Endothelial cells expressing CPE and vWF are involved in the Immunopathogenesis of primary biliary cholangitis

Abstract

Primary biliary cholangitis (PBC) is an immune-mediated, non-suppurative cholestatic liver disease. Liver sinusoidal endothelial cells (LSECs) play a pivotal role in maintaining hepatic immune tolerance. Emerging research indicates that under certain stimuli, LSECs can transition from a tolerogenic to an immunogenic role. We hypothesize that LSECs may be implicated in the pathogenesis of PBC. Single-cell RNA sequencing (scRNA-seq) data were initially analyzed using R statistical software and Cell Ranger. Differentially expressed genes and marker genes were identified using Seurat. Preliminary cell type identification was conducted with SingleR cell clustering. Differential gene expression was determined using t-tests, and significance analysis was performed with the Limma package. Pseudotime analysis was conducted with Monocle2. Compared to the control (CTR) group, the number of endothelial cells in the PBC group was significantly reduced (P < 0.05). Further analysis of these endothelial cells revealed seven distinct subpopulations, including a newly defined CPE⁺ vWF⁺ endothelial cell type. Interaction between CPE⁺ vW ⁺ endothelial cells and bile duct cells was mediated through the APP-CD74 axis. Expression levels of CD74 and MIF were significantly higher in patients with PBC compared to CTR. CD74, serving as a receptor for the pro-inflammatory cytokine MIF, may counteract the anti-inflammatory effects of glucocorticoids. Expression levels of PTPRC and CCL5 were positively correlated with hepatic inflammation and fibrosis severity and were elevated in patients with PBC. CPE⁺ vWF⁺ endothelial cells might play a promising role in contributing to bile duct cell injury in patients with PBC by upregulating the pro-inflammatory factor MIF and interacting with CD74. Additionally, another unidentified endothelial cell type was suggested to exacerbate biliary damage and fibrosis by upregulating PTPRC and CCL5 expression.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-07311-z.

Introduction

Primary biliary cholangitis (PBC) is an immune-mediated, non-suppurative cholestatic liver disease, characterized by small interlobular bile ducts destruction, progressive cholestasis, and finally cirrhosis and liver failure without appropriate treatment¹. PBC, which mainly develops in middle-aged women, has its unique disease-specific autoantibodies – antimitochondrial antibodies (AMAs)². Ursodeoxycholic acid (UDCA) is now the only first-line drug approved by the Food and Drug Administration (FDA). However, about 30–40% of PBC patients had inadequate biochemical responses to UDCA, with a high risk progressing to end-stage liver diseases³. Therefore, understanding the immunological, cytogenetical, and biological molecular mechanisms of PBC pathogenesis is critical in exploring an ideal treatment strategy in patients with a bad prognosis.

Single-cell RNA sequencing (scRNA-seq), a next-generation sequencing (NGS)-based approach, is emerging with the capacity of uncovering complex and rare cell population, revealing regulatory relationships among genes, and tracking the developmental trajectory of different cell lines⁴. This technology has its own unique advantages in bioinformatics mining and phenomenon explaining with high-throughput and long-read single-cell transcriptome analysis⁵. Many studies based on this technology has played their part in promoting the understanding of varies autoimmune liver diseases (ALDs).

Histologically, livers are constructed with hepatocytes, cholangiocytes, and varies non-parenchymal cells (NPCs) functioning in hexagonal-shaped lobules. Each lobule is consisted with a central vein, radial sinusoidal networks, and portal triads which includes arteries, veins and bile ducts. Blood vessels and lymphatics in lobules are lined with liver endothelial cells (LECs), working together with diverse immunological cells⁶. As the portal vein branches into terminal portal venules, the LECs transform into spindle-shaped, non-fenestrated cells with short microvilli. These venules transition into hepatic sinusoids, where the LECs are smooth, large, and contain many actin fibers. Liver sinusoidal endothelial cells (LSECs) comprise 50% of the nonparenchymal cells of the liver. The pressure drop between the hepatic arteriole and the sinusoid is moderated by a precapillary sphincter. In the periportal region of the sinusoid, LECs have large-diameter fenestrae. This setup ensures a smooth transition of blood flow from the high-pressure arterial system to the low-pressure sinusoidal system, protecting the delicate hepatic tissue from potential damage caused by sudden pressure changes⁷.

LSECs play a crucial role in both promoting immune tolerance and enhancing immune responses in the liver. LSECs can function in a tolerogenic manner under normal conditions, which is essential in preventing autoimmune responses by inducing regulatory T cell phenotypes and blocking immunogenic changes. However, the studies of hepatitis B virus (HBV) infection demonstrate that LSECs can switch from a tolerogenic to an immunogenic role. The ability of LSECs to promote CD8 + T cells and NK cells’ antiviral activity shows their flexibility and importance in immune responses⁸. The fenestrated morphology of LSECs has the ability of allowing these CD8 + T cells to travel through the fenestrae and exert their effector function. Besides, LSECs’ production of cytokines like IL-27 and TNFα and their interaction with molecules such as nucleotide-binding oligomerization domain 1 (NOD1) suggests that they play a role in shaping the immune environment in the liver^9–11. Moreover, LSECs acting as antigen-presenting cells (APCs) express major histocompatibility complex (MHC) class I and class II molecules, which are involved in presenting endogenous and exogenous antigens to CD8 + and CD4 + T cells, respectively^12,13. These immunological features of LSECs suggest their crucial roles in autoimmune liver diseases, where an aberrant immune response might be initiated or exacerbated.

Related studies have proved the immunological role of LSECs in pathogenesis of liver diseases. And sinusoidal fibrosis was identified to be an independent prognostic value beyond the Mayo Score in PBC. However, the specific role of LSECs in the pathogenesis of PBC remains unclear. A recent study utilizing scRNA-seq has identified a unique population of DUOX2⁺ ACE2⁺ small cholangiocytes as key pathogenic targets in PBC. Building on this study, we aim to investigate the histological, functional, and spatial characteristics of LSECs and their role in immune processes within PBC liver samples.

Methods

Data availability and processing of HRA002347

The raw data of 3’-scRNA-seq for control (CTR) and PBC liver tissues from human was downloaded from the PRJCA009122 project of the National Genomics Data Center, China National Center for Bioinformation (https://ngdc.cncb.ac.cn/), coded as HRA002347 in Genome Sequence Archive (GSA)¹⁴.

We analyzed scRNA-seq data using R statistical software (v. 4.1.2, The R Foundation for Statistical Computing, Vienna, Austria). Cell Ranger (https://support.10xgenomics.com/single-cell-gene-expression/software/overview/welcome), the official analysis software of 10x Genomics, were used to filter the original data, compare genomes, quantify transcripts, identify cells, and finally obtain the gene expression matrix of each cell. Subsequently, Seurat (https://satijalab.org/seurat/) was used for further cell filtration, quantitative standardization of cell expression, classification of cell subgroups, analysis of differentially expressed genes of each subgroup and marker gene screening¹⁵. Then we used SingleR to conduct preliminary cell type identification and single cell data mining for the cell clusters we identified.

Low-quality cells (gene number < 500; > 25% mitochondrial genes) were initially filtered, and the data were normalized. Batch effects were removed using the Harmony function (resolution = 0.2). Dimensions were reduced using Uniform Manifold Approximation and Projection (UMAP). We found marker genes for cell groups in datasets using the FindAllMarkers function. The output standards of marker gene are as follows: (1) Genes are expressed in more than 10% of the cells in target or CTR subpopulation; (2) P value ≤ 0.01; (3) The multiple of gene up-regulation ≥ 20.26.

Cell types were identified in each subpopulation based on the datasets of Human Primary Cell Atlas, including 36 main cell types. Cholangiocytes were identified according to the markers from previous referenced article¹⁴. In differentially expressed genes (DEGs)analysis, significance was analyzed using Limma package. The most significant changes in gene expression were analyzed with false discovery rates (FDR < 0.01) and absolute fold change (FC > 1). Using monocle2, a total of 2424 cells from 7 endothelial subclusters were selected for subsequent pseudo-time analysis.

Data availability and processing of GSE79850

Data of GSE79850 were acquired from National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE79850). This dataset contains RNA levels of 24 human liver biopsy, including 9 PBC patients of high risk who went on to require liver transplantation, 7 PBC patients of low risk who responded fully to UDCA, and 8 non-diseased CTR liver. T test was performed to identified genes expressed with significant different levels.

Results

Sequencing statistics and quality CTR

The statistical results of sequencing data of quality CTR processed with CellRanger and cell filtration results with Seurat are shown in Tables 1 and 2, respectively. Distribution of basic cell information of each sample before and after cell filtration was demonstrated in Fig. 1. Cells with poor quality usually have abnormally high mitochondrial gene expression. Mitochondrial gene expression can be used as a reference index to identify low-quality cells in the sample. Due to the abnormally high expression of mitochondrial gene, median genes per cell less than 1,000 and sequencing saturation lower than 60% of CTR4, we eliminated the sequencing data of this sample and performed the following analysis with the remaining 8 samples.

Table 1.

Quality CTR of sample sequencing data.

Sample	Estimated number of cells	Mean reads per cell	Median genes per cell	Sequencing saturation
CTR1	14,351	32,443	1090	68.2%
CTR2	6431	61,738	1618	71.9%
CTR3	7671	52,483	1339	73.8%
CTR4	17,703	23,786	888	57.1%
PBC1	9955	36,219	1377	63.4%
PBC2	12,536	30,509	1490	45.4%
PBC3	10,250	41,533	1494	73.5%
PBC4	5032	72,702	1752	79.2%
PBC5	17,791	26,813	1031	70.1%

Table 2.

Sample sequencing data of before and after cell filtration.

Sample	Cell number before filtration	Cell number after filtration	Percentage after filtration	Median of gene number per cell before filtration	Median of gene number per cell after filtration
CTR1	14,351	8258	57.54%	911	1062
CTR2	6431	4114	63.97%	1130	1292
CTR3	7671	5188	67.63%	1047	1136
CTR4	17,703	9481	53.56%	882	1107
PBC1	9955	7493	75.27%	1122	1157
PBC2	12,536	9042	72.13%	1490	1536
PBC3	10,250	8321	81.18%	1146	1169
PBC4	5032	4084	81.16%	1343	1382
PBC5	17,791	12,078	67.89%	897	942

Fig. 1.

Basic cell information of each sample before and after cell filtration. (a) Data distribution before filtration. (b) Data distribution after filtration. Left figure: the number of genes detected in a single cell in each sample (Y-axis). Middle figure: The total number of Unique Molecular Identifier (UMI) detected in individual cells of each sample (Y-axis) distribution. Right figure: Percentage (Y-axis) distribution of mitochondrial gene expression in individual cells of each sample.

Dimension-reduction and main cell-type clustering

We performed uniform manifold approximation and projection (UMAP) analysis for main cell-type clustering. 25 cell clusters (Fig. 2a) involving 12 cell lineages (Fig. 2b) were identified, separately displayed regarding CTR and PBC livers (Fig. 2c). Percentage of each cell lineage of 8 samples and in 2 groups were listed in Fig. 2d. Significant differences were identified in endothelial cells, hepatocytes, neutrophils, and plasmacytoid dendritic cell between PBC and CTR group (Fig. 2e). Every cell lineage has its specific maker genes (Fig. 2f).

Fig. 2.

Visualization of cell clustering. (a) 25 cell clusters identified in UMAP analysis of liver samples from 3 CTR and 5 PBC patients. (b) 11 cell lineages identified based on 25 cell clusters. (c) 11 cell lineages separated into CTR (left) and PBC (right) group. (d) Percentage of each cell lineage of 8 samples (right) of 2 groups (left). (e) Differences across 11 cell lineages between PBC and CTR group. (f) Representing marker genes of each cell lineage. *P < 0.05, **P < 0.01.

Significant differences were identified in four cell types

As shown in Fig. 2d, significant differences were identified in cholangiocytes, endothelial cells, plasmacytoid dendritic cells, and T lymphocytes. The first two cell types have less cell counts in PBC group, while the number of plasmacytoid dendritic cells and T lymphocytes are significantly higher compared to CTR group. The top 10 of up and down regulation genes in four cell types of PBC group compared to CTR group were displayed in Fig. 3, respectively. For the top significant up-regulated genes (significant = true, P value = 0.00, Q value = 0.00) in four cell types [gene number: cholangiocytes = 127 (Supplementary Data 1), endothelial cells = 696 (Supplementary Data 2), T lymphocytes = 254 (Supplementary Data 3), plasmacytoid dendritic cells = 11 (Supplementary Data 4), the top 6 results with the highest rich factor and strongest enrichment effect (Supplementary Fig. 1) were listed.

Fig. 3.

Top 10 of up and down regulated genes in four cell types of PBC group compared to CTR group. (a) Cholangiocytes. (b) Endothelial cells. (c) Plasmacytoid dendritic cells. (d) T lymphocytes.

Seven types of endothelial cells were identified in human livers

To explore the participation of different types of endothelial cells in PBC pathogenesis, we performed UMAP analysis on endothelial cells and 7 types of endothelial cells were identified (Fig. 4a). By separating endothelial cells into PBC and CTR group, we could find that except of endothelial cell 3 and 4, cell number of the remaining seven groups dramatically decreases in PBC group compared to CTR group (Fig. 4b and c). Heatmaps of top 10 expression genes in 7 types of endothelial cells were displayed in Fig. 4d.

Fig. 4.

Seven types of endothelial cells in human livers. (a) Seven cell clusters identified across all samples. (b) Seven cell clusters displayed in CTR and PBC groups. (c) Comparison of the percentages of seven cell clusters between CTR and PBC groups. (d) Heatmaps of top 10 expression genes in 7 types of endothelial cells.

CPE⁺ vWF⁺ endothelial cells are key in immunological involvement of PBC pathogenesis

To further elucidating physiological features of each cell type, literature review was performed and 9 genes (HEY1, RGCC, EDNRB, VWA1, ESM1, CD36, ACKR1, CCL21, PROX1) representing 5 cell types (arteries, angiogenic ECs, capillaries, vein, lymphatic ECs) were selected as canonical marker genes, respectively¹⁶. Gene average expression levels of 7 types of endothelial cells were marked as red to blue colors from maximum to minimum value. As shown in Fig. 5a, we could define endothelia 2 as arteries and angiogenic ECs, endothelia 3 as capillaries, endothelia 5 and 7 as veins, and endothelia 6 as lymphatic ECs. According to t-test analysis (Fig. 4c), lymphatic ECs were the only known endothelial cell type significantly destructed in liver samples of PBC patients. Besides, endothelia 2 and 5 highly expressed cell markers of liver sinusoidal endothelial cells (LSEC) - FLT4, LYVE1, STAB2, and CD36 in PBC group, while endothelia 6 mainly expressed markers in peri-central zone (WNT2, WNT9B, and LHX6)¹⁷. Compared to CTR group, endothelial 1 significantly decreased and endothelial 4 significantly increased in PBC patients, respectively, while these 2 cell types could not be defined by canonical marker genes. CPE⁺ vWF⁺ endothelial cells (endothelia 1) significantly decreased in PBC group, which were defined as an intermediate-state cell type in primary liver cancer¹⁸. Endothelia 4 with the marker gene PTPRC and CCL5 were significantly higher in PBC patients than CTR group, which were found to be positively correlated with the degree of liver inflammation and fibrosis¹⁹. Enrichment analysis of endothelia 4 identified dominant immunological features. After selecting the most significant different genes and ranking by the rich factor from high to low, the top 10 enrichment pathways of GO (Supplementary Fig. 2a) and KEGG^20–22 (Supplementary Fig. 2b) analysis were listed.

Fig. 5.

Endothelial cells definition and cell-cell interactions analysis. (a) Definitions of 7 types of endothelial cells. (b) Cell-cell interactions among endothelial cells, cholangiocytes and immune cells in CTR (left) and PBC (right) patients. (c) Genes involved in cell-cell interactions (left: CTR group, right: PBC group).

Next, we performed cell-cell interactions analysis among cholangiocytes, endothelial cells and immune cells. Results in Fig. 5b revealed a remarkable interaction between endothelial cells and cholangiocytes, and their interactions with mononuclear phagocytes (MPs) in CTR group. While in PBC patients, the interactions between cholangiocytes and other cells were weakened except for arteries, veins and CPE⁺ vWF⁺ endothelial cells. Figure 5c demonstrated genes involved in cell-cell interactions based on the cell clustered we identified. The two genes on the y-axis correspond to the two cell populations indicated on the x-axis, respectively. This result further proved the immunological role of CPE⁺ vWF⁺ endothelial cells and their interactions with cholangiocytes possibly through APP-CD74 and ALB-FcRn complex. Besides, APP-CD74 interactions were identified to be extensively high-regulated and reported to be related with biliary atresia²³.

Pseudo-time analysis revealed inflammation and destruction in PBC progression

Cells in the tree structure trajectory diagram are displayed according to the cell clustering results of monocle2. Different colors represent cells from distinct states. 5 cellular states (Fig. 6a left) were identified based on 7 endothelial cell subclusters (Fig. 6a middle) according to Seurat from CTR and PBC group (Fig. 6a right). The top 6 representative genes were shown in Fig. 6b, with annotation based on state and subcluster. Among them, TM4SF1 was reported to be endothelial cells with stem/progenitor potential in a study focusing on pulmonary arterial hypertension²⁴. By inferring the differentiation trajectory, we observed a shift in cellular composition, with classical endothelial cells diminishing and inflammatory-related endothelial cells (Endothelial 4, corresponding to state 5 in Fig. 6a, left) becoming more prominent. While these findings suggested potential endothelial remodeling during the progression of PBC, further multi-dimensional validation would be necessary to confirm the biological significance of the observed changes in endothelial subpopulations. Additionally, Fig. 6c displayed a heatmap of the most dynamically expressed genes along pseudo-time, and Fig. 6d illustrated changes in gene expression across disease progression.

Fig. 6.

Pseudo-time analysis of endothelial cells. (a) Five cell states (left) identified based on seven endothelial cell subclusters (middle) using Seurat from CTR and PBC groups (right). (b) Representative genes (top 6) associated with each state and cell type. (c) Heatmap showing expression of top-changed genes during disease progression. (d) Dynamic changes of endothelial subclusters along disease course.

Discussion

Previous studies proved that diverse T cells play a key role in the pathogenesis of PBC²⁵, including CD8 + T cell infiltration in the liver directly correlated with biliary ductular damage²⁶, and Tregs played in the loss of tolerance in PBC²⁷. A 100-fold increase in the frequency of PDC-E2–specific CD4 + T cells has been observed in the hilar lymph nodes and liver compared to PBMCs from patients with PBC²⁸, and a 10-fold increase in the frequency of PDC-E2-specific cytotoxic T lymphocytes has been reported in the liver compared to the blood²⁹. The immune function of endothelial cells has been extensively studied. In liver tissue, LSECs facilitate the adhesion of liver-resident immune cells and cross-prime naive CD8 T cells, leading to their differentiation into memory T cells, which migrate to secondary lymphoid tissues, offering protection upon re-encounter with antigens during infections, helping to maintain immune tolerance within the liver³⁰. This study mainly focusses on the immunological role of endothelial cells in the pathogenesis of PBC.

In our study, the expression of cholangiocytes, endothelial cells and plasmacytoid dendritic cells decreased significantly, and T cells increased significantly in PBC patients compared to CTR group. The mechanism of cholangiocytes in PBC has been extensively investigated, but related studies focusing on endothelial cells are still few. In this study, 7 types of endothelial cells were identified. All endothelial cell subclusters were classified as arterial, venous, capillary, or lymphatic ECs, except for two subclusters that remained undefined.

vWF is a multimeric protein secreted by endothelial cells and involved in hemostasis³¹. CPE⁺ vWF⁺ endothelial cells (endothelia 1) were identified to be an intermediate-state cell type in primary liver cancer, which were more highly expressed in tumor-peritumor junctional zone (J zone) than in tumor (T) or peritumor (P) zone¹⁸. This type of endothelial cells was functionally similar to the recently reported fetal-like endothelial cells in HCC, defined as tumor-Specific PLVAP + ECs in Human Fetal Liver³². In our analysis, this cell type was significantly decreased in liver samples of PBC patients compared to healthy controls, which suggests a destruction in immunological reaction and a possible negative correlation between PBC and primary liver cancer.

CD 74 is an HLA class II histocompatibility antigen gamma chain, which plays a pivotal role in MHC class II antigen processing by stabilizing peptide-free class II α/β heterodimers in a complex, and directing transport of the complex from the endoplasmic reticulum (ER) to the endosomal/lysosomal system, where the antigen processing and antigenic peptides binding to MHC class II takes place. CD 74 serves as a cell surface receptor for the cytokine MIF (macrophage migration inhibitory factor)³³. MIF is a pro-inflammatory cytokine, which is involved in the innate immune response to bacterial pathogens. The expression of MIF at sites of inflammation suggests a role as mediator in regulating the function of macrophages in host defense. MIF could counteract the anti-inflammatory activity of glucocorticoids by overriding glucocorticoid-mediated inhibition of cytokine secretion by lipopolysaccharide (LPS)-stimulated monocytes and to overcome glucocorticoid protection against inflammation³⁴. As we found that the expression of CD 74 and MIF (data not shown here) is significantly higher in PBC patients compared to CTR group, this mechanism may explain the destruction of CPE⁺ vWF⁺ endothelial cells and the fact that PBC does not respond well to glucocorticoids therapy.

The expression levels of PTPRC and CCL5 were found to be positively correlated with the degree of liver inflammation and fibrosis¹⁹ and were found to be highly expressed in other autoimmune diseases such as psoriasis³⁵. This type of endothelial cells was identified as endothelia 4 in Fig. 4b, in which expression levels were significantly higher in PBC patients than CTR group, suggesting its pro-inflammatory feature in the pathogenesis of PBC. GO analysis of endothelia 4 (Supplementary Fig. 1) revealed an enrichment in immune cells including NK cells and T cells, suggesting its immune mediated role in PBC pathogenesis.

This article further verified the original work published¹⁴ based on scRNA-seq data of liver samples in patients with PBC, and served as supplements in the perspective of endothelial cells during PBC pathogenesis. In this article, we identified a functional classification of endothelial cells and their possible role in PBC development and treatment. Further studies are required referring to mechanical exploration in molecular and cellular levels.

Conclusion

This study extends the findings of previous research that utilized scRNA-seq data from liver samples of patients with PBC [14], offering a focused examination of endothelial cells in the context of PBC pathogenesis. Through our analysis, we identified distinct subpopulations of endothelial cells and elucidated their potential roles in PBC development and therapeutic response. Specifically, we discovered a subset of CPE⁺ vWF⁺ endothelial cells that may exacerbate bile duct cell injury by upregulating the pro-inflammatory mediator MIF, which subsequently interacts with CD74. The ability of MIF to counteract the anti-inflammatory effects of glucocorticoids could provide insight into the resistance to hormone therapy observed in some PBC patients. Additionally, in endothelial 4, we detected significantly elevated expression levels of PTPRC and CCL5, suggesting their involvement in aggravating biliary injury and fibrosis. These results unveil new potential therapeutic targets for PBC treatment. However, further mechanistic studies at the molecular and cellular levels are warranted to deepen our understanding and validate these findings.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Review and project management: B.F. Writing original draft, data analysis, and literature review: LX.H. and ZX.Q. Data collection, validation, and revision: ZL.W. and XX.W. All authors reviewed the manuscript.

Data availability

Data of GSE79850 were acquired from National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc=GSE79850). All data used in this study were available in the original research. Data generated in this study were included in the main text and Additional files.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

All data used by this study were publicly available from participant studies with approvement of ethical standards committee related to human experimentation. No additional ethical approval was required in this study.