (i.p.) with αPD-1-(iRGD)₂ (5 mg/kg), αPD-1 (5 mg/kg) alone or with free iRGD (2.5μg), free iRGD peptide (2.5μg), 4 times over 12 days.

Histology analysis

Mouse main organ tissues (heart, liver, spleen, lung, kidney, Stomach) were prepared as 5 μm formalin-fixed, paraffin-embedded (FFPE) sections. Paraffin-embedded sections were dewaxed in water, sequentially immersed in xylene I for 10 min, xylene II for 10 min, absolute ethanol I for 5 min, absolute ethanol II for 5 min, 95% ethanol for 5 min, 95% ethanol for another 5 min, 80% ethanol for 5 min, and rinsed in tap water. Nuclei were stained with hematoxylin. Sections were immersed in hematoxylin staining solution for 3–8 min, followed by rinsing in tap water. Differentiation was achieved by briefly immersing the sections in 1% hydrochloric acid alcohol, followed by rinsing in tap water. Counterstaining was performed using 0.6% ammonia water for bluing, followed by rinsing under running water. Sections were stained in eosin solution for 1–3 min. Sections were dehydrated sequentially in 95% ethanol I for 5 min, 95% ethanol II for 5 min, absolute ethanol I for 5 min, absolute ethanol II for 5 min, xylene I for 5 min, and xylene II for 5 min. After dehydration, sections were briefly air-dried and mounted with neutral gum. Hematoxylin-eosin staining was done in mouse organ tissue sections, and treatment toxicity was analyzed.

Immunofluorescence staining

Subcutaneous tumors from MFC were processed into 5 μm formalin-fixed, paraffin-embedded (FFPE) sections. These sections were placed on a slide rack and baked in a 65°C oven for 1 to 1.5 h. Dewaxing involved immersing the sections in Xylene I for 10 min, Xylene II for another 10 min, and Xylene III for 5 min, followed by absolute ethanol I and II, each for 5 min, then 95% ethanol for 5 min, and finally 85% ethanol for another 5 min. Afterward, the sections were rinsed thoroughly with tap water and immersed in distilled water for five changes. Antigen retrieval was carried out by immersing the sections in boiling EDTA pH 9.0 solution and heating in a pressure cooker for 3 min after reaching full pressure. The sections were then allowed to cool naturally to room temperature before being rinsed with distilled water five times. To mark a circular boundary, the sections were dried by wiping off any residual water, then placed in a humid chamber with 5% BSA to cover the samples completely, and incubated at room temperature for 60 min. After incubation, the BSA was wiped away, and the sections were washed with PBST three times. Subsequently, 500× diluted anti-human IgG FITC (Abcam, #ab6854) and anti-murine CD8 PE (Abcam, #ab25498) were added, and the sections were incubated at room temperature in the dark for 1 h. Following this, the sections were washed with PBST for 5 min three times. The nuclei were then stained with DAPI, and the images were acquired using a fluorescent microscope (Leica).

Flow cytometry analysis

As previously described, single cell suspensions were prepared from tumor tissues and draining lymph nodes. For cell-surface markers detection, single-cell suspensions were stained with antibodies specific to markers including CD3, CD4, CD8, CD25, CD69, CD137, CD11b, CD11c, MHC II, CD39, NK1.1, F4/80, Ly6G, CD83, CD103, CD163, CD80, human IgG, PD-1, CD51, and NRP-1 for 30 min at 4°C. For intercellular markers, single cell suspensions were treated with fixation/Permeabilization Solution Kit (BD Biosciences) before incubating with antibodies targeting IFN-γ, GZMB, Ki67 and FOXP3. Stained cells were examined using BD Accuri C6 PLUS Flow Cytometry (BD Biosciences). All raw data were analyzed using FlowJo software (10.4, Tree Star). For absolute cell count, Precision Count Beads (Biolegend, #424902) were used and calculated according to the manufacturer’s protocol.

Cell sorting

Single cell suspensions were acquired as previously described and purified by a 67% Percoll gradient (800 g at 20 °C for 20 min) to enrich lymphocytes. Then, cell suspensions were stained with Zombie Aqu Fixable Viability Kit (#423101, Biolegend) and anti-CD45 FITC (#157607, Biolegend) for 30 min at 4°C. CD45⁺ Cell sorting was performed using the FACS Aria II (BD Biosciences) system.

Cell barcoding, multiplexing and scRNA-seq

Generally, one sample was barcoded with ClickTags by the following procedure: for each ClickTag preparation, 0.5 μL of NHS-TCO (1 mM in DMSO) and 20 μL of 25 μM Tz-oligo was thoroughly mixed and immediately pipetted into the sample. After 15 min of cell barcoding in the dark at room temperature on a rotating platform, the process was quenched by 10μL of quenching buffer (300 μM alkyne-Tz in FBS) for 5 min.⁵⁵ Then, cells were washed three times with DPBS and detected cell number and viability, before being pooled and loaded to a microwell chip targeting 15,000 cells on Singleron Matrix (GEXSCOPE Single Cell RNA-seq Kit, Singleron Biotechnologies, Nanjing, China). Libraries of scRNA-seq were prepared according to the manufacturer’s instructions (Singleron Biotechnologies, Nanjing, China). After amplification, cDNA and ClickTags were respectively separated by SPRI size-selection with 0.6× and 1.4× SPRI. ClickTag libraries were subsequently quantified (Qubit, Invitrogen) and amplified using primer SGR-beads-1/SGR-tag-1 and indexed by additional PCR with primer SGR-beads-2/SGR-tag-2. Final ClickTag libraries and transcriptome libraries were analyzed on a BioAnalyzer high-sensitivity DNA kit (Agilent) and sequenced on Illumina NovaSeq 6000. Specially, two-dimensional barcoding was enabled by mixing two different Tz-oligos (10 μL, 50 μM for each ClickTag).

Analysis of scRNA-seq data

Samples were demultiplexed and aligned using celescope to obtain a raw read count begin of barcodes corresponding to cells and features corresponding to detected genes. Read count matrices were processed, analyzed, and visualized in R (R Core Team, 2013) using Seurat v.4.

Quantification and statistical analysis

GraphPad Prism 7 software was used for statistical analysis of data and graphic representations. Although no statistical method was used to predetermine sample size, mouse numbers were taken into consideration for in vivo studies to ensure that biological effects would be detected, and to enable comparison between groups, and were determined based on results of preliminary experiments. For gene enrichment analysis, hypergeometric distribution test was used as indicated. For each dataset, the normality of the population and/or population residuals (Gaussian distribution) was confirmed using Shapiro–Wilk and/or D’Agostino–Pearson testing. For normal distributions, one-way ANOVA with Tukey’s multiple comparisons test or two-way ANOVA with Tukey’s multiple comparisons test was used to compare all groups with three or more treatments. When data were not normally distributed, a nonparametric test was used, either Kruskal–Wallis with Dunn’s post hoc test for multiple comparisons or Mann–Whitney test when two groups were compared. The error bars of data were presented as the means ± SEM or means ± SD as indicated. The p value of less than 0.05 was considered to be statistically significant. ns, not significant; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 and ∗∗∗∗p < 0.0001.

Published: June 5, 2024