Single-cell RNA sequencing reveals the mediatory role of cancer-associated fibroblast PTN in hepatitis B virus cirrhosis-HCC progression

Fibroblast clustering identification based on single-cell RNA-seq profiling

The single-cell RNA-seq profiling of GSE125449 dataset was used to group the cells. In PCA analysis, the appropriate number of principal components was evaluated for further dimensionality reduction of cell groups. Herein, we selected the top 18 PCAs for cell clustering (Figure S1), and all the cells could be classified into 12 cell types (Fig. 1A). Then, the 12 cell types were annotated to six cell subsets, including subsets 2 and 8 which were annotated as fibroblast (Fig. 1B). The six-cell subsets contained 2009 marker genes, including 213 and 216 genes in subsets 2 and 8, respectively. Furthermore, we performed differential expression analysis on marker genes across cell subsets and represented the top 10% differentially expressed marker genes in each subset on the heatmap (Fig. 1C). There were significant differences in marker gene expression across cell subsets, indicating that these marker genes are involved in the mechanism regulating cell differentiation.

Fig. 1figure 1

The subclass of cancer-associated fibroblast cells was obtained through analysis of HCC cell subset based on single-cell transcriptome data. (A) Cells were clustered into various types using the two-dimensional UMAP projection, each color represents the annotated phenotype of each types, (B) UMAP showing the categorization of 12 cell types into six cell subsets, including subset 2 and 8 which were designated as fibroblasts. (C) Heatmap showing 10% of marker genes with significant differences among 6 subsets. (D) The subdivision of cancer-associated fibroblast cells. (left) The UMAP showed the clustering results of fibroblast cells (3 subclasses, namely CAF0-2), and (right) the results of the notes of 3 fibroblast cells. (E) Dot plot showing expression level of marker genes in the 3 subclasses of fibroblast cells. (F) The trajectory analysis of six cell subsets. (G) The pseudotime analysis of six cell subsets

The fibroblasts in subsets 2 and 8 were subdivided into 3 subclasses, namely CAF0-2 (Fig. 1D). All 3 subclasses were annotated as fibroblast (Fig. 1D), indicating the accuracy of cell sorting by single-cell RNA-seq profiling. The 3 subclasses comprised 59 marker genes, including 7, 26, 26 marker genes in CAF0, CAF1, and CAF2, respectively (Fig. 1E). As each of the six cell subsets (B cells, CD8 + T cells, fibroblasts, endothelial cells, monocytes, and adipocytes) performs a different function, we described the cell trajectory of CAF cell subsets. Cell trajectory branching occurs because cells exhibit different gene expression patterns and perform different biological functions. Further analysis showed that the differentiation trajectories of the six cell subsets were remarkably different, with fibroblasts and endothelial cells forming a distinct branch from B cells and CD8 + T cells (Fig. 1F). In addition, the pseudotime of fibroblasts and endothelial cells was earlier than that of B cells and CD8 + T cells (Fig. 1G), implying that fibroblasts and endothelial cells may be present in the very early stage of HCC and play an important function in HCC.

Meanwhile, the GSE151530 dataset was used for validation with similar analytic methods. By selecting the top 17 PCAs (Figure S2A, B) for cell clustering, we obtained seven cell subsets, including epithelial cells and six-cell subsets in the GSE125449 dataset (Figure S2C). Similar to GSE125449, the fibroblasts of GSE151530 can be further divided into 3 subclasses (Figure S2D), with 23, 50, 115 marker genes for sub-cluster0, sub-cluster1, and sub-cluster2, respectively. The top 10% differentially expressed marker genes in the 3 subclasses are represented in Figure S2E. Subsequently, by comparing the marker genes of GSE151530 with those of GSE125449 datasets, we found the three CAF subclasses in GSE125449 dataset with the majority of overlapping genes with GSE151530 (Figure S2F-H). These results demonstrated the accuracy of CAF mining by single-cell transcriptome data and the important functions of CAF cell subsets and their marker genes in investigating HCC.

Association analysis between CAF0-2 and clinical information

Six cell subsets were determined from single-cell RNA-seq profiling, including cells of the fibroblast type (CAF0-2). Three subclasses (CAF0-2) of marker genes of CAF cells were proposed to study the relationship between CAF and the prognosis of HCC using the TCGA HCC dataset. The expression profiles of 59 marker genes were determined to separate TCGA HCC samples into four distinct clusters (FCluster1-4) (Fig. 2A). Importantly, the overall prognostic survival of the four clusters showed significant differences, with worst prognosis for FCluster1, and FCluster2 and 3 had the best prognosis (Fig. 2B).

Fig. 2figure 2

Three subclasses of CAFs associated with clinical information. (A) Hierarchical clustering of four distinct clusters (FCluster1-4) based on 57 marker genes from three subclasses (CAF0-2). (B) Kaplan–Meier plots showing the overall survival and progression-free survival among the patients. (C) Hazard ratio of CAF score was evaluated through the cox Hazard Scale Model. (D) The distribution of CAF0, CAF1 and CAF2 scores among the four distinct clusters (FCluster1-4). (E) Association of CAF2 with the HCC pathological stage, tissue samples, and pathological grade. Mixed: mixed sample

According to the marker genes of CAF0, CAF1 and CAF2 identified on the GSE125449 single-cell transcriptome, we evaluated the scores of TCGA HCC samples in these three subclasses, and took the average expression value of their marker genes as the CAF score in each HCC sample. Cox proportional hazards model was used to evaluate the relationship between CAF0-2 scores and overall survival. The results showed that the hazard ratios of CAF0-2 scores were all less than 1, indicating that they may be good prognostic factors, but only CAF2 had statistical significance (p = 0.033) (Figure C). Subsequently, we compared the CAF0-2 scores among the FCluster1-4, and found no significant difference between the score of CAF0 and those of the four groups of samples (p = 0.14). CAF1 score showed a gradual increase in FCluster1-4. CAF2 score for FCluster1 and 4 with poor prognosis was low but high in FCluster2 and 3 with good prognosis (Fig. 2D). In addition, CAF2 score was negatively correlated with pathological stage and pathological grade, and CAF2 score was highest in hepatocellular sample and lowest in mixed sample (Fig. 2E). However, we could not reveal any association between CAF2 score and tissue type, age, and sex of HCC (Figure S3). For CAF0 and 1, no significant relationship was observed between CAF score and clinical information (Figure S4A-L). These results indicated that CAF2 was an effective indicator for exploring the role of CAF in HCC.

Analysis of communication between six cell subsets

To obtain the interaction between CAF and other cell subsets, we analyzed the possible communication links between CAF0-2 and different cell subsets. CAF2 had the maximum interaction frequency with CAF0, CAF1, and endothelial cells, and some frequency communication with B cells and CD8 + T cells (Fig. 3A). Meanwhile, CAF2 had the highest weight intensity of interaction with B cells and CD8 + T cells (Fig. 3B). Moreover, we analyzed how each cell subset interacted with other cell populations. The results revealed that each of these cell subsets was associated with B cells and CD8 + T cells (Figure S5), especially CAF0-2 showed high frequency of interaction with B cells, CD8 + T cells, and endothelial cells (Figure S5A, E, G). Considering that B cells and CD8 + T cells are immune cells, CAF may be involved in the remodeling of the immune microenvironment.

Fig. 3figure 3

Communication links among the six cell subsets. The network of (A) frequency of interaction and weight intensity for the interaction among the six cell subsets

CAF2 communicated weight analysis of signaling pathways

Since CAF2 was identified as a significant factor in HCC, we intensively investigated the signaling pathways that contribute to CAF2 communication. The top 10 signaling pathways with the highest interaction probability and the most contribution to CAF2 communication were PTN, followed by MK, VEGF, GAS, PDGF, EDN, TWEAK, CCL, ANGPT and IGF (Fig. 4A). The interactions among six cell subsets which contributed the most to signaling pathways were CAF2, B cells, and CD8 + T cells (Fig. 4B). Subsequently, we further concentrated on the top five CAF2-related signaling pathways with the largest contribution. The results showed that CAF2 was more or less associated with B cells and CD8 + T cells in all CAF2-related signaling pathways (Fig. 4C-G). Particularly, communication between CAF2 and B cells had the highest weight, followed by the communication between CAF2 and CD8 + T cells to the contribution of PTN signaling pathways (Fig. 4E). These results suggested that PTN mediated the effects of CAF2 -on cancer-associated fibroblast in HCC, and the interaction between CAF2-B cells and CAF2-CD8 + T cells may significantly regulate the mediation of PTN.

Fig. 4figure 4

Analysis of CAF2-related signaling pathways. (A) The top 10 signaling pathways associated with CAF2. (B) The 10 signaling pathways linked to the interactions among six cell subsets. (C-G) The network of contribution weight of the top 5 CAF2-related signaling pathways (MK, VEGF, PTN signaling, GAS, PDGF) to the communication among six cell subsets

Analysis of key marker genes in signaling pathways

Since dysregulation of gene expression was considered to be the origin of CAFs, we calculated the average expression levels of marker genes of top five CAF2-related signaling pathways in cell subsets (Fig. 5A-E). The expression of PTN, SDC1, and NCL was highest in CAF2, B cells, and CD8 + T cells, respectively (Fig. 5C), which implied that PTN and SDC1 played an important role in the communication between CAF2 and B cells. SDC2 was also highly expressed on CAF1 and CAF0, suggesting potential as a key marker gene shared among CAF0-2 subclasses (Fig. 5C). Subsequently, we analyzed the contribution of marker genes interaction of top five CAF2-related signaling pathways to the communication between cell subsets. The results revealed that the interaction of PTN-SDC1 and PTN-NCL contributed the most to the CAF2-B cells interaction, and the interaction of PTN-NCL showing the most weight intensity on the interaction of CAF2-CD8 + T cells (Fig. 5F). Moreover, the interaction of PTN-NCL contributed to almost every communicated interaction between cell subsets (Fig. 5F). Therefore, the PTN-SDC1 and PTN-NCL interactions may regulate cancer-associated fibroblast in HCC by modulating B cells and CD8 + T cells, and PTN may be a novel mediator of CAF in HCC.

Fig. 5figure 5

Analysis of the key marker genes among the top five CAF2-related signaling pathways. (A-E) The average expression levels of marker genes of the top five CAF2-related signaling pathways among the six cell subsets. (F) The contribution of marker genes of the top five CAF2-related signaling pathways to the communication among six cell subsets

Expression of PTN among NAFLD, cirrhosis, and HCC

Liver fibrosis/cirrhosis is an important pathological change in the early stage of HCC. As hallmark features of TME in HCC, we postulated that PTN may mediate the role of CAF in cirrhosis-HCC progression. Hence, we compared the PTN expression level among NAFLD, cirrhosis, and HCC. The results showed that PTN was highly expressed in cirrhosis compared with NAFLD and HCC (Fig. 6A), suggesting its role in initiating liver fibrosis/cirrhosis. Additionally, the expression level of PTN showed an increasing trend in liver cirrhosis samples with disease stage (F0-F4) (Fig. 6B). In the NAFLD grouping based on NAS score, the expression level of PTN in the samples with NAS score = 7 and 8 was significantly higher than that in the other groups (Fig. 6C). In addition, we found that PTN was significantly higher in cirrhotic HCC than in non-cirrhotic HCC (Fig. 6D). Moreover, PTN was highly expressed in active viral replication chronic carrier HCC, followed by chronic carrier (CC) HCC when compared with non-HBV HCC (Fig. 6E). These results further suggested PTN as possible a novel mediator of CAF in HCC, especially for HBV related cirrhosis-HCC progression. Subsequently, we investigated the relationship between PTN expression and immune of TME, and found a significant correlation between the expression of PTN and the fraction of 11 cells. In particular, PTN had the most significant association with B cells memory and T cells CD8 (Figure S6), which further confirmed PTN’s role in mediation of B cells and CD8 + T cells in the cancer-associated fibroblast-related HCC.

Fig. 6figure 6

The expression of PTN in different tissues and histopathological stages. (A) The differential expression of PTN among NAFLD, cirrhosis, and HCC. (B) Expression of PTN among different stages of cirrhosis. (C) The expression level of PTN in the NAFLD groups based on NAS score. (D) The differential expression level of PTN between cirrhotic HCC and non-cirrhotic HCC. (E) The differential expression level of PTN among non-HBV HCC, chronic carrier HCC, and active viral replication chronic carrier HCC.

Verification of the modulatory role of PTN in HBV cirrhosis-HCC progression

Since HBV infection is an important cause of HCC and liver fibrosis/cirrhosis is the key intermediate link of HBV-HCC, we explored whether PTN can mediate the role of CAF in cirrhosis-HCC progression. Thus, we further investigated whether HBV infection affected the mediatory role of PTN on cancer-associated fibroblast-related HCC. Using plasmid pcDNA3.1-PTN used for overexpression of PTN, we verified PTN overexpression using qPCR and WB both in Hep3B and Huh7 cell lines (Figure S7A, Figure S8A). Hep3B was first marked with EDU and then was quantified using FCM and IF, which showed that the activity of cell in the overexpression of PTN was significantly increased (Fig. 7A, B). The CCK-8 (Fig. 7C), transwell assays (Figure S7B), and wound healing assay (Figure S7C) showed that the overexpression of PTN promoted the Hep3B cell line proliferation, invasion, and activity. However, we could not observe the distinct difference among these methods, including FCM (Figure S8B), CCK-8 (Figure S8C), wound healing assay (Figure S8D), transwell assays (Figure S8E) and IF (Figure S8F), when the overexpression of PTN was conducted in the Huh7 cell lines.

Fig. 7figure 7

The effect of PTN on hepatocarcinogenesis, and cirrhosis-HCC progression following HBV infection. The effect of PTN overexpression on proliferation of Hep3B cells as determined by (A) FCM, (B) IF, and (C) CCK-8. The proliferation of Huh7 cells transfected with pHBV1.3 or with PTN overexpression or normal plasmid as determined by (D) FCM, (E) IF, and (F) CCK-8.

Since Hep3B was HBV-positive cell whereas Huh7 was HBV-negative cell, we hypothesized that HBV infection may play an important role in initiating PTN-mediated tumor-associated fibrosis in HCC. Therefore, the Huh7 was transfected with pHBV1.3 containing complete HBV genome to simulate HBV-positive cell. As expected, the simultaneous transfection with plasmid pcDNA3.1-PTN and HBV significantly promoted the cell activity, whereas the effect was weakened when transfection with pcDNA3.1-PTN and HBV alone ((Fig. 7D, E). In addition, the CCK-8 (Fig. 7F), transwell assays (Figure S9A), and wound healing assay (Figure S9B) yielded similar results. These results provided evidence for PTN’s role as a novel mediator of cancer-associated fibroblast in HBV infection-caused cirrhosis-HCC progression.

留言 (0)

沒有登入
gif