Phenobarbital Induces Alterations in the Proteome of Hepatocytes and Mesenchymal Cells of Rat Livers

Preceding studies on the mode of action of non-genotoxic hepatocarcinogens (NGCs) have concentrated on alterations induced in hepatocytes (HCs). A potential role of non-parenchymal liver cells (NPCs) in NGC-driven hepatocarcinogenesis has been largely neglected so far. The aim of this study is to characterize NGC-induced alterations in the proteome profiles of HCs as well as NPCs. We chose the prototypic NGC phenobarbital (PB) which was applied to male rats for a period of 14 days. The livers of PB-treated rats were perfused by collagenase and the cell suspensions obtained were subjected to density gradient centrifugation to separate HCs from NPCs. In addition, HCs and NPC isolated from untreated animals were treated with PB in vitro. Proteome profiling was done by CHIP-HPLC and ion trap mass spectrometry. Proteome analyses of the in vivo experiments showed many of the PB effects previously described in HCs by other methods, e.g. induction of phase I and phase II drug metabolising enzymes. In NPCs proteins related to inflammation and immune regulation such as PAI-1 and S100-A10, ADP-ribosyl cyclase 1 and to cell migration such as kinesin-1 heavy chain, myosin regulatory light chain RLC-A and dihydropyrimidinase-related protein 1 were found to be induced, indicating major PB effects on these cells. Remarkably, in vitro treatment of HCs and NPCs with PB hardly reproduced the proteome alterations observed in vivo, indicating differences of NGC induced responses of cells at culture conditions compared to the intact organism. To conclude, the present study clearly demonstrated that PB induces proteome alterations not only in HCs but also in NPCs. Thus, any profound molecular understanding on the mode of action of NGCs has to consider effects on cells of the hepatic mesenchyme.


Introduction
Screening assays, which enable the early detection of potential carcinogenic activities, are of crucial importance for safe drug development strategies. Chemical compounds may cause cancer by directly affecting DNA and are thus called genotoxic carcinogens. This type of compounds is easily detectable as such by the application of well-established in vitro assays, such as Ames bacterial reverse mutation assay, mammalian forward mutation assays and detection of chromosomal aberrations. Furthermore, in vivo assays are routinely used which include rodent erythrocyte micronucleus assay, mammalian bone marrow chromosomal aberration assay and assays for somatic cell gene mutation in endogenous genes. Chemical carcinogens which do not affect DNA directly are called non-genotoxic carcinogens (NGCs) [1]. In contrast to genotoxic carcinogens, there are no sufficiently accurate and validated short-term assays that may allow detection of NGCs [2][3][4][5][6]. Currently employed assays necessitate long-term rodent carcinogenicity assays causing high efforts, costs and time requirement as major drawbacks. In order to overcome these problems, deeper insights into NGC-relevant mechanisms are urgently required, which may be obtained by the application of a screening technology such as proteome profiling.
According to current knowledge, a characteristic effect of many NGCs is a deviation of tissue homeostasis resulting in organ growth based on a dysbalance between cell replication and cell death by apoptosis [7,8]. This dysbalance acts also on mutated/initiated cells. By this mechanism NGCs enhance the selective proliferation of preneoplastic cells and exert tumour promoting effects. Possible molecular mechanisms of the tumour promoting effects of NGCs comprise epigenetic changes such as hypo-and hypermethylation of CpG sites, chromatin modifications, and miRNA regulated mechanisms [9], but also endocrine effects, inhibition of gap junctional intercellular communications, immune modulation, and/or profound disturbances in the epithelial-mesenchymal interactions [6]. It is known that during carcinogenesis the microenvironment may gain a pivotal role supporting preneoplastic and neoplastic cell growth via an altered vasculature, deviated immunological activities and altered interstitial extracellular matrix (ECM) [10,11]. Further possible modes of NGC actions in rodents may be cytotoxicity followed by regenerative growth and a pro-inflammatory status. Involvement of inflammatory mechanisms may be accompanied by enhanced production of reactive oxygen species (ROS) and reactive nitrogen species (RNS) [7,8]. These two species may have relevance in carcinogenesis via signalling function and possibly cause endogenous DNA damage, which may be responsible for a weak genotoxic potential of NGCs.
The most common target organ of NGCs in rodent models is the liver, i.e., about 40% of all NGC tested so far are hepatocarcinogens. Hitherto, research on the action of NGCs has been focusing mainly on hepatocytes (HCs), the major parenchymal cells of the liver which eventually give rise to liver cancer. A role of non-parenchymal liver cells (NPCs) in NGCdriven hepatocarcinogenesis has been mostly neglected, because these cells do not transform [12,13]. However, NPCs, which consists mainly of Kupffer, endothelial, and stellate cells, may also be targeted by NGCs and may contribute considerably to the selective proliferation of preneoplastic and neoplastic HCs via release of paracrine growth factors or other growth-enhancing stimuli. Here, as a model NGC we chose phenobarbital (PB), a barbiturate known to be a potent tumour promoter in rodent liver [14][15][16][17][18]. PB has been described to interact with the pregnane X receptor (PXR) and the constitutive androstane receptor (CAR) triggering a signal transduction cascade leading to an induction of cytochrome P450 genes such as members of the CYP2B and CYP3A subfamily [9,[19][20][21][22][23]. Furthermore, it was shown that PB acts through other mechanisms such as oxidative stress [24], which may correlate with the P450 induction [25], driving tumour promotion by inducing proliferation [26,27] in HCs, nonreceptor mediated endocrine modifications and inhibition of gap junction intercellular communications, regulating growth and differentiation [18].
Proteome profiling is a powerful technique to observe molecular consequences of drug action. Cells may respond to drug actions via the synthesis of new proteins. Cells synthesize proteins in order to overcome biological challenges. Therefore, the identification of drug-induced proteins may give important hints to better understand the way of action of drugs. Furthermore, if the drug-induced proteins display restricted expression patterns, they may be used as indicative marker proteins. For comprehensive investigation, we analysed subcellular fractions including the secretome of isolated HCs and NPCs separately applying LC-MS/MS analysis. We hypothesized that NGC-driven hepatocarcinogenesis may involve NPCs and specifically considered a potential contribution of inflammatory activities of this tissue compartment, as known for the pathogenesis of hepatocellular carcinoma in humans [28]. Therefore we treated HCs and NPCs with pro-inflammatory cytokines in vitro in order to identify proteome signatures characteristic for such events. Induction of such a signature by PB treatment could thus indicate the involvement of inflammatory processes in drug action. Furthermore, we investigated PB effects induced by in vivo treatment of animals in comparison to PB effects observed upon in vitro treatment of isolated primary cells, including both HCs and NPCs. This strategy provided the unique opportunity to differentiate direct drug effects on the isolated cell types in vitro from indirect drug effects modulated by complex epithelialmesenchymal interactions in the intact organism. This approach may thus give novel insights into the mode of action of this prototypic NGC. To conclude, the aim of the present study was to investigate proteome alterations in HCs as well as NPCs, which are caused by the non-genotoxic carcinogen phenobarbital (PB) in vitro as well as in vivo.

Animals and treatment
Male Wistar rats were obtained from and kept at the ''Division for Decentralized Biomedical Facilities of the Medical University of Vienna" under standardized SPF-conditions. Phenobarbital (PB, 5-Ethyl-5-phenyl-barbituric acid-sodium salt; Fluka, 04712) [29,30], admixed to drinking water, was administered to three rats which were 8 to 10 weeks old at a daily dose of 50mg/kg bodyweight. Two times a week the PB concentrations in the drinking water were adjusted to the body weights and the amount of water consumed. Five controls and four animals for the in vitro experiments were treated with tap water only. Animals were sacrificed by exsanguination under CO 2 -asphyxation during the liver perfusion. The experiments were approved by the ''Committee of Animal Protection of the Austrian Ministry of Sciences" (permission number 66009/157 II10b/2009) and performed according to Austrian regulations in accordance with the criteria outlined in the ''Guide for the Care and Use of Laboratory Animals" by the National Academy of Sciences.

Separation of liver cells and primary cultures
The following procedure was used for the in vitro and the in vivo experiments. To isolate and cultivate hepatocytes (HCs) and non-parenchymal cells (NPCs) at a functional state as naïve as possible, livers of rats were perfused with collagenase (Worthington CLS-2) as described before [31,32]. In the resulting cell suspension HCs were isolated from NPCs by low speed centrifugation followed by discontinuous density gradient centrifugation using Percoll [33]. The purity was found to be an average of 95.4% for hepatocyte fractions and 99.8% for NPC fractions. Cell preparations were used for further experimentation when the viability exceeded 90%, as determined by the trypan blue exclusion assay. HCs and NPCs were subsequently seeded on collagen-coated 6-well plates. HCs were seeded at a density of 4x10 5 cells/well in Williams' medium E (Invitrogen) supplemented with glutamax, HEPES, gentamycin, H2 mix, ascorbat and 10% FCS. NPCs were seeded at a density of 3-4x10 6 cells/well in RPMI 1640 medium (Gibco Ltd.) supplemented with gentamycin and 10% FCS at 37°C for 2 hours. To determine the purity of the isolated cell fractions, cells were counted using microscope pictures and the ImageJ software (National Institutes of Health). After an attachment period of 2 hours cells were switched to serum-free medium (Williams medium E and RPMI 1640 medium, both supplemented as above without the 10% FCS) and kept at 37°C for further 24 hours in order to collect cell supernatants.

In vitro treatment of primary HCs and NPCs
In addition to the procedure described above, the in vitro treatment of cultures commenced 2h after plating of cells deriving from four untreated rats. Two rats were used for the cytokine treatment, while the other two were used for the PB treatment. HCs were treated with 10ng/ml interleukin-1ß (R&D Systems) and 5ng/ml interleukin-6 (R&D Systems) [34] for 24 hours, which induce the acute phase plasma protein synthesis in HCs [35][36][37][38]. 10ng/milliliter lipopolysaccharide (LPS, Sigma-Aldrich) was applied to the medium of NPCs for 24 hours [39][40][41][42][43]. HCs and NPCs were treated with 1mM PB (Fluka) for 24 hours.

Cellular sub-fractionation and protein sample processing (see also 44)
The serum-free supernatants were sterile filtered using a 0.2µm cellulose acetate filter (Whatman). One part of this filtrate was precipitated by adding 5x volume of -20°C tempered p.a. ethanol (Merck) and subsequent storing at -20°C for at least overnight. The other part was directly stored at -80°C for subsequent analyses by ELISA. During all steps samples were kept on ice. For harvesting of cytoplasmic and nucleic protein fraction, cells were gathered in isotonic buffer (10mM HEPES/NaOH pH=7.4, 10mM NaCl, 3.5mM MgCl2, 1mM EGTA, 0.25M Sucrose and 0.5% Triton X-100) and protease inhibitor mix (1mM PMSF; aprotinin, leupeptin and pepstatin, [1µg/ml] each). The cells were disrupted by sheer stress caused by syringing the cell lysates through 23G needles. The cytoplasmic fraction, in the supernatant, was separated from nuclei and membrane proteins as well as debris by centrifugation at 2300xg and 4°C for 5min and was subsequently precipitated by ethanol tempered to -20°C. The nuclei protein fraction was extracted from the remaining residue via a 10min incubation with an extraction buffer (10mM Tris/HCl pH=7.4, 1mM EDTA, 500mM NaCl) which act through osmotic pressure followed by a 1:10 dilution with a NP40 buffer (10mM Tris/HCl pH=7.4, 1mM EDTA, 0.5% NP40), to reduce the final NaCl concentration, for 15min. Nucleic proteins were separated from debris by centrifugation at 2328xg and 4°C for 5min and precipitated by ethanol tempered to -20°C.
After precipitation all fractions were centrifuged at 4700xg and 4°C for 25min. The resulting protein pellets were dissolved in sample buffer (7.5M urea, 1.5M thiourea, 4% CHAPS, 0.05% SDS, 100mM DTT) in a volume according to the protein amount / pellet size.

Shotgun analysis -1D-SDS PAGE, in-gel tryptic digestion & MS analysis ('bottom up')
For 1D-SDS PAGE we used a 4% stacking and a 12% resolving polyacrylamide gel. Protein samples were loaded onto the gel and the electrophoresis ran until complete separation of a pre-stained molecular marker (Dual Color, Biorad, Hercules, CA) was visible. Gels were fixated with 50% methanol/10% acetic acid and MS compatible silver stained as described by Mortz, E. et al [45]. For tryptic digestion samples were cut into lanes to group proteins with a similar molecular weight, which improved the following LC-MS/MS analyses. These lanes were cubed and the proteins were destained, reduced and alkylated before digestion with trypsin (sequencing grad, Roche) at 37°C over night as described before [46]. After elution the peptide solutions were analysed by LC -MS/MS.
For reversed-phase chromatography we used a nano-flow LC (1100 Series LC system, Agilent) combined with the HPLC chip technology (Agilent). The chips consist of a 40nl Zorbax 300SB-C18 trapping column and a Zorbax 300SB-C18 (75µm x 150mm) separation column. The flow rate was 400nl/min, using a gradient from 0.2% formic acid and 3% ACN to 0.2% formic acid and 50% ACN over 40min or 60min for supernatants or the other two fractions, respectively.
Peptide identification was performed by MS/MS fragmentation analysis using an ion-trap mass spectrometer (XCT-Ultra, Agilent) combined with already described HPLC chips, which are eventually an orthogonal nanospray ion source (Agilent). For peak list-generation and spectrum identification, of the MS/MS data we used Spectrum Mill MS Proteomics Workbench software (Version A.03.03, Agilent). We searched our MS/MS data against the UniProtKB/Swiss-Prot protein database (Version 10 th August 2010; 519,348 entries). The settings were as follows: max. of 2 missed cleavages, minimum scored peak intensity (%SPI) of 70%, precursor mass tolerance of +/-1.5Da, product mass tolerance +/-0.7. Two types of modification were considered, both arising during sample preparation. Carbamidomethylation of cysteine (deliberately) was set as fixed modification and oxidized methionine (artefact) was set as post-translational modification.

Generating protein maps and their interpretation
The resulting peptide-protein assignment list was reviewed by hand considering the following parameters. Spectrum Mill peptide sequence score values (sequence matching probability) of >13 counted as valid sequence assignments. Peptide sequences were valid, when at least one peptide sequence scored as much and did not occur in other proteins of this cell type. An estimated error rate was calculated by searching the sequences against a non-sense reversed database. Furthermore we included peptides scoring between 9 and 13 only if precursor m/z value, retention time and MS2 pattern were found similarly in at least one of our previous experiments and the peptide was thereby scoring above 13. With respect to protein inference, we chose the smallest number of proteins required to explain all observed peptides as described for ProteinProphet [47]. As our protein identification algorithm includes manual selection, we cannot calculate an exact false discovery rate.
Proteome analysis using emPAI (exponentially modified protein abundance index) values [48], comparisons and PB specific alterations were recorded and interpreted using the GPDE (Griss Proteomics Database Engine), a database specifically engineered for the identification and characterization of marker proteins [49].

ELISA
Arginase-1 protein levels were determined by a kit (Uscn Life Science Inc.; Houston, TX 77036) according to the user manual.

Isolation and proteome profiling of primary rat hepatocytes and non-parenchymal cells
Primary rat cells were obtained from untreated rats and rats treated with PB in vivo. Cell isolation was performed by liver perfusion followed by separation into hepatocytes (HCs) and non-parenchymal cells (NPCs) (see Figure 1). Cells were cultured for 24 hours to allow accumulation of secreted proteins. Subsequently, cells were fractionated into cytoplasm and nuclear extract. The protein fractions were further separated by SDS-PAGE and forwarded to proteome profiling via LC-MS/MS as described.
We restricted our analyses to proteins which were identified by two peptides or more in two or more independent experiments. By this approach, we identified 1148 proteins in HCs and 1213 proteins in NPCs with an overlap of 966 proteins in both cell compartments. The 182 proteins being apparently specific for HCs indeed comprise a large number of known liver-specific proteins including, apolipoproteins and other serum proteins, cytochrome P450 isoenzymes, sulfotransferases, UDP-glucuronosyltransferases and many others (table 1-5, table S1 & S2). Furthermore, the 247 proteins identified in NPCs included many proteins known to be expressed in stromal cells. Representative for leukocytes are the surface antigens CD37, CD47, CD96 and CD166 as well as the chemokines CXCL1, CXCL2, and CCL6. Amongst known marker proteins for endothelial cells we identified endothelial cell-specific molecule 1, endothelial nitric oxide synthase and septin-2 (vascular endothelial cell specific protein 11), while MMP-3 and various collagens are characteristic for the stellate cells.

In vitro treatment of isolated primary cells
HCs were treated with IL-6 and NPCs with LPS, respectively. Upon treatment for 24 hours, the induction of several proinflammatory proteins was observed which confirms the responsiveness of these cells to stress stimuli under our experimental conditions ( Figure 2).
To investigate direct drug effects, primary untreated cells were treated with 1mM PB for 24 hours and analysed by proteome profiling. Remarkably, this treatment had hardly any measureable effect on the proteome composition of HC and NPCs. A single protein was apparently induced in HC and two proteins only in NPCs, respectively. Proteins indicating stress such as chaperones including the heat shock protein family, stress-induced phosphoprotein, heme oxygenase 1 and DNAJ homolog subfamily members were not or hardly induced. 13 proteins were found to be newly induced in the cytoplasmic fraction of HCs by PB, including phase I drug metabolizing enzymes such as amine oxidase, phase II enzymes such as UDP-glucuronosyltransferases and glutathione S-transferases, the chaperone rotamase D, glutathione synthetases and the proto-oncogen c-Raf. The induction of estradiol 17-beta-dehydrogenase 8 may indicate alterations in steroid hormone homeostasis. The induction of CYP2B, probably via CAR, is a known positive control for PB action. Table 1, 2 and 3 depict the most significant proteins found up-regulated in the secretome, cytoplasm or nuclear extract of HCs and NPCs in response to in vivo PB exposure. Proteins found up-regulated in the secretome of HCs are involved in the acute-phase response, inflammation response and the action of drugs. In the cytoplasm of HCs we found 82 proteins upregulated. These proteins preferentially act in the acute-phase response, cell growth, immune response, inflammatory response as well as in response to cell stress, oxidative stress and wounding. In the nuclear extract of HCs we found 182 proteins up-regulated. These proteins are involved in nucleosome assembly, nucleocytoplasmic transport, mRNA processing, translation, protein localization to the nucleus, RNA processing, removal of superoxide radicals, anti-gen processing and presenting, protein methylation and cell redox  homeostasis as well as are major nucleolar proteins of growing eukaryotic cells and proteins of the nuclear matrix. In NPCs, 12 proteins were found newly induced, comprising proteins related to inflammation such as PAI-1 and S100-A10, cell migration such as kinesin-1 heavy chain, myosin regulatory light chain RLC-A and dihydropyrimidinase-related protein 1 as well as altered immune cell functional state such as acetyl-CoA carboxylase 1 and ADP-ribosyl cyclase 1. Proteins found upregulated in the secretome of NPCs are involved in the acutephase response, inflammation response, and action of drugs and oxidative species. In the cytoplasm of NPCs we found 108 proteins up-regulated. These proteins exert function in tissue, acute phase, cell redox homeostasis, cell cycle, response to drug, cell adhesion, vesicle-mediated protein transport, stress response, positive regulation of MAPKKK cascade, cellular response to cell-matrix adhesion, response to ROS. In the nuclear extract of NPCs we found 78 proteins up-regulated which act on transcription regulation, mRNA processing, mRNA and protein transport, response to DNA damage stimulus and DNA repair. Table 4 and 5 depict selected proteins found only in untreated rats, which means a down-regulation of these proteins in HCs and NPCs in response to in vivo PB exposure. This rather stringent selection criterion for down-regulated proteins improves the reliability of these results. In HCs, 6 proteins were found only in untreated rats, including the E3 ubiquitin-protein ligase UBR4, a protein involved in in membrane morphogenesis and cytoskeletal organization as well as the glutamine synthetase, which catalyses the production of glutamine and 4-aminobutanoate (gammaaminobutyric acid, GABA). In NPCs, 28 proteins were found only in untreated rats, comprising proteins involved in cell adhesion, cell proliferation, liver development and protein transport.

In vivo treatment of animals with phenobarbital
Table S1 and S2 present the summarised proteome profiling results obtained with HCs (S1) and NPCs (S2) derived from PB treated rats (in vivo).

Pathway analysis of the in vivo data via Reactome
The alterations in the proteome were further analysed by Reactome, a public peer-reviewed pathway database from CSHL, OICR and EBI, which is cross-referenced to bioinformatics databases and allows the assignment of a given protein to one or more molecular pathways. When searching for the positively PB-induced proteome alterations, in both HCs and NPCs, proteins assigned to the categories of "gene expression", "metabolism of proteins" and "3`UTR-mediated translational regulation" were most prominent among the top listed events (see table 6 & 7). With respect to molecular pathways, in HCs "Peroxisomal lipid metabolism", "Class I MHC mediated antigen processing & presentation" and "Asparagine N-linked glycosylation" were listed on top (see table 8). In contrast, in NPCs "Protein folding", "Dissolution of Fibrin Clot" and "Platelet Adhesion to exposed collagen" were listed on top (see table 9).

ELISA verification of arginase-1 variations
To verify selected LC-MS/MS data in a quantitative fashion, we conducted an ELISA for rat arginase-1. Arginase-1 was chosen, because of its ability to diminish anti-tumour immunity by interfering with the activation of T-cells [50]. LC-MS/MS results indicated a PB induced decrease in protein secretion of arginase-1 by HCs in case of in vivo treated rats. The ELISA results confirmed our LC-MS/MS results as demonstrated in Figure S1. In the secretome arginase-1 concentration was found decreased by a factor of 2.3. In the cytoplasmic protein fraction, arginase-1 abundance was found induced by PB more than three-fold, which corresponds very well to the LC-MS/MS results.

Discussion
The aim of this study was to investigate molecular mechanisms induced by treatment of rats with the nongenotoxic carcinogen phenobarbital (PB) by means of proteome profiling. This approach was used to pinpoint crucial events not previously recognised by other technical approaches. Current mechanistic considerations on nongenotoxic carcinogenesis include altered cell-cell interactions, epigenetic changes endocrine effects, inhibition of gap junctional intercellular communications and immune modulation [6], which may be of crucial importance for initiation as well as promotion and progression [10]. However, the significance of an altered epithelial-mesenchymal dialogue and the role of NPCs for NGC-driven hepatocarcinogenesis have not been investigated so far. Teufehlofer et al has described that chemical compounds, including genotoxic hepatocarcinogens, may induce the superoxide radical production by Kupffer cells and may thus contribute to DNA damage and an increased occurrence of Table 3. Selected proteins found up-regulated in the nuclear extract protein fraction of HCs and NPCs isolated from rat livers, when treated with PB in vivo.  [12,25]. Furthermore, stromal liver cells may be induced to secrete survival factors, which may act as tumour promoters [52,53]. Therefore, an improved understanding of nongenotoxic compounds has to consider drug effects on stroma cells, even if these cells are not transformed to cancer cells. In order to assign molecular events caused by NGCs to the different cell types of the liver, we isolated primary cells by liver perfusion and separated them into parenchymal HCs and NPCs as described previously [54,55]. Here, we isolated cells from untreated and PB-treated animals in order to investigate in vivo effects. Cells obtained from untreated livers were also treated with PB in vitro. This experimental approach is limited mainly due to the de-differentiation of primary cells during prolonged in vitro cultivation. To avoid this issue, we chose a 24 hours treatment period to ensure a meaningful data interpretation. In the in vitro part, functional activation by interleukin-6 or LPS resulted in well detectable proteome alterations in HCs or NPCs (data not shown). Any cytotoxic substance will cause a cell stress response, which results in an increased expression of chaperones, especially the heat shock protein family. Under the present experimental conditions PB application in vitro hardly induced such a stress response and generally exerted marginal effects. A reason for that may be that the dedifferentiation of HCs in culture during the first 24 hours hampers the reaction of HCs to PB, namely the induction of drug metabolising enzymes. Nevertheless, proteins induced in vitro were also found to be induced in the in vivo experiments, some of which are presented in Figure 2A. Figure 3 highlights the diverse response of HCs and NPCs upon in vitro (A) and in vivo (B) PB treatment. This observation suggests that in vivo PB treatment has effects being more profound and largely different from those obtained by in vitro treatment. This may be due to the fact that the milieu in the intact organism is required to enable the full response of liver cells to NGCs and that the disrupted cell-cell interactions in vitro and the tendency towards de-differentiation under artificial culture conditions compromise such a response.

Hepatocytes / nuclear extract protein fraction
The presently observed PB-induced proteome alterations in the stromal cells of the liver indeed suggest profound functional alterations. There was strong induction of pleckstrin, which is a positive regulator of platelets causing an aggregation of these sensitive cells, indicating that PB may trigger a wound healing cascade. ADP-ribosyl cyclase 1 (CD38) is an immunity-related protein induced in NPCs upon PB action. This protein usually regulates B cell function by influencing the intra-cellular Ca 2+ concentration [56]. It was also demonstrated that it affects the migration capabilities of dendritic cells and as a consequence it also affects T cells [57].
The increased secretion of PAI-1, cathepsin L1, MMP-10 and V-CAM 1 indicates some inflammatory activation of these cells. Remarkably, however, a profound inflammatory activation as observed upon LPS-treatment was not evident (Figure 2). Cox-1 (Q63921), an important mediator of inflammatory signalling, was observed induced by PB in stroma cells in one animal only, which may be due to low protein concentration at the limit of detection. The expression of the inflammationrelated protein MX-1 [58] in the control stroma cells, however, indicates that the presently employed cell manipulation steps may also have altered the inflammatory activity state in an artificial way, rendering clear conclusions, with respect to inflammatory pathways, difficult. The present observations still indicate some complex regulatory actions of PB which are related to but still distinct from classical inflammatory activation warranting further investigation.
Many effects of PB on HCs, as observed in our in vivo experiments, have already been identified applying techniques other than proteomics, e.g. the induction of phase-I and phase-II drug-metabolising enzymes [59], redox-regulating enzymes Table 4. Selected proteins found down-regulated in HCs isolated from rat livers, when treated with PB in vivo.

Accession numberProtein name GO -biological process
Secreted protein fraction as well as the proto-oncogen c-RAF were observed by proteome profiling (table 1-3) [60][61][62]. As part of the drug metabolism, we found Cytochrome P450 2B1 and other isoforms, sulfotransferase 1A1 and glutathione S-transferase alpha-2 (GST-A2) up-regulated in HCs. Increased glutathione S-transferase levels may indicate increased oxidative stress [63]. It has been proposed that an accumulation of ROS has many effects on cells such as an increased proliferation, DNA mutation rates [64] and genetic instability [65]. Reproducing these and other well-known PB effects by our currently employed proteome profiling analyses supports the notion that this strategy was valid. Furthermore, we were able to observe a few molecular events which have not yet been described. This includes the induction of estradiol 17-beta-dehydrogenase 8 in HCs (table 2) and the down-regulation of estrogen sulfotransferase 1 in NPCs (table 5) by PB, which may indicate alterations of the estrogen metabolism. Estrogens are important growth factors and potential tumour promoters [66]   and may exert anti-inflammatory activities [67]. Whether PB may act via estrogen activity modulation deserves more detailed investigations, e.g. by including female rats in the PB study. Furthermore, ectonucleotide pyrophosphatase 3 (E-NPP 3) was found to be induced, which has been observed in association with neoplastic bile duct diseases [68].
The nuclear proteins presently observed to be induced by PB actually indicate that cells were finally exposed to DNA stress. This interpretation is supported by the PB-induced expression of DNA topoisomerase I, poly [ADP-ribose] polymerase 1 and sister chromatid cohesion protein PDS5 homolog B. These proteins are known to be involved in DNA damage sensing as well as DNA-repair. These findings were somewhat unexpected, as PB is known to be non-genotoxic. However, it may increase the hepatocellular production of ROS by Cyp450 induction or ROS production by activated NPCs and may thus evoke some marginal DNA repair activity. This demonstrates the potential difficulties to group chemical compounds unequivocally to the categories of genotoxic or non-genotoxic carcinogen.
Interestingly it was demonstrated that inflammation-related ROS formation and signalling may lead to carcinogenesis of Furthermore, under these conditions MMP activity may correlate with invasiveness as determined by Matrigel invasion assays [69]. An PB-induced formation of ROS is evidenced by the observed increase of ROS responding enzymes such as retinal dehydrogenase 1 and mitochondrial superoxide dismutase [Mn] in HCs and plasminogen activator inhibitor 1 (PAI-1) in NPCs. as well as by the induction of DNA repair proteins described above.
The present investigation of primary cells by proteome profiling may be considered as essential for the achievement of biological relevance of the employed model, but on the other hand, also accounts for the experimental limitation of the present study with respect to reliable quantification of potential marker proteins. Shotgun proteomics as presently employed using nano liquid chromatography and ion trap mass spectrometry may give a quite comprehensive overview to cell activities, but results in rather semi-quantitative data hardly accessible to stringent statistical analysis. However, the present observation of PB-induced proteome alterations especially in NPCs is a first but important step to improve our  understanding of the complex mode of action of non-genotoxic carcinogens. It is evident from the current results that the analysis of drug effects on isolated cell model systems will hardly represent the mode of action in vivo. Furthermore, the contribution of tumourpromoting molecules associated with ROS to the effect of PB is clearly evidenced by the induction of several marker proteins signifying oxidative stress and associated DNA damage. Only the complete comprehension of direct and indirect consequences will enable a strategy to identify a panel of marker molecules with sufficient specificity for the unequivocal indication of non-genotoxic carcinogen activity.   Supporting Information Figure S1. ELISA verification of arginase-1 variations. This figure depicts the arginase-1 variations in 1) secretome and 2) cytoplasm of HCs using the quantitative data from the ELISA and 3) the semi-quantitative data from the LC-MS/MS analyses (description of how to read this sort of presentation see figure  2), which are altered accordingly. Arginase-1 concentration decreases in the secretome and increases upon PB treatment of rats. Part 4) presents the values used to generate these figure, whereby the emPAI values were used for part 3).
(TIF) Table S1. Summary of proteome profiling results obtained with HCs derived from PB treated rats (in vivo). Column names, which are labelled 'analysis' and 'reference', refer to the experiments of the isolated primary cells, HCs and NPCs, deriving from rats treated in vivo with PB (analysis) and untreated animals (reference). "found", specificity of protein identification; accession, Uniprot accession numbers; name, protein names; "analysis_fractions", subcellular fractions in which the protein was identified in untreated samples; "analysis_peptides", number of distinct peptides identified per proteins in untreated samples; "analysis_nuclei_expcount", number of positive identifications in nuclear fractions compared to the total number of experiments; "analysis_nuclei_empai", calculated emPAI values in the nuclear fraction; "analysis_nuclei_empai_stdev", standard deviation thereof. These terms are used in the same way referring to the secretome as well as the cytoplasm. After the values obtained for the treated samples (analysis) all values are listed again referring to untreated samples (reference). The last three columns list difference values for the emPAI values obtained for nuclei, secretomes and cytoplasms, respectively. (XLSX )   Table S2. Summary of proteome profiling results obtained with NPCs derived from PB treated rats (in vivo). Column names, which are labelled 'analysis' and 'reference', refer to the experiments of the isolated primary cells, HCs and NPCs, deriving from rats treated in vivo with PB (analysis) and untreated animals (reference). "found", specificity of protein identification; accession, Uniprot accession numbers; name, protein names; "analysis_fractions", subcellular fractions in which the protein was identified in untreated samples; "analysis_peptides", number of distinct peptides identified per proteins in untreated samples; "analysis_nuclei_expcount", number of positive identifications in nuclear fractions compared to the total number of experiments; "analysis_nuclei_empai", calculated emPAI values in the nuclear fraction; "analysis_nuclei_empai_stdev", standard deviation thereof. These terms are used in the same way referring to the secretome as well as the cytoplasm. After the values obtained for the treated samples (analysis) all values are listed again referring to untreated samples (reference). The last three columns list difference values for the emPAI values obtained for nuclei, secretomes and cytoplasms, respectively. (XLSX)