One of the most frequently reported mechanisms of TGR5-mediated liver protection involves its anti-inflammatory properties (Figure 2), and the bulk of related studies in this field is growing [57]. As pointed out by several papers, BAs induce pro-inflammatory effects and this can be considered as crucial for mechanisms of liver injury during cholestasis [7, 12, 58, 59]. During cholestasis, a primary event would be that BAs activate cytokine production in hepatocytes, and neutrophils would secondarily contribute to liver injury [59]. Although the role of macrophages and Kupffer cells during cholestasis remains controversial, the anti-inflammatory action of TGR5 during cholestasis likely occurs at least in part in these cells, downstream an initiating effect of BA in hepatocytes (Figure 1) [5, 45, 46, 60]. In line, early and recent studies reported that TGR5 activation reduces the effect of lipopolysaccharide (LPS) on cytokine gene induction in rat Kupffer cells and mouse macrophages [60, 61]. Consequently, anti-inflammatory effects of TGR5 have been reported as protective in several rodent experimental models (see below) [40, 61–65]. From a signaling perspective in macrophages, TGR5 stimulation blunts NF-κB activation through mechanisms depending on cAMP production, involving an inhibition of Ikappa beta (IκB) phosphorylation and p65 nuclear translocation [61]. Anti-inflammatory effects of TGR5 engagement also involve impact on the AKT-mTOR complex 1 pathway, on CEBP-β induction, and on a complex modulation of the NLRP3 inflammasome activity [66–69]. We previously observed that cytokine induction was exacerbated in TGR5-KO as compared to WT mice after either PH or bile duct ligation (BDL), and that it enhanced cholestasis [70], favored hepatocyte necrosis, and delayed regeneration [16]. This was later confirmed by others [39, 71, 72], and reproduced by us [4, 5] in the BDL model. In line, TGR5-KO mice harbored exacerbated sensitivity to LPS-induced hepatic inflammation, while treatment with a TGR5 agonist improved WT mice [61]. In the more chronic Mdr2-KO mice model of primary sclerosing cholangitis (PSC), specific TGR5 stimulation alone did not improve the phenotype, in contrast with a dual FXR/TGR5 agonist [49]. As a whole in the lack of TGR5, mice harbored an excessive susceptibility to pro-inflammatory stimuli, not only in the liver but also outside this organ [64].

As a whole, TGR5 engagement during cholestasis or after liver injury likely tunes Kupffer cells and macrophage activation, with an impact on hepatocyte protection and proliferation allowing compensation of parenchymal cell loss [73, 74]. In line, it has been reported that TGR5 stimulation stabilized the M1/M2 ratio of macrophage populations towards an anti-inflammatory phenotype [75, 76]. Importantly, translational data suggested that such TGR5-dependent anti-inflammatory responses may also occur in human patients with liver failure, with an impact on clinical outcome [77]. In the same line, it has been recently reported that during PSC in humans, as well as in the Mdr2-KO murine model, TGR5 downregulation in biliary epithelial cells contributed to the pro-inflammatory phenotype observed in cholangiocytes during this disease [78].

During cholestatic liver diseases, especially when obstructive components occur within their pathophysiological course, the biliary epithelium integrity is obviously key to protect parenchymal cells (hepatocytes) from BA-induced injury. This may involve not only the replacement of dead cells but also the strengthening of barrier function [79]. In fact, the “blood-biliary barrier (BBB)” [80], namely the separation between blood and bile, is mainly formed by both epithelial liver cell types, hepatocytes and cholangiocytes, and especially involves their respective TJs that seal adjacent cells (Figure 1). Hepatocytes, due to their physiological position along the biliary path, are obviously frontally exposed to BA, but they share this frontline position in the BBB with the downstream layer of cholangiocytes that shields parenchyma from BA-induced damage. In the biliary epithelium as well as in any lining epithelial sheet, TJs constitute a regulated barrier to the diffusion of molecules in the paracellular space, allowing selective exchanges of small molecules (especially ions) between apical and basolateral microenvironments [81]. TJs are composed of occludin, claudins, and JAMs, which constitute the main core of plasma membrane proteins which associate with cytoplasmic actin-binding proteins [including zonula occludens (ZO) proteins] [81, 82]. In contiguous cells, extracellular domains of the transmembrane proteins interact with their counterparts and seal plasma membranes. Paracellular permeability is obviously a regulated process, although the mechanisms involved still remain poorly explored. It is reported that occludin, JAM-A, and claudin-4 phosphorylation can modulate paracellular permeability, although this aspect has been poorly studied in liver epithelial cells [83, 84]. In the liver, the role of selected TJ proteins expressed in hepatocytes in the regulation of bile secretion has been explored [85–87], in contrast with TJ proteins expressed in cholangiocytes of which pathophysiological impact is not strongly and specifically delineated. It is hypothesized that if inter-cholangiocyte TJs would be altered, bile duct walls would leak, resulting in bile-induced parenchymal injury. However, strong evidence confirming this pathophysiological view is still lacking [88–90], with a few exceptions. Indeed, severe liver disease is observed in patients with neonatal ichtyosis and sclerosing cholangitis (NISCH) syndrome, a primary (mutational) claudin-1 defect [91, 92], or with the more recently described mutations in the TJ-associated protein ZO-2 gene [93]. In line, TJP2 inactivation in both hepatocytes and cholangiocytes in mice resulted in altered biliary homeostasis, with exacerbated cholestatic features upon CA diet [94]. But still, the balanced influence of inter-cholangiocyte and inter-hepatocyte TJ protein defects during these diseases is not defined, in part because TJ protein repertoires differ in hepatocytes and cholangiocytes [95]. In mice also, double inactivation of ZO-1 and ZO-2 genes in hepatocytes resulted in profound alteration in hepatic tissue architecture and functions leading to lethality by 6 weeks of age [96]. During inflammatory cholestatic diseases, hepatic TJ can also be secondarily altered [88, 89, 97] by cytokines and endotoxin [97]. Interestingly, double deletion of β- and γ-catenins genes in hepatocytes in mice resulted in complete disorganization of TJ proteins and structure. Those mice displayed a complete loss of hepatocyte BBB, and as a consequence severe cholestasis-induced parenchymal injury [98]. Crucial impact of TJ integrity for liver recovery after cholestatic injury in the 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC) mouse model has also been reported, pointing out that vaso-intestinal peptide (VIP) secreted by periportal mesenchymal cells and cholangiocyte VIP receptor 1 were involved in a crucial crosstalk for TJP expression in this experimental setting [99]. As a whole, the literature points to a crucial hepatoprotective role of TJ in the liver, although it is better documented in mice than in human patients, and more specifically addressed in hepatocytes than in cholangiocytes.

The regulatory role of BA on paracellular permeability is mostly known as deleterious, because BAs enhance paracellular passage in intestinal [100] and respiratory epithelial cells [101], but signal transduction mechanisms and in particular the BA receptors likely involved were not reported. However, other reports established that TGR5-dependent BA signaling reinforces barriers in several murine tissues, i.e. intestinal, endothelial, retinal, and blood-brain barriers [102–105]. Presumably, BAs exert their effects on paracellular permeability through multiple pathways, possibly enhancing or reducing it as a function of the pathophysiological context and depending on the interested cell types. As the biliary tract is naturally exposed to high concentrations of BA, searching for an impact of BA signaling on cholangiocyte paracellular permeability in physiological and cholestatic conditions is thus highly relevant. In a recent study, we found that BA-induced TGR5 engagement protected the liver parenchyme against BA overload during cholestasis through a reduction of paracellular permeability [5]. More precisely TGR5 activation induces the PKCζ-dependent phosphorylation of the TJ protein JAM-A, which results in a protective modulation of the TJ barrier function (Figure 1). This was shown in two mice experimental models of cholestasis [BDL and 1 naphthyl isothiocyanate (ANIT) intoxication], and importantly, we provided data from patients with primary biliary cirrhosis (PBC) and PSC suggesting that the TJ protein JAM-A may be protective in the biliary epithelium during human cholestatic diseases, rising the therapeutic challenge of targeting this protein and TGR5 related pathways in the future.

Although scarce literature is available, it is suggested that the impact of TGR5 on bile composition operates mainly through cholangiocyte-dependent processes (Figure 2). As evoked above, the TGR5 effect on both CFTR and the anion exchanger 2 (AE2) mediates chloride/bicarbonate exchange across the apical membrane, promoting the formation of the so-called “bicarbonate umbrella”. Consequently, the alkaline environment enables the reduction of BA protonation and thus protects the cholangiocytes (bile ducts) and hepatocytes (liver parenchyma) from BA cytotoxicity in liver injury and cholestatic contexts [49, 106].

This might constitute an adaptive mechanism enhancing bile secretion and fluidity through which TGR5 would protect the liver from BA toxicity during repair [16] or cholestasis. TGR5 is also expressed on cholangiocyte cilia and may regulate bile flow through an effect on this organelle [107, 108], via a coupling with HCO₃^– secretion in bile [109]. Finally, TGR5 may modulate water handling in the biliary epithelium, although no direct evidence has been reported yet, while further studies would be necessary to eventually extend data published in kidney epithelial cells [110, 111].

It is also emphasized that the BA pool composition itself including its hydrophobicity is crucial to liver homeostasis [112, 113]. Excessive BA pool hydrophobicity is indeed reported to be noxious for liver repair and cholestatic contexts [56, 114–118]. This is best exemplified in the BSEP/abcb11^–/– mice, which exhibit non-progressive mild cholestasis [119], contrasting with the severe human progressive familial intrahepatic cholestasis type 2 (PFIC2, severe progressive cholestasis in children) phenotype. This discrepancy is, at least in part, due to the less hydrophobic (less toxic) BA pool observed in Bsep^–/– mice, as it is enriched in hyperhydroxylated BA [120]. These BA pool features are viewed as protective against cholestatic liver injury not only in mice [119], but also in children with progressive intrahepatic cholestasis [121]. Nice proof of concept of this protection has been reported in double Mdr2^–/– and Bsep^–/– mice, in which hyper-hydrophilic BA prevented liver damage caused by the lack of biliary phospholipids [122]. Recent studies also pointed out that an efficient activation of the BA receptors (entirely dependent on the BA pool composition) had a strong impact on alcoholic and non-alcoholic liver diseases [65, 123]. In line, Cyp8b1 inhibition, reducing BA pool hydrophobicity, prevented steatosis in high-fat diet-fed mice [124]. It has also been emphasized in a murine NASH preclinical model that the BA pool composition was dramatically altered, with poor signaling power on both FXR and TGR5, and that DCA diet complementation prevented liver disorders in this model [125]. Recently, the BA metabolizing enzyme Cyp2c70 which converts CDCA (a hydrophobic primary BA) into the more hydrophilic muricholic acid (MCA), has been identified as a master regulator of the BA pool composition in mice [126]. This enzyme is lacking in humans, explaining in part why the BA pool is more hydrophobic in humans than in mice. Interestingly, Cyp2c70^–/– mice exhibit a human-like more hydrophobic BA pool as compared with control animals, associated with liver inflammation [127] and altered FXR signaling [128]. More recently, it was reported that a cholangiopathy with biliary fibrosis developed in female Cyp2c70^–/– mice, a phenotype reversed upon UDCA treatment [129], reinforcing the fact that BA pool hydrophobicity had a crucial pathophysiological impact on liver disease outcome.

Others and we reported that BA pool composition was more hydrophobic in TGR5-KO than in WT mice [4, 16 , 47, 48, 130], as measured in bile, plasma, liver, and feces [4, 16]. More precisely, MCA the most hydrophilic primary BA in mice, and the MCA/CA ratio, were strongly reduced, whereas secondary (more hydrophobic) BAs were significantly overrepresented in TGR5-KO as compared with WT mice [16]. Our recent study explored different mechanistic hypotheses to support this phenotype, linking TGR5 with BA pool composition [4]. First, we did not find any significant direct impact of TGR5 on BA synthesis. Second, based on gut microbiota transfer experiments, we provided data showing that despite the lack of TGR5 was associated with an intestinal dysbiosis, this latest was not responsible for the observed more hydrophobic BA pool. This is in contrast with data indicating that FXR-mediated modification of the gut microbiota resulted in reshaping the BA pool (LCA enrichment) towards TGR5 stimulation and improved glucose tolerance through intestinal glucagon-like peptide-1 (GLP-1) secretion [131]. We finally focused on TGR5-mediated impact on the GB, as it is reported as the tissue displaying the highest TGR5 expression, at least in mice [132], suggesting that this receptor may have a physiological impact on GB functions, including GB relaxation as reported [48] and potentially BA reabsorption also known as the cholecysto-hepatic shunt [3]. This shunt is expected to allow a short circuit by which BAs from the GB return directly to the liver via the portal circulation without passing through the intestine [3], thereby restricting the amount of toxic secondary BA in the pool. Importantly, we found that in the lack of TGR5, defective GB dilatation resulted in reduced cholecysto-hepatic shunt, building a more hydrophobic BA pool. We also put forward TGR5-dependent hepato-protective properties of GB dilatation in the setting of obstructive cholestasis [4]. We showed that after cholecystectomy (as compared with spared GB), BDL mice had more severe necrotico-inflammatory liver injury, more bile duct dilatation, and reduced survival rate. Importantly this phenotype was not found in TGR5-KO mice, suggesting that TGR5-dependent GB dilatation was crucial in hepatoprotection upon obstructive cholestasis [4].