Introduction

The innate and adaptive immune systems are essential for immune surveillance of cancer.1 The interaction between the immune system and cancer cells is a continuous, dynamic, and intricate process.2 Because of the unique physiological functions of the liver, hepatocellular carcinoma (HCC) cells possess an inherent advantage in immune evasion. The deactivation of cytotoxic T lymphocytes (CTLs) represents a crucial characteristic of HCC.3 The function of CTLs is influenced by many factors in the tumor microenvironment (TME). In recent years, increasing evidence indicates that neutrophils can regulate T-cell function through direct or indirect mechanisms.4–6 Moreover, neutrophil extracellular traps (NETs) released by neutrophils have been reported to alter the activation threshold of T cells.7 However, the role and mechanisms of neutrophils and NETs in mediating dysfunctional CTLs in HCC remain unclear.

HCC is the sixth most common cancer worldwide, with poor prognosis and high mortality rates because of its chemotherapy and radiotherapy resistance.8 In recent years, immunotherapy represented by immune checkpoint inhibitors (ICIs) has revolutionized the treatment of a wide range of malignant tumors. ICIs have shown satisfactory results in clinical trials of various malignancies.9–12 However, patients with HCC have a low overall response rate to ICIs, with only 20% and 17% of patients with advanced HCC responding to nivolumab and pembrolizumab, respectively, according to the results of two phase II clinical trials CheckMate-040 and KEYNOTE-224.13 14 Therefore, uncovering the molecular mechanisms of how tumor cells evade immune cell killing is of great significance for improving the efficacy of immunotherapy and developing new immunotherapeutic strategies.

Due to its programmability and flexibility, CRISPR-mediated genome editing has become a powerful tool in cancer biology, and high-throughput screening methods based on its principle have become an effective means to search for new immune modulators and tumor immune targets.15–18 Compared with in vitro, in vivo screening can better reflect the interaction between tumor cells and their immune microenvironment. In this study, using CRISPR screening, we identified Gsk3a as a critical candidate target for immune evasion in HCC. Functional and mechanistic studies demonstrated that Gsk3a could inhibit CTL activity by inducing neutrophil chemotaxis and NETs formation. Increased expression of Gsk3a was detected in anti-programmed cell death protein-1 (PD-1) antibody non-responsive patients. Pharmacological inhibition of Gsk3a could enhance CTL function and further improve the efficacy of anti-PD-1 antibody. In summary, our study provided new insights into the immune evasion mechanisms of HCC cells, and revealed Gsk3a may be a novel therapeutic target for immunotherapy in HCC.

Materials and methodsCRISPR sgRNA library and screen in vivo

A mouse disease-related immune gene library was screened, and the construction of the gene library was previously validated in a prior study. The gene library information used for screening is derived from the article by Ji et al.18 The library consisted of 11,184 sgRNAs targeting 2,796 mouse genes corresponding to human diseases and the immune system, along with 816 non-targeting control sgRNAs. After synthesis and amplification of the DNA oligonucleotide library on a microarray, it was cloned into the lentiGuide-Puro vector to generate the disease-related immune gene library. The library was purified, and sequencing was performed to monitor the abundance changes of each sgRNA between the initial and final cell populations.

To generate cells with stable Cas9 expression, the lentiCas9-Blast (Addgene), pMD2.G (Addgene), and psPAX2 (Addgene) constructs were introduced into HEK 293 T cells for lentiviral packaging. Stable integrated Hepa1-6-Cas9 cells were selected using blasticidin (5 µg/mL). Cells infected with the virus containing the disease-related immune gene library were infected at a multiplicity of infection of 0.3 to ensure that each cell was infected with a single copy of the virus. After 48 hours of transduction, infected cells were selected with 5 µg/mL puromycin for 7 days. After 7 days, genomic DNA was extracted from a portion of the cells, while another portion was resuspended in phosphate-buffered saline (PBS) for transplantation. The transfected Hepa1-6-Cas9 cells containing the disease-related immune gene library were injected subcutaneously into C57BL/6 or NPG mice at a density of 4×106 cells per mouse for in vivo screening. The survival status of the mice and tumor size were monitored daily, and 2 weeks later, the mice were euthanized, and the tumors were dissected for further analysis.

The sgRNA sequences were amplified through two rounds of quantitative PCR (qPCR) and then subjected to sequencing using the HiSeq 2500 system (Illumina). The original FASTQ files were demultiplexed using Geneious V.8.0 (Biomatters). The constant regions of the sgRNA sequences were removed, and the read counts of each sgRNA per sample were normalized by the total read counts of each sample and subjected to logarithmic transformation. The MAGeCK analysis method was employed to quantify the abundance of sgRNAs in each sample. The proliferation-promoting and proliferation-inhibiting genes in C57BL/6 and NPG mice (n=5, each group) were ranked using robust rank aggregation (RRA), and key genes affecting immune adaptation were identified through cross-comparison. Please refer to online supplemental tables S1 and S2 for detailed information.

In vivo animal studies

To validate immune-related genes in vivo, subcutaneous xenograft models were established by injecting 4×106 different HCC cells subcutaneously into immunocompetent C57BL/6 mice and severely immunodeficient NPG mice (n=5, each group). Tumor volume was measured at intervals of 2–3 days when the subcutaneous tumors were macroscopically visible and calculated using the formula: Volume =(length×width2)/2. At the end of each experiment, mice were euthanized, and the tumors were dissected, weighed, and photographed. Subsequent immunofluorescence staining was performed after tumor fixation with 4% paraformaldehyde.

All animals were fed under standard conditions. The animal experiments were conducted in an specific-pathogen-free (SPF) grade laboratory and approved by the Animal Ethics Committee of Fudan University (2023-HSYY-295JZS, Shanghai, China).

The rest of the mouse studies are available in the online supplemental methods.

Flow cytometry analysis

Preparation of single-cell suspension: For flow cytometry of tumor cells in vivo, fresh mouse tumor tissue of appropriate size was dissected and mechanically separated using sterile ophthalmic scissors. The tumor fragments were then incubated at 37°C in serum-free Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with DNase I (0.1 mg/mL; Solarbio), collagenase I (1 mg/mL; Sigma-Aldrich), collagenase II (1 mg/mL; Sigma-Aldrich), and collagenase IV (1 mg/mL; Sigma-Aldrich) for 60 min with continuous stirring. The resulting single-cell suspension was passed through a 70 µm cell strainer (Miltenyi). Subsequently, red blood cells within the tumor were lysed for 5 min using red blood cell lysis buffer (Miltenyi) at room temperature. The lysed tumor cells were then centrifuged at 400 g for 5 min at 4°C, and the reaction was stopped by adding 5% fetal bovine serum (FBS) RPMI 1640 medium. For flow cytometry experiments in vitro, the suspended immune cells in the lower chamber of a Transwell system were collected and washed three times with PBS before being collected in centrifuge tubes.

The cells were washed with PBS and stained with BD Horizon Fixable Viability Stain 510 (BD Biosciences) at a 1:1,000 dilution in PBS for 15 min at 4°C. Afterward, the cells were blocked with a monoclonal antibody against CD16/32 (BioLegend) at 4°C for 15 min. For surface staining, cells were stained with fluorescently labeled surface protein antibodies in a staining buffer at 4°C for 30 min. For intracellular staining (interferon-gamma (IFN-γ), granzyme B (GZMB) and forkhead box protein P3 (Foxp3)), cells were fixed and permeabilized after surface staining. Follow the manufacturer’s instructions using Fix/Perm Buffer (BD, 562574) and Perm/Wash Buffer (BD, 562574) to fix and permeabilize the cells. After fixation and permeabilization, incubate the samples with the appropriate antibodies at 4°C for 30 min. The antibodies and concentrations used for staining are detailed in online supplemental table S3.

Flow cytometry data acquisition was performed using a CytoFLEX flow cytometer (Beckman Coulter), and data analysis was conducted using FlowJo software (V.10.8.1, TreeStar).

Human specimens

A cohort comprising 23 patients with HCC who underwent liver resection for primary onset was obtained from Huashan Hospital. According to the staging system for chronic liver disease, there were 16 cases in stages S0/1 and 7 cases in stage S4. The paraffin-fixed HCC tissues were obtained from 32 patients with HCC who underwent liver resection and received anti-PD-1 therapy at our institution. Response evaluation was performed every 3 months using the modified Response Evaluation Criteria in Solid Tumors. All samples were obtained in accordance with the Helsinki declaration, and written informed consent was obtained.

Statistical analysis

All data are presented as mean±SD. Statistical analysis was performed using Student’s t-test or one-way analysis of variance. Correlation analysis was conducted using Pearson’s correlation test. Survival data were analyzed using the Kaplan-Meier method and log-rank test. GraphPad statistical software (V.9.0) was used for all statistical analyses. Unless otherwise specified, all data were analyzed using two-tailed tests, and p<0.05 was considered statistically significant.

A detailed description of the methods used in this study is found in the online supplemental methods.

ResultCRISPR screening identified Gsk3a as a critical gene for immune evasion of HCC

To identify key genes regulating immune adaptability in HCC, we constructed an immune-related gene library consisting of 12,000 sgRNAs targeting 2,796 genes. The Hepa1-6 cells transduced with the library were, respectively, subcutaneously implanted into C57BL/6 (immunocompetent) and NPG mice (immunodeficient). After 2 weeks, the tumors were harvested from mice and subjected to high-throughput sgRNA sequencing (figure 1A). To reduce errors and improve the accuracy of the results, we excluded two samples because of low quality in next-generation sequencing (online supplemental figure S1A,B). Before implantation (day 0), the library expression of tumor cells followed a log-normal distribution. After implantation, significant changes in sgRNA expression were observed in the tumor tissues from both C57BL/6 and NPG mice models (online supplemental figure S1C). We ranked the negative selection scores obtained by the MAGeCK RRA algorithm after normalization. A normalized score >2 in the negative selection was defined as a promoting factor (figure 1B). The genes that could promote growth in both NPG and C57BL/6 mice were considered as oncogenes (red), while those that could only promote growth in C57BL/6, but not in NPG, were considered as immune evasion genes (blue) (figure 1C). Finally, we identified and characterized functionally important molecular targets for immune evasion in HCC cells (online supplemental table S2).

Figure 1

CRISPR screening identified Gsk3a as a critical gene for tumor immune evasion. (A) CRISPR in vivo screening schematic diagram. (B) Negative selection analysis of sgRNA abundance in transplanted tumors and control cells. Normalized score >2 was defined as a promoting factor. (C) Based on cross-validation analysis of negative selection in NPG mice and C57BL/6 mice, a group of oncogenes (red) and immune evasion genes (blue) can be identified. (D) STRING+MCODE identified core regulatory molecules in the interaction-regulated network of the immune evasion gene set. (E) Bar charts depicting the messenger RNA expression levels of GSK3A, GRB2, MAVS, IRAK1, and ILK in the TCGA database. (F) Overall survival analysis of GSK3A, GRB2, MAVS, IRAK1, and ILK expression in patients with liver cancer using TCGA database. (G) Correlation heatmap of GSK3A, GRB2, MAVS, IRAK1, and ILK expression with immune cell infiltration in the TIMER database. (H) Growth curves of stable knockdown cell lines of different genes in subcutaneous xenografts of C57BL/6 (n=5) and NPG mice (n=5). Data were presented as means±SD. **p<0.01; ****p<0.0001; ns, p≥0.05. MDSC, myeloid-derived suppressor cell; TIMER, Tumor Immune Estimation Resource; TCGA, The Cancer Genome Atlas; Treg, regulatory T cells.

We conducted STRING analysis on the gene set associated with immune evasion and employed the Molecular Complex Detection (MCODE) plugin to identify core regulatory genes (figure 1D). Given that our screening was performed at the mouse genome level, we further wanted to explore whether these genes were significantly altered at the level of human transcription and protein expression. Five genes (GSK3A, GRB2, MAVS, IRAK1, ILK) were selected for further investigation, which were validated to be overexpressed in the human HCC from The Cancer Genome Atlas (TCGA) transcriptome and The Human Protein Atlas proteome, and associated with a worse prognosis (figure 1E,F and online supplemental figure S1D). Immuno-estimation indicated a close relationship between the five enriched genes and altered immune cells component in the TME (figure 1G). We then knocked down the five genes with shRNA in murine HCC Hepa1-6 cells and adopted the subcutaneous implantation tumor models in mice of different immune-background, and confirmed varying degrees of growth inhibition only in immunocompetent C57BL/6 mice, but not immunodeficient NPG mice (figure 1H and online supplemental figure S2A–C). Among them, interfering Gsk3a exhibited the most significant inhibitory effect on tumor growth. Taken together, these findings indicate that Gsk3a is a critical gene for immune evasion of HCC.

The immune evasion effects of Gsk3a required the involvement of the tumor immune microenvironment

We then knocked-down (KD) or overexpressed (OE) Gsk3a in Hepa1-6 cells (sh-Gsk3a or Gsk3a-OE), and confirmed an impaired or increased tumor growth only in immunocompetent C57BL/6 mice (online supplemental figure S3A–C). In line with the unaltered tumor growth within an immunodeficient background, Gsk3a KD or OE also affected no proliferation, migration or apoptotic capacity of Hepa1-6 cells in vitro (figure 2A–C).

Figure 2

Gsk3a is associated with neutrophil infiltration and T-cell functional suppression. (A) In vitro cell growth curve of sh-Gsk3a or Gsk3a-OE Hepa1-6 cells. (B) Representative Transwell images showing the migration of sh-Gsk3a or Gsk3a-OE Hepa1-6 cells. (C) Representative histogram of annexin V positive sh-Gsk3a or Gsk3a-OE Hepa1-6 cells. (D–E) Representative images and bar plots of the percentage of immune cells in the tumor immune microenvironment analyzed by flow cytometry. For (D) Percentage of MDSC (Gr1+Ly6g−) and neutrophils (Gr1+Ly6g+) in the microenvironment. For (E) Percentage of CD4+ and CD8+ T cells in the microenvironment. (F–I) Flow cytometric analysis of the frequency of IFN-γ+ cells in CD8+ T cells (F) GZMB+ cells in CD8+ T cells (G) PD-1+ cells in CD8+ cells (H) and LAG3+ cells in CD8+ cells (I) isolated from TILs. (J) Representative immunofluorescence staining of CD8+ T-cell (green), GZMB+ cells (pink) and neutrophils (red) in vector controls and sh-Gsk3a Hepa1-6 tumors. Scare bar: 50 µm. All data were presented as means±SD. *p<0.05; **p<0.01; ****p<0.0001; ns, p≥0.05 MDSC, myeloid-derived suppressor cell; GZMB, granzyme B; IFN-γ, interferon-gamma; KO, knocked-down; MDSC, myeloid-derived suppressor cell; OE, overexpressed; PD-1, programmed cell death protein-1; TILs, tumor infiltrating lymphocytes.

We then assessed the difference of immune cells component in TME between murine tumor of Hepa1-6 sh-Gsk3a and control (pLKO.1) by flow cytometry. Only neutrophils (Gr+Ly6g+) were significantly decreased in the sh-Gsk3a group. There was no changes in the infiltration of the other immune cells, such as T cells (CD4+ and CD8+), myeloid-derived suppressor cells (MDSCs) (Gr+Ly6g−), macrophages (CD11b+F4/80+), B cells (CD45+B220+), natural killer (NK) cells (CD45+NK1.1+), or regulatory T cells (Tregs) (CD25+Foxp3+) (figure 2D and online supplemental figure S4A–D). Despite that the total number of T cells was not affected (figure 2E), a higher proportion of functional T cells (IFN-γ+CD8+ and GZMB+CD8+) companied with a conversely decreased exhausted T cells (PD-1+ and LAG3+) was observed in the sh-Gsk3a group compared with the control (figure 2F–I). The results suggested that Gsk3a KD enhanced the cytotoxic function of CTLs, but this was independent of altering the ratio of T-cell populations. Immunofluorescence also confirmed a positive correlation between Gsk3a expression and neutrophils along with a negative correlation between Gsk3a and effector T cells, respectively (figure 2J). These results uncovered an altered TME by Gsk3a with a reverse change of infiltrated neutrophils and functional T cells.

Gsk3a inhibited T-cell function by inducing neutrophil infiltration and chemotaxis

To assess how Gsk3a reforms the immunosuppressive TME, we established an in vitro co-culture system consisting of CD8+ T cells and neutrophils either alone or in combination in upper chamber, and Hepa1-6 cells in lower chamber, in which T-cell killing efficiency and leukocyte chemotaxis were observed simultaneously. The tumor-killing efficiency of T cells alone was not affected by Gsk3a KD, indicating that tumorous Gsk3a did not act directionally on T cells. However, when neutrophils were added, Hepa1-6 cells exhibited enhanced resistance to T-cell cytotoxicity, and this was impaired by Gsk3a KD (figure 3A). Gsk3a KD also reduced the tumor cells’ ability to recruit neutrophils without affecting T cells recruitment (figure 3B,C). Flow cytometry analysis revealed no significant difference in the proportion of functional T-cell subsets between Hepa1-6 sh-Gsk3a and the control group when T cells were added alone. However, when neutrophils were added to T cells, the proportion of functional T-cell subsets decreased in both, but less in the sh-Gsk3a group (figure 3D,E). For validation, we repeated the co-culture system using human T cells, neutrophils and two common human HCC cell lines Hep3B and MHCC-97H with different GSK3A level (online supplemental figure S5A), and observed consistent changes in neutrophil chemotaxis and cytotoxic T-cell killing efficiency after intervening GSK3A in corresponding human cell lines (online supplemental figure S5B–D).

Figure 3

Gsk3a inhibits T-cell function by inducing neutrophil chemotaxis. (A–E) In vitro chemotaxis co-culture cytotoxicity experiment. In vitro co-culture cytotoxicity system schematic and bar graphs depicting the survival of underlying tumor cells measured by CCK-8 assay (A). Crystal violet staining of the Transwell membrane in the co-culture cytotoxicity system shows the chemotaxis of CD8+ T cells (B) and neutrophils (C). Scare bar: 100 µm. Flow cytometric analysis of the frequency of IFN­-γ+ cells in CD8+ T cells (D) and GZMB+ cells in CD8+ T cells (E) harvested from the lower chamber. (F) Tumor weight and images of transplanted Gsk3a-OE Hepa1-6 cells in C57BL/6 mice following the treatment of α-Ly6g or isotype antibody (n=5). (G–H) Flow cytometric analysis of the frequency of neutrophils (G) and CD8+ T cells (H) from tumor immune microenvironment. (I–J) Flow cytometric analysis of the frequency of IFN­-γ+ cells in CD8+ T cells (I) and GZMB+ cells in CD8+ T cells (J) from tumor immune microenvironment. All data were presented as means±SD. *p<0.05; **p<0.01; ***p<0.001; ns, p≥0.05. CM, conditioned media; GZMB, granzyme B; IFN-γ, interferon-gamma; KO, knocked-down; OE, overexpressed.

Consistently, depleting neutrophils by α-Ly6g abrogated the promoting effects of Gsk3a-OE Hepa1-6 cells on tumor growth in vivo (figure 3F,G). Depleting neutrophils did not affect CD8+ T-cell infiltration but increased the proportion of functional T cells in the TME (figure 3H–J). These findings suggest that neutrophils play a critical role in mediating the Gsk3a-induced suppressive TME in HCC.

Tumor cells with altered Gsk3a expression can affect neutrophil self-chemotaxis and NETs formation

We conducted RNA sequencing (RNA-seq) on neutrophils treated with conditioned media (CM) from sh-Gsk3a or Gsk3a-OE Hepa1-6 cells to investigate how neutrophils are reshaped by tumorous Gsk3a, and found a significant change with 1,731 genes upregulated and 1,990 genes downregulated (online supplemental figure S6A). Gene Ontology analysis showed enrichment of inflammatory activation-related gene sets, including inflammatory response, neutrophil chemotaxis, and response to oxidative stress (online supplemental figure S6B). Notably, the neutrophil chemotaxis gene set was enriched, suggesting a potential self-amplifying chemotactic effect induced by Gsk3a (figure 4A and online supplemental figure S6B). We validated the upregulation of chemotactic genes (Cxcl1, Cxcl2, Cxcl3) in neutrophils with Gsk3a-OE Hepa1-6 cells CM (figure 4B). Neutrophils treated with CM from Gsk3a-OE Hepa1-6 cells exhibited enhanced recruitment compared with that from sh-Gsk3a ones (figure 4C).

Figure 4

Tumor cells with altered Gsk3a expression can affect neutrophil self-chemotaxis and NETs formation. (A) The heatmap shows an upregulated expression of the neutrophil chemotaxis gene set in neutrophils cultured with CM from Gsk3a-OE Hepa1-6 cells. (B) Cxcl1, Cxcl2 and Cxcl3 mRNA in murine neutrophils co-cultured with sh-Gsk3a and Gsk3a-OE Hepa1-6 cells for 24 hours were analyzed by qPCR. (C) Migration of murine neutrophils recruited by CM from sh-Gsk3a or Gsk3a-OE Hepa1-6 cells, or from neutrophils pretreated with indicated Hepa1-6 CM. (D) The heatmap shows an upregulated expression of NETs associated pathway gene set in neutrophils cultured with CM from Gsk3a-OE Hepa1-6 cells. (E) Representative immunofluorescence staining of NETs (labeled in CitH3) and neutrophils (labeled in MPO) in sh-Gsk3a or Gsk3a-OE Hepa1-6 tumors. Scare bar: 50 µm. (F) Representative immunofluorescence staining of NETs (labeled in SytoxGreen) and DNA (labeled in DAPI) in neutrophils cultured with Hep3B and MHCC-97H cells CM in vitro. Scare bar: 50 µm. (G) Representative immunofluorescence images demonstrate NETs (labeled in SytoxGreen) enveloping tumor cells (labeled in DAPI and >10 µm), impeding their contact with T cells (labeled in Dil). Scare bar: 20 µm. (H) Bar plots show the expression of Cd274 and Elane in RNA sequencing. (I) Representative immunofluorescence images show the expression of PD-L1 and NE within neutrophil-NETs structures. Scare bar: 10 µm. (J) The cell viability of pLKO.1 and sh-Gsk3a Hepa1-6 cells in the co-culture system with neutrophils and CD8+ T cells was assessed using the CCK-8 assay, with or without DNase I treatment. All data were presented as means±SD. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001; ns, p≥0.05.CM, conditioned media; KO, knocked-down; mRNA, messenger RNA; NE, neutrophil elastase; NET, neutrophil extracellular trap; OE, overexpressed; PD-L1, programmed death-ligand 1; qPCR, quantitative PCR.

Tumor-associated neutrophils (TANs) display diversity and can be categorized as antitumor (N1) or protumor (N2) phenotypes, but Gsk3a did not bias neutrophils toward neither N1 nor N2 (online supplemental figure S6C). However, Nos2, a key encoding nitric oxide synthase involved in NETs formation, was upregulated (online supplemental figure S6D).19 Indeed, tumorous Gsk3a increased NETs-related genes expression in neutrophils (figure 4D). Consistently, CM from murine or human HCC cells with sh-Gsk3a and Gsk3a-OE impaired or increased NETs formation from homologous neutrophils in vitro, respectively. And this was also confirmed by increased NETs formation in sh-Gsk3a or Gsk3a-OE Hepa1-6 subcutaneous tumors in vivo (figure 4E,F and online supplemental figure S6E). Our previous studies have demonstrated the crucial role of NETs in promoting HCC metastasis.20 To further illustrate how NETs contribute to immune invasion, we co-cultured neutrophils with Gsk3a-OE Hepa1-6 cells followed by T cells challenge to allow NETs formation, and found Gsk3a-induced NETs encasing tumor cells, hindering their interaction with cytotoxic CD8+ T cells (figure 4G). Moreover, RNA-seq and immunofluorescence staining revealed that NETs induced by Gsk3a-OE CM were equipped with elevated programmed death-ligand 1 (PD-L1) and less cytotoxic neutrophil elastase (NE),21 which further enhanced the immunosuppressive capacity (figure 4H,I). Finally, DNase I treatment degraded NETs, effectively rescuing the impaired tumor cells killing efficiency of T-cell induced by tumorous Gsk3a (figure 4J). We observed the same effect in human cell lines (online supplemental figure S6F).

Gsk3a promotes recruitment and NETs formation of neutrophil through leucine-rich α-2-glycoprotein 1

To investigate cellular communication of how tumorous Gsk3a impact neutrophils, we performed RNA-seq on sh-Gsk3a and pLKO.1 Hepa1-6 cells. We found that the expression of secreting factor Lrg1 was most significantly downregulated in sh-Gsk3a cells (figure 5A). Interestingly, classic neutrophil and immune cell chemokines exhibited no significant changes (online supplemental figure S7A). To confirm the correlation between Gsk3a and Lrg1 expression, we quantified Lrg1 expression intracellularly and extracellularly using qPCR and ELISA, showing reduced Lrg1 expression with Gsk3a (figure 5B,C). Moreover, the protein expression of leucine-rich α-2-glycoprotein 1 (LRG1) was also downregulated both in murine and human HCC cells by western blot (online supplemental figure S7B,C). Consistently, it was further confirmed that Lrg1 was induced by Gsk3a through transcription factors nuclear factor kappa B (NFκB) and signal transducer and activator of transcription 3 (STAT3) (figure 5D).22 23 Lrg1 has been reported to induce neutrophil chemotaxis and amplify the effect through autocrine secretion.24 To our expectation, in vitro addition of recombinant LRG1 restored the expression of neutrophil self-amplifying chemokines reduced in neutrophils treated with sh-Gsk3a CM (figure 5E). The impaired neutrophils chemotaxis and NETs formation after interfering Gsk3a/GSK3A in murine Hepa1-6 and human Hep3B cells were also restored by LRG1 (figure 5F,G and online supplemental figure S7D,E). In the HCC cells-T cells-neutrophils co-culture system, LRG1 addition restored the impaired survival of sh-Gsk3a tumor cells to the control level, which was nullified by DNase I digestion of NETs (figure 5H and online supplemental figure S7F). Flow cytometry analysis of CD8+ T cells transmigrating into the lower chamber revealed that LRG1 reduced the proportion of cytotoxic T cells (IFN-γ+CD8+T cell and GZMB+CD8+T cell), which was reversed by DNase I (figure 5I). In vivo, the addition of LRG1 increased tumor volume and weight of sh-Gsk3a tumors, which was counteracted by DNase I co-administration (figure 5J). DNase I reversed T-cell function inhibited by LRG1, while the increased the neutrophils infiltration induced by LRG1 and total T cells were not affected (figure 5K and online supplemental figure S7G,H).

Figure 5

Gsk3a promotes recruitment and NETs formation of neutrophil through LRG1. (A) The volcano plot of differentially expressed genes in sh-Gsk3a versus pLKO.1 Hepa1-6 cells. (B) qPCR validation of the differentially expressed gene Lrg1 in Hepa1-6 cells. (C) ELISA validation of LRG1 protein expression in Hepa1-6 cell culture supernatant. (D) Western blot validation of Gsk3a phosphorylation on the NFκB pathway and STAT3 pathway. (E) qPCR validation showed a concentration-dependent upregulation of chemokine expression (Cxcl1, Cxcl2, Cxcl3, Csf1, Csf2) with the supplementation of exogenous LRG1 protein. (F) Transwell assays demonstrated that reconstitution with recombinant LRG1 protein restored the reduced neutrophil chemotaxis caused by Gsk3a knockdown. Scare bar: 100 µm. (G) The representative immunofluorescence images showed that reconstitution with recombinant LRG1 protein restored the reduced NETs formation caused by Gsk3a knockdown. Scare bar: 100 µm. (H) The viability of sh-Gsk3a Hepa1-6 cells in the co-culture system of neutrophils and CD8+ T cells with recombinant LRG1 protein or combined DNase I treatment. (I) Flow cytometric analysis of the frequency of IFN­-γ+ cells in CD8+ T cells and GZMB+ cells in CD8+ T cells from the lower chamber after treatment with recombinant LRG1 protein or combined DNase I treatment. (J) In vitro images of sh-Gsk3a Hepa1-6 subcutaneous tumors in C57BL/6 mice treated with recombinant LRG1 protein or in combination with DNase I. (K) Flow cytometric analysis of the frequency of IFN-­γ+ cells in CD8+ T cells and GZMB+ cells in CD8+ T cells from tumor immune microenvironment. All data were presented as means±SD. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001; ns, p≥0.05.CM, conditioned media; qPCR, quantitative PCR; GZMB, granzyme B; IFN-γ, interferon-gamma; KO, knocked-down; LRG1, leucine-rich α-2-glycoprotein 1; mRNA, messenger RNA; NETs, neutrophil extracellular traps; OE, overexpressed; PBS, phosphate-buffered saline; NFκB, nuclear factor kappa B; STAT3, signal transducer and activator of transcription 3.

Blocking GSK3A enhances the therapeutic efficacy of anti-PD-1 monoclonal antibody treatment

Immunohistochemical staining on a tissue microarray containing 23 patients with HCC samples collected from Huashan Hospital was performed (figure 6A). The results showed that GSK3A was positively correlated with LRG1, MPO, CitH3, and GZMB, but not associated with CD8+ T-cell infiltration (figure 6B). Transcriptomic data from the TCGA database confirmed that both low expression of GSK3A and NETs-score had the best prognosis (figure 6C).