P-STAT3-dependent SPP1 transcription in cholangiocytes drives hepatic stellate cells activation in congenital hepatic fibrosis

Abstract

Congenital hepatic fibrosis (CHF) is a rare inherited liver disorder caused by mutations in the polycystic kidney and hepatic disease-1 gene (PKHD1), leading to progressive hepatic fibrosis. In our study, single-cell RNA sequencing of a CHF patient carrying compound heterozygous PKHD1 mutations revealed markedly intercellular communication between cholangiocytes and hepatic stellate cells (HSCs) via the secreted phosphoprotein 1 (SPP1)–integrin α8β1 axis. The CRISPR-Cas9-generated Pkhd1 knockout mouse model (Pkhd1^{del3-4/del3-4}) was generated and recapitulated bile duct hyperplasia and fibrosis. We validated that PKHD1 deficiency led to progressive upregulation of SPP1 in cholangiocytes. Co-culture experiments demonstrated that PKHD1-deficient cholangiocytes secrete OPN, activating PI3K/Akt signaling in HSCs and promoting their activation. Mechanistically, phosphorylated STAT3 (p-STAT3) was identified as a direct transcriptional activator of SPP1. Importantly, cholangiocyte-specific knockdown of Spp1 via AAV8 significantly attenuated hepatic fibrosis in vivo. Collectively, our findings demonstrate that reduced PKHD1 expression in cholangiocytes promotes hepatic fibrosis by activating the p-STAT3–SPP1 pathway, and targeting this axis may offer a potential therapeutic strategy.

Cholangiocytes directly activate hepatic stellate cells in congenital hepatic fibrosis through SPP1- dependent signaling, as revealed by single-cell transcriptomics and in vivo functional validation.

Introduction

Congenital hepatic fibrosis (CHF) is a rare inherited disorder characterized by polycystic liver, polycystic kidney, and progressive hepatic fibrosis, with an estimated incidence of 1 in 10,000 to 20,000 individuals^1,2. Hepatic fibrosis, the central pathological process leading to end-stage liver disease, poses a serious threat to patient survival owing to its complications, including ascites and hepatic encephalopathy, contributing to significant mortality and posing a major risk to prognosis^3,4. To date, no effective pharmacological therapies have been developed for this rare disease⁵.

Mutations in the polycystic kidney and hepatic disease-1 (PKHD1) gene are the most common genetic cause of CHF⁶. PKHD1 undergoes extensive alternative splicing, generating transcripts of variable lengths. Its longest isoform encodes fibrocystin (FPC), a 4074-amino-acid protein featuring a single transmembrane segment⁷. FPC is primarily expressed in intrahepatic cholangiocytes and renal tubular epithelial cells⁸. Although its precise biological function remains incompletely understood, loss of FPC in cholangiocytes leads to ductal plate malformation and is believed to play a critical role in maintaining the ciliary structure and epithelial cell polarity⁹. FPC deficiency is also associated with progressive hepatic fibrosis, and complications of portal hypertension represent a major cause of mortality in advanced CHF¹⁰. In patients with CHF, liver-related complications are the third leading cause of mortality¹¹.

Among the cellular events underlying hepatic fibrogenesis, aberrant activation of hepatic stellate cells (HSCs) is a key driver of extracellular matrix deposition^12,13. Previous studies have shown that the loss of FPC in cholangiocytes induces low-grade periductal inflammation, recruits macrophages via chemokine secretion, and promotes HSCs activation through M2 macrophage polarization, thereby contributing to fibrotic progression¹⁴. These findings underscore the role of macrophages in the pathogenesis of CHF. However, the hallmark lesions of CHF are FPC-deficient cholangiocytes, which likely serve as initiating cells. Evidence from other cholangiocyte-origin liver fibrotic diseases, such as cholestatic liver fibrosis, has shown that cholangiocyte-derived extracellular vesicles can directly promote HSCs differentiation and activation, supporting the notion that cholangiocytes may also contribute to HSCs activation through multiple mechanisms in CHF¹⁵. Therefore, the mechanisms by which FPC-deficient cholangiocytes in CHF secrete paracrine mediators that directly trigger HSCs activation and what those mediators might be remain to be elucidated.

To figure out the potential mechanism, we performed single-cell RNA sequencing (scRNA-seq) of liver tissue from one patient with CHF and two control individuals, followed by ligand–receptor interactome analysis to systematically map the communication network between cholangiocytes and HSCs. In a CRISPR-Cas9-generated Pkhd1 knockout mouse model, in which deletion of exons 3–4 results in loss of FPC expression, Pkhd1^{del3-4/del3-4} mice exhibit a spectrum of phenotypes that faithfully mirror the key pathological features observed in human CHF¹⁶. This model provides a platform to both validate the findings from scRNA-seq and investigate the mechanisms underlying HSCs activation.

Methods

Human liver specimens

Liver biopsies were obtained from one patient with CHF and two control individuals at the First Affiliated Hospital with Nanjing Medical University (Supplementary Table 1). The study was approved by the Institutional Ethics Committee of the First Affiliated Hospital with Nanjing Medical University (2024-SRFA-287), and written informed consent was obtained from all participants prior to surgery. All ethical regulations relevant to human research participants were followed.

Animal model

All mice used in this study were generated on a C57BL/6J genetic background by GemPharmatech Co., Ltd¹⁷. Animal experiments were approved by the Animal Ethics Committee of Nanjing Medical University (IACUC-2506036). We have complied with all relevant ethical regulations for animal use. Pkhd1 knockout mice were generated using a CRISPR-Cas9-mediated genome editing strategy targeting exons 3–4 of the Pkhd1 gene. Guide RNAs (gRNAs) (5′-AAGGCAAAGCCACTTCCAGG-3′ and 5′-TTTGGGTAGTCACATTGGTG-3′) were designed to specifically flank exons 3–4, and the Cas9/gRNA system was constructed and validated in vitro prior to embryo manipulation. Fertilized zygotes were obtained from superovulated C57BL/6J female mice mated with male mice of the same genetic background. The Cas9 mRNA and gRNAs were microinjected into the pronuclei of fertilized embryos using a micromanipulation system. Successfully injected embryos were subsequently transferred into the oviducts of pseudopregnant recipient females.

Genomic DNA from founder (F0) mice was extracted from tail biopsies and subjected to PCR amplification spanning the targeted region, followed by Sanger sequencing to confirm deletion of exons 3–4. Positive F0 mice carrying the desired deletion were crossed with wild-type C57BL/6J mice to generate heterozygous F1 offspring. F1 heterozygous mice (Pkhd1^del3-4/+) were intercrossed to obtain homozygous Pkhd1^{del3-4/del3-4} mice for subsequent experiments. The genotype was confirmed by polymerase chain reaction (PCR) analysis of tail DNA (Supplementary Table 2). All Mice were housed under specific pathogen-free (SPF) conditions.

Mice aged 1–12 months were used in this study. The number of mice used in each experimental group varied depending on the specific assay and ranged from 3 to 6 mice per group. No sex-dependent effects were observed in our study. Therefore, both male and female mice were included, with an approximately equal distribution between the two sexes in each group. For anesthesia, mice were administered pentobarbital sodium (100 mg/kg, intraperitoneal injection). For euthanasia, mice were euthanized by carbon dioxide inhalation followed by cervical dislocation.

AAV vector construction and in vivo administration

To achieve cholangiocyte-specific knockdown of Spp1 in vivo, a miR30-based shRNA targeting mouse Spp1 was designed, and a non-targeting sequence was used as a negative control. The synthesized miR30-shRNA (Spp1) cassette was cloned into an adeno-associated virus (AAV) expression vector under the control of the cholangiocyte-specific HNF1β promoter, generating the recombinant construct pAAV-HNF1β-ZsGreen1-miR30-shRNA (Spp1)-WPRE. Recombinant AAV particles were packaged and purified by a commercial vendor (OBiO Technology, Shanghai, China). For in vivo delivery, mice of 3 months were randomly injected via the tail vein with 100 μL of AAV8-shRNA virus at a titer of 5 × 10¹² viral genomes (vg)/mL, corresponding to a total dose of 5 × 10¹¹ vg per mouse. Other control mice received an equivalent dose of AAV carrying a non-targeting shRNA (Supplementary Table 3).

10 × single-cell RNA sequencing library preparation and sequencing

Fresh human liver tissue was collected through liver puncture, placed in a special preservation solution, and immediately sent to the Laboratory of Singleron Biotechnologies in Nanjing, China. The specimens were washed three times with Hanks balanced salt solution and then minced to obtain 1–2 mm pieces. These tissue pieces were then digested using 2 mL of the GEXSCOPETM tissue dissociation solution at 37 °C for 15 min. After digestion, the samples were filtered through 40-µm sterile strainers and then centrifuged at 320 × g for 5 min. The supernatant was discarded, and the pellet was treated with 2 mL of the GEXSCOPETM red blood cell lysis buffer (Singleron Biotechnologies) at 25 °C for 10 min to remove the red blood cells. Afterward, the quality of the samples was evaluated using a computer. Subsequently, single-cell suspensions were loaded onto microfluidic devices, and scRNA-seq libraries were constructed by following the GEXSCOPETM protocol using the GEXSCOPETM single-cell RNA library kit. Each prepared library was diluted to 4 nM, and all libraries were pooled for sequencing using a Illumina NovaSeq 6000 device (Illumina, San Diego, CA, USA) and 150 bp-long paired-end reads.

Data quality control and clustering

Raw sequencing data were processed using CeleScope (v1.9.0). Low-quality reads and adapters were removed using Cutadapt (v1.17), and reads were aligned to the reference genome using STAR (v2.6.1a). Gene expression matrices were generated using featureCounts (v2.0.1). Downstream analysis was performed using Scanpy (v1.8.2). Cells with fewer than 200 detected genes, excessive UMI counts (top 2%), or mitochondrial gene content >20% were excluded. Genes expressed in fewer than 5 cells were removed. After filtering, 24,224 cells were retained. Data were normalized and log-transformed, and the top 2000 highly variable genes were selected. PCA was performed, and clustering was conducted using the Louvain algorithm based on the top 20 principal components. UMAP was used for visualization.

Cell type annotation, differential expression, and functional analysis

Cell types were annotated based on canonical marker genes, the SynEcoSys database, and published literature. Clusters expressing markers from multiple cell types were considered potential doublets and were manually excluded. Differentially expressed genes (DEGs) were identified using Seurat (v3.1.2) with the Wilcoxon rank-sum test. Genes expressed in >10% of cells with an average log2 fold change >0.25 were considered significant. Functional enrichment analysis, including Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, was performed using the clusterProfiler package (v3.16.1). Adjusted p-values < 0.05 were considered statistically significant.

Cell-cell communication analysis

Cell-cell communication analysis was performed using CellChat. Ligand–receptor interactions were inferred based on normalized gene expression and annotated cell types. Communication probabilities were calculated based on ligand–receptor co-expression, and statistical significance was assessed using permutation testing. Significant interactions (p < 0.05) were retained, and interaction strength and network topology were analyzed and visualized using standard CellChat workflows.

Cell culture and transfection

Human intrahepatic biliary epithelial cells (HIBEpiC), mouse hepatic stellate cells (mHSCs), and the human hepatic stellate cell line LX2 were obtained from Procell Life Science & Technology Co., Ltd. (Wuhan, China). The Human intrahepatic biliary epithelial cells (HIBEpiC) were cultured at 37°C in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS). The hepatic stellate cell line LX2 was maintained in high-glucose DMEM containing 10% FBS. All the siRNAs used in this study were purchased from GenePharma (Nanjing, China). Cells at 70–90% confluency were transfected with siRNA or si-NC (negative control siRNA) using Lipofectamine 2000. Briefly, siRNA-Lipofectamine complexes were prepared in serum-free medium and added to the cells. Protein and mRNA extraction was performed 48–72 h post-transfection (Supplementary Table 4).

Isolation of primary cells from mice

Primary mouse intrahepatic biliary epithelial cells (MIB) were isolated from 3-month-old Pkhd1^{del3-4/del3-4} and Pkhd1^+/+ mice¹⁸. Briefly, after PBS perfusion via the inferior vena cava, the liver was excised, minced, and digested with collagenase D, collagenase II, and DNase I at 37 °C. Bile duct fragments were enriched by low-speed centrifugation and sedimentation. For single-cell dissociation, the ducts were incubated with TrypLE and DNase I, gently pipetted, and monitored microscopically. Digestion was stopped when most cells were dissociated. The suspension was then filtered, washed, and centrifuged. The cells were resuspended in washing medium, counted, and used for culture or downstream assays. Primary mouse hepatocytes were isolated using a two-step collagenase perfusion method. Briefly, the liver was first perfused in situ with 37 °C calcium- and magnesium-free Hanks’ balanced salt solution at a flow rate of 5 mL/min until the tissue became pale and the effluent was clear, followed by perfusion with prewarmed type IV collagenase for enzymatic digestion until the liver was swollen and softened. The liver was then excised and filtrated through a 70 μm cell strainer to remove debris. The cell suspension was subjected to low-speed centrifugation at 4 °C for enrichment, repeated three times. The final pellet was resuspended in complete DMEM, and cell viability was assessed by trypan blue exclusion prior to subsequent experiments.

Measurement of hepatic hydroxyproline content in mice

Mouse liver tissues (30–100 mg) were hydrolyzed in hydrolysis solution at 95 °C for 20 min. After cooling, pH was adjusted to 6.0–6.8, followed by volume adjustment, decolorization with activated charcoal, and centrifugation to obtain the supernatant. Hydroxyproline content was measured using a commercial kit (Nanjing Jiancheng Bioengineering Institute, China). Absorbance was read at 550 nm, and concentrations were calculated using a standard curve.

DNA pull-down

Biotin-labeled double-stranded DNA probes were synthesized and conjugated to streptavidin beads. Nuclear extracts were prepared from HIBEpiC and incubated with the probe–bead complexes in binding buffer supplemented with protease inhibitors to preserve protein activity. After incubation, the complexes were washed extensively to remove nonspecifically bound proteins, followed by elution of specifically bound proteins. The eluates were separated by SDS-PAGE, excised, digested, and analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) to identify proteins bound to the target DNA sequence, thus enabling the screening of potential transcription factors.

Dual-luciferase reporter assay

HIBEpiC cells were co-transfected with wild-type (WT) or mutant (MUT) luciferase reporter plasmids, along with either STAT3-overexpression plasmid (oe-STAT3) or negative control vector (oe-NC) plasmids for 48 h, followed by cell lysis (Supplementary Table 5). Luciferase activity was measured using the Dual-Luciferase Reporter Assay Kit (Vazyme, DL101). Firefly luciferase activity was normalized to the Renilla luciferase activity as an internal control. All plasmids used in this study were custom-constructed by Corues Biotechnology (Nanjing, China).

ChIP-qPCR

Chromatin immunoprecipitation (CHIP) followed by qPCR was performed using a ChIP assay kit (Wuhan GeneCreate Biological Engineering Company). Briefly, cross-linked chromatin was immunoprecipitated with the indicated antibody, and the enriched DNA fragments were purified and quantified by qPCR. Primers used for ChIP-qPCR are listed in Supplementary Table 6.

RT-qPCR

Total RNA was extracted from the cells or tissues and reverse-transcribed into cDNA. Real-time quantitative PCR was performed using the SYBR Green system with gene-specific primers. Amplification was conducted using a real-time PCR instrument, and Ct values were recorded. Relative gene expression was calculated using the 2^−ΔΔCt method, with GAPDH as the internal control. The PCR primers that were used are listed in Supplementary Table 7 and 8.

Western blot

Protein samples were extracted, quantified, and separated by SDS-PAGE, followed by transfer onto PVDF membranes using a wet transfer system. After transfer, membranes were incubated at room temperature and probed sequentially with the primary and secondary antibodies. Protein signals were detected using enhanced chemiluminescence (ECL). The antibodies used in this study are listed (Supplementary Table 9).

ELISA

OPN protein levels in cell supernatant and mouse plasma samples were measured using ELISA kits (BOSTER, EK0482 for human, EK4803 for mouse). Samples were added to antibody-coated plates, incubated with the detection reagents according to the assay protocol, and absorbance was measured to determine OPN concentrations based on standard curves.

Biochemical analysis

Serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were determined using liver function assay kits (Nanjing Jiancheng Bioengineering Institute, China). Serum samples were incubated with the corresponding reaction reagents, and enzyme activities were determined by measuring absorbance changes.

EdU assay

The EdU working solution was added to the cell culture at the specified concentration and incubated for the designated time. The cells were then fixed with 4% paraformaldehyde and permeabilized with 0.5% Triton X-100. A click reaction solution was used to label the EdU-incorporated proliferating cells. After DAPI nuclear staining and washing with PBS, cell proliferation was observed under a fluorescence microscope.

CCK-8 assay

Treated LX2 and HIBEpiC cells were seeded into 96-well plates, and 10 μL of CCK-8 reagent was added to each well. Cells were incubated for 1–4 h at 37 °C, and absorbance was measured at 450 nm using a microplate reader.

Immunohistochemistry

Paraffin-embedded tissue sections were deparaffinized and rehydrated, followed by antigen retrieval using citrate buffer solution. Endogenous peroxidase activity was blocked using 3% hydrogen peroxide. Sections were then blocked and incubated with primary antibodies overnight at 4 °C. The following day, HRP-conjugated secondary antibodies were applied at room temperature, followed by DAB chromogenic detection and hematoxylin counterstaining. After dehydration and mounting, the staining results were evaluated under a microscope to assess target protein expression and distribution.

Immunofluorescence (IF) staining

Cells were seeded in 24-well plates at a density of 4 × 10⁵ cells/well and cultured for 24 h to allow for adhesion. After incubation, the cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100, and blocked to prevent nonspecific binding. Primary antibodies were incubated overnight at 4 °C, followed by fluorescent secondary antibodies at room temperature in the dark. After nuclear staining with DAPI and washing with PBS, the samples were mounted with antifade medium and imaged by fluorescence microscopy. Quantification was performed using ImageJ software.

Image acquisition and quantification

IF and immunohistochemistry (IHC) images were acquired using a fluorescence microscope (Olympus, Japan) under identical exposure settings across all groups to ensure comparability. For each sample, at least three randomly selected, non-overlapping fields were captured, with systematic sampling to include both portal and central venous regions. All analyses were performed in a blinded manner to minimize bias. All quantitative analyses were performed using ImageJ software (NIH, USA).

For IHC, both the percentage of positive area and integrated optical density (IOD) were measured. For IF, fluorescence intensity was quantified and normalized to the number of DAPI-positive nuclei. For Sirius Red and Masson’s trichrome staining, fibrotic areas were quantified as the percentage of positively stained area relative to the total tissue area in each field of view. In addition, a standardized histological fibrosis scoring system was applied, with all sections independently evaluated by two experienced pathologists blinded to group allocation, and the average score used for statistical analysis.

Statistics and reproducibility

Quantitative data are presented as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism (version 8.0, GraphPad Software, San Diego, CA, USA). Normality of data distribution was assessed using the Shapiro–Wilk test. For comparisons between two groups, either the Student’s t test or the Mann–Whitney U test was applied, depending on data distribution. For comparisons among multiple groups, one-way ANOVA with Dunnett’s multiple comparisons test was used. For experiments involving two independent variables, two-way ANOVA with Tukey post hoc test was performed. Statistical significance was defined as p < 0.05. The specific statistical tests and sample sizes (n), defined as biologically independent samples or animals, are provided in the corresponding figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Clinical features of a CHF patient

A 34-year-old male was admitted to our hospital with an 11-month history of intermittent melena. The patient had previously undergone gastric variceal ligation due to gastrointestinal bleeding. Clinical biochemical tests on admission showed alanine aminotransferase (ALT) of 18 U/L, aspartate aminotransferase (AST) of 19.5 U/L, gamma-glutamyltransferase (GGT) of 152.7 U/L, alkaline phosphatase (ALP) of 68 U/L, and total bilirubin (TB) of 5.5 μmol/L. The total protein (TP) and albumin were 67.7 g/L and 41.2 g/L, respectively. Contrast-enhanced computed tomography (CECT) revealed cirrhosis with esophageal varices (Fig. S1a). Tests for viral hepatitis markers, antinuclear antibodies, and autoimmune liver disease antibodies were negative. Liver biopsy demonstrated peri-portal fibrosis and bile duct proliferation (Fig. S1b). After excluding common causes of cirrhosis, such as viral infection, drugs, alcohol, and cholestasis, the findings were attributed to CHF.

Given the potential risk of CHF, genetic whole-exome sequencing (WES) was performed to assist in diagnosis and guide clinical management. Two candidate variants in PKHD1 were identified, including a missense mutation (c.1559A>T, p.N520I) and a frameshift duplication (c.8746dupG) (Fig. S1c). Genetic analysis of this patient’s family revealed that the missense variant (c.1559A>T, p.N520I) was inherited from the father, whereas the frameshift duplication (c.8746dupG) was inherited from the mother, confirming a compound heterozygous state. The c.1559A>T substitution causes an amino acid change at position 520 from asparagine to isoleucine, while c.8746dupG results in a frameshift mutation causing premature termination and loss of function (LOF) of FPC. A pedigree analysis was also conducted to determine the genetic status of family members. His siblings inherited the maternal frameshift mutation (c.8746dupG), and his sons inherited both mutations, one from the paternal grandfather (c.1559A>T, p.N520I) and the other from the paternal grandmother (c.8746dupG) (Fig. S2a, b). Except for the patient, all family members were phenotypically normal. Accordingly, the patient was diagnosed with CHF caused by compound heterozygous PKHD1 mutations c.1559A>T and c.8746dupG.

Single-cell transcriptomic features of the CHF patient

scRNA-seq was performed on liver biopsies from one CHF patient and two controls. After quality control and filtering, a total of 24,224 cells were retained across all three samples. The clinical information of the CHF patient and the two controls was described in Supplementary Table 1. Notably, both control individuals harbored no mutations in the PKHD1 gene, serving as genetically appropriate controls. Based on the gene expression profiles, the cells were clustered into 24 groups (Fig. S3a). By matching marker genes to known hepatic cell types, 12 distinct cell types were annotated, including hepatocytes, HSCs, and cholangiocytes (Fig. 1a, b). Comparative analysis revealed a marked expansion of cholangiocytes, HSCs, and macrophages in the CHF group, with cholangiocytes comprising tenfold higher proportions than controls (Figs. 1c–e and S3b). PKHD1 gene was primarily expressed in cholangiocytes and significantly downregulated in the CHF patient (Fig. S3c, d). Upregulated genes in CHF cholangiocytes were associated with increased energy metabolism, enhanced protein synthesis, and mitochondrial function, whereas downregulated genes were related to embryonic development, signaling pathways, and lipid metabolism (Fig. S3e, f).

Fig. 1. Single-cell transcriptomic profiling of liver tissues from CHF and healthy control.

a Dot plot showing expression patterns of canonical marker genes used for cell type annotation across identified clusters. Dot size represents the proportion of expressing cells, and color intensity indicates average expression. b UMAP plot showing annotated cell types. c UMAP plot showing distribution of cells from CHF and HC groups. d Bar plot illustrating cell type composition in CHF and HC livers. e Proportion of cholangiocytes, HepSCs and MPs in CHF versus HC groups. f Global intercellular communication network inferred from ligand–receptor interactions. Nodes represent individual cell types, and edges indicate the number and strength of predicted interactions. g Dot plot summarizing the top 20 ligand–receptor pairs between cholangiocytes and HepSCs. Dot size indicates the number of interactions between specific cell pairs, and color represents interaction strength. h Violin plot showing the distribution of SPP1 expression across different cell types. i Comparing SPP1 expression levels between HC and CHF liver samples in different cell population. *P < 0.05, **P < 0.01, and ***P < 0.001.

CHF is characterized by initial extracellular matrix (ECM) accumulation around the bile ducts, suggesting a potential role of cholangiocytes in HSCs activation. Therefore, we performed CellChat analysis, which revealed strong interactions between cholangiocytes and HSCs in the CHF liver (Figs. 1f and S4a, b). We speculated that cholangiocytes activate HSCs through specific secreted proteins. Ligand–receptor interaction analysis showed that the top-ranked candidate ligand derived from cholangiocytes did not exhibit a significant difference between groups and was therefore not further investigated. In contrast, the second-ranked ligand, secreted phosphoprotein 1 (SPP1), was selected as a potential key mediator. SPP1 mRNA encodes osteopontin (OPN), a secreted phosphoglycoprotein (Fig. 1g). ScRNA-seq analysis showed that SPP1 expression was significantly upregulated and predominantly enriched in cholangiocytes from the CHF patient (Figs. 1h, i and S4c, d). These findings suggest that the SPP1-integrin α8β1 axis may play a critical role in mediating aberrant communication between cholangiocytes and HSCs.

Phenotypic characterization in Pkhd1^{del3-4/del3-4} mice

To validate the scRNA-seq findings and further explore the potential mechanism, we generated a murine model by deleting Pkhd1 exons 3–4 using the CRISPR-Cas9 system. This strategy leads to loss of FPC, the key pathogenic driver of CHF resulting from PKHD1 mutations (Fig. S5)¹⁹. Pkhd1^{del3-4/del3-4} mice have been widely used to recapitulate the phenotypic features of CHF¹⁷. Gross examination revealed multiple distinct hepatic cysts containing clear colorless fluid with a smooth surface (Fig. 2a). Serum biochemical analyses in Pkhd1^{del3-4/del3-4} and Pkhd1^+/+ mice at 3, 6, 9, and 12 months revealed a progressive increase in ALT, AST, urea, and creatinine levels, with significant elevations observed at 12 months (Fig. 2b). Hepatic H&E staining revealed multiple cystic dilatations in Pkhd1^{del3-4/del3-4} mice, with both the number and size of cysts increasing progressively with age (Fig. 2c). Spontaneous hepatic fibrosis developed in Pkhd1^{del3-4/del3-4} mice in an age-dependent manner, with marked fibrosis evident at 6 months of age (Figs. 2c–f and S6). IHC staining for CK19 further confirmed the biliary epithelial origin of these cystic structures, demonstrating extensive intrahepatic bile duct expansion in Pkhd1^{del3-4/del3-4} mice (Fig. 2g). Taken together, the Pkhd1^{del3-4/del3-4} mouse model was successfully established and closely mirrored the disease phenotypic characteristics observed in the CHF patient.

Fig. 2. PKHD1 deficiency induces liver fibrosis.

a Gross morphology of livers from Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice. b Serum biochemical parameters, including ALT, AST, urea, and creatinine, were measured in Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice (n = 5). c Representative liver sections from Pkhd1^{del3-4/del3-4} mice at 1, 3, and 6 months stained with H&E, Masson’s trichrome, and Sirius Red, indicating progressive hepatic fibrosis; quantification of Masson and Sirius Red-positive area (%) is shown (n = 5). d Measurement of hepatic hydroxyproline content in Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice (n = 5). e Relative mRNA expression levels of hepatic Col1a1 and Acta2 in of Pkhd1^{del3-4/del3-4} mice at 1, 3, and 6 months, normalized to Pkhd1^+/+ controls (n = 6). f Relative protein levels of collagen I and α-SMA in of Pkhd1^{del3-4/del3-4} mice at 1, 3, and 6 months, normalized to Pkhd1^+/+ controls (n = 3). g IHC staining of CK19 in Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice livers at 1, 3, and 6 months of age. Quantification of CK19-positive area (%) is shown below (n = 6). Data are mean ± SD. Data presented in (b–f) were analyzed via two-way ANOVA with Tukey post hoc test. Data in (g) were analyzed via one-way ANOVA with Dunnett’s multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

SPP1 is upregulated in the CHF liver and predominantly expressed by cholangiocytes

To validate the scRNA-seq results, IHC for OPN was first performed on liver tissues from patients with CHF and controls, which showed markedly increased OPN staining in the biliary regions of the CHF liver (Fig. 3a). Then, liver tissues from Pkhd1^{del3-4/del3-4} mice were analyzed, revealing age-dependent increases in hepatic Spp1 mRNA and OPN protein levels compared with controls (Fig. 3b, c). As a secreted protein, serum OPN levels were also elevated in Pkhd1^{del3-4/del3-4} mice (Fig. 3d). To exclude the possibility that increased Spp1 mRNA and OPN protein expression was solely due to cholangiocyte expansion, fluorescence co-localization analysis was performed. OPN fluorescence intensity, normalized to the number of CK19-positive cells, demonstrated significantly higher average OPN expression per cholangiocyte in Pkhd1^{del3-4/del3-4} mice than in Pkhd1^+/+ mice (Fig. 3e).

Fig. 3. Upregulation of SPP1 expression and secretion in PKHD1-deficient liver tissues and cholangiocytes.

a IHC staining of hepatic OPN from CHF patient and control (n = 4). b RT-qPCR analysis of Spp1 mRNA expression in liver tissues from Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice at 1, 3, and 6 months of age (n = 6). c WB plot of OPN protein levels in livers from Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice (n = 3). d Serum OPN (osteopontin, encoded by SPP1) concentrations in Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice measured by ELISA at 1, 3, and 6 months (n = 4). e IF staining of liver sections from Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice, showing OPN (green) and cholangiocyte marker CK19 (red). OPN fluorescence intensity was normalized to the number of CK19-positive cells (n = 4). f Relative mRNA expression of Spp1 in MIB-WT and MIB-KO (n = 4). g WB plot of OPN protein levels in lysates of MIB-WT and MIB-KO (n = 4). h Secreted OPN in supernatants from MIB-WT and MIB-KO cells were measured by ELISA (n = 4). i SPP1 mRNA levels in cells transfected with si-PKHD1 orsi-NC (n = 3). j OPN protein levels were upregulated in PKHD1-knowckdown cells (n = 3). k Secreted OPN levels were measured by ELISA in culture supernatants (n = 6). Data are mean ± SD. Data presented in (a, e–g, h) were analyzed via Mann–Whitney U tests. Data presented in (b–d) were analyzed via two-way ANOVA with Tukey post hoc test. Data in (i, j) were analyzed via one-way ANOVA with Dunnett’s multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

According to the scRNA-seq results of the CHF patient, macrophages also showed increased expression of SPP1 in CHF and were another major source of hepatic SPP1 mRNA and OPN protein expression. However, co-localization studies revealed that OPN showed almost no overlap with the macrophage marker F4/80 (Fig. S7).

Primary intrahepatic biliary epithelial cells were isolated and cultured from Pkhd1^{del3-4/del3-4} mice (mouse intrahepatic biliary epithelial cell-knock out, MIB-KO) and Pkhd1^+/+ mice (mouse intrahepatic biliary epithelial cell-wild type, MIB-WT) (Fig. S8). Spp1 mRNA and OPN protein expression and secretion were significantly increased in MIB-KO (Fig. 3f–h). Furthermore, siRNA-mediated silencing of PKHD1 in HIBEpiC resulted in significant upregulation of SPP1 mRNA and OPN protein, accompanied by increased secretion (Figs. 3i–k and S9a, b).

Loss of PKHD1 in cholangiocytes promotes HSCs activation through OPN secretion

Previous studies have established the PI3K/Akt signaling pathway as a key regulator of HSCs activation and a downstream effector of OPN stimulation, which was corroborated in mouse models (Fig. S10a). To further validate these findings at the cellular level, LX2 was treated with recombinant human osteopontin (rOPN). The rOPN stimulated LX2 proliferation in a dose-dependent manner (Fig. S10b). Notably, treatment with 50 nM rOPN significantly enhanced LX2 proliferation, accompanied by upregulation of Collagen I and α-SMA expression and activation of PI3K/Akt signaling pathway (Fig. S10c–e).

To define the regulatory role and underlying mechanism of cholangiocytes with reduced PKHD1 expression on HSCs, HIBEpiC transfected with si-PKHD1 or si-NC were co-cultured with LX2, separately (Fig. 4a). Compared with the si-NC group, LX2 exposed to si-PKHD1-treated HIBEpiC exhibited significantly increased expression of activation markers, accompanied by enhanced PI3K/Akt phosphorylation, as well as increased proliferation and migration (Figs. 4b, c, and S11a, b). We next sought to validate these findings in a murine system by co-culturing MIB-WT or MIB-KO with mouse hepatic stellate cells (mHSCs) (Fig. 4d). Consistently, MIB-KO co-culture induced upregulation of ECM and increased PI3K/Akt signal in mHSCs, together with enhanced proliferation and migration (Figs. 4e, f, and S11c, d).

Fig. 4. PKHD1-deficient cholangiocytes promote HSCs activation through the secretion of OPN.

a Schematic illustration of the co-culture system of HIBEpiC and LX2. b mRNA expression levels of ACTA2 and COL1A1 in LX2 cells after co-culture (n = 6). c protein levels of activation markers and PI3K/Akt pathway in LX2 after co-culture (n = 3). d Schematic illustration of the co-culture system of MIB and mHSCs. e mRNA expression levels of Acta2 and Col1a1 in mHSCs after co-culture (n = 6). f protein levels of activation markers and PI3K/Akt pathway in mHSCs after co-culture (n = 3). g Spp1 mRNA expression in MIB-KO (n = 4). h mRNA expression levels of Acta2 and Col1a1 in mHSCs co-cultured with Spp1-knockdown MIB cells (n = 6). i Protein levels of activation markers and PI3K/Akt pathway in mHSCs after co-culture (n = 3). Data are mean ± SD. Data presented in (b, c, e, f, h, i) were analyzed via two-way ANOVA with Tukey post hoc test. Data in (g) were analyzed via one-way ANOVA with Dunnett’s multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

To determine the contribution of cholangiocyte-derived OPN to HSCs activation, Spp1 was knocked down in MIB-KO prior to co-culture with mHSCs (Fig. 4g). Spp1 knockdown markedly attenuated ECM expression in mHSCs and suppressed PI3K/Akt pathway compared with the si-NC group, indicating that the pro-activation effect of PKHD1-deficient cholangiocytes was at least partly mediated by SPP1 (Fig. 4h, i).

Cholangiocyte-derived SPP1/OPN activates HSCs via integrin α8β1 mediated PI3K/Akt signaling

CellChat analysis identified integrin α8β1 (composed of ITGA8 and ITGB1) as the primary receptor for SPP1 in HSCs (Fig. 1f). ScRNA-seq further suggested that both ITGA8 and ITGB1 co-localized strongly with the HSCs activation marker ACTA2, indicating that SPP1 secreted by cholangiocytes may bind to integrin α8β1 on HSCs and promote their activation (Fig. S12a–d). In addition, we found that ITGA8 and ITGB1 are most highly expressed in HSCs within the cell populations of the liver (Fig. S12e, f).

Ligand–receptor interaction analysis indicated that OPN can potentially engage multiple integrins, including α9β1 and αvβ1 (Fig. 1g). Consistent with prior reports, OPN is also capable of interacting with αvβ3 and CD44²⁰. To identify the dominant OPN receptor in CHF, we assessed the protein expression of relevant integrin subunits in Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice. Although several integrin subunits were upregulated to varying extents, ITGA8 and ITGB1 exhibited the most pronounced increase (Fig. 5a). Multiplex IF staining was performed to further validate receptor localization. In liver tissues from Pkhd1^{del3-4/del3-4} mice, integrin subunits ITGA8 and ITGB1 were markedly increased and exhibited co-localization with the HSCs marker Desmin. Notably, this co-localization was predominantly observed in proximity to OPN-expressing cholangiocytes. In contrast, such co-localization was not detected in Pkhd1^+/+ mice (Fig. 5b).

Fig. 5. Cholangiocyte-derived SPP1/OPN promotes HSCs activation via the integrin α8β1-mediated PI3K/Akt signaling pathway.

a Protein expression of different integrin receptors in the liver of Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice. b Co-localization of ITGA8/ITGB1, OPN, and Desmin in mouse liver. c Protein expression levels of ITGA8 and ITGB1 in mHSCs after co-culture with MIB-WT or MIB-KO cells (n = 3). d Protein expression levels of ITGA8 and ITGB1 in mHSCs after co-culture with MIB-KO treated with si-NC or si-Spp1 (n = 3). e Protein levels of activation markers and PI3K/Akt signaling pathway in LX2 under different co-culture conditions (n = 3). Data are mean ± SD. Data presented in (a, c, d, e) were analyzed via two-way ANOVA with Tukey post hoc test. **P < 0.01, ***P < 0.001, and ****P < 0.0001.

In the co-culture system, we observed that compared with co-culture with MIB-WT, MIB-KO markedly increased the protein levels of ITGA8 and ITGB1 in mHSCs (Fig. 5c). Notably, knockdown of Spp1 in MIB-KO significantly reduced ITGA8 and ITGB1 protein expression in mHSCs (Fig. 5d).

To further delineate the roles of integrin α8β1 and the PI3K/Akt signaling pathway in HSCs activation, we performed loss-of-function experiments. In LX2, ITGA8 and/or ITGB1 were silenced prior to co-culture with HIBEpiC transfected with si-NC or si-PKHD1. Pre-silencing of ITGA8 and/or ITGB1 markedly attenuated the expression of activation markers in LX2 (Fig. 5e). In parallel, LX2 was treated with the PI3K/Akt inhibitor LY294002 (20 μM) before co-culture with si-NC- or si-PKHD1-transfected HIBEpiC. Inhibition of PI3K/Akt signaling significantly reduced the expression of ECM-related proteins and suppressed the activation phenotype of LX2 (Fig. 5e). Collectively, these findings demonstrate that integrin α8β1 and PI3K/Akt signaling are critically involved in cholangiocyte-driven HSC activation induced by PKHD1 deficiency.

Mechanism of increased SPP1 expression in PKHD1-deficient cholangiocytes

Since scRNA-seq reflected transcriptomic changes, we hypothesized that the upregulation of SPP1 was driven by changes in transcription factor activity in CHF. To validate this, we designed biotin-labeled probes specific to the SPP1 promoter region and performed a DNA pull-down to enrich the promoter-binding proteins (Fig. 6a). The pulled-down proteins were identified by mass spectrometry to screen for potential upstream transcriptional regulators of SPP1. Then we cross-referenced proteins identified by DNA pull-down with predicted SPP1 promoter-binding factors from the JASPAR database combined with transcription factor databases, resulting in seven putative regulators (Fig. 6b). Notably, the percentage of cells co-expressing STAT3 and SPP1 was significantly higher in CHF samples than in controls, whereas no significant differences were observed for the other six candidate factors (Figs. 6c and S13A). JASPAR analysis predicted specific STAT3 binding sites within the SPP1 promoter (Fig. S13b, c). These findings suggest that STAT3 may act as a key transcription factor that mediates SPP1 upregulation in CHF.

Fig. 6. Upregulation and enhanced nuclear translocation of p-STAT3 in PKHD1-deficient cholangiocytes.

a Biotinylated oligonucleotide pull-down assay using SPP1 promoter probes to identify candidate SPP1 promoter-binding factors. b Venn diagram illustrating the identification of seven putative transcriptional regulators of SPP1 through integration of DNA pull-down results with predicted promoter-binding factors from the JASPAR and other transcription factor databases. c The proportion of cholangiocytes co-expressing STAT3 and SPP1 was significantly increased in CHF samples relative to control. d IHC analysis of p-STAT3 protein levels in liver tissues from CHF and control patients. e p-STAT3 protein level in liver from Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice (n = 6). f IF staining of p-STAT3 (red) and OPN (green) in liver sections from Pkhd1^+/+ and Pkhd1^{del3-4/del3-4} mice. g p-STAT3 expression in MIB-WT and MIB-KO (n = 4). h p-STAT3 expression in in PKHD1-knock down HIBEpiC (n = 3). i Enhanced nuclear translocation of p-STAT3 in PKHD1-knock down HIBEpiC. Data are mean ± SD. Data presented in (e, g) were analyzed via Mann–Whitney U tests. Data in (h) were analyzed via one-way ANOVA with Dunnett’s multiple comparisons test. *P < 0.05, **P < 0.01.

STAT3 exerts its transcriptional activity mainly through phosphorylation at amino acid position 705 (Tyr705)²¹. We found that p-STAT3 levels were markedly elevated in liver tissues from patients with CHF compared with control (Fig. 6d). Consistently, Pkhd1^{del3-4/del3-4} mice exhibited increased p-STAT3 protein levels relative to Pkhd1^+/+ mice, while total STAT3 remained unchanged (Fig. 6e). IF analysis further demonstrated that p-STAT3 was predominantly localized to the nuclei of cholangiocytes and showed strong co-localization with OPN signals (Fig. 6f). This finding was further confirmed in MIB-KO and HIBEpiC with PKHD1 silencing, where nuclear translocation of p-STAT3 was also observed (Fig. 6g–i).

To investigate the potential role of p-STAT3 in regulating SPP1 transcription, the STAT3 phosphorylation inhibitor Stattic was used. To exclude potential cytotoxic effects of Stattic, cell viability was assessed in HIBEpiC using a CCK-8 assay, which showed a moderate reduction in cell viability at higher concentrations (Fig. S14a). At 5 μM, Stattic could significantly inhibit the p-STAT3 without affecting total STAT3 levels (Fig. 7a, b). Stattic could also suppress both the expression levels of SPP1 mRNA and OPN protein (Figs. 7c, d, and S14b). Inhibition of p-STAT3 by Stattic significantly attenuated the upregulation of SPP1 mRNA and OPN protein expression induced by PKHD1 knockdown, suggesting a central role for p-STAT3 in SPP1 transcription in PKHD1-deficient cholangiocytes (Fig. 7e, f).

Fig. 7. p-STAT3-mediated transcriptional activation of SPP1 in PKHD1-deficient liver tissues and cholangiocytes.

a–d Dose-dependent reduction in SPP1 mRNA and OPN protein level and p-STAT3 protein level following treatment with increasing concentrations of Stattic (0–10 μM) (n = 3). e SPP1 mRNA levels in cells transfected with si-NC or si-PKHD1, with or without Stattic treatment (n = 6). f OPN protein levels under transfected with si-NC or si-PKHD1, with or without Stattic treatment. g Schematic depicting putative p-STAT3 binding sites (P1–P4) within the SPP1 promoter region predicted by JASPAR database. h Dual-luciferase reporter assay assessing the activity of WT and MUT in cells transfected with oe-NC or oe-STAT3 (n = 6). i ChIP-qPCR analysis of p-STAT3 binding to SPP1 promoter regions (primers 1–4). IgG served as negative control (n = 3). Data are mean ± SD. Data in (b–f) were analyzed via one-way ANOVA with Dunnett’s multiple comparisons test. Data presented in (h) were analyzed via Student t tests. Data presented in (i) were analyzed via two-way ANOVA with Tukey post hoc test. Ns not significant, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Using the JASPAR database, seven potential p-STAT3 binding sites were identified within the SPP1 promoter region, which were subsequently consolidated into four distinct sites (Fig. S13c). Based on these sites, mutant SPP1 promoter plasmids were constructed (Fig. 7g). Wild-type (WT) and mutant (MUT) SPP1 promoter luciferase reporter plasmids were co-transfected with either oe-NC or oe-STAT3 into HIBEpiC, followed by dual-luciferase reporter assay. STAT3 overexpression significantly increased luciferase activity driven by the WT promoter, whereas no effect was observed with the MUT promoter, indicating that STAT3 specifically binds to the SPP1 promoter to activate transcription (Fig. 7h). To further delineate the specific binding sites of p-STAT3 in the SPP1 promoter region, we designed four primers to detect these four binding site regions. The results showed the effective binding sites of the SPP1 promoter region and STAT3 on sites 1, 2 and 4 (Fig. 7i). In summary, p-STAT3 is markedly upregulated in cholangiocytes of CHF, enhancing its transcriptional activation of the SPP1 promoter.

Inhibition of Spp1 in Pkhd1^{del3-4/del3-4} mice attenuates liver fibrosis progression

To further clarify the role of Spp1 in CHF mice, we constructed an AAV8-based shRNA system targeting Spp1, in which a cholangiocyte-specific HNF1β promoter was incorporated to enhance cholangiocytes specificity. The vector was delivered via tail vein injection to Pkhd1^{del3-4/del3-4} mice, enabling efficient and targeted suppression of Spp1 mRNA and OPN protein expression in cholangiocytes. (Fig. 8a). To verify the cell-type specificity of Spp1 knockdown, mouse primary hepatocyte (MPH) and MIB were isolated. In MPH, Spp1 mRNA and OPN protein expression showed no significant changes following sh-Spp1 treatment, whereas in MIB, sh-Spp1 markedly reduced both Spp1 mRNA and OPN protein expression (Fig. 8b–e). These findings were further corroborated by IF co-localization analysis in liver tissues (Fig. 8f). In addition, a significant reduction in hepatic Spp1 mRNA and OPN protein expression was confirmed following sh-Spp1 (Fig. 8g–i). Serum ELISA analysis further confirmed a significant reduction in circulating OPN levels (Fig. 8j).

Fig. 8. AAV8-mediated Spp1 silencing attenuates liver fibrosis in Pkhd1^{del3-4/del3-4} mice.

a Experimental schematic depicting in vivo administration of AAV8-sh-Spp1 or AAV8-sh-NC into 3-month-old Pkhd1^{del3-4/del3-4} mice. b Protein expression level of OPN in MIB. c Protein expression level of OPN in MPH. d Spp1 mRNA expression levels in MIB and MPH (n = 6). e Quantification of OPN protein expression (n = 3); f co-localization of the cholangiocyte marker CK19 and OPN, and the hepatocyte marker HNF4A and OPN. g Spp1 mRNA expression levels in liver tissues (n = 6); h protein expression level of OPN in liver tissues and its quantification (n = 3); i IHC analysis of OPN expression and distribution (n = 3). j Serum ELISA measurement of circulating OPN levels in mice (n = 6). k Representative images of Masson’s trichrome and Sirius Red staining of mouse liver tissues; l protein expression levels in liver tissues (n = 3); m mRNA expression levels of fibrosis markers in liver tissues (n = 6); n Western blot analysis of integrin α8β1/PI3K/Akt signaling pathway proteins (n = 3). Data are mean ± SD. Data presented in (d, e, m, n) were analyzed via two-way ANOVA with Tukey post hoc test. Data presented in (g, j) were analyzed via Student t tests. *P < 0.05, ***P < 0.001, and ****P < 0.0001.

Downregulation of Spp1 mRNA and OPN protein expression markedly attenuated liver fibrosis and ECM accumulation in sh-Spp-treated mice (Fig. 8k–n). Moreover, Spp1 knockdown inhibited the PI3K/Akt signaling pathway by downregulating the expression of hepatic integrin α8β1 expression (Fig. 8l). These findings indicate that the suppression of hepatic Spp1 expression ameliorates liver fibrosis progression in mice with CHF.

In summary, PKHD1-deficient cholangiocytes exhibit increased levels of p-STAT3, which enhanced transcriptional activation of the SPP1 promoter, leading to increased SPP1 mRNA and OPN protein expression. Secreted OPN binds to integrin α8β1 on HSCs, thereby activating HSCs through the PI3K/Akt pathway and contributing to the progression of liver fibrosis.

Discussion

In this study, we demonstrated the direct interaction between cholangiocytes and HSCs in the pathogenesis of CHF, which is mediated through SPP1. Our key findings were as follows: (1) cholangiocytes in the CHF model exhibit increased expression of SPP1 mRNA and OPN protein; (2) cholangiocyte-derived OPN promotes HSCs activation by binding to the integrin α8β1 receptor expressed on HSCs; (3) upregulation of SPP1 in cholangiocytes is driven by increased expression and enhanced nuclear translocation of the transcription factor p-STAT3; (4) suppression of hepatic Spp1 expression in Pkhd1^{del3-4/del3-4} mice attenuates fibrosis progression.

In this study, the patient carries a paternal missense variant (c.1559A>T, p.N520I) and a maternal frameshift duplication (c.8746dupG). The c.1559A>T substitution causes an amino acid change at position 520 from asparagine to isoleucine, while c.8746dupG introduces a frameshift predicted to cause premature termination and consequent loss of function²². Large cohort studies have established that loss of function (LOF) variants in PKHD1 are strongly associated with CHF, frequently observed in compound heterozygosity^23,24. The Pkhd1^{del3-4/del3-4} mouse line, generated by deletion of exons 3–4, results in a frameshift consistent with a functional knockout and has been recognized as a validated CHF model in prior reports¹⁶. In our study, Pkhd1^{del3-4/del3-4} mice recapitulate the hepatic manifestations observed in CHF patient, including bile duct proliferation and progressive portal fibrosis. To further assess systemic biochemical changes, we measured serum ALT, AST, urea, and creatinine at 3, 6, 9, and 12 months, and observed mild increases at 12 months. This is consistent with previous reports showing minimal changes in routine liver and renal function in Pkhd1^del4/del4 models before²⁵. These findings suggest that CHF primarily reflects cholangiocyte dysfunction and biliary remodeling, with biochemical abnormalities likely emerging only at advanced stages of disease progression, consistent with clinical observations that liver dysfunction in CHF patients typically becomes evident at later stages⁵. Taken together, these data support the Pkhd1^{del3-4/del3-4} mice as a valid model of CHF.

Given that PKHD1 expression in the liver is predominantly confined to cholangiocytes, elucidating the crosstalk between cholangiocytes and HSCs and identifying the key mediators are essential for understanding and targeting fibrosis progression. ScRNA-seq analysis enables the identification of transcriptomic alterations across distinct cell populations and facilitates the elucidation of intercellular communication networks using integrative analytical approaches²⁶. To clarify the mediators of HSCs activation and the communication between cholangiocytes and HSCs, we performed scRNA-seq on liver tissues from one patient with CHF and controls to identify cholangiocyte-derived factors driving fibrosis progression. Notably, we report that OPN is an upregulated secreted protein in PKHD1-deficient cholangiocytes.

SPP1 mRNA and OPN protein overexpression promotes HSCs activation, thereby contributing to the progression of liver fibrosis²⁷. Previous studies have demonstrated elevated OPN levels in metabolic dysfunction-associated steatohepatitis (MASH)-associated fibrosis, with serum OPN positively correlating with liver stiffness in cirrhosis patients, thereby serving as a noninvasive marker for fibrosis progression^28–30. This study demonstrates that increased SPP1 expression in cholangiocytes contributes to fibrosis in CHF. Using the successfully established mouse and cellular models, we confirmed that SPP1 expression is elevated in cholangiocytes lacking PKHD1, validating the findings from scRNA-seq of the CHF patient. Notably, in CHF mice, SPP1 expression increases progressively with age, which likely reflects the cumulative nature of disease progression. As PKHD1 deficiency induces ongoing cholangiocyte stress and bile duct malformation, the resulting chronic injury may contribute to gradually elevated SPP1 levels³¹. Additionally, age-dependent accumulation of portal inflammation and fibrotic signaling further promotes SPP1 upregulation, highlighting its role as a dynamic mediator of hepatic fibrosis in the context of CHF³².

Under physiological conditions, cholangiocytes and macrophages are the primary sources of OPN in the liver³³. However, our co-localization analysis revealed that in CHF mice, cholangiocytes, rather than macrophages, were the predominant cells that expressed and secreted OPN. Given the marked expansion of the cholangiocyte population in CHF, we next sought to determine whether the increased SPP1 expression was driven by hyperplasia or by per-cell upregulation within individual cholangiocytes. To address this, we first analyzed our scRNA-seq dataset, which demonstrated a significant increase in SPP1 expression at the single-cell level in cholangiocytes from CHF patient, indicating per-cell upregulation rather than solely an increase in cell abundance. Consistently, primary cholangiocytes isolated from Pkhd1^{del3-4/del3-4} mice exhibited significantly elevated Spp1 mRNA and OPN protein expression compared to Pkhd1^+/+. In parallel, IF analysis normalized to CK19-positive cholangiocytes showed increased OPN signal intensity per cell, providing additional spatial support for this observation.

To further investigate the molecular mechanisms driving SPP1 upregulation, a DNA pull-down assay coupled with proteomic analysis was conducted using the SPP1 promoter, which identified seven candidate transcription factors. Among these, p-STAT3 was confirmed to be a transcriptional activator of SPP1 in cholangiocytes. PKHD1 deficiency in cholangiocytes promoted STAT3 phosphorylation and subsequent nuclear translocation of p-STAT3, which augmented its binding to the SPP1 promoter, resulting in elevated SPP1 transcription and protein secretion. These findings are consistent with previous reports on STAT3 activation in PKHD1-deficient cells, in which IL-1β drives STAT3 phosphorylation and β-catenin, promoting its translocation to the nucleus³⁴. Previous studies have shown that both SPP1 and STAT3 signaling are commonly activated across different liver fibrosis models. However, t the cell type involved in this activation varies across different models. For instance, in bile duct ligation (BDL)- and CCl₄-induced liver fibrosis models, STAT3 activation in HSCs has been shown to drive their activation³⁵. In addition, STAT3 signaling in liver sinusoidal endothelial cells has also been reported to promote fibrogenesis in similar models³⁶. In MASH, hepatocytes have been identified as a major source of OPN, which in turn contributes to HSCs activation³⁰. In contrast, our findings demonstrate that the activation of the p-STAT3/OPN axis occurs in cholangiocytes. This suggests that, while the signaling axis may represent a general profibrotic mechanism, the cellular origin of its activation is context-dependent and may vary across different disease models.

SPP1/OPN is a well-recognized profibrotic mediator that has been reported to promote HSCs activation via receptors, such as integrins or CD44, involving the PI3K/Akt signaling pathway^33,37–39. In this study, scRNA-seq analysis revealed that cholangiocyte-derived SPP1 mediates HSCs activation through interaction with integrin α8β1 in HSCs. Furthermore, the co-culture of MIB-KO with Spp1 knockdown attenuated mHSCs activation, confirming the functional role of SPP1 in the fibrotic process.

To evaluate the therapeutic impact of Spp1 targeting in vivo, we employed an AAV8-based gene silencing strategy in the Pkhd1^{del3-4/del3-4} mice. Previous studies have shown that, while AAV8 efficiently transduces hepatocytes, it can also access and transduce a subset of cholangiocytes with prolonged systemic circulation, indicating its capacity to reach the biliary compartment^40,41. To further enhance cholangiocyte-directed targeting, we incorporated a cholangiocyte-associated HNF1β promoter into the AAV8 vector. HNF1β is highly expressed in cholangiocytes and plays a critical role in their development and maturation, while exhibiting minimal expression in hepatocytes^42–44. Therefore, this study incorporated a cholangiocyte-associated HNF1β promoter into the AAV8 vector to achieve cholangiocyte knockdown, thereby improving the specificity of intervention and interpretability of the results.

Owing to the rarity of CHF, the pathogenesis of fibrosis remains poorly understood and is postulated to be initiated by cholangiocytes¹⁴. Previous studies have shown that macrophages play a pivotal role in liver fibrosis. Cholangiocytes with FPC deficiency secrete chemokines that recruit macrophages and promote their polarization toward an M2 phenotype, thereby contributing to fibroblast activation and fibrogenesis¹⁴. Although previous studies have underscored macrophage-mediated signaling in fibrogenesis, our findings highlight a direct profibrotic crosstalk between cholangiocytes and HSCs. Notably, although clodronate-mediated macrophage depletion alleviated hepatic fibrosis in CHF mice, it did not result in complete resolution, suggesting the involvement of nonmacrophage cell types, such as cholangiocytes—in driving HSCs activation and fibrogenesis. Moreover, the discrepancy in intervention timing-clodronate administered at 3 months versus AAV8-sh-Spp1 delivered at 1 month in our study-may also account for the differential therapeutic outcomes.

A recent study using scRNA-seq of ARPKD-derived organoids demonstrated activation of TGF-β signaling in cholangiocytes, expansion of fibroblast populations, and upregulation of PDGFR-β pathway, key molecular events that contribute to the initiation and progression of hepatic fibrogenesis⁴⁵. While organoids offer valuable insights, their in vitro nature limited the ability to fully replicate the complexity of native tissue architecture. In contrast, our study integrates scRNA-seq data from liver tissues of a CHF patient and validation in PKHD1-deficient mouse models, providing in vivo evidence that supports and extends previous organoid-based findings.

A limitation of our study is that only one CHF patient sample was subjected to scRNA-seq. This is primarily due to the rarity of CHF patients, the limited availability of patient specimens, and the requirement for high tissue freshness in scRNA-seq. Nevertheless, the findings from scRNA-seq were validated by both mouse and cellular models, thereby providing robust supporting evidence for the conclusions.

In summary, in PKHD1-deficient cholangiocytes of CHF, the transcription factor p-STAT3 is aberrantly activated, leading to upregulation of SPP1 expression. OPN, the protein encoded by SPP1, subsequently acts in a paracrine manner by binding to integrin α8β1 receptors on HSCs, activating the downstream PI3K/Akt signaling pathway, thereby promoting HSCs activation and exacerbating liver fibrosis. In vivo, cholangiocyte-specific silencing of Spp1 inhibits the integrin α8β1/PI3K/Akt signaling axis and alleviates liver fibrosis.

Author contributions

Chuanlong Zhu and Mengfan Jiao designed this research; Mengfan Jiao, Ye Sun, and Rui Liu conducted experiments and wrote the manuscript; Xizhi Yin contributed to human specimens; Xiaoxue Hou, Hui Yuan, and Zixing Dai analyzed the data; Qingling Chen revised the manuscript. All authors read and approved the final manuscript.

Peer review

Peer review information

Communications Biology thanks Nidhi Jalan-Sakrikar and the other anonymous reviewer(s) for their contribution to the peer review of this work. Primary handling editors: Kaliya Georgieva.

Data availability

The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2025) in the National Genomics Data Center (Nucleic Acids Res 2025), China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA018715) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human. The quantitative Source data of all figures in the text can be found in Supplementary Data. The untrimmed original blot images are included in the Supplementary Information. Other data during the current study are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains Supplementary material available at https://doi.org/10.1038/s42003-026-10476-6.