Blocking PCSK9 suppresses hepatocellular carcinoma immune escape by decreasing FLI1-mediated SPP1 and PD-L1 expression

Abstract

Background

Proprotein convertase subtilisin/kexin type 9 (PCSK9) is a newly identified immunosuppressive regulator, but its mechanism of suppressing antitumor immunity remains ambiguous. This study aims to uncover the underlying mechanism by which PCSK9 promotes hepatocellular carcinoma (HCC) immune escape and to explore potential intervention strategies.

Methods

Co-culture assay assessed the cytotoxicity of CD8⁺ T cells against PCSK9-knockout HCC cells. Hepa1-6, H22, and HepG2 cells were used to establish HCC mouse models. Tumor microenvironment changes were evaluated using flow cytometry and single-cell RNA sequencing. Additionally, we developed a CRISPR adenine base editing (ABE) base editor and screened small molecule inhibitors for PCSK9 inhibition in HCC treatment.

Results

We found that PCSK9 was highly expressed and correlated with poor survival in patients with HCC. While PCSK9 deficiency did not affect HCC growth in vitro, it significantly enhanced CD8⁺ T cell-mediated selective killing in vitro and in vivo. This selective killing of PCSK9-deficient HCC cells could not be explained by existing theories related to major histocompatibility complex-I and T-cell receptor (TCR) degradation. Instead, our study revealed that PCSK9 knockout inhibited the expression of secreted phosphoprotein 1 (SPP1) and programmed death-ligand 1 (PD-L1) in HCC cells, and identified friend leukemia virus integration 1 (FLI1) as their co-transcription factor. Overexpression of FLI1 reversed the PCSK9 knockout-induced downregulation of SPP1 and PD-L1, thereby promoting HCC immune escape. Furthermore, PCSK9 upregulated FLI1 expression through the neurogenic locus notch homolog protein 3 (NOTCH3) pathway. Additionally, we designed an all-in-one ABE base editor with thyroxine-binding globulin promoter (ABE-TBG-PCSK9) to knock down PCSK9 and identified parecoxib as a small molecule inhibitor. We confirmed both approaches enhanced CD8⁺ T-cell antitumor activity, significantly inhibiting HCC tumor growth and prolonging mouse survival.

Conclusions

PCSK9 promoted HCC immune escape by upregulating SPP1 and PD-L1 via NOTCH3/FLI1 signaling. CRISPR ABE-mediated PCSK9 deficiency and PCSK9 inhibitor parecoxib may serve as effective strategies to inhibit HCC.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• In recent studies, proprotein convertase subtilisin/kexin type 9 (PCSK9) is highly specifically expressed in the liver and is also a biomarker for the treatment of liver cancer. It has been found that the absence of PCSK9 can enhance the antitumor effects of CD8⁺ T cells. However, the mechanisms reported so far cannot explain the specific killing effect of CD8⁺ T cells on tumor cells with PCSK9 deficiency. Meanwhile, the development of small molecule inhibitors targeting PCSK9 is also urgent.

WHAT THIS STUDY ADDS

• Here, we uncovered the mechanism by which CD8⁺ T cells selectively kill PCSK9-deficient tumor cells. Results showed that PCSK9 knockout suppressed the expression of secreted phosphoprotein 1 (SPP1) and programmed death-ligand 1 (PD-L1) by repressing their co-transcription factor friend leukemia virus integration 1 (FLI1), thereby enhancing CD8⁺ T cell-mediated cytotoxicity. Furthermore, PCSK9 upregulated FLI1 expression through the neurogenic locus notch homolog protein 3 (NOTCH3) pathway. Additionally, we developed a hepatocyte-specific all-in-one adenine base editing with thyroxine-binding globulin promoter (ABE-TBG-PCSK9) base editor and identified parecoxib as an inhibitor of PCSK9; both strategies significantly reverse hepatocellular carcinoma (HCC) immune escape. Our study highlights the potential of PCSK9 as a promising target for anti-HCC immunotherapy and presents two effective strategies for PCSK9 inhibition.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study provides experimental evidence for the application of PCSK9 in liver cancer treatment. Our findings indicate that the absence of PCSK9 enhances the CD8⁺ T-cell cytotoxicity by downregulating SPP1 and PD-L1 expression through the NOTCH3/FLI1 signaling pathway. In addition, the study highlights the feasibility of the all-in-one ABE-TBG-PCSK9 base editor as a targeted therapy. The study may also promote the development of PCSK9 small-molecule inhibitors to overcome the limitations of monoclonal antibody drugs and improve the treatment of liver cancer. This provides valuable information for clinical combination with other immune modulators.

Background

Hepatocellular carcinoma (HCC) is the second leading cause of cancer-related death worldwide. Current guidelines recommend surgical resection for early-stage HCC, while immune checkpoint inhibitors (ICIs) and vascular endothelial growth factor receptor (VEGFR) inhibitors are the first-line treatments for advanced HCC.^{1 2} Anti-programmed cell death protein-1 (PD-1) antibodies have achieved positive outcomes in the clinic, while only 20–30% of patients respond to anti-PD-1 and a large proportion of patients remain unresponsive.3,5 The immunosuppressive microenvironment has been reported to be critical in CD8⁺ T-cell dysfunction, immune escape and nonresponse to ICIs in patients with HCC, indicating the presence of alternative mechanisms of immune evasion.⁶

Proprotein convertase subtilisin/kexin type 9 (PCSK9) is a low-density lipoprotein receptor (LDLR) binding protein that promotes LDLR degradation in lysosomes, leading to increased low-density lipoprotein (LDL) levels in the blood and results in hyperlipidemia.⁷ Its inhibitors, evolocumab and alirocumab, have been approved for hyperlipidemia treatment.^{8 9} PCSK9 exhibits a broad expression profile, with the highest levels in the liver.¹⁰ Elevated PCSK9 expression is linked to poor prognosis in various cancers, and its knockdown has been shown to suppress liver metastasis of melanoma.⁷ PCSK9 has also been reported to promote tumor cell epithelial-mesenchymal transition and induce M2 macrophage polarization in tumor microenvironment (TME).^{11 12} Several studies have demonstrated that knockout of PCSK9 in mouse colon, breast, and melanoma cancer cells significantly reduced tumor growth and enhanced CD8⁺ T cell-mediated cytotoxicity in vivo.12,14 Notably, PCSK9 inhibitors evolocumab and alirocumab have been shown to synergize with anti-PD-1 therapy in breast cancer cell mouse models, suggesting that PCSK9 acts as a novel tumor-promoting and immunosuppressive regulator.¹⁵

In this study, we conducted both in vitro and in vivo HCC cell competition experiments, and all results consistently demonstrated that CD8⁺ T cells exert selective cytotoxicity against PCSK9-deficient HCC cells within the same co-culture wells or the same TME. Currently, two mechanisms have been reported for PCSK9’s inhibition of antitumor immunity: first, PCSK9 secreted by tumor cells binds to major histocompatibility complex (MHC)-I on the tumor cell membrane, promoting MHC-I endocytosis and degradation, which reduces antigen presentation and hinders CD8⁺ T-cell recognition; second, PCSK9 binds to the T-cell receptor (TCR) on CD8⁺ T cells in the TME, facilitating TCR endocytosis and degradation, thus impairing CD8⁺ T-cell function.^{15 16} However, in the competition experiments, both control and PCSK9 knockout (PCSK9-KO) tumor cells coexisted, with control cells also secreting PCSK9 into the TME. This secreted PCSK9 could indiscriminately lead to MHC-I degradation on both control and knockout tumor cells, as well as TCR degradation on CD8⁺ T cells. According to existing theories, this would result in overall suppression of CD8⁺ T-cell antitumor activity, and it is unlikely to produce a stronger selective killing effect on PCSK9-KO tumor cells. Therefore, these existing theories cannot explain the results of tumor competition growth experiments, suggesting the presence of other and more important mechanisms underlying PCSK9’s inhibition of CD8⁺ T-cell antitumor activity.

This study aims to reveal the underlying mechanism by which PCSK9 promotes HCC immune escape and to explore potential intervention strategies. We uncovered a novel mechanism underlying PCSK9 that drives antitumor immunosuppression that is distinct from existing theories. PCSK9 facilitates neurogenic locus notch homolog protein 3 (NOTCH3)/friend leukemia virus integration 1 (FLI1)-mediated upregulation of secreted phosphoprotein 1 (SPP1) and programmed death-ligand 1 (PD-L1) transcription, fostering PCSK9-expressing HCC cell immune escape. Additionally, we developed a hepatocyte-specific all-in-one adenine base editing (ABE) base editor and identified a small molecule inhibitor for PCSK9 inhibition, both of which potently enhance the efficacy of anti-HCC immunotherapy.

Methods

Cell culture

Mouse liver cancer cell lines H22 and Hepa1-6, human liver cancer HepG2 cells, and 293FT cells were purchased from the American Type Culture Collection and identified using short tandem repeat analysis. Hepa 1–6, HepG2, and 293FT cells were cultured in Dulbecco’s Modified Eagle Medium (Gibco) with 10% fetal bovine serum (FBS), 100 µg/mL penicillin, and 100 µg/mL streptomycin. H22 cells were maintained in Roswell Park Memorial Institute 1640 (Gibco) with 10% FBS, 100 µg/mL penicillin, and 100 µg/mL streptomycin. All cells were tested for Mycoplasma contamination using the Universal Mycoplasma Detection Kit (Vazyme, D101-01).

Animals and treatments

All animal experiments were approved by the Army Medical University Institutional Animal Use and Care Committee. C57BL/6 and Balb/c mice were purchased from Huafukang Biological Technology (Beijing, China). OT-1 transgenic mice (C57BL/6 background) were purchased from Cyagen company (Jiangsu, China). Alb-cre⁺ and Myc⁺ mice were obtained from GemPharmatech (Jiangsu, China) and crossed to generate Alb-cre⁺/Myc⁺ spontaneous liver cancer mice, with genotyping results shown in online supplemental figure S18. Mice were housed under controlled conditions at 25°C, 30–70% humidity, and a 12-hour light/dark cycle (08:00 to 20:00). Tumor models were established by injecting 6–8 weeks old mice subcutaneously with 1.0×10⁶ tumor cells. For all-in-one adenine base editing with thyroxine-binding globulin promoter (ABE-TBG-PCSK9) plasmid or PCSK9 inhibitor treatments, when tumors reached 300–500 mm³, plasmids encapsulated in Polyplus liposomes or parecoxib were administered intravenously or intraperitoneally, respectively. Tumor growth was monitored every other day, with size measured by caliper and calculated as volume=length×(width)²/2. In an experiment involving anti-PCSK9, about 200 µg of anti-PCSK9 monoclonal antibodies (evolocumab from Medchemexpress) were injected (intraperitoneally) on days 0, 2, 4 and 6. The endpoint was defined when tumor size reached 2,000 mm³ or if mice developed open skin lesions or lost more than 15% of their body weight.

Plasmid construction

Clustered regularly interspaced shot palindormic repeats/single-guide RNA (CRISPR/sgRNA) plasmid construction was performed based on established protocols for gene knockout.¹⁷ The sgRNA used in this study is listed in online supplemental table S1. Scramble-sgRNA or gene-targeting sgRNA was inserted into the LentiCRISPRv2 plasmid (Addegene, 52961) via BsmBI digestion. For gene overexpression, the coding sequences of PCSK9, FLI1, and LDLR were amplified by PCR and cloned into a lentiviral backbone (Addegene, 72263). For CRISPR-based VPR overexpression, the sgRNA of NOTCH3 was cloned into a lentiviral backbone (Addegene, 139090). For the luciferase plasmid, the promoter sequences of SPP1 and CD274 were amplified and incorporated into a pGL6 luciferase plasmid (Beyotime, D2102). For site mutation, the potential binding sites of FLI1 on SPP1 and CD274 promoters were mutated using the Fast Mutagenesis System (TRAN, FM111). All constructs were verified by Sanger sequencing. The all-in-one ABE-TBG-PCSK9 plasmid was generated by inserting the U6-driven gRNA scaffold and liver-specific TBG promoter-driven ABE 8.20.

Transient transfection and stable cell line construction

For transient transfection, H22 and Hepa 1–6 liver cancer cells were plated at a density of 1×10⁵ cells/well in a 12-well plate. After 24 hours, the cells were co-transfected with 1 µg of plasmid using Lipofectamine 3000 (Vazyme, TL301) according to the manufacturer’s instructions. After 8 hours of transfection, the cells were washed and cultured in fresh medium for 48 hours, then seeded into 96-well plates. Clones were expanded, and base editing was detected via Sanger sequencing.

For stable cell line construction, plasmids were co-transfected with lentivirus packaging plasmids VSVG (Addgene, 12259) and psPAX2 (Addgene, 12260) into HEK-293FT cells using Lipofectamine 3000 and OPTI-MEM media (Invitrogen, 31985070). Lentiviruses were harvested two times on day 3 and day 5, filtered with a 0.45 µm filter, and used to infect target liver cancer cells with 5 µg/mL polybrene (Sigma, H9268). After puromycin selection, surviving cells were collected, and knockout efficiency was confirmed by Sanger sequencing and western blotting.

Dual-luciferase reporter assay

HepG2 cells were seeded in 12-well plates (5×10⁵ cells per well) and transfected with the PCSK9-overexpression (PCSK9-OV) or control vector, along with the SPP1 or PD-L1 promoter-driven luciferase reporter plasmids. Renilla reporter plasmids were co-transfected as an internal control. After 24 hours transfection, cell lysates were harvested and analyzed using a dual-specific luciferase reporter assay system (Beyotime, RG027). Alternatively, FLI1 overexpression lentivirus-infected HepG2 cells were co-transfected with 0.5 µg of promoter-reporter plasmids and 0.5 µg of Renilla plasmids using Lipofectamine 3000. The same procedure was applied for mutation promoter plasmids.

CUT&Tag assay

The cleavage under targets and tagmentation (CUT&Tag) assay was performed according to the protocol of the Hyperactive Universal CUT&Tag Assay Kit (Vazyme, TD904). Briefly, 5×10⁵ cells were centrifuged in 1.5 mL eppendorf tube (EP) tubes at room temperature for 5 min at low speed of 600 g, and the supernatant was discarded. The 500 µL wash buffer was used to wash the cells two times, and the 100 µL wash buffer was used to resuspend the cells. The cells were transferred into eight tubes containing activated ConA Beads Pro, and incubated at room temperature for 10 min. After discarding the supernatant, 50 µL pre-cooled anti-FLI1 buffer was added. Immediately centrifuged and left overnight at 4°C. The next day, the second antibody was incubated for 1 hour. Then pA/G-Tnp Pro incubation was performed, TTBL fragmented DNA was used and Spike-in DNA was added. After processing the above samples with DNA Extract Beads Pro, DNA was extracted and used for PCR amplification. PCR products were examined by agarose gel electrophoresis.

Single-cell sequencing

Wild type (WT) and PCSK9-KO H22 liver cancer cells (1×10⁶) were subcutaneously injected into the right flank of Balb/c mice. 14 days later, three independent tumors were dissected and digested with collagenase II, IV (0.25%), and DNase I (0.05%) in 1640 medium for 30 min at 37°C. Lymphocytes were isolated using a 40% and 70% Percoll gradient, followed by CD45⁺ cell sorting via flow cytometry. A total of 1×10⁷ cells from each group were processed using the 10x Genomics Chromium Platform. Raw data were mapped to the mouse genome (mm10-3.0.0) with Cell Ranger. Downstream analyses were performed in R (V.4.3.3) using Seurat (V.5.0.3) following standard guidelines. Seurat’s default settings were applied for normalization, principal component analysis, t-distributed stochastic neighbor embedding, and clustering.¹⁸ Differential expression was analyzed using the Wilcoxon rank-sum test, with a fold change threshold of 1.0 and an adjusted p value cut-off of 0.05. Cell type-specific genes were used to annotate the clusters, as shown in online supplemental figure S8. Cell–cell communication analysis was performed using the CellChat (https://github.com/sqjin/CellChat) pipeline with its open-source R package.¹⁹ Tumor-infiltrating T cells were performed by ProjecTILs.²⁰

In vitro CD8⁺ T-cell killing assay

CD8⁺ T cells were isolated from the spleens of Balb/c and C57BL/6 mice using EasySep CD8 Magnetic Isolation Kit (STEMCELL, 19853), followed by activating with anti-CD3 and anti-CD28 for 2 days, and cultured for an additional 2 days. WT and PCSK9-KO H22 and Hepa1-6 cells (5×10³) were stained with calcein-AM and seeded into 96-well plates overnight. CD8⁺ T cells were then added to cancer cell cultures at different ratios, with interleukin-2 to maintain CD8⁺ T-cell activity. 72 hours later, the number of living cancer cells was counted under a fluorescence microscope and normalized to the control group. For the OT1-OVA killing assay, OT1⁺ CD8⁺ T cells were isolated from the spleens of OT1 mice and stimulated with 1 µg/mL OVA peptide for 3 days. WT and PCSK9-KO H22-OVA and Hepa1-6-OVA cells (5×10³) were stained with calcein-AM or expressed tdTomato, seeded into 96-well plates overnight in media containing 1 ng/mL interferon (IFN)-γ, and co-cultured with OT1⁺ CD8⁺ T cells for 24 hours. Viable cancer cells were then counted under a fluorescence microscope.

In vivo competition assay

Control and PCSK9-KO H22 cells stably expressed enhanced green fluorescent protein (EGFP) or tdTomato, respectively, and selected with puromycin for 1 week. Subsequently, about 1×10⁶ PCSK9-KO cells (tdTomato expressing) and 1×10⁶ control cells (EGFP expressing) were mixed and inoculated subcutaneously to Balb/c mice. Tumors were excised 7–21 days after inoculation. They were then cut into pieces and cells were passed through a 70 µm cell strainer (SPL Life Science, 93070). Cells were washed and resuspended in 2% FBS phosphate-buffered saline (PBS) (cold). The ratios of EGFP and tdTomato tumor cells were analyzed using the flow cytometer (Attune NxT, Thermo Fisher Scientific). The pieces of tumor tissue were removed and placed in a small box, dipped in an OCT embedding medium (Sakura, 4583), and slowly frozen in liquid nitrogen. The percentage of EGFP and tdTomato cells was observed under a fluorescence microscope after sectioning. For Sanger sequencing, the tumor cells were separated by a flow sorter (MoFlo XDP, Beckman Coulter).

RNA sequencing and analysis

RNA sequencing (RNA-seq) analyses were performed on three independent RNA samples from WT and PCSK9-KO H22 cells. RNA-seq libraries were generated from double-stranded complementary DNA synthesized from 10 ng of total RNA. Sequencing was performed by Shenzhen Huada Sequencing Center. Raw data were processed using the DESeq2 package in R Bioconductor for normalization and differential gene expression (DEG) analysis. Genes with an adjusted p value<0.05 and a fold change >1.0 were considered significant. Biological processes and Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling pathways were analyzed using the DAVID tool (https://david.ncifcrf.gov/).

Western blot analysis

Cells were harvested and lysed using lysis and Extraction Buffer (Beyotime, P0013). Protein concentration was determined with the BCA Protein Assay Kit (Beyotime, P0010). The samples were loaded on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE gels), followed by transfer to polyvinylidene difluoride (PVDF) membranes. Membranes were blocked with 5% bovine serum albumin (BSA) for 2 hours at room temperature and then incubated overnight at 4°C with primary antibodies against PCSK9, FLI1, SPP1, and PD-L1 (antibody details are listed in online supplemental table S2). After incubation with horseradish peroxidase (HRP)-linked secondary antibodies, the membranes were visualized using an ECL luminescent solution (Bio-Rad, 1705061) on a ChemiDoc Imaging System (Bio-Rad).

Quantitative real-time PCR

Total RNA was isolated from the liver cancer cells with the RNAiso (TAKARA, No.9109). The concentration of the RNA was measured by the microspectrophotometer (Nano-300, Allsheng). The primers used in this study are listed in online supplemental file 1. One-step real-time quantitative PCR (qPCR) was performed by the protocol (Vazyme, Q222). The program was run by the real-time PCR instrument (QuantStudio 3, Thermo Fisher Scientific).

Flow cytometry

After anesthetizing the mice, tumors were collected to prepare single-cell suspensions according to established protocols. Tumors from bearing mice were analyzed by flow cytometry when tumor size was around 500 mm³. Lymphocytes were isolated using a 40% and 70% Percoll gradient. Red blood cells were lysed by RBC lysis buffer (Beyotime, C3702) for 3 min. Cells were then cultured with a leukocyte activation cocktail (BD, 550583) and Monensin (BioLegend, 420701) for 4 hours to block cytokine secretion. Next, cells were fixed and permeabilized using the Cytofix/Cytoperm Fixation/Permeabilization Kit (BD, 554714) and incubated with antibodies in FACS buffer (PBS with 2% FBS) at 4°C in the dark for 30 min. Dead cells were excluded using the LIVE/DEAD Zombie UV Fixable Viability Kit (BioLegend, 423108). After washing two times with FACS buffer, cells were acquired on a BD LSRFortessa X-20 flow cytometer and analyzed using FlowJo software (V.10.6.2). Antibodies used for flow cytometry analysis are listed in online supplemental table S2.

Immunofluorescence

For slide preparation, the fixed tumor tissues of the mice were embedded, sectioned, dewaxed, and hydrated. The slides were blocked with 10% goat or donkey serum after epitope retrieval. Then, the slides were incubated with different primary antibodies overnight. The next day, slides were incubated with different fluorophore-labeled secondary antibodies were incubated. Finally, the slides were stained with 4’,6-diamidino-2-phenylindole (DAPI) and photographed under a fluorescence microscope (DM18, Leica).

Immunohistochemistry analysis

The immunohistochemistry (IHC) was performed as published protocol.²¹ The tumor or non-tumor tissues of patients with HCC and mice were subjected to IHC staining to detect the expression of PCSK9, FLI1, SPP1, and PD-L1. The antibodies used are listed in online supplemental table S2.

Off-target analysis

For the off-target analysis, potential off-targets were predicted by CasOT-1.0 and Cas-OFFinder (http://www.rgenome.net/cas-offinder/), allowing up to four mismatches in the target region. The primers for the potential off-targets are in online supplemental table S3. PCR products were amplified, purified with a PCR purification kit (QIAGEN, 28104) and analyzed by GBI (Beijing). The substitutions and indel frequencies were calculated by an R-based tool EditR (V.1.0.0) from Moriarity Lab (http://baseeditr.com/).

Datasets and bioinformatics analysis

The raw single-cell sequencing data (CRA019962) and RNA sequence data (CRA019963) have been deposited in the Genome Sequence Archive in National Genomics Data Center and are available at https://ngdc.cncb.ac.cn/gsa.²² The expressions of PCSK9 in pan-cancer were downloaded from Gene Expression Profiling Interactive Analysis. The Cancer Genome Atlas (TCGA) liver hepatocellular carcinoma datasets were downloaded from TCGA. GSE84005, GSE136319, and GSE183113 were obtained from the Gene Expression Omnibus (GEO) database. ALGGEN-PROMO (https://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dir DB=TF_8.3), hTFtarget (https://guolab.wchscu.cn/hTFtarget/#!/), and JASPAR (https://jaspar.elixir.no/) were used to predict the FLI1 binding sites on the promoter of SPP1 or PD-L1. HCC single-cell transcriptome data were obtained from the PRHCCdb database (https://db.cngb.org/PRHCCdb/).

Ethics approval

All animal study protocols were approved by the laboratory animal welfare and ethics committee of the Amy Medical University (AMUWEC20234895).

Statistical analysis

All values are presented as the mean±SEM. Each experiment was repeated at least three times with independent biological replicates. Statistical significance between two groups was determined using two-tailed unpaired Student’s t-tests. For comparisons among multiple groups, one-way or two-way analysis of variance was used. Survival curves were plotted using the Kaplan-Meier method, and significance was evaluated by the log-rank test. P values<0.05 were considered statistically significant. All statistical analyses were performed using GraphPad Prism V.8.0 software.

Results

PCSK9 deficiency in HCC cells activates CD8⁺ T-cell antitumor efficacy

To investigate the relationship between PCSK9 and patient with HCC prognosis, we first analyzed the RNA-seq data from the TCGA database and found that PCSK9 is highly expressed in HCC tissues compared with normal tissues, and is associated with poor overall survival in patients with HCC (online supplemental figure S1A-D). Moreover, we found that PCSK9 expression negatively correlates with cytotoxic lymphocyte infiltration into HCC tumors and positively correlates with known immunosuppressive molecules across TCGA patients with cancer (online supplemental figure S1E-H). We then knocked out (online supplemental figure S2) or overexpressed PCSK9 in mouse and human liver cancer cells, and found that PCSK9 knockout or PCSK9 overexpression did not affect the growth or apoptosis of HCC cells in vitro (online supplemental figure S3). However, when PCSK9 knockout cells were orthotopically or subcutaneously inoculated into immunocompetent Balb/c or C57BL/6 mice, their abilities to grow tumors were significantly suppressed compared with control cells (figure 1A–F). Furthermore, flow cytometry analyzed the tumor-infiltrating lymphocytes and confirmed that the number of tumor-infiltrating CD8⁺ T cells, the proportion of CD8⁺ T cells expressing IFN-γ and granzyme B (GZMB), and the ratio of cytotoxic CD8⁺ T lymphocytes/regulatory T (Treg) cells were significantly increased in PCSK9-KO H22 tumors (figure 1G,H and online supplemental figure S4). These increased CD8⁺ T-cell tumor infiltration and cytokine releases in PCSK9 knockout H22 tumors were further confirmed by immunofluorescence staining of tumor tissues (figure 1I). Our results validate that PCSK9 acts as an immunosuppressive molecule that negatively regulates antitumor immunity.

Figure 1. PCSK9 deficiency in HCC cells activates CD8⁺ T cell antitumor efficacy. (A–B) The bioluminescent images and tumor growth of control and PCSK9-knockout H22 cells in mouse orthotopic liver (n=7 in B). Nt-Ctrl: cells transfected with non-targeting control plasmids expressing CRISPR/Cas9 and scramble-sgRNA. (C–F) The tumor growth, bioluminescent images, tumor weight and overall survival of control and PCSK9-knockout H22 cells in Balb/c mice and Hepa 1–6 cells in C57BL/6 mice (n=10 in C, (E and F). (G) Flow cytometric analysis of the percentage of tumor-infiltrating CD8⁺ T cell (CD3⁺ CD8⁺), Treg (CD4⁺ Foxp3⁺), and CD8⁺ T cells expressing IFN-γ, GZMB (n=5). (H) The ratio of CTL to Treg cells. (I) Representative immunofluorescence images of CD8⁺ T, IFN-γ, and GZMB in tumors (n=3). Scale bar, 50 µm. Data are shown as mean±SEM. Statistics: two-way ANOVA (B and D), log-rank test (F), two-tailed unpaired student’s t-test (E, G, H and I). ANOVA, analysis of variance; CRISPR, clustered regularly interspaced shot palindromic repeats; CTL, cytotoxic CD8⁺ T lymphocytes; DAPI, 4’,6-diamidino-2-phenylindole; GZMB, granzyme B; HCC, hepatocellular carcinoma; IFN, interferon; KO, knockout; PCSK9, proprotein convertase subtilisin/kexin type 9; sgRNA, single-guide RNA; Treg, regulatory T cell.

CD8⁺ T cells selectively kill PCSK9-deficient HCC cells in vitro and in vivo

To ascertain the effects of the PCSK9-KO on the killing effects of CD8⁺ T cells, we performed co-culture assays combining PCSK9 knockout with either anti-PCSK9 treatment or LDLR overexpression. The results showed that anti-PCSK9 slightly enhanced CD8⁺ T-cell killing; however, it did not further enhance the cytotoxic effect of PCSK9 knockout (online supplemental figure S5A). Similarly, LDLR overexpression did not further enhance the killing effect of PCSK9-deficient cells, indicating that the enhanced killing effect is independent of LDLR (online supplemental figure S5B). To further confirm this selective killing effect on PCSK9-KO HCC cells, we performed competition experiments using dual fluorescently labeled HCC cells in vitro. Control H22 cells expressing EGFP (H22-WT-EGFP) and PCSK9-KO H22 cells (with a base G missing) expressing tdTomato (H22-KO-tdTomato) were mixed at a 1:2 ratio and co-cultured with CD8⁺ T cells for 72 hours. Results showed that CD8⁺ T cells exhibited selective cytotoxic effect on PCSK9-KO H22 cells, as evidenced by a marked reduction in the tdTomato⁺/EGFP⁺ cell ratio (figure 2A–D). This competition experiment was also conducted in vivo, where a 1:1 ratio of H22-WT-EGFP and H22-KO-tdTomato cells was mixed and subcutaneously inoculated into Balb/c mice. Fluorescence staining of tumor tissues at different time intervals revealed that the fluorescence intensity of EGFP-expressing control H22 cells gradually increased, while the fluorescence intensity of tdTomato-expressing PCSK9-KO H22 cells was progressively decreased, and even undetectable after 21 days of tumor growth (figure 2E–G). Furthermore, Sanger sequencing of sorted EGFP or tdTomato-positive HCC cells from tumors also confirmed that PCSK9-deficient HCC cells (with a base G missing) were more sensitive to host antitumor immune killing (figure 2H,I). These findings collectively suggest that CD8⁺ T cells exert a selective cytotoxic effect on PCSK9-deficient HCC cells.

Figure 2. CD8⁺ T cells selectively kill PCSK9 deficiency HCC cells in vitro and in vivo. (A) Schematic of the in vitro CD8⁺ T-cell competition experiment. (B–D) H22-WT-EGFP and H22-KO-tdTomato cells were 1:2 ratio mixed and co-cultured with different ratios of CD8⁺ T cells. The number of living cells was measured by fluorescence (n=3) and flow cytometry (n=6). (E–G) Schematic of the in vivo competition assay. H22-WT-EGFP and H22-KO-tdTomato cells were mixed in a 1:1 ratio, then subcutaneously injected into Balb/c mouse flanks. The fluorescence changes were determined at different time intervals (n=3). Scale bar, 75 µm. (H) Sanger sequencing of the H22-WT-EGFP and H22-KO-tdTomato cells before inoculation (n=3). (I) Sanger sequencing of tumor cells expressing EGFP or tdTomato separated by flow sorter (n=3). Statistics: one-way ANOVA (C, D, F, G and I), two-tailed unpaired Student’s t-test (H). ANOVA, analysis of variance; EGFP, enhanced green fluorescent protein; HCC, hepatocellular carcinoma; KO, knockout; PCSK9, proprotein convertase subtilisin/kexin type 9; WT, wild type.

ScRNA-seq revealed that PCSK9-KO activates CD8⁺ T cells in TME

Since the competition experiment results could not be explained by existing theories, we employed single-cell RNA sequencing (scRNA-seq) to investigate the key mechanism by which PCSK9 inhibits antitumor immunity. We first analyzed scRNA-seq data (GEO140228) from tumor and adjacent normal tissues of patients with HCC. Cells were annotated based on specific marker genes of different cell types (online supplemental figure S6A and B). The cell distribution analysis revealed a significantly lower proportion of effector CD8⁺ T cells in HCC tumors compared with adjacent normal tissues (online supplemental figure S6C). Additionally, ProjecTILs analysis also consistently observed a notable increase in exhausted CD8⁺ T cells in HCC tissues compared with normal tissues (online supplemental figure S6D). These findings suggest that HCC tissues are immunosuppressive microenvironments characterized by a high number of exhausted CD8⁺ T cells. To further investigate whether PCSK9 contributes to immunosuppression in patients with HCC, we divided the scRNA-seq data into two groups (PCSK9_high and PCSK9_low) based on PCSK9 expression in tumor cells. Results showed that samples with high PCSK9 expression had fewer tumor-infiltrating effector CD8⁺ T cells and natural killer cells (online supplemental figure S7A and B). Moreover, ProjecTILs analysis confirmed that high PCSK9 expression was associated with CD8⁺ T-cell exhaustion in HCC tissues (online supplemental figure S7C). Additionally, IHC and immunofluorescent (IF) staining of patients with HCC tumors also observed that areas with low PCSK9 expression exhibited higher CD8⁺ T-cell infiltration and abundant dead tumor cells (online supplemental figure S8).

To further explore the mechanisms underlying the immunosuppressive role of PCSK9, we performed scRNA-seq to analyze the control H22 and PCSK9-KO H22 transplanted tumors, and cells were annotated by specific marker genes (online supplemental figure S9A and B). The cell distribution analysis revealed that compared with the control group, tumor-infiltrating exhausted CD8⁺ T cells were markedly reduced, and CD8⁺ T cells, CD8⁺ T memory cells were increased in PCSK9-KO H22 tumors (figure 3A–D, online supplemental figure S9C). Collectively, these findings suggest that PCSK9-KO in HCC cells activates CD8⁺ T cells in TME.

Figure 3. PCSK9 enhances HCC immune escape by upregulating the expression of SPP1 and PD-L1. (A) The t-SNE plots of CD8⁺ T-cell subpopulations in control and PCSK9-KO groups. (B) Dot plot of specific marker genes used to annotate different cell clusters in (A). (C–D) Proportions of CD8⁺ T-cell subpopulations. (E) CellChat analysis of downregulated signals in PCSK9-KO tumors. (F) Venn diagram shows the co-expression genes in the GO, KEGG, and the immune checkpoints on the cancer cell membrane. (G) The expression of PCSK9 and the double positive (Do-pos.) of PCSK9 with SPP1 or PCSK9 with PD-L1 analyzed by scRNA-seq. (H, J) The mRNA levels of CD274 (encoding PD-L1) and SPP1 after PCSK9 knockout or overexpression (n=3). (I, K) PCSK9, SPP1 and PD-L1 protein levels were measured by western blot. (L–N) SPP1 and PD-L1 promoter-driven luciferase activities in HepG2 and Huh-7 cells transfected with PCSK9-OV or PCSK9-KO plasmids (n=4). Statistics: two-tailed unpaired Student’s t-test (H, J M and N). DCs, dendritic cells; GO, gene ontology; HCC, hepatocellular carcinoma; KEGG, kyoto ebcyclopedia of genes and genomes; KO, knockout; mRNA, messenger RNA; NK, natural killer; PCSK9, proprotein convertase subtilisin/kexin type 9; PCSK9-OV, PCSK9-overexpression; PD-L1, programmed death-ligand 1; scRNA-seq, single-cell RNA sequencing; SPP1, secreted phosphoprotein 1; t-SNE, t-distributed stochastic neighbor embedding; Treg, regulatory T cell.

PCSK9 enhances the immune escape of HCC cells by upregulating SPP1 and PD-L1

Previous reports indicated that PCSK9 inhibits CD8⁺ T-cell antigen presentation and activation by promoting the degradation of MHC-I and TCR.^{15 16} However, RNA-seq analysis of HCC cells treated with PCSK9 inhibitor alirocumab, shRNA, or knockout revealed no changes in MHC-I expression (online supplemental figure S10A-C). Similarly, scRNA-seq analysis of patient with HCC samples with high and low PCSK9 expression and our PCSK9-KO HCC tumor samples showed no significant effect of PCSK9 on TCR expression (online supplemental figure S10D and E). We then performed CellChat analysis to uncover the receptor-ligand communication changes between HCC cells and immune cells. Notably, SPP1/integrin receptor signaling was identified as the most prominently decreased communication pair in PCSK9-KO tumors, compared with control tumors (figure 3E). SPP1 has been reported as an immunosuppressive signal to inhibit CD8⁺ T activation and promote macrophage M2 polarization.23,25 Moreover, RNA-seq analysis comparing PCSK9-KO H22 cells to vector controls and pathway enrichment analyses revealed that genes associated with cell surface and cell adhesion were predominantly altered in PCSK9-KO cancer cells (online supplemental figure S11). Notably, Venn analysis of key genes in the cell surface and cell adhesion pathways from gene ontology (GO) and KEGG with known immune checkpoints showed that only PD-L1 was enriched (figure 3F).

Moreover, scRNA-seq analysis of patients with HCC revealed that the expression of PCSK9 was co-localized with SPP1 and PD-L1 (figure 3G). We also found that knocking out PCSK9 significantly reduced messenger RNA (mRNA) and protein levels of SPP1 and PD-L1, while overexpression of PCSK9 increased their expression (figure 3H–K). Using luciferase assays driven by SPP1 and PD-L1 promoter sequences, we confirmed that PCSK9 overexpression enhances, whereas PCSK9 knockout reduces, the transcription of SPP1 and PD-L1, indicating that SPP1 and PD-L1 are downstream targets of PCSK9 (figure 3L–N). Collectively, these findings suggested that PCSK9 promotes HCC cell immune escape by upregulating the transcription of SPP1 and PD-L1.

PCSK9 promotes the transcription of SPP1 and PD-L1 by elevating their co-transcription factor FLI1

Next, we focused on the transcript mechanism involved in upregulating the expression of SPP1 and PD-L1. We predicted the co-transcription factors on the promoter of SPP1 and PD-L1 from the hTFtarget and JASPR databases (figure 4A). 26 key transcription factors were identified, with FLI1 and EGR2 significantly downregulated, and CTCF significantly upregulated in PCSK9-KO cancer cells compared with control cells (figure 4B and online supplemental figure S12A). Correlation analysis revealed that FLI1 exhibited the highest positive correlation with SPP1 and PD-L1, compared with EGR2 and CTCF (online supplemental S12B-G). Moreover, Kaplan-Meier analysis demonstrated that patients with HCC with higher PCSK9/FLI1 levels had poorer survival rates (online supplemental S12H). Chromatin immunoprecipitation followed by sequencing (ChIP) and assay for transposase-accessible chromatin with sequencing (ATAC-seq) analyses revealed that FLI1 had the binding motifs on the promoter region of both SPP1 and PD-L1 (online supplemental figure S12I-K). Besides, ATAC-seq from GSE190959 suggested that FLI1 knockdown reduced its distribution in the transcription start site of the genome, and decreased the chromatin accessibility at multiple loci of SPP1 and CD274 promoters (online supplemental figure S12L-P).

Figure 4. FLI1 directly regulates the transcription of SPP1 and PD-L1. (A) co-TFs prediction of SPP1 and CD274 from JASPAR and hTFtarget databases. (B) The expression changes (log2FC) of the co-TFs, data from control and PCSK9-KO H22 cells RNA-seq. (C) The mRNA expression of FLI1 in control or PCSK9-OV or PCSK9-KO HepG2 cells (n=5). (D) The protein level of PCSK9 and FLI1 in vector control or PCSK9-OV or PCSK9-KO HepG2 cells. (E) The mRNA expression of SPP1 and CD274 in indicated HepG2 cells (n=4). (F) The protein expression of SPP1 and PD-L1 in indicated HepG2 cells. (G) The SPP1 and PD-L1 promoter-driven luciferase activities in HepG2 cells transfected with FLI1-OV plasmids (n=4). (H) CUT&Tag PCR analysis displayed that FLI1 binds to the SPP1 and PD-L1 promoters in HepG2 cells (n=4). (I) PCR amplification of precipitated chromatin fragments from CUT&Tag using different primers. (J) Luciferase reporter assay of SPP1 and PD-L1 promoters with different site mutation in FLI1-OV HepG2 cells (n=5). (K) ChIP assays displayed the directly binding of FLI1 on the SPP1 and CD274 promoters (n=5). Statistics: two-tailed unpaired Student’s t-test (C, G, H and K), one-way ANOVA (E, J). ANOVA, analysis of variance; ChIP, chromatin immunoprecipitation followed by sequencing; co-TFs, co-transcription factors; CUT&Tag, cleavage under targets and tagmentation; FLI1, friend leukemia virus integration 1; KO, knockout; mRNA, messenger RNA; PCSK9, proprotein convertase subtilisin/kexin type 9; PCSK9-OV, PCSK9-overexpression; PD-L1, programmed death-ligand 1; RNA-seq, RNA sequencing; SPP1, secreted phosphoprotein 1.

Moreover, our qPCR and western blot further confirmed that PCSK9-knockout decreased, while PCSK9 overexpression increased the expression of FLI1 (figure 4C,D). Overexpression of FLI1 elevated, while PCSK9-knockout reversed the mRNA and protein expression of both SPP1 and PD-L1 (figure 4E,F). Overexpression of FLI1 enhanced both SPP1 and CD274 promoters-driven luciferase activities (figure 4G). Additionally, we predicted that there were four and three FLI1-binding sites on the promoters of SPP1 and CD274, respectively (online supplemental figure S13 and 14). qPCR of anti-FLI1-immunoprecipitated DNA fragments showed that site 1 of both SPP1 and CD274 promoters had the highest DNA fragments enrichment indicating the most binding affinity of FLI1 (figure 4H,I). Furthermore, luciferase expression plasmids driven by SPP1 and CD274 promoters with the above putative FLI1 binding site mutations were constructed, respectively (online supplemental figure S15), and results showed that mutation of binding site 1 had the most significantly inhibitory effects on SPP1 and CD274 promoter activities (figure 4J). ChIP assay further confirmed that FLI1 could directly bind to the promoters of SPP1 and CD274 (figure 4K). The above data indicated that PCSK9 promotes the SPP1 and PD-L1 transcription through elevating their co-transcription factor FLI1.

PCSK9 promotes FLI1-mediated SPP1 and PD-L1 transcription via NOTCH3 pathway

Next, to further confirm whether FLI1 contributes to PCSK9-induced immune escape, we overexpressed FLI1 in PCSK9-KO H22 cells and inoculated into immunocompetent mice. The tumor growth curve showed that FLI1 overexpression abolished the tumor growth inhibition of PCSK9-KO (figure 5A–D). In addition, FLI1 overexpression reduced the release of IFN-γ and GZMB from tumor-infiltrating CD8⁺ T cells (figure 5E). Furthermore, the IF assay also showed that FLI1 overexpression inhibited the PCSK9-KO-mediated CD8⁺ T-cell activation in TME (figure 5F). Similarly, we also confirmed that FLI1-OV can reverse the PCSK9-KO-induced downregulation of SPP1 and PD-L1 by IHC (figure 5G). Moreover, the downregulated DEGs revealed that genes associated with the NOTCH pathway were predominantly altered in PCSK9-KO cancer cells (figure 5H). Additionally, the mRNA level of NOTCH3 was significantly reduced by PCSK9-KO, while the other subtypes of NOTCH did not change (figure 5I). Western blot confirmed that PCSK9-knockout decreased the expression of NOTCH3 and FLI1 (figure 5J). In addition, Western blot further confirmed that PCSK9-KO decreased, while NOTCH3-OV increased the expression of FLI1 (figure 5K). Finally, we combined PCSK9-KO with NOTCH3-OV in co-culture analysis. Results showed that NOTCH3-OV significantly reversed the enhanced killing effect of PCSK9-KO (figure 5L). The above data indicated that PCSK9 promotes the immune escape of HCC cells by upregulating SPP1 and PD-L1 through activation of the NOTCH3/FLI1 pathway.

Figure 5. PCSK9-KO enhanced antitumor immunity is reversed by FLI1 overexpression. (A–D) Effect of overexpression of FLI1 (FLI1-OV) and PCSK9-KO on HCC cell tumor growth and mouse survival. Schematic of the experiment protocol (A), the tumor size (n=10), overall survival (n=10), and tumor weight (n=5). (E) Flow cytometric analysis of the percentage of tumor-infiltrating CD8⁺ T cell, Treg, and CD8⁺ T cells expressing IFN-γ, GZMB and the ratio of CTL to Treg cells (n=5). (F) Representative immunofluorescence images of CD8, IFN-γ, and GZMB in the tumors. Scale bar, 75 µm. (G) Representative staining of SPP1, PD-L1 in tumors. Scale bar, 75 µm. (H) KEGG analysis of the downregulation DEGs in PCSK9-KO compared with Nt-Ctrl. group. (I) The mRNA expression of NOTCH1, NOTCH2, NOTCH3, and NOTCH4 in H22 and HepG2 cells after PCSK9-KO (n=3). (J) The protein level of NOTCH3 and FLI1 in Nt-Ctrl. and PCSK9-KO cancer cells. (K) The protein expression in PCSK9-KO and/or NOTCH3 overexpressing (NOTCH3-OV) H22 cells. (L) PCSK9-KO and/or NOTCH3-OV H22-OVA cells were 1:4 ratio co-cultured with OT-1 CD8⁺ T cells. The number of living cells was measured by fluorescence (n=3). Scale bar, 50 µm. Statistics: two-way ANOVA (B), log-rank test (C), two-tailed unpaired Student’s t-test (I), one-way ANOVA (D, E, F and L). ANOVA, analysis of variance; CTL, cytotoxic CD8⁺ T lymphocytes; DEG, differential gene expression; DAPI, 4’,6-diamidino-2-phenylindole; ECM, extracellular matrix; FLI1, friend leukemia virus integration 1; GZMB, granzyme B; HCC, hepatocellular carcinoma; IFN, interferon; KEGG, kyoto ebcyclopedia of genes and genomes; KO, knockout; mRNA, messenger RNA; NOTCH, neurogenic locus notch homolog; PCSK9, proprotein convertase subtilisin/kexin type 9; PCSK9-OV, PCSK9-overexpression; PD-L1, programmed death-ligand 1; SPP1, secreted phosphoprotein 1; S.C., subcutaneous; Treg, regulatory T cell.

Hepatocyte-specific all-in-one CRISPR/Cas9 adenine base editing mediated PCSK9 deficiency activates CD8⁺ T-cell antitumor immunity

Our above studies and previous reports showed that PCSK9 was a newly identified immunosuppressive inhibitor, and blocking its expression in HCC cells could enhance antitumor immunity. CRISPR base editors are an attractive gene-editing modality due to their precise targeted alterations without causing double-strand breaks, in contrast to CRISPR-Cas9 and other gene-editing nucleases.²⁶ Additionally, CRISPR base editor targeting PCSK9 for PCSK9 knockdown has been reported to be successful in reducing the blood LDL in primates.^{27 28} In this study, to generate higher transfection efficiency and lower off-target editing, we constructed an all-in-one CRISPR/Cas9 ABE-TBG-PCSK9 plasmid by engineering liver-specific TBG promoter-driven CRISPR/Cas9 ABE and a gRNA targeting the splice donor at the boundary of PCSK9 exon 1 (figure 6A). Targeting splice donor of PCSK9 exon 1 results in retention and read-through of intron 1, which will cause transcription stop owing to the presence of an in-frame TAG stop codon near the beginning of its first intron.²⁷ We found that the transfection efficiency of ABE-TBG-sgRNA plasmid was higher than the previously reported two plasmids system (figure 6B). Moreover, Sanger sequencing showed much higher editing efficiency of our all-in-one plasmid (~75%) than previously reported two plasmids system (~58%) after 72 hours transfection (figure 6C–F). Western blot further confirmed that the expression of PCSK9 was markedly decreased after editing by ABE-TBG-PCSK9 (figure 6G). Moreover, we encapsulated ABE-TBG-PCSK9 into polyplus liposomes and intravenously injected for in vivo H22 tumor treatment. Consistently, we observed ABE-TBG-PCSK9 significantly inhibited H22 tumor growth, and prolonged mice survival (figure 6H–K). Its antitumor efficacy is superior to that of anti-PCSK9 monoclonal antibodies (online supplemental figure S16A-C). Flow cytometry and IF assays also revealed its strong efficacy in active tumor-infiltrating of CD8⁺ T cells (figure 6L–N). Similarly, we also confirmed that ABE-TBG-PCSK9 treatment reduced the expression of FLI1, SPP1, and PD-L1 in TME (figure 6O). Additionally, nine top-ranked predicted off-target sites were selected for Sanger sequencing in cells with ABE-TBG-PCSK9. No base edits were detected at these sites in either ABE-TBG-PCSK9-treated or control cells (online supplemental figure S16D-F). The above results suggest that the all-in-one ABE-TBG-PCSK9 is a promising therapy for PCSK9 deficiency and HCC treatment.

Figure 6. CRISPR/Cas9 ABE-mediated PCSK9 knockdown inhibits HCC immune escape. (A) Schematic of the liver-specific TBG promoter-driven all-in-one CRISPR ABE base editor (ABE-TBG-sgRNA). (B) The transfection efficiency of two plasmids ABE system and our all-in-one plasmid. The yellow dot indicates the co-transfected. (C) The diagram shows the editing site (GT-AG) at the boundary of PCSK9 exon 1. (D–F) The editing efficiency of ABE-TBG-PCSK9 and two plasmids ABE system in H22 cells. (G) The protein level of PCSK9 in H22 cells after being transfected by ABE-TBG-PCSK9 or PCSK9 shRNA for 72 hours. (H–O) Effect of PCSK9 knockdown by ABE-TBG-PCSK9 on H22 cell immune escape (n=7). The tumor size (I–J), overall survival (K), tumor weight (L), population of tumor-infiltrating immune cells (M), representative staining images of tumors (N), representative immunohistochemistry staining of PCSK9, FLI1, SPP1, PD-L1 in tumors (O). Scale bar, 75 µm. Statistics: two-tailed unpaired Student’s t-test (F, L, M), and N), two-way ANOVA (J), log-rank test (K). ABE, adenine base editing; ANOVA, analysis of variance; CMV, cytomegalovirus immediate early promoter; DAPI, 4’,6-diamidino-2-phenylindole; FLI1, friend leukemia virus integration 1; HCC, hepatocellular carcinoma; PCSK9, proprotein convertase subtilisin/kexin type 9; PD-L1, programmed death-ligand 1; SPP1, secreted phosphoprotein 1; sgRNA, single-guide RNA; S.C., subcutaneous; TBG, thyroxine-binding globulin promoter.

Parecoxib is a PCSK9 transcriptional inhibitor

Due to the smooth surface of the PCSK9 protein, developing its direct binding small-molecule inhibitors is challenging.^{7 29} We constructed a PCSK9 promoter-driven tdTomato plasmid to screen its small molecule transcription inhibitor from a Food and Drug Administration-approved compound library (online supplemental figure S17A and B). We identified 24 potential PCSK9 inhibitors that could decrease the fluorescence intensity of tdTomato and the expression of PCSK9, in which parecoxib showed the strongest inhibitory effects (online supplemental figure S17C). We further confirmed that treating cells with parecoxib resulted in significant decreases in the mRNA and protein expression of PCSK9 in a dose-dependent manner (online supplemental figure S17D and E). Consistent with PCSK9-KO results, tumor cell growth and apoptosis were not altered by parecoxib treatment (online supplemental figure S17F-I). Since parecoxib is a COX-2 inhibitor, we constructed COX-2-deficient cells to verify its effect. Both mRNA and protein analyses showed that COX-2 deficiency did not alter PCSK9 expression (online supplemental figure S17J and K). In addition, our co-culture and tumor growth experiment results indicated that parecoxib-enhanced cell killing was independent of COX-2 (online supplemental figure S17L-O). However, parecoxib effectively inhibited H22 tumor growth and prolonged mouse survival (figure 7A–D). Flow cytometry and IF staining also revealed that parecoxib treatment enhanced the number and activation of tumor-infiltrating CD8⁺ T cells (figure 7E,F). IHC staining of tumor tissues confirmed that parecoxib inhibits the expression of PCSK9, FLI1, SPP1, and PD-L1 in vivo (figure 7G). Additionally, H&E staining of tumor sections showed obvious tumor cell death in the parecoxib treatment group (online supplemental figure S17P). Moreover, as a clinically approved medication, parecoxib had a high safety profile and showed no toxic effects on the heart, liver, spleen, lungs, or kidneys of mice (online supplemental figure S17Q). Blood biochemistry tests in mice revealed that parecoxib reduced indicators of liver and kidney toxicity and lowered serum LDL levels (online supplemental figure S17R).

Figure 7. PCSK9 small-molecule inhibitor parecoxib effectively inhibits H22 transplanted tumor and spontaneous liver cancer growth in vivo. (A–D) Effect of parecoxib on H22 transplanted tumor model (n=7). Schematic of experiment protocol of parecoxib treatment (A), tumor size (B), overall survival (C), and tumor weight (D). Flow cytometric analysis of the percentage of tumor-infiltrating CD8⁺ T cell, Treg, and CD8⁺ T cells expressing IFN-γ, GZMB (n=5) (E). Representative immunofluorescence images of CD8, IFN-γ, and GZMB in the tumors (F). Representative staining of PCSK9, FLI1, SPP1, PD-L1 in tumors (G). (H) Schematic of experiment protocol of spontaneous liver cancer mice treated with parecoxib (n=4). (I) The representative images of mouse livers. (J) The representative H&E staining of mouse livers. (K) Flow cytometric analysis of the percentage of tumor-infiltrating CD8⁺ T cell, Treg, and CD8⁺ T cells expressing IFN-γ and GZMB (n=4). (L) Representative staining images of CD8, IFN-γ and GZMB in the tumors. (M) Representative immunohistochemistry staining of PCSK9, FLI1, SPP1, PD-L1 in tumor isolated from mouse livers. Scale bar, 75 µm. Statistics: two-way ANOVA (B), log-rank test (C), two-tailed unpaired Student’s t-test (D, E, F, K and L). ANOVA, analysis of variance; FLI1, friend leukemia virus integration 1; GZMB, granzyme B; IFN, interferon; PCSK9, proprotein convertase subtilisin/kexin type 9; PD-L1, programmed death-ligand 1; SPP1, secreted phosphoprotein 1; S.C., subcutaneous; Treg, regulatory T cell.

Next, we evaluated the inhibition of PCSK9 expression and the anti-HCC efficacy of parecoxib in the Alb-Cre⁺/MYC⁺ spontaneous liver cancer model. The spontaneous liver cancer mice were generated using a previously reported strategy (online supplemental figure S18). Treatment with parecoxib reduced the number and size of liver tumors (figure 7H–J). Additionally, the number and function of tumor-infiltrating CD8⁺ T cells were enhanced in the parecoxib treatment group (figure 7K–L). Consistently, IHC staining confirmed that parecoxib effectively inhibited the expression of PCSK9, FLI1, SPP1, and PD-L1 in tumor tissues (figure 7M). These findings suggest that parecoxib is a newly identified PCSK9 transcriptional inhibitor that effectively reverses FLI1/SPP1/PD-L1-mediated CD8⁺ T-cell immunosuppression, thereby inhibiting HCC progression.

Discussion

PCSK9 is emerging as a critical molecule in both hyperlipidemia and tumor immunosuppression. Inhibiting PCSK9 effectively lowers blood LDL levels and simultaneously enhances antitumor immunity.^{7 10} The competition experiments of both previous reports and our study have demonstrated that CD8⁺ T cells exhibit stronger cytotoxicity against PCSK9-deficient tumor cells in vitro and in vivo.¹⁵ However, the existing theories that PCSK9 promotes MHC-I and TCR degradation cannot explain this selective killing effect within the same TME, suggesting the existence of alternative more important mechanisms underlying PCSK9-mediated immunosuppression.^{15 16} In this study, we used scRNA-seq to systematically analyze the PCSK9-driven immunosuppressive microenvironment in PCSK9-knockout HCC mouse models and human HCC samples. Our findings revealed that PCSK9 promotes HCC immune evasion through FLI1-mediated upregulation of SPP1 and PD-L1 via the NOTCH3 pathway. Additionally, we designed a hepatocyte-specific all-in-one ABE base editor (ABE-TBG-PCSK9) and screened a small molecule inhibitor (parecoxib) to target PCSK9 for anti-HCC immunotherapy.

SPP1, also known as osteopontin, is an extracellular matrix protein that signals through integrins to induce immunosuppression and matrix remodeling.³⁰ The SPP1-integrins axis accelerates cancer progression and immunosuppression.^{31 32} Cancer cells express SPP1 to interact with its receptors integrin beta subunits (ITGBs) and integrin alpha subunits (ITGAs), inhibiting CD8⁺ T-cell activation and orchestrating Th17 cell differentiation, thereby facilitating immune escape.33,35 Targeting SPP1 has been identified as a promising therapeutic strategy for HCC in combination with anti-PD-L1 therapy.³⁶ SPP1 is regulated by several transcription factors, including C/EBPβ, NF-κB, Runx2, and TGF-β.37,40 PD-L1, a well-known immune checkpoint, significantly inhibits T-cell activation and promotes immune evasion in various cancers.⁴¹ Its expression is regulated by transcription factors such as STAT1, NF-κB, c-Myc, AP-1, EGR1, and HIF-1α.⁴² In this study, we observed that knockout of PCSK9 inhibited, while overexpression increased, the transcription of both SPP1 and PD-L1. This suggests that the selective killing of CD8⁺ T cells in PCSK9-deficient HCC cells may result from their lower expression of SPP1 and PD-L1. FLI1, an erythroblast transformation-specific transcription factor with a conserved DNA-binding motif, plays roles in various physiological processes, including immunoregulation, epithelial-mesenchymal transition, cell differentiation, and development.⁴³ In the TME, FLI1 has been identified as an upstream transcription factor of signal transducer and activator of transcription 1 (STAT1) and matrix metalloproteinase 1 (MMP1), promoting cancer cell migration and immune escape.^{44 45} FLI1 regulates CREB-binding protein and STAT1, enhancing indoleamin 2,3-dioxygenase 1 (IDO1) transcription, which leads to increased kynurenine synthesis in tumor cells and contributes to CD8⁺ T-cell exhaustion.^{44 46} Furthermore, downregulation of FLI1 has been shown to improve the differentiation of effector T cells in cancers.⁴⁶ Notably, the NOTCH pathway is involved in the regulation of immunosuppressive signals.⁴⁷ Especially, NOTCH3 can specifically regulate PD-L1 via the mTOR pathway, thereby promoting immune evasion.⁴⁸ Our study reveals that FLI1 is a downstream target of PCSK9 and acts as a co-transcription factor for both SPP1 and PD-L1. Overexpression of FLI1 reversed the downregulation of SPP1 and PD-L1 induced by PCSK9 knockout, thereby abolishing the activation of CD8⁺ T cells and inhibiting tumor growth in PCSK9-KO cells. However, further investigation is required to ascertain the direct intracellular targets of PCAK9. Our study presents a novel mechanism of PCSK9-mediated immune escape in HCC cells and elucidates the underlying mechanism of selective CD8⁺ T-cell killing in PCSK9-deficient HCC cells.

Inhibiting PCSK9 expression not only suppresses tumorigenesis and development but also enhances antitumor immunity, demonstrating potent synergistic effects with immune checkpoint inhibitors.⁷ The PCSK9 inhibitor alirocumab is being tested in a clinical trial for non-small cell lung cancer in combination with the anti-PD-1 agent cemiplimab (NCT05553834). Another trial is recruiting patients with advanced colorectal cancer to receive PCSK9 inhibitors alongside standard first-line regimens (NCT06391905). Additionally, trials are underway for neoadjuvant chemoradiation combined with PD-1 and PCSK9 inhibitors in rectal cancer (NCT06304987). In particular, PCSK9 antibodies also warrant investigation as a potential therapeutic option for HCC. CRISPR-based editors offer a precise gene-editing method that avoids double-strand breaks and is considered safer than CRISPR-Cas9.²⁶ Targeting PCSK9 with an ABE base editor has effectively reduced blood LDL levels in primates.²⁷ However, the current dual plasmid base editor system exhibits low editing and silencing efficiency.⁴⁹ In this study, we designed a hepatocyte-specific all-in-one ABE base editor (ABE-TBG-PCSK9), which significantly improved PCSK9 editing efficiency, leading to effective knockdown of PCSK9 and enhanced CD8⁺ T cell-mediated anti-HCC immunity without off-target. Advanced, safe, and reliable nanomaterials may be further considered for the delivery of the all-in-one ABE base editor.^{50 51} Moreover, parecoxib, a COX-2-selective inhibitor approved for short-term perioperative pain control, has been shown to significantly activate CD8⁺ T cells in TME and enhance the antitumor response of anti-PD-1 antibodies in triple-negative breast cancer and lung cancer.^{52 53} Through high-throughput screening, we identified parecoxib as a transcriptional inhibitor of PCSK9, effectively activating tumor-infiltrating CD8⁺ T cells and suppressing immune escape in both transplantation and spontaneous liver cancer models.

In conclusion, our findings indicated that PCSK9 is highly expressed in HCC and negatively correlates with patient prognosis. It enhanced immune escape by promoting FLI1-mediated transcription of SPP1 and PD-L1 through the NOTCH3 pathway, thereby inhibiting tumor-infiltrating CD8⁺ T-cell activation. We developed the ABE-TBG-PCSK9 base editor and identified the inhibitor parecoxib, both of which downregulate PCSK9 and inhibit HCC immune escape. Thus, PCSK9 represented a valuable target for cancer immunotherapy, and its inhibition could serve as a promising complement to first-generation immunotherapies.

Data availability statement

Data are available upon reasonable request.