Disruption of FGL2 induces TFEB-dependent lysosomal degradation of PD-L1 and enhances the efficacy of anti-PD1 therapy in hepatocellular carcinoma.

Introduction
Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related mortality globally [1]. Although early detection is linked to better overall survival rates, most patients with HCC are diagnosed at unresectable advanced stages, leading to a 5-year survival rate of only 12% [2]. The immunosuppressive tumor microenvironment (TME) facilitates tumor progression, metastasis, and immunotherapy resistance in HCC. The intricate immunosuppressive network comprises various cellular components, such as regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs), along with inhibitory cytokines like transforming growth factor-β (TGF-β), interleukin (IL)-35, and IL-10 [3, 4]. Additionally, immune checkpoints, including programmed cell death-1 (PD1), programmed death ligand 1 (PD-L1) and cytotoxic T lymphocyte-associated protein 4 (CTLA4), play significant roles in this process. The expression of PD1 is mainly on activated T and B lymphocytes, while PD-L1 and programmed death ligand 2 (PD-L2), the ligands of PD1, are expressed by antigen-presenting cells and tumor cells [5–7]. Upon engagement with PD-L1/2, PD1 is thought to dampen T cell activity through various approaches, such as regulation of T cell receptor signaling, inhibition of T cell expansion, and suppression of cytokine production [8]. The function of these checkpoints is to keep immune responses within a physiological range, thereby protecting the host from autoimmune diseases. However, in cancer patients, malignant tumors often upregulate PD-L1 expression to evade immune attack [9, 10]. Gaining insights into the diverse regulatory mechanisms of PD-L1 may lead to the development of potential strategies for cancer treatment.
In the past few years, targeting PD1/PD-L1 axis with immune checkpoint blockade (ICB) has revolutionized cancer immunotherapy, but most patients with HCC do not obtain satisfactory clinical benefits [11, 12]. Adaptive immune resistance has been identified as a barrier to the success of anti-PD1/PD-L1 therapy [13]. For instance, upon reactivation by ICB, tumor antigen-specific T cells release interferon-γ (IFN-γ), which not only amplifies and sustains T cell functions but also induces a compensatory increase in PD-L1 expression on tumor cells, thereby restricting anti-tumor response [14]. Hence, blocking PD-L1 appears to be an effective combinatorial approach to potentiate anti-PD1 therapy. Supporting this point, Yi et al. [14] demonstrated that lenvatinib blocked fibroblast growth factor receptor 4 signaling, thereby promoting the ubiquitination and degradation of PD-L1 in HCC cells, which ultimately enhanced anti-PD1 immune responses. In addition, pharmacological inhibition of Gasdermin D was reported to reduce PD-L1 expression and synergize with anti-PD1 therapy [15]. These studies collectively underscore the mechanistic rationale and therapeutic potential of targeting PD-L1 to optimize immunotherapy outcomes.​
As a member of the fibrinogen-related protein superfamily, fibrinogen-like protein 2 (FGL2) is primarily expressed by macrophages, endothelial cells, T cells, and tumor cells [16]. It exhibits both prothrombinase activity and immunosuppressive functions. Previous studies have demonstrated that FGL2 facilitates tumor growth across multiple malignancies, including glioblastoma multiforme (GBM), lung cancer and colorectal cancer [17–19]. Researches on HCC from us and other teams reported that serum FGL2 levels were notably higher in HCC patients compared to individuals with chronic hepatitis B (CHB), suggesting a potential link between FGL2 and tumorigenesis [20, 21]. Our group further confirmed that blocking FGL2, either by using antibodies or through genetic deletion, decreased hepatoma load by boosting the activation of dendritic cells (DCs) and increasing the quantity and cytotoxicity of cytotoxic T lymphocytes (CTLs) within tumor tissues [22]. Additionally, another study indicated that FGL2 accelerated tumor progression by promoting MDSC accumulation [23]. In gliomas, FGL2 augments immunosuppression through elevating PD1 and CD39 expression on CD45+ immune cells [24]. Nevertheless, the involvement of FGL2 in the regulation of immune checkpoints in HCC remains unknown.
In the present study, we reported that FGL2 mediated tumor immune escape by inducing PD-L1 expression on hepatoma cells, along with PD1 and CTLA4 expression on CD4+ T cells and CD8+ T cells, which contributed to the progression of HCC. Mechanistically, FGL2 activated the mammalian target of rapamycin complex 1 (mTORC1) signaling by stabilizing the interaction between mTOR and Raptor. The activated mTORC1 then inhibited the nuclear translocation of transcription factor (TF) EB and subsequent lysosomal biogenesis, thereby​ diminishing PD-L1 degradation. Furthermore, intervention targeting FGL2 improved the efficacy of anti-PD1 therapy by enhancing the cytotoxicity of tumor-infiltrating CD8+ T cells in HCC.

Materials and methods

Immunohistochemistry
A tissue microarray (TMA) consisting of 85 valid liver cancer tissues and 73 nontumor liver tissues was purchased from Shanghai Outdo Biotech Company (Ethics number: SHYJS-CP-1601010). It should be noted that of these 85 patients with HCC, 1 had unavailable tumor-free survival time. The clinical information of patients was presented in supplementary Table 1. For multiplex immunohistochemistry (mIHC) staining, the tissue microarray was subjected to baking at 63℃ for 60 min and deparaffinizing through an automatic dyeing machine (LEICAST5020, LEICA). Subsequently, antigen retrieval was performed. Endogenous peroxidase activity was quenched using hydrogen peroxidase. After blocking, the TMA was incubated with specific antibodies as follows: FGL2 (1:500) (H00010875-M01, Abnova) and TFEB (1:100) (13372-1-AP, Proteintech). The slide was then washed with tris borate saline tween-20 (TBST) and incubated with the secondary antibody (SM802, ready-to-use, DAKO). Finally, the slide was counterstained with DAPI and enclosed with antifade mounting medium. The Tissue-FAXS system (TissueFAXS Spectra, TissueGnostics GmbH, Vienna Austria) was used to conduct the panoramic multispectral scan of the slide. Further data analysis was carried out using Strata-Quest analysis software (Version 7.1.129, TissueGnostics GmbH, Vienna Austria).
Mouse tumor tissues were fixed with 4% paraformaldehyde and embedded in paraffin. The paraffin-embedded tumor sections were stained for PD-L1(1:100) (64988, CST) at 4°C overnight. After washing with TBST, the sections were incubated with secondary antibody at room temperature for 45 minutes. The 3,3’-diaminobenzidine tetrachloride (DAB) was used to visualize the target protein. Finally, the slides were scored independently by two pathologists according to the staining intensity and the area of the protein expression on a scale from 0 to 3: 0 (negative), 1 (weak), 2 (medium), and 3 (strong).

Animals
Wild-type (WT) C57BL/6 mice between 4 and 6 weeks old were purchased from Beijing Vital River Laboratory Animal Technology Co. Ltd. (Beijing, China). Syngeneic Fgl2 gene knockout (Fgl2−/−) mice were constructed by Shanghai Model Organisms Center, Inc. (Shanghai, China). All mice were housed in microisolator cages in the animal experiment center of Tongji Hospital. Experimental protocols were performed according to the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and approved by the Tongji Hospital Animal Ethics Committee (Ethics number: TJH-202103002).

HCC models and animal treatment
To establish HCC models, Hepa1-6 cells were inoculated subcutaneously into the right inguinal fold region or orthotopically into the left liver lobe of WT and Fgl2-/- mice. The long and short diameters of the subcutaneous tumors were measured every 2–3 days. The tumor volume was calculated using the following formula: tumor volume = (long diameter × short diameter2)/2. For antibody-based drug interventions, tumor-challenged mice were randomly allocated into different treatment groups, ensuring that the initial tumor sizes were similar across groups. Depletion antibodies against CD3+ T cells (A2104, Selleck Chem) or corresponding IgG isotype (A2106, Selleck Chem) were administered intraperitoneally every 3 days at a dose of 200 µg per mouse. Anti-PD1 antibodies (100 ug, BE0146, Bio X Cell) or control IgG isotype (BE0089, Bio X Cell) were injected intraperitoneally every 4 days. At the endpoint of the experiment, all the mice were sacrificed. The tumors were weighed, and then analyzed by IHC and flow cytometry.

Cell culture and transfection
Human HCC cell lines (MHCC97H, Huh7), mouse HCC cell line (Hepa1-6) and human embryonic kidney (HEK) 293T cells were cultured in Dulbecco’s minimum essential medium (DMEM, Gibco) supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin/streptomycin solution. All cells were cultured at 37 °C in an atmosphere of 5% CO2.
HBLV-h-FGL2-GFP-PURO(OE-FGL2) and HBLV-GFP-PURO(Vector), Lenti-shFGL2(sh-FGL2) and negative control shRNA (sh-control) were constructed by Hanbio Biotech Co., Ltd. (Shanghai, China). To generate cells stably expressing FGL2 or FGL2 shRNA, HBLV-h-FGL2-GFP-PURO or Lenti-shFGL2 were transfected into MHCC97H and Huh7 cells at a multiplicity of 100. Three days post transfection, stable transfectants were selected in DMEM containing 10 µg/ml puromycin (P8230, Solarbio) for two days. Following 2–3 passages with puromycin, the cultured cells were utilized for subsequent experiments.
HA-tagged full-length mTOR and mTOR truncation mutants were subcloned into the pcDNA3.1 vector. All plasmids were verified by Sanger sequencing. TFEB, ATG5, and ATG7-specific siRNAs and control non-specific siRNA were purchased from General Biol Co. Ltd. (Anhui, China). The siRNA sequences were listed in the supplementary Table 2. Transfection of plasmids and siRNAs was performed using Lipofectamine 3000 (L3000008, Thermo Fisher) according to the manufacturer’s protocol.

T cell cytotoxicity assay
Human T cells were separated from peripheral blood mononuclear cells using the MojoSort Human CD3 T Cell Isolation Kit (480022, BioLegend) and activated with 100 ng/mL CD3 antibody (317325, BioLegend), 100 ng/mL CD28 antibody (302933, BioLegend), and 10 ng/mL IL-2 (589102, BioLegend). Then, MHCC97H cells transfected with FGL2 shRNA or control shRNA were cocultured with activated T cells in 12-well plates at a ratio of 1:10 for 48 h. The percentages of apoptotic tumor cells were examined by flow cytometry using the Annexin V-APC/7-AAD apoptosis kit (AP105, Multi Science) following the manufacturer’s protocols.

Flow cytometry analysis
Fresh mouse tissues were finely minced and digested into single-cell suspension using mouse tumor dissociation kits (130-096-730, Miltenyi Biotec). After red blood cell removal, single cells were suspended in FACS buffer and then incubated with fixable viability stain 700 (564997, BD) to exclude dead cells. Purified CD16/32 antibody (101302, BioLegend) was used to avoid nonspecific binding. Next, cells were stained with the following fluorescein-labeled antibodies in the dark: anti-CD45 (557659, BD), anti-CD3 (566494, BD), anti-CD4 (563232, BD), anti-CD8 (563068, BD), anti-Foxp3 (563101, BD), anti-CD25 (564023, BD), anti-PD1 (562671, BD), anti-CD152 (106314, Biolegend), anti-CD274 (564715, BD), anti-CD107a (564347, BD), anti-granzyme B (372216, Biolegend), anti-IL-4 (554436, BD) and anti-IFN-γ (557649, BD). For intracellular cytokine staining, cells were stimulated with the leukocyte activation cocktail containing GolgiPlug (550583, BD) for 6 h, followed by fixation and permeabilization using the transcription factor buffer set (562574, BD). Data were acquired using CytoFLEX flow cytometer (Beckman) and analyzed with FlowJo software.

Quantitative real-time Polymerase Chain Reaction (qPCR)
Total RNA was extracted from cells and mouse liver tumor tissues using TRIzol reagent (15596018, Invitrogen). Then, RNA was reverse-transcribed into complementary DNA using the ReverTra Ace qPCR RT Kit (FSQ-101, TOYOBO). The expression of target genes was detected using SYBR Green Real-Time Polymerase Chain Reaction Master Mix (QPK-201, TOYOBO) in a real-time PCR system (Bio-Rad). The primer sequences were listed in the supplementary Table 3.

Western blot analysis
Total protein was extracted from cells and mouse liver tumor tissues using RIPA lysis buffer (AR0102, Boster) supplemented with protease inhibitor cocktail (AR1182, Boster) and phosphatase inhibitor cocktail (K1015, Apexbio). The nuclear and cytoplasmic protein extraction kit (P0028, Beyotime) was used to separate cellular nuclear and cytoplasmic fractions. The protein was separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and then transferred to PVDF membranes. 5% skimmed milk was used to avoid nonspecific binding. Next, membranes were incubated with the following antibodies: FGL2 (1:500) (sc-100276, Santa Cruz Biotechnology), PD-L1 (1:4000) (66248-1-Ig, Proteintech), DDDDK-Tag (1:5000) (AE063, Abclonal), LAMP1 (1:10000) (67300-1-Ig, Proteintech), CTSD (1:10000) (21327-1-AP, Proteintech), TFEB (1:5000) (68632-1-Ig, Proteintech), phospho-TFEB (1:1000) (37681, CST), TFE3 (1:500) (A7936, Abclonal), MITF (1:1000) (13092-1-AP, Proteintech), GAPDH (1:50000) (60004-1-Ig, Proteintech), Histone H3 (1:5000) (17168-1-AP, Proteintech), p70S6K (1:1000) (2708, CST), phospho-p70S6K (1:1000) (9234, CST), 4EBP1 (1:1000) (9644, CST), phospho-4EBP1 (1:1000) (2855, CST), AKT (1:1000) (9272, CST), phospho-AKT (1:1000) (4060, CST), GSK3β(1:1000) (12456, CST), phospho-GSK3β (1:1000) (5558, CST), PKC (1:4000) (12919-1-AP, Proteintech), phospho-PKC (1:1000) (38938, CST), ERK1/2 (1:1000) (16443-1-AP, Proteintech), phospho-p44/42 MAPK (Erk1/2) (1:2000) (4370, CST), ATG5 (1:2000) (10181-2-AP, Proteintech), ATG7 (1:1000) (10088-2-AP, Proteintech), HRP goat anti-rabbit IgG (1:10000) (A21020, Abbkine) and HRP goat anti-mouse IgG (1:10000) (A21010, Abbkine). Densitometry was quantified using ImageLab software.

Co-immunoprecipitation
Total protein was extracted from MHCC97H cells and 293T cells using NP40 cell lysis buffer (AR0107, Boster). After precleaning with Protein A/G Magnetic Beads (HY-K0202, MedChemExpress), the protein was incubated with primary antibodies at 4 °C overnight. Next, prewashed magnetic beads were added and incubated with the cell lysates for 3 h at 4 °C. Following magnetic separation, the immune complex samples were collected, boiled with loading buffer for 5 min, and subjected to western blot analysis.

LysoTracker red staining
MHCC97H cells were incubated with 50 nmol/L LysoTracker Red DND-99 dye (L7528, Thermo Fisher) for 40 min at 37 °C to label lysosomes. After incubation, the culture medium was aspirated. Cells were then rapidly washed twice with pre-warmed phosphate-buffered saline (PBS) to remove unbound dye. Red fluorescence was detected using a fluorescence microscope (IX71, OLYMPUS). Fluorescence intensity was analyzed by ImageJ software.

Immunofluorescence analysis
MHCC97H cells were seeded in 6-well culture plates with a circular coverslip per well. After washing and blocking, the cells were incubated with the following primary antibodies: FGL2 (1:200) (H00010875-M01, Abnova), PD-L1 (1:200) (17952-1-AP, Proteintech), LAMP1 (1:200) (ab25630, Abcam), TFEB (1:200) (13372-1-AP, Proteintech) and mTOR (1:100) (66888-1-Ig, Proteintech). DAPI (G1012, Servicebio) was used for nuclear staining. Images were acquired with a fluorescence microscope (BX53, OLYMPUS, Japan).

Enzyme-Linked Immunosorbent Assay (ELISA)
The level of FGL2 in tumor tissue homogenate from mice was detected using the ELISA kit (SEA512Mu, Cloud-Clone, China) according to the manufacturer’s instructions.

Statistical analysis
All data were presented as mean ± standard error of the mean (SEM). For normally distributed data, Student’s t test was used for comparisons between two experimental groups. One-way ANOVA with Tukey’s correction was used for comparing three or more groups. Mann-Whitney U test was used for non-normally distributed data. The log-rank test was used to compare the survival rate. P < 0.05 was considered statistically significant. GraphPad Prism 9 software was used for statistical analysis.

Results

FGL2 knockout inhibited tumor growth and downregulated the expression of PD1, PD-L1 and CTLA4 in transplanted HCC models
To characterize the importance of FGL2 in HCC, we first assessed its prognostic value in a cohort consisting of 85 HCC patients. Kaplan-Meier analysis demonstrated that patients with higher FGL2 levels exhibited significantly poorer overall survival (OS) (Fig. 1B) and progression-free survival (PFS) (Fig. 1C) than those with lower FGL2 expression. In view of the pivotal role of immune checkpoints in tumor immunosuppression, we investigated the relevance of FGL2 and immune checkpoints in liver cancer patients using the Gene Expression Profiling Interactive Analysis (GEPIA) server. The results revealed a significant positive correlation between FGL2 and PD1 (Fig. 1D), PD-L1 (Fig. 1E), PD-L2 (Fig. 1F), CTLA4 (Fig. 1G), T cell immunoglobulin and immune receptor tyrosine-based inhibitory motif domain (TIGIT) (Fig. 1H), CD39 (Fig. 1I), T-cell immunoglobulin mucin 3 (TIM3) ( Fig. 1J), B- and T-lymphocyte attenuator (BTLA) (Fig. 1K) and lymphocyte activation gene-3 (LAG3) (Fig. 1L), as indicated by Pearson’s correlation coefficient.

To examine the association observed in human specimens, we constructed HCC models in WT and Fgl2−/− mice. In the orthotopic models, FGL2 levels were elevated in tumor tissues compared with the surrounding peri-tumor regions (Fig. S1A). Following FGL2 knockout, the concentration of FGL2 in tumor tissue homogenates was decreased (Fig. 2A). Meanwhile, FGL2 deficiency markedly reduced tumor burden (Fig. 2B). Immunohistochemical (IHC) staining showed that FGL2 knockout decreased tumorous expression of PD-L1 (Fig. 2C). In the TME, PD-L1 expression on tumor cells was considerably lower in Fgl2−/− mice compared to WT mice (Fig. 2D), while PD-L2 levels were comparable between the two groups (Fig. S1C). Concurrently, the expression of PD1 and CTLA4 on CD8+ T cells (Fig. S1D) and CD4+ T cells (Fig. S1E) was also markedly reduced following FGL2 knockout. In the subcutaneous HCC models, the experimental results closely mirrored those of the orthotopic tumor models. FGL2 knockout notably reduced tumor volume (Fig. 2F, G) and weight (Fig. 2H), along with the significant decrease of PD-L1 and PD-L2 on tumor cells (Fig. 2I and S1F), PD1 and CTLA4 expression on T cells (Fig. S1G, H). Overall, these results suggested that FGL2 augmented immunosuppression by prompting the expression of negative immune checkpoints, thereby facilitating HCC progression.
Since FGL2 promoted the expression of PD1, PD-L1 and CTLA4, which were potent inhibitors of T cell function, ​​we hypothesized that FGL2 exerted its pro-tumorigenic effects by suppressing T cell activity​​. To test this hypothesis, we depleted T cells using anti-CD3 antibodies (Fig. S2A) ​​and assessed the contribution of T cells to FGL2-driven HCC progression​​. The depletion efficacy was monitored through flow cytometry (FCM) (Fig. S2B-E). Notably, T cell depletion abolished the tumor suppression caused by FGL2 knockout (Fig. 2J-L), confirming that FGL2 drove HCC progression through CD3+ T cell-dependent mechanisms.

FGL2 upregulated PD-L1 expression in HCC cells and attenuated the cytotoxicity of T cells
We further conducted gain- and loss-of-function assays in Huh7 and MHCC97H cells to validate FGL2-mediated PD-L1 regulation in HCC. As depicted in Fig. 3A and B, FGL2 overexpression increased PD-L1 levels in both cell lines, while FGL2 knockdown diminished the abundance of PD-L1, as assessed by western blot (Fig. 3C) and immunofluorescence (Fig. 3D). Given that PD-L1 is a transmembrane protein with immunosuppressive activity in its extracellular region [25], the membrane PD-L1 levels were also determined. The findings indicated that membrane PD-L1 expression also decreased following FGL2 knockdown (Fig. 3E). IFN-γ is recognized as the most effective inducer of PD-L1 expression within the TME [26]. We then examined whether interference of FGL2 could abolish inductive PD-L1 expression. In accordance with prior research, both total PD-L1 and membrane PD-L1 were elevated following stimulation with IFN-γ. Nevertheless, this induction was greatly eliminated due to FGL2 deficiency (Fig. 3F, G). T-cell-mediated cytotoxicity against tumor cells can be hindered by the binding of PD1 on T cells to PD-L1 on tumor cells [27]. To examine whether FGL2 knockdown-mediated PD-L1 suppression affected T cell cytotoxicity, we co-cultured MHCC97H cells with activated CD3+ T cells. According to the apoptosis analysis, FGL2-silenced MHCC97H cells showed increased susceptibility to T-cell killing (Fig. 3H). But the apoptosis of MHCC97H cells was not influenced by FGL2 expression when they were cultured separately (Fig. S3). The combined data indicated that FGL2 influenced the ability of T cells to kill HCC cells by modulating PD-L1 expression.

FGL2 disruption induced lysosome-dependent degradation of PD-L1
We proceeded to clarify how FGL2 regulated PD-L1 expression in tumor cells. Real-time PCR analysis showed that FGL2 had little influence on PD-L1 transcription in either Huh7 (Fig. S4A) or MHCC97H cells (Fig. S4B, C), implying that FGL2-mediated PD-L1 regulation might occur at the protein level. To prove the concept, HCC cells were treated with the protein synthesis inhibitor cycloheximide (CHX). Compared to control cells, FGL2 overexpression delayed the degradation of PD-L1 (Fig. 4A), whereas FGL2 knockdown displayed the opposite effect (Fig. 4B). PD-L1 can be degraded via the ubiquitin-proteasomal system and lysosomal pathways [28]. Subsequently, MHCC97H cells were exposed to the proteasome inhibitor MG132 and the results showed that the downregulation of PD-L1 caused by FGL2 knockdown was unaffected (Fig. 4C). Besides, FGL2 had no impact on PD-L1 ubiquitination (Fig. S4D, E). The above results indicated that FGL2 delayed the degradation of PD-L1 independently of the proteasomal system.
We then focused on the autophagic-lysosomal pathways. The results demonstrated that the downregulation of PD-L1 induced by FGL2 knockdown was significantly restored by co-incubating with the lysosomal inhibitors NH4Cl or chloroquine (Fig. 4D, E), whereas the autophagy inhibitor 3-MA (Fig. S4F) or siRNAs targeting against ATG5 and ATG7 (Fig. S4G, H) showed no effect on PD-L1 expression. These data suggested that FGL2 mediated PD-L1 degradation through an autophagosome-independent lysosomal pathway. We further assessed the number of acidic lysosomes labeled by LysoTracker probe and found increased staining intensity in FGL2- knockdown cells (Fig. 4F). Concurrently, these cells exhibited higher levels of lysosome-associated membrane protein 1 (LAMP1), a widely employed marker of lysosomes, and cathepsin D (CTSD), one of the key lysosomal enzymes [29] (Fig. 4G), which reinforced our conclusion that FGL2 disruption triggered lysosome-dependent degradation of PD-L1.
The intracellular PD-L1 was reported to be transported between recycling endosomes and lysosomes to maintain its stability [30]. Next, we examined whether FGL2 could influence the subcellular localization of PD-L1. As illustrated in Fig. 4H, the co-localization of PD-L1 and lysosome marker LAMP1 was significantly elevated in cells with FGL2-low expression, suggesting that FGL2 knockdown enhanced the presence of PD-L1 in lysosomes. However, the localization of PD-L1 on recycling endosomes marked by RAB11 was only slightly affected (Fig. S4I). Collectively, these results revealed that FGL2 knockdown promoted PD-L1 degradation by inducing lysosomal biosynthesis and facilitating PD-L1 accumulation in lysosomes.

FGL2 disruption promoted the expression of nuclear TFEB
Accumulating evidence shows that some members of the microphthalmia family, including TFEB, TFE3 and microphthalmia-associated transcription factor (MITF), play key roles in promoting lysosomal biogenesis [31–33]. Therefore, we wondered whether FGL2 regulated lysosomal biogenesis in an MITF/TFE3/TFEB-dependent manner. Through subcellular fractionation experiments, we found that FGL2 disruption elevated nuclear TFEB expression, while the levels of MITF and TFE3 remained unchanged in both HCC cells (Fig. 5A, B and S5A) and mouse liver tumor tissues (Fig. 5D, E). And similar results were observed by immunofluorescence staining (Fig. 5C). Nuclear TFEB can directly bind to the coordinated lysosomal expression and regulation (CLEAR) element in the promoters of lysosomal-related genes, thereby coordinating the expression of these genes. In conjunction with the elevated nuclear TFEB level, the transcriptional expression of TFEB target genes, including Lamp1, Ctns and Ctsb, was obviously higher in the FGL2 low-expression group than that in the control group (Fig. 5F, G). TFEB siRNA was further employed to verify the role of TFEB in mediating FGL2 activity. As depicted in Fig. 5H and I, silencing TFEB markedly rescued the upregulated expression of LAMP1 and CTSD, along with the degradation of PD-L1, caused by FGL2 knockdown in MHCC97H cells. These data demonstrated that FGL2 disruption induced lysosomal PD-L1 degradation through facilitating the nuclear translocation of TFEB in HCC cells.
To evaluate the translational potential of TFEB in HCC, we analyzed its expression patterns in a human liver tissue microarray encompassing 85 tumor tissues and 73 matched non-malignant specimens. The results demonstrated increased TFEB levels in HCC tumors in comparison with adjacent non-tumor tissues (Fig. S5C). And this tumor-specific upregulation was recapitulated in orthotopic mouse HCC models (Fig. S5D), confirming the conserved dysregulation of TFEB across species. Notably, elevated nuclear TFEB levels were associated with prolonged PFS (Fig. 5K), although it did not correlate with OS (Fig. 5L). Future studies in larger clinical cohorts are warranted to validate the prognostic value of TFEB.

FGL2 interacted with mTOR kinase and affected the phosphorylation of the mTOR pathway
The phosphorylation status of TFEB dictates its subcellular distribution. Phosphorylated TFEB forms complexes with 14-3-3 protein, thereby retaining TFEB in the cytoplasm, while dephosphorylated TFEB is readily translocated into the nucleus to activate genes related to lysosomes and autophagy [34]. To investigate how FGL2 influenced TFEB nuclear localization, we screened the most recognized pathways related to TFEB phosphorylation, including mTOR, extracellular signal-regulated kinase 2 (ERK2), protein kinase C (PKC), and glycogen synthase kinase-3β (GSK-3β) [34, 35]. As shown in Fig. 6A, FGL2 knockdown significantly reduced the phosphorylation of p70 S6 kinase (p70S6K) and eIF4E-binding protein 1 (4EBP1), both of which were known as substrates of mTOR complex 1 (mTORC1). Furthermore, phosphorylation of mTOR at Ser2448, a site that correlates with mTORC1 activity, was also markedly decreased. However, FGL2 did not influence the phosphorylation status of AKT, an mTORC2 substrate, nor the phosphorylation status of ERK1/2, PKC, and GSK-3β (Fig. 6A and S6). We observed similar results in murine liver cancer tissues (Fig. 6B and S6). Consistent with suppressed mTORC1 activity upon FGL2 knockdown, the phosphorylation of TFEB at Ser211, a known mTORC1 target site, was significantly reduced (Fig. 6C). Moreover, reactivation of mTOR signaling using MHY1485 restored the phosphorylation of 4EBP1 and p70S6k, along with PD-L1 expression in FGL2-silenced MHCC97H cells (Fig. 6D, E). Taken together, these results suggested that FGL2 modulated PD-L1 expression via mTOR-TFEB axis.
To investigate the mechanisms by which FGL2 regulated the mTORC1 pathway, we first performed immunoprecipitation (IP) assays, which revealed a physical interaction between FGL2 and mTOR in MHCC97H cells (Fig. 6F). Supporting this observation, immunofluorescence staining illustrated that FGL2 and mTOR were co-localized in the cytoplasm (Fig. 6G). And this interaction was consistently observed in 293T cells co-transfected with FLAG-tagged FGL2 and HA-tagged mTOR (Fig. 6H). To gain structural insights into the FGL2-mTOR interaction, we employed the AlphaFold3 model, as previously described [36], to predict potential binding interfaces. The analysis identified multiple sites on mTOR (Glu1251, Lys1256, Ser2035, Asp2102, Pro2273, Asp2274, Asp2276) that might be involved in the interaction with FGL2 (Fig. S7). Based on the prediction and the domain architecture of mTOR, we generated three truncated mTOR constructs (1-900 aa, 900–1376 aa, and 1376–2549 aa) for binding domain mapping. Interestingly, although AlphaFold3 predictions suggested potential interaction sites within both the 900–1376 aa and 1376–2549 aa regions, subsequent co-IP assays indicated that FGL2 specifically interacted with the mTOR fragment encompassing residues 900–1376 aa (Fig. 6I). This discrepancy may reflect inherent limitations of AlphaFold3 model in capturing dynamic or cofactor-dependent binding conformations under physiological conditions.
Previous studies have established that mTORC1 activity is potentiated by enhanced mTOR-Raptor binding or disrupted mTOR-Deptor binding [37, 38]. Given the interaction between FGL2 and mTOR, we wondered whether FGL2 modulated mTORC1 activity by influencing the assembly of these core regulatory subunits. As shown in Fig. 6J, FGL2 knockdown significantly weakened the interaction between mTOR and Raptor, while the binding between mTOR and Deptor was unaffected. These findings suggested that FGL2 selectively stabilized mTOR–Raptor association, thereby mediating mTORC1 activation.

FGL2 Inhibition enhanced antitumor immune responses of anti-PD1 therapy in HCC
Previous studies showed that combinatorial strategies suppressing PD-L1 expression could improve the therapeutic outcomes of anti-PD1 antibodies [14, 15]. Based on the impact of FGL2 on the regulation of PD-L1, we proposed that FGL2 inhibition would act synergistically with PD1 blockade. To test this hypothesis, we employed CRISPR/Cas9 to specifically deplete FGL2 to simulate an as-yet-undeveloped FGL2 inhibitor for tumor immunotherapy studies. The results showed that both FGL2 knockout and anti-PD1 monotherapy significantly retarded tumor progression. Remarkably, FGL2 depletion combined with anti-PD1 antibodies exerted the most substantial reductions in tumor volume and weight in comparison with either treatment alone (Fig. 7A-C). Subsequent analysis of alterations in the TME revealed that FGL2 depletion consistently reduced tumoral PD-L1 levels with or without anti-PD1 treatment (Fig. 7D). Additionally, upregulated expression of IFN-γ, granzyme B, and CD107a on CD8+ T cells was identified in mice that received the combined treatment (Fig. 7E-G), suggesting that FGL2 inhibition amplified anti-PD1 efficacy by enhancing the cytotoxicity of tumor-infiltrated CD8+ T cells.
While conventional cancer immunotherapies emphasize CD8+ T cell-mediated cytotoxicity, emerging evidence highlights the essential functions of CD4+ T cells in modulating anti-tumor immune responses and even influencing ICB therapy outcomes [39]. Among CD4+ T cell subsets, FGL2 depletion and anti-PD1 monotherapy resulted in a remarkable decrease in the accumulation of Tregs, and this suppression was more pronounced in the combination treatment group (Fig. 7H). Furthermore, the dual therapy promoted Th1 cell infiltration, while Th2 cells remained largely unaffected (Fig. 7I, J). Collectively, these findings demonstrated that combining FGL2 inhibition with PD1 blockade synergistically controlled HCC growth by remodeling the tumor microenvironment toward an activated state.

Discussion
Blockade of the PD1/PD-L1 axis is currently considered as a leading approach for tumor immunotherapy. However, during the course of clinical treatment, some patients experience compensatory up-regulation of PD-L1, which ultimately leads to drug resistance [13]. Hence, it is vital to elucidate the regulatory mechanisms governing PD-L1 and seek combination therapies to block its expression, as this represents a promising approach to improve ICB efficacy [14].
Accumulating evidence indicates that the expression of PD-L1 is tightly controlled by the lysosome-degradation system. Huntingtin-interacting protein 1-related (HIP1R), a protein engaged in intracellular transport, has been reported to physically interact with PD-L1 and facilitate its delivery to lysosomes via an intrinsic sorting signal [30]. Small-molecule drugs, such as Tubeimoside-1 and SA-49, boost lysosome biogenesis to destabilize PD-L1 through lysosomal degradation [35, 40]. ​​Building upon this established link between lysosomal function and PD-L1 degradation, our study revealed that FGL2 disruption increased the amounts of lysosomes and lysosomal enzymes to accelerate its proteolysis. Functional experiments using a coculture system showed that FGL2-silenced tumor cells exhibited enhanced susceptibility to T-cell killing, implying that FGL2-mediated PD-L1 upregulation impaired the cytotoxicity of T cells. While FGL2 has previously been implicated in promoting HCC progression by inhibiting DC activity and increasing MDSC accumulation [22, 23], this study unveils a novel mechanism whereby tumor cell-intrinsic FGL2 exerts immunosuppressive effects through direct modulation of PD-L1 expression.
TFEB, a well-established master regulator of lysosome biogenesis [41], has been implicated in the development of tumors in the kidney, pancreas, lung and liver [42, 43]. In agreement with prior research, our results demonstrated a significant upregulation of TFEB in tumor tissues relative to normal liver tissues. The functional activation of TFEB as a transcription factor relies on its dephosphorylation and movement into the nucleus [34]. In our study, FGL2 knockdown significantly reduced cytoplasmic TFEB levels while increasing its nuclear accumulation in HCC cells. In contrast, we observed a general upregulation of TFEB levels in both cytoplasmic and nuclear fractions of liver tumor tissues from Fgl2-/- mice. This apparent discrepancy may stem from the high complexity of the TME, which comprises not only tumor cells but also densely infiltrating immune and stromal cells. Therefore, we propose that the regulatory mechanisms of FGL2 on TFEB may differ across cell types. Specifically, within tumor cells, FGL2 appears to primarily regulate the nucleocytoplasmic shuttling of TFEB, whereas in non-tumor cells, such as immune cells, it may promote TFEB expression. This hypothesis requires further experimental validation. Despite these contextual differences, a consistent finding from both in vitro and in vivo models is that FGL2 deficiency promotes the expression of nuclear TFEB. This indicates that FGL2 acts as a negative regulator, inhibiting the nuclear localization and subsequent transcriptional activity of TFEB. What’s more, treatment with TFEB siRNA abrogated​​ the degradation of PD-L1 induced by FGL2 knockdown, indicating that FGL2 regulated lysosomal PD-L1 degradation in a TFEB-dependent manner. Intriguingly, our clinical analysis revealed an inverse correlation between FGL2 and nuclear TFEB in terms of prognostic implications: high levels of FGL2 were associated with poor prognosis, whereas elevated nuclear accumulation of TFEB predicted a favorable outcome. A potential mechanism is that FGL2-mediated repression of TFEB activity diminishes lysosomal degradation of PD-L1, thereby promoting PD-L1 expression and augmenting immune evasion, which ultimately worsens the clinical outcomes.
Multiple upstream kinases, including mTOR, GSK3β, ERK2, AKT and PKC, have been implicated in phosphorylating TFEB, collectively orchestrating its transcriptional activity [34, 35]. Our analysis of these signaling pathways revealed that FGL2 consistently activated the mTORC1 signaling both in vitro and in vivo. Further immunoprecipitation experiments demonstrated that FGL2 interacted with mTOR, as evidenced by their colocalization in the cytoplasm of HCC cells. During liver tumorigenesis, the mTOR signaling cascade is involved in the regulation of various cellular processes in cancer cells, such as proliferation, invasion, and metabolism [44]. Given the interaction between FGL2 and mTOR, additional exploration is needed to determine the impact of FGL2 on these cellular features.
Although PD1/PD-L1 blockade is effective in disrupting the function of membrane PD-L1, the relocation of PD-L1 from subcellular organelles to the cell surface, along with its adaptive increase, could potentially undermine the effectiveness of antibody-based drugs [30, 45]. Under the circumstances, combination treatments that modify PD-L1 homeostasis by regulating its synthesis, transport, recycling and degradation are expected to improve the efficacy of checkpoint blockade. Here, we observed that FGL2 depletion plus anti-PD1 therapy reduced tumoral PD-L1 levels compared to anti-PD1 monotherapy, demonstrating that FGL2 depletion acted synergistically with PD1 blockade to limit the PD1/PD-L1 axis by depleting intracellular reservoirs of PD-L1. In addition, the combined treatment alleviated the infiltration of Tregs into TME. As a major immunosuppressive population, Tregs restrain effector T cell activity by secreting immunosuppressive substances and competing for nutrients and cytokines. Hence, it is plausible to speculate that the enhanced cytotoxic activity of CTLs induced by the combination therapy may partly be attributed to the inhibition of Treg-mediated immune suppression.
However, there are some limitations in our current study. We only constructed allograft models derived from Hepa1-6 in vivo experiments. It was necessary to test our observations using other models, such as spontaneous tumorigenic models and chemically induced hepatocarcinoma models. Apart from the expression of PD-L1 on hepatoma cells, PD1 and CTLA4 expression on T cells were also altered following FGL2 deficiency. However, the precise mechanisms are not elucidated. Considering that Fc gamma receptor IIB (FcγRIIB), the known receptor for FGL2, is expressed on CD8+ T cells [46, 47], we propose that FGL2 may engage with T cells via FcγRIIB on cell surface, thereby inducing the expression of these immune checkpoints. This hypothesis requires experimental validation in the future.

Conclusion
In this study, we reported that FGL2 elevated PD-L1 expression in hepatoma cells both in vitro and in vivo. Mechanistically, FGL2 activated the mTORC1 signaling and decreased TFEB nuclear accumulation, thereby inhibiting lysosome biogenesis and PD-L1 degradation. Preclinically, FGL2 inhibition enhanced the antitumor effects of anti-PD1 treatment by enhancing the cytotoxicity of tumor-infiltrating T cells in HCC (Fig. 8). These results uncovered a novel mechanism underlying PD-L1 regulation, and provided a new combinatorial therapeutic strategy for the treatment of HCC.

Supplementary Information