Introduction

Primary liver cancer is the fourth leading cause of cancer death worldwide.1 Hepatocellular carcinoma (HCC) accounts for the majority of primary liver cancers and has a very poor prognosis.2 3 Recently, immune checkpoint blockade (ICB) therapies, such as programmed death receptor 1 (PD-1) antibody therapy and programmed death ligand 1 (PD-L1) antibody therapy, have been approved therapy for patients with HCC.4–6 However, only limited survival benefits for patients with HCC have been observed owing to tumor immune evasion.7 8 Thus, it is meaningful to gain insight into the mechanism of tumor immune evasion to improve ICB therapy.

PD-L1 is often overexpressed on the surface of cancer cells.9 The binding of PD-L1 and PD-1 inhibits T-cell proliferation and activation, suppresses the function of cytotoxic T cells, and ultimately induces immune escape and tumor development, leading to immunotherapy failure.10 11 Clinically, PD-L1 expression level in tumors is an important biomarker to evaluate the efficacy of patients for ICB therapies, but currently, studies report that both PD-L1-positive and PD-L1-negative patients show promising responses to immunotherapy.12–14 Emerging evidence has shown that downregulating the PD-L1 expression in tumors can improve the efficacy of anti-PD-1 therapy.15–18 Increased PD-L1 expression is frequently used by tumor cells as a strategy to evade antitumor immune responses.19 Therefore, exploring the mechanism of aberrant PD-L1 expression in tumor is crucial to resolve the conflicting observations and enhance clinical responses to ICB therapy.

Emerging evidence has indicated that glycosylation of immune receptors and ligands plays a critical role in cancer immunity.20 Protein glycosylation consists mainly of N-glycosylation and O-glycosylation.21 Currently, the role of N-glycosylation-related genes in tumor immune escape has drawn increasing attention22–24; however, the functions of O-glycosylation-related genes in cancer progression remain unclear. Fucosylation is a type of glycosylation modification and fucosylated glycan structures are commonly present on the cell surface. There are 13 fucosyltransferase genes in the human genome. Protein O-fucosyltransferases (POFUTs) are located in the endoplasmic reticulum and catalyze O-linked fucosylation.25 Emerging evidence has suggested that the expression of POFUT1 is elevated in several human cancers,26 27 however, whether and how POFUT1 plays an essential role in HCC progression and immunosuppressive microenvironment remains unclear.

In this study, we have identified a previously unknown functional link between POFUT1 and immune evasion in HCC. POFUT1 stabilizes PD-L1 protein by preventing the tripartite motif containing 21 (TRIM21)-mediated degradation and ubiquitination, leading to HCC progression and immune microenvironment suppression. From a therapeutic viewpoint, our findings revealed that inhibition of POFUT1 synergized with anti-PD-1 therapy in the HCC mouse model. Thus, this study provides a potential therapeutic approach for HCC immunotherapy.

Materials and methods

Plasmids and reagents

PT3-EF1a-MYC-IRES-luciferase (MYC), px330-sg-p53 (sg-p53) and CMV-SB13 were kindly provided by Amaia Lujambio, Icahn School of Medicine at Mount Sinai. DNA fragments of POFUT1 and PD-L1 were cloned into MYC by replacing the luciferase sequence to generate MYC-POFUT1 and MYC-PD-L1. Additional information related to plasmids and reagents can be found in the online supplemental information.

Cell dissociation and flow cytometry analysis

Mouse tumor tissues were harvested and immediately placed into a centrifuge tube on ice. The tumor tissue was then washed with phosphate-buffered saline and cut into small pieces using scissors. Collagenase I was used for enzymatic digestion with constant shaking at 37°C for 20–30 min. The enzymatic reaction was subsequently terminated, and the sample was filtered through a 70 micron cell strainer to remove undissociated tissue fragments. For flow cytometry, approximately 1×10ˆ5 to 1×10ˆ6 cells per sample were stained and subsequently analyzed with a BD Fortessa flow cytometer (BD Biosciences).

Clinical samples and immunohistochemistry

The patient samples used in this study were 90 liver cancer tissue specimens collected between 2014 and 2018 from Shanghai Renji Hospital, affiliated with Shanghai Jiao Tong University School of Medicine, including 72 men and 18 women with an age range of 33–83 years. Tissue microarrays were prepared by Shanghai Zuocheng Bio company. Informed consent was obtained from each patient before sample collection. All research was conducted according to the principles outlined in the Helsinki declaration and the Istanbul declaration, and the use of these samples and informed consent were approved by the Ethics Committee of Renji Hospital, Shanghai Jiao Tong University School of Medicine (RA-2020–250). One case was excluded from statistical analysis due to sample detachment during tissue sectioning.

Immunohistochemical (IHC) staining was conducted in accordance with established protocols. Specifically, the process included deparaffinization and rehydration, followed by antigen retrieval using citrate buffer on paraffin-embedded sections. Subsequently, the sections were blocked and incubated with the corresponding primary antibody. Detection was achieved through the application of the 3,3’-diaminobenzidine substrate.

For estimating the number of positive cells: a box with an area of 0.04 mm2 was drawn using Aperio software, and the number of positive cells was counted at 20× magnification. The number was then multiplied by 25 to estimate the number per mm2.

For tissue microarrays, the extent of staining was quantified using a “Staining Area (%)” score (0=no positive staining; 1=1–25%; 2=26–50%; 3=51–75%; 4=76–100%), and staining intensity was evaluated with an “Intensity Score” (0=no staining; 1=weak; 2=mild; 3=moderate; 4=strong; 5=intense).

Animal experimentsHydrodynamic tail-vein injection

For the experiments to detect the functionality of POFUT1, we prepared a 2 mL solution of 0.9% sodium chloride containing 12 µg of pT3-EF1a-MYC-IRES-luciferase (MYC), 10 µg of px330-sg-p53 (sg-p53), 12 µg of a mixture of two lenti-CRISPR sgPofut1 (or 12 µg of control plasmid lenti-CRISPR V2), and 6 µg of the transposon SB13 transposase-encoding plasmid. For the experiments to detect the overexpression functionality of POFUT1, we prepared a 2 mL solution of 0.9% sodium chloride containing 12 µg of pT3-EF1a-MYC-POFUT1 (or 12 µg of MYC), 10 µg of sg-p53, and 6 µg of the transposon SB13 transposase-encoding plasmid. For the experiments to detect the functionality of PD-L1, the first group was prepared with a 2 mL solution of 0.9% sodium chloride containing 12 µg of pT3-EF1a-MYC-PD-L1 (or 12 µg of control plasmid MYC), 10 µg of sg-p53, 12 µg of a mixture of two lenti-CRISPR sgPofut1 (or 12 µg of control plasmid lenti-CRISPR V2), and 6 µg of the transposon SB13 transposase-encoding plasmid. The second group was prepared with a 2 mL solution of 0.9% sodium chloride containing 12 µg of pT3-EF1a-MYC-POFUT1 (or 12 µg of control plasmid MYC), 10 µg of sg-p53, 12 µg of a mixture of two lenti-CRISPR sgPdl1 (or 12 µg of control plasmid lenti-CRISPR V2), and 6 µg of the transposon SB13 transposase-encoding plasmid. The mice were intravenously injected with the 0.9% sodium chloride/plasmid mixture through the tail vein. The injection volume was 10% of their body weight, and the injection was completed within 5–7 s.

Tumor xenograft mouse models

Subcutaneous injections of 1×10ˆ6 Hepa1-6 cells were administered into the inguinal region of the mice. The order of subcutaneous tumor measurement is randomized. After approximately 2 weeks, mice were euthanized using cervical dislocation, and the tumors were excised for subsequent analysis. Tumor volume was calculated using the formula length×width2/2.

For the PD-1 treatment experiment, intraperitoneal injection of anti-PD-1 antibody (100 µg per mouse, Bio X Cell, BE0146) or IgG isotype control was performed every 3 days for a total of four injections. For depletion of CD8+ T cells in vivo, mice were intraperitoneally injected with anti-CD8α antibodies (200 µg per mouse, Bio X Cell, BE0061) or IgG isotype control 2 days before tumor implantation and two times per week thereafter for a total of five injections to ensure sustained depletion of CD8+ T cells during the experimental period.

Male C57 mice (5–6 weeks old, obtained from www.jh-labanimal.site/) were used in this study. Experimental animals that fail to receive successful hydrodynamic tail vein injection or subcutaneous tumor implantation, or die prematurely before the designated time after successful modeling (injecting the hydrodynamic tail vein too quickly can lead to the death of the mouse, occurring within 6 hours after injection, unrelated to tumor occurrence) are not included in the experimental statistics.

Each group of experimental animals has similar body weights, and the animals are randomly assigned to groups using a computer-based random sequence generator. Measurements are randomized in order. During tumor harvesting, one researcher erases grouping information, and another researcher who is unaware of the grouping is responsible for collecting and processing the samples as well as subsequent data analysis. All animal experiments were conducted in accordance with relevant guidelines and regulations and received approval from the Animal Experimental Committee of Renji Hospital, Shanghai Jiao Tong University School of Medicine (RJ20220720).

Statistical analysis

The statistical analysis was performed using GraphPad Prism V.9 software. Flow cytometry data were analyzed using FlowJo_V.10 software. The biological informatics analysis was conducted using the R language (V.4.1.1). Gene Set Enrichment Analysis (GSEA) was conducted using either R language or GSEA software. Quantification of western blot and IHC was performed using ImageJ software.

Data availability

RNA sequencing (RNA-seq) data are accessible at the NODE under accession number: OEP004799 (https://www.biosino.org/node/project/detail/OEP004799). GSE22058, GSE14520 and ICGC_JP data sets were downloaded from the HCCDB website (http://lifeome.net/database/hccdb/home.html).28 The Cancer Genome Atlas Liver hepatocellular carcinoma (TCGA_LIHC) data set was obtained through the UCSC website (https://xenabrowser.net/datapages/). The patients with Chinese HCC with hepatitis B virus (HBV) infection (CHCC_HBV) cohort was downloaded from the CPTAC website (https://proteomics.cancer.gov/programs/cptac).29 Tumor immune scores and microenvironment scores were evaluated and downloaded from the TIMER V.2.0 website (http://timer.cistrom.org/).30 The immune subtypes were obtained from Bagaev et al.31 The leukocyte and stromal fractions and the scores of lymphocyte infiltration, Th1 cells, Th2 cells and proliferation were obtained from Thorsson et al.32 The gene sets of expanded interferon (IFN)-G signature, a good response associated genes 1 and up in responders to anti-cytotoxic T-lymphocyte associated protein-4+anti-PD-1 were obtained from Cerezo-Walli et al.33 The genes related to IFN-G signaling, effector T-cell signaling, chemokine and antigen presentation pathway were obtained from Newell et al.34 All data used in this study are publicly available for download. All data are available on request from the authors.

Additional detailed Methods are included in online supplemental information.

ResultsPOFUT1 is upregulated in HCC and correlates with immunosuppressive microenvironment

To identify the O-glycosylation-related genes that play an important role in HCC immunosuppressive microenvironment, we first explored the genes both upregulated in TCGA_LIHC and CHCC_HBV_2019 (CHCC_HBV) cohorts (logFC>0.5, p value<0.05) and found that 29 of them were O-glycosylation-related genes (figure 1A). Second, we estimated the relationship between the 29 candidate genes and tumor immune microenvironment by xCell method in TCGA_LIHC and found the top enrichment of POFUT1 in the poor immune score (figure 1B). Similar results were obtained in the CHCC_HBV and three other HCC databases (HCCB1_GSE22058, HCCB6_GSE14520 and HCCB18_ICGC_JP) (figure 1C,D and online supplemental figure S1A). Moreover, we further examined the microenvironment score for the identified candidate genes and also found high enrichment of POFUT1 in the context of a poor microenvironment score (figure 1E and online supplemental figure S1B). Due to its substantial enrichment, POFUT1 was selected for further investigation into its functional link with HCC immunosuppressive microenvironment. We further investigated whether POFUT1 could regulate the infiltration and response of immune cells. We first performed the correlation analysis between POFUT1 expression and scores of immune cells based on xCell algorithm, and found that POFUT1 expression was negatively correlated with the infiltration of many antitumor-related immune cells in TCGA_LIHC (figure 1F). Additionally, immune-related analyses were performed on the POFUT1 high and low expression groups within the TCGA_LIHC cohort. The result indicated a decrease in the leukocyte fraction, and the scores of lymphocyte infiltration and Th1 cells were also diminished. In contrast, the scores of Th2 and proliferation were increased in the POFUT1 high-expression group (figure 1G). Moreover, we analyzed the tumor immune subtype proportion between POFUT1 high and low expression groups in the TCGA_LIHC cohort and found that POFUT1 high expression group tended to exhibit more D (immune-depleted) and F (fibrotic) phenotypes, conversely, more IE (immune-enriched/non-fibrotic) and IE/IF (immune-enriched/fibrotic) phenotype were found in the POFUT1 low expression group (figure 1H). GSEA also showed that immune-related pathways were enriched in POFUT1 low expression group (online supplemental figure S1C). We next analyzed the correlation between POFUT1 expression and the IFN-G signaling, effector T-cell signaling, chemokine and antigen presentation pathway in TCGA_LIHC and CHCC_HBV cohorts, and the universal negative relationship was found in the POFUT1 and above factors (figure 1I). Additionally, GSEA revealed that the expanded IFN-G signature was enriched in the POFUT1 low expression group (online supplemental figure S1D). Overall, these results suggest that POFUT1 high expression may contribute to the immunosuppressive microenvironment in HCC and the immune escape of tumor cells.

Figure 1

POFUT1 is highly expressed in HCC and is associated with the immunosuppressive microenvironment. (A) The Venn diagram showed that there were 29 genes upregulated in both TCGA_LIHC (n=369) and CHCC_HBV (n=159) (tumor vs normal, p value<0.05, log fold change >0.5) that intersected with the gene set GOBP_PROTEIN_O_LINKED_GLYCOSYLATION. (B–C) Volcano plot showing the distribution of immune score differences for 29 candidate genes grouped by expression level (high vs low) in TCGA_LIHC (n=369) and CHCC_HBV (n=159). (D–E) Volcano plot showing the distribution of immune score and microenvironment score differences for 29 candidate genes grouped by expression level (high vs low) in HCCB18_ICGC_JP (n=212). (F) The correlation heatmap for POFUT1 expression and scores of different types of immune cells in the HCC data sets (based on xCell algorithm). The Spearman coefficient was used to evaluate correlations. (G) Leukocyte fraction and stromal fraction, scores of lymphocyte infiltration, scores of Th1 and Th2 cells and score of proliferation with high or low POFUT1 in TCGA_LIHC (n=369). Data represent mean±SD. P values were calculated using Student’s t-test. (H) Distribution of immune subtypes with high or low POFUT1 in TCGA_LIHC (n=369). (I) Heatmap of correlation between POFUT1 expression and IFN-G signaling pathway, effector T-cell signaling pathway, chemokine pathway and MHC-I pathway in TCGA_LIHC (n=369) and CHCC_HBV (n=159). The Pearson coefficient was used to evaluate correlations. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. CHCC, Chinese HCC; D, immune depleted; F, fibrotic; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; IE, immune-enrichment/non-fibrotic; IE/IF, immune-enrichment/fibrotic; IFN, interferon; MHC, major histocompatibility complex; POFUT1, protein O-fucosyltransferase 1; TCGA_LIHC, The Cancer Genome Atlas Liver Hepatocellular Carcinoma.

Downregulation of POFUT1 inhibits HCC progression in vivo and improves tumor immunosuppressive microenvironment

To investigate the role of POFUT1 in HCC progression, we first constructed the Pofut1 knockdown cell line in mouse hepatoma cell Hepa1-6. The levels of Pofut1 protein and messenger RNA (mRNA) were determined by western blot and quantitative PCR, respectively (online supplemental figure S2A,B). Then, we injected the Hepa1-6 cells subcutaneously into C57BL/6 mice to confirm the function of Pofut1 in vivo. Genetic knockdown of Pofut1 expression significantly inhibited liver tumor growth in C57BL/6 mice (figure 2A,B). Next, we generated a mouse MYC/Trp53−/− HCC model35 36 following hydrodynamic tail vein injections of MYC, sg-p53 and CMV-SB13 vectors. We delivered a Pofut1 targeting sgRNA expressing vector (or control vector) in combination with MYC, sg-p53 and CMV-SB13 (figure 2C). sgPofut1 mice exhibited less aggressive HCC development 35 days after hydrodynamic injection, showing fewer tumors and smaller tumor diameters, compared with sgCtrl mice (figure 2D,E). Moreover, Pcna positive cells were significantly reduced in the sgPofut1 and shPofut1 tumors compared with the control group (figure 2F and online supplemental figure S2C). Thus, these findings demonstrate the crucial role of POFUT1 in promoting HCC growth in vivo.

Figure 2

POFUT1 knockdown inhibits HCC progression and improves tumor immunosuppressive microenvironment. (A–B) Representative tumor images, (B) tumor weight and tumor volume of subcutaneous Hepa1-6 tumors (n=6). Scale bar, 1 cm. Data presented is the median with the range from six biological replicates. (C–E) Schematic outline showing sgCtrl and sgPofut1 MYC/Trp53−/− HCC models by hydrodynamic plasmids injection. (D) Representative images of livers (n=6) from the indicated de novo mouse liver cancer models. (E) Tumor number per liver and max tumor diameter were measured. Scale bar, 1 cm. (F) Representative images of H&E, immunohistochemical Pofut1 and Pcna staining in the tumor tissues of mouse livers shown in (A) and (C). Scale bar, 50 µm. (G) Gating strategies for analysis of the tumor immune microenvironment by flow cytometry. (H) Relative proportions of lymphoid cells (CD45+), T cells (CD45+CD3+), CD8+ T cells (CD45+CD3+CD8+) and CD4+ T cells (CD45+CD3+CD4+) in tumor tissues were analyzed by flow cytometry (n=5). (I) Numbers of lymphoid cells (CD45+), T cells (CD45+CD3+), CD8+ T cells (CD45+CD3+CD8+) and CD4+ T cells (CD45+CD3+CD4+) per mg of tumor weight (cells per mg) (n=5). (J) Gating strategies for analysis of cytotoxic function of CD8+ T cells by flow cytometry. (K) Relative proportions of IFN-G+ and TNFA+ of CD8+ T cells in tumor tissues were analyzed by flow cytometry (n=5). (L) PD-1 expression on CD8+ T cells in tumor tissues was analyzed by flow cytometry (n=5). Data represent mean±SD. P values were calculated using Student’s t-test. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. HCC, hepatocellular carcinoma; IFN, interferon; MFI, median fluorescence intensity; PD-1, programmed death receptor 1; POFUT1, protein O-fucosyltransferase 1; TNF, tumor necrosis factor.

To assess the effects of POFUT1 on the immune microenvironment of HCC, mouse liver tumor tissues were collected for flow cytometry assay to characterize the impact of POFUT1 on different types of immune cells (figure 2G, online supplemental figure S2D,E). We found that the proportions and numbers of CD45+ cells (lymphoid cells), CD45+ CD3+ cells (T cells), CD8+ T cells and CD4+ T cells were significantly increased in the sgPofut1 mouse group (figure 2H and I). There were no differences in the proportions of natural killer (NK) and natural killer T (NKT) cells (online supplemental figure S2F). We further examined the characteristics of CD8+ T cells, and found that the cytotoxic function of CD8+ T cells was enhanced in sgPofut1 group, as evidenced by elevated production of cytotoxic molecules IFN-G and tumor necrosis factor alpha (TNFA) in the tumor-infiltrating CD8+ T cells (figure 2J and K). Moreover, the PD-1 expression on CD8+ T cells was lower in sgPofut1 group than in sgCtrl group (figure 2L and online supplemental figure S2G). In addition, immunofluorescence assays showed that the infiltration of CD8+ T cells into the tumor tissues was increased in sgPofut1 group (online supplemental figure S2H). Collectively, these results suggest that POFUT1 downregulation inhibits HCC progression by increasing immune cell infiltration and inhibiting CD8+ T-cell exhaustion.

POFUT1 overexpression promotes HCC progression and exacerbates immunosuppression

To further investigate the role of POFUT1 gain-of-function in vivo, we interrogate the effect of POFUT1 overexpression in MYC/Trp53−/− HCC model. POFUT1 gene was cloned into PT3-EF1A-MYC-IRES-luciferase by replacing the luciferase sequence, referred to as MYC-POFUT1 (figure 3A). Indeed, POFUT1 overexpression mouse liver tumor group exhibited higher liver weight, liver/body weight ratio, tumor number, max tumor diameter and tumor burden (figure 3B–3D and online supplemental figure S3A). Moreover, Pcna positive cells were also significantly increased in MYC-POFUT1 tumors (online supplemental figure S3B). In addition, we also harvested spleens from mice and the results showed that spleen weight and spleen/body weight ratio in MYC-POFUT1 group were higher than those in MYC group (online supplemental figure S3C). These findings suggest that overexpression of POFUT1 facilitates HCC growth in vivo by regulating the immune microenvironment.

Figure 3

POFUT1 promotes HCC development and exacerbates immunosuppression. (A–D) Schematic outline showing MYC and MYC-POFUT1 MYC/Trp53–/– HCC models by hydrodynamic plasmids injection. (B) Representative images of livers (n=8) from the indicated de novo mouse liver cancer models. (C) Liver weight, liver/body weight ratio, (D) tumor number per liver, max tumor diameter and tumor burden determined by the ratio of tumor area/liver area were measured. Scale bar, 1 cm. (E) Representative images of H&E, immunohistochemical Pofut1 and Pcna staining in the tumor tissues of mouse livers shown in (A). Scale bar, 50 µm. (F) Gating strategies for analysis of the tumor immune microenvironment by flow cytometry. (G) Relative proportions of lymphoid cells (CD45+), T cells (CD45+CD3+), CD8+ T cells (CD45+CD3+CD8+) and CD4+ T cells (CD45+CD3+CD4+) in tumor tissues were analyzed by flow cytometry (n=5). (H) Numbers of lymphoid cells (CD45+), T cells (CD45+CD3+), CD8+ T cells (CD45+CD3+CD8+) and CD4+ T cells (CD45+CD3+CD4+) per mg of tumor weight (cells per mg) (n=5). (I) Relative proportions of IFN-G+ and TNFA+ of CD8+ T cells in tumor tissues were analyzed by flow cytometry (n=5). (J) PD-1 expression on CD8+ T cells in tumor tissues was analyzed by flow cytometry (n=5). (K) Gene Set Enrichment Analysis of signature of genes upregulated in comparison of naive CD8 T cells versus PD-1 high CD8 T cells in MYC-POFUT1 and MYC groups in RNA sequencing. The gene set “NAÏVE_VS_PD1HIGH_CD8_TCELL_UP” is available in the C7 IMMUNESIGDB collection (GSE26495_NAIVE_VS_PD1HIGH_CD8_TCELL_UP) at http://www.msigdb.org. Data represent mean±SD. P values were calculated using Student’s t-test. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. FDR, false discovery rate; HCC, hepatocellular carcinoma; IFN, interferon; MFI, median fluorescence intensity; NES, normalized enrichment score; PD-1, programmed death receptor 1; POFUT1, protein O-fucosyltransferase 1; TNFA, tumor necrosis factor alpha.

Next, we collected liver tumors to revalidate the impact of POFUT1 on different types of immune cells (figure 3F and online supplemental figure S3D). The flow cytometry analysis also showed that the proportions and numbers of CD45+ cells, T cells, CD8+ T cells and CD4+ T cells in the MYC-POFUT1 group were lower than those in the MYC group (figure 3G,H). There were no differences in the proportions of NK and NKT cells (online supplemental figure S3E). Furthermore, we found that the production of cytotoxic molecules IFN-G and TNFA was decreased in the tumor-infiltrating CD8+ T cells of MYC-POFUT1 group (figure 3I). Moreover, the PD-1 expression on CD8+ T cells was higher in MYC-POFUT1 group than in MYC group (figure 3J and online supplemental figure S3F). We also observed overexpression of POFUT1 resulted in a significant decrease in CD8+ T cells infiltration (online supplemental figure S3G). Then we performed RNA-seq on liver tumor tissues from the MYC-POFUT1 and MYC groups, and GSEA analysis also revealed that the naive_versus_PD-1 high_CD8_T cell_upregulated gene set was significantly enriched in the MYC group (figure 3K), and the IFN response signaling gene sets were similarly enriched in the MYC group (online supplemental figure S3H). Together, these results suggest that POFUT1 promotes HCC progression and immune evasion.

POFUT1 interacts with PD-L1 to enhance its protein stability by inhibiting its ubiquitination and degradation

We analyzed RNA-seq data for differentially expressed gene analysis between the two groups (MYC-POFUT1 vs MYC) to explore the mechanism underlying POFUT1-mediated HCC immune evasion. We identified 180 differentially expressed genes using the intersections between the differentially expressed and immune-related genes in a Venn diagram. Then we performed the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of these genes and the most significantly enriched pathway related to immunity was the T-cell receptor (TCR) signaling pathway (figure 4A). GSEA also revealed that the TCR signaling pathway was significantly enriched in the MYC group (figure 4B). As described above, POFUT1 could upregulate the expression of PD-1 on CD8+ T cells in HCC and since the binding of PD-1 to its ligand, PD-L1, could inhibit TCR signaling pathway,9 37 we wondered whether POFUT1 inhibits TCR signaling pathway by upregulating PD-L1 expression in liver tumor cells. As expected, overexpression of POFUT1 could upregulate PD-L1 protein level in the Hepa1-6, Huh7 and non-HCC HEK293T cells (figure 4C and online supplemental figure S4A). Conversely, the knockdown of POFUT1 could downregulate the protein level of PD-L1 in the Hepa1-6, Huh7 and HEK293T cells (figure 4D and online supplemental figure S4B). Given the critical role of membrane PD-L1 in tumor immune escape,10 we further detected the PD-L1 level of cytomembrane using flow cytometry. We found that Pofut1 knockdown resulted in the downregulation of PD-L1, with a more significant decrease observed following IFN-G-induced PD-L1 expression (figure 4E and online supplemental figure S4C). Moreover, we also found the upregulation of membrane PD-L1 level in Hepa1-6 cells after overexpression of POFUT1 (online supplemental figure S4D). In addition, we detected the expression level of PD-L1 in vivo by flow cytometry, and found tumor cells in the sgPofut1 group showed lower PD-L1 expression level, while those in the MYC-POFUT1 group exhibited higher PD-L1 expression (figure 4F and online supplemental figure S4E). Similarly, we examined the PD-L1 level in the liver tumor tissues from our animal models by IHC and found POFUT1 protein level showed a positive correlation with PD-L1 protein level (online supplemental figure S4F-H). These results demonstrate that POFUT1 could upregulate PD-L1 protein expression in HCC cells and liver tumor tissues. To further verify the relationship between POFUT1 and PD-L1 expression in human HCC tissues, IHC staining was performed on 90 HCC samples to analyze the protein levels of POFUT1 and PD-L1. In HCC tissues, POFUT1 protein level showed a positive correlation with PD-L1 protein level (figure 4G and online supplemental figure S4I). These results indicate the clinical relevance of POFUT1 and PD-L1 expression.

Figure 4

POFUT1 interacts with PD-L1 to enhance its protein stability. (A) Venn diagram depicting the overlap of genes containing immune genes and the differential expressed genes (DEG) in MYC-POFUT1 versus MYC in RNA-seq (left). Common genes for KEGG pathway analysis (right). The gene set “immune-gene” is available at https://www.immport.org/home. (B) Gene Set Enrichment Analysis of T-cell receptor signaling pathway in MYC-POFUT1 and MYC groups in RNA-seq. (C) Western blot analysis of the levels of PD-L1 and POFUT1 in control and POFUT1 overexpression Hepa1-6 and Huh7 cells. (D) Western blot analysis of the levels of PD-L1 and POFUT1 in control and POFUT1 knockdown Hepa1-6 and Huh7 cells. (E) Expression level of PD-L1 on the cell membrane in Hepa1-6 cells with Pofut1 knockdown, with or without 24 hours incubation with IFN-G (n=3). (F) Expression level of PD-L1 on the cell membrane in tumor cells in vivo by flow cytometry (n=5). (G) The correlation between PD-L1 and POFUT1 IHC scores by correlation coefficient analysis in human HCC microarray (n=89). The Pearson coefficient was used to evaluate correlations. (H) Western blot analysis of PD-L1 protein in control and Pofut1 knockdown Hepa1-6 cells (top) and in control and POFUT1 overexpression Huh7 cells (bottom) treated with cycloheximide (CHX, 100 µg/mL). (I) Co-IP assay analyzed the interaction of endogenous PD-L1 with heterogenous expressed POFUT1-FLAG in HEK293T cells. (J) Co-IP assay analyzed the interaction of endogenous POFUT1 with heterogenous expressed PD-L1-FLAG in HEK293T cells. (K) In vitro GST pulldown assay analysis of the interaction of GST-PD-L1 and POFUT1. Black arrows indicated the Comassie blue staining of GST and GST-PD-L1. (L) Immunofluorescence assay showed PD-L1 co-localized with POFUT1 in the liver tumor tissue. Scale bar, 5 µm. (M) Western blot analysis of the levels of PD-L1 and POFUT1 in control, POFUT1-WT and POFUT1-MUT Hepa1-6 and Huh7 cells. (N–P) Ubiquitination assay of PD-L1 in HEK293T cells transfected with the indicated plasmids. Ubiquitinated PD-L1 was immunoprecipitated and subjected to western blot analysis with an antibody against ubiquitin. For E and F data represent mean±SD. P values were calculated using Student’s t-test. **p<0.01; ***p<0.001; ****p<0.0001. Co-IP, co-immunoprecipitation; FDR, false discovery rate; HCC, hepatocellular carcinoma; IFN, interferon; KEGG, kyoto encyclopedia of genes and genomes; MUT, mutant; NES, normalized enrichment score; PD-L1, programmed death ligand 1; POFUT1, protein O-fucosyltransferase 1; RNA-seq, RNA sequencing; WT, wild-type.

The upregulation of PD-L1 protein level may result from the activation of PD-L1 transcription and/or enhanced PD-L1 protein stability. First, we examined the mRNA level of PD-L1 in Hepa1-6 cells; however, the quantitative PCR showed that POFUT1 did not activate PD-L1 transcription (online supplemental figure S4J). Next, we wondered whether POFUT1 could enhance the stability of PD-L1 protein, the stability of PD-L1 was explored using the translation inhibitor cycloheximide, and we found that knockdown of POFUT1 could shorten the half-life of PD-L1 protein in the Hepa1-6 cells, moreover, ectopic expression of POFUT1 had an opposite result in the Huh7 cells (figure 4H, online supplemental figure S4K,L). These results indicate that POFUT1 enhances the stability of PD-L1 protein.

Next, we wondered whether POFUT1 might interact with PD-L1 to promote its stability. To this end, using the co-immunoprecipitation (Co-IP) assays, the interaction of endogenous PD-L1 with FLAG-POFUT1 in HEK293T cells was verified (figure 4I). Similarly, the interaction of endogenous POFUT1 with FLAG-PD-L1 was also found in HEK293T cells (figure 4J). Furthermore, GST pulldown assays also indicated that POFUT1 interacted with PD-L1 (figure 4K). Moreover, immunofluorescence assay revealed that endogenous PD-L1 co-localized with POFUT1 in the liver tumor tissue (figure 4L) and Huh7 cells (online supplemental figure S4M). Given POFUT1 is a key enzyme in O-fucose biosynthesis and O-fucosylation is a type of protein post-translational modification, we investigated the role of its enzymatic activity in PD-L1 upregulation. We first constructed POFUT1 wild-type (POFUT1-WT) plasmid and POFUT1 mutant (POFUT1-MUT) plasmid with Arg240 changed to Ala, which lost its enzyme activity.38 Western blot analysis showed that ectopic expression of POFUT1-MUT still could upregulate PD-L1 protein level in Hepa1-6, Huh7 and HEK293T cells (figure 4M and online supplemental figure S4N). In addition, we further detected the level of PD-L1 on the cell membrane and found a similar result in Hepa1-6 cells (online supplemental figure S4O). These results demonstrate that POFUT1 promotes PD-L1 protein stability independently of its protein-O-fucosyltransferase activity.

Given that the degradation of PD-L1 protein mainly involves ubiquitin proteasome or lysosome pathways,10 39 we wondered whether POFUT1 promotes the stability of PD-L1 protein by inhibiting its ubiquitination. To test this hypothesis, Co-IP assays were conducted using the anti-FLAG monoclonal antibody (mAb), followed by the detection of the ubiquitin-conjugated FLAG-PD-L1. Overexpression of POFUT1 markedly reduced the level of the ubiquitin-conjugated FLAG-PD-L1 (figure 4N). Conversely, knockdown of POFUT1 could increase the level of ubiquitin-conjugated FLAG-PD-L1 (figure 4O). In addition, we further confirmed that the role of POFUT1 inhibit the ubiquitination of PD-L1 protein independently of its protein-O-fucosyltransferase activity (figure 4P). These results showed that POFUT1 regulates PD-L1 expression via the ubiquitination pathway.

Together, these data suggest that POFUT1 upregulates the stability of PD-L1 by inhibiting its ubiquitination and protein degradation.

POFUT1 upregulates PD-L1 expression via preventing TRIM21-mediated PD-L1 ubiquitination

Given that POFUT1 promotes the protein stability of PD-L1 by inhibiting its ubiquitination, we wondered whether POFUT1 could impair some E3 ligase-mediated PD-L1 ubiquitination and degradation. To