PXR-mediated idiosyncratic drug-induced liver injury: mechanistic insights and targeting approaches

Abstract

Introduction:

The human liver is the center for drug metabolism and detoxification and is, therefore, constantly exposed to toxic chemicals. The loss of liver function as a result of this exposure is referred to as drug-induced liver injury (DILI). The pregnane X receptor (PXR) is the primary regulator of the hepatic drug-clearance system, which plays a critical role in mediating idiosyncratic DILI.

Areas covered:

This review is focused on common mechanisms of PXR-mediated DILI and on in vitro and in vivo models developed to predict and assess DILI. It also provides an update on the development of PXR antagonists that may manage PXR-mediated DILI.

Expert opinion:

DILI can be caused by many factors, and PXR is clearly linked to DILI. Although emerging data illustrate how PXR mediates DILI and how PXR activity can be modulated, many questions concerning the development of effective PXR modulators remain. Future research should be focused on determining the mechanisms regulating PXR functions in different cellular contexts.

1. Introduction

As the center of drug metabolism and toxin clearance in the human body, the liver is constantly exposed to endogenous and exogenous chemicals that may cause cell damage. Drug-induced liver injury (DILI) is a term that describes the loss of liver function caused by exposure to a drug or non-infectious agent [1]. DILI is divided into two categories: intrinsic and idiosyncratic [2]. Drugs that belong to the former category are toxic and induce liver injury in a direct, dose-dependent manner. By contrast, drugs that cause idiosyncratic liver damage are often not toxic by themselves, but their interaction with host or environmental factors, such as drug-metabolizing enzymes (DMEs), may result in toxic compounds that induce hepatotoxicity. Consequently, idiosyncratic DILI often presents as an unexpected, dose-independent event, which makes it a major safety concern in drug development [3].

Although the exact incidence of DILI worldwide remains unclear, various studies indicate that the incidence of DILI ranges from 1 to 25 in 100,000 persons per year, with the number being higher in countries using herbal medicines [4–7]. Despite DILI’s low frequency, its importance should not be underrated because it can lead to poor prognosis. In the United States, idiosyncratic DILI is responsible for approximately 10% of acute liver failure cases, and the patient survival rate can be as low as 27% without liver transplantation [8,9]. Additionally, DILI is difficult to predict or prevent, despite intense toxicology studies and clinical trials. A study of 462 post-marketing drug withdrawals showed that hepatotoxicity is the most common reason for drug withdrawal [10]. Even drugs that have been on the market for a long time can be withdrawn because of unexpected DILI: flupirtine, which served as a pain reliever for over 30 years, was withdrawn from the European market in 2018 because of idiosyncratic liver toxicity [11–13]. Therefore, DILI is a potential threat to both patient lives and the pharmaceutical industry.

A variety of factors contribute to the development of idiosyncratic DILI; formation of reactive metabolites is the most common cause [14]. Reactive metabolites can either directly damage cells by irreversibly binding to enzymes, structural proteins, and nucleic acids, or generate oxidative stress that leads to mitochondrial damage [15]. In addition, reactive metabolites may promote the release of inflammatory cytokines that elicit autoimmune responses [15–17]. Interestingly, the primary source of reactive metabolites is the hepatic drug-clearance system, which aims to detoxify and remove harmful xenobiotics. This system is like a double-edged sword because it needs to convert parent drugs into reactive intermediates to facilitate the detoxification of foreign compounds [14].

The pregnane X receptor (PXR) is a member of the nuclear receptor family that transcriptionally regulates the hepatic drug-clearance system, including DMEs that produce reactive metabolites [18]. PXR can be activated by a variety of structurally diverse chemicals owing to its large and flexible ligand-binding pocket. As a result, PXR serves as the master xenobiotic sensor in the liver, where it is mainly expressed. In the absence of xenobiotics, enzymes and transporters in the drug-clearance system are expressed at minimal levels. Upon activation by exogenous ligands, PXR binds DNA and induces the expression of critical components in the drug-clearance system to promote drug metabolism and toxin removal. The most prominent targets of PXR belong to the cytochrome P450 (CYP) 3A family of enzymes, which is responsible for the metabolism of over 50% of drugs on the market [19]. CYP3A enzymes catalyze the first step of drug metabolism, which incorporates oxygen or hydroxyl groups onto parent drugs to increase their reactivity. Consequently, PXR plays an important role in idiosyncratic DILI by regulating CYP3A enzymes, the major contributors of hepatic reactive metabolites.

In this review, we focus on PXR’s role in DILI and provide an update on PXR-mediated DILI in recent publications. We also discuss approaches and cell models used to predict and test PXR-mediated DILI and the possibility of targeting PXR to manage DILI.

2. Mechanisms of PXR-mediated DILI

Many studies have indicated a critical role for PXR in the development of idiosyncratic DILI by either showing direct evidence that PXR activation is associated with enhanced DILI [20–22] or demonstrating that PXR-regulated enzymes are responsible for triggering DILI [23–27]. Although detailed mechanisms vary from case to case, PXR-mediated DILI mechanisms can be categorized based on two criteria: dependence on PXR-regulated DMEs and the presence of drug-drug interactions (Figure 1).

Figure 1.

Mechanistic models for PXR-mediated DILI. Each quadrant contains an example of PXR-mediated DILI with a single drug or a drug-drug interaction that is dependent or independent of DMEs. In the upper left corner, acetaminophen (APAP) overdose leads to accumulation of the toxic intermediate NAPQI and glutathione (GSH) depletion. The subsequent oxidative stress can activate hPXR through poly(ADP-ribosyl)ation by PARP1 to induce CYP expression, thus exacerbating liver injury. In the upper right corner, co-administration of rifampicin (RIF) and ritonavir results in CYP3A4-dependent liver injury through generation of reactive metabolites and endoplasmic reticulum (ER) stress. In the bottom left corner, RIF-mediated activation of hPXR can induce expression of SREBP1 and its target genes, ultimately leading to lipid accumulation and steatosis. In the bottom right corner, co-administration of RIF and isoniazid results in upregulation of ALAS1, which produces hepatotoxic PPIX, and downregulation of ferrochelatase (FECH), which converts PPIX to heme. Consequently, PPIX accumulation in hepatocytes causes liver injury over time.

Most PXR-mediated DILI cases involve the upregulation of the hepatic drug-clearance system. In liver cells, the drug-clearance system metabolizes and detoxifies foreign compounds in two phases [28,29]. In the first phase of metabolism, lipophilic parent drugs are converted into more-hydrophilic reactive intermediates through oxidation or hydroxylation. The enzymes responsible for phase I reactions are predominantly CYP family enzymes [28,30]. Upon activation by an agonistic drug, human PXR (hPXR) upregulates the expression of CYP 2A, 2B, 2C, and 3A family enzymes to metabolize the drug into reactive intermediates that may be cytotoxic [14,29,31]. For example, lumiracoxib and troglitazone were both withdrawn from the market because of CYP-induced hepatotoxicity. The metabolism of both drugs by CYP2C and CYP3A4 enzymes produces reactive quinone intermediates [32,33], which are well-established cytotoxic agents that induce oxidative stress, mitochondrial lesion, and DNA damage [34]. In fact, CYP-dependent reactive metabolite formation is a common reason for DILI associated with many other drugs, including acetaminophen [35], diclofenac [36], and phenytoin [37].

In the second phase of drug metabolism, transferases conjugate reactive metabolites with polar functional groups to detoxify these compounds and increase their solubility in water, thereby allowing products to be exported into the urinary and biliary systems for excretion [14,31]. hPXR regulates the expression of several key phase II enzymes, including uridine-5′-diphosphate glucuronosyltransferase family 1 member A1 (UGT1A1) [38], sulfotransferases [39–41], and glutathione S-transferase [42]. Although products of phase II drug metabolism are generally less toxic and less reactive than those of phase I drug metabolism, some enzymes generate toxic metabolites that contribute to liver injury. For instance, UGT1A1 catalyzes the conjugation of glucuronate groups to drug metabolites [43]. The resulting glucuronidated products have long been postulated to non-specifically bind proteins and promote idiosyncratic DILI, but no conclusive evidence had been found [44]. However, a recent study demonstrated that trovafloxacin acyl-glucuronide, a metabolite of the antibiotic trovafloxacin, decreased cell viability and induced production of cytokines responsible for DILI, implicating UGT1A1 in the development of DILI [45]. Nonsteroidal anti-inflammatory drugs, such as naproxen, diclofenac, and ibuprofen, were also proposed to induce hepatotoxicity by the same mechanism [24,46].

Besides direct toxicity from the prescribed drug, unexpected drug-drug interactions can also elicit DILI via PXR activation. Although some compounds that are strong hPXR activators do not induce liver injury by themselves, co-administering these drugs with other therapeutics may lead to adverse drug reactions. Rifampicin (rifampin) is a potent hPXR agonist that is frequently used in anti-tuberculosis treatment. Despite its low toxicity, co-administration of rifampicin with certain drugs is restricted because rifampicin activates hPXR and increases the risk of DILI [47–49]. Accordingly, treating patients who have both tuberculosis and diseases such as human immunodeficiency virus (HIV) is especially challenging because co-therapy of rifampicin and anti-retroviral drugs is essential for recovery [47,50]. Multiple studies have reported that patients treated first with rifampicin or efavirenz (another PXR agonist) followed by ritonavir have a higher-than-usual incidence of DILI [51–54]. A subsequent study using a humanized mouse model showed that rifampicin triggers ritonavir-induced liver injury through a PXR-CYP3A4–dependent pathway (Figure 1) [55], further demonstrating PXR’s essential role in mediating DILI.

Apart from its central role in drug metabolism, PXR also regulates genes involved in various cellular pathways, including the cell cycle, inflammatory response, and angiogenesis pathways [56]. Therefore, overexpressing these genes upon PXR activation may lead to adverse drug reactions unrelated to drug metabolism, as in the case of anti-tuberculosis co-therapy with rifampicin and isoniazid (Figure 1). Activation of hPXR by rifampicin upregulates the expression of alanine synthase I (ALAS1), which is the rate-limiting enzyme that produces hepatotoxin protoporphyrin IX (PPIX), the toxic intermediate in heme biosynthesis [57]. Concurrently, isoniazid downregulates the expression of ferrochelatase, the enzyme that catalyzes the conversion of PPIX to heme [58]. As a result, PPIX accumulates in hepatocytes, where most hPXR is expressed, and causes liver injury over time. Accordingly, co-administering rifampicin and isoniazid led to higher PPIX concentrations and liver damage in mice expressing hPXR than in PXR-null or wild-type mice [21].

After the initial cellular injury, PXR-mediated DILI may occur through downstream events, such as mitochondrial damage and initiation of necrotic or apoptotic cell death [59]. For example, acetaminophen overdose results in glutathione depletion and accumulation of the toxic intermediate, N-acetyl-p-benzoquinone imine (NAPQI), that is generated by CYP1A2, CYP2E1, and CYP3A4 [35,60]. As NAPQI binds to liver proteins, its buildup leads to mitochondrial dysfunction and hepatocyte necrosis (Figure 1) [61]. In addition, the oxidative stress resulting from glutathione depletion activates poly (ADP-ribose) polymerase 1 (PARP1) in response to cell death initiation [62]. A recent study showed that PARP1 directly interacted with the mouse PXR (mPXR) ligand-binding domain to robustly activate mPXR through poly(ADP-ribosyl)ation, inducing its recruitment to the response elements at the Cyp3a11 promoter. This posttranslational modification of mPXR, in return, exacerbates acetaminophen metabolism and liver toxicity [63].

Beyond chemical disposition, PXR-mediated DILI may also occur through the disruption of lipid homeostasis. PXR activation can induce sterol regulatory element binding protein 1 (SREBP1) in human hepatocytes, leading to the expression of SREBP1 target genes and subsequent accumulation of liver triglycerides (Figure 1) [64]. Consequently, active PXR enhances liver lipid synthesis and fatty acid uptake, resulting in lipid accumulation and steatosis [65]. In further support of PXR disrupting lipid homeostasis, wild-type mice exposed to amprenavir, a widely used HIV protease inhibitor and potent agonist of both human and mouse PXR, showed significant increases in total cholesterol in plasma and atherogenic low-density lipoprotein cholesterol levels. By contrast, these changes were not apparent in PXR-deficient mice, indicating a PXR-dependent mechanism [66]. Consistent with in vivo studies, computational docking and site-directed mutagenesis approaches indicated direct binding of amprenavir to the ligand-binding pocket of hPXR [66]. In another study using mouse models, mPXR activation reduced peroxisome proliferator-activated receptor α (PPARα) activity, induced strong inhibition of plasma levels of hepatokine fibroblast growth factor 21, and suppressed more than 25 PPARα gene targets [67]. Taken together, these studies indicate that PXR activation by drugs can potentiate hepatic steatosis and liver injury.

3. Update on the examples of PXR-mediated DILI

Many clinically prescribed drugs promote DILI by activating hPXR, and their mechanisms have been summarized in a previous review article [68]. This section provides an update on new DILI cases associated with PXR since 2014.

3.1. Ritonavir

Ritonavir is an antiretroviral protease inhibitor commonly used to treat HIV infection; rarely, full doses of ritonavir produce clinically apparent liver injury [69–72]. The toxicology of ritonavir-induced liver injury is interesting, as ritonavir is both a potent agonist of hPXR and inhibitor of CYP3A enzymes [73–75]. Cell-based assays indicate that ritonavir activates hPXR and upregulates the expression of CYP2B6 and CYP3A4, but ritonavir also strongly inhibits CYP3A4 activity such that the inhibition supersedes induction, resulting in net lowering of CYP3A4 activity [74–77]. By inhibiting CYP3A enzymes, ritonavir prolongs the exposure time (i.e., plasma concentration-time curve or area under the curve) of co-administered drugs that rely on CYP3A metabolism. Although this effect is beneficial for increasing bioavailability of some co-administered drugs, such as saquinavir [78], the increased serum levels may also contribute to hepatotoxicity and liver injury [69,70]. The potential for serious, life-threatening reactions resulting from drug-drug interactions led to a black box warning on ritonavir [79,80].

Apart from its drug-drug interactions, ritonavir also has toxic intermediates that may contribute to liver injury. Despite being a potent CYP3A4 inhibitor, ritonavir is metabolized into reactive intermediates by CYP3A4 [81–83]. A recent study using humanized-PXR mouse models found that rifampicin-activated hPXR upregulates CYP3A4 expression and potentiates ritonavir hepatotoxicity [55]. Activation of the unfolded protein response by ritonavir has also been reported [84–86]. Such activation likely occurs following the initial cellular injury by toxic metabolite accumulation. Furthermore, ritonavir inhibition of efflux bile acid transporters, such as SLC51A and ABCB4, may be responsible for the cholestatic pattern of liver injury [86].

3.2. Bromuconazole

Bromuconazole is a fungicide that is used on food crops and fruits. Although the risk of toxicity and carcinogenicity in humans exposed to relevant levels of bromuconazole is considered low [87], chronic exposure in male rats leads to liver toxicity, followed by hepatocellular and cholangiocellular carcinomas [22,87,88]. These rats exhibited significant increases in serum levels and activities of liver enzymes, including alanine transaminase, aspartate transaminase, alkaline phosphatase, and acid phosphatase, which are indicative of hepatocellular injury [22,88]. Additionally, bromuconazole caused an increase in liver weights and necrobiotic changes (vacuolation and hepatocellular hypertrophy), likely the result of increased reactive oxygen species causing hepatic oxidative stress [22,88]. Interestingly, bromuconazole increases the expression and activity of rat PXR and its downstream target rat CYP3A1, which has high homology with human CYP3A4 [22,89]. Although increased reactive oxygen species production and toxic metabolites may cause bromuconazole-induced liver injury, studies in human model systems need to be conducted to better understand the metabolic pathway of bromuconazole and to identify potentially toxic metabolites.

3.3. Flupirtine

Flupirtine is a non-opioid analgesic with skeletal muscle–relaxing activity that was licensed in Europe, China, and Russia for the treatment of acute and chronic pain [11,90]. Although flupirtine-induced liver injury is rare, flupirtine was the most implicated drug in DILI cases in Germany, where it was mainly produced and prescribed. RUCAM (Roussel Uclaf Causality Assessment Method) was used to assess causality in patients with suspected DILI from many top-ranking drugs, including flupirtine [91,92]. Results obtained from the blood of patients with DILI confirmed by causality assessment for specific drugs can be used to guide further mechanistic studies. As mentioned earlier, flupirtine was withdrawn from the market in 2018 due to the persistent occurrence of severe liver injury and failure despite the attempts of European Medicines Agency to restrict its use to short-term acute pain [12]. The clinical presentations of flupirtine-induced idiosyncratic liver injury in six Caucasian patients included jaundice, hepatic encephalopathy, coagulopathy, and elevated serum aminotransferase levels. Upon discontinuation of flupirtine, their serum amino-transferase, alkaline phosphatase, and bilirubin levels fell rapidly and normalized after several weeks. Histopathological analysis of liver biopsies from these cases showed perivenular (zone 3) necrosis, which is commonly seen with intrinsic and idiopathic hepatotoxins, and an abundance of ceroid (or “pigment-laden”) macrophages [11].

Although flupirtine is not a substrate of the PXR-regulated efflux transporters ABCB1 and ABCC2 [93], PXR may be involved in detoxification of flupirtine through regulating expression of phase II drug-metabolizing enzymes. In the liver, flupirtine is first hydrolyzed by carboxylesterase 2 or arylacetamide deacetylase and then acetylated by N-acetyltransferases to form an acetylamino product named D13223. During this process, the hydrolyzed intermediate can be converted into a reactive metabolite via non-enzymatic conversion [94]. This metabolite, which is likely a quinone diimine, is the prime candidate for causing flupirtine’s hepatotoxicity, as quinones are highly reactive and well-known to cause cell toxicity [34,93,95]. In support of the hypothesis that the quinone diimine metabolite triggers hepatotoxicity, cytotoxicity was observed in HepG2 cells overexpressing carboxylesterase 2, but not in control cells, after flupirtine treatment [94]. As detoxification of quinone diimine can occur by glutathione conjugation, activation of PXR, which upregulates the expression of phase II enzyme glutathione S-transferase, may alleviate DILI caused by flupirtine.

4. In vitro and in vivo models developed to predict DILI

Traditionally, in vitro studies to assess DILI have relied on liver slice–based assays, liver microsomes, immortalized hepatic cell lines, and primary human hepatocytes (PHHs), but each of these methods has its limitations. Although liver slices retain tissue architecture and all liver cell types, their functionality rapidly declines when placed in static culture medium for a few days [96]. Liver microsomes, which are enzymatic preparations made from liver homogenates, are the model of choice to assess CYP- and UGT-related drug metabolism and to predict hepatic clearance. However, loss of enzyme activity after 1 hour necessitates shorter incubation time periods in microsomal-based studies, making liver microsomes inappropriate for assessing metabolism and clearance of drugs with slow turnover [97,98]. In contrast, immortalized hepatic cell lines, such as HepG2, Hep3B, Fa2N4 and HuH7, can be maintained in large quantities and for long periods of time; however, these cell lines do not maintain major liver functions and exhibit low hPXR inducibility and low CYP expression and activity [99,100]. Conversely, PHHs are scarce and not adaptable to long-term culture and repeat drug exposures. In addition, data from PHHs vary greatly due to interindividual differences in donor responses to drug exposures. A promising alternative to PHHs is the bipotent hepatic progenitor HepaRG cells, which differentiate into hepatocyte-like and cholangiocyte-like cells that retain hPXR expression levels and hPXR-mediated CYP activities that are similar to those of PHHs [101]. The major limitation of HepaRG cells is the long 28-day differentiation culture process. Induced pluripotent stem cells (iPSCs) are another appealing alternative to PHHs, as they are renewable, can be isolated from patients susceptible to DILI, and can be differentiated into hepatic cells [102].

Recently, more advanced in vitro PHH co-culture assay platforms have been introduced to address hepatic cell line limitations by enabling long-term maintenance and more complex drug mechanism studies. To better address interindividual responses to drug metabolism and systemic organ toxicity resulting from liver drug metabolism, investigators have developed a spectrum of engineered liver models that include static two-dimensional micropatterned co-cultures, three-dimensional (3D) static spheroids, perfused biochips, and precision-cut liver slices [103]. Another type of in vitro platform that was recently developed is the use of perfusable 3D bioreactor cell co-cultures that mimic natural tissue conditions by inducing zonal hepatic functions to allow for better nutrient exchange and waste product removal [96]. In these co-cultures, adult PHHs interact with various liver nonparenchymal cell types, such as Kupffer cells, liver sinusoidal endothelial cells, hepatic stellate cells, and biliary cells, which aid in increasing hepatic functions when exposed to flow [96]. These perfusable 3D human liver co-culture platforms are useful for studying PXR’s activation and regulation of its target genes for several weeks with chronic drug treatments [104]. For example, co-cultures of cryopreserved PHHs and Kupffer cells in the 3D microreactor platform recapitulated clinical observations better than traditional hepatic cultures, including suppression of CYP3A4 activity by interleukin 6 (IL-6) and the reversal by Tocilizumab, an anti-IL-6R monoclonal antibody used to treat rheumatoid arthritis [104].

Although in vitro systems express DMEs that may indicate possible drug disposition mechanisms, they are not suitable for modeling the multi-step, complex, tissue-dependent drug metabolism that occurs in the human body [105]. As a result, rodent models are still the gold standard for in vivo drug metabolism studies. Because species-specific differences in ligand activation of hPXR and mPXR are well documented, PXR-humanized mouse models have been developed to more accurately model human drug metabolism in vivo [106–108]. One humanized PXR mouse model was produced by knocking in hPXR cDNA at the ATG of murine Pxr in C57BL/6NTac-derived embryonic stem cells, effectively deleting the endogenous Pxr while placing hPXR expression under the control of the mouse Pxr promoter [109]. In addition, a double-humanized mouse model was generated by knocking in hPXR and human constitutive androstane receptor (hCAR) genes to investigate the hepatic regulation of CYP and UGT [110]. These models were further improved by replacing mouse Cyp genes with their human counterparts to better mimic human metabolic conditions [111]. For example, one model introduces the human CYP3A4, CYP3A7, PXR, and CAR genes into the Cyp3a-null (except for Cyp3a13) mouse [112], and another multiple-humanized mouse model incorporates human CYP2C9, CYP2D6, CYP3A4, CYP3A7, PXR, and CAR genes [113]. The major drawback of these humanized mouse models is that the expression of human PXR and CYP genes is still in the mouse context, which may lead to unexpected regulation that complicates data interpretation.

In the last few years, use of in silico simulations and high-content assays has rapidly increased to predict the risk factors of DILI in the early stages of drug design. Bioinformatics, which compensates for the shortcomings of in vitro and in vivo models, enables deeper exploration of DILI mechanisms, minimizes the time frame for choosing drug targets, and makes large biological datasets more manageable [114]. In addition, artificial intelligence technologies using advanced machine learning models are currently being developed and tested to better identify molecular markers of DILI and to classify drugs as DILI-causing or non–DILI-causing [115]. In 2016, the Food and Drug Administration (FDA) published DILIrank, the largest known database that ranks drugs on the basis of their potential to cause DILI, which was recently expanded to encompass additional drugs [115]. Despite these technological advances, more human-like in vivo and in vitro models that more closely simulate the liver are needed to further improve toxicity predictions during drug development.

5. Managing DILI by targeting PXR

Because PXR contributes to DILI by inducing the expression of DMEs, which may generate toxic metabolites, a reasonable approach to managing DILI is to lower PXR-mediated transcriptional activation. This goal can be achieved by chemically modifying drugs to minimize binding to or activation of hPXR [116]. However, this strategy is challenging because the ligand-binding pocket of PXR is highly flexible and can accommodate many structurally diverse compounds [117,118]. In some cases, small chemical modifications may turn a potent antagonist into an agonist [119]. An alternative approach is to develop inhibitors and antagonists that prevent ligand-dependent PXR activation. Co-administering these compounds would not only attenuate PXR-mediated DILI but also enhance therapeutic efficiency as drugs would be less likely to be metabolized by hepatic enzymes. Compared to inhibitors (inverse agonists) that hinder PXR function in the absence of ligand, antagonists are more advantageous because, by themselves, they have little effect on PXR’s basal activity.

Although many agonists are known to activate hPXR, significantly fewer hPXR inhibitors and antagonists have been reported. A review paper published in 2014 provided a comprehensive summary of these inhibitors/antagonists [68], many of which have various drawbacks that make them inappropriate for clinical use (Table 1). For instance, ketoconazole was reported to be a potent hPXR antagonist in vitro [120,121], but it fails to antagonize hPXR activation in vivo at non-toxic concentrations [122]. Camptothecin was reported to be a hPXR inhibitor with a sub-micromolar IC₅₀ [123], but it has poor water solubility and low stability under physiological conditions [124]. These drawbacks might be difficult to overcome by chemical modifications, as evidenced by the unexpected hPXR agonist activity of two camptothecin analogs, irinotecan and topotecan [123,125]. Consequently, a potent, specific, and non-toxic hPXR antagonist is still needed.

Table 1:

Summary of hPXR antagonists and their properties

Name	IC50	Toxicity	Selectivity	Major Genes Repressed	Reference
SPB03255	6.3 μM	ND	ND	ND	[143]
SPB00574	24.8 μM	ND	ND	ND	[143]
Leflunomide	6.8 μM	Increased risk of liver injury at >100 μM	ND, mainly used as anti-rheumatic drug	ND	[143,144]
Coumestrol	12 μM	ND	Acts as ER agonist and hCAR inverse agonist	CYP3A4, CYP2B6	[143,145]
Camptothecin	0.58 μM	Highly cytotoxic	Also inhibits hCAR, induces hVDR activity	CYP3A4	[123]
Polychlorinated biphenyls 197 (PCB 197)	0.6 μM (Ki)	Increased risk of cancers, developmental problems, and adverse reproductive effects	ND	CYP3A4, AhR, AGT1A1, MDR1	[134]
Ecteinascidin 743 (ET 743)	2 nM	Highly cytotoxic; hematologic and hepatic toxicity at high doses	Has other biological effects (e.g. binds DNA minor groove)	CYP3A4, MDR1	[133,135,146]
A-792611	≈ 2 μM	ND	Little or no activation of other nuclear receptors detected	CYP3A4, CYP2B6, CYP2C8, CYP2C9, MDR1	[147]
Sesamin	>30 μM	Low cytotoxicity	Also inhibits hCAR	CYP3A4	[148]
Ketoconazole	18.73 μM	Idiosyncratic organ toxicity and cellular cytotoxicity in vitro at >50 μM	Also inhibits hCAR, hGR, hAR, hLXR, hFXR	CYP3A4, MDR1	[121,149,150]
Metformin	>1 mM	Selective cytotoxicity	Also inhibits CYP3A4 interaction mediated by hVDR, hGR, hCAR	CYP3A4	[151,152]
Sulforaphane	12 – 14 μM	Low cytotoxicity	Little or no activation of other nuclear receptors detected	CYP3A4	[143,153]
SPA70	510 nM	Low cytotoxicity	Highly selective for hPXR	CYP3A4, MDR1, and many others	[119]
Pimecrolimus	1.2 μM	Non-toxic (up to 50 μM)	Slightly activates CAR1, CAR3	CYP2B6, CYP2C8 and many others, but not CYP3A4	[126]
Pazopanib	4.1 μM	Non-toxic (up to 50 μM)	Inhibits multiple nuclear receptors	AKR1B10	[126]
Resveratrol	ND	Low cytotoxicity	ND	CYP3A4	[128,131]

ND = Not Determined

Recently, high-throughput screening (HTS) of an FDA-approved drug library identified two drugs, pimecrolimus and pazopanib, as potent hPXR antagonists [126]. Pimecrolimus is an immuno-modulating drug that is used to treat inflammatory skin diseases [127]. Cell-based assays revealed that pimecrolimus antagonizes rifampicin-induced hPXR activation with an IC₅₀ of 1.2 μM, and proteolysis digestion indicated that the compound directly binds hPXR in its ligand-binding pocket. In PHHs, pimecrolimus reduced mRNA levels of most hPXR-regulated genes, except for CYP3A4, upon hPXR activation. Co-factor interaction assays indicated that the inhibition was achieved by blocking the recruitment of co-activator SRC-1. Notably, pimecrolimus demonstrated characteristics of both competitive and non-competitive antagonists, as low concentrations (<3 μM) only shifted the dose-dependent curve of rifampicin-induced hPXR activation; yet, 10 μM treatment affected the EC₅₀ and reduced the maximal effect of rifampicin-induced activation. However, this observation could be attributed to the lack of a complete dose-response curve.

The other hPXR antagonist identified by HTS, pazopanib, is a tyrosine kinase inhibitor used to treat renal cell carcinoma and soft-tissue sarcoma. The luciferase reporter assay showed that pazopanib antagonizes rifampicin-induced hPXR activation with an IC₅₀ of 4.1 μM [126]. Similar to pimecrolimus, pazopanib is not toxic up to 50 μM and inhibits hPXR activity by preventing the recruitment of co-activator SRC-1. Additionally, pazopanib acts as a non-competitive antagonist, as it affects the maximal effect, rather than the EC₅₀, of rifampicin-induced hPXR activation. However, pazopanib only reduced AKR1B10 gene expression in a cell-based assay, and there was no evidence to support direct binding between pazopanib and hPXR. Therefore, it is unclear whether pazopanib is a true hPXR antagonist.

Resveratrol is a natural product and food supplement that was shown to weakly antagonize PXR activation induced by rifampicin in human cells and by pregnenolone 16a-carbonitrile (PCN) in mice [128]. In the human LS174T cell line that overexpressed hPXR, treatment with >25 μM resveratrol significantly lowered rifampicin-induced CYP3A4 expression, but resveratrol alone had a minimal effect on hPXR basal activity. Similarly, 25 or 50 μM of resveratrol suppressed PCN-induced mPXR activation and subsequent Cyp3a11 expression [128]. A corresponding publication confirmed the interaction between resveratrol and hPXR by using time-resolved fluorescence energy transfer, but the authors proposed that resveratrol is a weak agonist that induces antagonistic effects by competing with strong agonists, such as rifampicin and PCN [129]. Despite its low potency, resveratrol’s low cytotoxicity and abundance in food products are appealing aspects of it serving as a potential hPXR antagonist [130,131].

SPA70 is another potent hPXR antagonist that was identified through HTS [119]. Combined cell-based and biochemical assays indicated that SPA70 binds and antagonizes hPXR with high affinity and selectivity (compared to 10 other nuclear receptors). In HepG2 cells, SPA70 inhibited rifampicin-induced hPXR activation with an IC₅₀ of 510 nM, although the compound itself only weakly inhibited basal activity of hPXR. Direct binding measurements revealed that SPA70 binds hPXR with a sub-micromolar IC₅₀. In addition to its high selectivity, SPA70 exhibited marginal cytotoxicity across multiple cell lines. Furthermore, SPA70 blocks the expression of a broad range of hPXR-regulated genes induced by three hPXR agonists, rifampicin, SR12813, and T090137. The antagonistic effects of SPA70 were also observed in a humanized PXR mouse model. Further mechanistic studies indicated that SPA70 enhances the interaction of hPXR with co-repressors, such as SMRT and mNCoR [119]. Combined structural, mutagenesis, and computational analyses suggest that SPA70 does not keep the AF-2 helix in an active conformation [119,132].

Compared to other reported antagonists, SPA70 is a better drug candidate for preventing and treating PXR-associated DILI. In addition to being well-characterized by biochemical, cell-based, and in vivo assays, SPA70 is the only antagonist that combines the desirable characteristics of high potency and selectivity with low toxicity. Among other known PXR antagonists, only ecteinascidin 743, polychlorinated biphenyls, and camptothecin demonstrated sub-micromolar IC₅₀ values (Table 1) [123,133,134]. However, these three compounds exhibit significant cytotoxicity and adverse health impacts, limiting their use as therapeutic drugs [134–136]. By comparison, SPA70 does not demonstrate significant toxicity under 30 μM in HepG2, HEK293, or Hepa 1–6 cell lines [119]. Pimecrolimus was shown to antagonize PXR activation with an IC₅₀ of 1.2 μM [126], but it does not inhibit the upregulation of CYP3A4, the major drug-metabolizing enzyme that contributes to DILI. Most other well-studied PXR antagonists, such as ketoconazole, lack either potency or selectivity. A recent in vivo study demonstrated the potential therapeutic effect of SPA70 on PXR-mediated DILI. Using a humanized mouse model, the authors demonstrated that the activation of hPXR by rifampicin promotes hemorrhagic shock–induced liver injury; however, SPA70 attenuated liver injury by antagonizing rifampicin-induced hPXR activation [137].

6. Conclusion

DILI is a relatively rare, but life-threatening adverse drug effect that impacts both patients and the pharmaceutical industry. Despite numerous studies, our understanding of idiosyncratic DILI is still incomplete. In the past two decades, several publications have indicated or suggested the role of PXR in mediating DILI. PXR is the primary xenobiotic sensor in the liver that regulates the expression of drug-metabolizing enzymes. However, enhanced drug metabolism as a result of PXR activation is a double-edged sword: while PXR activation promotes toxin clearance, it also leads to the accumulation of reactive intermediates that may be cytotoxic at high doses. In addition, PXR activation may cause unexpected drug-drug interactions because PXR also regulates other non-drug-metabolizing enzymes that are important for various cellular functions. Therefore, understanding the effects of drugs on PXR is important for the prediction and treatment of PXR-mediated DILI.

Recent development of in vivo and in vitro models facilitates better prediction and assessment of PXR-mediated DILI. Novel 3D cell cultures mimic the liver environment better than two-dimensional cultures do, and improved humanized mouse models enable simultaneous evaluation of hPXR activity and DILI. Bioinformatics approaches and databases such as DILIrank are instrumental in predicting DILI risk factors. All of these new tools and strategies will be beneficial for future studies investigating the mechanisms of PXR-mediated DILI.

PXR antagonists, which only inhibit agonist-induced PXR activity, are promising chemicals for attenuating PXR-mediated DILI. PXR antagonists not only prevent the development of DILI but also enhance the therapeutic effects of other drugs by suppressing the activity of hepatic drug metabolism. However, all known hPXR antagonists, except for SPA70, have drawbacks in either potency, cytotoxicity, or selectivity. Therefore, future efforts are needed to develop therapeutics that manage PXR-mediated DILI.

7. Expert opinion

Many additional works may contribute to our understanding of the effect, mechanisms, and approaches to manage DILI. For example, thorough characterization of a drug’s metabolites may help identify which are toxic and contribute to DILI. Additionally, developing novel tools, such as proteolysis-targeting chimeras, to degrade PXR might provide alternatives to inhibit PXR and a potential DILI therapeutic option.

In regard to PXR antagonists, how PXR antagonists bind and inhibit PXR activity warrants further investigation. Uncovering mechanistic information will greatly facilitate the development of new PXR antagonists. Although hPXR antagonists are promising drug candidates to manage PXR-mediated DILI, the development of potent, selective, and low-toxicity antagonists is challenging because most compounds serve as agonists of PXR. As a result, identifying new hPXR antagonists requires large-scale HTS that is resource-intensive. Alternatively, chemical modifications of existing antagonists may be used to improve their potency and selectivity. However, the structural basis of PXR antagonism remains largely unknown, and current attempts have not produced analogs with improved activity in inhibiting PXR activity [119,123]. Therefore, understanding the mechanism of antagonist action will greatly facilitate the design of novel PXR antagonists.

In contrast to PXR inhibition, drug-induced PXR activation is not always our enemy, even though it can lead to unexpected liver injury as a result of enhanced drug metabolism. When a drug is not metabolized by CYP enzymes, PXR activation and the subsequent upregulation of phase II DMEs may promote the removal of toxic metabolites. In the case of flupirtine, PXR activation may facilitate the detoxification of quinone by-products by promoting the expression of glutathione S-transferase. However, PXR inactivation by antagonists prolongs the exposure of liver cells to parent drugs, which may lead to “black box” effects, such as direct toxicity. Therefore, we should recognize and consider the positive role of PXR in DILI when developing co-therapy with PXR antagonists.

Most drug-induced hepatotoxicity is considered idiosyncratic, with highly affected individuals being predisposed to DILI. Consequently, it is challenging to predict the potential for DILI in early stages of drug design and discovery. Individual susceptibility to DILI is critically influenced by genetic, epigenetic, and environmental factors [138]. For this reason, certain DILI-causing drugs are only detected after wide administration within a population. Co-therapy by far poses the highest risk of PXR-related DILI because PXR, as a master regulator of xenobiotic metabolism, is highly promiscuous and binds to structurally diverse chemical compounds that may lead to activation or inhibition of its gene regulation activity. It is widely accepted that the initiating event in DILI is the metabolism of a parent drug, leading to accumulation of hepatotoxins. In this case, the level of PXR expression within hepatocytes of an individual plays a major role in the treatment outcomes.

PXR expression is highly influenced by an individual’s genetic makeup, environmental exposure, and presence of pathological conditions. The characterized PXR isoforms (PXR1, PXR2, PXR3 and PXR4) [139] are results of alternative promoters and mRNA splicing that lead to 5′-region and coding-region diversity. PXR1 is the major isoform used in most PXR studies. PXR1 and PXR2 can bind ligands and activate PXR target genes, but PXR3 and PXR4 do not induce gene expression [139]. In addition, PXR1 and PXR2 show variability in their promoter regions, their protein interactions, and their cancer cell phenotypes. A naturally occurring 6 base pair–deletion (−133)GAGAAG(−128) in PXR2’s promoter region has been reported to downregulate PXR activity in HepG2 cells, and investigations showed a correlation between reduced PXR activity and reduced CYP3A4 and MDR1 transporter expression [140]. Further research showed that this deletion could predispose individuals to hepatic carcinoma due to decreased liver detoxification ability. PXR4 results from an alternative transcription starting site downstream of a CpG island near exon 3 of the hPXR gene. This isoform, which is homologous to the ligand-binding domain of both PXR1 and PXR2, is capable to bind ligands and coregulators. However, it lacks gene transactivation activity despite being detected in the nucleus, therefore functioning as a dominant-negative for PXR1 in the liver [141]. A study on multiple human liver samples indicated high inter-individual variance in the expression level of PXR4, and it was reported to be associated with favorable outcomes of hepatocellular carcinoma treatment by resection [141]. Further research into regulation and function of PXR isoforms is necessary to enhance DILI prediction models.

We also want to point out that the mechanistic studies of idiosyncratic DILI are challenging, because many different mechanisms might be involved, including those mediated by the adaptive immune system. This is beyond the scope of this review, but has been discussed recently by Uetrecht [142].

Article highlights.

• Idiosyncratic DILI is a rare, but life-threatening adverse drug reaction caused by unexpected interaction between the drug and environmental factors.

• PXR plays a critical role in mediating DILI because of its regulatory function in the hepatic drug metabolism.

• When PXR is activated by xenobiotics, it induces the expression of drug-metabolizing enzymes (DMEs), which may lead to the accumulation of reactive metabolites of parent drugs that are cytotoxic at high doses.

• PXR activation can also trigger DME-independent downstream events that cause cytotoxicity.

• PXR antagonists, which only inhibit agonist-induced PXR activity, may manage PXR-mediated DILI.