IFNγ.

IFNγ is a cytokine that exists as a soluble dimer encoded by the IFNG gene. It is not produced in substantial amounts by hepatic parenchymal cells but is expressed and secreted by several immune cell types, including CD4+ T helper (Th) cells (i.e., Th1 cells), CD8+ cytotoxic T cells, natural killer T (NKT) cells, natural killer (NK) cells, and eosinophils. These cells can be activated to express and release IFNγ by cytokines such as interleukin (IL)-12, IL-18, and IL-27 (Trinchieri and Scott, 1995; Okamura et al., 1998; Batten and Ghilardi, 2007). Moreover, interaction among different types of leukocytes occurs; for example, using CD1 to present glycolipid stimuli, NKT cells can activate NK cells to produce IFNγ (Carnaud et al., 1999; Hayakawa et al., 2001). This stimulation involves IFNγ itself.

The biologic activities of IFNγ result mostly from binding to two transmembrane receptors that reside on hepatocytes and on certain nonparenchymal cells, including Kupffer cells. Ligation of interferon-γ receptors (IFNγRs) activates an intracellular signaling pathway involving Janus kinase (JAK) and signal transducer and activator of transcription (STAT). This activation results in the transcription of dozens of genes, the protein products of which produce a variety of cellular responses (Schroder et al., 2004). Many of these responses involve immune cells. For example, IFNγ can increase the activity of antigen-presenting macrophages, promote leukocyte adhesion required for migration into tissue, and enhance NK cell activation. Kupffer cells activated by IFNγ produce cytokines, including TNFα, which can modulate hepatocellular function, promote an inflammatory response, and participate in cell death (see below). Both CD8+ and CD4+ T cells have receptors for IFNγ (Whitmire et al., 2005). IFNγ promotes differentiation of naïve CD4+ (Th0) cells into Th1 cells and inhibits their differentiation into Th2 cells. Since Th1 cells produce IFNγ, this differentiation can enhance IFNγ secretion and provide positive feedback to increase Th1 differentiation.

In addition to these activities, IFNγ can inhibit proliferation of many types of cells. Proliferation of hepatocytes occurs slowly in normal liver but can increase markedly when the liver is stressed by partial hepatectomy or by exposure to pathogens or toxicants. Expression of IFNγRs increases on hepatocytes when liver injury occurs, and receptor ligation by IFNγ inhibits hepatocyte proliferation (Volpes et al., 1991; Dong et al., 2007). This occurs through inhibition of cell cycle progression involving STAT-1 activation of IFNγ-responsive genes. One result is the expression of interferon regulatory factor (IRF)-1, which in turn promotes expression of p53 (Kano et al., 1999; Sun et al., 2006). STAT-1 and p53 activate the promoter for p21, the expression of which leads to inhibition of S-phase progression and, consequently, to the inhibition of cell proliferation (reviewed in Tura et al., 2001 and Horras et al., 2011).

IFNγ can also induce cell death, including apoptosis of hepatocytes, by mechanisms independent of p53 (Kano et al., 1997). The mechanisms by which this occurs are not well understood (reviewed in Horras et al., 2011). According to one proposed mechanism, IFNγ acts through STAT-1 activation and IRF-1 induction to express inducible nitric oxide synthase (iNOS); this results in enhanced production of nitric oxide (NO), which under conditions of redox stress can initiate intracellular signaling that culminates in apoptosis. However, other cell death pathways, some not involving STAT-1, can also be activated by IFNγ, depending on cell type. Finally, IFNγ can cause expression of Fas ligand (FasL) on cells and thereby contribute to Fas-FasL–induced cell killing (Boselli et al., 2007). Thus, IFNγ-mediated cell killing may occur by several different mechanisms (Fig. 1).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

IFNγ-mediated hepatocyte killing by T cells. Both CD8+ T cells and CD4+ T cells as well as NK and other cells produce and release IFNγ, which can kill cells through activation of cell death pathways involving STAT-1 and IRF-1 activation (right). This killing can be enhanced by TNFα and/or by reactive oxygen species (ROS)/NO released from inflammatory cells, including macrophages such as Kupffer cells (Mphage). IFNγ can also kill hepatocytes via autocrine or paracrine activation of CD8+ T cells involving pathways initiated by Fas-FasL interaction or by formation of perforin pores that allow granzyme B translocation (left). During adaptive immune responses, CD8+ T cells recognize altered hepatocytes via receptors (TCRs), prompting release of IFNγ. See the text for references.

TNFα.

IFNγ often acts through interaction with TNFα. TNFα is a cytokine that can induce several cellular responses and plays a critical role in liver physiology. It is produced and released by a variety of immune cell types including, but not limited to, NK cells, macrophages, and T cells (both CD4+ and CD8+). TNFα signaling can initiate either hepatocyte proliferation or hepatocyte apoptosis, and an appropriate balance between these two conditions is critical to preserving homeostasis in the liver (Wullaert et al., 2007). TNFα binds to and activates two distinct plasma membrane receptors, TNFα receptor 1 (TNFR1, p55) and TNFα receptor 2 (TNFR2, p75). TNFR1 is expressed constitutively in most cell types, whereas the expression of TNFR2 is restricted mainly to immune cells. TNFR1 is responsible for initiating most of the biologic activities of TNFα (Chen and Goeddel, 2002).

Whether TNFα initiates intracellular signaling for cell survival and proliferation or for apoptosis depends on the state of the cell (Wajant et al., 2003). To initiate cell survival, TNFα binding to the extracellular domain of TNFR1 drives receptor trimerization followed by recruitment of a complex of adapter proteins, which can result in activation of nuclear factor-κB (NFκB). NFκB then translocates to the nucleus, where it promotes transcription of many genes involved in cell survival and proliferation and in inhibition of signaling for apoptosis (Wullaert et al., 2007).

When hepatocytes become stressed or damaged, TNFα can lead to activation of cell death signaling pathways. Activation of TNFR1 can recruit and activate procaspase-8, resulting in apoptosis via two possible routes. The first involves direct cleavage and activation of executioner caspase-3 and caspase-7, which cleave a number of proteins leading to apoptosis. The second, mitochondrial pathway can also be initiated by caspase-8 and entails cleavage and activation of the proapoptotic protein, Bcl-interacting protein (Bid). The truncated form of Bid translocates to the outer mitochondrial membrane where it facilitates formation of the mitochondrial permeability transition (MPT) pore. Formation of the MPT pore allows for release of cytochrome c into the cytosol where it interacts with apoptosis protease-associated factor (Apaf1) and procaspase-9, resulting in cleavage and activation of the latter. Activated caspase-9 subsequently cleaves and activates the executioner caspase-3 and caspase-7, which effect apoptosis (Green, 1998; Wullaert et al., 2007).

In addition to the pathways discussed above, TNFα signaling via TNFR1 can result in activation of the mitogen-activated protein kinases (MAPKs), c-Jun N-terminal kinase (JNK) and p38. Importantly, activation of these MAPKs can promote signaling for either cell survival or apoptosis depending on their subcellular localization, duration of activation, health state of the cell, and other factors (Cowan and Storey, 2003). For instance, when it is activated transiently, JNK, in particular, activates transcription factors that promote cell survival, including activator protein 1 (AP-1) and NFκB (Hasselblatt et al., 2007). However, when it is activated for a prolonged period of time, JNK can lead to activation of substrates that promote cell death, including p53. Specifically, phosphorylation of p53 by JNK promotes its stabilization and resistance to proteasomal degradation (Fuchs et al., 1998). Moreover, JNK can lead to activation of the transcription factor c-MYC, which can give rise to apoptosis under certain conditions (Hoffman and Liebermann, 2008). Finally, persistent activation of JNK can result in a decrease in mitochondrial membrane potential (by an unknown mechanism) leading to MPT, apoptosome formation, and activation of caspase-3 and can thereby bring about apoptosis (Gross et al., 1999).

IFNγ-TNFα Interaction.

Importantly, IFNγ and TNFα are sometimes incapable of causing the responses described above on their own or are required in very large concentrations to do so; however, a pronounced synergy between IFNγ and TNFα can lead to various responses at relevant cytokine concentrations. For example, IFNγ and TNFα can synergize with each other in causing DNA fragmentation and apoptosis in vitro in primary mouse hepatocytes (Morita et al., 1995). In addition, it has been suggested that IFNγ can synergize with TNFα and other inflammatory mediators to induce expression of the iNOS gene; as noted above, in the presence of redox stress, iNOS induction can lead to production of oxidizing species that promote hepatocyte apoptosis (Vodovotz et al., 2004; Fig. 1).

Although interaction between IFNγ and TNFα seems to be important in some IDILI models (addressed below), this interaction is incompletely understood at the molecular level. The IFNγ-mediated binding of STAT-1 to gamma activation sequences in DNA and the binding of IRF-1 to IFN-activated response elements in DNA were enhanced in a human epithelial cell line after cotreatment of the cells with IFNγ and TNFα, and this may be attributable, in part, to increased expression of the IFNγR (Robinson et al., 2003). TNFα enhanced IFNγ-stimulated JAK2 phosphorylation and activation in human sarcoma cells, and tyrosine phosphorylation of IFNγR chain 1 was elevated in cells cotreated with IFNγ and TNFα (Han et al., 1999). These results suggest that both the expression and activation of IFNγR can be enhanced by TNFα coexposure. Conversely, expression of TNFRs can be enhanced by IFNγ (Wang et al., 2006). Moreover, in human monoblastic Mono-Mac-6 cells stimulated with lipopolysaccharide (LPS), IFNγ prolonged TNFα expression (Lee and Sullivan, 2001). In microglial cells in vitro, IFNγ and TNFα cooperated in enhancing expression of iNOS and other prooxidant enzymes (Mir et al., 2009); the enhanced expression depended on MAPK kinase and on extracellular signal-regulated kinase (ERK) signaling that resulted in the release of TNFα (Mir et al., 2008). In addition, IFNγ activation of the JAK/STAT pathway potentiated TNFα-induced NFκB binding to DNA and activated IRF-1 needed for iNOS expression. In the AML-12 hepatocyte cell line, cotreatment with IFNγ and TNFα caused cell cycle arrest that was independent of apoptosis and mediated by p53 and NO (Brooling et al., 2005). Together, these results suggest that the IFNγ-TNFα interaction can involve 1) enhanced IFNγR expression and activation by TNFα exposure, 2) enhanced TNFR expression and prolonged TNFα expression by IFNγ exposure (Fig. 1), 3) potentiation by IFNγ of TNFα-induced NFκB binding to DNA, and 4) synergistic cell cycle arrest that is mediated by NO. However, most of these studies were conducted in extrahepatic, transformed cells so additional study is needed to understand the molecular mechanisms that underlie IFNγ-TNFα interactions in liver parenchymal and nonparenchymal cells and whether these interactions become significant only in stressed cells.

The Adaptive Immunity Hypothesis of IDILI.

The adaptive immunity hypothesis has remained for decades the most popular of the theories to explain IDILI. The classic thinking has been that a reactive metabolite of a drug binds to a protein, and the resulting adducted protein acts as a hapten that is recognized by and sensitizes the adaptive immune system. Drug rechallenge or continued drug exposure then precipitates an adaptive immune response that injures the liver (Fig. 2, upper right). The “pharmacological interaction” hypothesis is a recent modification, proposing that a drug might bind directly and reversibly with antigen-presenting molecules, stimulating a damaging immune response that does not require prior sensitization to the drug (Wuillemin et al., 2013).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Hypotheses regarding drug-cytokine interactions in IDILI pathogenesis. Evidence in humans and animals suggests that drugs can be bioactivated to reactive metabolites that can initiate an adaptive immune response (top right). Alternatively, initiation of an innate immune response through TLR activation (top left) can prompt a hepatotoxic interaction with drugs. Both of these immune responses result in the generation of mediators such TNFα and IFNγ. Some drugs also directly increase production of these cytokines by activated immune cells. At the level of the hepatocyte, IDILI-associated drugs can enhance hepatocellular sensitivity to the cytotoxic actions of TNFα and IFNγ (bottom center). In addition, IFNγ and at least some IDILI-associated drugs inhibit hepatocyte proliferation, thereby impairing the ability of the liver to adapt to drug-induced stress (bottom right). DAMP, damage associated molecular pattern molecule; HPC, hepatic parenchymal cell; polyI:C, polyinosinic-polycytidylic acid.

The observation that fever, skin rash, and eosinophilia accompany some IDILI reactions has been taken as evidence for an adaptive immune etiology. However, the most compelling evidence for adaptive immune system involvement in IDILI has arisen from recent studies in humans that revealed associations between human leukocyte antigen (HLA) polymorphisms and cases of IDILI for several drugs. For example, patients who suffered IDILI from amoxicillin/clavulanate had one or more HLA single nucleotide polymorphisms (SNPs), suggesting involvement of the adaptive immune system in the pathogenesis (Lucena et al., 2011). Most of the SNPs occurred in HLA class II genes. Importantly, although various HLA SNPs were associated with increased IDILI risk, the predictive value of the SNPs was very small; this suggests that some other genetic or environmental factor(s) is needed to precipitate IDILI, even in individuals with the associated HLA SNPs. Interestingly, in a study by Lucena et al. (2011), a SNP in the gene encoding TNFα was also found to be significantly associated with amoxicillin/clavulanate IDILI, suggesting the possibility of a role for inflammatory cytokines. Indeed, the linkage of the TNFα gene to the HLA-B locus raises the possibility that some IDILI reactions associated with various HLA-B alleles could be due to linkage with variations in the TNFα gene (Inoko and Trowsdale, 1987).

The adaptive immune response is complex and interacts with the innate immune system to effect killing of pathogens and damage to host tissue. Haptens comprising drug metabolite-protein adducts are endocytosed and degraded into peptides by antigen-presenting cells, which can then present specific peptides bound to HLA II molecules to CD4+ Th cells or to CD8+ T cells that express specific receptors for the peptide-HLA complex. This binding activates the T cells, and upon a second, costimulatory signal (provided by CD28, CD80, or CD86) and subsequent autocrine signals, the cells proliferate and differentiate. One antigen-presenting pathway involving CD8+ T cells results in differentiation into cytotoxic T cells, which can kill pathogen-infected cells by expressing cytotoxic proteins (see below). Another pathway results in differentiation of Th0 cells into either Th1 cells or Th2 cells. Differentiated Th1 cells produce several cytokines, a major one of which is IFNγ. Th2 cell differentiation results in a population of B cells that produce antibodies. Th2 cells also produce factors such as IL-3, IL-5, and granulocyte macrophage colony-stimulating factor that stimulate differentiation of myeloid precursor cells into eosinophils; these cells produce large amounts of IFNγ in response to TNFα, IL-12, or IL-4 (Spencer et al., 2009).

Killing of host cells by activated T cells can occur through several mechanisms. Perforin expressed by cytotoxic CD8+ T cells can incorporate into the plasma membranes of stressed host cells, forming a pore and allowing the passage of granzyme B into the cells, which initiates caspase-dependent cell death signaling (Fig. 1). Cytotoxic CD8+ T cells can also express FasL on their surfaces, which can bind to Fas on target cells to initiate killing through caspase-dependent apoptotic pathways. Cytotoxic CD8+ T cells also express IFNγRs, the activation of which promotes CD8+ T cell expansion as well as cell killing by these mechanisms (Whitmire et al., 2005). Alternatively, IFNγ produced by CD4+ Th1 or other cells can initiate cell death signaling by binding to an IFNγR on the surface of host cells (Fig. 1). Such binding initiates intracellular signaling involving STAT-1 activation of IFNγ-responsive genes, including the gene expressing IRF-1, which in turn induces iNOS and other genes that could be involved in initiation of cell death. IFNγ can also act indirectly by stimulating macrophages to release cytotoxic factors such as NO, reactive oxygen species and TNFα (Fig. 1); as noted above, TNFα acts synergistically with IFNγ in hepatocellular killing.

In addition to its ability to stimulate T cell–mediated killing, IFNγ might influence liver regeneration by inhibiting hepatocyte proliferation. In a study in partially hepatectomized rats, IFNγ stimulated major histocompatibility complex II antigen expression on Kupffer cells; the authors speculated that these cells then present hepatocytes as antigen to Th cells and cytotoxic T cells, which suppresses reparative hepatocyte proliferation (Sato et al., 1993).

Despite the longstanding popularity of the adaptive immunity hypothesis, no animal models have emerged in which substantial liver injury occurs from an adaptive immune response after sensitization with a drug that has caused IDILI in people (Ng et al., 2012; Metushi et al., 2015b). Animal models that implicate a role for adaptive immunity in IDILI responses to the drugs amodiaquine and halothane were recently developed; however, it is important to note that the liver injury produced in these models is mild and does not reflect the severity of liver injury that occurs in human IDILI responses to these drugs. Nevertheless, these models provide some insight concerning the mechanisms underlying adaptive immune-mediated IDILI responses. Metushi et al. (2015a) found in mice that depletion of NK cells attenuated the mild liver injury induced by amodiaquine exposure. When activated, NK cells release IFNγ, which is known to activate signaling pathways that lead to cell death. In addition, Chakraborty et al. (2015) found that depletion of CD4⁺ T cells, which also release IFNγ, protected mice from the delayed onset of halothane hepatitis. Accordingly, it is possible that IFNγ by itself or in the presence of other cytokines promotes hepatocellular killing in cases of human IDILI induced by amodiaquine or halothane.

Insight into which pathways predominate in T cell–mediated killing of hepatocytes has been provided by studies by Tiegs et al. in an animal model employing concanavalin A (Con A). This plant lectin is a T cell mitogen that causes T cell–dependent liver injury in mice. Antibody-mediated depletion of CD4+ T cells protected completely against liver injury from Con A, whereas depletion of CD8+ T cells did not prevent liver injury, suggesting the importance of Th1 cells (Tiegs, 1997; Tiegs et al., 1992; Cao et al., 1998). NKT cells also appear to contribute (Erhardt and Tiegs, 2010; Zhang et al., 2013). Con A treatment was associated with appearance of IFNγ and TNFα in plasma, and inactivation of macrophages or neutralizing either one of these cytokines prevented Con A–induced liver injury. Moreover, liver injury development correlated with IFNγ activation of STAT-1 and IRF-1 and with TNFα activation of the JNK pathway (Streetz et al., 2001; Hong et al., 2002). Findings in IL-27 conditional knockout mice supported the importance of dysregulated production of IFNγ by CD4+ T cells in Con A–induced liver injury in mice (Zhang et al., 2013). Interestingly, IL-5 derived from NKT cells led to maturation of eosinophils, which contributed to the liver injury in this model (Louis et al., 2002). These results suggest that IFNγ is critically important in mediating liver injury from Th1 cell activation and that it acts synergistically with TNFα produced by Kupffer cells.

Recent studies with 2,3,7,8-tetrachloro-dibenzo-p-dioxin showed that this environmental contaminant enhances Con A–induced liver injury in mice and that both IFNγ and NK cells, which produce IFNγ, are involved (Fullerton et al., 2013). This result indicates that xenobiotic agents can interact with Th1 cell–dependent pathways to cause hepatocellular killing by mechanisms involving IFNγ and IFNγ-producing cells and suggests the possibility that some drugs associated with IDILI might evoke similar responses.

The paucity of animal models employing IDILI-associated drugs has limited progress in understanding the contribution of the adaptive immune system to IDILI and the factors that might govern such reactions. Nevirapine used in the treatment of human immunodeficiency virus infections has caused skin and liver reactions in patients. Popovic et al. (2010) developed a model of nevirapine-induced skin injury in Brown Norway rats that is clearly mediated by adaptive immunity. The skin rash that developed depended on CD4+ T cells, and cells isolated from rechallenged rats released IFNγ and other cytokines. No liver injury developed in this animal model; however, if nevirapine-induced liver injury in humans arises from the same mechanism as the skin rash in rats, then IFNγ could be a player in the IDILI pathogenesis from this drug. Clearly, more animal models in which liver injury occurs from IDILI-associated drugs by an adaptive immune-mediated mechanism are needed to understand the importance of IFNγ, TNFα, and other mediators in such reactions.

The Multiple Determinant Hypothesis of IDILI.

It has been theorized that some IDILI reactions result from the intersection of several factors or events (i.e., “determinants”) that together precipitate hepatocellular necrosis (Li, 2002; Ulrich, 2007). Inasmuch as the probability of a reaction would be the product of the probabilities of each factor/event, and since the product would usually be very small, this hypothesis could explain why IDILI responses are typically rare. Important factors are proposed to relate to chemical properties of the drug, drug exposure and metabolism/bioactivation, and genetic and/or environmental factors that determine individual susceptibility to injury. Examples of likely genetic factors could include polymorphisms in drug-metabolizing enzymes or transporters that could lead to enhanced production of a toxic metabolite or altered hepatic accumulation of a drug or its metabolites, respectively. Other factors could include race/ethnicity, nutrition, preexisting chronic liver diseases, and differences in the intestinal microbiome. For example, recent studies indicate that alterations in the intestinal microbiome determine sensitivity of animals to liver injury from numerous hepatotoxiciants (Lv et al., 2014; Chiu et al., 2015; Dubey et al., 2015; Shen et al., 2015; Tian et al., 2015).

A murine model of halothane-induced liver injury has been developed based on this hypothesis. Halothane is metabolized to a reactive metabolite that binds covalently to cellular macromolecules, and the formation of halothane-protein adducts is thought to be required for liver injury. Among the known human risk factors for halothane hepatitis are female sex, middle age, and genetic predisposition (Inman and Mushin, 1974; Cousins et al., 1989). Fasting may be an additional risk factor, since all patients are fasted prior to surgery that requires general anesthesia. Halothane given to fasted, mature, female Balb/c mice caused pronounced liver injury (Dugan et al., 2010). Fed mice were less sensitive to the liver injury, male Balb/c mice and female C57Bl/6 mice were markedly resistant, and immature female Balb/c mice were less sensitive too. Moreover, isoflurane, which is not extensively metabolized to a reactive intermediate and does not share the human IDILI liability of halothane, failed to cause liver injury in the mouse model, indicating that the chemical structure of the anesthetic is a determinant of the hepatotoxic response. These results are consistent with the multiple determinant hypothesis, inasmuch as the confluence of factors known to increase risk in humans (female sex, mature age, halothane structure, genetics) was required to produce halothane hepatitis in mice.

In this murine model of halothane hepatitis, plasma IFNγ concentration was elevated 10-fold in halothane-treated females compared with similarly treated male mice or ovariectomized female mice, which were insensitive to injury (Dugan et al., 2011). IFNγ knockout mice were resistant to halothane-induced liver injury, indicating the importance of IFNγ in the pathogenesis. Halothane treatment increased the activation (CD69 expression) of NK cells, and inactivation of NK cells attenuated both the rise in plasma IFNγ and the liver injury. A recent study also implicated eosinophils in this model (Proctor et al., 2013); these cells are another potential source of IFNγ. Interestingly, TNFα concentration in plasma also rose as a result of halothane exposure. These results suggested that IFNγ released from NK cells and perhaps eosinophils plays an essential role in the development of severe halothane-induced hepatotoxicity in mice and raised the possibility of synergistic interaction between IFNγ and TNFα in the pathogenesis.

The Inflammatory Stress Hypothesis of IDILI.

The erratic temporal and dose-response relationships that characterize idiosyncratic reactions suggest that some event occurring during the course of therapy precipitates IDILI. If this is true, then the precipitating event must happen occasionally and irregularly to account for the infrequent and erratic occurrence of these reactions. Inflammatory cell infiltrates often characterize liver lesions in patients who suffer IDILI (e.g., see Khouri et al., 1987; Fukano et al., 2000; Murphy et al., 2000). This raised the possibility that some IDILI reactions might be explained by an episode of modest inflammation occurring during the course of drug therapy that interacts with some action of the drug to initiate liver injury. Such inflammatory episodes occur commonly in people and are associated with various diseases, infections, and intestinal translocation of inflammagenic bacterial products such as endotoxin (i.e., LPS). Indeed, episodes of mild, subclinical endotoxemia appear to be a normal occurrence in people (reviewed in Roth et al., 1997; Ganey and Roth, 2001; Ganey et al., 2004).

LPS and other inflammagens bind to “pattern recognition receptors” such as Toll-like receptors (TLRs) on cells of the innate immune system (Fig. 2, upper left). This initiates intracellular signaling that leads to activation of transcription factors and expression of inflammatory mediators such as TNFα and IFNγ (Arbour et al., 2000; Beutler, 2000). These mediators are essential in the defense against pathogens but, as noted above, they are also capable of altering homeostasis of host cells. It is well known that modest activation of the innate immune system by inflammagens such as small doses of LPS can markedly augment hepatotoxic responses to numerous chemicals, including some drugs (Ganey et al., 2004; Roth and Ganey, 2010). Accordingly, when an inflammatory episode of sufficient magnitude occurs during drug therapy, it could interact with a drug to render an individual susceptible to a hepatotoxic reaction that would not otherwise occur (i.e., an IDILI response). The episodic and variable nature of exposure to LPS and other inflammagens and genetic variations (e.g., in TLRs or genes encoding cytokines) that influence individual responses to inflammagens could explain individual susceptibility, the infrequency of idiosyncratic reactions, and their erratic temporal relationship to drug exposure.

The inflammatory stress hypothesis led to attempts to determine whether liver injury could be produced in animals by concurrent exposure to an IDILI-associated drug and an otherwise noninjurious inflammatory episode. Indeed, several drugs that cause human IDILI also caused liver injury in animals upon cotreatment with a small, nontoxic dose of LPS, whereas drugs without human IDILI liability failed to synergize with LPS to cause hepatotoxicity (summarized in Deng et al., 2009 and Shaw et al., 2010). It is of interest that drugs frequently associated with human IDILI are NSAIDs and antibiotics (i.e., drugs used in conditions associated with inflammation). Drug-LPS interaction models in rodents have now been developed with chlorpromazine, halothane, ranitidine, diclofenac, sulindac, amiodarone, doxorubicin (DOX), and TVX (see Deng et al., 2009 and Shaw et al., 2010). In each of these models, liver injury occurs from drug-LPS cotreatment but not from exposure to either agent alone. Thus, the inflammatory stress hypothesis has provided the first animal models in which pronounced liver injury occurs from numerous drugs associated with human IDILI.

An example is the TVX-LPS model of drug-inflammation interaction in mice. A robust hepatotoxic response occurred when a small (i.e., nonhepatotoxic but modestly inflammatory) dose of LPS was given to mice 3 hours after doses of TVX that were nontoxic by themselves (Shaw et al., 2007). Levofloxacin, a drug in the same pharmacological class that does not share the same propensity for causing IDILI, showed no synergistic interaction with LPS. Both IFNγ and TNFα were elevated in the plasma of TVX-LPS cotreated mice, and either 1) neutralization of TNFα with etanercept or knockout of either TNFR1 or TNFR2 or 2) knockout of the gene encoding IFNγ provided protection from hepatocellular necrosis (Shaw et al., 2009a,b,c). Interestingly, TVX enhanced LPS-stimulated TNFα production in the RAW 264.7 murine macrophage cell line in vitro (Poulsen et al., 2014a,b), suggesting that TVX can directly enhance macrophage activation and TNFα release.

Together, these results indicated that both IFNγ and TNFα are critical for the hepatopathogenesis of LPS-TVX interaction in vivo. Moreover, the liver injury in TVX-LPS cotreated mice depended on a modest prolongation of TNFα appearance in the plasma by TVX above that caused by LPS alone (Shaw et al., 2007). IFNγ knockout reduced plasma TNFα, and conversely neutralization of TNFα markedly reduced the appearance of IFNγ in plasma, suggesting that each cytokine amplified the production of the other in a dysregulated cycle (Shaw et al., 2009b).

Replacing LPS administration with TNFα administration in this murine model also led to hepatotoxic interaction with TVX (Shaw et al., 2009a). Interestingly, TNFα administration alone caused appearance of IFNγ in plasma, and coadministration of TVX enhanced this response. These findings emphasize the importance of TNFα in initiating IFNγ production and are consistent with studies in vitro, since TNFα, by acting through macrophages or other cell types, can stimulate IFNγ production by cultured NK cells, and conversely IFNγ can enhance TNFα production by macrophages (Berner et al., 2005; Makarenkova et al., 2005; Vila-del Sol et al., 2008).

Transcriptomic analysis of the livers of TVX-LPS cotreated mice revealed selective enhancement of expression of several genes involved in IFNγ signaling (Shaw et al., 2009b). These genes included IRF-1, which is involved in IFNγ-mediated apoptosis and inhibition of cell proliferation. In the human HepG2 hepatocyte cell line, TVX enhanced cell killing from exposure to a combination of IFNγ and TNFα, indicating that the IDILI-associated antibiotic increased the sensitivity of hepatocytes to killing from this cytokine combination (Maiuri et al., 2016b).

Studies in other inflammatory stress models also point to the importance of IFNγ. One study revealed that DOX-LPS cotreatment synergistically enhanced liver injury in rodents, and this enhancement depended on IFNγ (Hassan et al., 2008). Cheng et al. (2009) showed that the viral RNA mimetic and TLR3 agonist, polyinosinic-polycytidylic acid, markedly enhanced halothane-induced liver injury in mice. This was accompanied by activation of Kupffer cells and NK cells and upregulation of TNFα expression (Cheng et al., 2009). Although IFNγ was not evaluated in that study, it is produced by NK cells activated by polyinosinic-polycytidylic acid (Zhang et al., 2009), so that it seems likely that IFNγ interaction with TNFα contributes to injury in that model. We recently showed that IFNγ synergizes with TNFα in sensitizing hepatocytes in vitro to killing by a number of drugs that cause IDILI in people (Maiuri et al., 2015, 2016b). Together, these results suggest that IFNγ-TNFα interactions could be critical to IDILI pathogenesis from a number of IDILI-associated drugs.

The Failure-to-Adapt Hypothesis of IDILI.

A majority of human volunteers given a therapeutic dose of acetaminophen (APAP) daily for 2 weeks experienced early, modest increases in serum ALT activity, which subsequently subsided toward normal despite continued drug treatment (Watkins et al., 2006). An interpretation of this result was that APAP causes minor hepatotoxicity, to which the liver adapts over time, thereby returning to normal. It is possible that this injury-adaptation phenomenon occurs commonly with many drugs and that people who develop IDILI responses to a drug are those whose livers fail to adapt, permitting progression to fulminant injury (Watkins, 2005). Such adaptation to minor injury has been demonstrated in mice treated with amodiaquine, a drug that has caused IDILI in humans (Metushi et al., 2015b).

Stimulation of immune cells by TNFα or IFNγ leads to IL-10 expression, which in turn downregulates TNFα and IFNγ production, thereby preventing their organ-damaging actions. Failure of this regulation by IL-10 in susceptible patients could result in unrestricted and damaging cytotoxic actions of cytokines such as TNFα and IFNγ. In this regard, it is of interest that one study found worse outcomes from IDILI in patients who had IL-10 polymorphisms that resulted in lower plasma IL-10 concentrations (Pachkoria et al., 2008) and another study found an association between polymorphisms resulting in low plasma IL-10 and diclofenac-induced liver injury (Aithal et al., 2004). Accordingly, it is possible that an inability of IL-10 to control production of cytokines such as TNFα and IFNγ in susceptible patients could lead to hepatocellular death from these cytokines, which would present clinically as “failure to adapt” to their damaging effects.

Recovery from liver injury typically entails proliferation of hepatocytes. For example, after loss of liver tissue from partial hepatectomy, the organ responds with cell proliferation that restores liver mass and function. In a phenomenon that has been termed “autoprotection,” repeated, small doses of APAP given to rodents reduced liver injury when a larger, hepatotoxic dose was subsequently administered, and cell proliferation stimulated by the APAP pretreatments was thought to play a major role in the reduced hepatic sensitivity (Dalhoff et al., 2001). Indeed, tissue repair via cell proliferation has been shown to be critical to halting the progression of injury for numerous hepatotoxicants, including APAP, and inhibiting cell proliferation results in injury progression (Chanda et al., 1995; Mehendale, 2005). These observations suggest the possibility that failure to adapt to modest injury caused by drugs might be attributable at least in part to reduced hepatocellular proliferative ability in susceptible individuals, leading to injury progression and IDILI rather than adaptation (Fig. 2, bottom right).

As noted above, IFNγ is known to cause cell cycle arrest in hepatocytes in vitro and in vivo and might therefore play a role in inhibiting cell proliferation during treatment with certain drugs (Kano et al., 1999; Tura et al., 2001; Brooling et al., 2005; Sun et al., 2006; Dong et al., 2007). Moreover, numerous drugs that cause human IDILI inhibit cell proliferation in vitro. Examples include diclofenac (Rajabalian et al., 2009), sulindac (Chennamaneni et al., 2012), TVX (Beggs et al., 2015), halothane (Waxler et al., 1994), DOX (Supino et al., 1977), and chlorpromazine (Basta-Kaim et al., 2006). In addition, a series of NSAIDs associated with IDILI inhibited HepG2 cell proliferation in vitro, whereas an NSAID not associated with IDILI, aspirin, did not have this effect (Maiuri et al., 2014). This raises the possibility that a drug’s ability to inhibit proliferative repair could contribute to IDILI. Although speculative, of interest is the possibility that IFNγ produced during drug exposure might interact synergistically with direct, antiproliferative effects of a drug to inhibit hepatocellular regeneration, thereby prompting failure to adapt.

Although these cytokines might prompt adaptation failure by inhibiting hepatocyte proliferation, an effect on immune cells is also possible. There exists some evidence that hepatic failure to adapt might be due to a failure of immune tolerance in the liver (reviewed in Dara et al., 2016). Accordingly, cytokine-induced inhibition of proliferation of lymphocytes that are needed for immune tolerance could contribute to the initiation of severe IDILI.