Inflammatory factor CCL2 enhances the interaction between monocyte-macrophage cells and liver parenchymal cells to promote liver inflammation and fibrosis in biliary atresia

Abstract

Background

Liver fibrosis in biliary atresia (BA) progresses rapidly and has distinct characteristics; however, current studies have not identified effective prevention or treatment strategies to address this issue.

Methods

BA liver tissues with different degrees of liver fibrosis (n = 4), liver tissues of choledochal cyst (n = 2), and liver tissues of the normal control (NC) group (n = 2) were selected. Single-cell RNA sequencing (scRNA-seq), spatial transcriptomics (ST), and Mendelian randomization (MR) were integrated for analysis. The clinical data of the sequenced samples, GSE176189 and GSE122340, were used to verify the results.

Results

The level of inflammation in the severe fibrosis group was significantly higher than that in the mild fibrosis group (adjusted P < 0.0001). The results of MR showed that CCL2 had a causal relationship with BA (odds ratio (OR) = 1.70, confidence interval (CI): 1.19 to 2.43, P = 0.004, P _{false discovery rate (FDR)} = 0.117). The expression level of CCL2 in BA was significantly higher than that in NC (P < 0.001), and its expression level increased with the progression of fibrosis, mainly expressed in the central region of fibrosis. The pseudo-timing results of scRNA-seq showed that CXCL10 + intermediate monocytes may play a significant role in the early stages of fibrosis progression, while TREM2 + scar-associated macrophages may be more active in the later stages. OLR1 + M2 macrophages may represent a transitional state between the two cell types described above. The expression of CCL2 in these three cell subtypes was also higher than that in the others. CCL2 + monocyte-macrophage cells showed the strongest correlation with gamma-glutamyl transferase (R = 0.88, P = 0.0072). The interactions between CCL2 + monocyte-macrophage cells and hepatocytes, hepatic stellate cells, and bile duct epithelial cells were significantly upregulated in BA (P < 0.01). These interactions were more prominent in mild fibrosis than severe fibrosis (P < 0.01).

Conclusions

Severe liver fibrosis in BA is associated with a pronounced inflammatory response. CCL2 may be crucial in the occurrence and progression of liver fibrosis in BA. Targeting CCL2 + monocyte-macrophage cells by reducing their proportion or interaction with liver fibrosis-related cells may provide a potential treatment for liver fibrosis in BA.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12887-025-05984-z.

Background

Biliary atresia (BA) is characterized by progressive fibrosis and inflammatory destruction of intrahepatic and extrahepatic bile ducts. If left untreated, this condition leads to biliary cirrhosis and liver failure [1]. The etiology of BA remains unclear. Possible mechanisms include genetic variation, exposure to toxins, viral infections, autoimmune-mediated chronic inflammation, and bile duct lesions [2]. Although the Kasai procedure temporarily alleviates extrahepatic biliary obstruction, approximately 60–80% of children with liver fibrosis continue to progress [3]. Eventually, they develop cirrhosis and liver failure, which necessitates further liver transplantation [4].

In contrast to liver fibrosis associated with other liver diseases, liver fibrosis in BA is distinctive, occurring early and progressing rapidly. Furthermore, existing studies have not identified effective prevention and treatment strategies for the ongoing progression of postoperative liver fibrosis. Therefore, elucidating the mechanisms of liver fibrosis in BA and effectively controlling its progression is crucial for prolonging the survival of the native liver and avoiding the need for liver transplantation.

The rise of single-cell RNA sequencing (scRNA-seq) allows us to explore cellular heterogeneity and functional characteristics with a nuanced view, revealing the complex interactions and state changes of different cell types within a tissue. The scRNA-seq enables researchers to analyze gene expression at the level of individual cells, while spatial transcriptomics (ST) further captures the spatial distribution of gene expression and helps us understand the interaction of cells in tissues and the influence of their microenvironment. At the same time, Mendelian randomization (MR) is used as a causal inference method to reveal the causal relationship between biomarkers and health outcomes through genetic variants as instrumental variables (IVs).

In this study, BA samples with different degrees of liver fibrosis and control samples were selected for scRNA-seq and ST sequencing, combined with clinical data and MR analysis, to provide a multi-level perspective for BA research, explore the causal relationship between genes and phenotypes, and the functional differences of cells in different microenvironments, and promote the understanding of the occurrence and development of liver fibrosis in BA.

Methods

The flow chart of this study is shown in Figure S1.

Sample collection

This study collected liver tissues from four BA patients (type III) with varying degrees of liver fibrosis (F1: n = 1; F2: n = 1; F3: n = 1; F4: n = 1) as disease group, liver tissues from two patients with choledochal cysts (CC) as disease control (DC) group, and two adjacent liver tissues (one patient with hepatoblastoma and one patient with hepatic hemangioma) as normal control (NC) group. In the DC and NC groups, samples were selected from younger cases to minimize the impact of age as a confounding variable (Table S1).

Inclusion criteria for children with BA will be as follows:

1. BA will be confirmed by surgical intervention.

2. The Kasai procedure will be performed.

3. Informed consent will be obtained from the patient’s family members.

All the patients were treated at the Department of General Surgery at Tianjin Children’s Hospital from March 2023 to January 2024. All the patients underwent surgical intervention performed by the same surgical team. Clinical and pathological data were extracted from histopathology reports and electronic medical records from the Department of General Surgery, Tianjin Children’s Hospital. The scRNA-seq was conducted on liver tissue cells from BA, DC, and NC patients. ST was also performed on BA liver tissues and normal tissues adjacent to hepatoblastoma. Sequencing and clinical information for these samples are detailed in Table S1.

This study was approved by the Ethics Committee of Tianjin Children’s Hospital (2022-SYYJCYJ-008), and informed consent was obtained from the legal guardians of all patients. Liver samples from patients in both the BA and control groups were collected during the surgical procedures.

Experimental methods for sequencing data

Experimental methods for single-cell data

Single-cell suspension preparation

Tissue samples were dissociated into single-cell suspensions using the sCelLiVE™ Tissue Dissociation Solution. The resulting cell suspensions were diluted to an appropriate 2.5–3.5 × 10⁵ cells/mL concentration.

Single-cell isolation and labeling

The cell suspension was introduced into the SCOPE-chip™ microfluidic device, where single cells were isolated based on the principle of “Poisson distribution”. Cells were allowed to settle by gravity into the microvias of the chip, ensuring that only one cell was deposited into each microvia. Subsequently, millions of magnetic beads, each carrying unique cell barcodes, were added to the microtiter wells, ensuring that only one bead occupied each well. After cell lysis, the magnetic beads with unique barcodes and unique molecular identifiers (UMIs) captured the mRNA by binding to the poly(A) tails, effectively labeling both the cells and their corresponding mRNA.

Reverse transcription and amplification

The magnetic beads in the microarray were collected to reverse transcribe the captured mRNA into complementary DNA (cDNA), followed by amplification.

Single-cell sequencing library construction

The cDNA was fragmented and ligated to construct a sequencing library compatible with the Illumina sequencing platform.

Single-cell sequencing data processing and quality control

Raw reads were processed with fastQC and fastp to remove low quality reads. Poly-A tails and adaptor sequences were removed by cutadapt. After quality control, reads were mapped to the reference genome GRCh38 (ensembl version 92 annotation) using STAR. Gene counts and UMI counts were acquired by featureCounts software. Expression matrix files for subsequent analyses were generated based on gene counts and UMI counts.

Experimental methods for ST data

ST experiments were conducted following the manufacturer’s protocol (10x Genomics). In brief, fresh frozen tissue was OCT embedded. HE-stained sections of 5 μm thick paraffin-embedded liver tissue samples were selected, incubated with probes in designated areas, and then transferred to 10x air-transferred slides using the CytAssist system for capture and library preparation. The Space Ranger toolkit (v.1.2.1) with GRCh38 reference genomes was used to align and quantify the raw fastq files.

Data collection

Gene microarray data (GSE122340 [5] and GSE176189 [6]) will be downloaded from the Gene Expression Omnibus (GEO) database. GSE122340 (RNA-seq) contains 171 BA samples and 7 NC samples. GSE176189 (scRNA-seq) contains 9 BA samples and 3 CC samples. BA of genome-wide association study (GWAS) data set is the queue from Europe and the United States, including 499 patients with BA and 1928 cases of the control group, GWAS summary data can be accessed through https://www.ebi.ac.uk/biostudies/ (S-BSST182) [7]. A genome-wide protein quantitative trait locus (pQTL) study of 91 inflammation-related plasma proteins involved 14,824 European participants [8]. Detailed information regarding each dataset is presented in Table S2.

Single-cell data analysis

Data preprocessing

All sets of cells were normalized and scaled with the “Seurat” package (version 4.3.0.1) [9] with parameters nFeature_RNA > 300, nCount_RNA > 1000, and percent_mt < 20. Doublets were removed from the samples by the “DoubletFinder” package (version 2.0.3) [10].

Subsequently, the batch effect between samples was removed using the “harmony” package (version 0.1.1) [11] and the scRNA-seq data was normalized and scaled with the “Seurat” package (version 4.3.0.1) [9]. Cell subgroups were identified using the “FindNeighbors” and “FindClusters” functions. The unified manifold approximation projection (UMAP) method was utilized as a visualization technique for cell clustering [12]. In conjunction with literature data, manual annotation was employed for the annotation process (Table S3) [13–19].

Differential analysis and enrichment analysis

To identify differentially expressed genes (DEGs), the “FindAllMarkers” function was employed. The “clusterProfiler” package (version 4.8.3) was utilized to compute cell-enriched pathways [20]. Benjamini-Hochberg (BH) was used for multiple testing correction of DEGs identification and enrichment analysis.

Tissue heterogeneity analysis

To quantify the heterogeneity of cell types across different tissues, the study compared the observed and expected number of cells for each cluster in each tissue using the following formula: Ro/e = (observed/expected). The expected number of cells in each tissue’s cell clusters was calculated using the Chi-square test [21]. If Ro/e > 1, it indicates that the number of cells in a cell cluster in the tissue is higher than random expectation, that is, it shows enrichment; If Ro/e < 1, it indicates that the number of cells of a certain cell cluster in that tissue is lower than random expectation, that is, it shows depletion.

Scoring gene sets

To characterize the correlations between cells and fibrosis and inflammation, the study utilized the “AddModuleScore” function to quantify each cell’s score based on different gene sets. The Bonferroni method was used to correct P values for comparisons between more than two groups. Among them, 33 liver fibrosis-specific matrisome genes (LFMGs) were obtained from the study of Chen et al. [22]. The dataset for “Severe Scores” is the marker genes of the severe group. The inflammation and cytokine gene sets were obtained from the study of Ren et al. [23] (Table S4).

Single-cell pseudo-time analysis

The cytotrace2 function from the “CytoTRACE2” package (version 1.0.0) was used to calculate the differentiation potential scores for each cell to distinguish between cells with high and low differentiation potential and visualize these scores using UMAP plots. Cells with the highest differentiation scores were selected as potential starting points for trajectory inference. Then, based on the inferred starting points, the “monocle” package (version 2.28.0) [24] was used to analyze the lineage differentiation of cell subtypes and their potential developmental relationships.

Single-cell interaction analysis

The “CellChat” package (version 1.6.0) [25] was employed to compute intercellular communication relationships. Selected human data set, and through “identifyOverExpressedGenes” and “identifyOverExpressedInteractions” function to identify the excessive expression of genes and the interaction between cells. Then, the gene expression data were mapped into the PPI network by the “projectData” function. The “computeCommunProb” and “filterCommunication” functions were used to estimate the communication probability and infer the cellular communications network. The “computeCommunProbPathway” and “aggregateNet” algorithms were used to infer cell-to-cell communication between each cell type at the signaling pathway level. The “netAnalysis_computeCentrality” function was used to calculate the centrality of nodes in the intercellular communication network.

ST data analysis

Data preprocessing

The ST was normalized and scaled with the “Seurat” package (version 4.3.0.1) [9]. Clusters were identified using the “FindNeighbors” and “FindClusters” functions. The UMAP method was utilized as a visualization technique for clustering [12]. Combined with the opinions of pathological experts, the various clusters of ST were annotated.

Resolution enhancement technique for ST data based on BayesSpace

To improve the spatial resolution of ST, BayesSpace, a full Bayesian statistical approach, was employed by the “BayesSpace” package [26] (version 1.10.1) in this study. This approach utilizes information from spatial neighborhoods to enhance the resolution of spatial transcriptome data and perform cluster analysis.

ST scores based on cell population markers

To evaluate the expression of cell populations in the spatial transcriptome data, the “AUCell” package (version 1.22.0) was used to score the spatial transcriptome data.

ST pseudo-time analysis

The “monocle” package (version 2.28.0) [24] was used to analyze the lineage differentiation of cells and their potential developmental relationships based on the spatial location of tissues to understand the differentiation path and order of cells in spatial tissues.

MR analysis

The detailed analysis process of MR can be referred to in our previous work [27]. P _{false discovery rate (FDR)} < 0.2 in MR analysis suggests statistical significance [28]. The LDtrait function of LDlink (https://ldlink.nih.gov/?tab=home) was used to remove confounding factors [29]. The instrumental variables obtained from the screening are input into the website as rsid, and then the website gives SNP-related traits. To determine whether these traits were confounders (e.g., viral infection and maternal exposure to certain chemicals). SNPs associated with the confounders were removed from subsequent MR analyses.

Single-cell deconvolution analysis of bulk data

To further validate the accuracy of our results, we performed single-cell deconvolution analysis on the GSE122340 cohort using the CIBERSORT web tool (https://cibersortx.stanford.edu/). The data preprocessing methods for bulk RNA sequencing (RNA-seq) data were based on our previous research findings [30].

Statistical analysis

R software (version 4.3.1) was used for statistical analysis. For comparisons of two or more continuous variables, an unpaired student’s t-test was performed for data following a normal distribution, and the Wilcoxon test or Kruskal-Wallis’s test was performed for data not normally distributed. Spearman’s correlation coefficient measures the correlation of two variables. Two-sided Fisher’s exact test measured whether rates differed between groups. When multiple hypothesis tests (n ≥ 2) are conducted on the same dataset, appropriate multiple testing correction should be applied. Significance levels are marked by single, double, triple, and quadruple asterisks and ns (*, **, ***, and **** denote significance levels of 0.05, 0.01, 0.001, and 0.0001, respectively, and ns denotes no significance level). P < 0.05 was considered statistically significant if not specified.

Results

Cell quality control

Table S5 presents the number of cells before and after quality control for eight samples from sequencing and twelve samples from GSE176189. Figure S2A-B shows the identified doublet group, which was subsequently removed. The cloud and rain plots of cell quality control for both datasets are shown in Figure S2C-D.

Severe liver fibrosis in BA is associated with a pronounced inflammatory response

In the single-cell data from the eight samples, all cells were classified into thirteen clusters: T cells, natural killer (NK) cells, B cells, plasma cells, monocyte-macrophage (mono-macro) cells, dendritic cells (DCs), neutrophils, endothelial cells, hepatic stellate cells (HSCs), biliary epithelial cells (BECs), proliferating cells, erythroblasts, and hepatocytes (Fig. 1A). In the GSE176189 dataset, all cells were classified into fourteen clusters: T cells, NK cells, B cells, plasma cells, mono-macro cells, DCs, neutrophils, endothelial cells, HSCs, smooth muscle cells (SMCs), BECs, proliferating cells, erythroblasts, and hepatocytes (Figure S3A). Figures 1B and S3B showed each cluster’s bubble plots of marker genes.

Fig. 1.

Severe liver fibrosis in BA is associated with a pronounced inflammatory response. A UMAP for each cell in the sequenced samples. B Bubble plots of marker gene expression for each cell subgroup in the sequenced samples. C Ro/e plots for each NC, DC, mild, and severe cell subgroup. D Fibrosis, inflammation, severe, cytokine scores in NC, CC, mild, and severe groups. E AUCell score plots and box plots of HSCs from ST samples of BA and NC. F AUCell score plots and box plots of BECs from ST samples of BA and NC. G Fibrosis, inflammation, severe, and cytokine scores for different cell subgroups in the sequenced samples. BA, biliary atresia; UMAP, uniform manifold approximation and projection; NK, natural killer; mono-macro, monocyte-macrophage; DCs, dendritic cells; HSCs, hepatic stellate cells; BECs, biliary epithelial cells; NC, normal control; DC, disease control; ST, spatial transcriptomics

This study employed two distinct grouping models. The first model defined the BA group as group T and the control group as group C. The second model categorized two NC samples as the NC group, two CC samples as the CC group, F1 and F2 samples as the mild group, and F3 and F4 as the severe group. Ro/e results showed that both datasets enriched T cells, DCs, neutrophils, proliferating cells, and erythroblasts in the T group. B, plasma, and endothelial cells were all inhibited (Figure S3C-D). In the severe group, mono-macro cells, DCs, and neutrophils were enriched (Fig. 1C).

The results of the fibrosis score were consistent with the clinicopathological observations. The fibrosis score for group T was higher than that for group C in both datasets (P < 0.0001) (Figure S4A-B). Fibrosis score was significantly higher in the severe group than in the mild group, and in the mild group than in the DC and NC groups (Adjusted P < 0.0001) (Fig. 1D). Additionally, the severe score defined in this study indicated that the T group had higher scores than the C group across both datasets (P < 0.0001) (Figure S4A-B). However, since the characteristic genes of the severe fibrosis group were used as the reference dataset, the score was particularly prominent in the severe group and less effective in identifying the mild group (Fig. 1D). The results of inflammatory scores and cytokine scores in groups C and T were inconsistent in the two data sets (Figure S4A-B), which may have been influenced by using preoperative antibiotics. Regarding the use of preoperative antibiotics, for children with BA highly suspected by preoperative ultrasound, our center administers empirical antibiotics for 48–72 h prior to surgery to prevent postoperative cholangitis. For other conditions such as choledochal cysts, hepatic hemangiomas, and hepatoblastomas, prophylactic antibiotic treatment is given for 24 h before surgery. However, inflammation scores and cytokine scores were significantly higher in the severe group than in the mild group (Adjusted P < 0.0001) (Fig. 1D).

Figure S4C shows the pathological sections of the BA sample and the NC sample used for ST sequencing. The fibrotic and non-fibrotic regions were identified by categorizing each cluster of ST images in accordance with the pathological definition of fibrosis (Figure S4D). The results of ST indicated that hepatocytes were widely enriched in both BA and NC tissues; however, the degree of enrichment in the BA group was lower than that in the NC group (P < 0.001), which correlated closely with fibrosis in BA liver tissue (Figure S4E). Additionally, HSCs and BECs were more abundant in BA compared to NC (P < 0.001) and were predominantly enriched in the fibrotic areas of BA tissues (Fig. 1E-F). The enrichment levels of the four scores in BA tissues were significantly higher than those in NC tissues (P < 0.001). Additionally, the fibrosis score was a strong indicator of fibrotic areas, followed by the severe score. In contrast, the cytokine and inflammation scores had limited effectiveness in identifying fibrotic areas (Figure S4F).

Among all immune cells, neutrophils and mono-macro cells from both datasets had significantly higher fibrosis, severe, inflammation, and cytokine scores (Fig. 1G, Figure S4G).

The inflammatory factor CCL2 is related to BA fibrosis from the multi-omics perspective

To investigate the causal relationship between 91 inflammation-related plasma proteins and BA, a two-sample MR analysis was conducted. Three inflammation-related plasma proteins were identified as risk factors for BA: MMP-1 (odds ratio (OR) = 1.42, 95% confidence interval (CI): 1.13 to 1.77, P = 0.002, P _FDR = 0.113); CD40 (OR = 1.47, 95% CI: 1.15 to 1.89, P = 0.002, P _FDR = 0.113); MCP-1 (CCL2) (OR = 1.70, 95% CI: 1.19 to 2.43, P = 0.004, P _FDR = 0.117) (Fig. 2A).

Fig. 2.

The inflammatory factor CCL2 is related to BA fibrosis from the multi-omics perspective. A Forest plot of positive results from MR analysis of 91 inflammation-related plasma proteins and BA. B Bubble plot of CCL2 expression in NC, DC, mild, and severe groups of sequenced samples. C Bubble plot of CCL2 expression in C and T groups of the GSE176189 dataset. D Bubble plot of CCL2 expression in different cell subgroups of sequenced samples. E Bubble plot of CCL2 expression in various cell subsets of the GSE176189 dataset. F Feature plots based on BayesSpace enhancement and box plots of CCL2 from ST samples of BA and NC. G Visualization of ST pseudo-time trajectories in BA. H The expression curve of CCL2 in the ST of BA along with the pseudo-time trajectories. MR, Mendelian randomization; SNP, single nucleotide; OR, odds ratio; CI, confidence interval; FDR, false discovery rate; BA, biliary atresia. NK, natural killer; mono-macro, monocyte-macrophage; DCs, dendritic cells; HSCs, hepatic stellate cells; BECs, biliary epithelial cells; SMCs, smooth muscle cells; NC, normal control; DC, disease control; ST, spatial transcriptomics

The detailed data for the IVs are presented in Table S6. In general, a causal association is considered robust if statistical significance is observed in two additional MR tests (P < 0.05). Notably, the associations of MMP-1, CD40, and CCL2 were considered robust (Table S7). Heterogeneity was assessed using the Cochran Q-test. The P-values for MMP-1, CD40, and CCL2 were greater than 0.05, indicating the absence of significant heterogeneity (Table S7). Sensitivity analyses were conducted to detect horizontal pleiotropy and identify potential outliers. All associations were examined for potential horizontal pleiotropy using the MR-Egger intercept and the MR-PRESSO global test. All P-values in the results were greater than 0.05, suggesting the absence of significant horizontal pleiotropy (Table S7). To test Hypothesis 3 (IVs are unrelated to the outcome and influence the outcome solely through exposure), MR-Steiger analysis was performed. The results showed that in the reverse causality analysis, with BA as the exposure factor, the P-values for MMP-1, CD40, and CCL2 in MR-Steiger were all below 0.05 (Table S7). Although sensitivity analyses were conducted in this study to assess SNP estimates, we also examined whether all SNPs associated with MMP-1, CD40, and CCL2 were independent of BA risk factors (e.g., viral infection and maternal exposure to certain chemicals) to satisfy Assumption 2 (IVs must be independent of other confounders). The results showed that none of the SNPs were associated with any confounding factors (Table S8).

Our previous two-sample MR analysis indicated that three inflammatory cytokines—Eotaxin (CCL11), G-CSF (CSF3), and CCL2 have a causal relationship with BA and are its risk factors [27]. In this study, CCL2 was reconfirmed as a risk factor for BA.

Furthermore, the expression of CCL2 was higher in the T group compared to the C group in the GSE176189 dataset, and CCL2 expression increased with the progression of fibrosis (Fig. 2B-C). In addition, CCL2 was highly expressed in HSCs, BECs, and the monocyte-macrophage system (MPS) in both datasets (Fig. 2D-E). In the BA tissue, the expression level of CCL2 was significantly higher than that in the NC tissue (P < 0.001); notably, CCL2 was primarily expressed in the fibrotic regions (Fig. 2F). The results presented in Fig. 2G indicate that the developmental trajectory of cells in the fibrotic regions of the liver in BA progresses from the center towards the periphery. Furthermore, this study also found that the expression level of CCL2 in the central region is higher than that in the peripheral regions (Fig. 2H).

Changes in BA fibrosis-related subgroups in MPS

The results indicated that the MPS might have played a significant role in BA-related liver fibrosis and inflammatory responses. MPS were further classified, resulting in thirteen subgroups based on marker expression in eight sample datasets (Fig. 3A-B) and fourteen subgroups in twelve sample datasets (Figure S5A-B). The Ro/e plots, bar graphs of cell proportions, and density plots for different subgroups indicated that CXCL10 + intermediate monocytes were more enriched in the mild group but exhibited a suppressed state in the severe group (Fig. 3C, Figure S5C-D). Additionally, the proportions of TREM2 + scar-associated macrophages (SAMs) and OLR1 + M2 macrophages increased with the severity of fibrosis (Fig. 3C, Figure S5C-D). Within the fibrosis scores, inflammation scores, severity scores, and cytokine scores of both datasets, CXCL10 + intermediate monocytes exhibited the highest scores (Fig. 3D, Figure S5E). The KEGG pathway enrichment analysis for CXCL10 + intermediate monocytes in both datasets revealed significant enrichment in pathways closely related to inflammation and immune responses. These pathways included the NOD-like receptor signaling pathway, Toll-like receptor signaling pathway, chemokine signaling pathway, cytokine-cytokine receptor interaction, and viral protein interaction with cytokines and cytokine receptors (Figure S5F-G).

Fig. 3.

Changes in BA fibrosis-related subgroups in MPS. A UMAP for MPS subgroups in the sequenced samples. B Bubble plot of marker gene expression for each MPS subgroup in the sequenced sample. C Ro/e plots for each MPS subgroup of NC, DC, mild, and severe groups. D Fibrosis, inflammation, severe, and cytokine scores for different MPS subgroups in the sequenced samples. E The differentiation potential score UMAP of MPS subgroups, the lower the score, the greater the differentiation potential, the more likely it is the starting point of cell development. F Pseudo-time differentiation trajectories of all monocyte and macrophage subgroups in NC, DC, mild, and severe groups, with arrows pointing in the direction of differentiation. UMAP, uniform manifold approximation and projection; MPS, monocyte-macrophage system; Mono, monocyte; Macro, macrophage; Int, intermediate; KCs, Kupffer cells, SAMs, scar-associated macrophages; Pre DCs, Precursor dendritic cells; cDCs, conventional dendritic cells; NC, normal control; DC, disease control; diff, differentiation

The differentiation trajectory of mono-macro subgroups followed a path from monocytes to macrophages (Fig. 3E). The results indicated that as the severity of fibrosis increased, the number of monocytes differentiating into TREM2 + SAMs gradually increased (Fig. 3F). The above results suggested that CXCL10 + intermediate monocytes may play a significant role in the early stages of fibrosis progression, while TREM2 + SAMs may be more active in the later stages. Furthermore, OLR1 + M2 macrophages may represent a transitional state between CXCL10 + intermediate monocytes and TREM2 + SAMs.

Multi-angle validation from ST, bulk to clinical data

In the ST validation, CXCL10 + intermediate monocytes, OLR1 + M2 macrophages, and TREM2 + SAMs were primarily distributed in the fibrotic regions, and their quantities in BA tissues were greater than those in NC tissues (P < 0.001) (Figure S6A-C).

NK cells, T cells, B cells, plasma cells, neutrophils, and endothelial cells were classified into subgroups based on their primary markers (Figure S7A-H), and these were summarized in a UMAP plot to display all cell subgroups (Figure S6D). Validation using the GSE122340 dataset revealed that the enrichment of CXCL10 + intermediate monocytes, OLR1 + M2 macrophages, and TREM2 + SAMs in BA was significantly higher than the NC group (P < 0.0001) (Figure S6E).

Furthermore, an analysis combining clinical data found that CXCL10 + intermediate monocytes were positively correlated with DBIL (R = 0.86, P = 0.011) and TBIL (R = 0.79, P = 0.028) (Figure S6F-G), while TREM2 + SAMs and OLR1 + M2 macrophages also showed positive correlations with BA-related liver function indicators (Figure S6F).

Spatial localization of CCL2 + mono-macro cells and its clinical significance for liver fibrosis in BA

Subsequently, the expression of CCL2 in MPS was further explored, and found that it was primarily expressed in CXCL10 + intermediate monocytes, OLR1 + M2 macrophages, and TREM2 + SAMs (Fig. 4A). Additionally, the expression of CCL2 in these three cell types was significantly higher in the T group compared to the C group (Fig. 4B). The CCL2 expression in CXCL10 + intermediate monocytes showed a strong positive correlation with various BA-related liver function indicators, including alkaline phosphatase (ALP) (R = 0.73, P = 0.039), aspartate aminotransferase (AST) (R = 0.71, P = 0.05), alanine aminotransferase (ALT) (R = 0.85, P = 0.007), direct bilirubin.

Fig. 4.

Spatial localization of CCL2 + mono-macro cells and its clinical significance for liver fibrosis in BA. A Bubble plot of CCL2 expression in different MPS subgroups of sequenced samples. B Violin plot comparison of CCL2 expression in the C and T groups of different MPS subgroups. C UMAP of the three subgroups with high CCL2 expression. D UMAP of CCL2 + mono-macro cells and CCL2- mono-macro cells. E UMAP of CCL2 + mono-macro cells in all cell subgroups. F AUCell score plots and box plots of CCL2 + mono-macro cells from ST samples of BA and NC. G Ro/e plots for CCL2 + mono-macro cells, CCL2- mono-macro cells, and other MPS subgroups of NC, DC, mild, and severe groups. H Heat map of the correlation between the proportion of cells in CCL2 + mono-macro cells, CCL2- mono-macro cells, and other MPS subgroups and clinical information. I Scatter plot of the correlation between CCL2 + mono-macro cells and GGT. mono-macro, monocyte-macrophage; BA, biliary atresia; MPS, monocyte-macrophage system; UMAP, uniform manifold approximation and projection; NC, normal control; ST, spatial transcriptomics; GGT, gamma-glutamyl transpeptidase

(DBIL) (R = 0.90, P = 0.0021), gamma-glutamyl transferase (GGT) (R = 0.78, P = 0.022), indirect bilirubin (IBIL) (R = 0.76, P = 0.03), total bilirubin (TBIL) (R = 0.83, P = 0.011), and total bile acids (TBA) (R = 0.83, P = 0.011). This group exhibited the strongest association among all MPS subgroups and positively correlated with fibrosis levels (R = 0.75, P = 0.031) (Figure S8A-B).

The three subgroups with high CCL2 expression were further categorized into CCL2 + mono-macro cells and CCL2- mono-macro cells (Fig. 4C-D). The positioning of CCL2 + mono-macro cells among all cell populations is illustrated in Fig. 4E. Marker genes of CCL2 + mono-macro cells included CCL2, TREOLR1 + M2, CD9, SPP1 and SDS (Figure S9). The results of ST showed that CCL2 + mono-macro cells were significantly more enriched in BA tissues than NC tissues and were mainly distributed in fibrotic areas (Fig. 4F). The enrichment of cells in the fibrotic region was better than that of CXCL10 + intermediate monocytes, OLR1 + M2 macrophages, and TREM2 + SAMs (Figure S5A-C, Fig. 4F).

Ro/e plots showed that as the severity of fibrosis increased, the enrichment of CCL2 + mono-macro cells became more pronounced (Fig. 4G). Both CCL2 + mono-macro cells and CCL2- mono-macro cells were positively correlated with BA-related liver function indicators (such as ALT, AST, DBIL, TBA, TBIL) and the degree of liver fibrosis (Fig. 4H). Notably, CCL2 + mono-macro cells were positively correlated with GGT (R = 0.88, P = 0.0072), an important prognostic indicator for children with BA [31] suggesting that CCL2 + mono-macro cells may be closely related to the prognosis of BA (Fig. 4I).

Cell interaction between CCL2 + mono-macro cells and liver parenchymal cells

CCL2 + mono-macro cells, HSCs, hepatocytes, and BECs were selected for cell-to-cell interaction analysis. The results showed that ligand-receptor interactions involving CD45, ANNEXIN, ICAM, GRN, SEMA3, GAS, and MHC-1 were significantly enhanced in the BA group compared to the control group, and the aforementioned ligand-receptor interactions were also increased in severe group (Fig. 5A, Figure S10A). In addition, VTN, CHEMERIN, JAM, VCAM, CADM, VEGF, DESMOSOME, HGF, OCLN, TGFb, THBS, IL16, CALCR, which were significantly enhanced in BA, were also enhanced in mild group (Fig. 5A, Figure S10A). Enhanced interactions were observed among CCL2 + mono-macro cells, HSCs, hepatocytes, and BECs (Fig. 5B, Figure S10B). The cell interaction network showed that the interaction between cells was significantly enhanced in the BA group compared with the control group, and the interaction in the severe group was weaker than that in the mild group (Fig. 5C, Figure S10C).

Fig. 5.

Cell interaction between CCL2 + mono-macro cells and liver parenchymal cells. A The left panel shows the ratio plot, while the right panel presents the actual value comparison plot. When the ratio of the summed pathway probabilities between the BA group and the control group is < 0.95 with a P < 0.05 (rank-sum test), it indicates a significant increase in communication strength of that pathway in the control group (indicated by a red y-axis). Conversely, when the ratio of the summed pathway probabilities is > 1.05 with a P < 0.05, it indicates a significant increase in communication strength of that pathway in the BA group (indicated by a blue y-axis). A black y-axis indicates no significant difference in the pathway between the two groups. B Heatmap of differences in the number and strength of cell communication between BA and control. C Network plot of the number of cell interactions between BA and control. D-E Dot plot of the interaction between CCL2 + mono-macro cells and liver fibrosis-related cells in BA and control groups. BA, biliary atresia; mono-macro, monocyte-macrophage; HSCs, hepatic stellate cells; BECs, biliary epithelial cells

The BA group enhanced LGALS9-CD45 interaction from CCL2 + mono-macro cells to hepatocytes compared to the control group. Conversely, the ligand-receptor interactions from hepatocytes to CCL2 + mono-macro cells involving C3-C3AR1, and C3-(ITGAX + ITGB2) were also strengthened, with these interactions being more pronounced in the mild group than in the severe group (Figs. 5D-E and 6, Figure S10D-E).

Fig. 6.

Diagram of possible mechanisms of liver inflammation and fibrosis in BA. BA, biliary atresia. Created in https://BioRender.com

Compared to the control group, the ligand-receptor interaction of SPP1-(ITGA5 + ITGB1) between CCL2 + mono-macro cells and HSCs was enhanced in the BA group. Notably, SPP1-(ITGA5 + ITGB1) interaction was more pronounced in the mild group than in the severe group. Conversely, the ligand-receptor interaction of APP-CD74 from HSCs to CCL2 + mono-macro cells was also strengthened, with this interaction being more significant in the mild group compared to the severe group (Figs. 5D-E and 6, Figure S10D-E).

Compared to the control group, the ligand-receptor interaction of SPP1-(ITGAV + ITGB1) between CCL2 + mono-macro cells and BECs was also enhanced in the BA group, with this interaction being more pronounced in the mild group than in the severe group. Conversely, the ligand-receptor interaction of APP-CD74 from BECs to CCL2 + mono-macro cells was strengthened, and this interaction was similarly significant in both the severe and mild groups (Figs. 5D-E and 6, Figure S10D-E).

Discussion

BA is characterized by progressive liver fibrosis and inflammatory destruction, and its liver fibrosis progresses rapidly [32]. In this study, the association between liver fibrosis and inflammation in BA was initially explored using scRNA-seq on four samples with varying degrees of liver fibrosis and four control samples. The fibrosis score used in this study was consistent with the pathological and ST results, and the two inflammation-related scores (inflammation score and cytokine score) displayed a similar trend. Although the severe score was less effective than the fibrosis score, it still accurately reflected the characteristics of severe fibrosis. Then, we found that more severe fibrosis in BA children accompanied a stronger inflammatory response, consistent with findings from a recent study [33]. At present, it is still controversial whether the prophylactic use of antibiotics after Kasai procedure can reduce the risk of cholangitis after BA [34, 35]. Since inflammation is closely related to fibrosis in BA, the inflammatory response will drive the formation of fibrosis, and the stimulation of fibrosis will further release inflammatory factors [36]. Anti-inflammatory therapy not only reduces inflammation but also may play an anti-fibrosis role by inhibiting the fibrosis process mediated by inflammation, which provides certain evidence for the use of antibiotics after Kasai procedure.

Inflammatory factor CCL2 may play an important role in developing BA with progressive liver fibrosis [37]. Ramm et al. [38] found that bile acid-induced upregulation of liver-derived CCL2 expression leads to HSC recruitment and is a critical early event in developing liver fibrosis. Xiao et al. [6] found that BA organoids express higher levels of CCL2 upon TNFSF12 stimulation and promote monocyte chemotaxis through the CCL2-CCR2 axis. Blocking TNFRSF12A inhibits liver injury, inflammation, and bile duct features. We performed a multi-omics analysis by integrating scRNA-seq, ST, and MR data. We found a causal relationship between the inflammatory factor CCL2 and BA. They have chemotactic properties and are mainly expressed in MPS in immune cells. This finding was further validated in public databases. In addition, we found that the expression level of CCL2 increased with the degree of fibrosis, and CCL2 was mainly expressed in the central region of liver fibrosis in BA. Based on these results, we hypothesize that CCL2 in MPS may play a role in regulating liver fibrosis progression in BA.

Ye et al. [39] identified immune subsets that are present in the “fibrotic niche” (the region of scar), including “intermediate” CD14 + + CD16 + monocytes and scar-associated macrophages. This is consistent with the results of the present study. In addition, we found that CXCL10 + intermediate monocytes gradually transformed into TREM2 + SAMs with the progression of liver fibrosis in BA; the former may play a role in the early stage of fibrosis, and the latter may play a role in the late stage of fibrosis. We validated this result from Bulk, ST, and clinical data.

To further explore the regulatory role of inflammatory factor CCL2 in MPS on liver fibrosis in BA, we defined CCL2 + mono-macro cells and confirmed that this group of cells was most closely related to liver fibrosis in BA from multiple views from ST and clinical data. Among them, the proportion of CCL2 + mono-macro cells was significantly positively correlated with GGT. GGT is an indicator of biliary obstruction; the diagnosis of BA is considered when GGT > 100 U/L, and BA is highly suspected when GGT > 300 U/L [40, 41]. In addition, other studies have shown that preoperative GGT levels have a predictive effect on the prognosis of BA children, and high preoperative GGT levels indicate a better prognosis of BA children undergoing Kasai procedure [31, 42]. Therefore, it is important to further investigate the specific role of CCL2 + mono-macro cells in BA.

Galectin-9 (LGALS9), a β-galactosidase specific animal lectin, is thought to be involved in the whole process from chronic inflammation to fibrosis. A higher serum galectin-9 level predicts liver fibrosis progression, but its role in BA has not been previously reported [43]. Our results showed that CCL2 + mono-macro cells of BA may bind to CD45 on the surface of hepatocytes through LGALS9. This interaction may lead to dysfunction of hepatocytes and aggravation of liver inflammation. Hepatocytes release the complement component C3, which binds to its receptor C3AR1 and integrins (ITGAX and ITGB2) to release cytokines that further exacerbate the local inflammatory response. It can attract more immune cells to the inflammatory area and enhance their activity. Liver inflammation is a major risk factor for liver fibrosis. The inflammatory response drives the activation of HSCs, which, in turn, regulate immune mechanisms through the secretion of chemokines and cytokines or by transdifferentiating into matrix-producing myofibroblasts [44].

Sh et al. [45] showed that SPP1 is expressed in activated HSCs and suggested that upregulation of SPP1 may be a central pathway for HSCs activation. Whitingtondeng et al. [46] also showed that the expression of SPP1 was upregulated in BA interlobular bile duct epithelial cells and was associated with biliary hyperplasia and portal vein fibrosis. This study found that CCL2 + mono-macro cells in the BA group may act on HSCs and BECs through SPP1-(ITGAV + ITGB1). HSCs can be activated and secrete various cytokines to promote local inflammatory response. In addition, large amounts of extracellular matrix components are produced, which help to reconstruct the liver architecture but can lead to liver fibrosis if over-deposited. BECs are characterized by enhanced proliferation and secretion of cells (including bile acid and various cytokines). Our previous study showed that BECs in BA can promote cell proliferation and EMT through EGF/EGFR-ERK1/2 signaling when stimulated by EGF [47]. Therefore, the enhanced interaction between CCL2 + mono-macro cells and BECs may also promote the transformation of BECs into myofibroblasts through EMT, thereby participating in the process of liver fibrosis in BA.

Beta-amyloid (Aβ) has been observed around the intrahepatic bile ducts in BA [48]. A previous study also found that APP (the precursor of Aβ) is a potential fluid biomarker for BA and liver fibrosis [49]. Our results showed that HSCs and BECs interact with CCL2 + mono-macro cells with enhanced APP-CD74. Both APP and CD74 are related to cell migration [50, 51]. This may promote the migration of immune cells (CCL2 + mono-macro cells) to HSCs and BECs, accumulate in local areas, continuously activate the immune response, and aggravate liver inflammation and fibrosis.

These interactions were more pronounced in the mild group than in the severe group, possibly because cell function in the severe group was inhibited or altered due to pathological damage or chronic stimulation, resulting in a diminished interaction. We believe that diminishing these enhanced interactions (e.g., using inhibitors of the above targets) or reducing the proportion of CCL2+ monocyte-macrophage cells in the liver of BA patients may help reverse BA-related liver fibrosis, providing a potential therapeutic approach for anti-fibrotic therapy in BA.

This study has several strengths. (1) Previous scRNA-seq of BA children has not explained the fibrosis grade, so it is impossible to explore the liver fibrosis of BA from this perspective. In this study, scRNA-seq was performed on BA liver tissues with liver fibrosis grades F1 to F4. (2) This study performed scRNA-seq analysis on the sequencing data from our center, supported by ST and clinical data, which added clinical significance to our findings. (3) This study integrated results from MR analysis, providing a multi-faceted perspective on the study of BA. (4) This study validated some of our results using publicly available scRNA-seq and bulk RNA-seq data to enhance the credibility of our findings. However, this study also has certain limitations. Firstly, due to the low incidence of BA, obtaining BA samples based on fibrosis staging presents challenges [52, 53]. There was only one sample for each fibrosis grade, so conclusions about fibrosis progression should be interpreted with caution. To enhance the strength of the conclusions of this study, we defined fibrosis grade F1 and F2 as mild fibrosis and fibrosis grade F3 and F4 as severe fibrosis, avoiding having only one sample in each group. Secondly, the small sample size of this study may affect the robustness of the conclusions, so we validated the results using scRNA-seq and bulk RNA-seq data from public databases. In the future, we will continue to increase the sample size in each group to obtain more robust conclusions. Thirdly, as liver tissue samples from BA patients at our center can only be obtained once during the Kasai procedure, it is not feasible to conduct longitudinal trajectory analyses. Therefore, the findings of this study are based solely on cross-sectional differences, which represent one of the limitations of our work. Finally, the results of this study are only exploratory findings, and follow-up experiments are still needed to provide more reliable conclusions. In vitro, the role of CCL2 + monocyte-macrophage cells in the pathogenesis of liver fibrosis in BA can be validated using macrophages derived from human peripheral blood mononuclear cells in an in vitro system, or by employing a three-dimensional liver organoid model consisting of THLE-2 epithelial cells co-cultured with LX-2 cells. In vivo, the role of CCL2 + monocyte-macrophage cells in BA liver fibrosis can be further investigated using two common animal models of BA, the bile duct ligation rat model and the rhesus rotavirus mouse model.

Conclusions

In summary, our study integrated scRNA-seq, ST, clinical data, and MR analysis to explore the disease of BA from multiple perspectives. Our findings indicate a close relationship between inflammation and fibrosis in BA. Therefore, anti-inflammatory therapy not only alleviate inflammation but also may exert anti-fibrotic effects by inhibiting inflammation-mediated fibrogenic processes, providing a strong rationale for the use of antibiotics following Kasai procedure. Additionally, we discovered a causal relationship between the inflammatory factor CCL2 and BA, CCL2 + mono-macro cells are most closely associated with BA fibrosis. Finally, we summarize the intercellular interactions that may contribute to the progression of mild fibrosis in BA and propose that blocking these interactions or reducing the proportion of CCL2 + mono-macro cells may help reverse liver fibrosis in BA, providing potential targets for anti-liver fibrosis therapy in BA.

Glossary

Abbreviation

Full Spelling

BA

biliary atresia

scRNA-seq

single-cell RNA sequencing

ST

spatial transcriptomics

MR

Mendelian randomization

mono-macro

monocyte-macrophage

GGT

gamma-glutamyl transferase

HSCs

hepatic stellate cells

BECs

bile duct epithelial cells

IVs

instrumental variables

CC

choledochal cysts

DC

disease control

NC

normal control

UMIs

unique molecular identifiers

cDNA

complementary DNA

GEO

Gene Expression Omnibus

GWAS

genome-wide association study

pQTL

protein quantitative trait locus

UMAP

unified manifold approximation projection

DEGs

differentially expressed genes

LFMGs

liver fibrosis-specific matrisome genes

NK

natural killer

DCs

dendritic cells

SMCs

smooth muscle cells

OR

odds ratio

CI

confidence interval

FDR

false discovery rate

MPS

monocyte-macrophage system

SAMs

scar-associated macrophages

ALP

alkaline phosphatase

AST

aspartate aminotransferase

ALT

alanine aminotransferase

DBIL

direct bilirubin

IBIL

indirect bilirubin

TBIL

total bilirubin

TBA

total bile acids

RNA-seq

RNA sequencing

Authors’ contributions

Xin Li: Conceptualization, Resources, Writing - Original Draft, Supervision, Project administration, Funding acquisition. Shaowen Liu: Conceptualization, Methodology, Software, Validation, Formal analysis, Writing - Original Draft, Writing - Review & Editing, Visualization. Tengfei Li: Methodology, Software, Data Curation. Qianhui Yang: Writing - Review & Editing, Visualization. Yuqiang Chen: Investigation, Resources. Yu Meng: Writing - Review & Editing, Visualization. Zhiru Wang: Writing - Review & Editing. Yilin Zhao: Investigation, Resources. Shujian Zhang: Project administration. Liang Ge: Project administration. Alimujiang Abudureyimu: Project administration, Funding acquisition. Jianghua Zhan: Writing - Review & Editing, Supervision, Project administration, Funding acquisition.

Funding

This study was supported by grants from the Tianshan Talent Training Program (Grant No. 2023TSYCJC0052) and Xinjiang Uygur Autonomous Region Science Foundation Projects (2024D01A28).

Data availability

GSE122340, GSE176189, S-BSST182, GCST90274758 and GCST90274848 are available in a public, open-access repository. Data sources are included in the Methods section of the article. Further information is available from the corresponding author. The datasets generated and analysed during the current study are available in the OMIX database (https://ngdc.cncb.ac.cn/omix/) [OMIX008911].

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the principles outlined in the Declaration of Helsinki. Ethical approval was obtained from the Institutional Review Board (IRB) of Tianjin Children’s Hospital (Approval number: 2022-SYYJCYJ-008). Written informed consent was obtained from the guardians of all participants before enrollment in the study. We provided them with information about the purpose, procedures, and potential risks of the study and gave them the opportunity to ask questions. Participants were assured of the right to withdraw from the study at any time without consequence.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.