Hepatic Bile Acid Transporters and Drug-induced Hepatotoxicity

Abstract

Drug-induced liver injury (DILI) remains a major concern in drug development from a patient safety perspective because it is the leading cause of acute liver failure. One mechanism of DILI is altered bile acid homeostasis and involves several hepatic bile acid transporters. Functional impairment of some hepatic bile acid transporters by drugs, disease, or genetic mutations may lead to toxic accumulation of bile acids within hepatocytes and increase DILI susceptibility. This review focuses on the role of hepatic bile acid transporters in DILI. Model systems, primarily in vitro and modeling tools, such as DILIsym, used in assessing transporter-mediated DILI are discussed. Due to species differences in bile acid homeostasis and drug-transporter interactions, key aspects and challenges associated with the use of preclinical animal models for DILI assessment are emphasized. Learnings are highlighted from three case studies of hepatotoxic drugs: troglitazone, tolvaptan, and tyrosine kinase inhibitors (dasatinib, pazopanib, and sorafenib). The development of advanced in vitro models and novel biomarkers that can reliably predict DILI is critical and remains an important focus of ongoing investigations to minimize patient risk for liver-related adverse reactions associated with medication use.

Introduction

Hepatic transport proteins are membrane-spanning proteins that facilitate the movement of endogenous and exogenous substrates, such as ions, small molecules, peptides, lipids, and macromolecules across plasma membranes. Bile acids are synthesized in hepatocytes and excreted into bile by transport proteins. Approximately, 95% of bile acids are reabsorbed from the intestine by the apical sodium-dependent bile acid transporter (ASBT), transported into portal blood by ileal organic solute transporter (OST) α/β, and taken up into hepatocytes for subsequent metabolism and/or excretion. Bile acid transporters are crucial in this enterohepatic recycling process. The sodium taurocholate cotransporting polypeptide (NTCP) on the basolateral membrane of hepatocytes and the bile salt export pump (BSEP) on the apical (canalicular) membrane show high affinity for bile acids and are the primary hepatocyte transporters for bile acid uptake and excretion, respectively, in the healthy liver. Other transporters localized on the apical (multidrug resistance protein [MDR] 3 and multidrug resistance-associated protein 2 [MRP2]), and basolateral (organic anion transporting polypeptides [OATPs], MRP3, MRP4, and OSTα/β) membranes of hepatocytes may contribute to hepatic bile acid disposition and homeostasis (Figure 1).¹ The expression and function of these hepatic transporters are regulated by numerous mechanisms, including nuclear receptors, epigenetic gene regulation, microRNAs, alternative splicing, post-translational modifications, and trafficking, as detailed in a recent review by the International Transporter Consortium.²

Figure 1.

Key transporters involved in bile acid homeostasis in human hepatocytes and intestinal cells. Transporter-mediated biliary excretion of bile acids into the intestine, followed by reabsorption and transport back to hepatocytes via portal blood (enterohepatic recycling). ASBT, apical sodium-dependent bile acid transporter; BA, bile acid; BC, bile canaliculus; BSEP, bile salt export pump; MDR3, multidrug resistance protein 3; MRP, multidrug resistance-associated protein; NTCP, sodium taurocholate cotransporting polypeptide; OATP, organic anion transporting polypeptide; organic anion transporting polypeptide; OSTα/β, organic solute transporter alpha/beta; TJ, tight junction.

Drug-induced hepatotoxicity, frequently referred to as drug-induced liver injury (DILI), is an adverse reaction to any xenobiotic that is harmful or toxic to the liver. Drug-induced liver injury may be intrinsic, dose-dependent, and predictable (e.g., acetaminophen, bromfenac, cyclophosphamide, methotrexate) or idiosyncratic and unpredictable (e.g., amoxicillin-clavulanate, troglitazone, tolvaptan, tyrosine kinase inhibitors [TKIs]).³ Drug-induced liver injury occurrence is a major health safety concern, may be fatal, and remains a major challenge in drug development. Clinical diagnosis of DILI typically involves increased levels of serum biomarkers of liver function, including aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), and total bilirubin. Several clinical criteria and biomarkers are under investigation to diagnose and detect DILI. While patient-specific genetic factors may increase susceptibility to DILI, the general mechanisms are complex and multifactorial; biotransformation of drugs to reactive metabolites, mitochondrial dysfunction, oxidative stress, altered bile acid homeostasis, and immune response are involved.⁴

This review focuses on DILI caused by altered bile acid homeostasis. In subsequent sections, drug-mediated and disease-mediated impairment of hepatic bile acid transporter function are described as mechanisms of bile acid-mediated DILI. Species differences in bile acid regulation and drug-transporter interactions are emphasized as key considerations for evaluation of bile acid-mediated DILI. Challenges associated with in vitro and in silico model systems used for predicting transporter-mediated DILI liability are outlined. Learnings are summarized from three case studies of hepatotoxic drugs. Finally, perspectives and knowledge gaps in bile acid-mediated DILI risk assessment are discussed.

Mechanisms of DILI and the Role of Hepatic Transporters

BSEP inhibition is a well-studied mechanism of altered bile acid homeostasis and DILI. Several studies in the past decade have described an association between BSEP inhibition by drugs and/or metabolites and DILI occurrence. When assessing DILI risk in vitro, the ratio of the clinical drug exposure and BSEP inhibition potency (IC₅₀) are considered. Drugs with a total steady-state plasma concentration (C_ss or C_max) to BSEP IC₅₀ ratio ≥ 0.1 may be associated with DILI and require further investigation.⁵ Despite the complex physiological processes involved in bile acid homeostasis, BSEP-mediated bile acid excretion is critical to bile flow. Bile salt export pump actively excretes bile acids against a concentration gradient and is the rate-limiting step in hepatic bile acid excretion. BSEP inhibition leads to accumulation of bile acids in hepatocytes with or without impaired bile flow. Other hepatic transporters may also play a role in altered bile acid homeostasis and DILI. MRP2 on the canalicular membrane mediates the excretion of glucuronide-conjugated and sulfate-conjugated bile acids. Several drugs that inhibit BSEP also inhibit MRP2, although MRP2 inhibition provided no predictive benefit for DILI.⁶ MDR3-mediated biliary excretion of phosphatidylcholine drives mixed micelle formation with cholesterol and bile acids, which protects the biliary epithelia. Due to the essential role of MDR3 in bile formation, MDR3 inhibition may be associated with DILI. While some drugs are known MDR3 inhibitors, the role of MDR3 in DILI remains a knowledge gap. Basolateral transporters MRP3, MRP4, and OSTα/β provide an alternate route for bile acid excretion from hepatocytes and are important in maintaining bile acid homeostasis, especially when biliary excretion of bile acids is compromised; impaired function or inhibition of basolateral bile acid efflux transporters may exacerbate DILI. Inhibition of bile acid uptake transporters NTCP, OATP1B1, and OATP1B3 may also be hepatoprotective in the presence of BSEP inhibition, thereby limiting the intrahepatic accumulation of toxic bile acids.^7,8 Based on the Biopharmaceutics Drug Disposition Classification System (BDDCS), class 2 drugs (low solubility and high metabolism) have a higher likelihood of inhibiting BSEP and causing DILI.⁹

In addition to functional inhibition by drugs/metabolites, disruption of regulatory mechanisms involved in the expression of hepatic bile acid transporters must be considered in DILI. Farnesoid X receptor (FXR) is the master bile acid-sensing nuclear receptor that regulates the gene expression of BSEP, NTCP, OATP1B1, OATP1B3, MRP2, and OSTα/β. Bile acid-mediated FXR activation increases BSEP and OSTα/β expression, while lowering NTCP, thereby limiting the intrahepatic bile acid exposure by enhancing efflux and decreasing uptake.¹⁰ Therefore, bile acids act as signaling molecules capable of regulating their own disposition.¹¹ Due to this mechanistic FXR-mediated regulation of bile acid homeostasis, hepatotoxic drugs have been linked with FXR antagonism (e.g., trogitazone, bosentan, tolcapone, indomethacin, ibuprofen, ritonavir).¹² Importantly, the FXR agonist obeticholic acid, used in primary biliary cholangitis (PBC) treatment, increased OSTα/β and BSEP expression and function in human sandwich-cultured hepatocytes (SCH).¹³ Another nuclear receptor associated with DILI is pregnane X receptor (PXR); activation of PXR induces bile acid detoxification via phase I and II enzymes and elimination via MRP2 and MRP3.¹⁴ PXR activation by drugs may be hepatoprotective against bile acid accumulation since cholic acid (CA) feeding induced more hepatic damage in PXR knockout mice than wild-type mice; Mrp3 mRNA and activity were increased in response to pharmacological PXR activation.¹⁵ PXR-mediated DILI is primarily due to upregulation of metabolic pathways; lumiracoxib, troglitazone, acetaminophen, diclofenac, rifampin, and phenytoin are examples of hepatotoxic drugs that are human PXR agonists.¹⁶

Disease-mediated alterations in transporter expression may have profound effects on drug exposure and could predispose patients to DILI. Nonalcoholic fatty liver disease (NAFLD) and its advanced form, nonalcoholic steatohepatitis (NASH), are characterized by hepatic steatosis and inflammation. Conflicting reports show NAFLD may exacerbate or decrease susceptibility to acetaminophen-induced liver damage.^17,18 Based on proteomic data, OATP1B1 and OATP1B3 were decreased, whereas MRP3 and MRP4 were increased in liver tissue from NASH patients.^19,20 In addition, an increased amount of unglycosylated OATP1B1, OATP1B3, OATP2B1, NTCP, and MRP2 protein was discovered in NASH patients, leading to lower membrane localization and/or function of these transporters.²¹ These changes in transporter expression and activity were reflected in increased systemic exposure to morphine glucuronides and 99mTc-mebrofenin in NASH patients.^22,23 Similarly, in hepatitis C cirrhosis patients, protein abundance of NTCP, OATP2B1, MRP2, and MRP4 were decreased while OATP1B3 was increased,²⁴ and in alcoholic hepatitis patients, NTCP, OATP1B1, OATP1B3, OATP2B1, and MRP2 were decreased.²⁰ Pitavastatin and repaglinide showed increased systemic concentrations that correlated with decreased OATP1B1 in patients with cirrhosis Child-Pugh B score.²¹ OSTα/β is upregulated in various cholestatic liver diseases, such as PBC and NASH.²⁵ With advances in quantitative proteomics, disease-mediated alterations in hepatic transporter protein levels have been described that impact drug pharmacokinetics and may increase susceptibility of patients to DILI.²¹

Genetic mutations and polymorphisms of ABCB11 (gene encoding BSEP) cause rare cholestatic diseases, such as progressive familial intrahepatic cholestasis (PFIC2), benign recurrent intrahepatic cholestasis type 2 (BRIC2), and intrahepatic cholestasis of pregnancy (ICP).¹ Functional variants in ABCB4 (gene encoding MDR3) cause PFIC3 and are associated with an increased incidence of gallstone disease, liver cirrhosis, as well as gallbladder and bile duct carcinoma.²⁶ Similarly, mutations and loss-of-function polymorphisms in SLCO1B1/1B3 (genes encoding OATP1B1/1B3) or ABCC2 (gene encoding MRP2) lead to Rotor syndrome and Dubin-Johnson syndrome, respectively, both characterized by hyperbilirubinemia. These mutations affect transporter expression, trafficking, and function, and may increase DILI risk in patients. Notably, decreased ABCB11 expression was reported for the hepatotoxic compounds mithramycin and troglitazone in primary human hepatocytes.^27,28 In rat liver, estradiol 17β-D-glucuronide-induced cholestasis led to the retrieval of Bsep from the apical membrane and a subsequent decrease in Bsep-mediated biliary excretion.²⁹

Key Considerations in Assessing Drug Interactions with Bile Acid Transporters and Bile Acid-Mediated DILI Liability

Bile acids are amphipathic molecules and products of cholesterol catabolism generated in the liver by cytochrome P450 (CYP)-mediated oxidation. Hydrolysis of cholesterol to 7α-hydroxycholesterol by CYP7A1 is the rate-limiting step in the formation of bile acids. Subsequently, CA and chenodeoxycholic acid (CDCA) are synthesized in hepatocytes mainly by CYP8B1 and CYP27A1 enzymes. As a natural ligand, CDCA activates FXR to lower CYP7A1 and CYP8B1 expression and alters bile acid transporter expression to regulate intrahepatic bile acid exposure. Cholic acid and CDCA are conjugated with glycine and taurine by hepatic enzymes bile acid-CoA synthetase (BACS), and bile acid-CoA: amino acid N-acetyltransferase (BAAT).³⁰ Following BSEP-mediated biliary excretion, the human gut microbiome produces secondary bile acids deoxycholic acid (DCA), lithocholic acid (LCA), and ursodeoxycholate (UDCA), which are further conjugated with glycine and taurine.³¹ Key factors in assessing bile acid-mediated DILI liability using animal models are species differences in bile acid homeostasis, as well as bile acid transporter expression and function. Bile acid homeostasis differs between humans and preclinical species in multiple ways: (1) rodents have more polar bile acids due to formation of muricholic acids (MCAs) making the human bile acid pool more hydrophobic than rodents,³² (2) in humans, bile acid synthesis occurs via CYP7A1 and CYP8B1 that are regulated by FXR, whereas, CYP7A1-mediated bile acid synthesis is controlled by liver X receptor (LXR) in rodents,³³ (3) the plasma bile acid pool in humans is predominantly glycine-conjugated (~75%), whereas, mouse, rat, dog, and monkey plasma contain mostly taurine-conjugated bile acids,³⁴ (4) CDCA and CA are the primary bile acids in humans, whereas in rodents, CDCA and CA are further metabolized to MCAs and UDCA,³² (5) the gut microbiota differs between humans and preclinical animal models, (6) interspecies differences exist in hepatic bile acid transporter protein expression, function and inhibition, and (7) rats have no gall bladder to store bile. Owing to these remarkable differences, preclinical animal models are poor predictors of human bile acid-mediated DILI. For example, bile acid quantitation in troglitazone-treated rat and human SCH showed marked species differences in baseline levels; taurine-conjugated bile acids were higher in rat hepatocytes while glycine-conjugated bile acids were higher in human hepatocytes.³⁵ As detailed below, systems pharmacology modeling of troglitazone-mediated hepatotoxicity in rats and humans based only on bile acid effects using DILIsym accurately predicted species differences in troglitazone hepatotoxicity; simulations revealed that hepatic accumulation of toxic bile acids (CDCA and LCA) was ~10-fold higher in humans than rats.³⁶ Total bile acids were also higher in human hepatocytes, despite the presence of the rodent-specific bile acids, α- and β-tauromuricholic acid (α/β-TMCA) in rat hepatocytes.³⁵ Bosentan-mediated BSEP/Bsep inhibition has been reported to increase bile acids in humans and rats, however, hepatic injury is reported only in humans.³⁷ This is partially explained by differential inhibition of NTCP/Ntcp by bosentan⁸ and the more hydrophilic (less cytotoxic) bile acid pool in rats compared with humans. Some bile acid species are more hydrophobic and toxic than others following the rank order LCA > DCA > CDCA > CA > UDCA > MCA.^31,32 The toxicity of bile acids is due to their detergent-like properties that disrupt cell membranes, promote reactive oxygen species formation, and lead to apoptosis. Bile acid-mediated mitochondrial dysfunction, impaired electron transport, and depletion of adenosine triphosphate (ATP) also lead to necrosis.³⁸

Besides the cytotoxic potential and species differences in bile acid homeostasis, several important aspects need to be examined for accurate assessment of DILI risk. First, compensatory function of hepatic basolateral efflux transporters (MRP3, MRP4, and OSTα/β), substrate specificity of transporters, and the adaptive response of hepatocytes to elevated bile acid concentrations may differ across species.⁷ Furthermore, the in vivo drug exposure (total vs unbound) and the effect of protein binding on intracellular and extracellular drug concentrations must be considered. For in vitro safety evaluations, passive permeability, addition of protein to the incubation medium, and transporter inhibition mediated by the parent drug and/or the metabolite(s) are important factors to consider. Accurate DILI predictions require in vitro-to-in vivo extrapolation (IVIVE) of IC₅₀ results generated from in vitro systems; inaccurate conclusions may be drawn with limited data or high uncertainty with measurements.³⁹

Model Systems and Challenges to Assess Transporter-mediated DILI Liability

As discussed above, due to species differences in bile acid homeostasis, substrate specificity, and protein abundance of bile acid transporters, human bile acid-mediated DILI is poorly predicted by preclinical models. In a rat Bsep knockdown model, the three most abundant taurine-conjugated bile acids (tauro-CA [TCA] + TMCA + tauro-UDCA [TUDCA]) increased by 3-fold and 4.5-fold, and glycocholic acid (GCA) decreased by 90% in liver and plasma without hepatotoxicity. Due to several compensatory mechanisms, the rat liver was resistant to direct cytotoxicity from elevated bile acids making extrapolation to human DILI difficult.⁴⁰ In beagle dogs, CA and TCA were sensitive to OATP inhibition by simeprevir complicating their use as BSEP inhibition biomarkers.⁴¹

BSEP inhibition is most widely evaluated using overexpressing inside-out membrane vesicles to examine DILI risk. In some cases, a metabolite may be a more potent inhibitor of BSEP than the parent (e.g., troglitazone sulfate inhibits BSEP more potently than troglitazone).⁴² Membrane vesicles are not metabolically active and fail to capture BSEP inhibition from derived metabolites unless added directly in the assay. Similarly, vesicle-based assays assume competitive inhibition as the mechanism of action. Noncompetitive or indirect mechanisms of inhibition are not accurately measured using typical membrane vesicle screening assays and require additional analyses. In the case of tolvaptan, BSEP inhibition by the parent drug was characterized as noncompetitive, while the metabolite DM-4103 showed competitive BSEP inhibition.⁴³ Therefore, the inhibition potencies of the drug and metabolite(s), and the mechanism(s) of inhibition, are some inherent challenges with standard membrane vesicle assays. Numerous studies have revealed that the membrane vesicle assay can predict human DILI with modest sensitivity (~50%) and specificity (70%−80%) based on different BSEP IC₅₀ cut-off values.⁷ Due to the false-positive and false-negative predictions based on the membrane vesicle assay alone, a two-tiered in vitro approach was proposed to understand BSEP inhibition and DILI liabilities. Compounds with BSEP IC₅₀ ≤ 5 μM in membrane vesicle studies were further evaluated in a long-term human hepatocyte micropatterned coculture (MPCC) or HepatoPac system, which improved specificity (89%), albeit with low sensitivity (high false positive).⁴⁴

The quest for in vitro models suitable to examine bile acid transporter function and predict DILI with the goal of improving drug safety is ongoing. Liver toxicity is predicted only for ~50% of drugs during preclinical animal testing, and often DILI is not identified until clinical development.⁴⁵ This underscores the need for in vitro models that improve the accuracy and reliability of DILI prediction. Inhibition of hepatic uptake transporters is typically evaluated using overexpressing human embryonic kidney (HEK293) or Chinese hamster ovary (CHO) cell lines. Inhibition of hepatic efflux transporters is challenging to study without the formation of bile canaliculi and for compounds with low passive permeability that require active uptake into cells. Hepatic cell lines (HepaRG, HuH-7) often lack expression of functional bile acid transporters and the presence of bile canalicular-like networks. Differentiated HepaRG cells form bile canaliculi, express functional phase I CYP enzymes, phase II metabolic enzymes, and bile acid transporters, such as NTCP, OATP1B1, OATP2B1, BSEP, MRP2, and MRP3, after two weeks of culture.⁴⁶ Similarly, differentiated HuH-7 cells express NTCP, OATP2B1, BSEP, MRP2, and MRP3 with bile canalicular-like networks.⁴⁷ These hepatic cell lines are potentially useful in high-throughput screening of compounds for DILI. However, despite differentiation in vitro, these hepatocellular carcinoma cell lines may be functionally dissimilar to normal hepatocytes. SCH form bile canalicular networks with proper localization and function of apical and basolateral transporters. SCH are metabolically active and retain normal cellular and regulatory function required for expression, localization, and activity of hepatic transporters. SCH are a vital in vitro tool to evaluate DILI risk, especially for compounds with enzyme-transporter interplay.⁴⁸ SCH are widely used to evaluate DILI due to multifactorial mechanisms, including reactive metabolite formation, mitochondrial toxicity, and altered bile acid homeostasis.

Quantitative Systems Toxicology Modeling and Simulation

Considering the complexity of DILI, quantitative systems pharmacology (QSP)/toxicology (QST) models that integrate physiology/pathophysiology with drug exposure and drug-specific mechanistic toxicity data have been developed to predict DILI risk more accurately. DILIsym (https://www.simulations-plus.com/software/dilisym/) is one mechanistic, mathematical QST modeling approach. DILIsym incorporates submodules representing different mechanisms of toxicity as well as nonclinical and clinical data from various sources to predict liver response to a compound, mechanisms contributing to DILI, and potential risk factors for DILI in humans. Two examples using this software platform are briefly discussed below.

Learnings From Troglitazone Hepatotoxicity

Troglitazone, the first-in-class thiazolidinedione drug approved for the treatment of type 2 diabetes, was withdrawn from the market after cases of liver failure were reported in some patients. Numerous mechanisms of troglitazone hepatotoxicity have been reported, but it was unclear why hepatotoxicity occurred in only a small percentage of patients, and why preclinical species exhibited no evidence of liver injury. Using DILIsym, Yang et al³⁶ developed a physiologically based pharmacokinetic (PBPK) model to describe the systemic disposition and hepatic concentrations of troglitazone and its major metabolite, troglitazone sulfate, in humans and rats. Species-specific variability in the bile acid homeostasis sub-model of DILIsym also was incorporated. Simulations predicted that inhibition of bile acid transport by troglitazone and troglitazone sulfate induced delayed hepatotoxicity, consistent with clinical data, due to the accumulation of toxic bile acids in human but not rat hepatocytes. In addition, population analysis allowed identification of potential susceptibility factors for troglitazone-mediated DILI, including the maximum rate of LCA-sulfate biliary excretion and the maximum rate of LCA synthesis in the gut. Using the same modeling approach, the relative liver safety of pioglitazone in humans was correctly predicted. This application of QST modeling provided new insights into species differences in troglitazone-mediated DILI as well as the incidence and delayed presentation of troglitazone hepatotoxicity in humans.

Learnings From Tolvaptan-Associated Liver Injury

Tolvaptan, a selective vasopressin V₂-receptor antagonist used clinically in the treatment of hyponatremia, also slows the decline in kidney function in adults with autosomal dominant polycystic kidney disease (ADPKD). Tolvaptan and tolvaptan metabolites (DM-4103 and DM-4107) inhibit human hepatic bile acid transporters in membrane vesicle systems and human SCH.^43,49 However, no warnings of liver injury were reported with tolvaptan until clinical trials were conducted in patients with ADPKD. Preclinical studies in the PCK rat, a rodent model of polycystic kidney disease, revealed that the biliary excretion of Mrp2 and Oatp substrates was decreased, consistent with decreased protein abundance of these transporters.⁵⁰ The biliary clearance of tolvaptan was decreased significantly in isolated perfused livers from PCK compared with wild-type rats, and hepatocellular concentrations of tolvaptan, DM-4103 and DM-4107 were increased.⁵¹ DILIsym simulations in a virtual human population revealed that impaired biliary excretion of the tolvaptan metabolite DM-4103 substantially increased hepatocyte bile acid accumulation, decreased electron transport chain activity, reduced hepatic ATP concentrations, and increased the incidence of hepatotoxicity.⁵² In vitro studies using C-DILI^™, a cholestatic DILI assay, with human SCH and HepaRG cells confirmed that tolvaptan increased cytotoxicity markers when MRP2 was impaired by pharmacological inhibition or genetic knockout. In this case study, QST modeling predicted that reduced biliary excretion contributes to tolvaptan hepatotoxicity. Preclinical and in vitro data provided insights into disease-related mechanisms that may contribute to increased tolvaptan-associated liver injury.⁵²

Learnings From TKIs: Novel Mechanisms of DILI

Dasatinib, pazopanib, and sorafenib are TKIs that increase liver enzymes and serum biomarkers for hepatotoxicity in ~25% to 50% of patients.⁵³ Previous studies in rodent models identified mitochondrial toxicity, formation of reactive oxygen species, and oxidative stress as mechanisms of toxicity.^54,55 Formation of CYP3A4-mediated reactive metabolites, including a quinoneimine that can react with glutathione and cysteine residues of proteins, has been documented for dasatinib.⁵⁶ Dasatinib inhibited hepatic organic cation transporter 1 (OCT1) function by inhibiting phosphorylation via YES1 kinase.⁵⁷ Human leukocyte antigen (HLA) allele HLA-B*57:01 was reported to confer susceptibility to pazopanib-associated hepatotoxicity in cancer patients, implicating an immune-mediated mechanism of toxicity. However, this HLA allele accounted for ~10% of cases with elevated serum biomarkers, therefore, the presence of other concomitant mechanisms of toxicity was likely.⁵⁸ Sorafenib caused mitochondrial toxicity in HepG2 cells.^59,60 Key species differences were observed with dasatinib, pazopanib and sorafenib hepatotoxicity; dasatinib was more cytotoxic in primary rat hepatocytes than in human, while pazopanib and sorafenib were more cytotoxic in primary human hepatocytes than in rat.⁶¹ The effect of TKIs on hepatic bile acid homeostasis and transport had not been investigated in human SCH. Dasatinib, pazopanib, and sorafenib inhibited BSEP in membrane vesicle studies.^5,62 However, the mechanism(s) of hepatotoxicity associated with TKIs beyond BSEP inhibition were poorly understood.

As reported by Saran et al,⁶³ dasatinib, pazopanib and sorafenib showed bile acid-dependent toxicity at clinically relevant concentrations using the C-DILI^™ assay with human SCH. At these concentrations, the unbound cellular concentrations of dasatinib, pazopanib, and sorafenib were below the reported IC₅₀ for BSEP inhibition. CYP7A1 mRNA was specifically upregulated by dasatinib and pazopanib, which was consistent with increased total bile acid concentrations in culture medium and in human SCH. Protein abundance of NTCP was increased by dasatinib, pazopanib and sorafenib. The increase in NTCP protein abundance was consistent with increased function; dasatinib and pazopanib increased hepatocyte uptake clearance (CL_uptake) of TCA. The increase in CYP7A1 mRNA and NTCP protein abundance, which are novel mechanisms of dasatinib- and pazopanib-associated hepatotoxicity, were independent of FXR in human SCH. Sorafenib increased cellular glycochenodeoxycholate 3-O-glucuronide (GCDCA-3G), suggesting altered MRP function (Figure 2). Data suggested that the mechanism of CYP7A1 mRNA induction and increased bile acid synthesis by these hepatotoxic TKIs was ERK1/2 inhibition. This work explored DILI mechanisms beyond BSEP inhibition and highlighted the complex multifactorial nature of DILI.

Figure 2.

Mechanisms of cholestatic drug-induced liver injury (DILI) by tyrosine kinase inhibitors (TKIs; dasatinib, pazopanib, and sorafenib). Based on the C-DILI^™ assay, dasatinib, pazopanib, and sorafenib showed cholestatic toxicity at clinically relevant concentrations. Mechanisms of DILI investigated included bile salt export pump (BSEP) inhibition, cytochrome P450 7A1 (CYP7A1) mRNA induction, increased gene expression, protein abundance, and function of sodium taurocholate cotransporting polypeptide (NTCP), altered function of multidrug resistance-associated protein (MRPs), and farnesoid X receptor (FXR) antagonism. In sandwich-cultured human hepatocytes, dasatinib and pazopanib increased NTCP membrane protein abundance and uptake clearance (CL_uptake) of taurocholate (TCA) and increased bile acid synthesis via CYP7A1 mRNA induction; dasatinib promoted basolateral efflux of bile acids, presumably via MRP3 and MRP4 and pazopanib increased cellular bile acids; dasatinib reduced the biliary excretion index of TCA, and pazopanib promoted BSEP-mediated biliary excretion of TCA; sorafenib increased cellular levels of glycochenodeoxycholate 3-O-β-glucuronide (GCDCA-3G). FXR antagonism was evaluated using a reporter assay in HepG2 cells, where no antagonism was detected by the three TKIs. This figure was reproduced with permission from the Journal of Pharmacology and Experimental Therapeutics (JPET).

Perspectives and Knowledge Gaps

Currently, liver enzymes used as biomarkers of DILI (AST, ALT, and ALP) lack specificity and sensitivity. Novel biomarkers, such as circulating microRNA and bile acids, are under investigation to improve DILI detection. Liver-specific microRNA, miR-122, was elevated in the blood from early liver damage and is under investigation to predict, monitor, and classify DILI.⁶⁴ In recent years, several studies have reported distinct serum bile acid profiles that may serve as biomarkers to differentiate specific forms of liver injury. Elevated levels of serum CA corresponded with drug-induced hepatocellular injury in rats and conjugated bile acids (GCA and TCA) were increased with bile duct hyperplasia.⁶⁵ Metabolomic studies identified six bile acids that were significantly different in DILI patients compared with healthy controls; GCA and TCA correlated most closely with DILI severity.⁶⁶

More sophisticated, human relevant, in vitro models have been developed recently that show long-term metabolic viability, including MPCC (e.g., Hμrel hepatic co-cultures, HepatoPac). Organoid and spheroid 3D cultures, microfluidic liver models, and liver-on-a-chip models are being developed that may provide further insights into DILI. These advanced in vitro models offer multiple advantages in assessment of DILI liability by mechanisms such as mitochondrial toxicity and reactive metabolite formation. However, transporter expression, membrane localization, and function remain to be thoroughly validated in these systems.⁷ Furthermore, fidelity of regulatory mechanisms found in human hepatocytes needs to be evaluated in advanced in vitro models for future utility in transporter-mediated DILI.⁶⁷

In conclusion, owing to the complex mechanisms of bile acid-mediated DILI, key factors need to be considered when selecting an in vitro or in silico model to assess drug-transporter interaction and predict DILI liability, as highlighted for troglitazone, tolvaptan, and TKIs (dasatinib, pazopanib, and sorafenib). Future studies to identify novel biomarkers and optimize in vitro models will further advance our understanding of bile acid-mediated DILI and the development of safer medications.