Automated detection of hepatotoxic compounds in human hepatocytes using HepaRG cells and image-based analysis of mitochondrial dysfunction with JC-1 dye

Abstract

In this study, our goal was to develop an efficient in situ test adapted to screen hepatotoxicity of various chemicals, a process which remains challenging during the early phase of drug development. The test was based on functional human hepatocytes using the HepaRG cell line, and automation of quantitative fluorescence microscopy coupled with automated imaging analysis. Differentiated HepaRG cells express most of the specific liver functions at levels close to those found in primary human hepatocytes, including detoxifying enzymes and drug transporters. A triparametric analysis was first used to evaluate hepatocyte purity and differentiation status, mainly detoxication capacity of cells before toxicity testing. We demonstrated that culturing HepaRG cells at high density maintained high hepatocyte purity and differentiation level. Moreover, evidence was found that isolating hepatocytes from 2-week-old confluent cultures limited variations associated with an ageing process occurring over time in confluent cells. Then, we designed a toxicity test based on detection of early mitochondrial depolarisation associated with permeability transition (MPT) pore opening, using JC-1 as a metachromatic fluorescent dye. Maximal dye dimerization that would have been strongly hampered by efficient efflux due to the active, multidrug-resistant (MDR) pump was overcome by coupling JC-1 with the MDR inhibitor verapamil. Specificity of this test was demonstrated and its usefulness appeared directly dependent on conditions supporting hepatic cell competence. This new hepatotoxicity test adapted to automated, image-based detection should be useful to evaluate the early MPT event common to cell apoptosis and necrosis and simultaneously to detect involvement of the multidrug resistant pump with target drugs in a human hepatocyte environment.

Introduction

Prediction of hepatotoxicity and characterization of underlying mechanisms require investigations during the leading phase of optimization and the preclinical development of new drugs and chemicals. Evidence from studies over the past decade strongly suggests that many mechanisms contribute to drug-induced liver injury. This underlines the need for metabolically competent cellular models and the development of high content screening (HCS) strategies adapted to multiparametric analyses for relevant in vitro cytotoxicity testing (Abraham et al., 2008; Gomez-Lechon et al., 2008; O’Brien et al., 2006).

Human primary hepatocytes are considered as the most pertinent in vitro model. Unfortunately, scarce availability and inter-donor variabil- ity greatly hamper their use for routine comparative studies (Guguen- Guillouzo and Guillouzo, 2010). Immortalized human hepatic cell lines, such as HepG2 and Huh-7, have been proposed as an alternative to primary hepatocytes (Noor et al., 2009). However, these lines lack several liver specific functions which limit their use in toxicology (Wilkening et al., 2003). The new HepaRG hepatoma cell line overcomes these limitations. These cells express most of the liver specific genes at higher levels than the other cell lines, especially detoxifying enzymes as well as sinusoidal and canalicular drug transporters such as MDR1 and MRP3 (Aninat et al., 2006; Le Vee et al., 2006). Their functional stability has been demonstrated based on the maintenance of drug metabolizing enzyme activities and the response to prototypical inducers (Josse et al., 2008). However, their possible evolution and alteration from ageing due to culture time has not been yet evaluated. In addition, unique progenitor features carried out by HepaRG cells have been described (Cerec et al., 2007) but never considered as a possible cause of alterations in the differentiation status during culture time. Indeed, HepaRG cells exhibit bipotent properties leading to the formation of a mixed population composed of hepatocytes and biliary-like cells. Since both cell types share the capacity to reverse into hepatic progenitors when confluence is lost, it is important to take in account this phenomenon during toxicity testing.
In the present study our goals were i) to establish a highly reproducible model which maintains hepatocyte differentiation status and reproduces stable adult human primary hepatocyte monolayers, and ii) to develop a mutiparametric test establishing the functional capacities of hepatocyte monolayers to respond to xenobiotic toxic effects. This test was based on quantitative fluorescence microscopy coupled with automated imaging analysis. One of the major causes of hepatotoxicity involves the cytochrome P450 monooxygenase-dependent (CYPs) formation of reactive me- tabolites. Various CYPs are expressed in differentiated HepaRG cells (Antherieu et al., 2010), including CYP3A4 which plays a critical role in metabolism of about 50% of human drugs (Guengerich, 1997). Since CYP3A4 is expressed at a high level in differentiated HepaRG hepatocytes, it represents a useful parameter for functional qualifica- tion of HepaRG cells in hepatotoxicity testing. Coupling this liver- specific marker to morphological features, such as cytoskeleton F-actin deposition beneath the plasma membrane (Corlu et al., 1991), could be helpful to establish hepatocyte qualification of cell monolayers while numbering bile canaliculus structures might provide a cell polarity signature.

Besides cytotoxicity tests monitoring irreversible and late cell damage before death, more mechanistically-defined analyses are increasingly used. Mitochondria are now recognized as a main target of toxic xenobiotics, and mitochondrial permeability transition (MPT) is widely believed to be an early event common to necrosis and apoptosis (Kim et al., 2003). Indeed, permeability transition (PT) pore opening, which initiates MPT, leads to mitochondrial depolarisation, uncoupling of oxidative phosphorylation, ATP depletion and cell death (Bernardi et al., 1999; Karbowski and Youle, 2003). In situ analysis of mitochondrial dysfunction is mainly performed by using various mitotracker dyes (Bernardi et al., 2001; Reungpatthanaphong et al., 2003). JC-1 is a lipophilic, cationic fluorescent dye that can enter and accumulate into mitochondria of healthy cells as a dimeric fluorescent red form (Greer et al., 2009). Upon the onset of MPT, the membrane mitochondrial potential (Δψm) collapses and the JC-1 dye releases to cytoplasm in a monomeric form which fluoresces green.

A most important problem with the vast majority of fluorescent probes for ΔΨm is that their cellular accumulation can be drastically reduced due to efficient efflux by the multidrug-resistant (MDR) pump (Marques-Santos et al., 2003; Saengkhae et al., 2003) or following binding to CYP enzymes (Stresser et al., 2000). This led us to pay attention to the JC-1 dye charge since HepaRG hepatocytes are rich in MDR and CYP proteins (Antherieu et al., 2010).

This study was undertaken to assess an in vitro hepatotoxicity test based on pure human HepaRG hepatocyte monolayers which were functionally qualified by a triparametric analysis well adapted to automated imaging, and to evaluate MPT breakdown as an early event leading to cell death in these CYP and transporter-competent hepatocytes along with the accuracy of using the JC-1 mitotracker after overcoming its efflux through hepatic transporters.

Materials and methods

Chemicals

Staurosporin, aflatoxin B1, dimethylsulfoxide (DMSO), 5-bromo-4- chloro-3-indolyl β-D-galactoside (X-Gal), verapamil, ketoconazole, carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone (FCCP) and 3-[4,5-diméthylthiazol-2yl]-2,5-diphényltétrazolium bromide (MTT) were purchased from Sigma; 4-nitroquinoline N-oxide from Thermo Fisher Scientific; 5-bromo-2-deoxyuridine (BrdU) from Amersham and 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolo carbocyanine io- dide (JC-1 dye) from Invitrogen.

Cell cultures

The HepaRG cell line was established from a hepato-cholangiocar- cinoma tumor in a female patient suffering from hepatitis C virus infection (Gripon et al., 2002). Cells were cultured in a basal Williams’ E medium (Invitrogen) supplemented with 10% FCS (Perbio, France), 100 units/ml penicillin, 100 μg/ml streptomycin (Invitrogen, Carlsbad, CA), 5 μg/ml insulin (Sigma), 2 mM glutamine, and 50 μM hydrocor- tisone hemisuccinate (Upjohn Pharmacia, France) as previously described. The cells were passaged every 2 weeks at a density of 2.7 × 104 cells/cm2. Two weeks after seeding, hepatocyte differentia- tion was optimized by adding 2% DMSO to the basal medium for 2, 4 or 7 more weeks. Hepatocytes were selectively harvested from differen- tiated HepaRG cultures by a short time exposure (b 5 min) to 0.05% trypsin (Invitrogen) half diluted (0.025%) in Dulbecco’s Phosphate Buffered Saline (D-PBS) with Ca++ and Mg++ at room temperature and under microscope examination control. Trypsin neutralization was performed with the basal medium and discarded by centrifugation at 400g for 2 min. Isolated HepaRG hepatocytes were re-suspended in basal medium with or without DMSO, at densities ranging from 1.8 to 21 × 104 cells/cm2. Growth stimulation was performed by epithelial growth factor (EGF) (Peprotech, France) addition to the basal medium at cell seeding, at the concentration of 50 ng/ml.

Immunochemical staining analyses

Cell adhesion of isolated HepaRG hepatocytes was measured by counting of nuclei at different time points after seeding. Cells were washed with D-PBS, fixed with 90% ethanol/5% acetic acid during 20 min at 4 °C and then, nuclei were labelled by Hoechst dye 5 ng/ml (Promega) during 10 min. After a 24 h BrdU-incubation (6 μg/ml), DNA synthesis was measured using a Cell Proliferation kit (Amersham, Orsay, France). For immunolocalization of cytochrome CYP3A4 and F-actin, cells were fixed at 4 °C with 4% paraformaldehyde for 20 min and permeabilized for 30 min with 0.1% saponin in PBS containing 5% donkey serum. Cells were washed with PBS and incubated for 90 min with CYP 3A4 primary antibody (AB1254, Chemicon International) diluted at 1/800 in PBS containing 0.1% saponin and then, incubated for 45 min with FITC-labelled secondary antibody (Jackson ImmunoRe- search) diluted at 1/400 in 0.1% saponin. CYP3A4 expression was quantified using an arbitrary unit of fluorescent mass representing the combination of intensity and area adjusted to cell number calculated by nuclei counting (Hoechst). F-actin was labelled by phalloidin fluoprobe SR101 (200 U/ml) (Interchim) diluted at 1/100 for 20 min. Counting of canaliculus structures was performed by using focal phalloidin accumulation at the bile canaliculus poles and particular threshold calibration adapted to select canaliculus formations as objects.

Cytotoxicity measurements

JC-1 test. Mitochondrial membrane depolarization was detected by using JC-1 dye at the final concentration of 7 μM corresponding to the maximum solubility value. Two days after seeding, pure HepaRG hepatocytes were incubated with apoptotic inducers at increasing concentrations in the basal medium for 3, 10 or 24 h as indicated. Then, medium containing JC-1 dye was renewed and the first 30 min incubation was performed in a CO2 incubator at 37 °C, followed by a second charge up to 1 h. JC-1 dye specificity as mitochondrial ΔΨm- associated dimerization was assessed by using the uncoupling agent FCCP added to the culture 15 min before the two- charge JC-1 protocol. Verapamil was used to inhibit MDR transporter activity. Increased concentrations (25–100 μM) were assayed 15 min before one charge of JC-1 for 30 min. Ketoconazole, a CYP3A4 and to a lesser extent an MDR inhibitor (Huang et al., 2007), like verapamil was added 15 min before one charge of JC-1 for 30 min, but at concentrations ranging from 5 to 20 μM. JC-1 test results were expressed as the ratio of altered versus living cells, which was calculated from the ratio of green / red mass fluorescent cells detected at 514 nm and 585 nm excitation. IC50 values were estimated based on a graphic method.

AnnexinV test. Externalization of phosphatidylserine (PS) was mea- sured by annexin V coupled with Alexa® 488 (Vybrant apoptosis kit #2, Invitrogen). After treatment with various toxicants, cells were incubated with 1 μl of annexin V-Alexa®488 red, 0.05 μg of Hoechst blue for 15 min at 37 °C in a final volume of 30 μl of medium per well of 96 well-plates (MW). Apoptosis quantification was expressed as green mass per number of cells (Hoechst).

MTT assay. After treatment, cells were stained with 0.5 mg/ml MTT solution for 2 h, then formazan was dissolved in 100% DMSO and measured using optical density (540 nm).

Senescence β-galactosidase assay

Our senescence assay was designed to histochemically detect the pH6- dependent-β-galactosidase activity present only in senescent cells and not found in pre-senescent, quiescent or immortalized cells (Dimri et al., 1995). As a control, the lysosomal-pH4-dependent-β-galactosidase activity present in all cells was detected. Cells were fixed with 2% paraformaldehyde for 5 min, washed and incubated at 37 °C (without CO2) for 6 h with the fresh senescence-associated β-gal (SA-β-gal) stain solution containing 1 mg/ml of X-Gal in 40 mM citric acid/sodium phosphate buffer adjusted at pH 6 or pH 4 and added with 5 mM K4[Fe (CN)6], 5 mM K3[Fe(CN)6], 150 mM NaCl, 2 mM MgCl2. Results were expressed as arbitrary units of coloration mass representing the combination of intensity and area values.

Evaluation of CYP450 activities

A standard activity assay was employed, based on metabolite formation following incubation with a cocktail of five substrates at final concentrations: 200 μM phenacetin (CYP1A), 100 μM bupropion (CYP2B6), 50 μM midazolam (CYP3A4), 100 μM tolbutamide (CYP2C9) and 100 μM dextromethorphan (CYP2D6) in MEM medium during 2 h. Next, cells and medium half diluted with acetonitrile were separately kept at −80 °C until analysis by LCMS-MS. Enzyme activities were expressed as nanomoles of metabolites formed/h/mg total protein.

q PCR and sequence primers

Total RNA was obtained by trizol extraction with TRIzol® Reagent (Invitrogen). RNA quantity and purity were assessed with a Nanodrop ND-1000 spectrophotometer (Nyxor Biotech, Paris, France). One microgram of total RNA was reversed-transcribed into cDNA using the High-Capacity cDNA Archive kit (Applied Biosystems, Foster City, CA). RT-qPCR was performed by the fluorescent dye SYBR Green methodology using the SYBR Green PCR Master Mix and the ABI Prism 7000 (Applied Biosystems). Primer pairs for CYP3A4, CYP1A1, CYP2B6, CYP2C9, MDR1 and MRP2 are reported in Table 1. Amplifi- cation curves were read with the ABI Prism 7000 SDS Software using the comparative cycle threshold method and the relative quantifica- tion of the steady-state mRNA levels was normalized against 18S RNA.

Automated high content image acquisition and analysis

Images were taken with an upright microscope (AxioImager M1, Zeiss, France) equipped with LED based colibri illuminating system (Zeiss) and automated Märzhauser motorized stage allowing high content image capture with Axiovision 4.7 software (Zeiss, France). Images were a mosaic of four pictures per well of 96 MW. Image analyses were processed with the simple PCI software (C Imaging® Image Analysis System V 6.3, Compix, USA) allowing determination of the number and area of the different objects of interest and/or variations of fluorescence intensity. The same calibration parameters were applied for the batch of images obtained from one experiment. All experiments were performed in triplicate.

Statistical analysis

Data are presented as means±standard error of the means (S.E.M.). The Mann–Whitney test was used to test whether any condition was significantly different from control. p b 0.05 was interpreted as significant.

Results

Establishment of stable pure human hepatocyte cultures from the HepaRG cell line adapted to automated imaging analysis

Confluent differentiated HepaRG cells usually plated at 2.7 ×104 cells/ cm2 gave rise to undifferentiated progenitor cells which actively proliferated for 3–4 days (Fig. 1A a). They then reached confluence and committed early into either hepatocyte or biliary differentiation pathways (Fig. 1A b, c). After 2 weeks, a mixed population of hepatocyte-like cells surrounded by clear epithelial cells corresponding to biliary-like cells formed a confluent monolayer which could be stably maintained for several weeks in the presence of DMSO (Fig. 1A c). Hepatocyte colonies were recognized by occurrence of numerous bile canaliculus formations which are morphological structures characteristic of parenchymal cells (Fig. 1A d) (Corlu et al., 1991). In order to characterize the functional capacity of these hepatocytes, we developed a triparametric test using an automated imaging analysis. CYP3A4 expression was evaluated as a marker of the detoxication metabolism capacity. Frequency of local F-actin fibers accumulation beneath the plasma membrane at cellular poles delineating typical bile canaliculus structures was used to determine the stability of hepatocyte polarity. The third parameter aimed at quantifying the two previous markers by hepatocyte numbering using the nuclear Hoechst dye. By using 2 week-old confluent multilayered HepaRG cultures, we showed that this test was well suited to specifically detect and functionally characterize hepatocyte colonies from the mixed culture (Fig. 1B a).
However, it also evidenced multilayered hepatocyte organisation that was poorly adapted to imaging quantification.

Therefore, we decided to use selective hepatocyte isolation to yield pure hepatocyte cultures expressing the highest functional stability at the best imaging quality level. Selective isolation of hepatocyte colonies was reproducibly obtained using a short trypsin incubation of 2 week-old confluent differentiated HepaRG cells at room tempera- ture. Clumps of hepatocytes detached while most biliary-like cells remained attached to the plate (Fig. 1B b, c) so that hepatocyte suspensions reached a yield of 75–80% purity estimated by charac- teristic granular morphology of hepatocytes (Fig. 1B d). From homogeneous suspensions, hepatocytes were introduced into seeded at a density range of 5–60 × 103 cells per well in basal medium in 96 MW. Interestingly, the number of hepatocyte colonies with cuboidal shape at day 2 increased in parallel to increased cell density (Fig. 1C). In contrast, at low density these colonies failed to form, while widely dispersed hepatocytes and numerous elongated flat cells deprived of biliary poles rapidly appeared, corresponding to undiffer- entiated hepatic progenitors (Fig. 1C d). Functional capacity of pure hepatocytes was compared in low and high density culture conditions by counting CYP3A4-positive cells; this number dramatically de- creased within 2 days at the lowest densities while the maximum number of CYP3A4-positive cells was observed at the highest density values (Fig. 1C a, b). However, as shown in Fig. 1E, the mean value of CYP3A4 staining intensity was low even at the high density, although the enzyme remained inducible by DMSO and rifampicin (data not shown).

Therefore, we decided to combine DMSO effects with the cell density parameter. Hepatocyte suspensions were prepared in basal medium with or without 1% or 2% DMSO, and seeded at increasing cell density (Fig. 1D, E). Unexpectedly, DMSO did not prevent hepatocyte attachment at the 2 concentrations tested. In addition, regardless of the cell density conditions, DMSO increased the purity of the populations, which could yield up to 90% at day 1 post-seeding using the triparametric test. Cells kept their typical cuboidal and polarized morphology at day 2 (Fig. 1D). However, after 2 days, only hepatocytes seeded at densities equal to or higher than 30 × 103 cells per well expressed high CYP3A4 protein levels, and evidenced no or weak contamination by undifferentiated progenitors (Fig. 1D a, b, E). Nonetheless, a limitation in cell density was clearly observed (Fig. 1F). Indeed, while a good correlation was found between the number of CYP3A4-positive cells and the expected number of seeded cells up to a density of 40 × 103 cells, beyond this value a “plateau” was observed suggestive of accumulation of cells in multilayers which could not be analyzed with automated imaging technology. These results led us to select a cell density of approximately 40 × 103 cells per well in 96 MW, with addition of 2% DMSO at time of cell seeding for the following experiments. High magnification of cells in these conditions showed that almost all hepatocytes were CYP3A4 positive and polarised at day 2 (Fig. 1G). Enumeration of bile canaliculus structures by imaging analysis led to a mean of one structure per two hepatocytes in the colonies. In addition, CYP3A4, CYP1A1, CYP2D6, CYP2C9 and CYP2B6 expression and activity were evaluated in cells maintained in these new conditions. Fig. 2 showed no significant difference in CYP expression; a ratio of gene expression values from selective pure HepaRG hepatocytes and mixed cultures was close to 1 for all of them. Nor was any significant difference in their activity detected (Table 2, Fig. 2).

Influence of confluent HepaRG cell ageing on the hepatocyte monolayers response to a mitogen signal

We further questioned whether a long term differentiation period before hepatocyte isolation from confluent HepaRG cultures might influence establishment and functional properties of hepatocyte mono- layers. We used HepaRG cells maintained at confluence for 2, 4 and 7 weeks before hepatocyte selection and tested 2 biological properties:
i) Adhesion and spreading. We observed that hepatocytes purified from 2- and 4-week confluent HepaRG cultures attached more rapidly to plastic than those derived from 7 week-old cultures (Fig. 3A, C). Meanwhile, spreading, calculated as mean cell area of hepatocytes isolated from 2-week confluent HepaRG cultures at day 1, was 30% higher than mean values from the two oldest HepaRG cultures (Fig. 3B, D).
ii) Response to EGF mitotic signal. Fig. 3E showed that split hepatocytes were able to respond to EGF exposure regardless the length of differentiation period of confluent HepaRG cultures before cell isolation. However, the range reproducibility of these responses strongly varied from one culture to the other. Indeed, around 70% of cells from 2-week confluent HepaRG cultures entered S phase with high synchrony in 24 h post-seeding while less than 30% and 20% of the total number of hepatocytes were able to progress in cell cycle when coming from 4- and 7-week confluent-HepaRG- cultures respectively. In addition, the peak of responses was strongly delayed up to 48 h, with hepatocytes coming from the two oldest cultures.
We therefore, addressed the question of a correlation between these changes in hepatocyte functional responses and occurrence of ageing process in long term confluent HepaRG cells. We looked for increased pH6-dependent-β-galactosidase activity as one major ageing characteristic (Fig. 3F, H). As expected, this enzyme activity was near zero in 1 week culture, corresponding to cells just reaching confluence (Fig. 3G). An activity was clearly evidenced in 2 week- confluent HepaRG cultures which appeared 3-fold increased in 4 week-old confluent cells compared to the 2 week-old ones and even higher in 7 week-old cultures (4.5-fold). These results suggested an ageing process in long term confluent differentiated HepaRG cultures. They led us to preferentially choose 2 week-confluent-HepaRG cells for reproducible hepatocyte isolation and monolayer formation and for designing a toxicity test based on detection of mitochondrial dysfunction.

Monitoring the JC-1 dye loading in mitochondria of differentiated pure HepaRG hepatocytes in two charges

Alterations of mitochondrial dynamics are associated with MPT and ΔΨm changes. JC-1 dye was used as mitotracker to quantify MPT event. We determined the lowest dye/cell ratio and the appropriate probe concentrations to get maximal fluorescence variations for quantification. Hepatocyte monolayers at day 2 of culture were exposed to increasing concentrations of JC-1 for up to 3 h and compared to confluent hepatoma Huh-7 cells (Fig. 4A). Mitochondrial loading corresponding to JC-1 dimer formation was determined by occurrence of red fluorescence emission. A concentration of 2 μM was sufficient to get maximal mitochondrial loading in Huh-7 cells within 30 min of incubation as observed by the “plateau” of red fluorescent cells numbered by imaging analysis (Fig. 4A a–c). In contrast, JC-1 remained as monomers emitting green fluorescence in major HepaRG hepatocytes, even at increasing concentrations up to 8 μM and even with exposure times up to 180 min (Fig. 4A’ a–c). To circumvent this difficulty we assayed 2 successive loadings of dye at 7 μM, above which it became insoluble. The first charge was limited to 30 min while we performed a kinetic analysis to define the time of the second charge loading which would be needed for producing the red fluorescence associated with the multimeric form of the dye. An increased number of cells with yellow and red fluorescent mitochondria was observed with this second charge, reaching a “plateau” within 45 min of incu- bation (Fig. 4B a–c). We used this 2 charge protocol to demonstrate the specificity of JC-1 dye for mitochondrial ΔΨm-associated dimerization, using the classical uncoupling agent FCCP added to the culture 15 min before the start of the JC-1 test (Fig. 4C b). As expected, mitochondria exhibited a specific red fluorescence in control cultures (Fig. 4C a, c) whereas green fluorescence in FCCP-treated cells represented the baseline mitochondrial uncoupling (Fig. 4C b).

JC-1 metachromatic dye is a substrate for MDR transporters in HepaRG hepatocyte cultures

Next, we studied mechanism(s) responsible for the observation of hepatocyte monolayers being refractory to JC-1 dye loading. Our strategy was to seek a possible efflux of dye associated with high activity of transporters in differentiated HepaRG hepatocytes (Le Vee et al., 2006). We first confirmed the maintenance of MDR1 and MRP2 high expression levels in selective pure differentiated hepatocytes (Fig. 2A). We also observed that one charge of 7 μM of JC-1 for 30 min was sufficient to insert red fluorescent dimers into mitochondria in undifferentiated HepaRG progenitors (Fig. 4E). Then assessment of the contribution of MDR activity to slow down JC-1 loading in mature hepatocyte monolayers was performed by using the verapamil MDR blocker. Pre-incubation of cells with verapamil for 15 min before JC-1 addition permitted strong improvement of dye loading through inhibition of its MDR-mediated efflux out of cells (Fig. 4D a, b). Calculation of the green mass/red mass cell ratio in a dose-dependent assay of verapamil showed up to 3-fold increases of red cells with only one JC-1 charge in the presence of 50 μM verapamil (Fig. 4F). A second approach was also followed, taking into account a possible contribu- tion of CYP-dependent JC-1 metabolism. As a first-step, we used ketoconazole, a CYP3A4 and to a lesser extent MDR1 inhibitor, to
measure it. In a dose-dependent assay, accumulation of JC-1 both as monomers (green) and dimers (red) in mitochondria, appeared slightly increased in the presence of 10 μM ketoconazole (Fig. 3D a, c) so the green mass/red mass ratio remained unchanged compared to the control condition (Fig. 4F). However, dimer accumulation was weaker than in the presence of verapamil (Fig. 4D b, c). Together, these results suggested that MDR-dependent efflux of the mitotracker was the main mechanism responsible for differentiated hepatocyte refractoriness to JC-1 mitochondrial loading in monolayer cultures.

JC-1 metachromatic dye detects mitochondrial depolarization in altered HepaRG hepatocytes

The rationale of the approach was that the positively charged probe accumulated into the mitochondrial membranes of energized cells loaded with JC-1 in response to ΔΨm and displayed a bright red fluorescence, which was decreased by de-energization of altered mitochondria, therefore causing an increase of the green fluorescence that reflects MPT and intracellular monomers’ release. The effect of staurosporin, a reference apoptotic molecule, on JC-1 fluorescence was assayed. Two day-old hepatocyte monolayers were treated with increasing concentrations (0.3–2.5 μM) of staurosporin for 24 h (Fig. 5A). Then, cells were loaded by 2 charges of JC-1 as defined above. As expected, all cells from control cultures exhibited red fluorescence associated with functional mitochondria (Fig. 5A b). Treatment with staurosporin induced a disappearance of red fluorescence along with a diffusion of green fluorescence into the cytoplasm of damaged cells (Fig. 5A c). A dose-dependent effect was clearly observed showing IC50 values between 0.6 and 1 μM (Fig. 5A a). The robustness of the test was assessed by performing the same staurosporin dose-response assay in using verapamil treatment instead of a second charge of JC-1 dye. The IC50 value obtained was essentially identical, 0.5–1 μM (Fig. 5C). Finally, reliability of the JC-1 test in HepaRG hepatocyte monolayers was also assessed by comparing results to those obtained with Annexin V and MTT tests. Results of a dose-dependent effect of staurosporin showed a similar range of IC50 value (Fig. 5A, B, D). We therefore extended application of JC-1 test to two other toxic compounds, aflatoxin B1 and 4-nitroquinoline (Fig. 5E), chosen because of indirect and direct hepatic-dependent genotoxic mechanisms respectively. As expected, these two compounds were toxic to HepaRG hepatocytes. Particularly, 4-nitro-quinoline appeared highly toxic with evidence of mitochondrial alterations after 3 h exposure and at a concentration as low as 3.7 μM, while the pro-genotoxic aflatoxin B1 induced toxicity at a concentration of 50 μM following a brief, 3 h exposure. As expected, this toxicity was increased with a longer exposure time (10 h). Then, comparison of toxicity measured with the MTT test was performed using short -and long-term exposures. The JC-1 test appeared the most sensitive, being the only one able to detect toxicity as early as 3 h of exposure duration (Fig. 5F).

Discussion

Human hepatotoxicity testing is crucial. However, using conven- tional cytotoxicity assays it has not been sufficiently predictable because of its low concordance with either standard in vitro cytotoxicity screening assays (O’Brien et al., 2006) or regulatory animal studies. One reason could be that lethal events in late stages of toxicity are frequently chosen for testing while serious toxicities may not be lethal by themselves (Xu et al., 2004). Here, this work aims at developing and validating a robust and sensitive in vitro cell-based test to assess the human hepatotoxicity potential of drugs by using a new HCS technology and taking advantage of the features of the highly-differentiated HepaRG cell line. HSC was based on automated fluorescence microscopy and image analysis on cells in multi-well plate configuration, which facilitated measurement of multiple parameters, including liver-specific morphological and functional markers, and simultaneous detection of cytotoxic effects induced by different compounds at different doses. Efforts have long been undertaken to attain a high quality hepatic cell model for cytotoxicity testing. HepaRG cells are now widely used for studying drug kinetics, metabolic profile, induction/inhibition and prediction of drug–drug interactions (Aninat et al., 2006; Guillouzo and Guguen-Guillouzo, 2008). Indeed, differentiated HepaRG cell cultures are representative of mature human hepatocytes, mainly in the expression and activity levels of the major CYPs, various phase II enzymes, most transporters, and the key nuclear factors (CAR, PXR, PPARs). In addition, stability of their functional properties for several passages has been demonstrated, thus contrasting with the well known individual variability of human primary hepatocytes. However, these cells have preserved a unique progenitor bipotent property leading to the establishment of a mixed population (Cerec et al., 2007) that requires caution in their cultivation conditions. First, in order to get pure HepaRG hepatocyte cultures, a selection protocol was settled on, based on differential response of biliary/hepatocyte cells to proteases in absence of Ca2+, a strategy previously used for co-culture systems (Corlu et al., 1997). This hepatocyte selection provided the advantage of high hepatocyte purity (around 80%) mimicking isolated human primary hepatocytes and thereby allowing for i) comparative studies based on homogeneous and reproducible cell suspensions, critical mainly for transcriptomic and proteomic analyses; and ii) flat, single layers of parenchymal cells well adapted to imaging analyses by applying the cells at the right density.

We then demonstrated that cell density was critical to preserve the pure HepaRG hepatocyte functional stability, and we determined the ratio yielding the best functional capacity as evidenced by CYP3A4 quantification using imaging strategy. Differentiated hepatocytes are highly polarized epithelial cells with complex trafficking routes. A key event in the polarization of epithelial cells is the establishment of cell- cell contacts allowing formation of adherent junctions, which delineate membrane domains (Cereijido et al., 2008). These morphogenic characteristics deeply influence signalling pathways that promote specific functions of differentiated cells as well as those that negatively control proliferation and migration (Ishibe et al., 2006). High density promotes polarity and functional activity in pure HepaRG cell colonies in our conditions. In contrast, when cell-cell contacts are lost, morphogenic changes contribute to restore the sensitivity to cytokines and growth factors (Tanos and Rodriguez-Boulan, 2008). It is therefore, not surprising that at low density HepaRG hepatocytes rapidly underwent a phenotype reversion to progenitors associated with loss of specific functions and restoration of cell proliferation activity. In addition, combining DMSO and cell density effects added value for maintenance of high purity and functionality of mature HepaRG hepatocyte mono- layers. Purity resulted from DMSO toxicity specifically targeting biliary- like cells which are deprived of detoxication capacities. This toxicity occurred when confluence was incomplete or broken, both leading to a loss of interplay with hepatocytes. Purity also resulted from DMSO blockage of hepatocyte spreading and reversion to undifferentiated progenitors. Noteworthy, in contrast to normal human freshly isolated hepatocytes, DMSO preserved viability of mature HepaRG hepatocyte suspensions without delaying cell attachment and spreading. The property of DMSO in stabilizing adult hepatocyte cultures after attachment is well known (Isom et al., 1980; Gripon et al., 1993). Moreover, DMSO greatly favored cell polarity restoration and biliary poles functions while inhibiting proliferation. Its role as inducer of the detoxication function was also previously reported (Aninat et al., 2006; Dickins, 2004). This does not rule out the possibility of deleting it from the medium for drug testing as we did it in our experiments.

Finally, defining the suitable differentiation period of confluent HepaRG cultures before hepatocyte isolation and splitting appeared very important to assure highly reproducible hepatocyte monolayers. To our surprise, an ageing process was observed as early as 2 weeks of cell confluence as assessed by increased β-galactosidase activity and delayed response to mitogen signals. This process increased with confluence time, suggesting a link with increased number of highly mature hepatocytes unable to reverse to progenitors. Further analyses are needed to determine whether ageing is associated to accumula- tion of p21WAF1/CIP1 as described in HBG cells (Glaise et al., 1998) or a consequence of DNA damage which results in gradual telomerase activity loss (Kang et al., 1999). In any case, this parameter should be considered as a priority when getting reproducible cell functional capacities.
An improvement in reproducibility of the pure HepaRG hepatocyte model was achieved by use of the triparametric functional test associating hepatic cytoskeleton, polarity, and functional markers. This test based on imaging analysis may serve to functionally qualify the cells concurrent with cytotoxicity testing, a control which is crucial for data analyses and generally missing from protocols.

The last part of this work was focused on in situ hepatotoxicity testing using these highly functional pure HepaRG hepatocytes, and the alteration of mitochondrial transmembrane potential (MPT) was chosen for detecting early event in cell injury. Most techniques that have been developed were for the study of MPT and intramitochondrial [Ca+], these two parameters being considered as crucial for defining the role of mitochondria in cell death (Bernardi et al., 1999). The existence of a link between NAD(P)H, reactive oxygen species (ROS) and the PT pore was demonstrated in a variety of models of cell death and represents an early event in commitment to apoptosis and necrosis (Kroemer et al., 1998). Recent demonstration with HepaRG cells showed that MPT breakdown could occur before necrotic cell death and in absence of caspase activity or other evidence of apoptosis (McGill et al., 2011). Mitochondrial function in situ is essentially monitored by the use of cationic fluorescent probes that are accumulated in response to the mitochondrial transmembrane potential (mΔψ) in the mitochondrial matrix of living cells. However, JC-1 possesses the advantages of metachromatic living dyes specific for mitochondrial membranes and sensitive for mΔψ fluctuations; it allows measuring both polarisation (red) and depolarisation (green) (Smiley et al., 1991). Here, we applied this dye to different cell lines including HCT116 and Huh-7, and we observed great variations in its ability to accumulate in mitochondria and to form dimers (red), with particularly strong limitation in pure HepaRG hepatocytes. Several reports have shown a good correlation between cellular accumulation of this dye and transporter expression levels (Kuhnel et al., 1997), to the point that common assays of the multidrug resistance pump (MDR) inhibition were based on cellular accumulation of this dye (Chaoui et al., 2006; Van Leeuwen et al., 2009). JC-1 is known to be a substrate of transporters (MDR1) and numerous fluorescent probes. Moreover, it is established that differentiated HepaRG cells exhibit expression levels of both sinusoidal and canalicular drug transporters at least two-fold those of primary hepatocytes (Le Vee et al., 2006; Antherieu et al., 2010). These high levels of expression have been confirmed in selective pure hepatocytes for MDR1 and MRP2. This led us to postulate that JC-1 was actively effluxed by these transmem- brane efflux proteins. Indeed, we obtained a maximal dye charge in pure differentiated HepaRG hepatocytes when we achieved a saturation of dye in intact mitochondria with 2 successive loadings of JC-1 at limited dye concentration due to the poor solubility of this compound. Specificity of mitochondrial loading was confirmed by using the FCCP protonophore, an uncoupler of mitochondrial oxidative phosphoryla- tion and as expected, verapamil which actively blocks P glycoprotein and MRP transport activity, was found to be a powerful inhibitor of JC-1 efflux (Takano et al., 2009). To perform our toxicity testing in the best rapid conditions of dye loading, we decided to co-incubate JC-1 with verapamil once cells were treated with drugs. The robustness of this new combination of conditions was confirmed by observing the same IC50 values as those obtained with increasing doses of the reference apoptotic molecule staurosporin via either two or one loading in the presence of verapamil. These results did not rule out a possible contribution of CYPs in JC-1 metabolism. We used ketoconazole, a compound mainly known by scientists as CYP3A4 inhibitor to improve JC-1 accumulation. Ketoconazole appeared less efficient than verapamil in JC-1 loading suggesting that JC-1 is a poor CYP substrate. However, the weak effect of ketoconazole observed on dye accumulation could be attributed to its MDR1 inhibitor property (Salphati and Benet, 1998).
Finally, the efficiency of the test was attested by comparing these values to those obtained with the well known annexinV and MTT tests and extended to distinct prototypical hepatotoxic molecules. We showed that the same values were obtained with the annexinV and MTT tests, confirming the robustness of our new JC-1 test protocol. However, the JC-1 test did appear more sensitive since it was able to detect toxicity at earlier stage (3 h of exposure) than MTT.

In conclusion, evaluation of very early apoptotic effects of different hepatotoxic drugs can be reliably performed by using a quantitative fluorescence microscopy test coupling mitochondrial integrity and automated imaging analysis. Our hepatotoxicity drug test presents the advantages of i) preserving cellular functions by in situ examination, ii) using a hepatic cell model expressing most functions characteristic of liver, and iii) deriving benefit from multiparametric quantitative measurements which ascertain the quality level of the cells during the time of testing. Using JC-1 to simultaneously evaluate both multidrug resistant pump activity and apoptotic effects of target drugs in this HepaRG hepatocyte environment should provide new insights to further characterize molecular mechanisms compound 991 of drug toxicity and efficiently screen novel therapeutic molecules.