The procedures regarding human subjects were approved by the Ethic Committee of Fudan University School of Basic Medical Sciences (2022-C008). Formalin-fixed, paraffin-embedded liver slices from five PBC patients and five control livers from non-PBC patients were obtained from the Biorepository of Pathologic Specimens in the Department of Pathology, School of Basic Medical Sciences, Fudan University. All patient identification was removed, and the sections were re-numbered for annotation. Patient information is briefly summarized in Table 1.

Mice were housed in specific pathogen-free (SPF) facilities at Fudan University and Shanghai Model Organisms. All experimental protocols (2021-0302-051, 2023-0301-088) were approved by the Ethic Committee for Experimental Animal Use and Care, Fudan University School of Basic Medical Sciences, and performed in line with the NIH Guidelines for Experimental Animal Handling and Use, as well as the national, municipal, and university regulations. Breading pairs of global homozygous JCAD knockout C57BL/6J (JCAD-KO) mice were obtained from RIKEN Center for Life Science Technologies, Kobe, Hyogo, Japan [13], and utilized to generate offspring, and offspring chimeric mice were identified by PCR using primer pairs as reported [13]. Only male JCAD-KO mice were used in the present study in order to avoid gender difference in response to cholestatic insults. The wild-type (WT) C57BL6/J mice were used as controls. WT and JCAD-KO mice (at least 10 weeks of age) were subjected to sham or BDL without cholecystectomy for up to 14 days as we described previously [17,18]. Mice were sacrificed 2 weeks after BDL procedures, and liver tissue was collected and stored at –80°C for histopathologic, biochemical, and molecular analyses as reported in our previous study [19].

Hepatic stellate cell-specific conditional JCAD knockout mice (JCAD^flox/flox, GFAP-cre^+/+, named as HSC-JCAD-KO) were generated by crossing floxed JCAD mice containing the exon 3 open-reading frame of JCAD (generated in Shanghai Model Organisms, Shanghai, China) with the GFAP-cre/ERT2 strain mice (a generous gift from Prof. Yong Xu in Nanjing Medical University, Nanjing, China), as reported [20,21]. To activate creenzyme for JCAD deletion, 5- to 6-week-old male mice were injected daily with tamoxifen (TAX, 100 mg/kg) for 5 consecutive days, followed by a 7-day interval for the clearance of tamoxifen. After verification with genome typing and tamoxifen treatment (Supplementary Fig. 8A–D), mice with the correct genomic type and phenotype were subjected to BDL. Wild-type littermates (JCAD^flox/flox, GFAP-cre^-/-, defined as WT in short) were used as controls to determine whether JCAD deficiency in HSCs results in minimal or no fibrosis in homozygous mice. All experiments were performed in male mice. Primers for identification of mouse genome types were listed in Supplementary Table 1.

All data were expressed as mean±standard deviation (SD). SPSS software was used for statistical analysis. When two groups were compared, student t-test was used after normal distribution test. When the experimental design involves more than 2 groups, one-way ANOVA was used to compare between groups after normal distribution test, and multiple comparisons between two given groups were completed by the least significant difference (LSD) test. If normality or homogeneity of variance is not met, Kruskal-Walli and Hodges-Lehmann tests were used for multiple comparisons between two given groups. *, **, *** represent as P<0.05, 0.01 and 0.001; P-value<0.05 was considered as statistically significant.

Liver sections from control subjects without chronic cholestatic liver disease and patients with PBC were examined for counter-staining of JCAD with α-SMA protein. In control livers, positive α-SMA staining was localized to fibroblast-like cells in the portal triads as evidenced in H&E staining showing the bile ductule (Fig. 1, top panel). JCAD staining was largely negative in this area in control livers. However, immunofluorescent staining of serial sections revealed that JCAD was strongly positive in the biopsied tissue of PBC patients, and was overlapped in α-SMA-positive fibroblast-like cells (Fig. 1 white arrow), which illustrated that JCAD, a membrane conjunction protein, was expressed most in activated HSCs or possibly in PFs too in the portal triads when it was in an early phase of bile duct reaction. Moreover, positive JCAD staining was visualized in BECs with the ductular reactions in sections of biopsied tissue from PBC patients (Fig. 1 yellow arrow). These findings indicate that enhanced JCAD activity in BECs and HSCs may participate in the biliary ductular reaction and fibrosis of cholestatic disorders in human.

To study the potential role of JCAD in cholestatic insult, BDL was utilized as an animal model of cholestasis. Liver tissue was harvested 14 days after BDL, which resulted in acute obstructive cholestasis and portal fibrosis at this time point. The gross examination and histopathology showed a huge gallbladder, massive liver necrosis and profound fibrosis(Supplementary Fig. 1A). JCAD expression was examined in the liver after BDL. A significant increase in JCAD protein levels was detected by Western blotting analysis (Fig. 1B, C) in consistency with elevated mRNA levels compared to sham-operated mice (Supplementary Fig. 1B). Immunostaining confirmed that JCAD expression was increased in fibrotic liver after BDL compared to sham-operated mice (Fig. 1D). In agreement with the immunofluorescent staining of liver sections from PBC patients, JCAD expression in cholestatic livers was enhanced along with upregulated α-SMA, CK19 and YAP expression (Fig. 1B, C). Thus, it is tempting to speculate that JCAD is an important player in the regulation of biliary ductular reaction and fibrosis following BDL in mice or in cholestatic disorders in humans.

To test whether increased JCAD levels are critical for hepatic response, BDL was performed in WT and JCAD-KO mice. Liver histopathology revealed that the WT-BDL mice displayed increased focal necrosis, portal inflammation and bile leakage (Supplementary Fig. 1A). In JCAD-KO-BDL mice, the degree of focal necrosis was attenuated histopathologically compared to WT-BDL (Fig. 2A). The ratio of liver over body weight did not differ between WT and JCAD-KO mice after BDL. In contrast, JCAD-KO-BDL mice restored the reduction in liver weight and body weight compared to WT-BDL mice (Supplementary Fig. 2A). Serum levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were significantly elevated in WT-BDL mice; however, they were reduced remarkably in JCAD-KO-BDL mice (Supplementary Fig. 2B). Similarly, serum levels of bilirubin and bile acids were strikingly decreased in JCAD-KO-BDL mice compared to WT-BDL mice.

Effects of JCAD deficiency on intrahepatic bile duct mass were assessed with immunohistochemical staining for CK-19 (a specific marker of bile duct epithelial cells), and it was found that the positivity of CK-19 staining in JCAD-KO-sham mice was similar to WT-sham mice (Fig. 2B, Supplementary Fig. 3B). Comparing WT-BDL with WT-sham or JCAD-KO-sham mice revealed a large increase in ductal mass in WT-BDL mice; whereas JCAD-KO-BDL mice exhibited a significant decrease in intrahepatic ductal mass (Fig. 2B–D). In addition, JCAD deficiency was manifested with decreased intrahepatic bile duct mass, as evidenced by decreased Ki67-positive cell counts and 5-ethynyl-2’-deoxyuridine (EdU) incorporation (Fig. 2C, Supplementary Fig. 3A, B).

To further confirm whether JCAD modulates in vivo cholestatic fibrogenesis, fibrotic progression and collagen deposition were assessed. As shown in Figure 2E, 2F, BDL caused remarkable collagen accumulation in the liver compared to the sham-operated mice in Sirius red and Masson’s trichrome staining. In contrast, collagen deposition was dramatically reduced in JCAD-KO-BDL mice (Fig. 2E, F). Consistently, semiquantitative score of hepatic fibrosis and photometric analysis of Masson/Sirius red staining verified that the collagen fiber-deposited area was significantly increased in the BDL mice; however, it was remarkably decreased in JCAD-KO mice after BDL (Fig. 2G, Supplementary Fig. 3C). This finding was further confirmed by immunofluorescent staining for α-SMA, CK-19, and in situ EdU labeling (Fig. 2C, D). Liver hydroxyproline content was significantly decreased in JCAD-KO-BDL mice compared to WT-BDL mice (Fig. 2H). Moreover, Western blot images showed that α-SMA and CK-19 protein levels were elevated in BDL-induced fibrotic liver, and JCAD-KO dramatically reduced expression of these two markers at protein levels (Fig. 3A, B). Similarly, gene expression of α-SMA, TGF-β, CTGF, and procollagen type-1 was significantly up-regulated in WT-BDL mice compared to sham-operated WT; whereas, expression of these genes was much suppressed in JCAD-KO-BDL mice (Fig. 3C, Supplementary Fig. 3D). Given that TGF-β1 is a well-known fibrogenic cytokine, expression of TGF-β1 in the liver and secretion of TGF-β1 in serum from these mice were determined. As expected, the levels of TGF-β1 in hepatic homogenates and serum were remarkably decreased in JCAD-KO-BDL compared to WT-BDL mice (Fig. 3D). Collectively, these findings documented that JCAD deficiency mitigated bile ductular reaction and fibrosis after BDL in mice.

Moreover, the liver tissue was examined by transmission electron microscopy using semi-thin tissue sections, which enables the resolution and visualization of ultrastructural details. As shown in Figure 4I, hepatocytes from sham-operated WT mice exhibited a normal ultrastructure of the nucleus, mitochondria and lysosomes. The plasma membrane of these subcellular organelles is mostly smooth and flat except where it is enfolded to form the bile canaliculi (BC) in the bilateral faces, and there exist many microvilli projecting into the BC lumen (Fig. 4A-I, 4B-I, red arrow). HSCs are situated in the space of Disse, where innumerable microvilli from the apical face of hepatocytes extend into (Fig. 4C-I, 4D-I). The nuclei of stellate cells are oval or more or less elongated and frequently indented by lipid droplets in the cytoplasm (Fig. 4D-I). Liver sinusoidal endothelial cells (LSEC) form the lining of the sinusoidal wall, and their thin, flat projections of cytoplasm are spread out and contain fenestrae (Fig. 4C-III). On the contrary, ultrastructural features of BDL mice were characterized with cholestasis, activation of Kupffer cells and stellate cells, bile duct damage, ductular reaction, and hepatocellular damage and regeneration [22]. In hepatocytes from WT mice with BDL, many damaged mitochondria with broken cristae and increased density, vacuolated matrixes, and an increased number of primary lysosomes were observed (Fig. 4A-II, 4B-II), yet no clear changes in the nucleus and endoplasmic reticulum (ER) were seen. The bile canaliculus was massively expanded, and the microvilli disappeared (Fig. 4A-II, 4B-II, red arrow). Number of lipid droplets in HSCs was significantly reduced (Fig. 4C-II, 4D-II). As shown in Fig. 4III, global JCAD knock-out did not cause marked changes in liver ultrastructure at this age compared to WT mice. However, the ultrastructure was greatly improved in JCAD-KO-BDL mice compared to WT mice with BDL, representing as expansion of bile canaliculi was reduced with preservation of many microvilli (Fig. 4A-IV), the density of mitochondria was decreased (Fig. 4B-IV), and quiescent stellate cells were observed in the sinusoids (Fig. 4C-IV, 4D-IV). Taken together, these ultrastructural findings demonstrated that JCAD deficiency alleviated BDL-induced cholestatic insults and fibrogenic response in mice.

Based on the findings as shown above, it is to speculate that JCAD is essential for HSC activation. To test this hypothesis, JCAD localization was assayed in human hepatic HSCs by fluorescent immunohistochemistry. As shown in Figure 5A, JCAD was expressed in stellate cells and was increased in a same pattern as α-SMA, a well-known marker of activated HSCs. Positive staining of the JCAD protein was mainly located in the cytoplasm, and activated stellate cells were prone to be in a dendritic-shape. When JCAD expression was knocked-down by siRNA-JCAD transfection, HSCs tended to be in a quiescent stage with an oval shape. As expected, transfection of siRNA against JCAD was efficient, and it suppressed HSC activation as measured by fibrosis-associated genes and proteins, such as α-SMA, COL-1α1, TIMP1, and CTGF in human LX2 and HSC lines (Fig. 5B, C). Moreover, suppressing JCAD expression in these stellate cells counteracted TGFβ-stimulated proliferation by increasing the ratio of p-YAP over YAP (Fig. 5B, D), which is the key transcription factor in the Hippo signaling pathway, indicating that the activity of the Hippo-YAP signaling pathway was significantly boosted during the activation of HSCs.

To further investigate whether JCAD is directly involved in modulating HSC activation, primary HSCs were isolated from WT and JCAD-KO mice and plated on plastic dishes for culture-induced activation [23]. Morphologically, quiescent HSCs stored vitamin A as cytoplasmic lipid droplets; when activated by culture in plastic dishes or profibrogenic stimulus with 10 ng/mL of TGF-β, they transformed into myofibroblast-like cells characterized by the loss of vitamin A droplets. In freshly isolated HSCs from wild-type mice, there were many small lipid droplets in the cytoplasm, and lipid droplets disappeared on the fifth day of in vitro culture. Whereas, large lipid droplets remained in the cytoplasm at day 1, and still were visible till day 5 in HSCs isolated from JCAD-KO mice (Fig. 6A). The morphology of primary HSCs was consistent with the gene expression results (Fig. 6B). Moreover, primary HSCs, isolated from WT and JCAD-KO livers were activated with stimulation of TGF-β at 10 ng/mL. JCAD deprivation significantly abrogated HSC activation and reduced expression of proliferating cell nuclear antigen (Pcna) and cyclin D1 as markers of proliferation in the presence of TGF-β (Fig. 6C, D). Even in the absence of TGF-β stimulation, expression of α-SMA, Ctgf, Timp1, and procollagen type-1α was significantly suppressed in primary HSCs isolated from JCAD-KO-BDL mice compared to those from WT-BDL mice (Fig. 6C). In summary, these findings from in vitro or ex vivo experiments argued that JCAD is essential for HSC activation in mice with BDL.

In order to prove whether JCAD acts on HSCs through the Hippo-YAP signaling axis, a YAP/TEAD-based dual-luciferase reporter assay was utilized. As shown in Figure 7A, JCAD depletion reduced YAP/TAZ/TEAD transcriptional activity, while JCAD over-expression caused a significant induction of 8xGTIIC luciferase reporter activity, with YAP siRNA and HA-YAP transfection as negative and positive controls. Furthermore, JCAD was knocked-down using siRNA transfection in LX2 cells, and thereafter, they were treated with Lats-IN-1 one day after transfection. Lats-IN-1 is a potent and ATP-competitive inhibitor of Lats1/2 kinases and drives Yap-dependent proliferation in post-mitotic mammalian cells [24]. It was observed that with the effective JCAD inhibition, the degree of HSC activation was attenuated along with decreased YAP expression (Fig. 7B). Once Lats-IN-1 was supplemented, expression of YAP and α-SMA was increased in Ctrl (scrambled)-siRNA-transfected LX2 cells; whereas their expression was not significantly changed in JCAD-siRNA-transfected cells with blocking LATS1/2 kinase activity by Lats-IN-1 (Fig. 7B, C).

Since the findings from above-mentioned experiments indicated that JCAD is positively correlated with the expression of YAP, the function of YAP during HSC activation was evaluated with two different strategies: firstly, LX2 cells were treated with the pharmacological inhibitor of YAP: verteporfin (VP), that disrupts the YAP/TEAD complex. Treatment of LX2 cells with VP at the time of seeding abolished the induction of α-SMA and YAP expression 48 hours after seeding, and suppressed expression of TGF-β, TIMP1, CTGF genes, etc. (Fig. 7B, C). Secondly, the downregulation of YAP levels was achieved by siRNA transfection. As evidenced by immunofluorescent staining and Western blot analysis, YAP expression was in accordance with decreased α-SMA expression (Fig. 7D, E, Supplementary Fig. 4A). Control cells that were transfected with scrambled siRNAs resulted in robust expression of TGF-β, TIMP1, CTGF, and COL-1α1 genes, indicating that they were efficiently activated. In contrast, these genes were not increased in LX2 cells with YAP RNAi knockdown (Supplementary Fig. 4B). Collectively, these results demonstrated that YAP is essential for HSC activation, and JCAD facilitated HSC activation via the Hippo-YAP signaling axis.

Finally, in order to further prove the rule of JCAD in HSC activation, HSC-specific JCAD knock-out mice were subjected to sham or BDL surgical procedure. As shown in Western blotting and qRT-PCR analyses, JCAD levels were down-regulated in HSCs isolated from tamoxifen (TAX)-injected mice compared to corn oil-injected mice (Supplementary Fig. 7C). The H&E staining in liver sections indicated that conditional inactivation of JCAD in HSCs did not significantly affect the extent of liver injury (Fig. 8A) and serum ALT and AST levels were similar between WT and HSC-JCAD-KO mice (data not shown). However, liver fibrogenesis was appreciably ameliorated in HSC-JCAD-KO mice after BDL as evidenced by Masson’s trichrome (Fig. 8B) and Sirius red staining (Fig. 8C), showing less fibrogenesis in HSC-JCAD-KO mice compared to WT mice after a BDL procedure.

Second harmonic generation/two-photon excitation fluorescence (SHG/TPEF) microscopy with artificial intelligence (AI) analysis provides a standardized evaluation of liver fibrosis and collagen fiber quantitation on a continuous scale. This approach was employed to gain an in-depth understanding of fibrogenic process resulting from the BDL model. As shown in Figure 8D, 8E, HSC-JCAD-KO mice were undergone milder fibrogenesis than WT-mice. The percentage of fibrotic area was further determined in 5 separate regions: the portal tract, periportal (the area within 100 μm around the portal tract), transitional (zone 2), pericentral, and central vein areas (Supplementary Fig. 5A). The combination of periportal, transitional, and pericentral areas represents the perisinusoidal area with zone 2 being the largest component. Zonal analysis of fibrotic dynamics revealed that BDL procedure resulted in a large amount of collagen deposition in the entire liver, especially in the perisinusoidal area (Supplementary Fig. 5B, C); whereas there was less fibrotic deposition in the portal tract and central vein areas. Fibrotic extent was significantly mitigated in the HSC-specific JCAD-KO mice undergone BDL, especially in the perisinusoidal area (Supplementary Fig. 5B, C).

These observations were further verified by protein expression of α-SMA in the liver, which was highly elevated in WT-BDL mice, but alleviated in HSC-JCAD-KO-BDL mice (Fig. 8F, Supplementary Fig. 6A). Moreover, measurements of hydroxyproline levels verified that liver fibrosis was significantly attenuated in HSC-JCAD-KO mice compared to WT mice after BDL (Fig. 8G, P=0.0541). In consistency with global JCAD knockout mice, expression of fibrosis-associated genes in HSC-JCAD-KO mice after BDL was strikingly reduced (Supplementary Fig. 6B). Furthermore, HSCs were isolated 5 or 14 days after BDL, representing early and late time points of in situ HSC activation. Thereafter, expression of fibrotic genes was determined to assess extent of in situ HSC activation. As shown in Fig. 8H, Supplementary Fig. 6C, expression of Tgf-β, Ctgf, α-SMA, Col-1α1, Col-3α1 and Timp1 was markedly down-regulated in HSCs from HSC-JCAD-KO mice in comparison to those from WT mice. These findings further confirmed that JCAD was considerably essential for HSC activation in the progression of cholestatic fibrosis, implying that JCAD may be an alternative target to inhibit fibrogenesis.