Abstract Background The efficacy of immune checkpoint inhibitors (ICIs) for hepatocellular carcinoma (HCC) is limited by heterogeneity in individual responses to therapy. The heterogeneous phenotypes and crucial roles of cancer-associated fibroblasts (CAFs) in immunotherapy resistance remain largely unclear. Methods A specific CAF subset was identified by integrating comprehensive single-cell RNA sequencing, spatial transcriptomics and transcriptome profiling of patients with HCC with different responses to antiprogrammed cell death protein 1 (anti-PD-1) therapy. Mouse orthotopic HCC models and a coculture system were constructed, and cytometry by time-of-flight analysis was performed to investigate the functions and mechanisms of specific CAFs in the immune context of HCC. Results We identified a distinct flavin-containing monooxygenase 2 (FMO2)^+ CAF subset associated with a favorable response to anti-PD-1 therapy and better clinical outcomes. FMO2^+ CAFs increase anti-PD-1 treatment efficacy by promoting tertiary lymphoid structure formation and increasing the infiltration of CD8^+ T cells and M1-like macrophages through the C-C motif chemokine ligand 19 (CCL19)-C-C motif chemokine receptor 7 axis. Mechanistically, FMO2 promotes nuclear factor kappa B/p65-mediated CCL19 expression by competitively binding to glycogen synthase 1 (GYS1) with praja ring finger ubiquitin ligase 1 (PJA1), thereby suppressing the PJA1-mediated proteasomal degradation of GYS1. CCL19 treatment potentiated the therapeutic efficacy of anti-PD-1 therapy in mouse orthotopic HCC models. A favorable immunotherapy response was observed in patients with HCC with high serum levels of CCL19. Conclusions We identified a novel FMO2^+ CAF subset that serves as a critical regulator of microenvironmental immune properties and a predictive biomarker of the immunotherapy response in patients with HCC. CCL19 in combination with anti-PD-1 therapy may constitute a novel therapeutic strategy for HCC. Keywords: hepatocellular carcinoma, immunotherapy, biomarker __________________________________________________________________ WHAT IS ALREADY KNOWN ON THIS TOPIC * Patients with hepatocellular carcinoma (HCC) experience limited benefit from immune checkpoint inhibitors (ICIs) due to the heterogeneity in individual treatment responses. * Cancer-associated fibroblasts (CAFs), important components of the tumor microenvironment, have profound impacts on the immune features, therapeutic response and disease prognosis of patients with HCC; however, crucial CAF subsets and well-validated markers that can be prognostic classifiers and reveal microenvironmental characteristics remain largely unknown. WHAT THIS STUDY ADDS * By employing multiomic analysis, we identified a distinct flavin-containing monooxygenase 2 (FMO2)^+ CAF subset associated with a favorable response to antiprogrammed cell death protein 1 (anti-PD-1) therapy and better clinical outcomes in patients with HCC. * FMO2^+ CAFs increase anti-PD-1 treatment efficacy by promoting tertiary lymphoid structure formation and increasing the infiltration of CD8^+ T cells and M1-like macrophages through the C-C motif chemokine ligand 19-C-C motif chemokine receptor 7 axis. * Our in vivo study revealed that treatment with the recombinant CCL19 protein potentiated the efficacy of anti-PD-1 therapy. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY * This study identifies a novel FMO2^+ CAF subset as a critical regulator of the tumor immune microenvironment properties in HCC and provides robust preclinical evidence that CCL19 is a predictive biomarker of the immunotherapy response and a promising intervention to improve the efficacy of anti-PD-1 therapy. Introduction Hepatocellular carcinoma (HCC) is one of the most common malignancies and leading causes of cancer-related death worldwide because of its general late diagnosis and limited therapeutic options.[51]^1 The advent of immune checkpoint inhibitors (ICIs) has led to significant advancements in cancer treatment, and atezolizumab plus bevacizumab has been used as the first-line therapy for advanced HCC.[52]^2 Despite these advances, fewer than 20% of patients with HCC experience clinical benefits from immunotherapy, suggesting high heterogeneity in therapeutic efficacy among individuals.[53]^3 This finding underscores the urgent need to improve the efficacy of ICIs and to develop reliable predictive biomarkers that can guide treatment strategies and patient stratification. The tumor microenvironment (TME) enables immune escape and immune-mediated tumor cell killing, thereby exerting a profound influence on the efficacy of immunotherapy.[54]^4 5 Cancer-associated fibroblasts (CAFs), a crucial component of the TME, play a dual role in the TME: on the one hand, they contribute to the formation of an immunosuppressive TME by coordinating extracellular matrix (ECM) remodeling or inhibiting the activity of immune effector cells and recruiting immunosuppressive cells by secreting growth factors and cytokines, thereby inducing immune escape, tumor metastasis and therapy resistance[55]^6; on the other hand, CAFs can enhance the activation of T cells through mechanisms associated with major histocompatibility complex class II expression, thereby improving the efficacy of immunotherapies.[56]7,[57]11 Thus, a deeper understanding of the molecular characteristics of CAFs and the precise identification of critical CAF subsets are essential for reversing the immunosuppressive TME, increasing the efficacy of immunotherapy and predicting the prognosis of patients with HCC. However, an unmet need is the identification of CAF markers that can reliably serve as prognostic classifiers or reveal the microenvironmental characteristics of HCC. Flavin-containing monooxygenase 2 (FMO2), a member of the flavin-containing monooxygenase family, has been reported to be associated with the clinical outcomes of several cancers, including breast cancer, lung adenocarcinoma and ovarian cancer.[58]12,[59]14 Giannitrapani et al[60]^15 reported that FMO2 may be associated with the sorafenib response in patients with HCC. However, the expression profile and major role of FMO2 within the intricate TME of HCC remain to be further elucidated. In this study, we employed single-cell RNA sequencing (scRNA-seq) to identify critical subsets of CAFs associated with the immunotherapy response and prognosis of patients with HCC. Our analysis revealed that FMO2 was predominantly expressed in CAFs within the HCC microenvironment. Notably, patients with HCC with high tumor infiltration levels of FMO2^+ CAFs had better immunotherapy responses and longer survival durations. Furthermore, HCC tumors enriched with FMO2^+ CAFs exhibited an immunologically ‘hot’ phenotype characterized by increased infiltration of CD8^+ T cells and M1-like macrophages through the C-C motif chemokine ligand 19-C-C motif chemokine receptor 7 (CCL19-CCR7) axis. Finally, the increase in CCL19 signaling could sensitize HCC tumors to antiprogrammed cell death protein 1 (anti-PD-1) therapy in a preclinical model. This study provides a novel illustration of the immunological relevance and clinical significance of FMO2^+ CAFs in HCC. Methods Patients and specimens Cohort 1 containing 160 tumor and paired peri-tumor tissues obtained from patients with HCC who underwent curative resection from May 2002 to December 2006 at Zhongshan Hospital, Fudan University and was used for immunofluorescence staining, immunohistochemistry (IHC) and survival analyses. The inclusion criteria and follow-up procedures were performed as previously described.[61]^16 Cohort 2 consisted of 54 tumor tissues obtained from patients who received anti-PD-1 therapy at Zhongshan Hospital, Fudan University for immunofluorescence staining. The characteristics of 54 patients with HCC are summarized in [62]online supplemental table S1. Cohort 3 containing pretreatment serum samples from 68 patients with HCC receiving anti-PD-1 therapy was used for ELISA analysis. Additionally, transcriptome data of tumor tissues from 20 patients with HCC with different responses to anti-PD-1 therapy were obtained from our previous studies.[63]^16 17 Tumor response was assessed according to the modified Response Evaluation Criteria in Solid Tumors,[64]^18 and responders and non-responders were defined as previously described.[65]^16 Animal studies For the orthotopic tumor model used to investigate the role of FMO2^+ CAFs in the efficacy of anti-PD-1 therapy in vivo, Hepa1-6 cells (5×10^6 cells in 50 μL phosphate-buffered saline (PBS)) co-injected with or without the same number of FMO2^−overexpressing or control CAFs were injected under the liver capsules of mice (C57BL/6 male mice aged 5 weeks). An ultrasound imaging platform and mouse MRI system were used to measure the tumors. For antibody intervention, mice were treated intraperitoneally with 200 μg anti-PD-1 antibody (Bio X Cell) or isotype control twice a week for 2 weeks. At the end point of experiments, mice were sacrificed and tumors were dissected for further analyses. For the orthotopic tumor model to explore the effect of CCL19 on the efficacy of anti-PD-1 therapy in vivo, Hepa1-6 cells (5×10^6 cells in 100 μL PBS) were subcutaneously injected into the right inguinal fold regions of C57BL/6 mice. After 2 weeks, subcutaneous tumors were resected and cut into 1 mm^3 pieces, which were subsequently implanted into the liver parenchyma of C57BL/6 mice that were anesthetized via anesthesia. Antibody intervention was performed as described above. For cytokine receptor blockade, mice were injected intraperitoneally with 5 mg/kg CCR7 antagonist NSC658586 (MedChemExpress) or vehicle control daily for 14 days. For CCL19 intervention, mice were injected intraperitoneally with doses of 0.5 μg recombinant murine CCL19 (PeproTech) or vehicle control two times per week. At the end point of experiments, mice were sacrificed and tumors were dissected for further analyses. All mice were maintained under specific pathogen-free conditions and routinely monitored. All anesthesia or euthanasia methods were conducted in accordance with the Animal Research: Reporting of In Vivo Experiments guidelines. Cell lines and cell culture The human HCC cell line Huh7, human hepatic stellate cell (HSC) line LX-2, murine HCC cell line Hepa1-6, monocytic-leukemia cell line THP-1 and HEK293T cells were purchased from the Stem Cell Bank, Chinese Academy of Science (Shanghai, China). The murine HSC line JS-1 was obtained from the Liver Cancer Institute, Zhongshan Hospital, Fudan University. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco) or RPMI 1640 medium (Gibco) supplemented with 10% fetal bovine serum (Gibco) and 100 U/L penicillin/streptomycin (Gibco). Bone marrow-derived macrophages (BMDMs) were cultured in DMEM containing 50 ng/mL recombinant mouse macrophage colony-stimulating factor protein (R&D Systems) for 6–7 days before further experiments. Cells were maintained at 37℃ in a humidified incubator with 5% CO[2]. Public database analyses scRNA-seq data of patients with liver cancer cohort ([66]GSE125449 and [67]GSE202642) was obtained from the Gene Expression Omnibus (GEO) website.[68]^10 19 Raw count matrices were merged and analyzed using the Seurat R package.[69]^20 Based on the quality control metrics suggested in the Scanpy tutorial,[70]^21 cells with <200 genes expressed were filtered out. Cells expressing >6000 genes and >10% mitochondrial genes were also removed to ensure only high-quality cells were used in the downstream analyses. Genes expressed in less than three cells were also filtered out of the analysis. Scaled and centered read counts were used as gene expression for further analysis. Subsequently, cell annotation was performed by examining highly expressed marker genes between clusters as well as literature-derived and database-derived cell markers ([71]online supplemental figure S1A). Marker genes for each cell cluster were determined by using the FindAllMarkers function of the Seurat pipeline. The differentially expressed genes between FMO2^+ and FMO2^− fibroblasts were determined by using the FindMarkers function of the Seurat pipeline. In addition, we used CellChat[72]^22 to examine the molecular interaction networks among various cell types. CellChat serves as a quantitative tool for inferring and analyzing intercellular communication networks derived from scRNA-seq data. The findings from CellChat were used to uncover the incoming communication patterns of target cells and the outgoing communication patterns of secreting cells. Ligand-receptor pairs identified by CellChat with a p value <0.05 were deemed significant interacting molecules across different subsets. For spatial transcriptomics (ST) analysis, we used the R package Seurat for handling and illustrating ST data. ST data of patients with liver cancer cohort ([73]GSE238264 and HRA000437) were obtained from the GEO website and Genome Sequence Archive. To ascertain the proportions of cell subsets within each captured spot, we conducted SPOTlight deconvolution analysis following the guidelines provided at [74]https://github.com/MarcElosua/SPOTlight. Subsequently, we identified the marker gene matrix for each cell subset from our scRNA-seq dataset using Seurat’s FindAllMarkers function with the criteria logfc.threshold=1 and min.pct=0.8. This marker gene matrix, along with the ST data, was inputted into the spotlight_deconvolution function using its default settings. The outcomes of the deconvolution were integrated into the histological images using the spatial_scatterpie function, where each spot was labeled according to its subset. Following this, we applied SpatialFeaturePlot to depict the infiltration scores for each subset. Pearson’s correlation analysis was used to calculate the infiltration relationship of different subsets with each spot. All statistical analyses were conducted using R software (V.4.3.2). Statistical analysis All analyses were performed with GraphPad Prism 9 and R software (V.4.3.2). Numerical variables were shown as mean±SD and categorical variables as n (%). Differences between groups were tested using Student’s t-test or one-way analysis of variance with a post hoc least significant difference test, and constituent ratios were compared using Pearson’s χ^2 test or Fisher’s exact test. The Kaplan-Meier method with the log-rank test was used for survival analysis. A two-tailed p value <0.05 was considered statistically significant. More details of materials and methods are provided in the [75]online supplemental materials. Results FMO2^+ CAFs are correlated with the response to anti-PD-1 therapy and a better prognosis for patients with HCC We employed a public scRNA-seq dataset ([76]GSE125449) for bioinformatics analyses to comprehensively decipher the heterogeneity of fibroblasts in the TME of HCC. After the t-distributed stochastic neighbor embedding dimensional reduction analysis and copy number variation analysis, we successfully obtained 18 cell subclusters and annotated them into seven cell types using canonical marker genes, including malignant cells, endothelial cells, CD8^+ T cells, B cells, fibroblasts, pericytes, epithelial cells and macrophages ([77]figure 1A and [78]online supplemental figure S1A). Clustering analysis revealed significantly different transcriptional profiles among these cell subclusters ([79]online supplemental figure S1B). We extracted 492 characteristic marker genes of fibroblast subclusters for further analysis. We confirmed that these genes were fibroblast specific, as evidenced by the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes analyses, which indicated that they were primarily responsible for matrix construction and focal adhesion ([80]online supplemental figure S1C,D). Figure 1. FMO2^+ CAF subset correlated with anti-PD-1 therapy response and prognosis in HCC. (A) t-SNE plot for different clusters of single cells from patients with HCC identified in the [81]GSE125449 cohort. (B) Heatmap of top 20 differentially expressed genes between anti-PD-1 therapy responders and non-responders. (C) Venn diagram of the overlapping potential fibroblast-specific genes associated with the anti-PD-1 therapy response. (D) t-SNE plots of FMO2 expression level in all clusters. (E) Distribution of FMO2^+/FMO2 CAFs in HCC samples based on ST cohort ([82]GSE238264). (F) Immunofluorescence staining of FMO2 and ACTA2 in R and NRs. Scale bar: 50 μm. (G) IHC staining and statistical analysis of FMO2 expression in HCC tissues and peritumor tissues (n=160). Scale bar: 100 μm. (H) OS curve based on FMO2 expression in our patient cohort (n=160). (I) Immunofluorescence staining of FMO2 and ACTA2 in our patient cohort (n=160). Representative images from FMO2^high hHCC and FMO2^low hHCC were shown. Scale bar: 100 μm. (J) OS curve for patients with HCC with high and low levels of FMO2^+ CAFs. **p<0.01, Student’s t-test. Anti-PD-1, antiprogrammed cell death protein 1; CAF, cancer-associated fibroblast; FMO2, flavin-containing monooxygenase 2; HCC, hepatocellular carcinoma; IHC, immunohistochemistry; NR, non-responder; OS, overall survival; R, responder; ST, spatial transcriptomics; t-SNE, t-distributed stochastic neighbor embedding. [83]Figure 1 [84]Open in a new tab We extracted 205 fibroblast-specific genes associated with the anti-PD-1 therapy response and performed a univariate Cox proportional hazards regression analysis of these genes in the TCGA-LIHC cohort to identify which fibroblast-specific genes have an essential effect on the response to anti-PD-1 therapy and the prognosis of patients with HCC ([85]figure 1B,C). Genes with significant prognostic value (p<0.05) were further analyzed using the least absolute shrinkage and selection operator model ([86]online supplemental figure S2A). This process identified seven potential fibroblast-specific genes, and we focused primarily on FMO2, which was highly expressed only in fibroblasts (cluster 12), in our further investigations ([87]figure 1D and [88]online supplemental figure S2B,C). This finding was further supported by data from the [89]GSE202642 cohort, which included seven patients with HCC ([90]online supplemental figure S2D,E). Compared with FMO2^− CAFs, FMO2^+ CAFs presented characteristics of inflammatory CAFs and highly infiltrated HCC in patients with a favorable response to immunotherapy ([91]figure 1E and [92]online supplemental figure S2F). We conducted immunofluorescence staining of tumor tissues from 54 patients with HCC receiving anti-PD-1 therapy (cohort 2) to further elucidate the relationships among FMO2^+ CAFs, the response to anti-PD-1 therapy and the clinical outcomes of patients with HCC ([93]online supplemental table S1). Notably, FMO2^+ CAFs were predominantly present in anti-PD-1 therapy responders ([94]figure 1F and [95]online supplemental figure S3). Analyses of the TCGA-LIHC cohort revealed that FMO2 expression was significantly increased in peritumor tissues and that high expression of FMO2 was associated with better tumor differentiation and longer overall survival (OS) in patients with HCC ([96]online supplemental figure S4). Consistently, in our cohort of 160 patients with HCC (cohort 1), FMO2 was expressed primarily in peritumor tissues, and high expression of FMO2 was linked to longer OS ([97]figure 1G,H). We further stratified patients with HCC based on the infiltration level of FMO2^+ CAFs, as determined by immunofluorescence staining ([98]figure 1I). As shown in [99]figure 1J, patients with high FMO2^+ CAF infiltration experienced longer OS than those with low infiltration levels. These results indicated that FMO2^+ CAFs are associated with a positive response to anti-PD-1 therapy, and their presence in tumor tissues indicates longer OS of patients with HCC. FMO2^+ CAFs enhance the efficacy of anti-PD-1 therapy and regulate the immune landscape of the HCC microenvironment HSCs are the principal source of CAFs in the liver.[100]^10 23 We induced HSCs to differentiate into CAFs via coculture with HCC cells in vitro to verify the effect of FMO2^+ CAFs on the efficacy of anti-PD-1 therapy ([101]online supplemental figure S5A). On coculture with HCC cells, HSCs exhibited characteristics of CAFs, as evidenced by altered transcript levels of CAF markers ([102]online supplemental figure S5B–E). We subsequently established orthotopic HCC in mouse models bearing hepatic tumors formed from Hepa1-6 cells mixed with FMO2-overexpressing or control CAFs and treated the mice with an anti-PD-1 monoclonal antibody (mAb) or IgG2a ([103]figure 2A and [104]online supplemental figure S5F). At the study end point, we observed that the vector-expressing CAF group exhibited significantly increased tumor growth and resistance to the anti-PD-1 mAb compared with the Hepa1-6 cell-only group. In contrast, tumor growth was significantly suppressed in the FMO2-overexpressing CAF group compared with the other groups, and administration of the anti-PD-1 mAb exhibited the greatest antitumor efficacy in the FMO2-overexpressing CAF group ([105]figure 2B,C). Flow cytometry analysis revealed a marked increase in the infiltration and activity (granzyme B^+) of CD8^+ T cells in the FMO2-overexpressing group ([106]figure 2D). Figure 2. FMO2^+ CAFs enhance the efficacy of anti-PD-1 therapy and regulate the immune landscape of the HCC microenvironment. (A) Representative images of the orthotopic tumors at the study end point (five mice per group). Scale bar: 1 cm. (B) Tumor growth curves of the orthotopic tumors in each group. (C) The tumor volume and tumor weight of each group at the study end point. (D) Flow cytometry analysis of CD8^+ and granzyme B^+ CD8^+ T cells in the indicated groups. (E) t-SNE plot of different cell clusters in the tumors from FMO2^+ CAFs low and high infiltration groups. (F) Proportion of the different immune cell types among CD45^+ cells in the two groups. (G) Heatmap showing 39 markers expression in each cell cluster. (H) The expression of CD86, HLA-DR, CD206 and CD163 among macrophages in the two groups. (I) Left: t-SNE plot of cell clusters colored by the expression level of granzyme B. Right: proportion of granzyme B^+ CD8^+ T cells among CD8^+ T cells in the two groups. *P<0.05, **p<0.01, ***p<0.001 and NS, not significant, one-way analysis of variance with a post hoc least significant difference test (or Student’s t-test in part F (H, I). Anti-PD-1, antiprogrammed cell death protein 1; CAF, cancer-associated fibroblast; FMO2, flavin-containing monooxygenase 2; HCC, hepatocellular carcinoma; t-SNE, t-distributed stochastic neighbor embedding. [107]Figure 2 [108]Open in a new tab We further elucidated how FMO2^+ CAFs reshaped the immune landscape in HCC by profiling immune cell infiltration and functional status between HCC tumors with high and low infiltration of FMO2^+ CAFs using cytometry by time-of-flight (CyTOF) ([109]online supplemental figure S6). A clustering analysis based on 39 markers identified 32 cell clusters ([110]figure 2E and [111]online supplemental figure S7A). We observed a significant increase in CD8^+ T-cell and macrophage infiltration in HCC tumors with high levels of intratumoral FMO2^+ CAFs ([112]figure 2F). Further analysis of the expression of these 39 markers revealed that macrophages exhibited immunosuppressive characteristics in tumors with low levels of intratumoral FMO2^+ CAFs ([113]figure 2G,H and [114]online supplemental figure S7B). Additionally, CD8^+ T cells displayed elevated levels of granzyme B and CCR7 expression in tumors with high levels of intratumoral FMO2^+ CAFs ([115]figure 2I and [116]online supplemental figure S7C). We also examined the associations of intratumoral FMO2 expression with the immune landscape in the TCGA-LIHC cohort. Clustering analysis of ssGSEA data indicated that tumors with high FMO2 expression presented greater immunological activity than those with low FMO2 expression ([117]online supplemental figure S7D). Consistent with these findings, the CIBERSORT analysis of the TCGA-LIHC cohort showed a positive correlation between neoplastic FMO2 levels and the infiltration of CD8^+ T cells and M1 macrophages ([118]online supplemental figure S7E). Taken together, these findings suggest that FMO2^+ CAFs increase the efficacy of anti-PD-1 therapy and may contribute to an immune-active TME by influencing the infiltration of CD8^+ T cells and macrophage polarization. FMO2^+ CAFs promote M1-like polarization and infiltration of macrophages and CD8^+ T-cell chemotaxis in HCC We established an in vitro coculture system comprising CAFs and macrophages to further determine the effect of FMO2^+ CAFs on macrophages ([119]figure 3A). Gain-of-function studies showed that FMO2-overexpressing CAFs facilitated the M1-like polarization of differentiated THP-1 macrophages, as evidenced by altered transcript levels of M1 markers (CD80, NOS2, CXCL9 and TNF-α) and M2 markers (CD206, ARG1, TGF-β1 and CD163), as well as changes in the membrane expression of NOS2, CD80 and CD206 ([120]figure 3B,C). Similarly, increased expression of NOS2 and CD80 and decreased expression of CD206 were observed in BMDMs and primary human monocytes cocultured with FMO2-overexpressing murine CAFs ([121]figure 3D and [122]online supplemental figure S7F). In vivo experiments confirmed that tumors with FMO2-overexpressing CAFs exhibited increased infiltration of M1-like macrophages ([123]online supplemental figure S7G,H). Moreover, a chemotaxis assay indicated that conditioned medium from the coculture system containing FMO2-overexpressing CAFs drove the chemotactic migration of macrophages and CD8^+ T cells ([124]figure 3E,F). Figure 3. FMO2^+ CAFs promote CD8^+ T cells chemotaxis and M1-like polarization and infiltration of macrophages in HCC. (A) Schematic of in vitro coculture system for CAFs and PMA-treated THP-1 cells. (B) qRT-PCR analysis of the transcription level of M1 and M2 markers in THP-1-differentiated macrophages cocultured with the indicated CAFs. (C) Flow cytometry analysis of the expression levels of NOS2, CD80 and CD206 in THP-1-differentiated macrophages cocultured with the indicated CAFs. (D) Flow cytometry analysis of the expression levels of NOS2, CD80 and CD206 in BMDMs cocultured with the indicated CAFs. (E) Chemotaxis assays showing migration abilities of THP-1-differentiated macrophages toward the supernatants from the indicated coculture systems. Scale bar: 100 μm. (F) Chemotaxis assays showing migration abilities of CD8^+ T cells toward the supernatants from the indicated coculture systems. (G) Representative IHC staining images showing the expressions of FMO2, CD80 and CD8 in our patient cohort. Scale bar: 100 μm. (H) OS curves for patients with HCC with FMO2/CD80 and FMO2/CD8 co-expression. *P<0.05, **p<0.01, ***p<0.001, Student’s t-test. Experiments were repeated three times, and results were presented as the mean±SD. Anti-PD-1, antiprogrammed cell death protein 1; BMDM, bone marrow-derived macrophage; CAF, cancer-associated fibroblast; FMO2, flavin-containing monooxygenase 2; HCC, hepatocellular carcinoma; IHC, immunohistochemistry; OS, overall survival. [125]Figure 3 [126]Open in a new tab Additionally, as illustrated by the IHC staining of cohort 1, tumors with high FMO2 expression presented increased CD8 and CD80 expression levels ([127]figure 3G and [128]online supplemental table S2). High FMO2 expression in peri-tumor tissues did not correlate with tumor CD8 and CD80 expression levels, which was further validated in the TCGA-LIHC cohort ([129]online supplemental figure S8). Notably, FMO2^+ CAF^high/CD8^high and FMO2^+ CAF^high/CD80^high patients experienced the longest OS ([130]figure 3H). Collectively, our findings indicate that FMO2^+ CAFs promote macrophage infiltration and M1-like polarization and the chemotactic ability of CD8^+ T cells, leading to a positive effect on the prognosis of patients. FMO2^+ CAFs promote CD8^+ T-cell chemotaxis and M1-like polarization and infiltration of macrophages via the CCL19-CCR7 axis Given that FMO2^+ CAFs play an essential role in regulating CD8^+ T cells and macrophages, we investigated the interactions among them. We employed CellChat to analyse intercellular communication ([131]online supplemental figure S9). Our analysis revealed significant ligand-receptor pairs between these cells, with CCL19-CCR7 emerging as the most likely exclusive pair responsible for the binding of FMO2^+ CAFs to CD8^+ T cells and macrophages ([132]figure 4A and [133]online supplemental figure S10A). This finding was further validated in the [134]GSE202642 cohort ([135]online supplemental figure S10C). By integrating the ST dataset[136]^24 with the scRNA-seq dataset, we found that regions with higher infiltration scores of FMO2^+ CAFs corresponded with increased CCR7^+ CD8^+ T-cell infiltration scores and CCR7^+ macrophage infiltration scores ([137]figure 4B). Additionally, a bubble plot indicated that CCL19 predominantly originated from FMO2^+ CAFs ([138]online supplemental figure S10B). Quantitative reverse-transcription PCR and ELISA also confirmed elevated CCL19 expression and secretion in FMO2-overexpressing CAFs compared with control CAFs ([139]online supplemental figure S11A,B). In addition, we observed that CCL19 was significantly upregulated in anti-PD-1 therapy responders ([140]online supplemental figure S11C). Multiplex immunofluorescence staining of human HCC tissues showed the substantial infiltration of macrophages and CD8^+ T cells around CCL19^+FMO2^+ CAFs ([141]figure 4C). Previous studies have described the pivotal role of the CCL19-CCR7 axis in mediating tertiary lymphoid structure (TLS) formation and CD8^+ T-cell chemotaxis, as well as macrophage infiltration and polarization within solid tumors.[142]24,[143]31 We used ST to identify the FMO2^+ fibroblast region, TLS (identified by the 50-gene signature[144]^24) and CCL19 expression region. As shown in [145]online supplementalfigure S11D, the expression of CCL19 decreased with increasing distance from FMO2^+ fibroblasts, which was accompanied by a reduction in the TLS score. Moreover, immunofluorescence analyses revealed TLS formation around FMO2^+ CAFs in HCC tumors with high levels of intratumoral FMO2^+ CAFs ([146]online supplemental figure S11E). Next, we investigated whether the CCL19-CCR7 axis mediates the effects of FMO2^+ CAFs on CD8^+ T cells and macrophages. We observed that the knockdown of CCL19 in CAFs or the addition of a CCR7 inhibitor (CCR7-Cmp2105) significantly abrogated FMO2 overexpression-mediated macrophage chemotaxis and M1-like polarization, as well as the chemotactic ability of CD8^+ T cells ([147]figure 4D–F and [148]online supplemental figure S12A–D). We further validated the effects of FMO2^+ CAFs on macrophages and CD8^+ T cells through the CCL19-CCR7 axis by performing an orthotopic tumor implantation assay. The results showed that the knockdown of CCL19 in FMO2-overexpressing CAFs or the addition of a CCR7 antagonist (NSC658586) significantly attenuated the tumor-suppressive effect of FMO2-overexpressing CAFs and reduced the FMO2-induced accumulation of M1-like macrophages and CD8^+ T cells within the tumor ([149]online supplemental figure S13). In addition, in vivo experiments confirmed the absence of TLS formation around FMO2^+ CAFs following CCL19 knockdown ([150]online supplemental figure S14). Overall, these observations indicated that FMO2^+ CAFs promoted TLS formation and CD8^+ T-cell chemotaxis, as well as the infiltration and M1-like polarization of macrophages via the CCL19-CCR7 axis. Figure 4. FMO2^+ CAFs promote CD8^+ T cells chemotaxis and M1-like polarization and infiltration of macrophages via the CCL19-CCR7 axis. (A) Dot plot showing ligand-receptor pairs between FMO2^+/FMO2^- CAFs and macrophages and CD8^+ T cells. (B) Distribution and Pearson’s correlation analysis of FMO2^+/FMO2^- CAFs, CCR7^+ macrophages and CCR7^+ CD8^+ T cells in HCC samples based on ST analyses. (C) Immunofluorescence staining of FMO2, ACTA2, CD8, CD68 and CCL19 in our patient cohort. Scale bar, left: 100 μm; right: 20 μm. (D) Flow cytometry analysis of the expression levels of NOS2, CD80 and CD206 in THP-1-differentiated macrophages cocultured with the indicated CAFs. (E) Chemotaxis assays showing migration abilities of THP-1-differentiated macrophages toward the supernatants from the indicated coculture systems. Scale bar: 100 μm. (F) Chemotaxis assays showing migration abilities of CD8^+ T cells toward the supernatants from the indicated coculture systems. *P<0.05, ***p<0.001 and NS, not significant, one-way analysis of variance with a post hoc least significant difference test (or Student’s t-test in parts C–D). Experiments were repeated three times, and results were presented as the mean±SD. CAF, cancer-associated fibroblast; CCL19-CCR7, C-C motif chemokine ligand 19-C-C motif chemokine receptor 7; FMO2, flavin-containing monooxygenase 2; HCC, hepatocellular carcinoma; ST, spatial transcriptomics. [151]Figure 4 [152]Open in a new tab FMO2 facilitates nuclear factor kappa B/p65-mediated CCL19 transcription by interacting with glycogen synthase 1 in CAFs We employed scRNA-seq ([153]GSE125449) to analyze the differences between FMO2^+ and FMO2^− CAFs and to further investigate the biological signaling pathway involved in the regulation of CCL19 by FMO2 in CAFs and identified 213 downregulated genes and 550 upregulated genes in FMO2^+ CAFs compared with FMO2^− CAFs ([154]onlinesupplemental figure S15A [155]table S3). The GO analysis revealed that these genes participate in the regulation of the immune response and leukocyte activation and function as cytokines and chemoattractants ([156]online supplemental figure S15B). Notably, the pathway enrichment analysis of the upregulated genes revealed that the nuclear factor kappa B (NF-κB)/p65 signaling pathway, which is known to regulate CCL19 transcription, was significantly activated in FMO2^+ CAFs[157]32,[158]34 ([159]figure 5A). Furthermore, treatment with a p-p65 inhibitor (PTD-p65-P1 peptide TFA) significantly abolished the FMO2 overexpression-induced increase in CCL19 transcription and secretion in CAFs, and the activation of the NF-κB/p65 signaling pathway reversed the reduction in CCL19 expression associated with FMO2 deficiency ([160]figure 5B and [161]online supplemental figure S15C–F). These results suggest that the NF-κB/p65 pathway is responsible for FMO2 overexpression-mediated CCL19 transcription in CAFs. Figure 5. FMO2 binds GYS1 to facilitate NF-κB/p65-mediated CCL19 transcription. (A) KEGG analysis of differentially expressed genes between FMO2^+/FMO2^− CAFs. (B) Top: quantitative reverse-transcription PCR analyses of CCL19 transcriptional levels in the indicated CAFs treated with or without p-p65 inhibitor (PTD-p65-P1 Peptide TFA, 150 μM for 12 hours). Bottom: western blot analysis of FMO2, and phosphorylated and non-phosphorylated p65 in the indicated CAFs. (C–D) IP and immunofluorescence analyses showing the interactions and colocalization of FMO2 with GYS1 in CAFs. Scale bar: 10 μm. (E) AlphaFold 3 predicted the structural conformation of FMO2 and GYS1 and the interaction structural domains. (F) Flag-FMO2 FL or fragments were cotransfected with Myc-GYS1 FL or fragments in HEK293 T cells, and the interaction structural domains were detected by western blot analysis. (G) Western blot analysis of FMO2, GYS1 and phosphorylated and non-phosphorylated p65 in the indicated CAFs. *P<0.05, ***p<0.001 and NS, not significant, one-way analysis of variance with a post hoc least significant difference test. Experiments were repeated three times, and results were presented as the mean±SD. CAF, cancer-associated fibroblast; CCL19, C-C motif chemokine ligand 19; FMO2, flavin-containing monooxygenase 2; GYS1, glycogen synthase 1; IP, immunoprecipitation; KEGG, Kyoto Encyclopedia of Genes and Genomes; NF-κB, nuclear factor kappa B. [162]Figure 5 [163]Open in a new tab We performed a mass spectrometry analysis of FMO2-overexpressing CAFs to further examine the molecular mechanism by which FMO2 activates the NF-κB/p65 pathway and found that glycogen synthase 1 (GYS1) is a potential FMO2-interacting protein that has been reported to activate the NF-κB/p65 pathway[164]^35 ([165]online supplemental figure S16A). Coimmunoprecipitation and immunofluorescence analyses further confirmed the interaction and colocalization of FMO2 with GYS1 in CAFs ([166]figure 5C,D). In addition, the GYS1 protein was expressed at high levels in FMO2-overexpressing CAFs, whereas its mRNA levels remained unchanged ([167]online supplemental figure S16B,C). Moreover, we used AlphaFold 3 to predict the structural domains involved in the interaction between FMO2 and GYS1 and found that the structural domain consisting of amino acids 131–175 of FMO2 was involved in binding the structural domain of GYS1 consisting of amino acids 344–652 ([168]figure 5E,F). We determined whether GYS1 participates in the FMO2-mediated activation of the NF-κB/p65 pathway by transfecting CAFs with a small interfering RNA targeting GYS1 to disrupt the expression of GYS1. We found that GYS1 silencing significantly abrogated the effects of FMO2 overexpression on NF-κB/p65 pathways, as well as on CCL19 transcription and secretion in CAFs ([169]figure 5G and [170]online supplemental figure S16D,E). The overexpression of GYS1 effectively restored the decrease in CCL19 levels caused by FMO2 deficiency ([171]online supplemental figure S16F–H). In addition, GYS1 silencing in CAFs was sufficient to abolish the effects of FMO2-overexpressing CAFs on macrophages and CD8^+ T cells ([172]online supplemental figure S16I–K). These results indicate that FMO2 induces NF-κB/p65-mediated CCL19 transcription in a GYS1-dependent manner. FMO2 interacts with GYS1 in a mutually exclusive manner with praja ring finger ubiquitin ligase 1 to attenuate GYS1 ubiquitination The finding that FMO2 increases the GYS1 protein level but does not affect the mRNA level suggested that FMO2 may regulate GYS1 expression at the post-transcriptional level ([173]online supplemental figure S16B,C). Therefore, we examined the effect of FMO2 on GYS1 protein stability. The half-life of the GYS1 protein was shortened in control CAFs, whereas FMO2 overexpression led to a prolonged half-life, suggesting that FMO2 inhibits GYS1 degradation ([174]figure 6A). The ubiquitin-proteasome system and the autophagy-lysosome pathway are the two main mechanisms responsible for intracellular protein degradation.[175]^36 We treated CAFs with the proteasome inhibitor MG132 and the lysosome inhibitor chloroquine to clarify the pathway involved in FMO2-mediated GYS1 regulation. The results showed that the decrease in the GYS1 protein level caused by FMO2 knockdown was effectively reversed by treatment with the proteasome inhibitor MG132 but not with the lysosome inhibitor chloroquine, indicating that FMO2 inhibits the proteasomal degradation of GYS1 ([176]figure 6B). Moreover, the overexpression of FMO2 significantly reduced the ubiquitination levels of endogenous and exogenous GYS1 in cells ([177]figure 6C). These findings revealed that FMO2 participates in the ubiquitin-mediated degradation of GYS1. Figure 6. FMO2 attenuates the proteasomal degradation of GYS1 by competing with PJA1. (A) Western blot analysis of the effect of FMO2 on the half-life of GYS1 in CAFs treated with CHX (20 mM) for the indicated time periods. (B) Western blot analysis of the effect of FMO2 on the protein level of GYS1 in CAFs treated with MG132 (20 μM for 6 hours) or chloroquine (20 μM for 6 hours). (C) Left: western blot analysis of the effect of FMO2 on the ubiquitination level of GYS1 in CAFs. Right: IP analyses of cell lysates from HEK293T cells transfected with Flag-FMO2, HA-Ubiquitin and Myc-GYS1. (D) IP analyses showing the interactions of GYS1 with PJA1 in CAFs. (E–F) Western blot analyses of the protein and ubiquitination levels of GYS1 in PJA1 knockdown cells transfected with WT or LD Flag-shRNA-resistant (r) PJA1. (G) A pulse-chase analysis showing the half-life of GYS1 in CAFs with FMO2 and/or PJA1 knockdown. (H) IP analyses of the binding levels of FMO2 and PJA1 to GYS1 in HEK293T cells transfected with different doses of HA-PJA1 in the presence of MG132. (I) Top: schematic depicting the flow of sequential IP analyses. Bottom: IP analyses showing interactions of GYS1 with FMO2/PJA1 but no binding of FMO2 to PJA1. (J) IP analyses of cell lysates from HEK293T cells transfected with Myc-GYS1, Flag-FMO2 and HA-PJA1. CAF, cancer-associated fibroblast; CHX, Cycloheximide; GYS1, glycogen synthase 1; IP, immunoprecipitation; LD, ligase-dead; PJA1, praja ring finger ubiquitin ligase 1; WT, wild type. [178]Figure 6 [179]Open in a new tab We performed a mass spectrometry analysis of the immunoprecipitates of GYS1 to identify the E3 ubiquitin ligase responsible for the ubiquitination of GYS1 and found that GYS1 binds to praja ring finger ubiquitin ligase 1 (PJA1) ([180]figure 6D and [181]online supplemental figure S17A). Moreover, PJA1 knockdown increased GYS1 protein levels and reduced the level of ubiquitinated GYS1 in CAFs ([182]online supplemental figure S17B,C). The expression of recombinant shRNA-resistant wild-type but not ligase-dead PJA1 eliminated the elevated protein levels and inhibition of GYS1 ubiquitination induced by FMO2 knockdown, suggesting that the E3 ligase activity of PJA1 is required for GYS1 ubiquitination ([183]figure 6E,F). Notably, PJA1 knockdown markedly abolished the shortened half-life of GYS1 caused by FMO2 deficiency, indicating that the impact of FMO on GYS1 stabilization is associated with PJA1 ([184]figure 6G). Moreover, increasing the amount of PJA1 transfected gradually increased the interaction between PJA1 and GYS1 and decreased the binding affinity of FMO2 for GYS1, suggesting that PJA1 diminishes the binding between FMO2 and GYS1 in a dose-dependent manner ([185]figure 6H). We performed sequential immunoprecipitation assays to further investigate the interaction patterns of FMO2, GYS1 and PJA1 and found that the binding of GYS1 to FMO2 prevented its interaction with PJA1, which was further verified in coimmunoprecipitation assays ([186]figure 6I,J). Taken together, these results indicate that FMO2 competitively binds to GYS1 with PJA1 to attenuate GYS1 ubiquitination. CCL19 improves the efficacy of HCC immunotherapy in combination with anti-PD-1 therapy We performed an IHC analysis to examine the expression of FMO2, GYS1, p65, p-p65 and CCL19 in patients with HCC and further verify the role of the FMO2/GYS1/p65/CCL19 axis in HCC (Cohort 1) ([187]figure 7A). The results revealed positive correlations between FMO2, GYS1, p-p65 and CCL19 levels, which was consistent with findings in animal models ([188]figure 7B and [189]online supplemental figure S18). Given the limited baseline infiltration of FMO2^+ CAFs in most HCC tumors and their critical role in sensitizing HCC to anti-PD-1 therapy via CCL19 signaling, we next investigated whether exogenous CCL19 treatment and anti-PD-1 therapy could exert synergistic effects in HCC models. We constructed mouse orthotopic HCC models and treated them with CCL19 and/or an anti-PD-1 antibody ([190]figure 7C). At the study end point, more significant tumor growth limitations were observed in the mice treated with combination therapy than in the other groups ([191]figure 7D,E). Strikingly, the combination therapy induced a significant immune response with increased infiltration of granzyme B^+ CD8^+ T cells and CD86^+ macrophages into tumors ([192]figure 7F). No significant hepatic or renal toxicities were observed at the study end points for either monotherapy or the combination therapy ([193]figure 7G). Figure 7. CCL19 improves the efficacy of HCC immunotherapy in combination with anti-PD-1 therapy. (A) Representative IHC staining images showing the expressions of FMO2, GYS1, CCL19, phosphorylated and non-phosphorylated p65 in our patient cohort. Scale bar: 100 μm. (B) Correlation between immunohistochemical scores of indicated proteins in our patient cohort. (C) Representative images of the orthotopic tumors at the study end point (five mice per group). Scale bar: 1 cm. (D) Tumor growth curves of the orthotopic tumors in each group. (E) The tumor volume and tumor weight of each group at the study end point. (F) Flow cytometry analysis of CD11b^+ F4/80^+ 86^+ macrophages and granzyme B^+ CD8^+ T cells in each group. (G) Liver and kidney functions of mice with HCC tumors in each group at the study end point. (H) Left: representative abdominal MRI images of patients with HCC before and after anti-PD-1 antibody therapy. Right top: ELISA analysis of serum CCL19 levels in patients with HCC (cohort 3; n=68) with (n=34) and without (n=34) response to anti-PD-1 antibody therapy. Right bottom: using the median serum CCL19 level (1504 pg/mL) as the cut-off value, the proportion of responders and non-responders in different serum CCL19 levels. *P<0.05, **p<0.01, ***p<0.001, one-way analysis of variance with a post hoc least significant difference test (or Student’s t-test in part G). Experiments were repeated three times, and results were presented as the mean±SD. Anti-PD-1, antiprogrammed cell death protein 1; CCL19, C-C motif chemokine ligand 19; FMO2, flavin-containing monooxygenase 2; GYS1, glycogen synthase 1; HCC, hepatocellular carcinoma; NS, not significant. [194]Figure 7 [195]Open in a new tab Given that the serum CCL19 level has been identified as a predictive biomarker of the response to anti-PD-1 therapy in patients with triple-negative breast cancer,[196]^37 we next investigated the association between the serum CCL19 level and the immunotherapy response in patients with HCC. ELISA of cohort 3 revealed that patients with HCC who responded to immunotherapy had higher serum CCL19 levels than non-responsive patients ([197]figure 7H). Hence, these results suggest that the serum CCL19 level is a non-invasive predictive biomarker of the immunotherapy response in patients with HCC. In summary, our results revealed an essential role of CAF-intrinsic FMO2 in remodeling the tumor immune microenvironment through the GYS1/p65/CCL19 axis, thus providing a theoretical foundation that CCL19 treatment may be an effective approach to increase the efficacy of immunotherapy for HCC ([198]figure 8). Figure 8. Schematic diagram showing the regulatory role of FMO2^+ CAFs in the immune landscapes of HCC. CAFs intrinsic-FMO2 can compete with the E3 ligase PJA1 to bind GYS1. Thus, FMO2 attenuates PJA1-mediated ubiquitination and degradation of GYS1 in CAFs, which promotes the phosphorylation of p65 and the transcription of CCL19. Enhanced CCL19 signaling promotes tumor infiltration and M1-like polarization of macrophages, as well as TLS formation and the chemotactic capacity of CD8^+ T cells in HCC. CAF, cancer-associated fibroblast; CCL19, C-C motif chemokine ligand 19; FMO2, flavin-containing monooxygenase 2; GYS1, glycogen synthase 1; HCC, hepatocellular carcinoma; PJA1, praja ring finger ubiquitin ligase 1. [199]Figure 8 [200]Open in a new tab Discussion HCC is a malignant tumor with a poor immunotherapy response and high heterogeneity. A hallmark feature of the TME of HCC is the abundance of CAFs, which secrete a range of cytokines, chemokines and growth factors that directly or indirectly impact HCC progression.[201]^8 38 Several studies have used scRNA-seq to analyze the functional and phenotypic characteristics of CAFs in HCC.[202]^10 11 39 However, due to the lack of a detailed analysis of cohorts receiving immunotherapy, the relationship between CAFs and the immunotherapy response in HCC remains to be elucidated. Here, we identified a previously unknown CAF subset characterized by high expression of FMO2, which was associated with good survival outcomes and the immunotherapy response in patients with HCC. Previous studies have described the association of FMO2 with the clinical prognosis of patients with various cancers, but its exact function in the TME remains elusive.[203]12,[204]14 Our study revealed that FMO2 was predominantly expressed in CAFs and that the infiltration level of FMO2^+ CAFs was positively correlated with the survival duration of patients with HCC. In addition, patients with HCC who responded to anti-PD-1 therapy exhibited high infiltration of FMO2^+ CAFs. In a mouse orthotopic HCC model, we further confirmed that FMO2^+ CAFs increased the efficacy of anti-PD-1 therapy by increasing the infiltration and activation of CD8^+ T cells. The multidimensional interactions between CAFs and infiltrating immune cells in the TME have been shown to modulate the immune microenvironment and influence antitumor immune responses.[205]^40 41 By performing a high-throughput CyTOF analysis, we observed a significant increase in activated CD8^+ T-cell and M1-like macrophage infiltration in HCC tumors with high infiltration of FMO2^+ CAFs. Furthermore, we determined the main mechanisms by which FMO2 affects the infiltration and M1-like polarization of macrophages, as well as the chemotactic capacity of CD8^+ T cells by regulating the CCL19 signaling pathway in CAFs. Moreover, we found that FMO2^+ fibroblasts may facilitate TLS formation by establishing a CCL19 gradient. Although CCL19 has been classically attributed to myeloid cells, emerging evidence has revealed the tissue-specific cellular origins of CCL19.[206]42,[207]44 For example, CAFs in colorectal cancer liver metastases can secrete CCL19 to drive TLS formation.[208]^45 Therefore, we hypothesized that FMO2 in CAFs may play an important role in promoting TLS formation, CD8^+ T-cell chemotaxis and M1-like macrophage infiltration mainly through CCL19 signaling during immunotherapy. Moreover, we determined that the NF-κB/p65 signaling pathway, a well-known contributor to CCL19 transcription,[209]^46 47 was essential for the FMO2-mediated expression of CCL19 in CAFs. GYS1 can regulate the phosphorylation of p65 to activate the NF-κB/p65 signaling pathway.[210]^35 Our results indicated that FMO2 can bind GYS1 directly and positively regulate only the GYS1 protein level. However, the regulatory mechanisms of GYS1 post-translational modifications remain unclear. Here, we revealed a previously undescribed mechanism of GYS1 ubiquitination mediated by PJA1 and found that FMO2 can competitively bind to GYS1 with PJA1 to suppress the ubiquitylation-mediated proteasomal degradation of GYS1. In recent years, ICIs in combination with drugs that evoke the effector functions of immune cells have become important combination immunotherapy strategies.[211]^48 A recent study reported that a local injection of CCL19-expressing mesenchymal stem cells could enhance CD8^+ T-cell cytotoxicity and immunotherapy efficacy in solid tumors.[212]^49 Our results indicated that CCL19 treatment is an effective intervention to improve the efficacy of anti-PD-1 therapy against HCC, as evidenced by the optimal antitumor effect of the combined therapy in preclinical models. Similarly, the combined CCL19 treatment and anti-PD-1 antibodies have been shown to limit tumor growth and augment effective CD8^+ T-cell immunity in solid tumors.[213]^37 These studies and our results pave the way for future research that may further characterize CCL19 treatment and apply it to immunotherapy for cancer. Intriguingly, we also observed a potential association between the serum CCL19 level and immunotherapy response in patients with HCC, which suggests that the serum CCL19 level is a non-invasive predictive biomarker for the immunotherapy response in patients with HCC. One limitation of our study is that we did not elucidate the impact of FMO2^+ CAFs on tumor cells. Given the complexity and diversity of the TME, the underlying mechanisms may be multifaceted, necessitating further basic research through both in vivo and in vitro experiments. Additionally, the relationship between FMO2^+ CAFs and the immunotherapy response was assessed in a small-sample cohort of patients with HCC receiving immunotherapy; thus, we aim to conduct a large-scale, multicenter patient cohort study to obtain a more comprehensive understanding of this association and the implications for treatment in the future. If the results are validated in a large-scale, multicenter prospective patient cohort, CCL19 may play a crucial role in guiding stratification and treatment selection for patients with HCC, warranting further exploration across a broader range of cancers. In conclusion, the present study defines a novel FMO2^+ CAF subset that modulates the immune properties of the TME through the GYS1/p65/CCL19 axis, thereby establishing CCL19 as a predictive biomarker of the immunotherapy response and a promising intervention to improve the efficacy of anti-PD-1 therapy. Supplementary material online supplemental file 1 [214]jitc-13-5-s001.docx^ (18.3MB, docx) DOI: 10.1136/jitc-2025-011648 online supplemental file 2 [215]jitc-13-5-s002.xlsx^ (786.3KB, xlsx) DOI: 10.1136/jitc-2025-011648 Footnotes Funding: This work was funded by the National Natural Science Foundation of China (82102959, 82473390, 82272774, 82072666, 82172799, 82403703), China National Postdoctoral Program for Innovative Talents (BX20240089), China Postdoctoral Science Foundation (2024M750544), Zhejiang Provincial Natural Science Foundation of China (LQN25H160024) and Medical and Health Science and Technology Plan Project of Zhejiang Province (2025699757), Noncommunicable Chronic Diseases-National Science and Technology Major Project (2024ZD0520400, 2024ZD0520401). Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Ethics approval: The study was reviewed and approved by the Ethics Committee of Zhongshan Hospital, Fudan University, Shanghai, China (ID: Y2023-260). The study was performed in accordance with the principles set by the Declaration of Helsinki. All patients signed written informed consent for therapy and study inclusion. The animal study was approved by the Institutional Animal Care and Use Committee of Zhongshan Hospital, Fudan University (ID: 2023-242). Data availability statement No data are available. References