Potential of connexin 32 as a predictive marker for drug-induced cholestatic liver injury in a collagen vitrigel-culture model of HepG2-NIAS cells, a new subline of HepG2 cells, with bile canaliculus-like structures

Abstract

Cholestatic drug-induced liver injury (DILI) is caused by the aberrant excretion of bile acids (BAs) from hepatocytes via bile canaliculus-like structures (BCLSs) into the bile ducts. The precise in vitro evaluation method for cholestatic DILI has not been established due to a lack of specific markers and cell resources. We previously reported that HepG2-NIAS cells cultured on a collagen vitrigel (CV) membrane formed BCLSs with high protein expression of transporters involved in the excretion of BAs, including bile salt export pump (BSEP). In this study, the potential of connexin (Cx) 32, a component of gap junction, as a predictive marker for cholestatic DILI was investigated using a CV-culture model of HepG2-NIAS cells. The cells were treated with 7 drugs with different DILI-risk levels, and cell toxicity and Cx32 expression were evaluated. Cell toxicity was significantly increased not only by high DILI-risk drugs (troglitazone and cyclosporine A) but also by chlorpromazine with low DILI-risk. Furthermore, cell toxicity of troglitazone was not enhanced by a co-treatment with taurocholate, suggesting the low involvement of inhibition of BA excretion via BSEP in cholestatic DILI. In contrast, the total protein expression of Cx32 and co-localization of Cx32 and F-actin, which is composed of BCLSs, were significantly increased only by high DILI-risk drugs. Treatment with high DILI-risk drugs also induced the increased protein expression of zonula occludens (ZO)-1, which supports BCLSs concerted with Cx32. These results suggest that Cx32 expression in the CV-culture model of HepG2-NIAS cells may be a prominent predictive marker for cholestatic DILI.

INTRODUCTION

Drug-induced liver injury (DILI) has been one of the leading causes of failure in both clinical and post-approval stages of drug development (Norman, 2020). Precise prediction of the DILI risk of candidate drugs at an early preclinical stage is indispensable for efficient drug development. DILI is classified into hepatocellular, cholestatic, or mixed patterns, and they are caused by various mechanisms (Allison et al., 2023). One of the major causes of cholestatic DILI is an accumulation of bile acids (BAs) such as taurocholic acid (TA), cholic acid, chenodeoxycholic acid, and glycochenodeoxycholic acid into parenchymal hepatocytes. BAs produced by hepatocytes are delivered to bile canaliculus-like structures (BCLSs) between hepatocytes and excreted into the bile ducts via transporters, including bile salt export pump (BSEP). Thus, the cholestatic DILI risk of candidate chemicals was assessed by their BSEP inhibition activity using BSEP-expressing vesicles in previous studies (Dawson et al., 2012; Morgan et al., 2013). The disadvantage of this simplified cell-free system is that it cannot be applied to the evaluation of metabolites derived from candidate chemicals and the involvement of other mechanisms apart from BSEP inhibition. Therefore, there is an urgent need for a novel in vitro prediction method for cholestatic DILI.

Cells subjected to the evaluation of cholestatic DILI should express both high growth ability and hepatic functions such as cytochrome P450 (CYP) activity, formation of BCLSs, and transporter activity (Ozawa et al., 2021). Representative cell resources are primary human or rat hepatocytes, human induced pluripotent stem cell (hiPSC)-derived hepatocyte-like cells, and human hepatoma cells. Primary human hepatocytes are the gold standard model for the assessment of drug metabolism, excretion, and toxicity in a preclinical study due to their high hepatic function. However, they have lot-to-lot variation, poor proliferative potential, and lose most of the hepatic function, such as albumin secretion and urea synthesis within a week (Chen et al., 2021). To maintain hepatic function for a long time, various culture systems have been fabricated; for example, a co-culture system of hepatocytes and stromal cells and a sandwich culture system of hepatocytes, in which hepatocytes are embedded between a collagen substratum and collagen or Matrigel overlay (Khetani and Bhatia, 2008; Olsavsky Goyak et al., 2010). hiPSC-derived hepatocyte-like cells are a prominent cell resource while results of in vitro assay using them are low reproducibility due to a lack of standard differentiation procedure (Gerbal-Chaloin et al., 2014). Among human hepatoma cells, HepaRG cells are a popular liver model. HepaRG cells differentiated by long-term treatment with 2% dimethyl sulfoxide show high metabolic activity equivalent to primary human hepatocytes (Gripon et al., 2002). The disadvantage of this cell line is that their use is strictly limited, making it difficult to apply them to a screening of numerous candidate chemicals. HepG2 cells, another widely used human hepatoma cell line, show very low mRNA expression levels of CYP families and transporters compared to cryopreserved primary human hepatocytes although they have high proliferative ability (Ulvestad et al., 2013).

Recently, we established HepG2-NIAS cells from HepG2 cells registered in the RIKEN cell bank (RCB) with the cell line number of 1648 (HepG2-RCB1648 cells) by repeatedly subculturing for a long period using our culture technique for diminishing the occasion of cellular damage. HepG2-NIAS cells expressed high CYP3A4 activity, transporter protein expression, and potential for the formation of BCLSs shortly after oxygenation culture via a collagen vitrigel (CV) membrane (Takezawa and Uzu, 2022). These characteristics were hardly observed in their parental HepG2-RCB1648 cells. Furthermore, fluorescein metabolized from fluorescein diacetate accumulated into BCLSs in HepG2-NIAS cells was selectively excreted toward the apical side of cholangiocarcinoma cells when they were co-cultured, suggesting the high activity of BA excretion in HepG2-NIAS cells (Tokito et al., 2024; Uzu and Takezawa, 2023). These results suggest that the CV-culture model of HepG2-NIAS cells is suitable for the assessment of cholestatic DILI.

Meanwhile, Uzu et al. previously reported that the protein expression of connexin (Cx) 43, a component of a gap junction (GJ), was decreased in malignant mesothelioma cells by treatment with an anti-cancer agent (Uzu et al., 2017). GJ allows a direct transfer of several second messengers and small hydrophilic metabolites between adjacent cells, including Ca²⁺, 2’, 3’-cyclic GMP-AMP (cGAMP), microRNA, and glucose, which contributes to the maintenance of cellular homeostasis (Chen et al., 2016; Kandouz and Batist, 2010; Rouach et al., 2008; Zong et al., 2016). There are 21 known Cx subtypes in the human genome (Söhl et al., 2005). Cx43 and Cx32 are ubiquitously expressed in various tissues, including the liver, although some isoforms are expressed only in specific ones (Leithe et al., 2006). Cx32 and Cx26 are abundant in hepatocytes. Aberrant BCLS dilation and bile secretion were observed in the liver of a Cx32-deficient mouse although Cx26 expression remained (Temme et al., 2001). This observation suggests that Cx32 especially plays a key role in liver function. Furthermore, Cx32 protein expression was reduced in the liver in a rat with chronic liver injury induced by treatment with carbon tetrachloride (Nakata et al., 1996). Considering these findings, it is rational that Cx32 protein expression may be changed in hepatocytes exposed to hepatotoxic drugs, which causes BCLS dysfunction and leads to cholestatic DILI.

In this study, we aimed to verify the potential of Cx32 as a predictive marker for cholestatic DILI utilizing the CV-culture model of HepG2-NIAS cells. We found that the expression of Cx32 was altered only by treatment with drugs with high DILI risk while cell toxicity assessed by lactate dehydrogenase leakage was detected by treatment with drugs with both high and low DILI risk.

MATERIALS AND METHODS

Reagents

4′,6-diamidino-2-phenylindole (DAPI) (cat# 19178-91), LDH Cytotoxicity Assay Kit (cat# 18250-35), and taurocholic acid (TA) (cat# 32729-61) were purchased from Nacalai Tesque (Kyoto, Japan). Dulbecco’s Modified Eagle Medium (DMEM) (cat# 11885084) and rhodamine phalloidin (cat# R415) were purchased from ThermoFisher Scientific (Waltham, MA, USA). Fetal bovine serum (FBS) (cat# 175012) was purchased from Nichirei Biosciences (Tokyo, Japan). All other reagents not specified above were commercial products of the highest grade.

In this study, the following drugs were used as test compounds: atropine (Atr) (cat# 03533-11) and cyclosporine A (CsA) (cat# 12281-44) purchased from Nacalai Tesque, fenofibrate (Fen) (cat# F6020-5G) and piroxicam (Pir) (cat# P5654-1G) purchased from Sigma-Aldrich (St. Louis, MO, USA), miglitol (Mig) (cat# 15014) purchased from Cayman Chemical (Ann Arbor, MI, USA), and troglitazone (Tro) (cat# 209-19481) purchased from Wako Pure Chemicals (Osaka, Japan). The DILI risk level of each drug is shown in Table 1.

Table 1. Drugs used in this study.

Drug	FDA drug labeling	FDA DILI concern	Cmax, u in humans (µM)
Troglitazone (Tro)	Withdrawn	Most concern	0.064
Cyclosporin A (CsA)	Warnings and precautions	Most concern	0.054
Fenofibrate (Fen)	Warnings and precautions	Less concern	0.249
Piroxicam (Pir)	Warnings and precautions	Less concern	0.105
Chlorpromazine (Cpz)	Adverse reactions	Less concern	0.021
Miglitol (Mig)	No match	No concern	5.7
Atropine (Atr)	No match	No concern	0.0017

Drugs used in this study. Maximum unbound concentrations in plasma (C_{max, u}) for Tro, CsA, Fen, Pir, and Cpz are from the following reports: Dawson et al., 2012; Takemura et al., 2016, and those for Mig and Atr are calculated by multiplying unbound fraction and maximum concentrations in plasma from the following reports: Jain et al., 2011; Mubaslat et al., 2022.

Cell culture

HepG2-NIAS cells (cell line number: RCB4679) were obtained from RIKEN BioResource Center (Tsukuba, Japan). They were cultured as previously described (Takezawa and Uzu, 2022). Briefly, they were cultured in a DMEM containing 10% heat-inactivated FBS, 20 mM N-2-hydroxyethylpipearadine-N’-2-ethansulfonic acid (HEPES), 100 units/mL penicillin, and 100 μg/mL streptomycin at 37°C in a humidified atmosphere of 5% CO₂ and 95% air. The culture medium was changed every other day.

Preparation of CVM chambers

A chamber (ad-MED Vitrigel^®2) with 0.5 cm² or 1.0 cm² of a collagen xerogel membrane (CXM) containing 0.25 mg or 0.5 mg type-I collagen was purchased from Kanto Chemical Co., Inc. (Tokyo, Japan). The chamber with 0.5 cm² or 1.0 cm² of a CXM was set in a well of a 24-well plate or 12-well plate, respectively. The CXM with 0.5 cm² or 1.0 cm² was immersed in a culture medium by pouring 0.75 mL or 1.5 mL outside and 0.25 mL or 0.5 mL inside the chamber in the well for 10 min to convert the xerogel into a vitrigel immediately before use, respectively, resulting in the preparation of a chamber with a collagen vitrigel membrane (CVM) of about 20 μm in thickness. For culturing on the liquid-solid interface, the CVM chamber after removing the culture media was transferred onto the center of a silicone-coated polyethylene terephthalate (PET) film in a round shape with a diameter of 21 mm and dried for 30 min in a clean bench to obtain the chamber of which the bottom surface-weakly attached to the film as shown in the supplemental figure of a previous paper (Takezawa and Uzu, 2022).

Oxygenation culture via a CVM

In this study, 2 types of culture methods were used. HepG2-NIAS cells were cultured for 2 days on the liquid-solid interface and for an additional 1 day on the liquid-gas interface (a previous method using a CVM with 1.0 cm²) or cultured for 3 days on the liquid-gas interface (a novel method using a CVM with 0.5 cm²). The cells were seeded in a CVM chamber with a silicone-coated PET film pre-set in a culture dish with a diameter of 60 mm at an initial cell density of 5.0 x 10⁴ cells/cm² in the previous method and were seeded in a CVM pre-set in a well of a 24-well plate at an initial cell density of 7.0 x 10⁴ cells/cm² in the novel method. After 2 days, the silicone-coated PET film was gently detached from the bottom surface of the CVM chamber by pouring 5.0 mL of culture medium into the outside of the chambers, and subsequently, the CVM chamber in which cells uniformly proliferated was transferred to an empty well of a 12-well plate and the cells were cultured for an additional 1 day on the liquid-gas interface. The culture media were changed every day in both culture methods.

Cell toxicity

HepG2-NIAS cells subjected to the novel oxygenation culture method were treated with 50 μM test compounds described in Table 1, 200 μM TA, or their vehicle. After 24 hr, the activity of lactate dehydrogenase (LDH) released from damaged cells (LDH_sample) was measured using the LDH Cytotoxicity Assay Kit following the manufacturer’s instructions. Cell toxicity was expressed as a percentage of maximum LDH activity (LDH_lysis) measured in control cells treated with lysis solution included in the assay kit. The following equation was used:

(LDH_sample-LDH_vehicle)/(LDH_lysis-LDH_blank)×100 (%)

Formation of BCLSs and protein expression of Cx32 and zonula occludens (ZO)-1

HepG2-NIAS cells subjected to the novel oxygenation culture method were treated with 50 μM test compounds described in Table 1 or their vehicle. After 24 hr, the cells were fixed in 4% paraformaldehyde solution for 15 min at room temperature and washed with phosphate-buffered saline (PBS) twice. Subsequently, they were permeabilized in PBS containing 0.1% Triton X-100 for 15 min at room temperature and washed with PBS twice. The cells with CVM were isolated from the plastic frame of the chamber using an appropriate disposable biopsy punch and transferred onto a glass slide. Then, the specimens were blocked with PBS containing 3% bovine serum albumin (BSA) for 30 min at room temperature. For detecting Cx32 and ZO-1, the specimens were incubated with first antibodies (1:100) prepared in PBS containing 1% BSA overnight at 4°C, followed by washing with PBS three times. The following primary antibodies were used: rabbit polyclonal antibody anti-Cx32 (cat# 10450-1-AP, Proteintech, Rosemont, IL, USA) and mouse monoclonal antibody anti-ZO-1 (cat# 33-9100, ThermoFisher Scientific). Subsequently, they were incubated with secondary antibodies (1:200) and rhodamine phalloidin (1:400) for F-actin staining prepared in PBS for 1 hr at room temperature, followed by washing with PBS three times. The following secondary antibodies were used: goat anti-rabbit IgG conjugated to Alexa Fluor 488 (cat# A11008, ThermoFisher Scientific) and goat anti-mouse IgG conjugated to Alexa Fluor 647 (cat# A21236, ThermoFisher Scientific). Cell nuclei were counterstained with DAPI. Stained specimens were observed with a laser-scanning confocal microscope (model Zeiss LSM 780; Carl Zeiss microimaging GmbH, Jena, Germany). The obtained images were analyzed by ZEN blue edition (Carl Zeiss microimaging GmbH) to automatically calculate the colocalization rate of F-actin and Cx32/ZO-1. Also, cell nuclei stained with DAPI were counted and the total fluorescent intensity of Cx32/ZO-1 was analyzed by ImageJ2 (National Institutes of Health, Bethesda, MD, USA). The total expression of Cx32/ZO-1 per cell nucleus in each observed area was calculated by dividing the total fluorescent intensity of Cx32/ZO-1 by the number of cell nucleus.

Quantitative reverse transcription polymerase chain reaction (RT-qPCR)

HepG2-NIAS cells subjected to the novel oxygenation culture method were treated with 50 μM Tro, CsA, or their vehicle. After 24 hr, total RNA was isolated using ISOGEN II (NIPPON GENE, Tokyo, Japan), according to the manufacturer’ instructions. Complementary DNA was prepared using ReverTra Ace® qPCR RT Master Mix with gDNA Remover (Toyobo, Osaka, Japan), according to the manufacturer’ instructions. qPCR was performed using THUNDERBIRD® Next SYBR® qPCR Mix (Toyobo) and a CFX Duet Real-Time PCR System (Bio-Rad, Hercules, CA, USA). The following primers were used: GAPDH, (sense) 5’-GCACCGTCAAGGCTGAGAAC-3’ and (anti-sense) 5’-TGGTGAAGACGCCAGTGGA-3’, and GJB1, (sense) 5’-TTCTACTGCCATTGGCCGAG-3’ and (anti-sense) 5’-CTCTATGTGTTGCTGGTGAG-3’. The PCR reaction was performed by the following thermal profile: initial melting at 95°C for 3 min, followed by 40 cycles of melting at 95°C for 10 sec, and annealing at 60°C for 30 sec. The ΔΔCt method was used for relative quantification of mRNA with normalization to GAPDH.

Statistical analysis

Statistical analyses were performed using GraphPad Prism10 (GraphPad Software, San Diego, CA, USA). Results are shown as the mean ± standard deviation (SD). Statistical differences between groups were evaluated using a one-way analysis of variance followed by Dunnett’s test.

RESULTS

Comparison of BCLS formation of HepG2-NIAS cells subjected to previous and novel oxygenation culture methods

HepG2-NIAS cells were cultured on the liquid-solid interface for 2 days and then on the liquid-gas interface for 1 day (i.e., the previous method) or cultured on the liquid-gas interface for 3 days (i.e., the novel method) and their F-actin, a component of BCLSs, was stained. The growth rate was almost the same and reached confluency after a 3-day culture of HepG2-NIAS cells in both culture methods (Fig. 1). The cells formed a monolayer and F-actin was uniformly expressed, suggesting BCLS formation was induced in the cells subjected to the both culture methods. Therefore, HepG2-NIAS cells were cultured using the novel culture method in the following experiments.

Fig. 1

Observation of BCLS networks in HepG2-NIAS cells subjected to a previous (a, c) or a novel (b, d) oxygenation culture using a CVM chamber. The cells were fixed on day 3, and their F-actin (a, b) and nuclei (c, d) were stained with rhodamine phalloidin and DAPI, respectively. Scale bar, 50 μm.

Cell toxicity of test compounds

HepG2-NIAS cells subjected to the novel culture method were incubated with 7 test compounds with different DILI risk levels (Table 1) for 24 hr and cell toxicity was assessed. Treatment with Tro and CsA, both of which are strong cholestatic DILI inducers, caused a significant increase in LDH leakage (Fig. 2). Treatment with Cpz, which is ranked as a weak DILI inducer, caused the same level of cell toxicity as that with Tro and CsA. These results suggest that cell toxicity assessed by LDH leakage does not always reflect the cholestatic DILI risks of test compounds.

Fig. 2

Cell toxicity of test compounds. HepG2-NIAS cells subjected to a novel oxygenation culture using a CVM chamber were treated with test compounds (50 μM) or vehicle (Cont) for 24 hr. Cell toxicity was calculated using LDH leakage based on the equation described in the Materials and Methods section. Each graph represents the mean ± SD with individual data points (n = 4-8). **P < 0.01, ***P < 0.001, ****P < 0.0001 corresponds to the statistical difference between Cont.

Cell toxicity of the combination of Tro and TA

TA is a major substrate for BSEP and Tro is a strong BSEP inhibitor with IC₅₀ of 0.5-2.7 μM (Dawson et al., 2012; Morgan et al., 2013; Zhang et al., 2016). The combination of 200 μM TA and 50 μM Tro is expected to exhibit enhanced toxicity compared to Tro or TA alone if an accumulation of BAs into hepatocytes by BSEP inhibition largely contributes to the cholestatic DILI. It was confirmed that 200 μM TA alone did not affect cell viability (Fig. 3). The same level of cell toxicity was detected by the combination treatment with TA and Tro and the single treatment with Tro, suggesting that the accumulation of TA into HepG2-NIAS cells by BSEP inhibition does not contribute to the cell toxicity of Tro.

Fig. 3

Combination effect of Tro and TA. HepG2-NIAS cells subjected to a novel oxygenation culture using a CVM chamber were treated with Tro (50 μM) and/or TA (200 μM), or their vehicle (Cont) for 24 hr. Cell toxicity was calculated using LDH leakage based on the equation described in the Materials and Methods section. Each graph represents the mean ± SD with individual data points (n = 5).

The changes in Cx32 expression by treatment with test compounds

Cx32 expression was observed by immunofluorescent staining. Treatment with Tro and CsA induced a significant increase in total Cx32 expression and Cx32 colocalized with F-actin (Fig. 4). These changes were not observed in treatment with other drugs with less or no risk of DILI (Fig. 4B), suggesting that the changes of Cx32 expression may become a precise marker for the cholestatic DILI. Also, treatment with Tro and CsA induced a significant decrease in GJB1 (coding Cx32) expression (Fig. 5).

Fig. 4

Effect of test compounds on Cx32 expression. HepG2-NIAS cells subjected to a novel oxygenation culture using a CVM chamber were treated with test compounds (50 μM) or vehicle (Cont) for 24 hr. (A) After fixation, the cells were stained with specific antibodies for Cx32 followed by staining with rhodamine phalloidin and DAPI to detect F-actin and nuclei, respectively. Arrowheads represent typical merged areas of Cx32 and F-actin. Scale bar, 10 μm. Colocalization rate of Cx32 and F-actin (B) and total Cx32 intensity (C) were quantified. Total Cx32 expression was calculated by dividing the total fluorescent intensity of Cx32 by the number of cell nucleus. Each graph represents the mean ± SD with individual data points (n = 8-17). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 corresponds to the statistical difference between Cont.

Fig. 5

Effect of Tro and CsA on GJB1 (coding Cx32) expression. HepG2-NIAS cells subjected to a novel oxygenation culture using a CVM chamber were treated with Tro or CsA (50 μM) or vehicle (Cont) for 24 hr. After RNA extraction, RT-qPCR was performed. The expression level of GJB1 was normalized to GAPDH expression. Each graph represents the mean ± SD with individual data points (n = 3). ****P < 0.0001 corresponds to the statistical difference between Cont.

The changes in ZO-1 expression by treatment with Tro and CsA

It is reported that Cx43 directly interacts with tight junction protein ZO-1 and their complex modulates F-actin dynamics (Chen et al., 2015). It is plausible that the changes in Cx32 expression caused the changes in ZO-1 expression. As expected, treatment with Tro and CsA induced a significant increase in total ZO-1 expression and ZO-1 colocalized with F-actin (Fig. 6). These results suggest that the changes in the interaction among Cx32, ZO-1, and F-actin might be involved in the induction of cholestatic DILI.

Fig. 6

Effect of Tro and CsA on ZO-1 expression. HepG2-NIAS cells subjected to a novel oxygenation culture using a CVM chamber were treated with Tro or CsA (50 μM) or vehicle (Cont) for 24 hr. After fixation, the cells were stained with specific antibodies for ZO-1 followed by staining with rhodamine phalloidin and DAPI to detect F-actin and nuclei, respectively. Arrowheads represent typical merged areas of ZO-1 and F-actin. Scale bar, 10 μm. Colocalization rate of ZO-1 and F-actin (B) and total ZO-1 intensity (C) were quantified. Total ZO-1 expression was calculated by dividing the total fluorescent intensity of ZO-1 by the number of cell nucleus. Each graph represents the mean ± SD with individual data points (n = 8-10). **P < 0.01, ***P < 0.001 corresponds to the statistical difference between Cont.

DISCUSSION

The plasma membrane of hepatocytes in vivo is categorized into three types of domains and has different protein expressions and functions: sinusoidal domain, lateral domain, and bile canalicular domain (Maurice et al., 1988). Bile canaliculi in vivo periodically dilate and contract due to the changes in the phosphorylation status of myosin light chain 2 and actin distribution (Sharanek et al., 2016), resulting in the unidirectional transportation of BAs from the cytosol of hepatocytes into the bile duct. Meanwhile, BCLSs in vitro are formed by the bile canalicular domain rich in drug transporters such as BSEP and supported by F-actin (actin polymer), tight junction-related proteins such as ZO-1, and GJ protein Cx32 (Song et al., 1996; Sudo et al., 2005). Considering these backgrounds, we focused on Cx32 as a key regulator for maintaining the structure and function of BCLSs and assumed that the expression or function of four transmembrane proteins (BSEP, F-actin, ZO-1, and Cx32) may be sensitively changed by treatment with drugs inducing cholestatic DILI.

The focus of this study is to show the possibility of Cx32 as a novel predictive marker for cholestatic DILI. First, we investigated the effect of the change of the protocol for fabricating the HepG2-NIAS cell culture model (Fig. 1). We previously reported that the cells cultured on the liquid-solid interface for 2 days and then on the liquid-gas interface for 1 day (i.e., the previous method) rapidly activated their hepatic function, including BCLS formation and drug transporter expression (Takezawa and Uzu, 2022). The disadvantage of this culture method is the complexity of transferring a CVM chamber from the liquid-solid interface to the liquid-gas interface. The liquid-solid interface was essential for the cells to uniformly spread on the original CVM because the membrane was warped (Oshikata-Miyazaki and Takezawa, 2016; Takezawa et al., 2004). However, a novel CVM chamber (ad-MED vitrigel^®2) was designed to pull the membrane tight. In this study, it was confirmed that the cells cultured in an ad-MED vitrigel^®2 chamber on the liquid-gas interface for 3 days (i.e., the novel method) formed uniform BCLS networks same as the previous one. Then, we investigated the correlation between cell toxicity and DILI risk using HepG2-NIAS cells cultured by the novel method (Fig. 2). As expected, treatment with Tro and CsA, both of which are strongly related to the cholestatic DILI, caused approximately 10% LDH leakage when that induced by the lysis solution treatment was taken as 100%. In contrast, treatment with Cpz, which is categorized as “less DILI concern”, caused the same level of cell toxicity. Cpz is a therapeutic agent for schizophrenia via mainly antagonizing dopamine D₂ receptors. It also blocks adrenaline α₁, acetylcholine M₃, and histamine H₁ receptors. It is reported that natural herb-induced hepatotoxicity is related to the inhibitory effect on signal transduction via dopamine D₂ and acetylcholine M₃ receptors (Li et al., 2021). Therefore, it is reasonable that the pharmacological effect of Cpz as a multi-receptor antagonist caused cell death. Next, the combination effect of Tro and TA was investigated to confirm the contribution of BA accumulation via BSEP inhibition to the toxic effect of Tro (Fig. 3). There was no difference in the cell toxicity between single treatment with Tro and cotreatment with TA and Tro although we expected the increase of the cell toxicity of Tro by TA. This result suggests that TA accumulation via BSEP inhibition was not involved in the cell toxicity of Tro.

Then, we observed Cx32 localization 24 hr after drug treatment (Fig. 4). In vehicle-treated (control) cells, Cx32 was weakly detected as dots near the plasma membrane or in the cytosol. Treatment with Tro and CsA strongly increased Cx32 expression in the whole cell, resulting in a significant increase in the colocalization rate of Cx32 and F-actin. Drugs categorized as “less or no DILI concern” did not modulate Cx32 expression. These results suggest that the change in Cx32 expression may be related to the cholestatic DILI. Treatment with Tro and CsA for 24 hr caused a significant decrease in GJB1 expression (Fig. 5). It is expected that these drugs promote GJB1 transcription at an earlier time point, resulting in the increase in Cx32 protein synthesis and the suppression of GJB1 transcription by negative feedback. The effect of Tro and CsA on the time-dependent expression of Cx32 protein and mRNA should be investigated in a future study.

It is expected that increased Cx32 altered the expression and/or function of BCLS component proteins. We focused on a tight junction protein ZO-1, which is well known to directly interact with Cx protein (Fig. 6). As expected, total ZO-1 expression level and colocalization with F-actin were increased by treatment with Tro and CsA. It is reported that Cx43 knockdown reduced ZO-1 expression in rat retinal epithelial cells while transfection with Cx32 enhanced ZO-1 expression in mouse hepatocytes, suggesting that Cx regulates ZO-1 expression (Kojima et al., 2002; Tien et al., 2013). That mechanism has not been elucidated, however, there is a possibility that Cx32 localizing at the nucleus enhances ZO-1 transcription. It is reported that Cx43 exists in two forms: the full-length molecule, which localizes at the plasma membrane and forms GJ, and smaller carboxy-terminal fragments, which localize at the nucleus and regulate gene transcription (Kotini et al., 2018; Xiong et al., 2024). In this study, Cx32 upregulated by Tro and CsA was detected in the nucleus, therefore, they could be smaller carboxy-terminal fragments and directly activate the ZO-1 transcription.

This is the first report that revealed the potential of Cx32 as a predictive marker for cholestatic DILI. However, the limitation of the current study is that assays were performed at a fixed concentration (50 μM) of drugs. In previous studies, the effect of 50 μM drugs with cholestatic DILI risks on BCLS function or formation was evaluated in HepaRG cells or primary rat hepatocytes, respectively (Burbank et al., 2016; Takemura et al., 2016). Therefore, the condition of drug treatment is considered to be reasonable although it is 10 to 1000 times higher than the clinically relevant concentration for each drug (Table. 1). The dose-response change of Cx32 expression should be investigated in a future study. Also, the predictive performance must be confirmed by increasing the number of test compounds.

Moreover, the evaluation method needs to be sophisticated aiming for industrial use. Observation by immunofluorescent staining needs time and complicated procedure, resulting in low throughput and reproducibility. Considering the result of the current study that drugs with high cholestatic DILI risk modulated the expression of cell-cell junction proteins including Cx32 and ZO-1, we expect that the measurement of transepithelial electrical resistance (TEER) value is useful for the detection of cholestatic DILI. Takezawa and collaborators established the Vitrigel-Eye Irritancy Test method registered as OECD Test No. 494, in which the eye irritancy of test chemicals is judged by TEER value change of the CV-culture model of HCE-T cells (a human corneal epithelium-derived cell strain) (OECD, 2019; Yamaguchi et al., 2013). The measurement of TEER value can supply high throughput and reproducible results because it just requires insertion of the electrode into a CVM chamber. Thus, we are going to measure the TEER value of the CV-culture model of HepG2-NIAS cells exposed to drugs with high/low cholestatic DILI risks in a future study. We hope that the current study will contribute to the toxicological assays in drug development studies.

ACKNOWLEDGMENTS

This work was supported by the grant to M. Uzu from The Mochida Memorial Foundation for Medical and Pharmaceutical Research, and by the grant to T. Takezawa for the collaborative research between Kake Educational Institution and Kanto Chemical Co., Inc. from Kanto Chemical Co., Inc.

Conflict of interest

The authors declare that there is no conflict of interest.

REFERENCES

•  Allison,  R.,  Guraka,  A.,  Shawa,  I.T.,  Tripathi,  G.,  Moritz,  W. and  Kermanizadeh,  A. (2023): Drug induced liver injury - a 2023 update. J. Toxicol. Environ. Health B Crit. Rev., 26, 442-467.
•  Burbank,  M.G.,  Burban,  A.,  Sharanek,  A.,  Weaver,  R.J.,  Guguen-Guillouzo,  C. and  Guillouzo,  A. (2016): Early alterations of bile canaliculi dynamics and the rho kinase/myosin light chain kinase pathway are characteristics of drug-induced intrahepatic cholestasis. Drug Metab. Dispos., 44, 1780-1793.
•  Chen,  C.H.,  Mayo,  J.N.,  Gourdie,  R.G.,  Johnstone,  S.R.,  Isakson,  B.E. and  Bearden,  S.E. (2015): The connexin 43/ZO-1 complex regulates cerebral endothelial F-actin architecture and migration. Am. J. Physiol. Cell Physiol., 309, C600-C607.
•  Chen,  Q.,  Boire,  A.,  Jin,  X.,  Valiente,  M.,  Er,  E.E.,  Lopez-Soto,  A.,  Jacob,  L.,  Patwa,  R.,  Shah,  H.,  Xu,  K.,  Cross,  J.R. and  Massagué,  J. (2016): Carcinoma-astrocyte gap junctions promote brain metastasis by cGAMP transfer. Nature, 533, 493-498.
•  Chen,  Y.,  Tang,  D.,  Wu,  H.,  Wu,  Y.,  Yuan,  T.,  Zhang,  H.,  Jiao,  Y.,  Yu,  W. and  Yan,  H. (2021): Assessment of long-term functional maintenance of primary human hepatocytes to predict drug-induced hepatoxicity in vitro. Arch. Toxicol., 95, 2431-2442.
•  Dawson,  S.,  Stahl,  S.,  Paul,  N.,  Barber,  J. and  Kenna,  J.G. (2012): In vitro inhibition of the bile salt export pump correlates with risk of cholestatic drug-induced liver injury in humans. Drug Metab. Dispos., 40, 130-138.
•  Gerbal-Chaloin,  S.,  Funakoshi,  N.,  Caillaud,  A.,  Gondeau,  C.,  Champon,  B. and  Si-Tayeb,  K. (2014): Human induced pluripotent stem cells in hepatology: beyond the proof of concept. Am. J. Pathol., 184, 332-347.
•  Gripon,  P.,  Rumin,  S.,  Urban,  S.,  Le Seyec,  J.,  Glaise,  D.,  Cannie,  I.,  Guyomard,  C.,  Lucas,  J.,  Trepo,  C. and  Guguen-Guillouzo,  C. (2002): Infection of a human hepatoma cell line by hepatitis B virus. Proc. Natl. Acad. Sci. USA, 99, 15655-15660.
•  Jain,  R.,  Lukram,  O. and  Dwivedi,  A. (2011): Ultra-performance liquid chromatography electrospray ionization-tandem mass spectrometry method for the estimation of miglitol in human plasma using metformin as the internal standard. Drug Test. Anal., 3, 255-262.
•  Kandouz,  M. and  Batist,  G. (2010): Gap junctions and connexins as therapeutic targets in cancer. Expert Opin. Ther. Targets, 14, 681-692.
•  Khetani,  S.R. and  Bhatia,  S.N. (2008): Microscale culture of human liver cells for drug development. Nat. Biotechnol., 26, 120-126.
•  Kojima,  T.,  Spray,  D.C.,  Kokai,  Y.,  Chiba,  H.,  Mochizuki,  Y. and  Sawada,  N. (2002): Cx32 formation and/or Cx32-mediated intercellular communication induces expression and function of tight junctions in hepatocytic cell line. Exp. Cell Res., 276, 40-51.
•  Kotini,  M.,  Barriga,  E.H.,  Leslie,  J.,  Gentzel,  M.,  Rauschenberger,  V.,  Schambony,  A. and  Mayor,  R. (2018): Gap junction protein Connexin-43 is a direct transcriptional regulator of N-cadherin in vivo. Nat. Commun., 9, 3846.
•  Leithe,  E.,  Sirnes,  S.,  Omori,  Y. and  Rivedal,  E. (2006): Downregulation of gap junctions in cancer cells. Crit. Rev. Oncog., 12, 225-256.
•  Li,  S.,  Wang,  Y.,  Li,  C.,  Yang,  N.,  Yu,  H.,  Zhou,  W.,  Chen,  S.,  Yang,  S. and  Li,  Y. (2021): Study on Hepatotoxicity of Rhubarb Based on Metabolomics and Network Pharmacology. Drug Des. Devel. Ther., 15, 1883-1902.
•  Maurice,  M.,  Rogier,  E.,  Cassio,  D. and  Feldmann,  G. (1988): Formation of plasma membrane domains in rat hepatocytes and hepatoma cell lines in culture. J. Cell Sci., 90, 79-92.
•  Morgan,  R.E.,  van Staden,  C.J.,  Chen,  Y.,  Kalyanaraman,  N.,  Kalanzi,  J.,  Dunn,  R.T. 2nd,  Afshari,  C.A. and  Hamadeh,  H.K. (2013): A multifactorial approach to hepatobiliary transporter assessment enables improved therapeutic compound development. Toxicol. Sci., 136, 216-241.
•  Mubaslat,  O.,  Fitzpatrick,  M.,  McLachlan,  A.J. and  Lambert,  T. (2022): Pharmacokinetics and Effects on Saliva Flow of Sublingual and Oral Atropine in Clozapine-Treated and Healthy Adults: An Interventional Cross-Over Study. Psychiatry Clin. Psychopharmacol., 32, 17-27.
•  Nakata,  Y.,  Iwai,  M.,  Kimura,  S. and  Shimazu,  T. (1996): Prolonged decrease in hepatic connexin32 in chronic liver injury induced by carbon tetrachloride in rats. J. Hepatol., 25, 529-537.
•  Norman,  B.H. (2020): Drug Induced Liver Injury (DILI). Mechanisms and Medicinal Chemistry Avoidance/Mitigation Strategies. J. Med. Chem., 63, 11397-11419.
• OECD. (2019): Test No. 494: Vitrigel-Eye Irritancy Test Method for Identifying Chemicals Not Requiring Classification and Labelling for Eye Irritation or Serious Eye Damage. OECD Publishing, Paris, France.
•  Olsavsky Goyak,  K.M.,  Laurenzana,  E.M. and  Omiecinski,  C.J. (2010): Hepatocyte differentiation. Methods Mol. Biol., 640, 115-138.
•  Oshikata-Miyazaki,  A. and  Takezawa,  T. (2016): Development of an oxygenation culture method for activating the liver-specific functions of HepG2 cells utilizing a collagen vitrigel membrane chamber. Cytotechnology, 68, 1801-1811.
•  Ozawa,  S.,  Miura,  T.,  Terashima,  J.,  Habano,  W. and  Ishida,  S. (2021): Recent Progress in Prediction Systems for Drug-induced Liver Injury Using In vitro Cell Culture. Drug Metab. Lett., 14, 25-40.
•  Rouach,  N.,  Koulakoff,  A.,  Abudara,  V.,  Willecke,  K. and  Giaume,  C. (2008): Astroglial metabolic networks sustain hippocampal synaptic transmission. Science, 322, 1551-1555.
•  Sharanek,  A.,  Burban,  A.,  Burbank,  M.,  Le Guevel,  R.,  Li,  R.,  Guillouzo,  A. and  Guguen-Guillouzo,  C. (2016): Rho-kinase/myosin light chain kinase pathway plays a key role in the impairment of bile canaliculi dynamics induced by cholestatic drugs. Sci. Rep., 6, 24709.
•  Söhl,  G.,  Maxeiner,  S. and  Willecke,  K. (2005): Expression and functions of neuronal gap junctions. Nat. Rev. Neurosci., 6, 191-200.
•  Song,  J.Y.,  Van Marle,  J.,  Van Noorden,  C.J. and  Frederiks,  W.F. (1996): Redistribution of Ca²⁺, Mg²⁺-ATPase activity in relation to alterations of the cytoskeleton and tight junctions in hepatocytes of cholestatic rat liver. Eur. J. Cell Biol., 71, 277-285.
•  Sudo,  R.,  Kohara,  H.,  Mitaka,  T.,  Ikeda,  M. and  Tanishita,  K. (2005): Coordinated movement of bile canalicular networks reconstructed by rat small hepatocytes. Ann. Biomed. Eng., 33, 696-708.
•  Takemura,  A.,  Izaki,  A.,  Sekine,  S. and  Ito,  K. (2016): Inhibition of bile canalicular network formation in rat sandwich cultured hepatocytes by drugs associated with risk of severe liver injury. Toxicol. In Vitro, 35, 121-130.
•  Takezawa,  T.,  Ozaki,  K.,  Nitani,  A.,  Takabayashi,  C. and  Shimo-Oka,  T. (2004): Collagen vitrigel: a novel scaffold that can facilitate a three-dimensional culture for reconstructing organoids. Cell Transplant., 13, 463-473.
•  Takezawa,  T. and  Uzu,  M. (2022): HepG2-NIAS cells, a new subline of HepG2 cells that can enhance not only CYP3A4 activity but also expression of drug transporters and form bile canaliculus-like networks by the oxygenation culture via a collagen vitrigel membrane. J. Toxicol. Sci., 47, 39-50.
•  Temme,  A.,  Stümpel,  F.,  Söhl,  G.,  Rieber,  E.P.,  Jungermann,  K.,  Willecke,  K. and  Ott,  T. (2001): Dilated bile canaliculi and attenuated decrease of nerve-dependent bile secretion in connexin32-deficient mouse liver. Pflugers Arch., 442, 961-966.
•  Tien,  T.,  Barrette,  K.F.,  Chronopoulos,  A. and  Roy,  S. (2013): Effects of high glucose-induced Cx43 downregulation on occludin and ZO-1 expression and tight junction barrier function in retinal endothelial cells. Invest. Ophthalmol. Vis. Sci., 54, 6518-6525.
•  Tokito,  F.,  Kiyofuji,  M.,  Choi,  H.,  Nishikawa,  M.,  Takezawa,  T. and  Sakai,  Y. (2024): Modulation of hepatic cellular tight junctions via coculture with cholangiocytes enables non-destructive bile recovery. J. Biosci. Bioeng., 137, 403-411.
•  Ulvestad,  M.,  Nordell,  P.,  Asplund,  A.,  Rehnström,  M.,  Jacobsson,  S.,  Holmgren,  G.,  Davidson,  L.,  Brolén,  G.,  Edsbagge,  J.,  Björquist,  P.,  Küppers-Munther,  B. and  Andersson,  T.B. (2013): Drug metabolizing enzyme and transporter protein profiles of hepatocytes derived from human embryonic and induced pluripotent stem cells. Biochem. Pharmacol., 86, 691-702.
•  Uzu,  M.,  Sato,  H.,  Shimizu,  A.,  Shibata,  Y.,  Ueno,  K. and  Hisaka,  A. (2017): Connexin 43 enhances Bax activation via JNK activation in sunitinib-induced apoptosis in mesothelioma cells. J. Pharmacol. Sci., 134, 101-107.
•  Uzu,  M. and  Takezawa,  T. (2023): Novel cell culture technology to harvest hepatic metabolites accumulated in the bile canaliculus-like structures between HepG2-NIAS cells utilizing crosstalk with TFK-1 cells. Nano Biomed., 15, 75-87.
•  Xiong,  X.,  Chen,  W.,  Chen,  C.,  Wu,  Q. and  He,  C. (2024): Analysis of the function and therapeutic strategy of connexin 43 from its subcellular localization. Biochimie, 218, 1-7.
•  Yamaguchi,  H.,  Kojima,  H. and  Takezawa,  T. (2013): Vitrigel-eye irritancy test method using HCE-T cells. Toxicol. Sci., 135, 347-355.
•  Zhang,  J.,  He,  K.,  Cai,  L.,  Chen,  Y.C.,  Yang,  Y.,  Shi,  Q.,  Woolf,  T.F.,  Ge,  W.,  Guo,  L.,  Borlak,  J. and  Tong,  W. (2016): Inhibition of bile salt transport by drugs associated with liver injury in primary hepatocytes from human, monkey, dog, rat, and mouse. Chem. Biol. Interact., 255, 45-54.
•  Zong,  L.,  Zhu,  Y.,  Liang,  R. and  Zhao,  H.B. (2016): Gap junction mediated miRNA intercellular transfer and gene regulation: A novel mechanism for intercellular genetic communication. Sci. Rep., 6, 19884.