proteomics in the search for mechanisms and biomarkers of drug-induced hepatotoxicity

Toxicology in Vitro 26 (2012) 373–385

Contents lists available at SciVerse ScienceDirect

Toxicology in Vitro

journal homepage: www.elsevier .com/locate / toxinvi t

Review

Proteomics in the search for mechanisms and biomarkersof drug-induced hepatotoxicity

Anke Van Summeren a,b,c,⇑, Johan Renes a,c, Joost H.M. van Delft b,c, Jos C.S. Kleinjans b,c,Edwin C.M. Mariman a,c

a Department of Human Biology, Maastricht University, P.O. Box 616, 6200 MD Maastricht, The Netherlandsb Department of Toxicogenomics, Maastricht University, P.O. Box 616, 6200 MD Maastricht, The Netherlandsc Netherlands Toxicogenomics Centre (NTC), The Netherlands

a r t i c l e i n f o a b s t r a c t

Article history:Received 6 October 2011Accepted 9 January 2012Available online 17 January 2012

Keywords:ProteomicsLiverHepatotoxicityBiomarkersIn vivoIn vitro

0887-2333/$ - see front matter � 2012 Elsevier Ltd. Adoi:10.1016/j.tiv.2012.01.012

Abbreviations: 2-DE, two-dimensional gel electropserum; LC–MS/MS, liquid chromatography tandem mMALDI, matrix assisted laser desorption ionization; M⇑ Corresponding author. Tel.: +31 (0) 43 3881509;

E-mail address: a.vansummeren@maastrichtunive

The safety assessment for pharmaceuticals includes in vivo repeated dose toxicity tests in laboratoryanimals. These in vivo studies often generate false negative results and unexpected toxicity. Theappearance of this unexpected toxicity is one of the major reasons for the drawback of a drug fromthe market. The liver is often a target organ in toxicology since it is responsible for the metabolismand elimination of chemical compounds. Therefore, there is need for new screening methods whichclassify hepatotoxic compounds earlier in development. This will lead to safer drugs and a more effi-cient drug discovery process. Furthermore, these new screening methods are preferably in vitro testsystems, aiming at reducing the use of laboratory animals. In this review the possibilities of proteomicsand its promising results for improving current predictive and mechanistic toxicological studies aredescribed. Biomarkers or protein panels for hepatotoxic mechanisms, which reflect the in vivo situation,need to be identified to allow a better toxicity screening. Therefore, in vivo studies and in vitro cellmodels are discussed and evaluated with regard to the protein expression of their metabolic enzymes,their similarities with liver, their use for analyzing toxicological mechanisms and hepatotoxicityscreening. Studies in which proteomics are combined with other omics-technologies are also presented.The results from these integrated data analyses can be used for the development of improved panels ofbiomarkers for toxicity screening.

� 2012 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3742. Proteomics techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374

2.1. Gel-based proteomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
2.1.1. Two-dimensional gel electrophoresis (2-DE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3742.1.2. Blue native gel electrophoresis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
2.2. Liquid chromatography tandem mass spectrometry (LC–MS/MS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
2.2.1. Isotope-code labeling shotgun technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3762.2.2. Label-free quantification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
2.3. Targeted proteomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
3. Proteomic studies on hepatotoxicity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
3.1. In vivo. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
3.1.1. Tissue proteomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3763.1.2. Proteomics of cellular fractionations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
ll rights reserved.

horesis; DIGE, differential gel electrophoresis; ESI, electrospray ionization; ER, endoplasmatic reticulum; FCS, fetal calfass spectrometry; LTQ-FTICR, Linear Trap Quadrupole Fourier Transform Ion Cyclotron Resonance mass spectrometry;RM, multiple reaction monitoring; PHH, primary human hepatocytes; SRM, selective reaction monitoring.

fax: +31 (0) 43 3670976.rsity.nl (A. Van Summeren).

http://dx.doi.org/10.1016/j.tiv.2012.01.012

mailto:[email protected]

http://dx.doi.org/10.1016/j.tiv.2012.01.012

http://www.sciencedirect.com/science/journal/08872333

http://www.elsevier.com/locate/toxinvit

374 A. Van Summeren et al. / Toxicology in Vitro 26 (2012) 373–385

3.2. In vitro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
3.2.1. Precision cut liver slices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3783.2.2. Primary hepatocytes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3783.2.3. Cell lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3793.2.4. Stem cell-derived hepatocytes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3803.2.5. In vitro secretome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
4. Proteomics in relation to other omics techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3815. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381

Conflict of interest statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383

1. Introduction

In the development of new drugs the appearance of unexpectedtoxicity is one of the major reasons for the withdrawal of the prod-uct from the market. With respect to toxicity testing hepatotoxicityis most prominent, because after oral intake the compounds will betransported to the liver, where they will be metabolized to be elim-inated. Currently hepatotoxicity is evaluated in 28 day in vivo re-peated-dose toxicity tests by analysis of clinical parameters,hematology, and histopathology. However, these parameters areoften insensitive and can generate false negative results (Olsonet al., 1998; Suter et al., 2010). Consequently unexpected hepato-toxicity often appears in the clinical trials or even when the prod-uct is already on the market (Hartung, 2009). This unexpectedhepatotoxicity involves dose-dependent reactions, which slippedthrough the pre-clinical trials. However, it includes also idiosyn-cratic drug reactions. These are rare, unpredictable and mostly im-mune-mediated reactions appearing only in a large population(Knowles et al., 2000; Uetrecht, 2008). This unexpected hepatotox-icity in the development of new drugs emphasizes the need for no-vel screening methods that address toxicological hazards early inthe developmental process (Amacher, 2010). These new screeningmethods are preferably applied on in vitro test systems to reducethe number of laboratory animals. The omics-technologies have al-ready shown promising results for improving the current toxicitytests, in particular transcriptomic based screenings. However,studying the transcription level of a gene only gives a rather roughestimate of its corresponding protein expression level. With pro-teomics the functional molecules, are investigated which may re-veal new biomarkers and toxicity signatures for preclinical safetyassessment and disease diagnostics (Amacher, 2010). As such abetter understanding of the underlying molecular mechanisms-of-action can be achieved. Proteome technologies used for predic-tive or mechanistic toxicological research with acute or long-termexposure to toxicants is known as toxicoproteomics, which wasfirst introduced by Wetmore and Merrick (2004). With proteomicsnot only protein expression is monitored, but also post-transla-tional modifications and protein interactions. These cannot bemonitored by transcriptomics but are of substantial value forobtaining full insight in toxicity mechanisms. Moreover, unlikeDNA and RNA, many proteins are secreted into body fluids or cellculture medium (secretome). In the search for toxicity biomarkersthese secreted proteins may play a significant role, especially in li-ver research. The liver is responsible for the production and secre-tion of a large variety of plasma proteins. Since the liver is also amajor target for drug-induced toxicity, changes in secreted proteinprofiles can reveal relevant toxicity signatures.

Drug-induced hepatotoxicity may have several features, forexample: hepatic steatosis, cholestasis, inflammation, fibrosis,(steato)hepatitis and cytotoxicity caused by oxidative stressnecrosis and apoptosis. Classification of hepatotoxicity is difficultbecause some injuries may occur simultaneously like steatosis

and hepatitis. With proteomics individual proteins or proteinpanels reflecting defined hepatotoxic mechanisms can beidentified that improves the possibilities for classification.However, developing an in vitro toxicity testing assay based onprotein analysis requires the validation of proteome responsesof hepatotoxic compounds against in vivo data (Slany et al.,2010). Omics techniques provide the opportunity to measure sim-ilar endpoints of compound-induced changes, which enablesin vitro to in vivo comparison. In contrast, conventional toxicolog-ical parameters require the analysis of histopathological or bloodchemistry parameters, which are often not applicable to in vitromodels.

In a systems biology approach conventional toxicological as-says, proteomics and other omics-applications can be integrated.This information will contribute to the development of an alterna-tive model for hepatotoxicity testing characterized by a better pre-diction of human liver toxicity and a reduction of the number oflaboratory animals to be used. Furthermore, the application of highthroughput screening methods will lead to more cost-effectivetoxicity testing. In this review we will summarize strategies andresults of proteomic in vivo and in vitro investigations on drug-induced hepatotoxicity.

2. Proteomics techniques

The objective of differential proteomics is to separate, quantifyand identify the proteins and peptides in a complex mixture inorder to reveal differences of protein expression between experi-mental conditions. The main current technologies applied forprotein separation are either gel-based (one- or two-dimensionalgel electrophoresis 1-DE or 2-DE), gel-free techniques (liquidchromatography tandem mass spectrometry (LC–MS/MS)) or acombination of both (Fig. 1) (Barrier and Mirkes, 2005). These ap-proaches are complementary since they focus on subsets of pro-teins that are only partially overlapping. The separation methodsare combined with tandem mass spectrometry for the identifica-tion of proteins. Here is chosen to focus on 2-DE and LC–MS/MS, since they are widely used in the studies discussed in this re-view. The pros and cons of both technologies are summarized inTable 1.

2.1. Gel-based proteomics

2.1.1. Two-dimensional gel electrophoresis (2-DE)In 2-DE, the protein sample is separated first by iso-electric

focusing on an immobilized pH gradient gel under the influenceof an electric field. The proteins are focused in the gel accordingto their isoelectric point. Secondly, the proteins are separatedaccording to their molecular mass by SDS–polyacrylamide gel elec-trophoresis. The visualization of proteins after separation can beachieved by non-specific protein staining either with Coomassie

Fig. 1. Overview of the basic principles of DIGE and LC–MS/MS the two mainly used proteomics techniques.

Table 1Pros and cons of gel based.

Gel based proteomics Gel-free proteomics

+ – Accurate and relatively easy quantification– Indication of molecular weight and pI so post-translational

modifications are visualized

– High throughput– Suitable for most proteins

� – Unsuitable for all proteins (depending their concentration,hydrophobicity, pI and molecular weight.)

– High abundant proteins can cover low abundant ones by co-migration (which can be solved by fractionation)

– Matching and comparison of the gel images is timeconsuming

– Quantification is more challenging which require expensive isotope labels (however, label-free methods are available), specific software, and expertise to analyze data

– High abundant peptides can cover low abundant ones by co-elution (which can be solved byfractionation)

A. Van Summeren et al. / Toxicology in Vitro 26 (2012) 373–385 375

blue, silver, or with fluorescent dyes that are currently used mostfrequently (Gorg et al., 2004). Fluorescent dyes have a high sensi-tivity and cover a large dynamic range. After staining the gels arescanned and the protein spots are quantified according their spotintensities. Based on these spot intensities differential proteinspots are determined. The main technical drawback of traditional2-DE is the comparison of various gels corresponding to individualsamples. Because gels are treated separately the risk for technicalvariation such as alterations in protein migration patterns andstaining variability is high.

The intrinsic technical variations of classical 2-DE has beenstrongly reduced by the introduction of differential gel electropho-resis (DIGE). With this technology proteins are labeled with fluo-rescent dyes prior to separation (Karp et al., 2008). Threedifferent fluorescent dyes are available, that allow labeling of twoprotein samples and one internal standard, which is a mixture ofan equal amount of each protein sample. These labeled proteinsamples are mixed and separated on one gel according to standard2-DE methods. Subsequently, the gel is scanned and analyzed bydedicated software. The possibility to run two different protein


samples on a single gel as well as the introduction of an internalstandard on each gel considerably reduces the gel-to-gel variabil-ity, thereby improving the classical 2-DE. By this approach theDIGE technology has overcome the limitations of classical 2-DE.For identification of the differentially expressed proteins the pro-tein spots are picked from the gel, digested and introduced intothe mass spectrometer by electrospray ionization (ESI) or spottedon a target plate for matrix assisted laser desorption ionization(MALDI) mass spectrometry. Overall, the strength of the 2-DE ap-proach is determined by the accurate and relatively easy quantifi-cation of the differentially expressed proteins. Moreover, 2-DEindicates the molecular weight and pI of proteins. So, different pro-tein iso-forms are easily visualized. Despite the major improve-ment by the DIGE technology, the 2_DE approach is stillhampered by co-migration of different proteins on the gel, andhas limited applicability for hydrophobic (membrane) proteinsand proteins with an extreme pI. Moreover, accurate matchingand comparison of the gel images for protein quantification is atime consuming process. This is one of the reasons why shotgunproteomics (Section 2.2) gained more interest in the last decade.

2.1.2. Blue native gel electrophoresisAs stated, the 2-DE approach is less suitable for separation of

more hydrophobic membrane proteins. To overcome this limita-tion one can consider blue native polyacrylamide gel electrophore-sis (BN-PAGE). In BN-PAGE, the first dimension separation isperformed under non-denaturating conditions, which preserveprotein–protein interactions and protein complexes, like the oxi-dative phosphorylation system. Therefore BN-PAGE can provideinformation about the size, number, stoichiometry, protein compo-sition and relative abundance of protein complexes.

2.2. Liquid chromatography tandem mass spectrometry (LC–MS/MS)

As an alternative for gel-based protein separation, gel-free anal-ysis of complex protein samples by LC–MS/MS has become popularin the last decade, particular the so-called shotgun proteomics ap-proach. With this method protein samples are first digested intocomplex mixtures of peptides. The peptides are subsequently sep-arated by reversed-phase high-performance LC which is directlycoupled to a mass spectrometer by an ESI interface. As an alterna-tive the separated peptides can be automatically spotted on a MAL-DI target plate for MS/MS identification.

Shotgun proteomics enables the detection of more hydrophobicproteins and is considered as high throughput, since thousands ofpeptides are simultaneously analyzed. However, due to the com-plexity of the samples, the protein quantification is more challeng-ing. As such, both label-based and label free protein quantificationmethods have been established for LC–MS/MS-mediated proteomeanalyses.

2.2.1. Isotope-code labeling shotgun technologiesAs gel-free counterpart for DIGE, isotope-coded labeling shot-

gun technologies were developed. Proteins are isotope-labeledprior to separation, e.g. by iTRAQ-Isobaric Tags or by stable isotopelabeling with amino acids in cell culture (SILAC). Quantificationand comparison are performed by detection of the introduced labelduring MS analysis (Wilm, 2009). Isotope-coded labeling shotgunapproaches are considered as accurate technologies for proteinquantification. Though, they require expensive isotope labels, spe-cific software, and expertise to analyze data (Neilson et al., 2011).

2.2.2. Label-free quantificationLess expensive quantification methods are label free approaches

that rely on spectral counting or signal intensity measurementduring MS. With spectral counting it is assumed that when more

abundant peptides are selected for fragmentation, they will pro-duce a higher abundance of MS/MS spectra, corresponding withthe protein amount (Neilson et al., 2011). However, spectral count-ing is rather susceptible to variation. Signal intensity measurementor area under the curve (AUC) relies on the ion abundances at spe-cific retention times for the given ionized peptides, often referredto as ion counts. Partly due to the availability of optimized soft-ware, the label-free quantification methods are currently gainingmore popularity, compared to the more expensive label-basedmethods (Neilson et al., 2011).

2.3. Targeted proteomics

Although both 2-DE and shotgun proteomics are successfultechniques to identify candidate biomarkers, validation is requiredto ensure that those proteins are specific for a certain biologicalcondition. Usually such validations target the protein of interestby applying antibody-based techniques like Western blotting,ELISA or immunohistochemistry. These convenient techniquesare widely accepted for the verification and validation of biomark-ers. However, when multiple markers or samples need to be ana-lyzed, they are rather expensive and time consuming. Moreover,suitable antibodies are not available for all potential biomarkers.Therefore, multiple reaction monitoring (MRM) or selective reac-tion monitoring (SRM) was developed. With this targeted MS ap-proach only the known peptides of interest are selected, whileothers are ignored which allows the parallel analysis of specificmultiple peptide transitions (Meng and Veenstra, 2011). ThereforeMRM offers the high throughput and specificity required for theverification of multiple candidate biomarkers. MRM requires a tri-ple quadrupole mass spectrometer, where in the first quadrupole(Q1) the known precursor ion is isolated, in Q2 the ion is frag-mented and in Q3 the optimum fragment ions are monitored(Meng and Veenstra, 2011). Recently proof-of-principle has beendelivered by reliable detection of CYP450 enzymes and Udp-glu-curonosyltransferases in liver by means of MRM (Sakamoto et al.,2011).

3. Proteomic studies on hepatotoxicity

3.1. In vivo

In order to be recognized as safe for human consumption orexposure, new compounds need to be tested on rodents or othermammals. In the pharmaceutical industry rats are the preferredmodel because of their size, ease of manipulation, breeding charac-teristics, short life span (±3 years) and because of the in-depthknowledge of the model gained through many years of experience(Aitman et al., 2008). In vivo studies usually involve 28 or 90 dayrepeated-dose toxicity tests to evaluate chronic effects, organ tox-icity and to establish a non-observable effect level. During these re-peated-dose toxicity tests the animals are observed for indicationsof toxicity. Afterwards, necropsy, blood analysis and histopathol-ogy of the organs of the animals are performed.

With in vivo proteomics studies a further insight in the hepato-toxic mechanisms, individual proteins or protein panels reflectivefor hepatotoxicity can be revealed. Together with other omics tech-nologies, proteomics techniques have the potential to detect toxi-cological effects at earlier time points and lower doses comparedas the conventional toxicity assays.

3.1.1. Tissue proteomicsProteomics investigation towards acetaminophen toxicity in

mice was performed by Fountoulakis et al. (2000). The liver sam-ples of this study contained 35 modified proteins after acetamino-


phen treatment. Some of these proteins were known targets forcovalent modification of N-acetyl-p-benzoquinoneimine, which isthe most toxic metabolite of acetaminophen (Fountoulakis et al.,2000). This demonstrates that proteomics can identify protein tar-gets for toxic compounds in vivo. For a focused search of these pro-tein targets, Koen et al. used radioactively labeled compounds todetect targets for the reactive metabolites of bromobenzene (Koenet al., 2007). Mice were treated with 14C-bromobezene. 2-DE wasapplied to separate the liver proteins and bromobenzene-targetproteins were visualized after measuring the radioactivity of thespots in the gel. In total 33 unique protein targets for bromoben-zene were detected including glutathione S-transferases, proteindisulfide isomerases and liver fatty acid-binding protein (Koenet al., 2007).

In addition, proteomics can be used to provide insight in toxic-ity mechanisms. One example is the study of Low et al. were thi-oacetamide-induced hepatotoxicity in rat liver was examined(Low et al., 2004). This study showed a down-regulation of en-zymes from the metabolic pathways of fatty acid beta-oxidation,branched chain amino acids and methionine breakdown. Further-more, an up-regulation of proteins related to oxidative stress andlipid peroxidation was detected. Protein expression in rat liverafter administration of hydrazine, another model toxin for hepato-toxicity, was also investigated (Kleno et al., 2004). Here, the hydra-zine treatment induced altered expression of proteins related tolipid metabolism, calcium homeostasis, thyroid hormone path-ways and stress response (Kleno et al., 2004). As part of the EU Inn-omed Predtox Project, liver samples from rats treated 14 days withtroglitazone were analyzed with DIGE (Boitier et al., 2011). Thisstudy showed 104 deregulated protein spots, from which 55 spotswere identified by Maldi TOF/TOF. These proteins belong mostly tothe pathways of fatty acid metabolism, PPARa/RXR activation, oxi-dative stress and cholesterol biosynthesis. These affected pathwayswere also found deregulated in similar transcriptomics experi-ments (Boitier et al., 2011).

To generate marker panels for carcinogenicity and genotoxicityproteome analysis was used in an in vivo 28-day repeated dosestudy of 63 chemical compounds (Yamanaka et al., 2007). The pro-teome of rat livers was analyzed by DIGE and the carcinogeniccharacteristic proteins were classified. In this study, 79.3 % of thegenotoxic compounds and 76.5% of the non-genotoxic compoundscould be correctly classified (Yamanaka et al., 2007). In a compara-ble gene expression signature for predicting non-genotoxichepatocarcinogens, Fielden et al. (2007) found an accuracy of63–69%, whereas the gene expression signature accuracy forpredicting non-genotoxic carcinogens was between 55% and 64%(Nie et al., 2006). Although these accuracies still involve a highnumber of false negative results, proteomics techniques showpromising results in the detection of toxicity signatures.

3.1.2. Proteomics of cellular fractionationsIn vivo proteomics studies like those described above investi-

gated the overall cellular protein expression. However, when thetotal proteome of a cell type is examined by 2-DE, low abundantdifferential proteins are usually overshadowed by high abundantproteins. Furthermore, hydrophobic proteins, e.g. membrane pro-teins can be absent from the cell lysate because they are difficultto dissolve. Low abundant or hydrophobic proteins can be enrichedby fractionation of the proteome in subcellular fractions like mito-chondria, endoplasmatic reticulum, microsomes or the cell mem-brane (Brunet et al., 2003).

3.1.2.1. Endoplasmatic reticulum and microsomes. The endoplasmicreticulum (ER) is a key organelle for protein secretion and is in-volved in the synthesis of both proteins and lipids (Lavoie and Pai-ement, 2008). Moreover, the ER is involved in the detoxification of

xenobiotic compounds, which explains its attractivity for toxico-logical studies (Lavoie and Paiement, 2008). To acquire in-depthknowledge of the cellular and ER protein expression, Zgoda et al.(2006) has fractionated the proteome of liver samples into a cyto-solic and a microsome fraction. Microsomes are small vesicles de-rived from the ER when the cells are mechanically disrupted andcontain high amounts of cytochrome P450 metabolizing enzymes(Seliskar and Rozman, 2007). The liver samples were derived frommice treated with phenobarbital and 3-methylcholanthrene andthe fractions were subjected to 2-DE (Zgoda et al., 2006). Despitethe fractionation, the cytosolic fraction revealed more informationthan the subcellular microsome fraction. The microsomal changesafter treatment with phenobarbital and 3-methylcholanthrenewere quite similar. While the cytosolic response after phenobarbi-tal and 3-methylcholanthrene treatment could be clearly distin-guished from each other (Zgoda et al., 2006). A high amount ofCYP450 enzymes was expected in the microsomal fraction, how-ever only two CYP450 enzymes were detected. This can be ex-plained that CYP450 enzymes are membrane-associated proteinswhich are difficult to analyze by means of 2-DE. Therefore, shotgunproteomics is probably more suitable for analyzing the differentialexpression of cytochrome P450 metabolizing enzymes (Galeva andAltermann, 2002; Nisar et al., 2004). For example, microsomes iso-lated from carbon tetrachloride (CCl4)-treated rat liver were usedto analyze the changes in protein expression of the CYP450 en-zymes (Jia et al., 2007). From a 1-DE gel the area from 45 to66 kDa was excised, since this corresponds with the molecularweight of most CYP450 proteins. Subsequently an acetylation sta-ble isotopic labeling method coupled with Linear Trap QuadrupoleFourier Transform Ion Cyclotron Resonance mass spectrometry(LTQ-FTICR) was applied. With this approach the researchers wereable to identify and quantify 17 CYP450 proteins. Among them, theexpression of 2C11, 3A2, and 2E1 was down-regulated, while thatof 2C6, 2B2, and 2B1 was up-regulated (Jia et al., 2007).

3.1.2.2. Mitochondria. Mitochondria are responsible for the oxida-tion of fatty acids into acetyl-CoA through the b-oxidation. Impair-ment of this b-oxidation process, either drug-induced or throughlong-term imbalance between food intake and energy expenditure,will lead to fat accumulation in hepatocytes and ultimately steato-sis. Therefore, analyzing mitochondrial proteins can reveal mecha-nistic details in the development of drug-induced steatosis,although mitochondrial dysfunction is not exclusive for steatosis.Bailey et al. described 2-DE and BN-PAGE to analyze mitochondrialproteins in alcoholic liver disease (Bailey et al., 2008). Both gel-based techniques were used to study the mitochondrial liver pro-teome of rats with ethanol-induced hepatotoxicity (Venkatramanet al., 2004). In total 43 differentially expressed proteins were iden-tified, comprising enzymes of the b-oxidation cycle, nuclear en-coded subunits of the oxidative phosphorylation system,mitochondrial chaperones and enzymes of amino acid metabolism.Furthermore a decrease of several polypeptides of the respiratorycomplexes in the ethanol treated rats was observed.

3.1.2.3. Plasma/serum proteome. Since most proteins in blood aresecreted by the liver, plasma/serum is a potential source for hepa-totoxicity biomarkers, represented by aberrantly secreted proteinsor proteins leaked from the liver due to injury. This is demon-strated by the current liver function tests were the enzymaticactivity of alanine aminotransferase, aspartate aminotransferase,gamma glutamic transpeptidase, lactic dehydrogenase and alkalinephosphatase are measured in blood. In addition, blood is easilyaccessible and practical for repeated sampling and analysis. Re-cently, protein expression in Z24-treated and untreated rat liverswas studied in parallel with the plasma proteome from the animals(Wang et al., 2010). Z24 is a synthetic anti-angiogenic compound


that inhibits growth and metastasis of tumors. However, it wasshown that Z24 induces hepatotoxicity in rodents. The rats wereadministered once daily with Z24 for 5 days at doses of 0, 50,100, or 200 mg/kg per day. Twenty-four hours after the finaladministration, blood samples and whole livers were collected.From these samples differentially expressed proteins were ana-lyzed with 2-DE and MALDI-TOF/TOF MS, 22 non-redundant liverproteins and 11 plasma proteins were found differentially ex-pressed. These proteins are involved in several important meta-bolic pathways, including carbohydrate, lipid, amino acid, andenergy metabolism, biotransformation, and apoptosis. Most ofthe identified hepatic proteins are located in mitochondria, whereZ24 also increased the ROS production and decreased the NAD(P)Hlevels. Therefore, it was concluded that Z24 inhibits the aerobiccarbohydrate oxidation, fatty acid b-oxidation, and oxidative phos-phorylation pathways resulting in mitochondrial dysfunction andapoptosis-mediated cell death. In addition, potential biomarkersfor Z24-induced hepatotoxicity, namely fetub protein and argini-nosuccinate synthase were detected in the plasma (Wang et al.,2010). These two proteins were also found differentially expressedtogether with several other proteins in the serum of rats 24 h afteracetaminophen exposure (Merrick et al., 2006).

3.2. In vitro

The current in vivo repeated-dose toxicity studies which includehepatotoxicity screening have only a limited reliability and involvenumerous animals. Therefore, properties of the possible non-ani-mal-based models and their ability to classify toxic compoundsneed to be evaluated as possible replacements. The most com-monly used in vitro models in hepatotoxicity studies are cell lineslike HepG2, HepaRG and primary hepatocytes or liver slices fromseveral species. Most frequently used species are rat, mouse andhuman (Farkas et al., 2005; Kienhuis et al., 2009, 2007; Mathijset al., 2009; Slany et al., 2010; Zwickl et al., 2005). In addition, re-cent developments in stem cell-derived hepatocytes have led totheir use in toxicological investigations (Greenhough et al., 2010).The baseline expression of the biotransformation genes in severalcell models (primary human hepatocytes, HepG2, HepaRG andstem-cell-derived hepatocytes) were compared with human liverand their biotransformation pathways were statistically ranked(Jennen et al., 2010). Both primary hepatocytes and HepaRG cellsshowed an expression pattern of the biotransformation genes sim-ilar to human liver samples, whereas the expression pattern ofHepG2 did not. The two analyzed stem-cell-derived hepatocyteshad only a few biotransformation genes with an expression com-parable with those from the normal liver samples and primaryhepatocytes (Jennen et al., 2010). According to the ranking test ofthe meta pathway for biotransformation, the liver cell modelscan be placed in following order: human liver, primary human

Fig. 2. Primary mouse hepatocytes (A) after isolation, (B) after a recuperation periodnetworks (indicated by arrows) will be formed.

hepatocytes, HepaRG > HepG2 > human stem cell models (Jennenet al., 2010). This study can be complemented with the results fromthe baseline protein expression of the cell models. Below, the basalprotein expression of the most commonly used in vitro models isdiscussed, including the results of hepatotoxic proteome studies.

3.2.1. Precision cut liver slicesHepatotoxicity is often the result of multi-cellular processes in

the liver. In precision cut liver slices the three-dimensional struc-ture of the liver is preserved and contains all the cell types presentin the liver in vivo. The presence of other cells than hepatocytes,like Stellate and Kupffer cells, adds a significant value to this modelas these cells often play an important role in the mediation of hep-atotoxicity (Guyot et al., 2006). On the other hand, the presence ofdifferent cell types will increase the variation in the study; conse-quently the slices must be isolated in a controlled manner. Fortheir use in toxicological research it is essential that these liverslices retain their metabolic activity to ensure the biotransforma-tion of xenobiotic compounds. Unfortunately the use of precisioncut liver slices is hampered due to their short life span, the poorpenetration of compounds to the inner cell layers, the limitedavailability of donors for human liver slices and the great inter-individual variation, of which the latter can be minimized by usingsufficient biological replicas. So far, studies on the toxicologicalproteome of liver slices have not been published, probably due tothe fact that liver slices are not commonly used. Therefore, thereis still limited knowledge about the intercellular interactions in li-ver toxicity mechanisms. Hence, if their life span, which is nowaround 16 hours, could be increased, the liver slice model wouldbecome increasingly important for hepatotoxicity testing in thefuture.

3.2.2. Primary hepatocytesPrimary hepatocytes are considered as ‘the gold standard’ for

in vitro xenobiotic metabolism and toxicity studies. They expressboth Phase I and Phase II enzymes, which are necessary for bio-transformation of xenobiotic compounds, and maintain other liverspecific functions. Considering these characteristics, primary hepa-tocytes will probably predict toxicity for xenobiotic compounds ina better way as compared to cell lines (Wang et al., 2002). Seglen(1976) introduced the two-step collagenase perfusion to obtainprimary hepatocytes from rats, a technique that is now increas-ingly used. The primary hepatocytes need to be cultured in a sand-wich configuration of extracellular matrix proteins, like collagen I.This sandwich configuration mimics the three-dimensional struc-ture of the liver that retains the original functions of the hepato-cytes and increases their life span (Dunn et al., 1991). Afterisolation, a recuperation period of 42 h needs to be taken into ac-count before the hepatocytes can be exposed to any toxicologicalcompound. In this recuperation period, gap junctions as well as

of 16 h and (C) after a recuperation period of 42 h. After 42 h the bile canalicular


bile canalicular networks will be formed between the hepatocytes(Mathijs et al., 2009), as can be seen in Fig. 2.

3.2.2.1. Rodent hepatocytes. The rat is the most frequently used ro-dent species for the preparation of primary hepatocytes. Until nowmost in vivo toxicity studies are performed in rats, therefore theprotein expression measured in vivo can directly be linked to pro-teome data from primary hepatocytes. For the prediction of drug-induced hepatotoxicity several studies have been conducted toinvestigate differential protein expression after administration ofhepatotoxic compounds (acetaminophen, amiodarone, tetracyclineand carbon tetrachloride) to rats in vivo or to isolated primary rathepatocytes (Kikkawa et al., 2005, 2006; Yamamoto et al., 2005,2006). The liver extract of rats exposed to hepatotoxicants con-tained differentially expressed liver-specific proteins (Yamamotoet al., 2006). Similar in vitro experiments with primary rat hepato-cytes revealed comparable differentially expressed proteins relatedto oxidative stress and mitochondrial metabolism regulation(Kikkawa et al., 2005; Yamamoto et al., 2005). These pathways of-ten indicate the development of necrosis. The results of thesein vitro studies appeared to be well in line with the in vivo studies(Kikkawa et al., 2006; Yamamoto et al., 2006). The chemically in-duced hepatotoxicity in primary rat hepatocytes was investigatedby exposing cells to eight different hepatotoxic compounds (afla-toxin B1, cadmium, N-methyl-N0-nitro-N-nitrosoguanidine, methylmethanesulfonate, carbon tetrachloride, N,N-dimethylformamide,vinyl acetate and acetaminophen) (Farkas and Tannenbaum,2005). Significant decreases in urea and albumin secretion werefound after treatments with Aflatoxin B1 and the direct acting tox-icants; however for the indirect toxicants no significant toxicitywas measured. The protein expression of the CYPs 1A, 2B, 3A2and 2E1 was determined by Western blotting, showing that thehepatocytes maintained CYP1A, 2B, 3A2 but gradually lost CYP2E1.

CYP2E1 is one of the main enzymes responsible for the meta-bolic activation of acetaminophen (Rumack, 2002), carbon tetra-chloride (Weber et al., 2003), and N,N-dimethylformamide(Tolando et al., 2001). The partial loss of CYP2E1 can explain theresistance of the hepatocytes to these indirect compounds.

Kikkawa et al. compared data from toxicological tests obtainedwith primary rat hepatocytes and an in vivo liver study (Kikkawaet al., 2006). Proteome analysis of the in vivo study revealed severalproteins related to oxidative stress and mitochondrial metabolism-regulation, which had been previously detected in primary rathepatocytes. With immunohistochemistry, oxidative stress-relatedproteins were detected in vivo prior to appearance of compound-specific histopathological changes. For that reason oxidativestress-related proteins are considered as possible biomarkers.Moreover, it shows that proteomics studies on primary rat hepato-cytes are useful to detect hepatotoxicity at an early stage (Kikkawaet al., 2006). A compound in preclinical development, referred asCDA, was found to induce hepatocellular steatosis in vivo. In a firstexperiment using DIGE analysis, it was shown that 250 mg/kg CDAinduced hepatotoxicity in rats 6 h after treatment (Meneses-Lor-ente et al., 2004). Several identified differential proteins could beassociated with known toxicological mechanisms involved in liversteatosis. These protein changes occurred before the onset of clin-ical biochemistry parameters, suggesting that these proteins couldrepresent potential early markers of CDA-induced hepatotoxicity(Meneses-Lorente et al., 2004). In an extended proteome experi-ment it was investigated whether this information was also repre-sented in primary hepatocytes (Meneses-Lorente et al., 2006).Pathway analysis of the differential proteins in hepatocytesshowed that CDA treatment predominantly affected cell deathand cellular assembly and organization. This included alterationsin secreted proteins, endoplasmic reticulum and mitochondrialchaperones, antioxidant proteins, and enzymes involved in fatty

acid biosynthesis. As such, the proteomic signature of in vitroCDA treatment showed a good correlation with the in vivo study(Meneses-Lorente et al., 2006).

The primary rat hepatocytes and to a lesser extent primary hu-man hepatocytes, are well-established in vitro systems. Primarymouse hepatocytes, however, are less commonly used in toxicitystudies although transgenic mouse models provide an opportunityto study specific toxicity mechanisms. Moreover, compared to pri-mary rat hepatocytes, primary mouse hepatocytes are more stableand remain their metabolic competence for a longer period (Math-ijs et al., 2009). Primary mouse hepatocytes from different strainsshowed only little differences between cellular function andexpression of liver specific enzymes. Furthermore, the cytotoxicityof three model compounds acetaminophen, WY-14,643 and rifam-pin did not differ much between the strains (Martinez et al., 2010).Therefore, primary mouse hepatocytes have not only potential fortoxicity screening, they can also serve as an alternative for popula-tion diversity (Martinez et al., 2010).

3.2.2.2. Primary human hepatocytes. Similar to human liver slices,studies with primary human hepatocytes are limited by the scar-city of suitable liver samples. Furthermore, they are hindered bya great inter-individual variation (Donato et al., 2008). The isola-tion of human hepatocytes is also based on the two-step collage-nase perfusion of Seglen (1976) but adapted to the ex vivotreatment of human liver from a donor (Pichard et al., 2006). Untilnow the proteome of primary human hepatocytes has not fre-quently been evaluated in hepatotoxicity studies. One studyshowed the effects of bezafibrate on the proteome of primary hu-man hepatocytes (Alvergnas et al., 2011). In contrast to primary rathepatocytes, the expression of CYP4A was not induced in primaryhuman hepatocytes. This study indicates the differences of thedrug-metabolizing enzymes between species, and emphasizesthe use of primary human hepatocytes to overcome issues relatedto interspecies variation (Alvergnas et al., 2011).

3.2.3. Cell linesLiver cell lines are commonly used in hepatocellular and toxic-

ity studies (Choi et al., 2010; Slany et al., 2010; Thome-Kromeret al., 2003; Van Summeren et al., 2011). These models are a rela-tively simple system, since they are easy to maintain and cultivate.Several studies compared cell lines with primary cells. Most ofthese studies are based on the comparison of gene expression pro-files at the transcriptome level which poorly reflect the differencesat the proteome level (Pan et al., 2009). Slany et al. (2010) applied2-DE analysis and shotgun proteomics to compare cellular and se-creted proteins from primary human hepatocytes with the hepa-toma cells HepG2 and Hep3B. Both 2-DE and shotgun proteomicsof the cellular proteome lead to the conclusion that HepG2 showsmore liver specific proteins than Hep3B, while Hep3B displayedmore commonalities to skin fibroblasts. The cellular proteome re-vealed 104 liver specific proteins in primary human hepatocytes,20 in HepG2, and only six in Hep3B. With shotgun proteomics,32 detoxifying proteins were found in primary human hepatocytesand only four in HepG2 and two in Hep3B. This illustrates the lossof the detoxifying capacities in hepatic cells lines compared to pri-mary hepatocytes. With respect to secreted proteins, 46 out of 72proteins identified in the secretome of primary hepatocytes wereplasma proteins characteristic for hepatocytes. In HepG2 this was55 out of 139, while in Hep3B this was only 24 out of 72 (Slanyet al., 2010). Analyses of the secretome showed that HepG2 cellssecrete a considerable amount of plasma proteins such as albumin,serotransferrin, apolipoproteins and fibrinogen, which is a charac-teristic function of hepatocytes (Slany et al., 2010). Hep3B cellshowever secreted only a relatively small amount of plasma pro-


teins, while other secreted proteins were not characteristic for livercells (Slany et al., 2010).

The suitability of HepG2 as an in vitro cell system for toxicityscreening was investigated by comparing the differential proteinexpression of HepG2 cells with the in vivo protein expression inrats after administration of six hepatotoxicants (1-naphthyl-isothi-ocyanate, indomethacine, acetaminophen, cisplatin, tetracyclineand dimethyl-nitrosamine) and four non-toxic compounds (aspi-rin, isoproterenol, phenylephrine and DMSO) (Thome-Kromeret al., 2003). This comparative proteome analysis demonstratedthe functional differences between HepG2 cells and the in vivo sit-uation. The in vivo rat system rather reflected the full, althoughspecies-specifically biased, dynamics of a liver which could notbe completely replaced by HepG2 cells. Nevertheless, 8 of 13 po-tential toxicity marker proteins found in rat liver were also de-tected in HepG2 (Thome-Kromer et al., 2003). In addition, HepG2cells showed the quality of an in vitro testing system for detectinghepatotoxicity in an early stage of drug discovery (Thome-Kromeret al., 2003). Recently we have investigated drug-induced hepato-toxicity in HepG2 induced by amiodarone, cyclosporine A and acet-aminophen, which induce steatosis, cholestasis and necrosis,respectively (Van Summeren et al., 2011). The changes in proteinexpressions in HepG2 after exposure to these test compounds werestudied using DIGE. In this study cyclosporin A treatment wasresponsible for most differentially expressed proteins. The identi-fied differential proteins showed that cyclosporine A may induceER stress and disturbs the ER-Golgi transport, resulting in alteredvesicle mediated transport and protein secretion. Moreover, path-ways related to cholestatic mechanisms were affected. Therefore,our findings indicate that HepG2 have distinctive characteristicsenabling the assessment of cholestatic properties of novel com-pounds (Van Summeren et al., 2011).

The proteome of the mouse hepatoma cell line Hepa1-6 wascompared with primary mouse hepatocytes (Pan et al., 2009). Bio-informatic analysis of both proteomes showed a deficient mito-chondrial activity in Hepa1-6 cells, reflecting re-arrangement ofmetabolic pathways, drastically up-regulated cell cycle-associatedfunctions, and largely inactive drug metabolizing enzymes (Panet al., 2009). From the comparative studies between primary hepa-tocytes and hepatic cell lines it can be concluded that hepatomacell lines still have liver-specific functions and can be a valuabletool in toxicity screening. Nevertheless researchers should beaware of the fading of several specific functions, such as activityof cytochrome P450 enzymes due to the immortalization of thecells. This is demonstrated in the hierarchical clustering analysisof the expressed proteins of nine hepatocellular carcinoma celllines and primary human hepatocytes from five individuals, whereall cell lines were distinguished from the primary hepatocytes (Fu-jii et al., 2006). Therefore these cell lines will not always be suc-cessful in hepatotoxicity screening, particularly with respect toxenobiotics which need biotransformation/activation before theybecome toxic.

3.2.4. Stem cell-derived hepatocytesRecently, methodologies have been worked out for obtaining

hepatocyte-like cells derived from pluripotent stem cells originat-ing either from embryos (human embryonic stem cells, hESCs) orfrom non embryonic tissues (induced pluripotent stem cells, iPS-Cs). The hESC cells are derived from the inner cell mass of a blasto-cyst (Thomson et al., 1998). The iPSC cells are derived through thetreatment of fibroblasts, from healthy or patient donors, with stemcell factors (Sox2, Klf4, Oct4 and c-Myc) (Greenhough et al., 2010;Takahashi and Yamanaka, 2006). Both hESCs and iPSCs are capableof self-renewal and differentiation into hepatocytes. The use of iPS-Cs cells has the advantage that it is ethically more acceptable sinceit does not require human blastocysts. The use of pluripotent stem

cells is subjected to inter-individual genetic variation, howeveriPSCs from patient donors (with a known genotype) can be usedfor the development of polymorphic and disease libraries (Greenh-ough et al., 2010). Recently the use of hESC in proteomics has beenreviewed by Hughes et al. (2011). However the use of stem cell-de-rived hepatocytes is still limited without published hepatotoxicitystudies on the proteome of these cells so far. Nevertheless, thesemodels show promising results for their metabolic capacities andthus may become increasingly important for toxicity screening inthe future.

3.2.5. In vitro secretomeThe secretome of cells and tissues is a rich source of potential

markers, since it reflects a broad variety of pathological conditions.Moreover, secreted biomarkers are readily accessible and cantherefore easily be measured in the cell culture medium. One ofthe major functions of the liver is the production of serum proteins.Consequently, when specific protein markers for hepatotoxicitycan be found in the secretome of hepatocytes, it is likely that thesemarkers are present in serum and available for diagnosing hepato-toxicity in patients or subjects from clinical trials. However, theanalysis of proteins secreted by cultured cells is often a challengebecause they can be masked by non-secreted proteins. Cultivationof cells is unavoidably accompanied by cell death. Consequently,significant amounts of cytoplasmic proteins may be released intothe medium. By selective inhibition of secretion pathways it is pos-sible to discriminate secreted from non-secreted proteins. Theinhibition of these pathways can be obtained for example with Bre-feldin A that is known to inhibit protein secretion by interferingwith the function of the Golgi apparatus. Alternatively, incubationat 20 �C inhibits both the ER/Golgi-dependent and independentpathways (Wang et al., 2004). The differentially expressed proteinsbetween the not inhibited and the inhibited conditions are consid-ered as genuinely secreted. Ideally this is verified with bioinfor-matics tools. To improve the analysis of secreted proteins, Zwicklet al. (2005) have used metabolic labeling of proteins that are syn-thesized during a limited incubation period of HepG2 and humanliver slices and combined it with 2-DE. Whereas fluorescent stain-ing of the gel led to the detection of a large number of proteins de-rived from residual plasma and dead cells, the autoradiographsselectively displayed genuinely secreted proteins. Therefore thistechnique can improve the specific detection of secreted proteins(Zwickl et al., 2005).

Recently, the proteins secreted by HepG2/C3A cells in responseto ethanol exposure were investigated (Lewis et al., 2010). All dif-ferentially expressed proteins from this study are related withknown in vivo effects of ethanol exposure. These effects vary fromapoptosis and inflammation to cell leakage from disturbed cells,indicating that this model can be used for the identification of po-tential toxicity markers (Lewis et al., 2010). Moreover, the value ofthe secretome of HepG2 is confirmed by a study of Choi et al.(2010) where potential biomarkers for the genotoxic compounddi(2-ethylhexyl)phthalate were found in the secretome of exposedHepG2 cells (Choi et al., 2010). In this study, the proteins were ana-lyzed by 2-DE using two different pI ranges (4–7 and 6–9) com-bined to a large size 2-DE gel. This revealed 35 differentialproteins belonging to several functional groups (Choi et al.,2010). Based on the differentially expressed proteins, di(2-ethyl-hexyl)phthalate was found to affect the formation of cell structure,apoptosis, and tumor progression.

In the study of Farkas et al. (2005) the secretome of primary rathepatocytes exposed to aflatoxin B1 showed a decreased expres-sion of a2-macroglobulin and a1-antitrypsin. Patients with a1-antitrypsin deficiency have an increased risk for liver carcinomaand cirrhosis; these symptoms are also seen with aflatoxin B1intoxication. Therefore, it is likely that the decreased expression


of these protease inhibitors in the medium is linked to an impair-ment of the liver.

In another study the medium of human immortalized hepato-cytes with an over-expression of the CYP3A4, the most commontype of CYP450 enzyme in human liver, was studied after exposureto hepatotoxic (troglitazone, ciglitazone, farglitazar and ritonavir)and non-hepatotoxic compounds (DMSO, indinavir, rosiglitazoneand tesaglitazar) (Gao et al., 2004). The cells were incubated withthe compounds for 20 h and afterwards the conditioned mediumwas analyzed with LC–MS/MS. This analysis revealed two proteins,referred to as BMS-PTX-265 and BMS-PTX-837, that were signifi-cantly increased in the secretome of the cells treated with eachof the hepatotoxic compounds. The response of these two proteinsto an expanded set of 20 compounds was further analyzed. For all20 drugs, elevations of BMS-PTX-265 correlated exactly with theknown safety profile; whereas changes in BMS-PTX-837 correctlypredicted the safety profile from 19 of the 20 drugs (one false neg-ative) (Gao et al., 2004). This study exemplifies that proteomicsand in particularly secretome analysis can reveal biomarkers forhepatotoxicity screening.

4. Proteomics in relation to other omics techniques

The primary effects of toxicants on gene expression are trig-gered upon binding to receptors, whereas secondary effects ongene expression are to be expected as a result of changes in cellularprocesses (Heijne et al., 2003). Therefore, to obtain a complete pic-ture of the toxicological processes transcriptomics, proteomics andmetabolomics should be studied in parallel. A systems toxicologyapproach, which included proteomics (2-DE), metabolomics (1HNMR spectroscopy) and transcriptomics (micro-array), was usedto analyze methapyrilene-induced hepatotoxicity in the rat (Craiget al., 2006). Methapyrilene induced changes on gene, proteinand metabolite level and affected the following pathways: stress,urea cycle, glucose-, lipid-, choline and phenylalanine-metabolism.

Bromobenzene-induced hepatotoxicity was analyzed in a com-bined transcriptomics and proteomics experiment (Heijne et al.,2003). By using principle component analyses the treated samplescould be clearly distinguished from control samples both on thetranscriptome and the proteome level. However, when both datawere compared to each other only a modest overlap between thetranscriptomics and proteomics data was found. This was partiallybecause with proteomics usually a lower amount of deregulatedproteins are identified compared to the significantly altered genesfound with microarray analysis. Moreover, some proteins arestored in the cytoplasma and released upon the insult before theirtranscription is initiated to refill the cytoplasmic store. Anotherexplanation is that differentially expressed proteins detected withproteomics are often differential due to posttranslational modifica-tions, which do not appear at the transcriptome level.

The gene and protein expression in rat liver after acetamino-phen administration was simultaneously investigated (Rueppet al., 2002). Several proteins that changed due to acetaminophentoxicity could be identified. Also in this study the transcriptomicsand proteomics profiling generally did not detect the simultaneousexpression of the mRNA’s and the corresponding proteins. How-ever, the pathways identified by transcriptomics and proteomicsprofiling told a similar story (Ruepp et al., 2002). Despite the widerange of omics-data analyses and the different ways of interpretingthese results, the overall outcome is usually independent of thetools applied (Suter et al., 2010).

Several public or private partnerships have made an effort tointegrate omics data and techniques for a better mechanisticunderstanding of toxicity. The Chemical Effects in Biological Sys-tems (CEBS) is an integrated public repository for toxicogenomics

data, which allows the user to integrate various data types and al-lows studies on the early prediction of key toxicities. It is availableat http://cebs.niehs.nih.gov and described by Waters et al. (2008).This database includes 27 toxicogenomics studies from which 22are of rat, four are of mice and one is of Caenorhabditis elegans(Waters et al., 2008). However, 26 studies are based on transcripto-mics data and only one contains proteomics data.

In the EU Framework 6 project: Predictive Toxicology (PredTox),the effects of 16 different compounds were analyzed by severalomics techniques together with conventional toxicological param-eters (Suter et al., 2010). The compounds used in this project in-cluded 14 pharmaceuticals which had been discontinued atcertain stages of preclinical development due to toxicological find-ings in liver and/or kidney after 2- to 4-weeks in systemic rat stud-ies. Beside those 14 proprietary drug candidates, two referencetoxic compounds (gentamicin and troglitazone) where used (Suteret al., 2010). These studies were focused on hepato- and nephro-toxicity caused by the selected compounds. From this study liver,kidney, blood and urine samples were collected for transcripto-mics, proteomics and metabolomics. The proteomics studies wereperformed using DIGE, which experienced a low throughput. Incontrast, the transcriptomics studies were done with micro-arrayswhich allowed investigation of the expression levels of thousandsof genes. However, proteome analysis could identify toxicologicalmechanisms and revealed similar results as the transcriptomeanalysis (Boitier et al., 2011). For the NMR analysis urine was themost informative biofluid which could discriminate betweenmetabolites that were confirmed by histopathology. Compared toproteomics and metabolomics, target organ transcriptomics wasthe most revealing omics approach for the generation of mechanis-tic models (Matheis et al., 2011; Suter et al., 2010). In the Liver Tox-icity Biomarker Study (LTBS), the biochemical effects of drugswhich can cause hepatotoxicity were studied (McBurney et al.,2009). Eventually five compound pairs, with each a similar thera-peutic mechanism, will be analyzed in a 28-day rat study in whichliver and body fluids are collected in order to perform transcripto-mics, proteomics and metabolomics. The preliminary results fromthe first phase of the project revealed two compounds (entacaponeand tolcapone) with promising marker analytes (McBurney et al.,2009). The LTBS study illustrates the value of the broad-spectrum,molecular systems analysis to study hepatotoxic effects.

5. Conclusions

Biological processes are normally driven by proteins; thisemphasizes the need of proteomics investigations in mechanisticand predictive toxicology studies. Currently, either gel-based orgel-free techniques are frequently used. With DIGE the technicallimitations of the classical 2-DE are improved, allowing a quantita-tive analysis of the differentially expressed proteins. The introduc-tion of label free quantification techniques will lead to an increaseduse of shotgun proteomics. Moreover, the high throughput ob-tained by shotgun proteomics will lead to an increased use of pro-teomics in general. Advantages and disadvantages of bothapproaches are summarized in Table 1.

Proteomics investigations revealed promising results upon theclassification of hepatotoxic compounds, and showed the opportu-nities for the identification of protein biomarkers underlying thisclassification (Yamanaka et al., 2007). However, the detection ofidiosyncratic hepatotoxicants with the currently available in vitromethods will remain challenging, since these reactions are unpre-dictable and mostly immune-mediated. For non-idiosyncratichepatotoxicants, proteomics can indeed be used to gain insightinto the mechanistic processes of drug-induced hepatotoxicity.Fractionation of the total cellular proteome into the subproteomes

http://cebs.niehs.nih.gov

Table 2Summary of hepatotoxic proteome studies.

Species Compounds Cells/organelles Technique Observations Ref.

In vivoMouse Acetaminophen Total liver 2-DE Identification of known NAPQI targets Fountoulakis et al.

(2000)Mouse 14C-Bromobenzene Total liver 2-DE Identification of bromobenzene targets, e.g.

glutathione S-transferases, protein disulfideisomerases and liver fatty acid-bindingprotein

Koen et al. (2007)

Rat Thioacetamide Total liver 2-DE Down-regulation of enzymes from fatty acidb-oxidation, branched chain amino acids andmethionine breakdown

Low et al. (2004)

Rat Hydrazine Total liver DIGE Differential expression of proteins from lipidmetabolism, calcium homeostasis, thyroidhormone pathways and stress response

Kleno et al. (2004)

Rat Troglitazone Total liver DIGE Differential expression of proteins from fattyacid metabolism, PPARa/RXR activation,oxidative stress and cholesterol biosynthesis

Boitier et al.(2011)

Rat Acetaminophen amiodarone tetracycline Total liver 2-DE Proteins related to oxidative stress responseand energy metabolism

Yamamoto et al.(2006)

Rat 63 Chemical compounds non-genotoxic Total liver DIGE 79.3% genotoxic compounds classified76.5% non-genotoxic compounds classified

Yamanaka et al.(2007)

Rat CDA Total liver DIGE Differential expression of proteins from fattyacid b-oxidation and of secreted proteins

Meneses-Lorenteet al. (2004)

Rat AcetaminophenAmiodaroneTetracyclineCCl4

Total liver 2-DE Detection of proteins related to oxidativestress and mitochondrial metabolism-regulation

Kikkawa et al.(2006)

Mouse Pentobarbital3-Methylcholanthrene

ER fraction from liver 2-DE Differential expression of proteins frombiotransformation pathways, cytoskeletonand general metabolism

Zgoda et al. (2006)

Rat CCl4 ER fraction from liver LCTQ-FTICR

Down-regulation of CYP 2C11, 3A2 and 2E1Up-regulation of CYP 2C6, 2B2 and 2B1

Jia et al. (2007)

Rat Ethanol Mitochondria fractionfrom liver

BN-PAGE Differential expression of proteins from fattyacid b-oxidation, oxidative phosphorylation,mitochondrial chaperones and amino acidmetabolism

Venkatraman et al.(2004)

Rat Z24 Plasma 2-DE Differential expression of proteins frombiotransformation, apoptosis, carbohydrate,lipid amino acid and energy metabolism

Wang et al. (2010)

In vitroRat Acetaminophen

AmiodaroneTetracycline

Primary hepatocytes 2-DE Differential expression of proteins from celldeath, lipid a carbohydrate metabolism

Yamamoto et al.(2005)

Rat AcetaminophenAmiodaroneTetracyclineCCl4

Primary hepatocytes 2-DE Differential expression of proteins whichrelate to oxidative stress and metabolism-regulation

Kikkawa et al.(2005)

Rat Aflatoxin B1

CadmiumN-methyl-N0-nitro-N-nitrosoguanidineMethyl methane sulfonateCCl4

N,N-DimethylformamideVinyl acetateAcetaminophen

Primary hepatocytes LC–MS/MS

Maintenance of CYP1A, 2B, 3A2, partial loss ofCYP2E1

Farkas andTannenbaum(2005)

Rat CDA Primary hepatocytes DIGE Differential expression of secreted proteins,ER-proteins, mitochondrial chaperones, anti-oxidant proteins and proteins for fatty acidsynthesis

Meneses-Lorenteet al. (2006)

Human Bezafibrate Primary hepatocytes 2D-LC/Maldi-TOF

BEZA treatment modulated lipid and fattyacid metabolism/transport and cellular stress

Alvergnas et al.(2011)

Humanand rat

1-Naphthyl-isothiocyanate, indomethacine,acetaminophen, cisplatin, tetracycline,dimethyl-nitrosamine aspirin, isoproterenol,phenylephrine, DMSO

HepG2 and total liver 2-DE Thirteen potential toxicity marker proteinswere found in rat liver from which 8 werealso detected in HepG2These proteins are related to amino acidbiosynthesis, fatty acid metabolism andxenobiotic metabolic processes.

Thome-Kromeret al. (2003)

Human AcetaminophenAmiodaroneCyclosporin A

HepG2 DIGE Differential expression of secreted proteinsand ER-Golgi transport proteins

van Summerenet al. (2011)

Human Ethanol Secretome of HepG2/C3A

LC–MS Differential expression of proteins fromapoptosis, inflammation and cell leakage

Lewis et al. (2010)

Human Di(2-ethylhexyl)phthalate Secretome of HepG2 2-DE Differential expression of proteins from cellstructure, apoptosis and tumor progression

Choi et al. (2010)


Table 2 (continued)

Species Compounds Cells/organelles Technique Observations Ref.

Human Troglitazone, ciglitazone, farglitazar, ritonavirDMSO indinavir, rosiglitazone, tesaglitazar

Secretome of humanimmortalizedhepatocytes, withCYP3A4 over expression

LC–MS/MS

Detection of two possible biomarkersreferred to as BMS-PTX-265 and BMS-PTX-837

Gao et al. (2004)

Rat Aflatoxin B1 Secretome of primaryhepatocytes

LC–MS/MS

Decreased expression of a2-macroglobulinand a1-antitrypsin

Farkas et al. (2005)


of organelles and the secretome provide a more detailed view.Most affected pathways in the reported hepatotoxic proteomestudies are carbohydrate, lipid and energy metabolism, stress re-sponse and apoptosis (Table 2). In addition calcium homeostasisand membrane trafficking appear to be affected too. Drug-inducedinhibition of carbohydrate, lipid and energy metabolism pathwayscan be linked to the development of steatosis. Fat accumulation inhepatocytes or steatosis occurs after impaired oxidation of fattyacids to acetyl-CoA through the b-oxidation reaction in the mito-chondria. Cholestasis, characterized by bile accumulation was seentogether with an inhibition of the calcium homeostasis and varioustransport mechanisms, e.g. membrane trafficking (Erlinger, 1997;Reichen et al., 1985). The differentially expressed proteins of stressresponse and apoptotic pathways are clearly a result of cellular in-jury caused by hepatotoxicants.

With proteome maps of the several cell models it is possible toidentify the characteristics of, and differences between thesein vitro models and with this their usefulness in toxicity studiescan be assessed. The in vitro cell models discussed in this paperhave shown to have a good prospective for toxicity testing in thefuture; probably the secretome of these cell models will becomea special matter of interest. The secretome of these hepatic cellmodels contains a rich source of potential biomarkers which canbe easily measured in the medium and can be made applicablefor testing in plasma of patients during clinical trials and drugtherapy.

Despite these promising results of the toxicoproteomics ap-proach, the development of a panel of biomarkers will requirethe testing of several well-characterized model hepatotoxicants.The differential protein expression induced by particular com-pounds can be combined together in a database as is already donefor transcriptomics data by several initiatives (Iconix, Gene Logicand Toxicogenomics Project Japan) (Kienhuis et al., 2011). By test-ing classified compounds, common patterns of toxicity can be dis-tinguished from compound-specific mechanisms. The value ofproteome data can be increased by comparison with data fromcomplementary transcriptomics and metabolomics experimentsin a systems biology approach.

Conflict of interest statement

The authors have declared no conflict of interest.

Acknowledgements

This work was supported by the Netherlands Genomics Initia-tive/Netherlands Organization for Scientific Research (NWO),050-060-510.

References

Aitman, T.J., Critser, J.K., Cuppen, E., Dominiczak, A., Fernandez-Suarez, X.M., Flint, J.,Gauguier, D., Geurts, A.M., Gould, M., Harris, P.C., Holmdahl, R., Hubner, N.,Izsvak, Z., Jacob, H.J., Kuramoto, T., Kwitek, A.E., Marrone, A., Mashimo, T.,Moreno, C., Mullins, J., Mullins, L., Olsson, T., Pravenec, M., Riley, L., Saar, K.,Serikawa, T., Shull, J.D., Szpirer, C., Twigger, S.N., Voigt, B., Worley, K., 2008.Progress and prospects in rat genetics: a community view. Nat. Genet. 40, 516–522.

Alvergnas, M., Rouleau, A., Lucchi, G., Heyd, B., Ducoroy, P., Richert, L., Martin, H.,2011. Proteomic mapping of bezafibrate-treated human hepatocytes in primaryculture using two-dimensional liquid chromatography. Toxicol. Lett. 201, 123–129.

Amacher, D.E., 2010. The discovery and development of proteomic safetybiomarkers for the detection of drug-induced liver toxicity. Toxicol. Appl.Pharmacol. 245, 134–142.

Bailey, S.M., Andringa, K.K., Landar, A., Darley-Usmar, V.M., 2008. Proteomicapproaches to identify and characterize alterations to the mitochondrialproteome in alcoholic liver disease. Methods Mol. Biol. 447, 369–380.

Barrier, M., Mirkes, P.E., 2005. Proteomics in developmental toxicology. Reprod.Toxicol. 19, 291–304.

Boitier, E., Amberg, A., Barbie, V., Blichenberg, A., Brandenburg, A., Gmuender, H.,Gruhler, A., McCarthy, D., Meyer, K., Riefke, B., Raschke, M., Schoonen, W.,Sieber, M., Suter, L., Thomas, C.E., Sajot, N., 2011. A comparative integratedtranscript analysis and functional characterization of differential mechanismsfor induction of liver hypertrophy in the rat. Toxicol. Appl. Pharmacol. 252, 85–96.

Brunet, S., Thibault, P., Gagnon, E., Kearney, P., Bergeron, J.J., Desjardins, M., 2003.Organelle proteomics: looking at less to see more. Trends Cell Biol. 13, 629–638.

Choi, S., Park, S.Y., Jeong, J., Cho, E., Phark, S., Lee, M., Kwak, D., Lim, J.Y., Jung, W.W.,Sul, D., 2010. Identification of toxicological biomarkers of di(2-ethylhexyl)phthalate in proteins secreted by HepG2 cells using proteomic analysis.Proteomics 10, 1831–1846.

Craig, A., Sidaway, J., Holmes, E., Orton, T., Jackson, D., Rowlinson, R., Nickson, J.,Tonge, R., Wilson, I., Nicholson, J., 2006. Systems toxicology: integratedgenomic, proteomic and metabonomic analysis of methapyrilene inducedhepatotoxicity in the rat. J. Proteome Res. 5, 1586–1601.

Donato, M.T., Lahoz, A., Castell, J.V., Gomez-Lechon, M.J., 2008. Cell lines: a tool forin vitro drug metabolism studies. Curr. Drug Metab. 9, 1–11.

Dunn, J.C., Tompkins, R.G., Yarmush, M.L., 1991. Long-term in vitro function of adulthepatocytes in a collagen sandwich configuration. Biotechnol. Prog. 7, 237–245.

Erlinger, S., 1997. Drug-induced cholestasis. J. Hepatol. 26 (Suppl. 1), 1–4.Farkas, D., Tannenbaum, S.R., 2005. Characterization of chemically induced

hepatotoxicity in collagen sandwiches of rat hepatocytes. Toxicol. Sci. 85,927–934.

Farkas, D., Bhat, V.B., Mandapati, S., Wishnok, J.S., Tannenbaum, S.R., 2005.Characterization of the secreted proteome of rat hepatocytes cultured incollagen sandwiches. Chem. Res. Toxicol. 18, 1132–1139.

Fielden, M.R., Brennan, R., Gollub, J., 2007. A gene expression biomarker providesearly prediction and mechanistic assessment of hepatic tumor induction bynongenotoxic chemicals. Toxicol. Sci. 99, 90–100.

Fountoulakis, M., Berndt, P., Boelsterli, U.A., Crameri, F., Winter, M., Albertini, S.,Suter, L., 2000. Two-dimensional database of mouse liver proteins: changes inhepatic protein levels following treatment with acetaminophen or its nontoxicregioisomer 3-acetamidophenol. Electrophoresis 21, 2148–2161.

Fujii, K., Kondo, T., Yokoo, H., Okano, T., Yamada, M., Yamada, T., Iwatsuki, K.,Hirohashi, S., 2006. Database of two-dimensional polyacrylamide gelelectrophoresis of proteins labeled with CyDye DIGE Fluor saturation dye.Proteomics 6, 1640–1653.

Galeva, N., Altermann, M., 2002. Comparison of one-dimensional and two-dimensional gel electrophoresis as a separation tool for proteomic analysis ofrat liver microsomes: cytochromes P450 and other membrane proteins.Proteomics 2, 713–722.

Gao, J., Ann Garulacan, L., Storm, S.M., Hefta, S.A., Opiteck, G.J., Lin, J.H., Moulin, F.,Dambach, D.M., 2004. Identification of in vitro protein biomarkers ofidiosyncratic liver toxicity. Toxicol. In Vitro 18, 533–541.

Gorg, A., Weiss, W., Dunn, M.J., 2004. Current two-dimensional electrophoresistechnology for proteomics. Proteomics 4, 3665–3685.

Greenhough, S., Medine, C.N., Hay, D.C., 2010. Pluripotent stem cell derivedhepatocyte like cells and their potential in toxicity screening. Toxicology 278,250–255.

Guyot, C., Lepreux, S., Combe, C., Doudnikoff, E., Bioulac-Sage, P., Balabaud, C.,Desmouliere, A., 2006. Hepatic fibrosis and cirrhosis: the (myo)fibroblastic cellsubpopulations involved. Int. J. Biochem. Cell Biol. 38, 135–151.

Hartung, T., 2009. Toxicology for the twenty-first century. Nature 460, 208–212.Heijne, W.H., Stierum, R.H., Slijper, M., van Bladeren, P.J., van Ommen, B., 2003.

Toxicogenomics of bromobenzene hepatotoxicity: a combined transcriptomicsand proteomics approach. Biochem. Pharmacol. 65, 857–875.

Hughes, C.S., Nuhn, A.A., Postovit, L.M., Lajoie, G.A., 2011. Proteomics of humanembryonic stem cells. Proteomics 11, 675–690.

Jennen, D.G., Gaj, S., Giesbertz, P.J., van Delft, J.H., Evelo, C.T., Kleinjans, J.C., 2010.Biotransformation pathway maps in WikiPathways enable direct visualization


of drug metabolism related expression changes. Drug Discov Today 15, 851–858.

Jia, N., Liu, X., Wen, J., Qian, L., Qian, X., Wu, Y., Fan, G., 2007. A proteomic method foranalysis of CYP450s protein expression changes in carbon tetrachloride inducedmale rat liver microsomes. Toxicology 237, 1–11.

Karp, N.A., Feret, R., Rubtsov, D.V., Lilley, K.S., 2008. Comparison of DIGE and post-stained gel electrophoresis with both traditional and SameSpots analysis forquantitative proteomics. Proteomics 8, 948–960.

Kienhuis, A.S., Wortelboer, H.M., Maas, W.J., van Herwijnen, M., Kleinjans, J.C., vanDelft, J.H., Stierum, R.H., 2007. A sandwich-cultured rat hepatocyte system withincreased metabolic competence evaluated by gene expression profiling.Toxicol. In Vitro 21, 892–901.

Kienhuis, A.S., van de Poll, M.C., Dejong, C.H., Gottschalk, R., van Herwijnen, M.,Boorsma, A., Kleinjans, J.C., Stierum, R.H., van Delft, J.H., 2009. Atoxicogenomics-based parallelogram approach to evaluate the relevance ofcoumarin-induced responses in primary human hepatocytes in vitro forhumans in vivo. Toxicol. In Vitro 23, 1163–1169.

Kienhuis, A.S., Bessems, J.G., Pennings, J.L., Driessen, M., Luijten, M., van Delft, J.H.,Peijnenburg, A.A., van der Ven, L.T., 2011. Application of toxicogenomics inhepatic systems toxicology for risk assessment: acetaminophen as a case study.Toxicol. Appl. Pharmacol. 250, 96–107.

Kikkawa, R., Yamamoto, T., Fukushima, T., Yamada, H., Horii, I., 2005. Investigationof a hepatotoxicity screening system in primary cell cultures –‘‘whatbiomarkers would need to be addressed to estimate toxicity in conventionaland new approaches?’’. J. Toxicol. Sci. 30, 61–72.

Kikkawa, R., Fujikawa, M., Yamamoto, T., Hamada, Y., Yamada, H., Horii, I., 2006. Invivo hepatotoxicity study of rats in comparison with in vitro hepatotoxicityscreening system. J. Toxicol. Sci. 31, 23–34.

Kleno, T.G., Leonardsen, L.R., Kjeldal, H.O., Laursen, S.M., Jensen, O.N., Baunsgaard,D., 2004. Mechanisms of hydrazine toxicity in rat liver investigated byproteomics and multivariate data analysis. Proteomics 4, 868–880.

Knowles, S.R., Uetrecht, J., Shear, N.H., 2000. Idiosyncratic drug reactions: thereactive metabolite syndromes. Lancet 356, 1587–1591.

Koen, Y.M., Gogichaeva, N.V., Alterman, M.A., Hanzlik, R.P., 2007. A proteomicanalysis of bromobenzene reactive metabolite targets in rat liver cytosol in vivo.Chem. Res. Toxicol. 20, 511–519.

Lavoie, C., Paiement, J., 2008. Topology of molecular machines of the endoplasmicreticulum: a compilation of proteomics and cytological data. Histochem. CellBiol. 129, 117–128.

Lewis, J.A., Dennis, W.E., Hadix, J., Jackson, D.A., 2010. Analysis of secreted proteinsas an in vitro model for discovery of liver toxicity markers. J. Proteome Res. 9,5794–5802.

Low, T.Y., Leow, C.K., Salto-Tellez, M., Chung, M.C., 2004. A proteomic analysis ofthioacetamide-induced hepatotoxicity and cirrhosis in rat livers. Proteomics 4,3960–3974.

Martinez, S.M., Bradford, B.U., Soldatow, V.Y., Kosyk, O., Sandot, A., Witek, R., Kaiser,R., Stewart, T., Amaral, K., Freeman, K., Black, C., LeCluyse, E.L., Ferguson, S.S.,Rusyn, I., 2010. Evaluation of an in vitro toxicogenetic mouse model forhepatotoxicity. Toxicol. Appl. Pharmacol. 249, 208–216.

Matheis, K.A., Com, E., Gautier, J.C., Guerreiro, N., Brandenburg, A., Gmuender, H.,Sposny, A., Hewitt, P., Amberg, A., Boernsen, O., Riefke, B., Hoffmann, D., Mally,A., Kalkuhl, A., Suter, L., Dieterle, F., Staedtler, F., 2011. Cross-study and cross-omics comparisons of three nephrotoxic compounds reveal mechanisticinsights and new candidate biomarkers. Toxicol. Appl. Pharmacol. 252, 112–122.

Mathijs, K., Kienhuis, A.S., Brauers, K.J., Jennen, D.G., Lahoz, A., Kleinjans, J.C., vanDelft, J.H., 2009. Assessing the metabolic competence of sandwich-culturedmouse primary hepatocytes. Drug Metab. Dispos. 37, 1305–1311.

McBurney, R.N., Hines, W.M., Von Tungeln, L.S., Schnackenberg, L.K., Beger, R.D.,Moland, C.L., Han, T., Fuscoe, J.C., Chang, C.W., Chen, J.J., Su, Z., Fan, X.H., Tong,W., Booth, S.A., Balasubramanian, R., Courchesne, P.L., Campbell, J.M., Graber, A.,Guo, Y., Juhasz, P.J., Li, T.Y., Lynch, M.D., Morel, N.M., Plasterer, T.N., Takach, E.J.,Zeng, C., Beland, F.A., 2009. The liver toxicity biomarker study: phase I designand preliminary results. Toxicol. Pathol. 37, 52–64.

Meneses-Lorente, G., Guest, P.C., Lawrence, J., Muniappa, N., Knowles, M.R., Skynner,H.A., Salim, K., Cristea, I., Mortishire-Smith, R., Gaskell, S.J., Watt, A., 2004. Aproteomic investigation of drug-induced steatosis in rat liver. Chem. Res.Toxicol. 17, 605–612.

Meneses-Lorente, G., Watt, A., Salim, K., Gaskell, S.J., Muniappa, N., Lawrence, J.,Guest, P.C., 2006. Identification of early proteomic markers for hepatic steatosis.Chem. Res. Toxicol. 19, 986–998.

Meng, Z., Veenstra, T.D., 2011. Targeted mass spectrometry approaches for proteinbiomarker verification. J Proteomics.

Merrick, B.A., Bruno, M.E., Madenspacher, J.H., Wetmore, B.A., Foley, J., Pieper, R.,Zhao, M., Makusky, A.J., McGrath, A.M., Zhou, J.X., Taylor, J., Tomer, K.B., 2006.Alterations in the rat serum proteome during liver injury from acetaminophenexposure. J. Pharmacol. Exp. Ther. 318, 792–802.

Neilson, K.A., Ali, N.A., Muralidharan, S., Mirzaei, M., Mariani, M., Assadourian, G.,Lee, A., van Sluyter, S.C., Haynes, P.A., 2011. Less label, more free: approaches inlabel-free quantitative mass spectrometry. Proteomics 11, 535–553.

Nie, A.Y., McMillian, M., Parker, J.B., Leone, A., Bryant, S., Yieh, L., Bittner, A., Nelson,J., Carmen, A., Wan, J., Lord, P.G., 2006. Predictive toxicogenomics approachesreveal underlying molecular mechanisms of nongenotoxic carcinogenicity. Mol.Carcinog. 45, 914–933.

Nisar, S., Lane, C.S., Wilderspin, A.F., Welham, K.J., Griffiths, W.J., Patterson, L.H.,2004. A proteomic approach to the identification of cytochrome P450 isoforms

in male and female rat liver by nanoscale liquid chromatography-electrosprayionization–tandem mass spectrometry. Drug Metab. Dispos. 32, 382–386.

Olson, H., Betton, G., Stritar, J., Robinson, D., 1998. The predictivity of the toxicity ofpharmaceuticals in humans from animal data – an interim assessment. Toxicol.Lett. 102–103, 535–538.

Pan, C., Kumar, C., Bohl, S., Klingmueller, U., Mann, M., 2009. Comparative proteomicphenotyping of cell lines and primary cells to assess preservation of cell type-specific functions. Mol. Cell. Proteomics 8, 443–450.

Pichard, L., Raulet, E., Fabre, G., Ferrini, J.B., Ourlin, J.C., Maurel, P., 2006. Humanhepatocyte culture. Methods Mol. Biol. 320, 283–293.

Reichen, J., Berr, F., Le, M., Warren, G.H., 1985. Characterization of calciumdeprivation-induced cholestasis in the perfused rat liver. Am. J. Physiol. 249,G48–57.

Ruepp, S.U., Tonge, R.P., Shaw, J., Wallis, N., Pognan, F., 2002. Genomics andproteomics analysis of acetaminophen toxicity in mouse liver. Toxicol. Sci. 65,135–150.

Rumack, B.H., 2002. Acetaminophen hepatotoxicity: the first 35 years. J. Toxicol.Clin. Toxicol. 40, 3–20.

Sakamoto, A., Matsumaru, T., Ishiguro, N., Schaefer, O., Ohtsuki, S., Inoue, T.,Kawakami, H., Terasaki, T., 2011. Reliability and robustness of simultaneousabsolute quantification of drug transporters, cytochrome P450 enzymes, andudp-glucuronosyltranferases in human liver tissue by multiplexed MRM/selected reaction monitoring mode tandem mass spectrometry with nano-liquid chromatography. J. Pharm. Sci. 100, 4037–4043.

Seglen, P.O., 1976. Preparation of isolated rat liver cells. Methods Cell Biol. 13, 29–83.

Seliskar, M., Rozman, D., 2007. Mammalian cytochromes P450-importance of tissuespecificity. Biochim. Biophys. Acta 1770, 458–466.

Slany, A., Haudek, V.J., Zwickl, H., Gundacker, N.C., Grusch, M., Weiss, T.S., Seir, K.,Rodgarkia-Dara, C., Hellerbrand, C., Gerner, C., 2010. Cell characterization byproteome profiling applied to primary hepatocytes and hepatocyte cell linesHep-G2 and Hep-3B. J. Proteome Res. 9, 6–21.

Suter, L., Schroeder, S., Meyer, K., Gautier, J.C., Amberg, A., Wendt, M., Gmuender, H.,Mally, A., Boitier, E., Ellinger-Ziegelbauer, H., Matheis, K., Pfannkuch, F., 2010.EU Framework 6 Project: Predictive Toxicology (PredTox)-overview andoutcome. Toxicol. Appl. Pharmacol. 252, 73–84.

Takahashi, K., Yamanaka, S., 2006. Induction of pluripotent stem cells from mouseembryonic and adult fibroblast cultures by defined factors. Cell 126, 663–676.

Thome-Kromer, B., Bonk, I., Klatt, M., Nebrich, G., Taufmann, M., Bryant, S., Wacker,U., Kopke, A., 2003. Toward the identification of liver toxicity markers: aproteome study in human cell culture and rats. Proteomics 3, 1835–1862.

Thomson, J.A., Itskovitz-Eldor, J., Shapiro, S.S., Waknitz, M.A., Swiergiel, J.J., Marshall,V.S., Jones, J.M., 1998. Embryonic stem cell lines derived from humanblastocysts. Science 282, 1145–1147.

Tolando, R., Zanovello, A., Ferrara, R., Iley, J.N., Manno, M., 2001. Inactivation of ratliver cytochrome P450 (P450) by N,N-dimethylformamide and N,N-dimethylacetamide. Toxicol. Lett. 124, 101–111.

Uetrecht, J., 2008. Idiosyncratic drug reactions: past, present, and future. Chem. Res.Toxicol. 21, 84–92.

Van Summeren, A., Renes, J., Bouwman, F.G., Noben, J.P., van Delft, J.H., Kleinjans,J.C., Mariman, E.C., 2011. Proteomics investigations of drug-inducedhepatotoxicity in HepG2 cells. Toxicol. Sci. 120, 109–122.

Venkatraman, A., Landar, A., Davis, A.J., Chamlee, L., Sanderson, T., Kim, H., Page, G.,Pompilius, M., Ballinger, S., Darley-Usmar, V., Bailey, S.M., 2004. Modification ofthe mitochondrial proteome in response to the stress of ethanol-dependenthepatotoxicity. J. Biol. Chem. 279, 22092–22101.

Wang, K., Shindoh, H., Inoue, T., Horii, I., 2002. Advantages of in vitro cytotoxicitytesting by using primary rat hepatocytes in comparison with established celllines. J. Toxicol. Sci. 27, 229–237.

Wang, P., Mariman, E., Keijer, J., Bouwman, F., Noben, J.P., Robben, J., Renes, J., 2004.Profiling of the secreted proteins during 3T3-L1 adipocyte differentiation leadsto the identification of novel adipokines. Cell. Mol. Life Sci. 61, 2405–2417.

Wang, Y., Yang, B., Wu, C., Zheng, Z., Yuan, Y., Hu, Z., Ma, H., Li, S., Liao, M., Wang, Q.,2010. Plasma and liver proteomic analysis of 3Z-3-[(1H-pyrrol-2-yl)-methylidene]-1-(1-piperidinylmethyl)-1,3–2H-indol-2-one-inducedhepatotoxicity in Wistar rats. Proteomics 10, 2927–2941.

Waters, M., Stasiewicz, S., Merrick, B.A., Tomer, K., Bushel, P., Paules, R., Stegman, N.,Nehls, G., Yost, K.J., Johnson, C.H., Gustafson, S.F., Xirasagar, S., Xiao, N., Huang,C.C., Boyer, P., Chan, D.D., Pan, Q., Gong, H., Taylor, J., Choi, D., Rashid, A., Ahmed,A., Howle, R., Selkirk, J., Tennant, R., Fostel, J., 2008. CEBS – Chemical Effects inBiological Systems: a public data repository integrating study design andtoxicity data with microarray and proteomics data. Nucleic Acids Res. 36,D892–900.

Weber, L.W., Boll, M., Stampfl, A., 2003. Hepatotoxicity and mechanism of action ofhaloalkanes: carbon tetrachloride as a toxicological model. Crit. Rev. Toxicol. 33,105–136.

Wetmore, B.A., Merrick, B.A., 2004. Toxicoproteomics: proteomics applied totoxicology and pathology. Toxicol. Pathol. 32, 619–642.

Wilm, M., 2009. Quantitative proteomics in biological research. Proteomics 9, 4590–4605.

Yamamoto, T., Kikkawa, R., Yamada, H., Horii, I., 2005. Identification of oxidativestress-related proteins for predictive screening of hepatotoxicity using aproteomic approach. J. Toxicol. Sci. 30, 213–227.

Yamamoto, T., Kikkawa, R., Yamada, H., Horii, I., 2006. Investigation of proteomicbiomarkers in in vivo hepatotoxicity study of rat liver: toxicity differentiation inhepatotoxicants. J. Toxicol. Sci. 31, 49–60.


Yamanaka, H., Yakabe, Y., Saito, K., Sekijima, M., Shirai, T., 2007. Quantitativeproteomic analysis of rat liver for carcinogenicity prediction in a 28-dayrepeated dose study. Proteomics 7, 781–795.

Zgoda, V., Tikhonova, O., Viglinskaya, A., Serebriakova, M., Lisitsa, A., Archakov, A.,2006. Proteomic profiles of induced hepatotoxicity at the subcellular level.Proteomics 6, 4662–4670.

Zwickl, H., Traxler, E., Staettner, S., Parzefall, W., Grasl-Kraupp, B., Karner, J., Schulte-Hermann, R., Gerner, C., 2005. A novel technique to specifically analyze thesecretome of cells and tissues. Electrophoresis 26, 2779–2785.

proteomics in the search for mechanisms and biomarkers of drug-induced hepatotoxicity

Documents