Prophylactic but not therapeutic OCA dosing impedes fibrosis
Previous studies on various animal models revealed that FXR agonists exert anti-fibrotic effects36,37,38,39, however, clinical trials revealed only modest efficacy in humans. Notably, OCA is not effective against liver fibrosis in PBC patients21,22 and only a quarter of NASH patients, despite statistical significance, showed improvement in liver fibrosis in a phase III clinical study19. Although there are diverse causes underlining the discrepancy between preclinical and clinical results, a big concern is that FXR agonists in most preclinical animal models were administered in a prophylactic manner, at a stage when there is no apparent fibrotic changes in the liver, which is totally different from the practical treatment of human patients. To address this concern, the effects of OCA were compared in liver fibrosis between prophylactic and therapeutic administration (Fig. 1a). As expected, prophylactic but not therapeutic administration of OCA significantly reduced serum ALT levels (Fig. 1b). Masson and Sirius red staining of liver section revealed a significant increase in the fibrotic surface upon CCl4 treatment. Compared with the CCl4-treated group, the prophylactic arm showed marked reduction in fibrotic surface, while the therapeutic arm showed marginal reduction (Fig. 1c). In line with the histological analysis, results from the mRNA expression of pro-fibrotic genes (including Acta2, Col1a1, Col1a2, and Tgfb1) further demonstrated that prophylactic but not therapeutic administration of OCA were effective against liver fibrosis (Fig. 1d). Consistently, mRNA levels of various pro-fibrotic genes in HSCs isolated from the mice with prophylactic but not therapeutic treatment of OCA showed dramatic reduction (Supplementary Fig. 1). Since loss of LDs is a hallmark of HSC activation, the lipid contents in HSCs were measured. The contents of retinoic acid (RA), triglycerides (TG), cholesterol (CHO) in HSCs from CCl4-treated mice were all significantly reduced, and prophylactic but not therapeutic treatment of OCA reversed the loss of these lipids (Supplementary Fig. 1). These findings were validated in BDL-induced liver fibrosis (Fig. 1e). In line with the results in CCl4-treated mice, prophylactic but not therapeutic administration of OCA showed anti-fibrotic effects in BDL mice, as revealed by serum ALT and AST levels (Fig. 1f), histological analysis (Fig. 1g), as well as the expression of pro-fibrotic genes (Fig. 1h). Results from primary HSCs also demonstrated that prophylactic but not therapeutic administration of OCA prevented HSC activation and LD loss (Supplementary Fig. 1).
NASH is a major pathological driver of liver fibrosis. We thus further tested this concept in two typical NASH models. Mice were fed with high fat plus high CHO diet and fructose/sucrose water (HFHC) for 16 weeks (Fig. 2a). HFHC mice were characterized with increased serum ALT and AST levels, pronounced steatosis, inflammation, ballooning, and fibrosis, and enhanced expression of pro-fibrotic genes (Fig. 2b–d). Prophylactic (from the 9th week) but not therapeutic (from the 13th week) administration of OCA significantly reduced serum ALT and AST levels (Fig. 2b), and ameliorated liver fibrosis (Fig. 2c, d). Consistently, fibrotic gene expressions and the contents of RA, TG, and CHO in isolated HSCs were reduced by prophylactic but not therapeutic treatment of OCA (Supplementary Fig. 2). Similar results were observed in methionine and choline-deficient diet (MCD)-induced NASH model (Fig. 2e–h, Supplementary Fig. 2). These results indicate that the results collected from prophylactic dosing of OCA that is widely applied in preclinical studies might not be suitable for direct translation to the clinic.
OCA is effective in quiescent but not activated HSCs
Because HSC activation is a hallmark of liver fibrosis, we hypothesized that the differentiated responses of quiescent and activated HSCs to FXR agonists may underlie the discrepancy between prophylactic and therapeutic treatments. To this end, primary HSCs were isolated from vehicle-treated or CCl4-treated mice and cultured in vitro to promote their auto-activation. Significant inhibition of pro-fibrotic gene transcription upon OCA treatment was observed in primary HSCs from vehicle-treated mice, while marginal inhibition was observed in those from CCl4-treated mice (Fig. 3a). OCA prevented LD loss in HSCs from vehicle-treated mice as revealed by analysis of RA, TG, and CHO, and the lipid staining with Bodipy and Nile red, while no obvious effect was observed in cells isolated from CCl4-treated rats (Fig. 3b, c). Similar results were obtained from the study of HSCs of BDL mice. OCA reduced the mRNA levels of pro-fibrotic genes and prevented HSC LD loss from sham-operated mice but not BDL mice (Fig. 3d–f). Consistently, marginal inhibition in pro-fibrotic gene expression as revealed by mRNA levels after OCA treatment was observed in HSC-T6 cells with facilitated activation by TGFβ1 (Supplementary Fig. 3). Together, these results indicate that OCA is effective in quiescent but not pre-activated HSCs.
SUMOylation underlies reduced FXR activity in activated HSCs
Because FXR is a transcriptional factor, we reasoned that its transactivity might be compromised in activated HSCs. The FXR target gene Shp mRNA expression in HSCs from healthy mice was significantly increased after OCA administration, while its induction by OCA was increased but attenuated in CCl4-treated or BDL-treated mice (Fig. 4a). Similar results were obtained from the analysis of other FXR agonists, including GW4064 and WAY-362450 in HSC-T6 cells treated with vehicle or TGFβ1 (Supplementary Fig. 4). In addition, primary human HSCs from healthy donors, were more responsive to OCA stimulation as compared to HSCs from NASH patients (Fig. 4a). These results strongly support that the function of FXR is gradually lost in the process of HSCs activation. We first asked whether the protein levels of FXR in HSCs are reduced as found in hepatocytes23. Surprisingly, the mRNA and protein levels of FXR remained nearly unchanged during the activation of HSCs (Supplementary Fig. 5a, b).
Since transcriptional activities of NRs may also be modulated by post translational modifications (PTMs)40,41, the PTMs of FXR were explored. Results from Co-IP assay showed that FXR SUMOylation was gradually enhanced during the activation of primary HSCs, phosphorylation of FXR was suppressed in highly activated HSCs, while acetylation of FXR remained unchanged (Supplementary Fig. 5c). The molecular masses of FXR and SUMO1 are about 55 and 15 kDa, respectively, and thus the SUMOylated FXR would be detected at about 70 kDa by both anti-FXR and anti-SUMO1 antibodies. Western blot assays were conducted to detect the SUMOylated form of FXR protein. In accordance with the results from Co-IP, SUMOylated FXR was obviously elevated in HSCs from CCl4-treaed and BDL-operated mice than that in HSCs from control mice (Supplementary Fig. 5d). The elevation of FXR SUMOylation in activated HSCs was further validated by use of a SUMOylation ELISA kit (Supplementary Fig. 5e and Fig. 4b). Notably, the SUMOylation of FXR was found significantly higher in primary human HSCs from NASH patients compared to healthy donors (Fig. 4b). Mammals express three SUMO proteins that can be divided into two families, SUMO1 and closely related SUMO2/341,42. In vitro SUMOylation assay demonstrated that both SUMO1 and SUMO2/3 could be attached to recombinant FXR-GFP protein (Supplementary Fig. 5f), in line with previous reports43,44. Overexpression of either SUMO1 or SUMO2 in HSCs significantly enhanced FXR SUMOylation. However, knock-down of Sumo1 but not Sumo2 in HSCs reduced FXR SUMOylation (Supplementary Fig. 5g–j). Furthermore, SUMO1 overexpression resulted in not only increased FXR SUMOylation, but also reduced response to OCA treatment, as demonstrated by analysis of Shp mRNA levels (Fig. 4c, d, Supplementary Fig. 5k). Lys122, Lys275, and Glu277 of FXR had been previously identified as the SUMO consensus sites43. In line with previous reports, single mutation of K122R, K275R, or E277A reduced FXR SUMOylation, while triple mutations of these sites almost completely abolished SUMO conjugation (Supplementary Fig. 5l). Analysis of the transcriptional activity of these mutants by Shp expression also demonstrated that SUMOylation at Lys122, Lys275, and Glu277 of FXR drastically repressed its transactivity (Supplementary Fig. 5m). The triple mutant form of FXR was resistant to SUMOylation-caused loss in transcriptional activity in primary HSCs (Fig. 4e, f, Supplementary Fig. 5n). Together, these results strongly support that SUMOylation is a pivotal factor regulating FXR transactivity.
SUMOylation inhibitor restores FXR activity and function
Because SUMOylation determines FXR transactivation, we supposed that a SUMOylation inhibitor would synergize with FXR agonists in suppressing HSC activation. To this end, a panel of SUMOylation inhibitors were screened. Results from SUMOylation ELISA kit assays, Co-IP assays, and western blot assays demonstrated that both GA and SP could significantly inhibit SUMOylation of FXR (Fig. 5a, b, Supplementary Fig. 5o). OCA efficiently upregulated the FXR target gene Shp in the presence of SP (Fig. 5c). SP treatment also restored FXR transactivation in activated HSCs when exposed to other FXR agonists including GW4064 and WAY-362450 (Supplementary Fig. 6). Additionally, treatment with GA, another SUMOylation inhibitor, also restored FXR transactivation in activated HSCs (Supplementary Fig. 6).
We next tested whether SUMOylation inhibitors could restore FXR function of inhibiting HSC activation. Primary HSCs were isolated and cultured for 4 days in the presence or absence of SP. In culture-activated HSCs, OCA alone was insufficient in increasing the storage of lipids and decreasing the pro-fibrotic biomarkers (Fig. 5d–f). In contrast, a combination of OCA and SP increased the lipid storage and decreased all pro-fibrotic biomarkers (Fig. 5d–f). The enhanced effects against HSC activation were also observed for other FXR agonists combined with SP, as well as the combination of GA and OCA (Supplementary Fig. 6c). Together, these results support that SUMOylation inhibition is capable of restoring FXR activity in activated HSCs and thereby inhibiting the fibrotic change of HSCs.
SUMOylation inhibitor synergizes with OCA against fibrosis
Since SUMOylation inhibitors synergize with FXR agonists in inhibiting HSCs activation, these inhibitors may also potentiate the therapeutic FXR agonist decrease of liver fibrosis. Mice were injected with SP upon after CCl4, and then treated with OCA two weeks later (Fig. 6a). Serum ALT and AST levels were dramatically reduced upon OCA treatment together with SP (Fig. 6b), but not by OCA or SP alone. Histological analysis also demonstrated that co-administration of OCA and SP reduced ECM accumulation and fibrosis development (Fig. 6c). Consistently, the mRNA levels from pro-fibrotic genes were all reduced upon combined treatment with OCA and SP (Fig. 6d). Primary HSCs were isolated to further validate the anti-fibrotic effects of this combination. As expected, SP treatment significantly inhibited FXR SUMOylation in HSCs (Supplementary Fig. 7). In the presence of SP, OCA significantly down-regulated the mRNA levels of pro-fibrotic genes, up-regulated Shp mRNA, and prevented LD loss (Supplementary Fig. 7). Similar results were obtained from BDL-induced fibrosis model. SP treatment significantly synergized with OCA in attenuating liver fibrosis as demonstrated by serum aminotransferases, histological analysis, and the mRNA levels of pro-fibrotic genes (Fig. 6e–h). Moreover, in freshly isolated HSCs, a combination of SP and OCA strongly down-regulated mRNA levels of pro-fibrotic genes, up-regulated Shp mRNA and prevented LD loss (Supplementary Fig. 7).
In the clinic, NASH is a pivotal pathological cause in promoting liver fibrosis and a panel of FXR agonists has been developing for NASH fibrosis. We thus further validate the effects of SP and OCA combination against hepatic fibrosis in NASH models induced by HFHC diet as well as MCD diet. Individual administration of OCA showed marginal effects on serum aminotransferase levels, pathological improvement, and fibrotic gene expressions. In contrast, when combined with SP, OCA significantly reduced serum levels of aminotransferases and improved liver histological features including steatosis, inflammatory infiltration, and ballooning (Fig. 7b, c). Moreover, the combination of OCA and SP reduced ECM accumulation and mRNA levels of pro-fibrotic genes (Fig. 7c, d). In agreement, this combination also reduced mRNA levels of pro-fibrotic genes and restored lipid contents in primary HSCs (Supplementary Fig. 8). As expected, combined OCA and SP treatment also significantly impeded fibrotic development in MCD-induced NASH model (Fig. 7f–h, Supplementary Fig. 8). These results collectively support the view that a combination of SUMOylation inhibitors and FXR agonists could be a promising therapeutic approach to treat liver fibrosis caused by toxin, cholestasis, and NASH.
FXR agonists stabilize lipid droplet via regulating Plin1
We next asked how SUMOylation inhibition, via restoration of FXR function, can synergize with FXR agonists in inhibiting HSCs activation and decreasing fibrosis. HSCs activation was associated with decreased lipid accumulation (Supplementary Fig. 9)45, and thus FXR agonists might inhibit HSCs activation via stabilizing LD. Cultured HSC-T6 cells were loaded with ROH and FAs to promote lipid accumulation and LD formation. As expected, cells loaded with lipids showed decreased αSMA staining and mRNA levels of pro-fibrotic genes (Supplementary Fig. 9), indicating that preventing LD loss may contribute to inhibiting HSC activation. In culture-activated primary HSCs, OCA pre-treatment was able to prevent LD loss (Supplementary Fig. 10). Since LD degradation is largely caused by the loss of LD-associated proteins46,47, the expression profiles of those proteins in culture-activated HSCs was analyzed. Surveying mRNA expression of Plin family members revealed that the expression of Plin1, but not other Plins, could be up-regulated by OCA, as well as two other FXR agonists (Supplementary Fig. 10). Primary HSCs from Fxr−/− mice, in comparison with that from WT mice, were characterized with reduced mRNA levels of both Shp and Plin1, enhanced levels of various pro-fibrotic genes, and decreased lipid content (Supplementary Fig. 10).
Next, the role of Plin1 in LD stabilization and HSC activation was explored. Fresh HSCs were transfected with Ctrl empty or Plin1 expression plasmids 12 h after seeding, and then cultured for 1, 4, or 7 days (Supplementary Fig. 10h). Cellular neutral lipid content analysis showed that Plin1 overexpression increased lipid contents and alleviated their rapid loss (Fig. 8a). Additionally, cells transfected with Plin1 exhibited decreased levels of various pro-fibrotic genes (Fig. 8b). Staining analysis by Bodipy, Nile red, and αSMA further confirmed the role of Plin1 overexpression in LD stabilization and HSC activation (Fig. 8c). Moreover, fresh HSCs were transfected with Plin1 siRNA to further validate its role in LD storage and HSC activation. Fresh HSCs were transfected with respective siRNA 12 h after seeding and then cultured for 1, 4, or 7 days (Supplementary Fig. 10i). As expected, Plin1 deficiency reduced the lipid accumulation and promoted HSC activation (Fig. 8d–f). These results support the view that Plin1 plays a crucial role in maintaining LD stabilization and preventing HSC activation.
The possibility that FXR agonists inhibit HSCs activation via regulating Plin1 was examined using freshly isolated HSCs transfected with scramble or Plin1-specific siRNA and treated with OCA (Supplementary Fig. 10j). OCA treatment significantly increased the storage of lipids in ctrl siRNA-treated cells, while this effect was abolished in Plin1 siRNA-treated cells (Fig. 9a, b). Consistently, the effects in decreasing α-SMA level and other pro-fibrotic biomarkers were also abolished in Plin1 siRNA-treated cells (Fig. 9b, c). Together, these results indicate OCA may inhibit HSC activation via regulating Plin1. OCA is an FXR agonist, and thus the question arises whether Plin1 is a direct FXR target gene. FXR antagonist and siRNA interference largely abolished the effect of OCA in upregulating Plin1 (Fig. 9d). Based on previous studies, most functional binding sites (FXRE) identified in FXR target genes correspond to two inverted repeats spaced by one nucleotide as exemplified by the IR-1. The typical IR-1 elements including GGGTGAATAACCT and GGGTCAGTGACCT48. Analysis of the proximal promoter of rat Plin1 gene identified a potential IR-1 (5′-GTGGCAATCACCT-3′) located 1363–1375 bp upstream of the transcription start site. We then cloned this putative Plin1 gene promoter and evaluated their regulation by FXR. Transactivation of the Plin1 gene promoter by FXR was evaluated by luciferase reporter gene assays. As expected, the Plin1 gene promoter was transactivated by FXR in the presence of OCA (Fig. 9e) in HSC-T6 cells. Considering HSCs during the process of promoter transfection have been already activated and thereby compromising the response to FXR agonists, the reporter gene assays were conducted in HSCs in the presence of SP and in AML-12 cells. As expected, much stronger responses to OCA treatment were observed in these conditions (Fig. 9e). BLI assays were conducted to confirm the recruitment of recombinant FXR protein onto the Plin1 gene promoter. With the increase of FXR protein concentration, the association between FXR protein and the Plin1 gene promoter was enhanced (Fig. 9f). Moreover, this association was obviously reinforced in the presence of OCA and impaired by the mutation in this putative IR-1 region (Fig. 9f), supporting that FXR protein binds to IR-1 located in Plin1 promoter. To further validate this association between endogenous FXR and Plin1 promoter in cells, ChIP assay was further performed to detect whether endogenous FXR binds to Plin1 promoter. As expected, Plin1 was successfully identified from the ChIP assay, and the levels of Plin1 promoter recruited to FXR was significantly elevated following OCA treatment (Fig. 9g). These results support the contention that Plin1 is a direct FXR target gene.
FXR SUMOylation represses Plin1 regulation in activated HSCs
The present results showed that the increased SUMOylation of FXR in activated HSCs is an important causal factor restricting the functional benefits of FXR agonists against HSC activation and thereby liver fibrosis. It is thus reasonable to predict that the effects of FXR agonists on upregulating Plin1 may be compromised by increased FXR SUMOylation. In line with the results of the typical FXR target gene Shp (Fig. 3a), OCA significantly upregulated Plin1 in quiescent but not activated HSCs (Fig. 10a). Similar results were observed in reporter gene assays (Fig. 10b). Additionally, OCA treatment failed to upregulate Plin1 in cells transfected with the SUMO1 expression vector plasmid (Fig. 10c). In contrast, the SUMOylation inhibitor SP significantly restored the upregulation of Plin1 by OCA in activated HSCs (Fig. 10d). Moreover, in freshly isolated HSCs from CCl4-, BDL-, or NASH-induced fibrotic mice, therapeutic administration of OCA could up-regulate Plin1 expression in the presence of SP (Fig. 10e–h). Collectively, these results suggest that the increased FXR SUMOylation may compromise the functional benefits of FXR agonists in activating Plin1 and thereafter the efficacies in inhibiting HSC activation and fibrotic development.