XCL1/Glypican-3 Fusion Gene Immunization Generates Potent Antitumor Cellular Immunity and Enhances Anti–PD-1 Efficacy

Over the past decade, several clinically effective approaches of antitumor immunotherapy have been developed. The effectiveness of distinct immune-based strategies demonstrates the promise of profound anticancer effects when active T cells recognize their cognate antigens, which are expressed in cancer cells (1, 2). Clinically, CD8⁺ T-cell tumor infiltration predicts patient responsiveness to checkpoint blockade, which impedes inhibitory molecules on activated T cells and renew their anticancer function (2). Although massive T-cell infiltrations have been observed in patients with hepatocellular carcinoma (HCC), the clinical benefit of immune checkpoint blockade is obtained only in a subset of patients (3, 4). This limited antitumor effect may be due to the lack of functional tumor-specific T cells in patients with HCC (2, 5–7).

Cancer vaccines amplify existing antitumor responses or prime naïve T cells to elicit effector T-cell function in patients through immunization, which might improve immune checkpoint blockade (8). Antigen-specific CD8⁺ cytotoxic T lymphocytes (CTL) play critical roles against cancers (9). The uptake, processing, and cross-presentation of extracellular tumor antigens with MHC-I by professional antigen-presenting cells, mainly dendritic cells (DC), is essential for generating effective CTLs (8, 10). Therefore, various types of DC-based vaccines have been developed (8), such as directly targeting antigens to DCs in vivo via several types of C-type lectin receptors (CLR), including DEC-205, DC-SIGN, Clec9A, Clec12A, and mannose receptor (8, 11–15). In murine transplanted tumors, the DC-based vaccines by targeting some CLRs show protective effects against tumor growth (12, 13, 15). The effectiveness of these vaccines in primary cancer systems remains unknown. After tumor development, tumor-specific T cells within the immunosuppressive tumor microenvironment can lose functionality and DCs can acquire defects in tumor antigen processing/presentation (7).

Distinct subtypes of DCs have varying abilities to prime and activate distinct T cells (7, 8, 16). On the basis of ontogenies, classical DCs in mice are divided into two subsets: classical type 1 DCs (cDC1), including murine lymphoid CD8α⁺ DCs and migratory CD103^high DCs, and classical type II DCs (cDC2), including lymphoid CD4⁺ DCs and migratory CD11b⁺ DCs (17). Many studies have identified mouse cDC1 as the most efficient cell type to cross-present cellular-associated antigens to initiate CD8⁺ T cells (16, 18–20). The homolog and functional equivalent of mouse CD8α⁺ DCs is identified as CD141 (BDCA3)⁺XCR1⁺ DCs in humans (21, 22). The XCR1 chemokine receptor is selectively expressed on these DC subtypes (23, 24). Targeting microbial antigens to CD8α⁺ DCs via XCR1 induces antigen-specific CTLs and antibodies that protect mice from a lethal challenge with the influenza virus (25, 26). Thus, delivering tumor antigens to CD8α⁺ DCs via XCL1/XCR1 interactions may be a promising strategy for cancer vaccine designs.

Glypican-3 (GPC3), which is expressed in most HCCs, but is not detectable in normal hepatocytes and benign liver diseases, has been recognized as a target in HCC immunotherapy (27). Phase I and II clinical trials of HLA-A2- and HLA-A24–restricted GPC3-peptide vaccines show that the frequency of GPC3-specific CTLs is correlated with patient overall survival (28, 29). Delivering the HCC-associated GPC3 to CD8α⁺ DCs may be an effective HCC immunotherapy. Here, we constructed XCL1/GPC3 fusion molecules to target mouse CD8α⁺ and human CD141⁺ DCs. We aimed to induce GPC3-specific CD8⁺ T cells that could eliminate GPC3-expressing HCC inhibiting tumor development.

Study protocols involving mice were approved by the Institutional Animal Care and Use Committee (NCC2014A011) at the National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences (NCC/CH, CAMS, Beijing, China). The usage of subcutaneous lymph nodes (LN) from 14 patients with papillary thyroid carcinoma who received surgery therapy (30), and tumor tissues and the autologous LNs from 3 patients with HCC was approved by the Institutional Review Board/Ethics Committee (NCC2015YZ-22) of NCC/CH, CAMS following the guidelines issued by the Ministry of Science and Technology of China. The freshly removed tissues were stored in 10 ml of PRMI 1640 medium and transferred into laboratory within 30 minutes.

C57BL/6J mice (JAX:000664) were purchased from Vital River Laboratory Animal Technology. C57BL/6 congenic strain CD45.1 mice (JAX:002014) were purchased from The Jackson Laboratory. NOD-Prkdc^scid IL2rg^tm1 (NSG) mice (JAX:005557) were purchased from Biocytogen. C57BL/6 background OVA-specific CD8 TCR-transgenic (OT-I) mice (JAX:003831) were kindly provided by Dr. Bo Huang at the Institute of Basic Medical Sciences, CAMS. The three strains of mice used in the current study were male, 8 to 10 weeks old. C57BL/6J-Tg (AlblHBV) 44Bri/J mice (JAX:002226), which persistently express within hepatocytes and secret HBV surface antigen from hepatocytes into serum (31), were purchased from Health Science Center at Peking University (Beijing, China). One male AlblHBV mouse was mated with 2 females at 10 to 14 weeks old. When the female was pregnant, it was moved to a separate cage.

The cell lines of HEK293T (CRL-11268, received October 2010), 174xCEM.T2 (T2, CRL-1992, received April 2018), and HepG2 (HB-8065, received April 2014), which is HLA-A2⁺ and expresses GPC3 at mild levels, were purchased from ATCC. The original cells from ATCC were labeled as passage 0 (P0), cultured, and passaged according to ATCC instructions. The P5 cells were aliquoted and stored in liquid nitrogen for further usage. All the cells used in the experiments were less than P8 and authenticated by monitoring cell morphology and growth curve analysis. Hepa/GPC3 was constructed using Hepa1-6 cells in the laboratory (32). Mycoplasma testing was performed by PCR every 3 months and was negative for the entire course of this study. The cDNA clones of murine Xcl1, human GPC3, and E2crimson were all purchased from Sino-Biological. The Endotoxin-Free DNA Purification Kit was purchased from Tiangen Biotech.

The qRT-PCR analysis was performed using SYBR Premix Ex Taq (Takara) on an Applied Biosystems 7500 Real-Time PCR system (Life Technologies) using the primers listed in Supplementary Table S1. The primers were synthesized in SinoGenoMax. Total mRNA from the cells or liver tissues were isolated using TRIzol (Thermo Fisher Scientific) following the manufacturer's protocol. The cDNA was synthesized using PrimeScript RT Reagents (Takara). The PCR conditions were 3 minutes at 95°C, followed by 2-step cycles of 5 seconds at 95°C and 35 seconds at 60°C for total of 40 rounds, then 5 minutes at 72°C. Each sample was determined in duplicates. The relative mRNA levels were determined with GAPDH as control as reported previously (32).

We used the amino acid sequences of human GPC3 (Gene ID: 2719) in all fusion molecules because it has 95% homology with murine GPC3 (27). On the basis of the gene sequences of murine Xcl1 (mXcl1, Gene ID: 16963) and GPC3, we synthesized two types of mXcl1/GPC3 fusion genes, namely mXcl1-GPC3 and GPC3-mXcl1. By replacing mXcl1 with human XCL1 (hXCL1, Gene ID: 6375), we constructed hXCL1-GPC3 and GPC3-hXCL1. To replace GPC3, we constructed mXcl1-OVA using OVA (Gene ID: AUD54808.1), and mXcl1-E2crimson and E2crimson-mXcl1 using E2crimson, which encodes a red fluorescent protein. Gene synthesis was conducted at GenScript, and all fusion genes were cloned into pcDNA3.1 (-) for fusion protein expression.

We used the RaptorX structure prediction server to predict the structure of the fusion proteins, and ZDOCK server to predict the interactions of XCL1-fusion proteins with their corresponding XCR1 following the previously published protocols (33, 34). Briefly, from the menu of RaptorX homepage (http://raptorx.uchicago.edu/) “RaptorX Structure Prediction” was selected. In the “Job Identification” section, we supplied a job name and an e-mail address, provided the fusion protein sequences in FASTA format in the “Sequences” section, and pressed “Submit”. From the menu of ZDOCK server homepage (http://zdock.umassmed.edu/), “ZDOCK (3.0.2)” was selected. The fusion protein structure in PDB format was filled in the “Input Protein 1,” and the mouse XCR1 or human XCR1 in PDB format was filled in the “Input Protein 2.” After pressing the “Submit” button, the theoretically contacting residues of each submitted protein were selected. Results with a hyperlink were received via the e-mail provided. JMol in the results pages is used for molecular visualization.

HEK293T cells were cultured with DMEM supplemented with 10% FBS in a 37°C humidified incubator with 5% CO₂. In each well of a 6-well plate, 4 × 10⁵ HEK293T cells were added and cultured in 2 mL of completed DMEM for one day. When the cells achieved 70% to 80% confluency, 2.5 μg of each gene construct in 2.5 μL Tris-EDTA buffer was mixed with 8 μL of Lipofectamine 2000 (Invitrogen) and added into each well. The transfected cells were washed with PBS 68 hours later. We applied immunocytochemistry and immunoblot analyses to detect fusion protein expression in transfected cells using an anti-GPC3 mAb (R&D Systems). For immunocytochemistry analyses, single-cell slides were made with 50,000 cells per slide after centrifugation at 700 rpm × 5 minutes using Thermo Shandon Cytospin. Slides were fixed with cold acetone and stained with anti-GPC3 mAb. Images were captured and analyzed using ScanScope system (Aperio). For immunoblot analyses, we added 100 μL of cell lysis buffer (150 mmol/L NaCl, 1.5 mmol/L MgCl₂, 2 mmol/L EDTA, 2 mmol/L DTT, 10 mmol/L NaF, 12.5 mmol/L β-glycerophosphate, 1 mmol/L Na₃VO₄, 1 mmol/L PMSF, 1% Triton X-100, 0.1% sodium deoxycholate, and Roche Protease Inhibitor cocktail) into each well and incubated on ice for 30 minutes. After centrifugation at 12,000 rpm × 20 minutes at 4°C, the protein in the supernatant was quantified using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). A total of 40 μg protein was loaded into an SDS-PAGE gek, followed by immunoblotting with GPC3 mAb. β-Actin antibodies were used as loading controls. Images were captured and analyzed using Amersham Imager 600 (GE Healthcare).

By detecting the surrogate E2crimson expression, we monitored the real-time expression and location of the fusion protein in vivo after plasmid injection as reported previously (35). In brief, 250 μg of bupivacaine hydrochloride in a 50 μL volume of normal saline was injected intramuscularly into mice hind-legs. Twenty-four hours later, 100 μg of mXcl1-E2crimson or E2crimson plasmid DNA in 100 μL Tris-EDTA buffer were injected into the same sites. At different time points, the muscles with the injected plasmids and their draining LNs were collected, embedded in Tissue-Tek OCT compound (Sakura Finetech), and sectioned. After being fixed in cold acetone for 10 minutes, the sections were treated with PBS containing 1% BSA for 1 hour and stained with FITC-conjugated anti-mouse CD11c. In some experiments, the CD11c⁺ cells in the draining LNs were isolated using CD11c Microbeads (Miltenyi Biotec) and single-cell slides were prepared using Thermo Shandon Cytospin. Slides were fixed with 4% polyformaldehyde and stained with FITC-conjugated anti-mouse CD8α. The image acquisitions and analysis were performed using a Leica microscope and software.

HEK293T cells transfected with different gene constructs were collected 68 hours after transfection, and cell lysates were prepared by 3 cycles of freeze-thawing in a sterilized way. Total cellular proteins were quantified using Pierce BCA Protein Assay Kit after centrifugation at 12,000 × g for 20 minutes at 4°C. Chemotaxis assay was conducted based on a previous report (36). In brief, part of subcutaneous LNs was obtained from the patients with papillary thyroid carcinoma. The mouse spleens or human LNs were treated with collagenase D, and single-cell suspensions were prepared as reported previously (37). In the upper chamber of a Transwell insert (5-μm pore size), 1 × 10⁶ immune cells in 100 μL were added. In the lower chamber, 250 μg of transfected cellular proteins in 600 μL were added. Medium containing 100 ng/mL recombinant hXCL1 (PeproTech) was used as the positive control, and medium only was used as the negative control. After being cultured at 37°C for 90 minutes, cells that migrated into lower chambers were collected and counted. The percentages of migrated murine CD8α⁺ or human CD141⁺ DCs were analyzed by flow cytometry (FCM).

The information for the antibodies used is provided in Supplementary Table S2. All staining and washes were performed in ice-cold, freshly prepared FCM staining buffer, PBS containing 0.1% BSA, and 2 mmol/L EDTA at pH = 7.4. All staining incubations were conducted on ice. To determine the murine CD8α⁺ DCs and human CD141⁺ DCs that migrated in the chemotaxis assay, the cells collected from lower chambers were resuspended into 100 μL staining buffer with the concentration of 1–3 × 10⁷ cells/mL. After the addition of 1 μL anti-mouse CD16/CD32 and incubation for 15 minutes, 20 μL of antibody cocktail diluted in staining buffer was added without washing. For staining of murine CD8α⁺ DCs, the antibody cocktail contained anti-mouse I-A/I-E, CD11c, and CD8α was used. For staining of human CD141⁺ DCs, the antibody cocktail contained anti-human HLA-DR, CD11c, and CD141 was used. To determine the target effect of expressed mXCL1-GPC3 proteins in vivo, draining LNs were collected on D6 after mXCL1-GPC3 plasmid injection. Single-cell suspensions (37) were stained with anti-mouse CD11c, MHC-II, CD8α, and FITC-labeled anti-His tags (Easybio). To determine the cells in the mouse liver, the intrahepatic lymphocytes (IHL) were isolated as reported previously (32). The single-cell suspension was stained with anti-mouse CD45, CD3, and NK1.1. To determine the intracellular cytokine production, the cells were first stained with antibodies against their surface makers, followed by fixation and permeabilization using the Intracellular Fixation & Permeabilization Buffer set purchased from Thermo Fisher Scientific. After being washed the cells were stained with PE/Cy7-conjugated anti-mouse IFNγ or with PE-conjugated anti-mouse Granzyme B (GrzB). Data were acquired using an LSR-II (Becton Dickinson) and analyzed using FlowJo software (Tree Star Inc.).

To quantify IFNγ, GrzB, CCL5, CXCL9, and IL18 in liver tissues, interstitial liquid was prepared as reported previously (32). In brief, each 100 mg tissue sample was cut into small pieces in 400 μL of ice-cold normal saline and incubated on ice for 15 minutes. Total protein levels of liver interstitial liquid were measured by the Pierce BCA Protein Assay Kit. The cytokine concentrations in the liver interstitial fluid were measured using commercialized ELISA Kits according to the manufacturer's instructions.

IL12 production from cell supernatant was quantified using commercialized ELISA Kits listed in Supplementary Table S2. The mouse XCR1⁺ DCs or human XCR1⁺ DCs were chemoattracted as described in the “chemotaxis assay” section, and stimulated with the cell lysates (containing 250 μg protein) from XCL1-GPC3–transfected or GPC3-transfected cells for 72 hours.

Male CD45.1 mice were immunized with 10 μg of mXcl1-OVA or OVA plasmid in 10 μL Tris-EDTA buffer with a low pressure-accelerated gene gun (BioWare) at 40 psi following the previously reported protocol (38). CD45.2⁺ OT-I mouse T cells were isolated using mouse CD8α⁺ T Cell Isolation Kit (Cat: 130-096-543, Miltenyi Biotec) from splenocytes, and labeled with 5 μmol/L CFSE and transferred via tail vein injection into immunized CD45.1 mice five days after the immunization as described previously (12). Each mouse was bled from tail vein (about 150 μL into 20 μL of 5 mmol/L EDTA solution) on D2, D4, and sacrificed on D7 after the transfer of OT-I T cells. Whole blood was treated with Lysing Solution following the manufacturer's instructions (Cat: GAS010, Thermo Fisher Scientific). When the mice were sacrificed, the draining LNs from the immunized mice were collected to prepare the single-cell suspensions as reported previously (37). The cells were resuspended into FCM staining buffer at 1–3 × 10⁷ cells/mL. In a 100 μL cell suspension volume, the antibody cocktail of 1 μL PE-conjugated anti-mouse CD45.1 and 1 μL APC-conjugated CD45.2 was added. OT-I T-cell proliferation was determined by FCM after gating CD45.2⁺ cells as reported previously (12).

Antigen-specific cytotoxicity assays were performed as described previously (39, 40). In brief, to determine the antigen-specific cytotoxicity in the murine system, we isolated IHLs from immunized mice as previously reported and used Hepa/GPC3 cells as the targets (32). To determine the antigen-specific cytotoxicity in the human system, we prepared lymphocyte suspensions from freshly removed LNs as described in the “chemotaxis assay” section. This cell suspension was stained with PE-labeled anti-HLA-A2 and analyzed by FCM and the HLA-A2⁺ donors were selected. As described in the aforementioned chemotaxis assay section, migrated cells that were chemoattracted by hXCL1-GPC3– or GPC3-transfected cell lysates were used to stimulate the autologous T cells. The effectors were generated after being cocultured with the ratio of 20 autologous T cells to one migrated cell in the RPMI1640 medium supplemented with 10% FBS for 5 days. The targets were HepG2 or T2 cells pulsed with 10 μmol/L GPC3_144-152 peptide.

Diethylnitrosamine (DEN) was diluted in normal saline at the 2.5 mg/mL. Autochthonous liver cancer was induced as described in our previous report by administrating 25 mg/kg DEN via intraperitoneal injection to 2-week-old male AlblHBV mice (32). The immunization was divided into 3 injections with 2-week intervals. For each injection, 10 μg of plasmid DNA in 10 μL of Tris-EDTA buffer was delivered to the shaved right lateral flanks of mice using a low pressure-accelerated gene gun at 40 psi as reported previously (38). Two days before each injection, serum levels of alanine transaminase (ALT) in each mouse were measured using reagents from Biosino BioTechnology and determined with spectrophotometer (V-1000D, MAPADA Instruments Co., Ltd.).

Each AlblHBV mouse received 1 × 10⁶ Hepa/GPC3 cells in 200 μL of PBS containing 0.1% BSA subcutaneously. Perpendicular tumor diameters were measured using a Vernier caliper (Ohaus Scale Corp), and tumor volume was calculated using the following formula: length × width² × 0.5 following the previously published report (32). When measurable tumor nodules formed, each mouse received 1 × 10⁶ splenic T cells intravenously as described above that were prepared from the plasmid-immunized mice. On the following day after T-cell transfer, some mice (n = 5) received 200 μg of rat anti-mouse PD-1, and some received 200 μg of rat IgG. In some experiments, the transferred T cells were labeled with CFSE. The tumors were removed 5 days later and embedded in Tissue-Tek OCT compound. After being frozen-sectioned, the slides were fixed with cold acetone, mounted Vectashield^R Mounting Medium with DAPI (Vector Laboratories). The image acquisitions and analysis were performed using a Leica immunofluorescent microscope and software.

We prepared HCC patient-derived xenografts (PDX) using freshly removed tumor tissues according to a previous report (41). In brief, in a 200 μL solution of DMEM mixed with Matrigel matrix, five 1-mm³ tumor fragments were resuspended, and inoculated subcutaneously into NSG mice. The patient LNs were obtained at the same time of HCC tissue collection. Autologous T cells were prepared and stimulated as described in the aforementioned antigen-specific cytotoxicity assay. Each PDX mouse (n = 3) received 1 × 10⁶ T cells in 100 μL of PBS containing 0.1% BSA on D6 when tumors became palpable by tail vein injection, or 5 × 10⁵ T cells in 25 μL of PBS containing 0.1% BSA on day 8 by subcutaneous injection in the vicinity of the xenograft after implantation. Tumor incidence and volume were measured as described above at different time points.

Two-tailed t tests were used to compare differences between groups by using GraphPad Prism 5 version 5.1. P values less than 0.05 were considered statistically significant.

To target murine CD8α⁺CD11c⁺XCR1⁺ DCs, we constructed mXCL1-GPC3, in which the mXCL1 C-terminus was linked with the GPC3 N-terminus, and GPC3-mXCL1, in which the mXCL1 N-terminus was linked with the GPC3 C-terminus. The linker contained 11 amino acids of (glycine)₅ -serine- (glycine)₅. We replaced mXCL1 with hXCL1 to generate hXCL1-GPC3 and GPC3-hXCL1 constructs for targeting the human CD141⁺CD11c⁺XCR1⁺ DCs (Fig. 1A). At amino acid positions 560–580, GPC3 contains a glycosylphosphatidylinositol anchor, that anchors the N-terminus extracellular domain to the cell membrane (27); we deleted it in all constructs to get better secretion of the fusion proteins. A sequence that encodes six amino acids of histidine (42) was inserted in the C-terminal regions of the fusion molecules to facilitate the characterization. The matured fusion proteins have a total of 639 amino acids with the molecular weight of approximately 72 kDa. We constructed mXCL1-OVA to validate its targeting function of the fusion protein on the specified DCs in vivo (Fig. 1A). To trace the real-time expression and location in vivo after fusion gene injection, we replaced the GPC3 with E2crimson to construct mXCL1-E2crimson and E2crimson-mXCL1 (Supplementary Fig. S1A).

The 3D structural models of fusion proteins were analyzed using the RaptorX structure prediction server (http://raptorx.uchicago.edu/). Chemokine XCL1 was well exposed in each of the constructs regardless of mXCL1 or hXCL1 linked to GPC3 at its N-terminus or C-terminus (Fig. 1B). The ZDOCK server (http://zdock.umassmed.edu/) was used for the interaction analysis and indicated that all the fusion proteins could bind to their corresponding XCR1 either when XCL1 was at the N-terminus or C-terminus linked with GPC3 (Fig. 1C). Similar results were obtained after analysis of mXCL1-E2crimson, E2crimson-mXCL1, and mXCL1-OVA (Supplementary Fig. S1B and S1C).

After HEK293T cells were transfected with mXcl1-E2crimson or E2crimson-mXcl1, red fluorescent protein was detectable at 24 hours and peaked at 65 to 72 hours after transfection. Fluorescence intensity showed no differences between cells that were transfected with mXcl1-E2crimson or E2crimson-mXcl1 (Supplementary Fig. S2).

We then analyzed the expression of XCL1/GPC3 fusion proteins 68 hours after transfection. Immunocytochemistry staining with anti-GPC3 detected expressed proteins within the cytoplasm (Fig. 2A). Immunoblot analysis using anti-GPC3 staining showed that the fusion proteins were approximately 72 kDa, which was consistent with the predicted molecular weight (Fig. 2B).

We conducted a chemotaxis assay on murine CD8α⁺ DCs (n = 5) to validate the chemoattractant activity of XCL1 in the expressed fusion proteins. Compared with the numbers of migrated CD8α⁺ DCs (MHC-II⁺CD11c⁺CD8α⁺) when medium only or when the GPC3-transfected cell lysate was added in the lower chambers, the numbers of migrated CD8α⁺ DCs increased more than 3-fold when mXcl1/GPC3-transfected cell lysates were added to the lower chambers. Notably, more CD8α⁺ DCs were detected when the mXcl1-GPC3–transfected cell lysate was added into the lower chambers in comparison with the same amount of GPC3-mXcl1–transfected cell lysate (Fig. 2C). These results suggested that better chemoattractant activity occurs when the mXCL1 C-terminus is linked to GPC3.

We then confirmed the protein expression after hXCL1-GPC3 gene transfection (Fig. 2A and B). By using human LN cells (5 donors), we validated the chemoattractant activity of hXCL1-GPC3–transfected cell lysates on human CD141⁺ DCs (Fig. 2D). Therefore, we mainly used mXcl1-GPC3 and hXCL1-GPC3 for further analysis.

To confirm the targeting effect of our constructs on LN DCs in vivo, we injected the mXcl1-E2crimson or E2crimson plasmid into mouse hind leg muscles. At the sites of plasmid injection E2crimson was detected on day one (D1) and lasted until D8, and no differences of the positive cell numbers and fluorescent intensity were observed (Supplementary Fig. S3A). In the draining LNs of mice that received the mXcl1-E2crimson plasmid, the red protein was detectable on D2, peaked on D6, and was still detected on D8 (Supplementary Fig. S3B). However, in the draining LNs of mice that received the E2crimson plasmid, the protein was barely detected (Fig. 3A). Further analysis showed that E2crimson mainly colocalized with CD11c⁺ DCs (Fig. 3B). We further isolated CD11c⁺ DCs from draining LNs and stained them with FITC-labeled anti-mouse CD8α. E2crimson was mainly detected in CD8α⁺CD11c⁺ cells from mice that received the mXcl1-E2crimson plasmid (Fig. 3C).

We analyzed the capacity of LN DCs to take up the fused GPC3 tumor antigen. Six days after mXcl1-GPC3 plasmid injection, GPC3 protein was detected in more than 40% of CD8α⁺ DCs and about 15% of CD8α⁻ DCs of the draining LNs. However, only a small proportion of CD8α⁺ DCs (2.1%) was found to uptake the protein when mice received the same amount of the GPC3 plasmid (Fig. 3D). IL12 production from DCs is crucial for generation of Th1 immunity (43), which is important for antitumor immunotherapy (9). Increased production of IL12 was detected both in the mouse and human DCs when they were chemoattracted and stimulated by the mXcl1-GPC3- and hXCL1-GPC3–transfected cell lysates compared with the same amount of GPC3-transfected cell lysates (Fig. 3E).

The abovementioned results suggested that the XCL1-fusion protein efficiently targeted LN DCs, facilitating their uptake of delivered protein and promoting their IL12 production for Th1 immunity induction. Thus, we constructed the mXCL1-OVA fusion protein (Fig. 1) to validate the targeting function on CD8α⁺ DCs in vivo. CD45.1 mice were immunized with the same amount of the OVA or mXcl1-OVA plasmid and each immunized mouse received 1 × 10⁶ CFSE-labeled OT-I T cells. More potent T-cell proliferation was detected in mXcl1-OVA–immunized mice than in OVA-immunized mice (Fig. 3F). When the immunized mice were sacrificed seven days after receiving OT-I T cells, 66% of OT-I T cells in draining LNs of mXcl1-OVA–immunized mice were proliferating, whereas only 25% of the transferred T cells were proliferating in the OVA-immunized mice (Fig. 3G).

We used a murine autochthonous liver cancer model with an HBV background, which is considered similar to human HCC, to examine if mXCL1-GPC3 is an effective tumor vaccine to impede liver cancer development. The DEN-injected mice were immunized with the same amount of mXcl1-GPC3 or GPC3 plasmid at the same time points (Fig. 4A). The elevation of serum ALT levels, which reflect the status of hepatocyte damaged (31), did not occur in the mice that were immunized with mXcl1-GPC3 (G1, G3). Nevertheless, body weights of the mice were lower when they were immunized with mXcl1-GPC3 plasmid in comparison with the unimmunized and GPC3-immunized mice (Supplementary Fig. S5).

Starting from week 6, when malignant hepatocyte clusters formed, the mice were immunized with different plasmids (G1, G3). By week 16, all GPC3-immunized (G3, n = 4) and unimmunized mice (G5, n = 7) developed liver tumors. However, none (0/6) of mXcl1-GPC3–immunized mice (G1) had liver tumor. By week 22, tumor loads decreased significantly in the mXcl1-GPC3–immunized mice (G1) compared with the GPC3-immunized (G3) and unimmunized (G5) mice (Fig. 4B). We also immunized the DEN-injected mice starting from week 14 when small liver cancer nodules developed. In comparison with the GPC3-immunized and unimmunized mice, significantly fewer tumor numbers with decreased tumor size were observed in mXcl1-GPC3–immunized mice by week 22 (Fig. 4B).

All the tumor nodules were solid and protruded on the liver surface in unimmunized mice and GPC3-immunized mice. However, in mXcl1-GPC3–immunized mice, the tumor nodules were no longer protruded and a liquid halo surrounded the nodules (Fig. 4C). Analyses of qRT-PCR and immunochemistry showed that the GPC3 expression in the tumor nodules of mXcl1-GPC3–immunized mice was significantly lower than those in GPC3-immunized mice (Fig. 4D). These results suggested that mXcl1-GPC3 immunization induced effective immunity that eliminated most of GPC3-expressing tumor cells.

We isolated IHLs from immunized mice (Fig. 4A) and restimulated them with purified GPC3 protein in vitro. The percentage and total numbers of CD8⁺ IFNγ-producing and CD8⁺ GrzB-producing T cells significantly increased in mXcl1-GPC3–immunized mice compared with GPC3-immunized and unimmunized mice (Fig. 5A). When IHLs were cocultured with Hepa/GPC3 cells, significantly higher cytotoxicity was detected in the cells mixed with IHLs from mXcl1-GPC3–immunized mice in comparison with the IHLs from GPC3-immunized mice (Fig. 5B).

We failed to detect an increase of serum IFNγ in immunized mice. However, compared with GPC3-immunized and unimmunized mice, there was significantly higher IFNγ and GrzB detected in the liver intercellular fluid of mXcl1-GPC3–immunized mice. The concentration of CCL5, CXCL9, and IL18, which are involved in recruiting and enhancing the activities of natural killer (NK) and NKT cells (19, 36), was significantly higher in the livers of mXcl1-GPC3–immunized mice compared with GPC3-immunized and unimmunized mice. The mXcl1 mRNA expression of tumors was also higher in mXcl1-GPC3–immunized compared with GPC3-immunized mice (Fig. 5C). NK and NKT cells are abundant in liver and are important components involved in eradiating established liver tumors (5). Analysis showed that the percentages of NK (NK1.1⁺CD3⁻) and NKT (NK1.1⁺CD3⁺) cells among the IHLs were increased in mXcl1-GPC3-immunized mice in comparison of GPC3-immunized mice (Fig. 5D).

Among the IHLs of mXcl1-GPC3–immunized mice, PD-1 expression on CD3⁺ T cells isolated from tumor nodules was higher compared with T cells isolated from nodule-free tissues. However, no difference was observed in CD3⁻ cells (Fig. 6A). We then determined whether one dose of anti–PD-1 enhanced the antitumor activities of the tumor-infiltrated IHLs. Unimmunized mice were subcutaneously inoculated with 1 × 10⁶ Hepa/GPC3 cells. When measurable tumor nodules formed each mouse intravenously received 1 × 10⁶ T cells, which were isolated from mXcl1-GPC3–immunized mice. By D12 and D15, the average tumor volume significantly decreased and the tumor disappeared in 1 of 5 mice that received the immunized T cells. Mice administrated with one dose of anti–PD-1 after receiving the immunized T cells demonstrated tumor nodule disappearance in all mice by D15 (Fig. 6B).

The immunized T cells were labeled with CFSE and transferred to the Hepa/GPC3 tumor–bearing mice. When tumor nodules were removed 5 days after the T-cell transfer, more CFSE-labeled T cells were detected within the tumor tissues after anti–PD-1 administration. Notably, within the Hepa/GPC3 tumor tissues only the T cells isolated from mXcl1-GPC3–immunized mice were detected (Fig. 6C).

Human subcutaneous LNs were obtained when the patients with papillary thyroid carcinoma received surgery therapy (30). Using HLA-A2⁺ donors (3 donors; Supplementary Fig. S7), we analyzed the capability of hXCL1-GPC3 fusion proteins to activate autologous T cells against HepG2, which is an HCC cell line with HLA-A2⁺ and mild expression of GPC3 (Supplementary Fig. S8; ref. 44). LN DCs were chemoattracted by hXCL1-GPC3– or GPC3-transfected cell lysates (Fig. 2D) and then cocultured with autologous T cells for 5 days. The cytotoxicity against HepG2 by T cells activated by hXCL1-GPC3–conditioned DCs was significantly higher than the T cells activated by GPC3-conditioned DCs (Fig. 7A). We also used T2 cells pulsed with the GPC3_144-152 peptide as targets. The antigen-specific cytotoxicity toward T2 cells was only observed by T cells that were activated by hXCL1-GPC3–conditioned DCs (Fig. 7A), indicating the generation of GPC3-specific CTLs.

To extend the potential of XCL1-GPC3 fusion molecule in humans, we tested the effect of this molecule in the HCC-PDX model. We obtained fresh, surgically removed HCC tissues that expressed GPC3 (Fig. 7B) and transplanted them into NSG immunodeficient mice to generate a human HCC-PDX model (repeated with 3 HCC donors; Supplementary Fig. S8). At the same time, LNs from the same donors were obtained. The migrated cells that were chemoattracted by hXCL1-GPC3– or by GPC3-transfected cell lysates (Fig. 2D) were then cocultured with autologous T cells for 5 days. When mice developed measurable tumors, activated autologous T cells were then injected intravenously into the PDX mice on day 6 (Fig. 7C). All xenografts disappeared in the mice 10 days after they received the T cells that were stimulated with the DCs chemoattracted by hXCL1-GPC3 (Fig. 7C). We also tested the effects of local injections on day 8 after xenografts were larger than 100 mm³ (Fig. 7D). Xenografts disappeared 2 weeks after transfer of T cells that were stimulated by DCs chemoattracted by hXCL1-GPC3.

Classical type I DCs (cDC1; murine lymphoid XCR1⁺CD8α⁺, migratory XCR1⁺CD103^high DCs, and human XCR1⁺CD141⁺ DCs) have been recognized as professional inducers of antigen-specific CD8⁺ T cells (16, 18, 19). Given the selective expression of XCR1 on these cDC1s, we hypothesized that the XCL1-GPC3 fusion protein may be an efficient modality to induce tumor-specific CTLs. This study indicated that linking the XCL1 chemokine at its C-terminus to GPC3 at its N-terminus could deliver the tumor antigen efficiently to murine XCR1⁺CD8α⁺ DCs and to human XCR1⁺CD141⁺ DCs. These targeted DCs induced the generation of GPC3-specific CD8⁺ T cells, which eradicated GPC3-expressing tumor cells both in murine and human systems. In a murine autochthonous liver cancer model, tumor development was markedly hampered after therapeutic immunization with the mXcl1-GPC3 plasmid. XCL1-GPC3 may be a promising cancer vaccine for treating patients with HCC. The antitumor effect of mXcl1-GPC3 immunization was further enhanced by administration of the PD-1 antibody. Thus, XCL1-GPC3 vaccination may improve the efficacy of checkpoint inhibitors in HCC by inducing “high-quality” GPC3-specific T cells.

When contemplating DC in vivo targeting for tumor vaccination, one should consider the expressions of targeted receptors on the other cells. DCs express particular endocytic CLRs, which are also detected in macrophages, monocyte-derived DCs, and some other immune cells (11). DEC-205 is the most well-known CLR studied for the induction of CD4 and CD8 T-cell responses (11, 12, 14). Delivering antigens directly to mature DCs using anti-DEC205 in vivo has demonstrated efficient activation of T-cell responses (11, 12, 14). In this study, we used a similar concept. DEC-205 is expressed on cDC1s, skin Langerhans cells in mice, and a wide variety of leukocytes in humans (11). Thus, T-cell response may be suboptimal when an antigen bound to anti–DEC-205 is administered, because only some cancer patients benefit from the immunization (14). The efficacy of murine cDC1s in generating CD8⁺ T-cell responses has been well documented (16, 18, 19). However, the surface markers of human functional homolog of mouse cDC1s are different, which express CD141⁺/BDCA3^high (21, 22). The DC subsets in both mice and humans selectively express the XCR1 chemokine receptor (17, 23, 24). Here, we targeted XCR1-expressing DCs by fusing XCL1 to a tumor antigen. We aimed to deliver the antigen mainly to these types of cells and trigger their activation via the XCL1/XCR1 interaction. Our results showed that linking XCL1 at its C-terminus to GPC3 at its N-terminus reserved the chemoattractant activity of XCL1. The antigen fused to XCL1 improved the uptake of the antigen and enhanced presentation of antigen by DCs. These findings confirm various reports in other study systems, which fuse XCL1 to mCherry, hemagglutinin of influenza A virus, or OVA (25, 26, 45). When XCL1-GPC3 was taken up by XCR1⁺DCs, the cells produced more IL12, which favors Th1 antitumor immunity (10, 43). Potent T-cell proliferation was detected after the XCL1 was fused to GPC3, and the Xcl1-GPC3–immunized mice generated more CD8⁺ IFNγ-producing and CD8⁺ GrzB-producing T cells. Nevertheless, the vaccination efficacy to human cancers should be further investigated.

Cross-talk between different types of immune cells is an important aspect of antitumor immunity (19, 20). Innate lymphocytes, like NK and NKT cells, are enriched in the liver and can be activated directly by cDC1s and cytokines, mainly IL12 and IL18 (19, 46). Activated NK cells and antigen-specific CTLs further secrete several chemokines, including XCL1, CXCL9/10, and CCL5, inducing the local recruitment of cDC1s/CTLs/NK/NKT cells and leading to the subsequent production of IFNγ, GrzB, and perforin-dependent cytotoxicity (19, 20, 24). In mXcl1-GPC3–immunized mice, liver infiltration of GPC3-specific CTLs as well as NK and NKT cells increased compared with GPC3-immunized and unimmunized mice. Although we failed to detect the increased production of IFNγ in the sera of immunized mice, mXcl1-GPC3–immunized mice had significantly higher concentrations of CCL5, CXCL9, and IL18 in the liver, which locally recruit and activate NK and NKT cells (19, 36). In the isolated IHLs of mXcl1-GPC3–immunized mice, the mRNA expression level of Xcl1, which is mainly produced by activated CTLs and NK cells (20, 23, 24), was significantly higher than that in IHLs of GPC3-immunized mice. Therefore, the GPC3-specific CD8⁺ T-cell generation after mXcl1-GPC3 immunization and their infiltration into the tumor bed may be able to initiate antitumor immunity and enhance the activities mediated by other immune cells. In line with our observation, several other studies have demonstrated the importance of cross-talk between the antitumor effector cells (CTLs and NK cells) and cDC1s in effective antitumor immunity (20, 47).

Draining LNs organize immune responses by bringing DCs, which carry antigens, and T cells in close proximity, allowing them to active and promote T-cell proliferation. Our previous studies have indicated that promoting antigen-carrying monocyte-derived DCs to draining LNs enhances the antigen-specific T-cell immune response; however, the antigen-specific CTL responses are suboptimal (32, 48). Other studies have demonstrated that cDC1s are adept at taking up dead tumor cells and transporting the tumor antigens to draining LNs to cross-prime CD8⁺ T cells (49). These studies suggested that CCR7-dependent migration of CD103^high DCs, which express XCR1 and are classified as migratory cDC1s, is necessary for the generation of tumoral T-cell infiltration (49). Therefore, delivering tumor antigen directly to these specified lymph organ-resident DCs, could be another potential approach to elicit productive CTLs. Our constructed XCL1-GPC3 fusion molecules efficiently targeted CD8α⁺ DCs in draining LNs, which facilitated “on-site” antigen presentation and induced de novo generation of GPC3-specific CTLs, thereby eliminating GPC3-expressing tumor cells.

Antitumor efficacy relies on effective T-cell entry into the tumor bed, which can be stopped when PD-1/PD-L1 is overexpressed (47). The presence of tumor-specific CTLs is the prerequisite for PD-1/PD-L1 immunotherapy, and their infiltration plays critical roles in cancer treatment (9). Thus, mXcl1-GPC3 immunization could improve the response of solid tumors to checkpoint inhibitors when an inhibitory microenvironment is present by enhancing the number of tumor-specific CTLs that enter tumor legions. We utilized NSG immunodeficient mice to generate a HCC-PDX model and confirmed that the expressed hXCL1-GPC3 could induce potent antitumor T cells that inhibited HCC growth. The HCC-PDX model data generated here provided evidence for further studies in patients. Thus, XCL1-GPC3 might be a promising cancer vaccine to compensate for the deficiency of the checkpoint blockades in HCC immunotherapy.

No potential conflicts of interest were disclosed.

Conception and design: C. Qu

Development of methodology: K. Chen, Z. Wu, H. Zhao, C. Ma, C. Qu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Chen, Z. Wu, H. Zhao, Y. Wang, Y. Ge, Z. Li, C. An, Y. Liu, F. Wang, X. Bi, J. Cai, C. Qu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Chen, Z. Wu, Y. Wang, Y. Ge, C. Qu

Writing, review, and/or revision of the manuscript: K. Chen, C. Qu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Wang, C. Qu

Study supervision: K. Chen, H. Wang, J. Cai, C. Qu

Other (participated in some study design): K. Chen

This work was supported by National Natural Science Foundation of China (no. 81571620 and 81161120495, to C. Qu), State Key Projects Specialized for Infectious Diseases (2017ZX1020121101-006, to C. Qu), and Chinese Academy of Medical Sciences Innovative Medicine (no. 2016-I2M-1-007 and 2019-I2M-2-004, to C. Qu, and 2017-I2M-4-002, to H. Zhao). The sponsors had no role in the study design, data collection, data analysis, data interpretation, or writing of the manuscript.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.