Jian Jiang

General discussion

Adverse drug reactions (ADRs) are frequent events leading to a substantial mortality and morbidity (the estimated 197,000 deaths per year) in Europe [1]. Improving drug safety and providing protection of patients against ADRs are of great concern [2]. As the liver has a crucial function in the chemical biotransformation, bioactivation, and excretion, it is also the organ with the highest susceptibility to drug-induced damage [3]. In total, nearly 10% of all suspected ADRs are related to drug-induced liver injury (DILI) and account for a considerable proportion of fatal cases of ADRs [4]. Although the incidence of DILI is relatively low, it is the leading cause of acute liver failure in western countries [5, 6]. Due to the lack of objective diagnostic tests, the relatively low frequency of occurrence and a high risk of not being observed in preclinical or clinical trials, DILI does often not emerge until a drug has been put on the market [7]. Thus, as a major cause of drug withdrawal and disapproval, DILI is a worrying issue for not only patients and gastroenterologists but also regulatory agencies and pharmaceutical companies [8].

Despite intensive efforts to develop efficient biomarkers helping to predict the risk of DILI in the early stage of drug development, no specific biomarker has been identified [9]. Biochemical endpoints obtained from animal studies are currently used in drug development to predict potential human toxicity [10]. However, due to the interspecies differences in drug metabolism and pharmacokinetics, animals may respond differently to drugs than humans do [11, 12]. Although the testing of one drug can require up to 800 animals, the animal experimental scale may still be too small to detect any rare events before final approval of a new drug [13, 14]. Furthermore, a large variation in susceptibility to DILI may occur among humans. This can be the consequence of differences in human genetic variability in drug metabolizing enzymes, age, sex and preexisting liver diseases. These factors may also contribute to the induction of idiosyncratic reactions. These factors are difficult to be accounted for in animal models and make idiosyncratic adverse drug reactions less likely to be detected during the Phase I and Phase II trials of a new drug.

Monitoring the safety of drugs is a key focus of the U.S. Food and Drug Administration (FDA) and in 2010, the FDA launched the initiative regarding advancing regulatory science for public health [15] [16]. Among several topics, modernizing safety testing has been advocated as an important part of this initiative and it was proposed to accelerate drug development and improve drug safety assessment by integrating in vitro and in silico methodologies [15]. Therefore, the development of improved models to study the cause and molecular mechanisms involved in DILI is a key priority in order to advance drug safety. In this context, the incorporation of human-cell based in vitro models and in silico approaches holds great potential.

The overall goal of the work presented in this thesis is to better understand the underlying mechanisms of drug-induced liver toxicity using transcriptome responses induced in different in vitro hepatocyte models, which may enhance effectiveness of post-marketing drug safety evaluation and improve assessment of safety during early drug development. The main objectives are to 1) create an overview of the currently available human-cell based in vitro models for studying hepatic toxicity; 2) investigate the role of mitochondrial damage in the occurrence of oxidative stress and the development of DILI; 3) evaluate the relevance of HepG2 as an in vitro model for studying the mechanisms of DILI; 4) identify the common pathways that are responsible for inflammation-related drug idiosyncrasy; and 5) establish the role of hepatic inflammation in the susceptibility to drug-induced toxicity. Here, we summarize and integrate the main findings from all the previous chapters of this thesis.

Human-cell based in vitro models for studying hepatic toxicity

Several human hepatocyte-based in vitro systems are currently available in order to test liver toxicity. Since each model has its advantages and disadvantages, we discussed in Chapter 2 the benefits and limitations of these models. HepG2 and the HepaRG are the most commonly used immortalized hepatic cell lines for toxicity assessments derived from human cancer patients. Due to their single-donor genetic characteristics, the use of these two models is not subjected to interindividual differences occurring between experiments [17]. As an easy-to-handle and inexpensive cell line, HepG2 cells are often used for high-throughput screening approaches. HepG2 cells also display sufficient sensitivity to detect the exposure-induced oxidative stress, disturbed membrane integrity and cytotoxicity [18]. However, the decreased expression of several cytochrome P450 enzymes (CYP450s) and transporters in HepG2 limits the relevance of this model in hepatotoxicity studies focusing on drug transport and metabolism [19]. The highly differentiated HepaRG cell line, on the other hand, expresses considerably higher levels of drug transporters and CYP450s [20-22], but a high metabolic capacity and inducibility of CYP450s in in vitro models is not directly correlated to a high sensitivity of detection of hepatotoxicity [19].

Primary human hepatocytes (PHH), especially when cultured between two layers of an extracellular matrix (ECM), are considered as the ‘gold standard’ for in vitro evaluation of drug toxicity [19]. In the liver, hepatocytes are highly polarized cells possessing the apical and basal surfaces [23, 24]. Normal membrane polarity is a prerequisite for PHH to maintain the liver-specific functions [25]. When growing between two ECM layers, PHH can better maintain their normal polarized morphology, which improves the specific liver functions [25]. Compared to the immortalized cell models, PHH have a more in vivo-like morphology, increased cell viability [26] and improved metabolic competence [27]. These advances favor this model for investigating drug metabolism and transport, hepatobiliary disposition, drug-drug interaction as well as interindividual variance among humans [28]. A comparative study has shown that a 3-donor pooled PHH cell system displayed a higher sensitivity to hepatotoxicity induced by certain drugs, as compared to the cell line-based in vitro models mentioned above. However, the drug sensitivity of PHH was donor-dependent and did not reach the desired sensitivity (less than 44% sensitivity in PHH) [19]. Commercially available PHH have diverse characteristics among distinct lots [29, 30], which imply that over time, the same toxicity data may not be reproduced [30, 31]. Moreover, the top layer of gelled ECM creates a transfer barrier and prevents test compounds from reaching the PHH [32, 33]. Other limitations, such as the lack of interactions between cells [34] and the loss of phenotypic and function features over a prolonged culturing time [25], also restrict the applications of the sandwich-cultured PHH model in toxicity research.

In a 3D spheroid culture, PHH more closely resemble the in vivo situation and can maintain phenotypic stability, viability and hepatocyte-specific functions for a long-term period [35, 36]. A 3D multicellular system, containing multiple hepatic cell types, can reflect the cross-talk between hepatocytes and other liver-derived non-parenchymal cells and allow the study of cytokine profiles and the impact of cytokines on hepatocyte functions following drug exposures [27]. However, the current spheroid technique has difficulties associated with the downstream experiments since each spheroid generally consists of only 10,000 cells [37]. Similar to the PHH model, microtissues-derived from multiple donors also suffer from poor reproducibility of results due to the large interindividual differences among human donors of hepatic cells. Furthermore, the high cost of the microtissues also limits the broad application of this more advanced model in toxicity assessment. Based on this review, we selected the appropriate models for different research topics in the following studies.

The role of mitochondrial damage in the development of DILI

In Chapter 3, we explored the role of mitochondrial damage in the occurrence of oxidative stress and the development of DILI through analyzing the whole-genome gene expression responses to drug exposure. Acetaminophen (APAP), a widely used analgesic and antipyretic drug, was used as a model compound to investigate the mechanisms of drug-induced hepatotoxicity in HepG2 cells. After an overdose, APAP is known to induce oxidative stress in the liver, but the source is unclear. To identify the origin of APAP-induced oxidative stress, we exposed HepG2 cells to a low- and a toxic-dose APAP for 12, 24, 48 and 72h. Gene expression responses in HepG2 cells were analyzed and mitochondrial ROS formation and ATP yield were measured. The microarray analysis showed that APAP disrupted the expression of key genes involved in the assembly, stability, and structural integrity of the mitochondrial electron-transport chain (ETC) complexes. The gene expression changes could result in the disturbance of the mitochondrial energy metabolism and the increase ROS production. Meanwhile, the expression of mitochondria-specific antioxidant SOD2 was significantly inhibited by the toxic-dose of APAP, which could result in reduced ROS scavenging capacity. These alterations may lead to disequilibrium of the ROS homeostasis and the induction of oxidative stress in mitochondria. Indeed, these toxic-dose APAP-induced gene expression changes are associated with the measurable increase in mitochondrial ROS generation and a decrease in ATP production. Therefore, based on the whole-genome expression data, we suggest that the gene expression changes in ETC-coding genes and the subsequent increase in the mitochondrial ROS formation are major sources of the toxic-dose APAP-associated oxidative stress and cytotoxicity in HepG2 cells.

The relevance of HepG2 as an in vitro model for studying the mechanisms of DILI

As an inexpensive and easy-to-handle human hepatocellular carcinoma cell line, HepG2 cells are widely used as in vivo model for studying mechanisms of drug-induced hepatotoxicity. In Chapter 3, we have demonstrated the molecular mechanisms of APAP-induced hepatotoxicity in HepG2 cells. However, this in vitro model has been criticized for its low metabolic capacity and cancer cell characteristics, potentially making it less suitable to study mechanisms of drug-induced toxicity. To establish the relevance of our findings, we compared the APAP-induced genome-wide expression changes observed in HepG2 with those induced in cryopreserved human hepatocytes in sandwich culture (Chapter 4). The responses of CYP450 genes encoding the key APAP-metabolizing enzymes to the high-dose APAP treatment were indeed weaker in the HepG2 cells as compared to the PHH. However, our data also suggest alternative mechanisms that may contribute to the development of APAP-induced cytotoxicity in HepG2 cells. The numbers of DEGs were highly dependent on the number and the combination of donors. The DEGs of all the 3 donors resulted in the largest number of DEGs and these DEGs covered the majority of DEGs (90-98%) calculated based on every two-donor combinations. After exposure to the high-dose APAP, approximately one-third of the differentially expressed genes (DEGs) were overlapping between HepG2 and 3-donor pooled PHH. Analysis of the cellular location of these DEGs revealed that mitochondria were the primary target of the high-dose APAP treatment in both cell models. Similarly, the 10 mM APAP treatment significantly affected the expression of genes participating in the mitochondrial respiratory activity- and mitochondrial structural maintenance-associated processes in both models. Furthermore, the direction changes and functions of the genes were generally quite similar between HepG2 and PHH cells. Compared to the HepG2 cells, the high-dose APAP down-regulated the expression of more genes involved in the hepatic amino acid- and lipid- metabolisms in PHH. Due to metabolic reprogramming is a primary feature of cancer cells [38, 39], the differences in the metabolism related genes may be attributed to hepatocellular carcinoma characteristics of HepG2 cells. Nevertheless, compared to PHH, the HepG2 cells showed higher sensitivity to the APAP-induced cell death at both the transcriptome and the phenotypic levels, probably because of the transfer barrier produced by the top-layer of gelled-collagen when hepatocytes cultured in a sandwich configuration. Overall, our data demonstrate that, despite its reduced metabolic capacity, studies carried out with the HepG2 cells could produce relevant data for studying the mechanisms of DILI.

Pathways responsible for inflammation-related drug idiosyncrasy

After introduction on the market, some drugs have shown the ability to induce idiosyncratic hepatotoxic reactions. Idiosyncratic drug toxicity is a rare adverse event that only affects a small fraction (less than 5%) of people. Although rare, the idiosyncratic drug hepatotoxicity accounts for about 10% of all cases [40]. This type of ADRs is unpredictable because it does not show clear dose-response characteristics [41] and only occurs in certain susceptible individuals [42]. Therefore, the drug idiosyncrasy may remain undetected during the drug development processes until approval and marketing of the drug [43]. As idiosyncrasy is difficult to study, the mechanisms involved remain unclear.

Recently, animal studies have shown an association between inflammatory stress and the occurrence of idiosyncratic drug reactions [44-47]. It has been shown that ceramides accumulate in response to an inflammatory stimulus [48]. Ceramides are bioactive sphingolipids that act as lipid messengers mediating various cellular processes including apoptosis [49]. We hypothesize that the altered ceramide metabolism under the inflammatory conditions contributes to the idiosyncratic drug-induced hepatotoxicity. To test the hypothesis, we exposed HepG2 cells with six compounds (at high- and low-dose), including three idiosyncratic (I) and three non-idiosyncratic (N) compounds, with (I+ and N+) or without (I- and N-) a cytokine mix for 24h. Mass spectrometry was applied to investigate the concentrations of ceramides and flow cytometry was used to determine the apoptosis rate of each treatment. The genome-wide responses of these two groups of compounds were compared to their matching DMSO controls under inflammatory or normal conditions to reveal the combined effect of cytokines and idiosyncratic drugs. The results demonstrated that the immune- and stimulus-response associated biological processes were significantly influenced by the I+ treatments. The apoptosis-related pathways only showed up-regulation following exposure to the high-dose I+. Additionally, the pathways of ceramide signaling, ER stress, and NF-kB activation were actively involved in the high-dose I+-induced cell death. By studying the expression of genes included in these pathways and the concentrations of a variety of ceramides, we concluded that, at high dose, inflammatory cytokines interact with the idiosyncratic drugs to disrupt the expression of ceramide metabolism-associated genes, initiate a dynamic disequilibrium in ceramides/sphingolipids balance and eventually lead to an increased ceramide synthesis. Subsequently, the measurable elevations in intracellular ceramide levels altered the expression of ER stress- and JNK signaling cascade-related genes, which induced ER stress and JNK pathway activation and ultimately prompted the mitochondria-independent apoptosis in hepatocytes. Indeed, the high-dose I+ exposure significantly increased the concentrations of several ceramides and subsequently induced about 17% apoptosis 24 h after treatment. Therefore, by combining the metabolomics and transcriptomics responses, our study has provided a general mechanism for this inflammation-associated idiosyncratic drug reaction.

Hepatic inflammation and susceptibility to drug-induced toxicity

It has been reported repeatedly that inflammatory factors may induce or aggravate hepatotoxic responses to chemicals and drugs [50]. Also, clinical reports have pointed out that the minimum doses of APAP that induce liver toxicity vary strongly among individuals [51, 52]. Hepatic inflammatory reactions have been associated with the increased APAP susceptibility [52]. Apart from the genetic variation of CYP450 enzymes, the release of inflammatory cytokines has been proven to play a key role in APAP toxicity [53, 54]. Therefore, we investigated the role of inflammatory factors in the pathogenesis of APAP-induced hepatotoxicity (Chapter 6). Previous in vivo experiments have demonstrated that the lipopolysaccharides (LPS) challenge, which triggers Kupffer cells (KCs) to secrete inflammatory cytokines [55], renders the liver more susceptible to a variety of hepatotoxicants, including APAP [45]. We used an in vitro 3D co-culture model containing both PHH and KCs to investigate the mechanism of the inflammation-associated increase in APAP sensitivity. After exposure to LPS and different concentrations of APAP for 24h, the cytotoxicity and cytokine production profile were measured and the genome-wide gene expression patterns were analyzed.

Indeed, the LPS stimulation stimulated the production of IL-8, a pro-inflammatory cytokine [56], and IL-6, a cytokine regulates the hepatic regeneration [57], and the co-exposures significantly enhanced the cytotoxicity in this model. The combined treatments of LPS and median- and high-dose APAP significantly increased the IL-8 secretion but reduced the release of IL-6. Without LPS, the extensive suppression of ETC-OXPHOS complex- and antioxidant enzyme-coding genes occurred after the high-dose APAP exposure, which could lead to diminished ATP synthesis, increased ROS production and decreased antioxidant ability and may ultimately cause disequilibrium in ROS homeostasis and the occurrence of oxidative stress. However, in the presence of LPS, even the median-dose APAP is sufficient to induce down-regulation of genes encoding the ETC subunits and several antioxidants, indicating a decreased ATP production and an increased mitochondrial oxidative damage. At the median-dose, APAP inhibited the expression of Fcγ receptor-coding genes, suggesting a reduced KC phagocytosis. In the presence of LPS, this could result in a prolonged clearance of LPS and cell debris released from the damaged cells. The accumulated LPS and cell remnants, alternatively, over activated the TLR4 receptor on the surface of KCs to trigger a severe inflammatory response that further accelerate the progression of liver damage. In conclusion, using this model, we identified potential mechanisms that may explain the elevated APAP sensitivity, which may be useful in developing guidelines for preventing APAP-induced liver failure in patients with pre-existing hepatic inflammation as well as the healthy individuals with recurrent therapeutic doses of APAP intake.

Conclusions

Using toxicogenomics analysis, we are able to identify several molecular mechanisms responsible for the development of DILI. We have demonstrated that in vitro human-based cell models are suitable to study DILI mechanisms. The studies presented in this thesis point out that mitochondria, especially the mitochondrial respiratory function, are primary targets for APAP-induced hepatotoxicity. The suppression of expression of ETC subunit-coding genes occur repetitively among different cell models, ranging from the HepG2 cell line to the 3D co-cultured human hepatic microtissues, after exposure to the high-dose APAP. In addition, our data demonstrate that hepatic inflammation plays a crucial role in the development of idiosyncratic drug reactions and the accumulation of ceramides in hepatocytes and the subsequent ER stress are involved in this process. These results may be useful for early detection of drug-induced idiosyncratic liver toxicity and help improve drug safety monitoring. Furthermore, our data also illustrate that inflammatory conditions increase susceptibility to APAP, which may be helpful in the development of personalized therapeutic strategies for patients with pre-existing liver inflammation. The HepG2 cells can be used for mechanistic studies of drug-induced hepatotoxicity unless the amino acid- and lipid-metabolism are involved in the research questions.

Overall, this thesis contributes to the improvement of drug safety by providing new insights into the molecular mechanisms of drug toxicity that may help develop strategies to predict and prevent DILI

Limitations and Future perspectives

Over the past few decades, despite the numerous effects devoted to the study of DILI, the understanding of the mechanisms of the adverse drug reactions remains incomplete. With increasing demands for the reduction and replacement of animal use in research and testing, proper and reliable in vitro models derived from human liver cells are urgently needed. In this thesis, promising results are achieved for mechanistic explorations of DILI using different in vitro cell models generated from the human hepatic cell line and 3D (co-)cultured primary human liver cells. However, it should be taken into account that, as a complex disorder, multiple risk factors, including both human genetic variants and environmental risk factors, are associated with increased susceptibility to DILI among populations.

Recently, large interindividual variability of in response to drugs has been observed using PHH derived models. For instance, the gene expression and activities of CYP450s before and after compound exposures as well as the resulted genotoxicity and cytotoxicity vary significantly among PHH- and liver slice-derived from different donors [58, 59]. Indeed, we also observed extensive interindividual differences in APAP-treated PHH derived from 3 donors. Therefore, the interindividual variation in response to the exposure to drugs is a relevant issue during the study of DILI. In future studies, this issue should be taken into account and discussed.

Apart from the PHH and KCs, other nonparenchymal liver cells, including T/B lymphocytes, natural killer cells and natural killer T cells belonging to the adaptive and the innate immune systems, also play important roles during the development of DILI [60]. A more advanced in vitro co-culture model that also incorporates these cell types may be valuable in further improving the relevance of the in vitro assessment of drug-induced toxicity. In vitro models using induced pluripotent stem cell (iPSC)-derived hepatocyte-like cells [61], may overcome the scarcity of human tissues and popularize the application of this model in modeling diseases and drug development. Nevertheless, the generation of iPSCs, using retroviral or lentiviral systems, may cause genetic aberration which may influence the biological activities associated with exposures [62]. In addition, the assessment of this in vitro model may become more complicated due to a number of variables, such as the use of small molecules and single or multiple vectors and sources of cells during the production and reprogramming of iPSCs [63]. Nowadays, progress has been made to develop virus-free iPSCs in order to fulfill the needs of clinical application. Along the gathering of new and improved reprogramming technologies and differentiation protocols, the human iPSCs hold great promise in studying different types of liver diseases. With the combination of 3D culturing conditions, iPSCs may create more elaborate cell models that provide a unique opportunity for researchers to gather patient-specific information [61, 64]. This may accelerate the process of developing personalized drug therapies and eventually improve drug safety monitoring.