An Integrated Pathogenetic Model of Primary Biliary Cholangitis

Abstract

The pathogenesis of primary biliary cholangitis (PBC) is not fully understood. Despite recent progress, many aspects require further clarification. Thus, PBC is regarded as an autoimmune disease, but immunosuppressive treatment, which is effective in other autoimmune diseases, is not working in the case of PBC. Moreover, there are controversies over the pathogenetic role of anti-mitochondrial antibodies as mitochondria are present in all cells but only cholangiocytes are damaged. In this review, all the proposed models and factors that have been involved in the pathogenesis of PBC are presented. They include mechanisms such as dysregulated autophagy, senescence, apoptosis, impairment of the protective bicarbonate umbrella, immunological abnormalities, the dysbiosis of gut microbiota, and the role of bile acids. Genetics of PBC and epigenetic transcriptional modifications are also presented. Data supporting molecular mimicry and the viral etiology of PBC are analyzed. Finally, an integrated model is proposed based on interactions of the factors that may participate in PBC pathogenesis. Therefore, the purpose of this review is to provide a unifying presentation of the various aspects of PBC pathophysiology, which will allow for a better understanding of this multifaceted disease. New treatment targets may also be identified in such a holistic model.

1. Introduction

Primary biliary cholangitis (PBC) is a cholestatic disease characterized by the gradual destruction of small intrahepatic biliary ductules and biliary epithelial cells (BECs) associated with the appearance of anti-mitochondrial antibodies (AMAs) that may progress to cirrhosis. However, a uniform, all-inclusive definition of the word disease that could be universally accepted is still lacking. A recent proposition has been accepted as the Disease Signature, with the inclusion of causes and symptoms and occasionally treatment responses [1].

In that respect, PBC can hardly be characterized as a disease. Hence, the continuous change in definitions from intrahepatic non-suppurative destructive cholangitis to primary biliary cirrhosis and the contemporary primary biliary cholangitis.

There is a global variation in disease epidemiology across Europe, North America, and the Asia-Pacific region according to data collected over the last 20 years. Both the incidence and prevalence of PBC increase worldwide with an incidence rate of 0.23–5.31/100,000 and a prevalence of 1.91–40.2/100,000 [2,3,4]. Epidemiological data are still incomplete in China. Estimations report that the prevalence of PBC in China is 20.5/100,000, which is second only to Japan [5]. Despite the increased prevalence, the mortality is decreased, possibly as a result of better treatment [6,7]. Interestingly, a geographical clustering of the disease is observed in North America and Europe. In the north of England, a high spatial distribution of cases was reported, with clusters in urban areas (up to 13 cases/km²). The investigators attributed their results to environmental factors [8]. Geographic clustering was observed in First Nations of British Columbia, where the disease was as high as one in four within generations of several families [9,10]. Researchers also identified clusters of high prevalence in the genetically homogeneous population of the island of Crete. Differences were found between the eastern and western parts of the island, with high prevalence in the east (122 patients per 50 km²) and low prevalence in the west (11 patients per 50 km²) [11]. Indigenous Canadians have a significantly higher prevalence and more severe disease compared with Canadians of European ancestry [12].

The diagnosis of PBC must satisfy two out of the three criteria: (a) Presence of cholestatic biochemical markers, (b) presence of anti-mitochondrial antibodies (AMAs), or other PBC-specific nuclear antibodies, such as sp100 or sp210, in AMA-negative cases, and (c) compatible histology in liver biopsy [13]. AMAs recognize the pyruvate dehydrogenase E2 protein (PDC-E2) of the inner mitochondrial membrane. It has been suggested that the abnormal expression of PDC-E2 and the subsequent production of AMAs are the initiating events of autoimmunity [14]. Anti-sp100 and anti-gp210 are specific for the disease and correlate with clinical severity [15]. An interplay of genetic and environmental factors is implicated as the driving force for the breach of tolerance and autoimmune-mediated cellular death that follows. However, many gaps in PBC pathogenesis require explanation. Mitochondria are present in every cell, but only BECs are injured in PBC. In addition, AMAs have diagnostic but not prognostic value. AMA titers and subtypes do not correlate with disease severity and prognosis [16,17]. Moreover, although corticosteroid administration or immunosuppression attenuates most autoimmune diseases, inhibiting the autoimmune attack, PBC does not respond to these drugs in similarity with type 1 diabetes mellitus and Grave’s disease. It is possible though that in the case of diabetes, this is only true for advanced stages, when most insulin-producing cells have been destroyed [7,18]. In contrast, PBC patients respond to treatment with a bile acid such as ursodeoxycholic acid and a nuclear receptor agonist such as obeticholic acid [2,6]. In other autoimmune diseases, cytokine production plays a major role in pathogenesis. Treatments targeting cytokines have failed in PBC, despite genome-wide association study (GWAS) results [19]. Most importantly, the Witebsky criteria of autoimmune etiology have not been fulfilled [20]. PBC has not been completely reproduced in animal models with the administration of AMA or autoreactive T cells. PDC-E2 and adjuvant infusion generates AMAs but no biliary pathology. AMAs are not necessary in the development of PBC as AMA-negative PBC is responsible for 5–10% of PBC cases [21] with similar clinical and biochemical courses. Moreover, some AMA-positive individuals do not develop PBC. Because of the multiplicity of environmental, ethnic, and genetic factors of unknown significance that are implicated in the pathogenesis of PBC, the autoimmune etiology has been widely accepted. However, the evidence to support this theory is inadequate [22]. In this review, therefore, the different pathogenetic proposals will be analyzed and an integrative model encompassing most of the available data will be presented.

2. Genetics and Epigenetics

Monozygotic twins have an increased frequency of the disease compared to dizygotic twins. PBC is also more frequent in other family members of patients, suggesting that genetic factors are implicated in the pathogenesis of the disease. However, not all people with a genetic predisposition will eventually proceed to PBC, indicating that genetics alone are not enough for induction of the disease. Nonetheless, genes encoding elements of both the innate and adaptive immunity, such as complement deficiency-specific HLA class II alleles, and the IL-12 cytokine have been associated with PBC [23]. Additional evidence that implicates genetics in PBC pathogenesis are the findings in first-degree relatives (FDRs) of patients with PBC. In the Cretan population in Greece, FDRs of PBC patients had a high AMA prevalence independently associated with past urinary tract infections. However, PBC-specific ANAs were not identified in FDRs [24]. In another study, lipid abnormalities and autoimmune disorders were frequent not only in patients as expected but also in their FDRs [25].

The most frequent genetic associations in PBC are loci of the major histocompatibility complex (MHC) [26]. More than 40 HLA-associated variants have been identified so far, and many more are under investigation. PBC is linked to several risk haplotypes such as DRB1*08:01-DQA1*04:01-DQB1*04:02 and DRB1*04:04-DQB1*03:02 in populations in Europe. In the same populations, many protective haplotypes such as DRB1*11:01-DQA1*05:01-DQB1*03:01 and DRB1*15:01-DQA1*01:02-DQB1*0602 were identified [27]. Details of HLA associations with PBC can be found in a recent review [28]. X monosomy was also found in female PBC patients, mostly in B and T cells, a finding that may partly explain the female dominance in PBC [29]. Research has also linked 44 genetic variants outside the MHC region to PBC. Most alleles implicated in PBC have been identified from GWASs [30,31,32,33,34,35,36]. An extensive description of the significance of each of these variants has recently been presented [37]. In the most recent multi-national GWAS, genetic information from 10,516 people with PBC and 20,772 healthy individuals were analyzed. The researchers identified 56 significant loci (20 novel) including 46 in European, 13 in Asian, and 41 in combined cohorts. Candidate genes at newly identified loci include FCRL3, INAVA, PRDM1, IRF7, CCR6, CD226, and IL12RB1, which are involved in immune responses. The importance of pathways such as the pattern recognition receptors (PRRs), TNFa signaling, and differentiation of T helper (Th1 and Th17) cells were recognized as a result of this genetic analysis [38]. However, GWASs have problems. Detection of rare and ultra-rare variants with small effects is impossible with GWASs [39]. This may explain the so-called missing heritability in PBC. Thus, a report from China established an association of 10 variants with PBC in three PBC families localized in the HLA-DRB1 gene not previously detected by GWASs [40]. Another Chinese study demonstrated that the transcription factor Myocyte Enhancer Factor 2D (MEF2D) and the DNA repair gene Poly (ADP-Ribose) Polymerase 2 (PARP2) are genes implicated in PBC pathogenesis. These findings were based on sequencing of 90 subjects, including 30 PBC cases [41]. A different problem has also emerged. IL-12 was upheld as a risk locus for PBC in GWASs [42,43]. However, a proof-of-concept report demonstrated that treatment of PBC patients with the IL-12 monoclonal antibody ustekinumab failed to reach the primary target set in the study [44].

Epigenetics

The word epigenetics was first used by Waddington to summarize the links between gene and protein expression [45]. Today, epigenetics is generally defined as the study of changes in gene function that are mitotically and/or meiotically heritable without alterations in the DNA sequence [46,47].

Increasing evidence indicates that epigenetic changes are implicated in the pathogenesis of PBC. Epigenetic modifications include methylation of CpG DNA regions, post-translational modifications of histone proteins, and non-coding RNAs. In addition, telomere dysregulation in BECs may participate in the induction of the disease [48,49,50].

GWAS analysis of peripheral blood mononuclear cells in PBC identified constant methylation profile changes in 60 gene regions. Interestingly, hypermethylation occurred only in diseased twins, and not in non-affected twins. Researchers identified 51 genes in the X chromosome [51]. High CpG methylation in the promoter regions of AE2α, AE2b1, and AE2b2 in PBC livers was reported as associated with a reduction in AE2 messenger RNA levels [52].

Hypermethylation within gene promoters is always associated with gene silencing, while hypermethylation within intragenic regions is mostly related to gene hyper-expression [53]. This is true for PBC as well. On the other hand, a significantly decreased methylation of the CD40L promoter in CD4+ T cells of PBC patients was reported, leading to hyper-expression of CD40L mRNA. It was also reported that Immunoglobulin M levels in serum were inversely correlated with promoter methylation [54]. Hypermethylation on chromosome X was the most common finding in methylation patterns in monozygotic twins discordant for PBC [49,55]. A recent study demonstrated that the imbalance on the Treg/Th17 axis observed in PBC was influenced by FoxP3 hypermethylation. It was also shown that FoxP3 demethylation in a murine model of PBC restored the Treg/Th17 balance and attenuated liver lesions and inflammation [56].

MicroRNAs (miRNAs) are the most studied epigenetic modifiers in humans. Researchers found that 35 out of 377 miRNAs examined were differentially expressed in PBC livers compared to normal livers. These miRNAs target genes of apoptosis, oxidative stress, and inflammation, which participate in the induction and progression of PBC [57]. The expression of certain co-stimulatory molecules such as CD86 and CD80 in peripheral antigen-presenting cells (APCs) is regulated by plasma-derived exosomal miR-451a and miR-642a-3p, which are upregulated in PBC patients [58]. Among all miRNAs studied so far, miR-506-3p (miR-506) has attracted attention as a modifier of AE2 expression, which is complimentary to a human AE2-UTR-3’ mRNA sequence. Binding of mir506 resulted in downregulation of the Na-independent Cl/HCO3 exchange activity and reduced secretin-stimulated secretion. The miR-506/AE2 axis is therefore involved in the biliary bicarbonate secretory system in PBC [59,60]. MiR-506 was upregulated in PBC livers compared to normal livers [59]. Most importantly, it was specifically concentrated in BECs compared not only to healthy controls but also compared to primary sclerosing cholangitis livers [59,61]. Another miR-506 target relevant to PBC pathogenesis is inositol 1,4,5-trisphosphate receptor 3 (InsP3R3), which is a calcium (Ca2+) release channel at the endoplasmic reticulum (ER) membrane of cholangiocytes. InsP3R3 mRNA contains two miR-506 binding sites, which modify its expression [59]. Increased intracellular InsP3 levels induced by acetylcholine in cholangiocytes lead to increased cytoplasmatic Ca2+ levels, which results in bicarbonate secretion via AE2 [62,63]. In PBC, InsP3R3 expression is reduced in cholangiocytes [64], leading to reduced biliary bicarbonate secretion [64].

Other miRNAs acting on BECs are possibly implicated in PBC. In particular, miR-26a targets protein EZH2, a protein known to be involved in BECs’ senescence in PBC [65,66]. MiR-21 was also significantly upregulated in PBC livers, participating in necroptosis of BECs [67,68]. Repression of miR-425 in CD4+T cells of PBC patients markedly induced pro-inflammatory cytokines such as IL-2 and IFN-γ through N-Ras upregulation in the TCR signaling pathway [69].

Apart from miRNAs, long non-coding RNAs are important epigenetic modifiers. Thus, exosomal lncRNA H19 induces hepatic stellate cell (HSC) activation, bile duct proliferation, and hepatic fibrosis in cholestasis [70,71]. Detailed descriptions of the role of ncRNAs in PBC have recently been published [72].

Histone modifications in PBC have not been adequately studied. However, β-arrestin 1, a protein that increases histone acetylation in CD4+T cells, was upregulated in T lymphocytes from patients with PBC and correlated with disease severity. Interestingly, β-arrestin1 overexpression also modifies the expression of genes implicated in immune response, such as IL-17 and IFN-γ, which are deregulated in patients with PBC [73].

Epigenetics may explain why much of the heritability of PBC is still unaccountable. Missing heritability may be due to the fact that several epigenetic modifications cannot be detected in GWASs. Thus, polymorphisms of solute carrier family 4 member 2(SLC4A2)/anion exchanger 2 (AE2) genes were correlated with disease severity in a case–control study but not in GWASs [49].

Detailed presentations of miRNAs, DNA methylation, and histone modifications in PBC were recently published [72,74].

3. The Autophagy Theory

Autophagy is the recycling mechanism of the cell to preserve cellular homeostasis. Proteins, lipids, damaged organelles, and pathogens are degraded in the lysosomes and the molecules produced are re-used [75]. The significance of autophagy research was recognized by two Nobel Prizes, one to Christian De Duve, for the introduction of lysosomes in biology, and the second to Yoshinori Ohsumi, for his pioneer working on the mechanisms of autophagy [76]. Autophagy is in fact a series of phosphorylations and dephosphorylations. The key regulators of autophagy are three kinases, the mammalian target of rapamycin (mTOR), the Unc-51-like autophagy-activating kinase (ULK1), and the AMP-dependent protein kinase (AMPK) [77].

There are convincing data to suggest that impaired autophagy is a critical factor in the pathogenesis of PBC [78]. Initial reports showed that damaged BECs highly expressed the autophagy marker light chain 3β (LC3B) in association with the autophagy protein p62/sequestosome-1, suggesting the impairment of autophagy [79,80,81,82]. Initially, this finding was believed to represent augmented autophagy [83], but it proved to be an abnormal sequestration of autophagosomes due to autophagy dysregulation. Dysregulated autophagy may be implicated in the abnormal expression of mitochondrial antigens and also in BECs’ senescence in PBC [66,78,79,80,83,84,85,86,87,88,89] as the expression of PDC-E2 was co-localized with LC3B [86]. Importantly, autophagy is also involved in the intracellular antigen processing required for the association of the antigen with the MHC I and II molecules [90,91,92] and subsequent presentation to APC cells. This could also lead to a direct antigen recognition by PDC-E2-specific cytotoxic T cells [81,93]. The co-localization of the mitochondrial protein PDC-E2 with the autophagy protein LC3B indicated that in addition to the mainstream autophagy, mitophagy is also implicated. Mitochondria-derived vesicles (MDVs) facilitate mitochondrial antigen presentation as autoantigens [94].

Autophagy is also implicated in the fibrotic stages of advanced PBC. However, the end result depends on the individual liver cells involved. Autophagy inhibits fibrosis acting in hepatocytes, macrophages, and liver sinusoidal endothelial cells (LSECs). Autophagy decreases the production of pro-inflammatory cytokines by macrophages and LSECs, limiting hepatocyte injury. On the contrary, autophagy in HSCs increases fibrogenesis by increasing lipophagy, thus providing the energy required for HSC activation. Moreover, the p62 loss induced by autophagy leads to impairment of the vitamin D receptor/retinoid X receptor (VDR/RXR) complex, which is vital for the preservation of HSC quiescence [95,96].

It should be noted that ROS produced by toxic BAs may inhibit the activation of ATG4 protease, an enzyme critical in the regulation of autophagy. The ATG4 protease splices LC3/ATG8 to create cytoplasmic LC3-I, which binds to phosphatidylethanolamine (PE). The lipid version of LC3 (LC3-II) binds to the autophagosomal membrane, inducing autophagy. Another central inducer of autophagy, the 5′ AMP-activated protein kinase (AMPK) pathway, is activated by the hypoxia-inducible factor (HIF) under hypoxia, and autophagy is also initiated [97]. It was recently reported that autophagy is dysregulated in human cholestasis because hydrophobic bile acids induce Rubicon, followed by repression of the fusion of autophagosomes with lysosomes and inhibition of the final stage of autophagy, that of lysosomal degradation. Rubicon was upregulated after the administration of obeticholic acid (OCA), while inhibition of Rubicon reversed the dysregulation of autophagy [98].

4. The Senescence Theory

Cellular senescence implies that the ability of a cell to proliferate is being arrested in the G1/S phase of the cell cycle. Senescent cells do not react to external stimuli, but they continue to be metabolically active. Senescent BECs secrete senescence-associated secretory phenotypes (SASPs) [99,100] participating in PBC pathogenesis. SASPs consist of a variety of secreted molecules including cytokines such as IL-6 and IL-8, chemokines such as CCL2, plasminogen activator inhibitor 1, and matrix metalloproteinases (MMPs) [101,102]. They induce and recruit CD4+ T helper (Th) cells [42]. Their composition depends on the causative factor and the cell of origin. Additionally, SASPs are implicated in the regulation of the immune response, and they are regulated by many transcription factors such as C/EBPβ, NF-κB, and GATA4 [103]. SASPs also influence the progression of fibrosis in cholangiopathies [78,104].

Cellular senescence is demonstrated in BECs of the injured bile ductules observed in PBC [79,85]. Senescent BECs secrete chemotactic and inflammatory cytokines [101,102]. It is suggested that SASP cytokines/chemokines such as CCL2 and CX3CL facilitate cellular senescence in neighboring BECs, leading to the accumulation of senescent BECs in the ductules [105,106]. They also facilitate senescence of mesenchymal cells around the ductules promoting fibrosis [87,107]. Moreover, C-X-C motif chemokine ligand-11 (CXCL-11) and CCL-20 from senescent BECs were markers of the response status after UDCA treatment of patients [108].

Senescence is initiated by the activation of the N-Ras pathway in persistent cellular injury [109]. The exact reason for increased senescence in BECs is not clear. The activity of Cyclin-dependent kinase inhibitor p21 (p21WAF1/Cip1) and Cyclin-dependent kinase inhibitor 2A (p16INK4) initiates the senescence of BECs in PBC [79,84]. Oxidative stress reduces the expression of the gene bmi1, which is a suppressor of p16INK4 and may be implicated in the induction of cellular senescence of BECs in PBC [66].

A study of cholangiocytes in which senescence was induced with glycochenodeoxycholic acid (GCDC) revealed that interferon-induced tetrapeptide repeat 3 (IFIT3), which is implicated in antiviral innate immunity, was among the upregulated genes. Knockdown of IFIT3 in cholangiocyte senescence led to senescence markers p16 and p21, associated with increased apoptosis of BECs. Increased IFIT3 was also demonstrated in bile ductules of PBC patients [110]. The transcription factor ETS proto-oncogene 1 (ETS1) was also reported to increase p16 expression and promoted cholangiocyte senescence [111]. The Hippo-yes-associated protein (YAP) activity was markedly reduced in senescent BECs. BECs’ senescence and apoptosis were increased and the proliferation activity was decreased with a deletion of YAP in BECs [112].

The degree of senescent BECs in small bile ducts and bile ductules is positively correlated with the stage and activity of PBC, while the increased expression of the senescent marker p16 INK4a in bile ductules was associated with a reduced response to UDCA [113]. In line with the finding of senescence transmission to cellular bystanders, a recent observation may add evidence of the development of fibrosis in PBC. Induction of senescence to hepatocytes, and overproduction of PDGFa by them, may activate human HSCs, indicating the direct implications of senescent hepatocytes in PBC fibrosis [114].

Interestingly, senescent cholangiocytes identified in a murine model of early PBC were linked to the overactivation of the secretin/secretin receptor (Sct/SR) pathway. SASPs from these senescent cholangiocytes activated Kupffer cells and HSCs in a paracrine manner, leading to local inflammation and liver fibrosis [115]. Interestingly, experimental evidence has shown that although BECs’ increased senescence is reversed to normal after the deletion of SR, the opposite happens in HSCs, where SR deletion increases their senescence [116].

Senescence of BECs is mostly identified in the advanced stage of PBC, although it was also found in early human PBC. Therefore, it is reasonable that autophagy precedes the senescence of BECs [117].

Another factor that favors senescence is the ductular reaction, which is the appearance of immature biliary epithelial cells. This is an ominous prognostic sign [118]. The ductular reaction, apart for its involvement in BECs’ senescence, participates in the activation of HSCs as well [119].

Abnormalities of lysosomal enzymes were recently reported in patients with PBC. Cathepsin B1 was significantly increased in early PBC and reduced in later stages. Cathepsin D was also very high in early PBC and less so in later stages compared to normal controls. The high levels of lysosomal enzymes in early PBC are in accordance with increased senescence as SASPs contain high levels of lysosomal hydrolases. Interestingly, treatment with UDCA restored the abnormal values [120].

5. Apoptosis in PBC

In addition to senescence, BECs’ apoptosis is also involved in the pathogenesis of PBC [121,122,123]. Apoptosis is a mechanism of programmed cell death without the release of intracellular elements. Apoptosis is regulated by a series of pro-and anti-apoptotic proteins and a series of caspases sequentially activated. After an initial signal, the initiator caspases 8 and 9 are activated, initiating the activation of the executioner caspases 3, 6, and 7. Two apoptotic pathways are operative: the intrinsic pathway, induced by intracellular triggers such as oxidative stress of mitochondria or the unfolded protein response (UPR), and the extrinsic pathway, triggered by the interaction between death receptors and their ligands, such as TNFα, FasL, and TRAIL (tumor necrosis factor-related apoptosis-inducing ligand). Hypoxia also induces apoptosis through the intrinsic pathway [124,125].

Apoptosis of BECs in PBC was reported by many investigators. Increased deoxyribonucleic acid (DNA) fragmentation and high levels of perforin, granzyme B, and TNF-related apoptosis-inducing ligand (TRAIL) have been shown in BECs from PBC cases [126,127,128]. Increased BEC apoptosis has been also reported in PBC as opposed to other cholestatic diseases with similar degrees of inflammation [122,123,127,128]. Toxic bile acids (BAs) involved in the pathogenesis of PBC can induce mitochondrial dysfunction through oxidative stress. Liberated mitochondrial cytochrome c activates caspase 9, which activates the executioner caspases of apoptosis in BECs [129].

Apoptosis may explain the presence of AMAs in the serum of PBC patients. Immunologically active PDC-E2 was found in cholangiocytes after apoptosis, which was resistant to cleavage by caspases [130]. Localization of PDC-E2 in the apoptotic bodies allows for its recognition by immune cells and production of AMAs. PDC-E2 was translocated in the surface of apoptotic blebs from human BECs, exposing an epitope that is not present in other cells. Furthermore, apoptotic bodies released from BECs in combination with AMAs provoked the production and secretion of many pro-inflammatory cytokines from PBC macrophages [131,132]. The loss of recognition of this epitope in other cells under normal conditions is due to a modification of a PDC-E2 sulfhydryl group by glutathione. Upregulation of Bcl-2, which appears to inhibit protein glutathiolation or the reduction in glutathione, reversed the loss, suggesting that glutathiolation of PDC-E2 is the reason for loss of recognition [130]. There are more reports on the participation of apoptosis in the pathogenesis of PBC. High hepatic levels of activated caspases 3 and 8 were found in PBC, suggesting the augmentation of apoptosis. Deletion of caspase 8 ameliorated apoptosis [133]. Pan-caspase inhibitors in the bile duct ligated cholestatic model attenuated apoptosis and HSC activation and increased survival [134]. Other mitochondrial antigens, including the E2 subunit of the oxo-glutarate dehydrogenase complex (OGCD-E2) and the E2 subunit of the branched chain 2-oxo acid dehydrogenase complex (BCOADC-E2), were also present in apoptotic bodies from BECs incubated with toxic bile acids [135]. The significance of apoptosis in cholangiopathies was recently shown in studies on ductular reactive cells. In a murine model, deletion of both TRAIL and multidrug resistance 2 (MDR2) induced an excessive ductular reaction and promoted fibrosis [136]. Taken together, these studies demonstrate the significance of apoptosis in PBC.

Apoptosis is an early event in PBC, as demonstrated in a mouse autoimmune cholangitis (AIC) model. BECs in AIC mice overexpress TNFa, several chemokines such as CXCL9 and CXCL10, and the toll-like receptor 2 (TLR2). Activation of TLR2 enhanced apoptosis and CXCL10 production. Interestingly, this was associated with impaired gut microbiota. Dysbiosis with abundant Firmicutes initiated the apoptosis of BECs via TLR2 signaling [137]. Apoptosis is possibly a target of action of UDCA in PBC treatment. UDCA protects BECs from apoptosis induced by hydrophobic bile acids through several pro-survival mechanisms [138]. High levels of CXCL9 and CXCL10 were found in the liver tissue and serum of PBC patients. Treatment with UDCA significantly reduced serum levels of these chemokines [139].

6. The Bicarbonate Umbrella Theory

BECs secrete bicarbonate (HCO3-) through the anion exchange 2 (AE2) and generate the so-called luminal alkaline “umbrella” [140]. The important role of the chloride/bicarbonate exchanger AE2 was realized after the report that early-stage PBC patients had reduced expression of the SLC4A2 gene in BECs [141]. SLC4A2 encodes for AE2. A reduction in SLC4A2 caused alkalization of cytosol, followed by the induction of soluble adenylyl cyclase (sAC) expression [142]. BECs became more sensitive toward toxic bile salts such as GCDC [143]. The activity of sAC is regulated by increased intracellular bicarbonate concentrations. Therefore, sAC is a very sensitive pH sensor, able to translate even minute pH alterations into the regulation of activity of cellular proteins [144]. AE2 is localized at the apex of BECs and is the main biliary bicarbonate extruder [145,146,147]. An intact glycocalyx at the apex of BECs and the normal intracellular sAC are required to maintain the biliary umbrella in addition to the normal function of AE2 [142,148,149].

The biliary epithelium has a dual defense system consisting of bicarbonate secretion and the maintenance of a tight barrier. Passive diffusion of bile acid conjugates inside BECs depends on pH. Decreased AE2 activity reduces bicarbonate secretion, leading to cytosolic accumulation of bicarbonate followed by increased sAC activity and increased entrance of bile acids into BECs, which further increases sAC activity. This in turn modulates bile-salt-induced apoptosis (BSIA). Inhibition of sAC corrects sensitization to BAs and completely prevents BSIA [143,150,151]. Another mediator of the bicarbonate umbrella is the secretin (Sct)/secretin receptor (SR) axis, which modulates both the cystic fibrosis transmembrane receptor (CFTR) and AE2, increasing choleresis [152].

AE2 expression was markedly decreased in liver biopsies and peripheral blood mononuclear cells in PBC patients [153,154]. Furthermore, cultured BECs from PBC patients had decreased AE2 activity compared to normal human BECs [155]. GCDC decreased AE2 activity in BECs through upregulation of ROS production, which enhanced the senescence of BEC. Moreover, the downregulation of AE2 levels by either GCDC or an AE2 inhibitor led to increased secretion of IL-6, IL-8, and CXCL10 from BECs in response to toll-like receptor ligands [156]. Furthermore, studies using labeled HCO3- showed that PBC patients had defective secretin-stimulated biliary bicarbonate secretion, which was restored after UDCA treatment [157]. Inhibition of Sct/SR function in a late-stage murine PBC model caused dysregulation of the bicarbonate umbrella and decreased mucin production by BECs. Sct treatment reversed the cholangiocyte secretory function, augmented bicarbonate and mucin secretion, and ameliorated liver fibrosis [152]. The AE2-sAC axis might additionally be implicated in the impaired immune response in PBC. Interestingly, sAC is localized in both activated CD4+ T cells and CD8+ T cells [158]. The murine model Ae2a, b-/- spontaneously develops several PBC-like features, such as portal infiltration by CD4+ and CD8+ T cells and bile duct injury [159]. In this model, CD8+ T cells show signs of increased proliferation and activation [160]. CD8+ T cells show suppression of the programmed cell death-1 (PD-1) checkpoint expression, followed by decreased apoptosis, and increased immune-mediated cholangitis [161]. Importantly, AE2 downregulation is associated with deranged autophagy and BECs’ senescence, which are characteristics of PBC, as mentioned above [162]. These data support the hypothesis that decreased AE2 expression in PBC is followed by a defective umbrella, leading to BECs’ damage and secondary breach of tolerance and autoimmune responses against BECs [60]. Lysosomes are more acidic in females than in males. Therefore, reduced AE2 activity probably impairs mitophagy more in females than in males. This may account, at least in part, for the increased female susceptibility to PBC [163]. Additional explanations for the female preponderance were recently reviewed [164]. No single nucleotide polymorphisms (SNPs) were identified to the AE2 locus so far, supporting the suggestion that the AE2 impairment as observed in PBC is the result of epigenetic modifications. The current evidence supports miR-506 as such an epigenetic target. Thus, miRNA 506, which targets AE2 mRNA, is upregulated in BECs from patients with PBC, leading to decreased AE2 expression and impaired biliary secretions, as mentioned before [59]. Cultured PBC BECs express higher levels of miRNA-506 and lower Cl−/HCO3 exchange activity, which can be restored by anti-miR-506 oligonucleotides [59]. Several pro-inflammatory cytokines such as IL-8, IL-12, and IL-17, which are increased in the livers of PBC patients, upregulate the expression of miR-506 in BECs, while miR-506 initiates PBC-like features in BECs and increases immune activation. miR-506 also renders BECs more sensitive to cell death induced by the toxic bile acid CDCA or GCDCA [156,165]. Moreover, the overexpression of miR-506 upregulates the expression of inflammatory, fibrotic, and senescence genes [48].

The miR-506-AE2-sAC Model for PBC

An interesting hypothesis has been formulated by the current data. The unknown initial insult epigenetically activates a normally silenced gene such as the X-linked miR506, which binds to the 3′UTR of AE2 mRNA and represses the translation of AE2 followed by impairment of the bicarbonate umbrella. Genetically susceptible females harboring two X-linked alleles have a greater possibility of developing PBC than males. The resultant elevated cellular bicarbonate upregulates sAC activity, which sensitizes BECs to apoptosis by protonated BAs and other toxic metabolites in small concentrations that under normal circumstances would be harmless. Apoptosis of cholangiocytes results in a leaky biliary epithelium and overflow of bile leads to portal inflammation. The barrier function of BECs is damaged, resulting in inflammation and oxidative stress. Production of inflammatory cytokines further upregulates miR-506 expression, closing a vicious cycle and eventually causing senescence of BECs, vanishing bile duct syndrome, and biliary fibrosis [166,167]. Moreover, in cholangiocytes, Ca2+ signaling is controlled by inositol 1,4,5 trisphosphate (InsP3), which promotes Ca2+ release from the endoplasmic reticulum (ER) by binding to the inositol 1,4,5-trisphosphate receptor (InsP3R). MiR506 directly inhibits InsP3, leading to a reduction in Ca2+, followed by the derangement of BEC secretion and cholestasis [165,166] (Figure 1).

7. Bile Acids

Primary bile acids chenodeoxycholic acid (CDCA) and cholic acid (CA) are synthesized in the hepatocytes from cholesterol and are secreted in the bile through the bile salt export protein (BSEP), after conjugation in the liver with glycine or taurine. Almost 95% are reabsorbed by the enterocyte apical sodium-dependent bile acid transporter (ASBT). The remaining bile acids enter the colon and interact with gut microbes, forming secondary bile acids such as deoxycholic acid (DCA), lithocholic acid (LCA), and ursodeoxycholic acid (UDCA) [168,169,170].

The evidence indicates that bile salts have several functions according to their concentration. At concentrations of 15–25 μM, they act as signal molecules; at 50–200 μM, they induce apoptosis; at around 200 μM, they are inflammation mediators, acting as danger-associated molecular patterns (DAMPs), which lead to the activation of the NLRP3 inflammasome; at 200–2000 μM, they lead to necrosis; and at above 2000 μM, they act as detergents [171,172,173].

The mechanisms that are responsible for bile-acid-induced liver damage include mitochondrial and endoplasmic reticulum oxidative stress, apoptosis, and their detergent action on the liver cell plasma membrane [174,175,176]. Experimental studies in murine models additionally indicate that reactive oxygen species (ROS) are an important mechanism of BEC injury in PBC. UDCA is a strong ROS scavenger, attenuating mitochondrial oxidative stress and lipid peroxidation [177,178,179]. Hydrophobic bile acids, such as GCDCA, reduce the expression of AE2 in BECs by inducing mitochondrial ROS and enhancing BECs’ senescence, as mentioned before [156]. In line with these reports, an earlier study showed that antioxidant substances such as vitamin E and retinol were reduced in PBC patients compared to normal controls [180]. A later study found normal levels of vitamin E but a significant reduction in other antioxidants such as plasma total glutathione, serum selenium, and vitamin A [181]. In contrast, an increased corrected total antioxidant capacity was reported in early PBC [182], which may be explained as a compensatory but insufficient increase in antioxidants to compensate for augmented ROS production. Interestingly, an earlier report indicated that a decrease in the intracellular amounts of the antioxidant glutathione led to upregulated degradation of the anti-apoptotic bcl-2 protein and increased BECs’ apoptosis [183].

Bile acids are implicated in the regulation of immune functions. Thus, increased concentrations of BAs transform liver dendritic cells (DCs) from a tolerogenic to an activated phenotype able to activate T lymphocytes [184]. Furthermore, BAs are involved in Th17 cells’ infiltration by upregulating cytokine and chemokine secretion by macrophages [185]. Certain BAs directly regulate adaptive immunity by modifying the balance of Th17 and Treg cells. In a murine experiment, two derivatives of lithocholic acid (LCA), namely 3-oxoLCA and isoalloLCA, were mediators of this effect. The first prevented Th17 differentiation by binding to the retinoid-related orphan receptor-γt (RORγt). The second expanded Treg differentiation [186]. Interestingly, human gut microbes were found to convert LCA into 3-oxoLCA and isolithocholic acid [187].

8. Gut Microbiota

Gut microbiota is involved in the metabolism of BAs, affecting the composition of the bile acid pool. At the same time, BAs also interfere with the composition of gut microbiota through their antimicrobial activity. The fact that bile acids (BAs) are modified by microbes in the intestinal lumen suggested that the gut microbiota may be implicated in the induction and progression of PBC. Recent studies indicated that in PBC, conjugated BAs were reduced and unconjugated, while BAs were relatively increased. In addition, the conversion of primary to secondary BAs was decreased, suggesting a deranged microbial metabolism of BAs. Late-stage PBC patients have more abnormalities of the BA composition compared with early-stage disease. Moreover, the number of secondary BAs was inversely correlated with PBC gut microbial genera such as Veillonella and Klebsiella, while it was positively correlated with control microbes such as Faecalibacterium and Oscillospira. Treatment with UDCA showed a marked increase in the taurine-metabolizing Bilophila spp., which were strongly associated with decreased taurine-conjugated BAs [188].

However, the mechanisms by which gut microbiota affect PBC have not been clarified. A recent study found that the class Coriobacteriia is linked to the risk of PBC, whereas the class Deltaproteobacteria forms a protective factor [189]. In another recent study, the abundances of the genera Clostridium innocuum, Butyricicoccus, and Erysipelatoclostridium were inversely correlated with the risk of PBC, which means that an increased relative abundance decreased the risk of PBC [190]. It has been suggested that dysbiosis in gut microbiota impairs the intestinal barrier and alters the production of microbial metabolites, which in turn disrupt immune homeostasis [191]. Such a gut barrier dysfunction was demonstrated in the murine model dnTGF RIITLR2-/- of PBC, allowing for bacterial translocation into the portal circulation and exacerbation of hepatic T-cell-mediated cholangitis [192]. An important defect identified in PBC patients is the increased intestinal permeability, which was demonstrated by an increased sucrose excretion test [193], increased serum LPS levels, and higher antibody titers of lipoteichoic acid [194,195]. Alterations of gut microbiota combined with increased gut permeability favor pathogens’ translocation to the liver, which may cause damage through the LPS/TLR4 pathway and NF-kB activation. ultimately producing pro-inflammatory cytokines [196]. Intestinal dysbiosis and increased toxic BA levels play a significant role in the pathogenesis of PBC [197,198]. BAs modulate several signaling metabolic pathways by acting on the farnesoid X receptor (FXR) and TGR5 [199,200].

In patients with PBC, fecal bacteria alterations such as reduced alpha diversity and an increase in the genus Weisella are associated with advanced fibrosis [201]. PBC patients had an increased abundance of the order Lactobacilli and reduced levels of the beneficial order Clostridium [202]. The abundance of beneficial bacteria is generally reduced in PBC, while the abundance of opportunistic pathogens is increased. The genera Sphingomonas, Pseudomonas, and Acinetobacter were significantly increased in PBC, and the genus Leptotrichia was significantly reduced [203]. Fecal microbiota of patients with PBC and healthy individuals were transplanted into a murine PBC model. PBC feces increased the serum alkaline phosphatase, total bile acid content, and liver injury in the PBC model. Upregulation of liver immune and signal transduction pathways was also observed in PBC transplanted animals but not in disease controls. The mice gut microbiota and metabolome differed between PBC and healthy fecal transplantation and were similar to those of their donors. Importantly, transplantation from other disease control patients did not affect the recipient animals [204].

In the bile duct ligation model, the administration of Lactobacillus acidophilus ameliorated hepatic injury, reduced liver total BAs, and increased fecal total BAs. Enhanced excretion occurred due to increased unconjugated BAs as L. acidophilus increased fecal bile salt hydrolases. A clinical trial reported on the administration of L. acidophilus in cholestasis patients. Treatment improved liver biochemistry and was inversely correlated with aminotransferases, alkaline phosphatase, γGt, and total bilirubin. Clinical values were related to alterations in BAs and gut microbiota composition [205].

Short chain fatty acids (SCFAs) are molecules that are associated with the gut microbiota. SCFAs are usually beneficial to the host. They promote nutrition of the intestinal epithelium and support the normal gut barrier function. They also modify the immune response as they increase the ability of macrophages to kill bacteria. They promote the differentiation of T cells into Tregs and ameliorate the inflammatory response induced by bacterial pathogens [206,207,208,209]. It was reasonable, therefore, that their role in PBC was investigated. However, the results are controversial in clinical studies. A significant reduction in butyrate-producing beneficial bacteria was reported in UDCA non-responder patients [202]. Treatment of PBC patients with the bile acid sequestrant cholestyramine showed that high responders enriched Lachnospiraceae species, producing SCFAs in the microbiota [210]. On the other hand, Klebsiella pneumonia was only increased in the partially responders’ group [211]. In contrast, it was reported that PBC patients with fibrosis had increased levels of total fecal SCFAs and acetate compared to non-fibrotic patients, suggesting an association of SCFA with fibrosis [212]. Further investigations are required.

A detailed description of microbiotal changes has been recently published [196].

9. Environment

Bacterial infection and xenobiotics are considered potential environmental factors that may initiate PBC. Case–control studies have repeatedly reported that urinary tract infections caused by Escherichia are associated with PBC. The composition of PDC-E2 from E. coli is similar to human PDC-E2. Another bacterium that has been associated with PBC is Novosphingobium aromaticivorans, which produces lipoylated proteins able to react with sera from PBC patients.

Xenobiotics are also potential environmental factors that may be implicated in PBC. Reports indicate that frequent use of nail polish may predispose individuals to PBC according to epidemiological observations. The 2-octynamide, a derivative of 2-octynoic acid present in cosmetics and some chewing gums, was strongly reactive with AMA+ sera of PBC patients. The alkynamide 2-nonyamide, also a product of an alkynoic acid, is another xenobiotic that is reactive with AMA-positive PBC sera [213,214,215]. It is also found in cosmetics.

It should be noted that conclusions from experimental animal models using infection with N. aromaticivorans did not support its specificity in the induction of PBC [216,217]. A recent study frequently identified bacterial genera such as Sphingomonas panacis, Providencia, and Cutibacterium in the portal areas from livers of PBC patients. All tested samples had polymerase chain reaction bands for S. panacis [218].

However, an argument against the etiological role of bacteria in PBC is that improvement of environmental hygiene is associated with an increased prevalence of PBC, contrary to expectations. Nonetheless, bacterial infection may be an important risk factor, especially in female patients. The increasing prevalence of PBC, especially in males, may indeed be related to the increasing exposure to xenobiotics in both genders [22].

Interesting data on environmental factors are provided by recent studies in different areas of the world. Thus, in Korea, smoking and a family history of PBC were related to PBC. A high abortion rate and less full-term deliveries were also related to PBC in women. Interestingly, mild to moderate alcohol use was inversely associated with PBC [219]. In Japan, poor hygiene in children, smoking, and hair dye were risk factors for PBC [220]. In Greece, not only active smoking but also passive smoking were risk factors for advanced fibrosis after adjusting for sex, age, BMI, and alcohol consumption. For every pack-year increase in smoking, there was a 3.2 times higher likelihood of advanced fibrosis [221]. A recent meta-analysis has confirmed that smoking is related to the risk of advanced liver fibrosis in PBC [222]. On the other hand, for people living in the northeast of England and north Cumbria, a risk factor for PBC was the high level of cadmium in urban areas with a history of coal mining [223].

9.1. Molecular Mimicry

Molecular mimicry refers to the fact that self-antigens are simply innocent victims of the immune system reactions against microbial or viral peptides that display a molecular similarity with the host antigens. The possible roles of Novosphingobium aromaticivorans (also called Sphingomonas) and E. coli have been described before [203,224].

The available evidence also demonstrates that xenobiotics are capable of imitating self-antigens or modifying endogenous proteins to form neo-antigens [225].

A common drug such as acetaminophen can modify lipoic acid in PDC-E2, promoting loss of tolerance and the initiation of PBC [226]. In this line, changes in IFN-γ expression may also participate in PBC pathogenesis. A mouse model of PBC was created with post-transcriptional deregulation of IFN-γ. A female dominant cholangitis with portal tract granulomas, increased levels of BAs, increases in serum IgM, and the production of AMAs and antibodies to gp210 are characteristics of this model [227,228].

9.2. The Viral Etiology of PBC

Viruses have been proposed by earlier reports as predisposing cross-reactive agents in PBC [229,230,231]. The human beta retrovirus (HBRV) has been associated with the induction of cholangitis and AMA production in PBC patients, while the mouse mammary tumor virus (MMTV) has been linked to autoimmune biliary disease in murine models [232]. Direct linking of the viral infection with a complex multifactorial disease such as PBC is difficult, although most of the modified Koch’s postulates and Hill’s modified criteria are fulfilled [20]. In vitro HBRV stimulates the appearance of mitochondrial antigens on the surface of BECs. More importantly, antiretroviral therapy has indicated that repression of the viral load was accompanied by a sustained biochemical response. Gastrointestinal side effects in PBC were increased in comparison to patients with HIV [233]. Moreover, evidence of HBVR infection in PBC patients has been reported through integration studies in lymphoid tissue and BECs. Lymph node homogenates from PBC patients were co-cultivated with a breast cancer cell line, which showed that a transmissible beta retrovirus is present in PBC [234]. Additional evidence of a transmissible virus was offered by the infection of BECs [234,235].

Recurrent primary biliary cholangitis (rPBC) frequently follows liver transplantation and is accompanied by increased mortality. It has been suggested that rPBC resembles an infectious disease since increasing immunosuppression with tacrolimus leads to an earlier recurrence of increased severity [236]. In accordance with previous evidence, two patients with rPBC were treated with tenofovir/emtricitabine-based regimens combined with either lopinavir or raltegravir. Both biochemistry and histology were improved but the antiretroviral therapy was associated with several side effects [237].

10. Immunology Theory

Innate and adaptive immunity are involved in the pathogenesis of PBC.

10.1. Innate Immunity

10.1.1. Monocytes and Macrophages

Kupffer cells (KCs) are a self-proliferating indigenous population of macrophages in the liver and are distinct from the recruited bone-marrow-derived macrophages (BMMos) originating from circulating Ly6Chi monocytes or from CD14+ monocytes, which are the human analog of the mice Ly6Chi [238]. Monocytes can be classified into the classical CD14+CD16- and the non-classical CD14+CD16+ populations. Increased levels of the non-classical monocytes are found in the peripheral blood and livers of CLD patients and correlate with disease severity by participating in liver inflammation and fibrosis [239].

An impaired monocyte/macrophage function in PBC accompanied by accumulation of these cells in the liver has been shown [240]. Increased circulating levels of non-classical monocytes were also demonstrated in PBC patients in relation to the levels of aminotransferases, alkaline phosphatase, and γGt levels [241]. CD68-positive monocytes were present in the biliary epithelial layer and periductal tissue of PBC patients, but not in control livers [242]. In vitro stimulation of TLR2,3,4,5,9 receptors of PBC monocytes with the appropriate ligands showed higher production of pro-inflammatory cytokines such as IL-1β, IL-6, and TNF-a compared to normal monocytes [243]. Additionally, the expression of TLR4 and its negative regulator RP105 were impaired in PBC monocytes stimulated with LPS, leading also to increased production of several pro-inflammatory cytokines [244].

In accordance with these findings, higher levels of monocyte chemokines were shown in PBC livers [245]. In a PBC murine model of autoimmune cholangitis, Ly6Chi monocytes were increased in the portal tracts, but Kupffer cells were reduced. Repression of monocyte recruitment through deletion of the C–C chemokine receptor 2 (CCR2) attenuated autoimmune cholangitis [246]. The AMA-PDC-E2 complexes polarize monocyte-derived macrophages into M1 phenotype and generate a burst of pro-inflammatory cytokines including IL12p40, perpetuating inflammation and bile duct injury [131]. These findings verify the important role of IL-12 in PBC [43,247]. Endotoxin, which favors M1 polarization, is increased in BECs of PBC patients [248]. Additionally, CD40L, which mediates inflammation after binding to its receptor CD40, is increased in PBC macrophages. LPS is the main stimulator of CD40L [249]. Interestingly, it was reported that monocytes turn into an anti-inflammatory phenotype in UDCA responders but retain the pro-inflammatory phenotype in non-responders [250].

Kupffer cells in PBC patients also polarize into an M1 phenotype and localize in the peri-portal area, which is positively associated with liver inflammation [251].

An interesting finding was recently reported. The Clostridium metabolite P-Cresol Sulfate (PCS), a tyrosine product, attenuated PBC inflammation. Oral administration of tyrosine to a model of murine PBC increased PCS and decreased inflammatory mediators, while Kupffer cells polarized into the M2 phenotype, suggesting that a food intervention might be justified in human PBC [252].

10.1.2. Biliary Innate Immunity

Several TLRs linked to intracellular adaptor molecules such as myeloid differentiation factor 88 (MyD88) and receptor-associated kinase-1 (IRAK-1) are expressed in biliary epithelium [253,254]. Activation of MyD88 and IRAK-1 activates NF-kB, leading to the production of pro-inflammatory molecules [242,255,256,257,258,259]. These chemokines induce the recruitment of T cells, macrophages, neutrophils, and NK cells, leading to persistent inflammation. Liver NK cells are markedly increased in PBC patients. Furthermore, liver NK cells from PBC patients demonstrate high cytotoxicity against autologous BEC [260] and injure BECs at high NK/BEC ratios. In turn, injured BECs release antigens, which activate T cells in association with APC, as mentioned before [261,262].

MAIT cells are a class of innate cells participating in PBC pathogenesis. Activated MAIT cells are increased in the livers of PBC patients. Cholic acid regulates the MAIT cell function by inducing IL-7 production in hepatocytes. This is an indication that a bile acid signaling pathway may be the connection between cholangiocyte damage and innate immunity [263].

The expression of the chemokine receptors CCR6 and CXCR6 allows MAIT cells to be recruited around bile ductules in response to CCL20 and CXCL16 produced by BECs [262]. Additionally, persistent stimulation of MAIT cells by IL-12 leads to increased IL-17, IFN-γ, TNF-α, perforin, and granzyme B production, followed by damage to BECs and induction of fibrosis by activating HSCs [264,265].

In contrast to LPS and other bacterial PAMPs, double-stranded RNA (dsRNA) binds to TLR3, leading to the activation of interferon regulatory factor 3 (IRF-3) and NF-kB [266]. The creation of a cytokine network that follows around ductules participates in the pathogenesis of fibrous cholangiopathies [255,256,258,267,268]. An additional signaling pathway triggered by TLRs is the activation of Ras and the production of pro-inflammatory cytokines and chemokines, which lead to cellular senescence [109,269]. Persistent LPS stimulation of BECs also initiates the production of mediators of cellular senescence such as p16INK4a and p21WAF1/Cip in cholangiocytes [109]. N-Ras is activated in BECs after both acute and persistent LPS stimulation. Acute stimulation induces the activation of innate immunity, while persistent stimulation triggers cellular senescence. In chronic biliary diseases, persistence predominates.

BECs can also function as APCs. Human leukocyte antigen (HLA) class I molecules are constitutively expressed in human BECs, implicated in CD8+ cytotoxic activity [262]. HLA class II molecules are expressed after stimulation in immune-mediated diseases such as PBC. Normal and diseased BECs express molecules such as ICAM1, CD40, and CD44, which mediate the interaction between BECs and T cells. BECs can present antigens and activate unconventional T cells, such as NKT cells, via the CD1d molecule, expressed by normal BECs, but CD1d is downregulated in PBC [270]. Human BECs phagocytose autologous apoptotic BECs, accompanied by upregulation of the chemokines CCL2 (attracting monocytes) and CXCL8 (attracting neutrophils). BECs express the phagocytosis receptors and phosphatidylserine receptors (PSRs), indicating they are capable of phagocytosing apoptotic BECs through the ‘eat-me’ phosphatidylserine signals [271].

10.2. Adaptive Immunity

Adaptive immunity has been at the center of mechanisms in the pathogenesis of PBC [228,272,273,274,275]. APCs, including BECs, dendritic cells (DCs), macrophages, and B cells, present elements of microorganisms, xenobiotics, and even apoptotic BECs through pattern recognition receptors (PRRs) to T lymphocytes, to initiate the inflammatory response [276]. After antigen presentation, T cells differentiate into Th1, Th2, Tfh, or Th17 cells or into CD8+ cytotoxic T lymphocytes. B lymphocytes are also stimulated into plasma cells to produce antibodies [277]. At the same time, Tregs and B regulatory cells control this pro-inflammatory response [278]. APCs in the bile ducts are continually exposed to various PAMPs that arrive from the gut. Therefore, tolerance has been developed to maintain immunological homeostasis [279]. However, molecular mimicry can induce a breach of tolerance in PBC.

The participation of Tregs in the breach of tolerance has been investigated. Patients with PBC have reduced liver Tregs in comparison to controls. Importantly, reduced CD4+CD25+ Tregs were also identified in first-degree female relatives of PBC patients. However, functional studies failed to demonstrate a universal PBC Treg abnormality. The number of Tregs was lower in PBC portal tracts when compared with chronic HCV and autoimmune hepatitis cases [280]. Tregs from PBC patients react to low-dose IL-12, acquiring a Th1-like phenotype with decreased suppressive ability and increased production of IFNγ. A rapid phosphorylation of STAT4 on Tregs was observed, providing further evidence for the implication of the IL12/IL-12Rβ2/STAT4 pathway for Tregs in PBC pathogenesis [281]. As mentioned above, one of the consequences induced by the loss of tolerance is the presentation of epitopes of the 2-oxo-acid dehydrogenase complex of the inner mitochondrial membranes to APC cells and generation of autoantibody-secreting plasma cells [168]. This response is amplified by the relative lack of functional CD4 Tregs [282].

AMAs were considered not only surrogate markers of PBC but also participants in disease pathogenesis [50]. All mitochondrial antigens recognized by AMAs have been identified and cloned [283,284]. They have only one dominant epitope within the lipoyl binding site [285,286]. In PBC, uptake of AMA–PDC-E2 immune complexes by APCs promotes the activation of PDC-E2-specific CD8+ cytotoxic lymphocytes (CTLs). The fact that CD4 T cells, CD8+ CTLs, and B cells respond to a PDC-E2 epitope was interpreted as an indication that PDC-E2 autoantigens have a central role in pathogenesis, even if AMAs do not directly damage BECs [287]. AMAs appear before clinical disease by at least a decade [288,289,290]. It should be noted that molecular mimics of the lipoyl binding site of PDC-E2 react with AMAs [22,214,274,291]. Murine administration of such xenobiotic mimics initiates AMAs of similar specificity with antibodies [292,293]. The presence of AMAs in FDRs of PBC patients that do not acquire PBC is a strong argument against a pathogenic role of AMAs [294]. The fact that there are no clinical differences between AMA-positive and AMA-negative PBC patients also supports arguments against a fundamental role of AMAs in pathogenesis [295]. One may also wonder why AMAs wait for years before causing the disease if they are indeed involved in pathogenesis. It should also be stressed that although PBC experimental models allow for the detection of the initial immunopathogenic events, none represents all the characteristics of human PBC.

In PBC, destruction of BECs is mediated by CTLs, which are significantly increased in PBC livers [296]. Nonetheless, in the IL-2 receptor (CD25)-negative murine model, the reduction in CTLs ameliorates but does not eliminate ductular damage, suggesting that this is not the only destructive mechanism [297,298]. A child with a deficiency of the α subunit of CD25 in peripheral lymphocytes acquired a PBC-like disease [299], indicating the importance of dysfunctional CD4+CD25+ Tregs in autoimmunity. On the other hand, increased levels of IL-12/Th1 and IL-23/Th17 have been demonstrated in the livers of PBC patients, mainly located around the injured bile ducts [42,300]. Moreover, a shift from the Th1 to the Th17 infiltrate was observed in advanced PBC. This might indicate that Th1 cells are important at the onset of disease, but Th17 cells may be responsible for the perpetuation of the disease [42]. A recent paper provided additional evidence for the role of IL-17. Increased numbers of invariant natural killer T (iNKT) cells as well as CD3+ CD56+ αGalcer-CD1d tetramer- T cells were demonstrated in PBC patients. These cellular subsets produced increased amounts of IL-17A and were positively correlated with disease severity [301]. However, in a murine PBC model, it was IL-21 produced by Th17 cells and not IL-17A that was related to fibrosis. IL-21 promoted liver inflammation and fibrosis by increasing the numbers of CTLs [302].

Further evidence for the role of CD8+ and Tregs is provided by a different PBC murine model, namely the dominant-negative transforming growth factor (TGF)-β receptor II (dnTGF-βRII) mouse [303]. These mice express a mutated TGF-βRII in both CD4+ and CD8+ T cells, which inhibits signal transduction [304]. They present common characteristics with human PBC, such as the presence of AMAs, CD4+, and CD8+ T cells, infiltration of portal tracts, and increased levels of IFN-γ, and pro-inflammatory cytokines [303,305]. CTLs seem to be essential in cholangitis development [306], while the loss of tolerance in this model is due to a defective suppressor function of Tregs [307].

TGF-β is a cytokine with extensive involvement in immune regulation. It is implicated in the preservation of immune tolerance [308]. TGF-β initiates both the differentiation [309,310] and preservation [307] of T-regs, and in association with IL-6, it initiates the differentiation of the pro-inflammatory Th17 cells [311,312,313]. Experiments in murine models suggest that impaired TGF-β signaling is implicated in the pathogenesis of PBC, through dysregulation of Tregs [303]. In line with this, an increased TGF-β3 isoform was observed both in peripheral and the hepatic vein blood in PBC patients. This increase was evident from the early stages, suggesting that this is a disease-specific finding and not the result of the cirrhosis. TGF-β1 was reduced in all cirrhotics, irrespective of etiology, as verified in the blood of hepatic veins. Furthermore, cholangiocytes, portal lymphocytes, and sinusoidal cells were identified as the origin of liver TGF-β isoforms. Hepatocytes also strongly expressed TGF-β3 in PBC patients [314]. Alterations in TGF-β isoforms may shift the balance between Tregs and Th17 cells towards the pro-inflammatory Th17 cells, assisting the CTL-induced destructive mechanism, while functionally defective T-regs are not able to control inflammation. TGF-β1 is involved in FoxP3 expression and the regulatory function of CD4 cells [315]. This hypothesis is further supported by the finding that mice with a mutation of the gene encoding FoxP3 develop a PBC-like cholangitis with high levels of IL-17 and IL-23 produced by Th17 cells [316]. Experimental evidence further supports the significance of TGF-β3 in the pathogenesis of PBC. TGF-β3 reduces collagen synthesis and tissue inhibitor of metalloproteinases-1 (TIMP1) expression, and it upregulates matrix metalloproteinase-9 in the carbon tetrachloride model of liver fibrosis [317]. This finding may explain how the progression of fibrosis in PBC is delayed over several years in comparison with other chronic liver diseases. TGF-β3 may be the connection for the Tregs/Th17 deregulation observed in PBC.

As mentioned before, one of the characteristics of PBC is the infiltration of portal tracts with CD4+ and CD8+ lymphocytes. A possible explanation for the recruitment of T cells in the liver in PBC patients was proposed. The CXC chemokines CXCL9 and CXCL10 attract CXCR3- T lymphocytes. Increased levels were identified in the serum of PBC patients in association with increased numbers of CXCR3+ peripheral and portal tract lymphocytes [318]. These chemokines and their receptors recruit Tregs and Th17 and Tc17+ lymphocytes in the liver [319,320]. Importantly, increased levels of these chemokines, but not of CXCR3+ lymphocytes, were demonstrated in first-degree female relatives of patients [318]. UDCA treatment significantly reduced serum chemokines, possibly indicating a new mechanism of action for UDCA [139]. More importantly, the evidence indicates that CXCL10 is a pro-fibrotic chemokine, which participates in the crosstalk between hepatocytes, HSCs, and immune cells such as NK cells. CXCL10 inhibition could be a treatment possibility for fibrosis of PBC [321]. CXCR3, on the other hand, is not only involved in chemotaxis but is also implicated in T-cell differentiation and may play a central role in the maintenance of tolerance. It has been suggested that CXCR3 reduction initiates autoimmune cholangitis through pathogenic CD8+ T cells [322]. Collectively, all cytokines and chemokines produced by BECs are designated as cholangiokines. Cholangiokines modulate the hepatic environment. Quiescence-maintaining cholangiokines support liver homeostasis, whereas cholangiokines secreted by BECs after senescence or injury mediate fibrogenesis, ductular reaction, and inflammation, finally leading to carcinogenesis [323].

The recent interest in the role of the checkpoints and their inhibitors in the regulation of the immune response has attracted interest in earlier studies of immune checkpoints in PBC. PD-1 is expressed on activated T cells and B cells. In a mouse model of PBC, PD-1/PD-L1 interaction resulted in apoptosis of activated liver CTLs at an early animal age, indicating that programmed cell death-ligand 1 (PD-L1) expression mediated by IL-10 and TGF-β plays a fundamental role in T-cell tolerance. Older animals, however, showed a downregulation of PD-1 and apoptosis, followed by the accumulation of cytotoxic CD8(+) T cells and cholangitis [161].

In PBC patients, PD-1 was markedly expressed on liver-infiltrating T cells around damaged bile ducts, but the mRNA levels of PD-1 and PD-L1 were downregulated in the peripheral blood [324]. The PD-1 ligands were modulated by IFN-γ in peripheral blood mononuclear cells of PBC patients [325]. Recently, it was reported that PD-1 was downregulated in peripheral CD8+ cells, while PD-L1 in human BECs was also downregulated. The results therefore were increased CTL cytotoxicity and enhanced BEC apoptosis [326].

Another immune checkpoint pathway that involves two receptors, the CD226 and the T cell Ig and ITIM domain (TIGIT), was investigated in PBC. TIGIT is expressed on effector and memory T cells. It transduces a negative signal after binding to its ligand CD155, leading to repression of the immune response. CD226 is a competitor for the same ligand and transfers an activation signal regulating the immune system [327]. PBC patients had increased proportions of peripheral CD8+ T and CD4+ T cells expressing either CD226 or TIGIT in comparison with other liver diseases or normal controls. It was suggested that CD226/TIGIT immune checkpoint imbalance is implicated in the pathophysiology of PBC [328].

The therapeutic effects of checkpoint inhibitors (ICPs) have not been tested in PBC. A case report described a melanoma patient with established overlap syndrome of PBC/autoimmune hepatitis who was treated with pembrolizumab, which proved that ICP administration is safe in patients with PBC [329]. There seems to be a justification for clinical trials to verify the safety and efficacy of ICPs in reducing biliary injury and cirrhosis in PBC [330].

A simplified summary of the immunological theory is presented in Figure 2.

11. The Endothelin Theory

Three peptides are included under the designation of endothelin (ET), namely ET-1, ET-2, and ET-3. They bind and activate their specific receptors, ET-A and ET-B [331]. ET-A mediates vasoconstriction and inflammation, whereas ET-B initiates vasodilation and represses inflammation [332]. Both ET-1 and ET-2 bind to ET-A, and BECs are an important source of ET-1 in the bile duct ligation (BDL) model. TGF-β1 stimulates BECs to produce ET-1 in this model [333].

Recent evidence indicates that endothelins and their receptors regulate liver disease progression [334,335]. Earlier studies showed that ET-A inhibition decreased liver fibrosis in the BDL model [336], whereas inhibition of ET-B increased portal pressure in normal mice [337]. Inhibition of ET-A or ET-B alone had no effect on the increased portal pressure of BDL rats, but the combined ET-A and ET-B blockade ameliorated portal pressure in cirrhotic rats [338]. However, a recent study did not confirm these findings, as in human cirrhotics, inhibition of ET-A reduced the portal pressure [339]. A recent report demonstrated that dual endothelin receptor antagonism led to markedly increased pro-fibrotic and pro-inflammatory markers due to increased synthesis and deposition of chondroitin sulfate [340].

The Mdr2-/- mice model of primary sclerosing cholangitis had increased biliary ET-1/ET-2/ET-A expression, but human PSC had only increased ET-1 and ET-A expression. Binding of ETs in ETA in BECs led to upregulated production of IL-6 and TGF-β1, leading to macrophage activation, inflammation, and fibrosis. In vitro, ET-1/ET-A caused BECs’ senescence [341].

Increased serum levels of ET-2 were described in PBC patients. This was a disease-specific finding since it was not found in other chronic liver diseases with or without cirrhosis. Moreover, ET-2 levels were increased from the early stages of the disease and, most importantly, they were restored in PBC patients after UDCA treatment [342].

Based on this observation, a hypothesis for the pathogenesis of PBC was proposed [343]. It was suggested that the initial insult in genetically susceptible patients is the infection with either a retrovirus or a pathogen such as N. aromaticivorans. Uptake of foreign antigens was mediated by a receptor that has been described in endothelial cells, while lipoteicholic acid, which is a strong antigen of Gram-positive bacteria, was identified in LSECs of PBC patients [344]. It was also suggested that ET-2 may activate Kupffer cells to secrete pro-inflammatory cytokines based on an earlier finding in mice peritoneal macrophages [345]. Furthermore, PAMPs binding on the TLRs of Kupffer cells lead to additional secretion of TNFa, IL-1, and IL-6. Moreover, ET-2 has a peptide homology with CXC chemokines and is a strong chemoattractant for macrophages through their ET-B receptor [346]. It was shown that 30% of the cells infiltrating portal tracts in PBC are macrophages and are mostly localized around injured bile ducts [347]. Activated macrophages differentiate into epithelioid cells [348], leading to granuloma formation. Endothelins are also responsible for the contraction of HSCs and the early development of portal hypertension in PBC [349]. In accordance with this hypothesis, PBC-specific antibodies against liver sinusoidal cells were identified in PBC [350], suggesting that sinusoidal cells may participate in PBC pathogenesis. Moreover, polymorphisms related to LSECs were identified in PBC. Both eNOS intron4 VNTR and eNOS exon7 G894T SNPs were associated with an increased risk of PBC, while endothelin-1 rs2071942 “A” and rs5370 “T” alleles had a link with disease progression [351].

An important aspect of this hypothesis is the association with the vascular supply of bile ductules. They receive blood only from hepatic arteries through a capillary called the peribiliary vascular plexus (PVP), which drains into the sinusoids. This characteristic vascular network is the background of why small bile ductules are implicated in ischemic injury [352,353,354,355]. Endothelins (ETs) and nitric oxide are the main regulators of circulation in the PVP [356]. The damage of BECs in PBC is initially due to ischemia through ET2-induced vasoconstriction. Ischemia then leads to apoptosis of BECs with the formation of apoptotic bodies containing epitopes, presented in the peri-biliary APC-expressing MHC II molecules. This uptake might be genetically regulated or alternatively may be initiated by pro-inflammatory cytokines [357]. This hypothesis could explain the already suggested similarities between PBC and graft vs. host disease (GVHD) [358,359,360,361], which is also related to endothelial cell damage [362]. In GVHD, high levels of ET1 were histochemically shown in endothelial cells days before the occurrence of GVHD, indicating a pathogenetic significance [363]. TGF-β abnormalities have also been included in this model. TGF-β1 modulates FoxP3 and the regulatory activity of Tregs [315]. We postulated that TGF-β3 impairs the regulatory function of Tregs.

In summary, the proposed hypothesis integrated the role of infective agents and the similarity of PBC with GVHD. In this model, AMA production is not a pathogenetic factor, but rather a secondary event in BEC damage. The existence of AMA-negative PBC patients is also explained.

A recent report corroborates this hypothesis. The increase in ET-2 was confirmed in the livers and sera from patients with PBC and in a murine model. This was associated with a parallel upregulation of ET-B. Importantly, overexpression of ET-B enhanced liver injury and triggered the production of TNFa and IFNγ, followed by an increase in aminotransferases, alkaline phosphatase, and AMA-M2 in the sera of PBC mice. This response was abrogated after the depletion of ET-B. The pro-inflammatory responses were related to upregulation of the G-protein-coupled receptor kinase-2 (GRK2) and activation of the NF-κB [364].

Comprehensive reviews of PBC pathogenesis have been recently published [168,365,366,367].

12. Interaction of Pathogenetic Factors in PBC

There is substantial evidence that several of the proposed pathogenetic factors implicated in PBC are closely interrelated.

An interplay exists between autophagy and senescence. The role of autophagy in senescence seems to be a double-edged sword. Thus, autophagy may attenuate senescence, degrading the causative factors. On the opposite end, reports have demonstrated that autophagy may promote SASP formation by producing useful building blocks such as amino acids and nucleotides required by SASPs [97,368]. Mitophagy and cell senescence also interact. Defective mitophagy promotes cellular senescence, and restoration of mitophagy delays senescence [369]. The regulation of senescence by mitophagy is in part independent of alterations in general autophagy [370,371].

Endoplasmic reticulum (ER) stress seems to be the reason behind deregulated autophagy initiated by hydrophobic bile acids, such as GCDC. A reduced ‘biliary bicarbonate umbrella’ additionally increases the toxicity of GCDC to BECs. UDCA, which is a FXR antagonist, increases the adaptive capacity of the ER. Pre-treatment with UDCA significantly repressed ER stress, and it improved impaired autophagy and senescence induced by GCDC [78]. Obeticholic acid (OCA), a second-line treatment of PBC, is an FXR agonist that inhibits the synthesis of endogenous bile acids [372,373]. OCA impairs autophagy, which is not consistent with the existing evidence showing a deterioration of cholestasis when autophagy is inhibited. Lysosomal inhibition by chloroquine or deletion of Atg7 and Atg5 augmented experimental cholestasis [374]. Therefore, OCA should have additional effects to compensate for its negative effect on autophagy.

There is also an interplay between an impaired biliary bicarbonate umbrella with autophagy and cellular senescence in PBC. The expression of AE2 was significantly increased in cultured BECs after short exposure of BECs to GCDC, but it was markedly decreased in senescent BECs induced by GCDC. Deletion of AE2 was associated with deregulated autophagy, increased cell surface expression of PDC-E2, and senescence in BECs. The findings were confirmed in the livers of PBC patients [375].

Apoptosis and senescence are also interconnected. In most cells, apoptosis and senescence seem to exclude each other. Studies on human fibroblasts demonstrated that overexpression of the anti-apoptotic protein Bcl-2 is vital for the switching from apoptosis to senescence [376]. Histone modifications of chromatin alter the ratio of Bcl-2/Bax gene expression in senescence in favor of Bcl-2, contributing thus to the apoptosis-resistant phenotype [377,378,379]. Also important for the crosstalk between apoptosis and senescence is the ability to stabilize p53. Senescent cells are unable to undergo p53-dependent apoptosis [380,381]. Apoptosis resistance in senescence is also mediated by the inhibition of p53 via overexpression of the cell cycle inhibitor p21 [382,383] or inhibition of the effector caspase 3, which is a critical step of apoptosis [384]. Moreover, overexpression of telomerase did not protect cells from senescence, but made cells more resistant to stress-induced apoptosis [385]. Additionally, there is evidence that IL-6 favors senescence through inhibition of the intrinsic pathway of apoptosis and via upregulation of the pro-survival nuclear translocation of NF-κB [386]. Recent evidence indicated that the transcription factor ETS1 and the histone acetyltransferase EP300 epigenetically increase the anti-apoptotic factor BCL-xL, inhibiting apoptosis in senescent BECs [387,388].

It is important to clarify the actual sequence of events, as PBC progression usually takes many years. The pioneer work from the group of Sasaki has shown that autophagy precedes senescence in PBC after immunohistochemical evaluation of both processes in PBC livers. Data also indicate that an impaired bicarbonate umbrella may precede abnormalities of autophagy, and cellular senescence in BECs. Aberrant expression of mitochondrial antigens may follow [78,112,117].

Finally, there is also a connection between senescence and macrophage infiltration in PBC, suggested by findings in human primary sclerosing cholangitis (PSC). This disease is closely associated with macrophage recruitment. Reduction in peribiliary macrophage recruitment attenuates liver damage and fibrosis in a murine model [389]. As mentioned before, portal granulomas found in PBC consist of epithelioid cells of macrophage origin. As mentioned before, senescent BECs activate the N-Ras signaling pathway, leading to the production of mediators such as IL-6, IL-8, and CCL2, which recruit M1 and M2 macrophages. Senescent BECs also release extracellular vesicles (EVs) containing DAMPs, which bind to the RAGE receptors expressed in macrophages, leading to macrophage activation and infiltration [390]. Among DAMPs, S100A11 is a constituent of EVs released by BECs derived from Mdr2-/- mice [391]. S100A1 is an agonist of RAGE [392] and is highly expressed in macrophages. Upon binding to S100A1, BMMos differentiate into a pro-inflammatory M1 polarization, through an NF-kB-dependent mechanism [391].

13. The Integrated Theory

Pathogenetic models proposed so far consider the breach of T-cell tolerance to mitochondrial epitopes as a critical event in the pathogenesis of PBC. As mentioned before, the appearance of AMAs does not damage BECs [393]. The molecular mimicry model seems to be a more logical approach compared with the rather obscure concept of autoimmunity [394].

An integrated model of PBC pathogenesis is proposed by summarizing the earlier and most recent data, as analyzed above. According to this model, the initial event in PBC is an infection by B human retrovirus (BHRV) of a genetically susceptible person [20]. This is more plausible than bacterial infections due to the long latency period of retroviruses before clinically evident disease. This would explain the finding of immunological abnormalities in patients long before the biochemical and pathological changes of PBC appear. Moreover, a protracted infection of healthy BECs could more reasonably explain the continuous progression of the disease over the years as compared to continuation of a rather peculiar autoimmunity not responding to traditional immunosuppressive treatment. On the other hand, genetic predisposition in such a model may be localized in more than one locus, and it is tempting that a combination of genetic abnormalities should co-exist rather than a simple anomaly. Intestinal dysbiosis and several microbiota abnormalities such as increased Firmicutes [137] may also add to the initial insult by altering the bile acid equilibrium, increasing the levels of toxic bile acids such as GCDC.

In this model, a central initial defect in PBC is the increased production of ET-2 by endothelial cells and possibly by BECs infected by the virus [343]. ET-2 causes contraction of HSCs, leading to early portal hypertension. Most importantly, ET2 leads to ischemia of BEC through constriction of the PVP. The resultant hypoxia of BECs [124] causes ROS generation and autophagy and mitophagy dysregulation, leading to apoptosis or senescence and gradual development of the vanishing bile duct syndrome with further retention of bile acids. Oxidative stress increases senescence by downregulating bmi1, which is a suppressor of P16INK4. Moreover, ET-2 binding to ET-A on the surface of BECs also drives the senescence of BECs [341]. Senescence is the second fundamental point in this model as it causes the appearance of SASPs, which in turn affect Kupffer cells, bone-marrow-derived macrophages (BMMs), and HSCs. SASPs overproducing IL-6 and CCL2 recruit BMM and activate Kupffer cells to produce pro-inflammatory cytokines. BMMs are the progenitors of epithelioid granulomas regularly found in PBC. Senescent BECs also release extracellular vesicles (EVs) containing DAMPs that bind to the RAGE receptors expressed in macrophages, leading to macrophage activation. In addition, SASPs overproducing PDGFa, Sct/SR, and TGF-β activate HSCs to increase fibrosis [115]. Overproduction of CCL2 and CX3CL1 chemokines initiate, senescence in neighboring BECs and hepatocytes [105]. Switching between senescence and apoptosis depends on the upregulation of Bcl-2 and the key protein IFIT3. Decreased IFIT3 leads to apoptosis, while upregulation of IFIT3 leads to senescence [110]. Upregulation of YAP also leads to senescence [112]. The third fundamental point is the reduction in AE2 and the resultant decrease in the bicarbonate umbrella, which is associated with dysregulated autophagy and mitophagy. HBRV infecting BECs upregulates miR 506, which further downregulates AE2 RNA and reduces the umbrella. At the same time, increased intracellular bicarbonate increases sAc, leading also to apoptosis or senescence of BECs by increasing the sensitivity of BECs to BA injury [166]. This is the fourth fundamental defect in this model. On the other hand, dysregulated autophagy and apoptosis generates mitochondrial antigens that are transported by apoptotic bodies and other EVs to liver APCs, leading to AMA production through B-cell transformation into plasma cells. AMAs in turn may attract T cells, enhancing a similar attraction by Kupffer cells and BMMs through the production of CXCL9 and CXCL10 chemokines [139]. The fourth fundamental defect is the production of ROS driven by ET-2-induced hypoxia of BECs and ER stress induced by the abnormal entrance of protonated bile acid in BECs. ROS generation is further increased by Kupffer cells through ingestion of BEC apoptotic bodies. ROS drive Kupffer cells and attract macrophages to produce pro-inflammatory cytokines and TGF-β. Increased production of the isoform TGF-β3 in PBC may impair fibrosis as it has been shown that TGF-β3 decreases collagen synthesis [317]. This may explain the relative delayed fibrosis and development of cirrhosis over many years compared to other chronic liver diseases such as alcoholic or viral disease. BECs are also attacked by immunological mechanisms involving T cells. T cells are recruited by chemokines produced by macrophages. Cytotoxic CD8+ cells are the main effector cells, assisted by Th1/IL-12 and Th17 cells, iNKT cells, and NK cells. CD8+ cells express reduced levels of PD-1 checkpoints accompanied by reduced expression of PD-L1 on the surface of BECs, resulting in diminished elimination of CD8+ cells by immune checkpoints. Increased levels of IL-17 and IL-23 are also produced by IL-17 and iNKT cells [301]. Differences in TGF-β isoforms are the connecting link leading to Treg/Th17 dysregulations reported in PBC. Increased local amounts of TGF-β3 (accompanied by the relative lack of TGF-β1) in association with high levels of IL-6 shift the balance towards the pro-inflammatory Th17 cells assisting the CD8+ cytotoxic destructive mechanism in conjunction with defective Tregs. Death of BECs by apoptosis, ferroptosis, and cytotoxic CD8+ T cells finally lead to the vanishing bile duct syndrome characteristic of the advanced stages of PBC.

Figure 3 diagrammatically presents the proposed model.

14. Conclusions

Primary biliary cholangitis, or chronic non-suppurative destructive cholangitis, is a disease with a complicated and not fully delineated pathogenesis. It is generally regarded as an autoimmune disease, but immunosuppressive treatment is not effective. During the last two decades, considerable progress has been made and many pathogenetic factors have been described. We presented all available proposals without any prioritization as we think that all disclose parts of the pathogenesis of this complex disease. Dysregulated autophagy, BEC senescence and apoptosis, and impairment of their protective bicarbonate umbrella have been proposed as important elements of pathogenesis. The role of toxic bile acids and dysbiosis of gut microbiota are also incriminated, acting on genetically susceptible patients. Environmental factors including xenobiotics may also participate. Epigenetic factors such as the effects of miR506 are implicated, explaining the female preponderance of cases. The result of these suggestions has switched the hypothesis from autoimmunity towards molecular mimicry as the fundamental event of PBC pathogenesis. Clearly, the current data indicate that the immune destruction of BECs and the production of AMAs are the result rather than the primary events leading to the disease. However, the initiating insult is still elusive. Moreover, there is no satisfactory explanation for relatives of patients with immunological abnormalities who do not develop the disease or see experience progression of the disease for very long periods over decades. The lack of a response to anti-interleukin treatments also has no explanation in the current immunological models. However, it is explained by a model suggesting that interleukins are merely a secondary arm in a more complicated mechanism. We proposed, therefore, an integrated model that includes the different factors suggested so far. The important aspect of this model is the suggestion that the initial insult is the infection by the human B retrovirus of BECs and possibly LSECs, a fact that is missing from previous models that bypass the initial insult. Retroviruses may be present for years before manifestations of the disease. Furthermore, based on the finding, specifically for PBC, of increased ET-2 production, we propose that the initial event is the vasoconstriction of the peribiliary vascular plexus and the resultant ischemia and hypoxia. Dysregulated autophagy, senescence, apoptosis, the impaired bicarbonate umbrella, and immunological abnormalities follow the initial event. Moreover, we propose that susceptible patients may in fact have more than one genetic abnormality, and therefore it is the sum of them and not a single one that characterizes the genetic profile of PBC patients. The new aspects of this model require further research, but therapeutic trials can start as drugs are already available. The integrated model implies that at least two new approaches in the treatment of PBC should be adequately tested. Anti-retroviral therapy and endothelin inhibitors may be tested as novel treatments of the disease, alone or in combination with UDCA. Finally, the proposed model tries to identify both the initial insult and the initial events, which may help to understand the pathophysiology of PBC in a holistic way, further initiating future research.