Exploring the potential of liver microphysiological systems of varied configurations to model cholestatic chemical effects
Katharina S. Nitsche, Courtney Sakolish, Paul L. Carmichael, Philip Hewitt, Piyush Bajaj, Stephen S. Ferguson, Sarah M. Lloyd, Sarah S. Wilson, Hans Bouwmeester, Ivan Rusyn
TL;DR
This study compares different liver models to see how well they can predict cholestasis caused by chemicals.
Contribution
The study identifies that PhysioMimix LC12 with PHH is most effective in detecting cholestatic effects.
Findings
HepaRG and PHH showed comparable liver function markers in 2D and PhysioMimix LC12.
Bile acid synthesis was highest with PHH in PhysioMimix LC12.
Cholestatic agents reduced bile acid release only in PhysioMimix LC12, with PHH showing more consistent responses.
Abstract
Human in vitro liver tissue models have evolved to maintain hallmarks of hepatocellular function for extended periods with potential to model aspects of cholestasis for drug and chemical safety applications. Microphysiological systems (MPS) have been suggested as promising new approaches to model liver physiology and predict chemical-induced cholestasis in humans. This study comprehensively compared both basal function and toxicant-induced effects in 2D cultures and three liver MPS (i.e., 2-lane OrganoPlate, 3-lane OrganoPlate and PhysioMimix LC12) that were seeded with either HepaRG cells, primary human hepatocytes (PHH), or human induced pluripotent stem cell (iPSC)-derived hepatocytes. PHH and iPSC-derived hepatocytes (iHeps) were tested up to 7 days while HepaRG were evaluated over 30 days. Albumin, urea, CYP3A4 activity, and bile acids were measured. HepaRG and PHH showed…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6- —http://dx.doi.org/10.13039/100000066National Institute of Environmental Health Sciences
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
Topics3D Printing in Biomedical Research · Liver physiology and pathology · Pharmacogenetics and Drug Metabolism
Introduction
The liver has various physiological functions including production of proteins, bile acids, hormones and other biologically active molecules; it also plays a critical role in metabolism for most orally-administered xenobiotics (Almazroo et al. 2017). As all orally ingested and absorbed drugs and chemicals pass through the liver, this organ is particularly susceptible to their toxic effects (Rusyn et al. 2021). Drug-induced liver injury (DILI) is a major challenge for drug development (Andrade et al. 2019; Thakkar et al. 2020). Despite extensive profiling of drug candidates using both in vitro and in vivo models, DILI continues to be one of the top reasons for post-market drug withdrawals and black box warnings of new molecular entities (Olson et al. 2000; Sistare et al. 2016). The complexity of potential DILI mechanisms and cell–cell interactions in the liver (Kaplowitz 2005), or biological differences between species, cannot be fully recapitulated in conventional cell-based models (Soldatow et al. 2013). Thus, there is high demand for innovative approaches to devise in vitro models that can maintain long-term functionality for metabolic and transporter pathways to evaluate different DILI mechanisms (Ishida 2020; Vinken 2018). Also, while most clinical DILI cases are hepatocellular in nature, cholestatic and mixed types represent significant proportions, often comprising 20–40% of cases, each depending on the population and source of data (Pinazo-Bandera et al. 2023). In the past decade, many complex in vitro models for studying the effects of drugs and chemicals on the liver have been developed (Godoy et al. 2013; Ribeiro et al. 2019). While these novel approaches, especially microphysiological systems (MPS), aim to mimic human hepatic physiology and integrate media flow and a broader array of hepatic cell types (Gough et al. 2021; Mehta et al. 2024), only few have effectively modelled cholestatic DILI (Donato et al. 2022; Vinken 2018). Currently, sandwich and spheroid cultures of primary human hepatocytes (PHH) are considered one of best performing in vitro models to detect cholestatic compounds due to their longevity, metabolic capacity, and ability to closely mimic in vivo 3D cellular configuration as well as bile ductules (Vinken 2018).
Commercial availability of various MPS devices has increased their use in academic research and for drug and chemical safety evaluation. However, these platforms are yet to be used extensively for regulatory studies and decision making (Anklam et al. 2022; Marx et al. 2020; Nitsche et al. 2022). One of the challenges to wider acceptance of these models has been the lack of confidence in their robustness and reproducibility, especially the choice of cells that are used in the MPS (Rusyn et al. 2022). For example, previous comparative studies have shown that donor and cell source (e.g., primary vs induced pluripotent stem cell (iPSC)-derived hepatocytes vs liver-derived immortalized cells) may have a disproportionate impact on the “performance” of liver MPS (Kato et al. 2022; Lim et al. 2023; Negi et al. 2025; Sakolish et al. 2021). One solution to address the challenge of inter-donor variability with primary hepatocytes is to use HepaRG cells, a human hepatoma-derived cell line that has gained prominence as a surrogate for PHH in liver research (Lubberstedt et al. 2011). For toxicology studies, HepaRG cells are typically differentiated over multiple weeks of growth and differentiation into hepatocyte-like and biliary epithelial-like cells that display morphological and functional characteristics similar to adult PHH—they express and functionally display a broad spectrum of drug-metabolizing enzymes, transporters, and nuclear receptors (Nelson et al. 2017; Tascher et al. 2019). Because HepaRG can be passaged to some extent, cryopreserved, and re-differentiated to maintain stable hepatocellular functionality over weeks to months in culture, they are used as alternatives to PHH with longer-term utility for repeated drug exposure experiments, especially if concerns around donor variability and costs are relevant (Bell et al. 2017; Jackson et al. 2016; Kopp et al. 2024; McGill et al. 2011; Stanley and Wolf 2022).
One of the promising applications of HepaRG in mechanistic toxicology is studies of cholestatic liver injury given their capacity to differentiate into both hepatocyte and cholangiocyte lineages (Gijbels et al. 2019; Hendriks et al. 2016; Koga et al. 2023; Maerten et al. 2025; Rodrigues et al. 2018). Cholestasis is a form of DILI that is of high relevance to both drugs and non-pharmaceutical compounds, in which bile acid metabolism has a key role (Vilas-Boas et al. 2019, 2020). Different triggers and molecular events have been identified that can lead to bile acid accumulation in hepatocytes (Gijbels and Vinken 2019), most of these depend on maintaining the function of the canalicular transporters (Canet et al. 2015). HepaRG have been shown to be a sensible model because they express requisite transporters, often differentiate as mixed cell populations that include cholangiocytes, and can recapitulate key events in the adverse outcome networks involved in cholestasis (Gijbels et al. 2021, 2020). While most studies with HepaRG are done in traditional static monolayer (2D) cultures or, more recently, in spheroids and involve 7–14 day culture periods, recent work in flow-based MPS showed that HepaRG functionality can be maintained for 2–8 weeks (Boul et al. 2021; Duivenvoorde et al. 2021; Kopp et al. 2024). In these studies, MPS-cultured HepaRG showed robust CYP activity, albumin production, and responses to drugs.
A number of previous studies compared the utility of HepaRG cells to that of PHH and other liver-like cells in the context of drug and chemical safety studies (de Bruijn et al. 2022; Gerets et al. 2012; Lubberstedt et al. 2011; Sharanek et al. 2015; Szabo et al. 2013). However, few studies compared 2D and MPS models and different cell types, especially in the context of modeling functional aspects of bile acid composition related to cholestatic DILI. Thus, to evaluate robustness of cell and culture method (e.g., 2D vs. MPS) and their suitability as in vitro models of cholestatic DILI, we conducted a side-by-side comparison of PHH, HepaRG, and iPSC-derived hepatocytes (iHeps) in 2D and several MPS (constant perfusion-based PhysioMimix LC12 and intermittent flow-based Organoplate 2-late 96 and 3-lane 40). The experiments were conducted with PHH (in 2D and PhysioMimix LC12), iHeps (in 2D and OrganoPlate models), and HepaRG cells (in all models). Basal hepatocellular function was evaluated for up to 7 (2D and OrganoPlate models, all cell types) or 30 (PhysioMimix LC12, HepaRG cells) days. In addition, effects of three compounds known to induce cholestasis in vivo—bosentan (BOS), 2-Octynoic acid (2-Oct) and alpha-naphthyl isocyanate (ANIT)—were evaluated at the end of each study to investigate the proficiency of different MPS configurations to recapitulate known cholestatic effects.
Materials and methods
Chemicals and reagents
Bosentan hydrate (BOS; CAS#157212–55-0), 2-Octynoic acid (2-Oct; CAS# 5663-96-7), alpha-naphthyl isothiocyanate (ANIT; CAS# 551-06-4) were purchased from Millipore Sigma (Burlington, MA). Bile acids used as standards in LC–MS/MS method were individually purchased from Sigma-Aldrich (Zwijndrecht, The Netherlands; cholic acid, CAS# 81-25-4 and chenodeoxycholic acid, CAS# 474-25-9) along with their taurine and glycine conjugated derivatives (glycochenodeoxycholic acid CAS# 640-79-9; glycocholic acid, CAS# 475-31-0; taurochenodeoxycholic acid, CAS# 516-35-8 and taurocholic acid, CAS# 81-24-3).
The microfluidic MPS models used in this project were PhysioMimix LC12 from CN Bio (Cambridge, UK) and OrganoPlate® 2-lane 96 and 3-lane 40 from Mimetas (Leiden, The Netherlands). The PhysioMimix LC12 has the footprint of a traditional micro-well plate and contains 12 microfluidic chambers (Docci et al. 2022). Each chamber contains a media reservoir and a culture well consisting of a scaffold and filter plastic that is held in place with a retaining ring. Cells are seeded onto and cultured within the scaffold (Lim et al. 2023). The plate is attached to a pneumatic docking station, containing micropumps that introduce recirculating media flow that can be adjusted using the PhysioMimix LC12 controller. The OrganoPlates have the size of a traditional micro-well plate where the OrganoPlate® 2-lane 96 plate contains 96 individual chips with two lanes (Bircsak et al. 2021) and OrganoPlate® 3-lane 40 plate contains 40 individual chips with three lanes (Trietsch et al. 2017). In these models, one lane is filled with a gel (in the case of liver models, cells are placed in the gel channel) while the remaining channels can be perfused by slowly rocking the plate from side to side as detailed elsewhere (Bircsak et al. 2021; Trietsch et al. 2017). Black-walled, clear-bottom, tissue culture-treated 96-well plates (Cat# 3603; Corning, Corning, NY) were used for static 2D cell culture experiments.
Cells and cell culture reagents
Cells used in these studies are listed in Table 1. HepaRG cells were purchased in differentiated, cryopreserved format from ThermoFisher (Cat #HPRGC10, Lot HPRG116347-TA10; Waltham, MA). HepaRG Thawing and Plating medium consisted of William’s E media (Cat# A1217601, ThermoFisher) with general purpose supplements for thawing and plating (ThermoFisher, Cat# HPRG670, lot HPRG670006). HepaRG maintenance medium, was used for cell culture from day 1 until the end of each experiment, consisted of William’s E media with Maintenance/Metabolism Medium supplement (ThermoFisher; Cat# HPRG620, lot HPRG620011).Table 1. Experimental details on each experiment’s platforms, cells, cell density and duration. See Fig. 1 for the schematic and timeline of each experimentExperiment #Tested cellsVendor (Catalogue #)Cell culture platformPlating cell densityCell culture durationExp 1HepaRG ()ThermoFisher#HPR116347-TA10PhysioMimix LC12600 K cells/chip7 days96-well plate100 K cells/wellExp 2Primary human hepatocytesThermoFisher#HU8373PhysioMimix LC12600 K cells/chip7 days96-well plate100 K cells/wellExp 3HepaRG ()ThermoFisher#HPR116347-TA10PhysioMimix LC12600 K cells/chip30 days96-well plate100 K cells/wellExp 4HepaRG ()ThermoFisher#HPR116347-TA10OrganoPlate 2-lane 9620 K cells/chip7 days96-well plate100 K cells/welliHep 2.0FujiFilm-CDI#103,934OrganoPlate 2-lane 9620 K cells/chip96-well plate100 K cells/wellExp 5HepaRG ()ThermoFisher#HPR116347-TA10OrganoPlate 2-lane 9620 K cells/chip7 daysOrganoPlate 3-lane 4020 K cells/chip96-well plate100 K cells/welliHep 2.0FujiFilm-CDI#103,934OrganoPlate 2-lane 9620 K cells/chipOrganoPlate 3-lane 4020 K cells/chip96-well plate100 K cells/well(*) HepaRG cells were purchased pre-differentiated and maintained in 1.6% DMSO-containing maintenance medium as detailed in Methods
Single-donor PHH were obtained from ThermoFisher (HMCPIS, Lot #HU8373). PHH plating medium was used to thaw and seed cells. It consisted of William’s E medium (A1217601, ThermoFisher) with thawing and plating supplements (3.6% cocktail A, 5% FBS, 10 nM dexamethasone (CM3000, ThermoFisher). PHH maintenance medium was used for cell culture from day 1 until the end of each experiment; it consisted of William’s E medium with PHH maintenance supplements (4% cocktail B, 1 nM dexamethasone; CM4000, ThermoFisher).
Human iHeps (iCell Hepatocytes 2.0, iHeps 2.0) were purchased from FujiFilm-Cellular Dynamics International (C1023, lot# 103934; Santa Ana, CA). iCell Hep plating medium consisted of RPM 1640 phenol red-free media (11835030, ThermoFisher) with 2% B-27 supplement (17504044, ThermoFisher), 100 nM dexamethasone (D9184, Millipore Sigma), 25 µg/mL gentamicin (1574044, ThermoFisher), and 20 ng/mL oncostatin M(295-OM-010, R&D Systems, Minneapolis, MN); it was used for pre-differentiation of iHeps according to the manufacturer’s protocol. iHep maintenance medium consisted of RPMI 1640 phenol- red free media, 2% B-27, 100 nM dexamethasone, and 25 μg/mL gentamicin; it was used for cell culture in the OrganoPlate 2-lane 96 and OrganoPlate 3-lane 40 devices and in the 96-well plates.
HepaRG and PHH culture in the PhysioMimix LC12 and 96-well plates
The day when HepaRG or PHHs were seeded into PhysioMimix LC12 plates was defined as day 0 (Fig. 1). HepaRGs and PHHs were cultured in the PhysioMimix LC12 following the manufacturer’s protocols. In brief, PHHs were thawed into CHRM (CM7000, ThermoFisher) and HepaRG were thawed into Thawing and Plating medium, centrifuged at 100 × g for 10 min, resuspended in plating medium (specific for each cell type as detailed above) and then seeded into the PhysioMimix LC12 scaffold at a seeding density of ~ 600,000 cells/chip as detailed elsewhere (Lim et al. 2023). The flow was initiated at 1 µL/s. After overnight incubation, the medium was exchanged to cell-specific maintenance medium. Media samples were exchanged every 1–2 days by aspirating and replacing all liquid in the reservoir (1.8 mL/chip).Fig. 1. General study design for experiments performed using human hepatocytes in monoculture. The experiments were conducted with PHH (in 2D and PhysioMimix LC12), iHeps (in 2D and OrganoPlate models) and HepaRG cells (in all models). Basal liver function and cholestatic injury was evaluated for up to 7 (2D, PhysioMimix LC12 and OrganoPlate models) or 30 (PhysioMimix LC12) days. Timing of media changes with or without CYP3A4 assay and cholestatic injury induction treatment with Bosentan (BOS, 25 µM), 2-Octynoic acid (2-Oct, 70 µM) and alpha-naphthyl isocyanate (ANIT, 50 µM) are indicated (see symbol legend in inset). Media samples were collected for LDH, albumin and urea assays on several days (see symbol legend in inset). A Study 1 and 2 with HepaRG and PHH in 2D and PhysioMimix LC12 over 7 days. B Study 3 with HepaRG in 2D and PhysioMimix LC12 extended to 30 days. C Study 4 and 5 with HepaRG and iHep 2.0 in 2D and Organoplate 2- and 3-lane over 7 days
For the HepaRG 2D model, uncoated 96-well plates were used, whereas for PHH 2D model, 96-well plates were first coated with 50 μg/mL collagen I (CB354249, Corning) and 200 μg/mL fibronectin (F1141, Sigma-Aldrich, St. Louis, MO) and incubated at 37 °C, 5% CO_2_ for 1 h. Thawed HepaRG cells and PHHs were collected by centrifugation (100 × g, 10 min), resuspended in corresponding plating medium, and seeded at a density of 100,000 cells/well. After overnight incubation, culture medium was exchanged to the corresponding maintenance medium for each cell type. The total well volume was 200 μL, and plates were incubated at 37 °C, 5% CO_2_, with medium refreshed in all wells every 1–2 days (see Fig. 1).
For evaluation of bile acid profiles as biomarkers of cholestatic injury (Experiments 1 and 2; Table 1 and Fig. 1), HepaRG and PHH were exposed on culture days 5 and 6 to either 25 µM BOS, 70 µM 2-Oct, or 50 µM ANIT (dissolved in DMSO for the final vehicle concentration of 0.1% v/v). After 24 h, medium was collected and stored at -80 °C until further analysis. For longer duration studies in HepaRG (Experiment 3; Table 1 and Fig. 1) cells were exposed on culture day 28 and 29 and media samples were collected after 24 h.
HepaRG culture in the OrganoPlate 2-lane 96 and OrganoPlate 3-lane 40
The day when HepaRG were seeded into OrganoPlate® 2-lane 96 and 3-lane 40 devices was defined as Day 0 (Fig. 1). In brief, HepaRGs were thawed in Thawing and Plating medium, centrifuged at 100 × g for 10 min, and resuspended into 4 mg/mL collagen solution (Cultrex 3-D Culture Matrix Rat Collagen-I, 3447–020-01, R&D Systems; 5 mg/mL type 1 collagen, 1 M HEPES, 37 g/L sodium bicarbonate at a ratio of 8:1:1, respectively) at a density of 10,000 cells/µL. The HepaRG/collagen suspension (2 μL/chip) was kept on ice and gently injected into the inlet of the gel channel of each chip using an electronic pipettor. After injection, the plates were placed at 37 °C, 5% CO_2_ for 15 min to allow polymerization of the matrix. After incubation, 50 μL of maintenance medium was added into inlets and outlets of gel-free channels. The plates were then placed on the perfusion rocker platform (Mimetas) set to cycle every 8 min to a maximum angle of approximately 7° to induce gravity-driven liquid flow through the perfusion channels. Perfusion channel medium was collected, and fresh medium added every 1–2 days by aspirating all liquid from each chip and adding 50 μL of the fresh medium to inlet- and outlet openings.
iHeps culture in the OrganoPlate 2-lane 96, OrganoPlate 3-lane 40, and 96- well plates
The day when the cells were seeded into OrganoPlate® 2-lane 96 and 3-lane 40 devices and into 96-well plates was defined as Day 0. Before seeding, iHeps were pre-differentiated according to the manufacturer’s protocols. In brief, iPSC-derived iCell hepatocytes were thawed and counted to be seeded at a density of 2.5 × 10^6^ viable cells/well in plating medium on a 6-well plate pre-coated with type 1 collagen (Cellcoat®, 657950-005, Greiner Bio-One, Monroe, NC). Cells were allowed to attach over 16 h; unattached cells were removed with a medium change. iPSC-derived hepatocyte cells were pre-differentiated over 5 days with daily plating medium changes as detailed elsewhere (Grimm et al. 2015). For seeding in the OrganoPlate® 2-lane 96 and 3-lane 40 devices, we followed previously reported procedures (Bircsak et al. 2021; Kato et al. 2022). Briefly, the differentiated cell clusters were collected after treatment with StemPro Accutase (ThermoFisher, Cat# A1110501) and centrifugation (200 × g, 3 min). Cell clusters were resuspended into 3.33 mg/mL collagen solution (Cultrex 3-D Culture Matrix Rat Collagen-I, 3447–020-01, R&D Systems; 5 mg/mL type 1 collagen, 1 M HEPES, 37 g/L sodium bicarbonate at a ratio of 4:1:1, respectively) at a density of 10,000 cells/µL gel. The iHep/collagen suspension (2 μL/chip) was kept on ice and gently injected into the inlet of the gel channel of each device using an electronic pipettor. After injection, the plates were placed at 37 °C, 5% CO_2_ for 15 min to allow polymerization of the matrix. After incubation, 50 μL of maintenance medium was added into inlets and outlets of gel-free channels. The plates were then placed on the perfusion rocker platform (Mimetas) set to cycle every 8 min to a maximum angle of 7° to induce gravity-driven liquid flow through the perfusion channels. Perfusion channel medium was collected, and fresh medium added every 1–2 days by aspirating all liquid from each chip and adding 50 μL of fresh medium to inlet- and outlet openings.
Biochemical assays
Cell culture media samples collected in these experiments were evaluated for a variety of biomarkers. The ELISA assays were used for measurements of albumin (E88–129, Bethyl Laboratories, Montgomery, TX), urea (EIABUN, ThermoFisher and Stanbio, Cat# 0580–250) and lactate dehydrogenase (LDH, ab102526, Abcam, Cambridge, UK). Assays were conducted according to the manufacturer’s protocols. The P450 Glo 3A4 with Luciferin-IPA assay (V9001, ProMega, Madison, WI) was conducted using the manufacturer’s protocol with the IPA substrate added to wells/perfusion channels for 1 h prior to medium collection and luminescence plate reading using SpectraMax iD3 (Molecular Devices, San Jose, CA).
Bile acid profiling using liquid chromatography tandem mass spectrometry (LC–MS/MS)
Cell culture media samples (200 µL) were mixed with 200 µL internal standard (Chenodeoxycholic acid-D4, 330259W-100UG, Merck, Darmstadt, Germany) solution in acetonitrile on ice. Samples were vortexed and centrifuged (12,000 × g, 10 min, room temperature) to allow for protein precipitation. The supernatant was collected from each sample, lyophilized overnight in a Christ Alpha 1–2 LD plus freeze- dryer, and reconstituted in 20 µL methanol. The reconstituted samples were transferred to glass autosampler vials with 120 µL inserts and stored at − 20 °C prior to LC–MS/MS analysis.
Bile acid analysis (de Bruijn et al. 2022) was performed on a triple quadrupole LC–MS/MS system (model LCMS-8045; Shimadzu, Kawasaki City, Japan) for two primary unconjugated bile acids (Cholic acid and Chenodeoxycholic acid) and along with their taurine and glycine conjugated derivatives (glycochenodeoxycholic acid; glycocholic acid, taurochenodeoxycholic acid and taurocholic acid). Standards (in MeOH) and samples were separated on a Kinetex C18 column (1.7 μm × 100A × 50 mm × 2.1 mm; 00B-4475-AN; Phenomenex, Torrence, CA) with a guard precolumn (C18 2.1 mm; AJ0–8782; Phenomenex) using an ultra-high performance liquid chromatography system with gradient elution. MilliQ water with 0.01% formic acid served as mobile phase A, while methanol/acetonitrile (50% v/v) was used as mobile phase B. Data acquisition and processing were conducted using LabSolutions software (Shimadzu). The limit of detection (LOD) for each BA was defined as the lowest concentration yielding a signal-to-noise ratio greater than 5, while the limit of quantification (LOQ) was set at the lowest concentration with a signal-to-noise ratio above 10. The MS parameters, LODs, LOQs and representative chromatograms are provided in See Table S1 and Figures S1–S7.
Statistical analyses
General statistical analyses were conducted using GraphPad Prism 10 (San Diego, CA). Statistical significance (p < 0.05 was selected as threshold) was tested using paired or unpaired t-test with Welch’s correction or one-way ANOVA with Tukey’s multiple comparisons test, as indicated in Figure legends.
Results
This study was designed to explore the utility of PHH, HepaRG, and iHeps to evaluate the cholestatic potential of drugs and chemicals when cultured in different MPS configurations (Fig. 1 and Table 1). Specifically, we aimed to determine (i) whether different liver cell sources and model combinations demonstrate advantages in terms of hepatocellular function, (ii) the functionality of different liver cells in these models over time, and (iii) if known cholestatic DILI can be replicated in these models, especially under conditions of long-term cell culture (i.e., 30 days). We evaluated three liver MPS – the constant perfusion-based PhysioMimix LC12 and the intermittent flow-based Organoplate 2-lane 96 and 3-lane 40 models—and compared them to a conventional 2D culture method. Our previous studies showed that liver cell functionality generally declines after the first 7 days in culture in both OrganoPlate 2-lane 96 (iHeps) and PhysioMimix LC12 (PHH from different donors) (Kato et al. 2022; Lim et al. 2023); therefore, experiments were generally limited to 7 days. By contrast, previous studies from other laboratories showed that HepaRG maintained robust function in microfluidic conditions for 14 days and longer (Boul et al. 2021; Jang et al. 2019). Therefore, one experiment with HepaRG was extended to 30 days.
For the comparison of basal liver function across cell types and models, we evaluated secretion of albumin and urea, as well as activity of CYP3A4 (Fig. 2). Three liver cell types were compared among different MPS and traditional 96-well plate (2D) culture. In 2D cultures, PHH functional biomarkers (albumin and urea) were similar to those reported in our previous study using the same PHH donor (Lim et al. 2023). This result demonstrates reproducibility of the data across experiments and operators if the same protocol and donor cells are used. HepaRG function was similar to that of PHH for the first 7 days. Interestingly, a drop in albumin and urea secretion by HepaRG was observed in the second week with a rebound in weeks 3 and 4—an effect similar to that previously reported for spheroid cultures of HepaRG (Gunness et al. 2013). The hepatocellular marker secretion and CYP3A4 activity of iHeps was low in comparison to both PHH and HepaRG, commensurate with previous report (Lim et al. 2023).Fig. 2. Comparison of basal function across different cell sources and models. Albumin (A), urea (B) secretion and CYP3A4 activity (C) were tested in 2D (HepaRG, PHH, iHep 2.0), PhysioMimix LC12 (HepaRG and PHH) and OrganoPlate 2-lane 96 and 3-lane 40 (HepaRG and iHep 2.0) over 7 (all cell sources) and 30 days (HepaRG only). Data shown are mean ± SD (n = 3–6 per condition). “n.d.” indicates conditions where levels of urea were below the limit of quantitation. Cell types are labelled as indicated in the legend insets
In PhysioMimix LC12, only PHH and HepaRG were compared because iHeps do not thrive in this type of MPS (Lim et al. 2023). Again, the function of PHH over the 7 days in culture in PhysioMimix LC12 MPS was similar to that reported previously on the same donor (Lim et al. 2023). Albumin levels in HepaRG cells were similar to those of PHH in the first 7 days, then decreased but rebounded by week 3 in culture. The levels were largely similar to those reported previously for HepaRG in either static or perfused models – 10–20 μg/day/10^6^ cells as seen in Fig. 2A (Ehrlich et al. 2019; Gunness et al. 2013; Kopp et al. 2024). Urea production of PHH was ~ 10-times higher than that of HepaRG in the first days of fluidic cultures but the levels equalized by day 5 and HepaRG cells maintained the level of urea production for up to 30 days. CYP3A4 activity also was comparable between cell types in the first week, and remained stable over 30 days in HepaRG, commensurate with previous reports (Duivenvoorde et al. 2021).
In OrganoPlate models, we tested HepaRG and iHeps, PHH do not perform well in this type of MPS (Kato et al. 2022). Experiments were performed over 7 days after plating and the function of both HepaRG and iHeps was generally low regardless of whether cells were cultured in a 2-lane or 3-lane version of the OrganoPlate. These results are similar to previous reports of iHeps cultured in OrganoPlate 2-lane 96 (Kato et al. 2022) and HepaRG cultured in OrganoPlate 3-lane 40 (Jang et al. 2019). Overall, evaluation of these three liver cells sources across models revealed that PHH and HepaRG maintain comparable basal hepatocellular function over 7-day culture in 2D and PhysioMimix LC12 formats.
To compare how well PHH, HepaRG, and iHeps could recapitulate cholestatic liver injury in vitro, we used three compounds from different use categories that are known to cause liver injury via direct or indirect cholestatic mechanisms—BOS, ANIT and 2-Oct. BOS is a known inhibitor of bile salt export pump (BSEP) and other canalicular transporters leading to intrahepatic bile acid accumulation (Fattinger et al. 2001). As human C_max_ of BOS is 7.4 μM (Gutierrez et al. 2013); we tested it at 25 μM which is ~ 3 × C_max_. ANIT is a pesticide that is both a direct cytotoxic agent to cholangiocytes and an inhibitor of bile acid transporters (Yang et al. 2017). There is no data on human C_max_ for ANIT. One study in rats reported a C_max_ of 5 μM after a single oral administration of 100 mg/kg (Jean et al. 1995). Therefore, we tested ANIT at 50 μM, a concentration shown to exert effects on bile acid transporters in vitro while not causing cell death (Orsler et al. 1999). 2-Oct is a flavoring agent not known to have direct effects on the liver, but it has been suggested to be associated with biliary injury in vivo via immune-mediated mechanisms (Amano et al. 2005). There are no studies reporting plasma levels of 2-Oct in either humans or animals. However, our selected concentrations of 70 μM is in the non-cytotoxic range based on previous studies with HepaRG (Gijbels et al. 2021).
Figure 3 shows the effects of BOS (25 μM) on different liver cell sources in 2D and MPS. When cells were treated on days 5 and 6 (Fig. 3A) and biomarkers evaluated on day 7, the most pronounced effect was the increase in CYP3A4 activity in PHH and HepaRG cells cultured in either the 96-well plates or in PhysioMimix LC12. This effect was expected because BOS is a ligand to the human pregnane X receptor (PXR) leading to increased transcription of CYP3A4 (Dingemanse and van Giersbergen 2004). Interestingly, while the basal activity of CYP3A4 towards the end of the 7-day culture was much higher in both PHH and HepaRG cultured in PhysioMimix LC12 than in 2D cultures (Fig. 2), the induction was significant in both models and cell types and was about 2 × greater (6.1 vs 2.5 and 5.1 vs 1.9 fold) in 2D as compared to the MPS in both cell types. No such effect on CYP3A4 was observed in both OrganoPlate models, neither with iHeps nor with HepaRG. No significant cytotoxicity, or effects on albumin or urea synthesis, was observed in 2D or PhysioMimix LC12 MPS. A modest but significant decrease in LDH and albumin release was observed in OrganoPlate 3-lane 40 with HepaRG, but these effects were not found with other liver cells or models and the functional significance of these effects is unknown. In HepaRG cultured over 30 days (Fig. 3B), BOS had effects only on CYP3A4 induction; the response in both 2D and PhysioMimix LC12 was nearly identical to that observed after 7 days in culture.Fig. 3. Assessment of the known in vivo cholestasis inducer Bosentan (BOS, 25 µM) on hepatocyte function across different cell sources and models. A Changes in CYP3A4 enzyme activity, albumin and urea secretion and LDH release were quantified on day 7 following a 48-h daily treatment with vehicle control (0.1% DMSO) and BOS in 2D (96-well Plate, all cell sources), PhysioMimix LC12 (LC12; HepaRG and PHH) and OrganoPlate 2-lane and 3-lane (OP 2L and OP 3L; HepaRG). B Changes in CYP3A4 enzyme activity, albumin and urea secretion and LDH release were quantified on day 30 following a 48-h daily treatment with vehicle control (0.1% DMSO) and BOS in 2D and PhysioMimix LC12 (HepaRG only). Shown are box-and-whiskers plots where vertical line is median and the box is 25th to 75th percentile. The whiskers are min–max values and all individual data points are shown. Numbers indicate fold difference between pairs as indicated by brackets; numbers in blue and bold font indicate significant differences (p < 0.0332) within a pair as determined by one-way ANOVA with Tukey’s multiple comparison test
In the experiments with ANIT (50 μM), we observed effects similar to those with BOS—significant induction of CYP3A4 in the absence of cytotoxicity or effects on synthetic function of liver cells (Fig. 4). CYP3A4 induction was observed only in 2D and PhysioMimix with both PHH and HepaRG, both after 7 (Fig. 4A) and 30 (Fig. 4B) days in culture. No such effect was seen in either OrganoPlate models. While there are no studies of direct effects of ANIT on CYP3A4 or PXR, it has been shown to significantly decrease the expression and function of critical hepatic bile acid transporters (Li et al. 2016) leading to intra-cellular accumulation of bile acids that can induce the expression and activity of CYP3A4 (Minegishi et al. 2024). Figure 5 shows the results of studies with 2-Oct (70 μM). No changes in most biomarkers were observed, with the exception of a significant but minor decrease in CYP3A4 activity in some liver cells and models. There is currently no evidence that 2-Oct induces, inhibits, or is metabolized by CYP3A4. It is also not known to act through the PXR pathway. These results are commensurate with the lack of direct effects on hepatocytes.Fig. 4. Assessment of the known in vivo cholestasis inducer alpha-naphthyl isocyanate (ANIT, 50 µM) on hepatocyte function across different cell sources and models. A Changes in CYP3A4 enzyme activity, albumin and urea secretion and LDH release were quantified on day 7 following a 48-h daily treatment with vehicle control (0.1% DMSO) and ANIT in 2D (96-well Plate, all cell sources), PhysioMimix LC12 (LC12; HepaRG and PHH) and OrganoPlate 2-lane and 3-lane (OP 2L and OP 3L; HepaRG). B Changes in CYP3A4 enzyme activity, albumin and urea secretion and LDH release were quantified on day 30 following a 48-h daily treatment with vehicle control (0.1% DMSO) and ANIT in 2D and PhysioMimix LC12 (HepaRG only). See legend to Fig. 3 for explanation of the graph type and numbersFig. 5Assessment of the known in vivo cholestasis inducer 2-octynoic acid (2-Oct, 70 µM) on hepatocyte function across different cell sources and models. A Changes in CYP3A4 enzyme activity, albumin and urea secretion and LDH release were quantified on day 7 following a 48-h daily treatment with vehicle control (0.1% DMSO) and 2OCT in 2D (96-well Plate, all cell sources), PhysioMimix LC12 (LC12; HepaRG and PHH) and OrganoPlate 2-lane and 3-lane (OP 2L and OP 3L; HepaRG). B Changes in CYP3A4 enzyme activity, albumin and urea secretion and LDH release were quantified on day 30 following a 48-h daily treatment with vehicle control (0.1% DMSO) and 2OCT in 2D and PhysioMimix LC12 (HepaRG only). See legend to Fig. 3 for explanation of the graph type and numbers
To expand on cholestatic readouts in these studies, we measured concentrations of several bile acids in cell culture media in the experiments with PHH and HepaRG cells in 2D and PhysioMimix LC12 MPS at the end of the 7 or 30 day culture (Fig. 6 and Figures S5–S10). We evaluated unmodified, as well as glycine and taurine conjugated, primary bile acids—cholic acid (CA) and chenodeoxycholic acid (CDCA). These bile acids and their conjugated derivates are known to constitute the majority of the bile acids produced by hepatocytes in culture (Einarsson et al. 2000). Figure 6A shows that after 7 days, total bile acid levels were the lowest in PHH cultured in 2D and highest in PHH cultured in PhysioMimix LC12—a difference of 27-fold. HepaRG showed intermediate levels of bile acid production; the levels in PhysioMimix LC12 were greater than those in 2D but the difference was not significant. While the levels of total bile acid production was comparable in PHH and HepaRG cultured in 2D, the profile of individual conjugated bile acids was very different—PHH produced primarily glycine-conjugated bile acids while HepaRG produced taurine-conjugated bile acids (Figures S8–S10), a result similar to previous reports (de Bruijn et al. 2022; Sharanek et al. 2015).Fig. 6. Quantification of total bile acid secretion in cell culture medium under basal and chemical-treated condition in 2D and PhysioMimix LC12. A Basal levels of total bile acids in cell culture medium for HepaRG and PHH cultured in 2D and LC12 after 7 days. B, C Effects of known cholestatic chemicals bosentan (BOS, 25 µM), alpha-naphthyl isocyanate (ANIT, 50 µM) and 2-octynoic acid (2-Oct, 70 µM) on total bile acid levels in cell culture medium in HepaRG (B) or PHH (C) cultures after 7 days of culture. D Basal levels of total bile acids in cell culture medium for HepaRG cultured in 2D and LC12 after 7 or 30 days. E Effects of known cholestatic chemicals bosentan (BOS, 25 µM), alpha-naphthyl isocyanate (ANIT, 50 µM) and 2-octynoic acid (2-Oct, 70 µM) on total bile acid levels in cell culture medium in HepaRG cultures after 30 days. Individual bile acid data are included in Supplemental Files (Figs. S8–S10). Shown are box-and-whiskers plots where vertical line is median and the box is 25th to 75th percentile. The whiskers are min–max values and all individual data points are shown. Asterisks indicate significant differences (*p < 0.0332; ***p < 0.0002; and ****p < 0.0001) within a pair as determined by one-way ANOVA with Tukey’s multiple comparison test
When chemical treatment effects on bile acid secretion were compared, significant effects were observed with both HepaRG (Fig. 6B) and PHH (Fig. 6C), but only in PhysioMimix LC12 MPS. Interestingly, all 3 compounds resulted in a significant (about twofold) decrease in bile acid concentrations in cell culture media in experiments with PHH. By contrast, only ANIT and 2-Oct, but not BOS, resulted in a significant effect in HepaRG. When HepaRG cell cultures of 7 and 30 days were compared (Fig. 6D), it is noteworthy that at both time points the levels in PhysioMimix LC12 were higher than those in 2D, albeit only 7 day differences were significant, as noted above. Total bile acid levels in the medium remained at approximately the same levels over 30 days in culture. When HepaRG were treated with three test agents starting on day 28 (Fig. 1B), trends in bile acid levels in the medium (Fig. 6E) were largely similar to those after 7 days (Fig. 6B). Specifically, in 2D, chemical-treated levels were slightly lower, but not significantly, as compared to controls. In PhysioMimix LC12, all three test compounds decreased bile acid levels in the medium, but only 2-Oct had a significant twofold reduction compared to controls and ANIT did not have a significant effect.
Discussion
PHH remain the gold standard for the most physiologically accurate metabolism and toxicity studies in short-term culture conditions (LeCluyse et al. 2012) and in long-term cultures when assembled into spheroids (Bell et al. 2016). HepaRG, a human hepatic bipotent progenitor cell line (Parent et al. 2004), has been shown to closely mimic hallmarks of hepatocellular function (e.g., signal transduction, metabolic functions) across a range of in vitro tissue culture platforms, particularly in 3D and flow-based systems where their xenobiotic metabolism functions have been found to be comparable to, or even exceed that of PHH for certain endpoints (Nelson et al. 2017; Ramaiahgari et al. 2017). Most comparative studies of PHH and HepaRG have been conducted in traditional 2D culture conditions where PHH are known to lose the majority of their xenobiotic metabolism (Jackson et al. 2016; Smith et al. 2012). Both similarities and differences in gene expression and xenobiotic metabolism and transport functions have been reported between these two cell types (de Bruijn et al. 2022; Duivenvoorde et al. 2021; Gerets et al. 2012; Lubberstedt et al. 2011; Sharanek et al. 2015; Szabo et al. 2013). PHH remain a model of choice for later-stage drug safety evaluations, while HepaRG have become a commonly used model for drug-drug interaction studies and safety studies of non-pharmaceuticals (e.g., genetic toxicology), especially in cases where long-term cultures with proliferative/repair capacity are needed to simulate repeated-dose effects. HepaRG provide advantages with reduced lot-to-lot variability in a consistent genetic background compared to PHH for both liver enzyme induction and metabolism applications (Jackson et al. 2016).
Because liver MPS aim to both increase human relevance and extend the functionality of liver cells in vitro (Gough et al. 2021; Ribeiro et al. 2019), comparative studies of different cell types and models are needed to demonstrate the value and utility of different choices of cell sources for studies of hepatotoxicity, especially for potential cholestatic liver injury (Baudy et al. 2020). The microfluidic-enabled 3D culture conditions have been shown to positively affect both PHH and HepaRG cell behavior in several important ways as compared to static 2D cultures. Most liver MPS use PHH as the closest approximation of human liver tissue, and the advantages of flow-based conditions have been previously reviewed (Gough et al. 2021; Ribeiro et al. 2019). Even though fewer studies evaluated HepaRG cultured under flow, improved stability of liver-like gene expression and functional enzyme activities have been reported over extended periods (up to 8 weeks). For example, more than two-thirds of key genes in static cultured HepaRG changed significantly over time, while under microfluidic conditions HepaRG maintained a more consistent gene expression profile, especially for cytochrome P450 enzymes and transporters, indicating a more robust and reliable model for longer-term in vitro studies of the liver (Duivenvoorde et al. 2021). Still, few studies have evaluated the long-term viability and functionality of various liver cell sources across different MPS, especially those that can be used in medium-throughput format. Accordingly, our study provides direct comparison of PHH, HepaRG and iHeps across several culture modalities and timepoints, as well as in response to treatment with cholestatic agents that act via different mechanisms.
We found that the function of iHeps was low in both 2D and OrganoPlate models, confirming that these cells are not sufficiently differentiated to emulate PHH-like hepatocellular functionality even in MPS and further optimization of their function is needed (Monckton et al. 2022). With respect to the basal function of HepaRG in various culture conditions, our data are highly concordant with previous reports of comparable hepatocellular functional (e.g., albumin and urea) and metabolic (CYP3A4 and bile acid conjugation) markers comparing HepaRG and PHH cultured in traditional 2D format, for both short-term and extended duration (Lubberstedt et al. 2011; Sharanek et al. 2015; Tascher et al. 2019). Albumin production was approximately 10–15 μg/day/1 million cells in both cell types for most of the first week in culture, representing levels approaching lower bounds of that for in vivo human liver (Baudy et al. 2020). In the experiments with PhysioMimix LC12, HepaRG were also comparable with PHH in terms of albumin and urea synthesis, the functionality of both cell types was even greater than that in 2D. However, levels of albumin were about 5-times lower in both OrganoPlate models. Previous studies with HepaRG in dynamic flow MPS reported high and consistent functionality for up to 8 weeks (Darnell et al. 2011). By contrast, one study in OrganoPlate 2- and 3-lane models, where media flow is intermittent, reported a gradual decline in albumin production over 14 days in culture (Jang et al. 2019), which may also be impacted by differences in serum content and cellular composition variations (e.g., donor preparations with varied proportions of non-parenchymal cells). In this study, we used pre-differentiated HepaRG maintained under high DMSO-containing (1.6%) differentiation media while other studies differentiated HepaRGs “on-MPS”. Still, we reason that our data in PhysioMimix LC12 and studies of others in dynamic flow conditions (Boul et al. 2021; Darnell et al. 2011) show that long-lasting functionality of HepaRG can be maintained to a higher level under conditions of constant media perfusion and differentiation. Lower functionality of HepaRG in the OrganoPlate models may be related to limited re-establishment of cell–cell interaction domains in these gel-based culture configurations, or limited perfusability with gel-based platforms that may lower surface-to-media-volume ratio and precludes sufficient nutrient/waste exchange (Mavris and Hansen 2021).
Our data are concordant with previous reports that HepaRG models are largely comparable to PHH with respect to hepatocellular functionality in short-term studies (Gerets et al. 2012). We also found that HepaRG can be used for longer-term cultures, especially in PhysioMimix LC12 where the stability of high drug metabolism function, for both CYP activity and bile acid conjugation, is essential to model chemical-induced disease states (Baudy et al. 2020; Vinken 2018). The latter is especially important for studies of repeated exposure to low doses of consumer use chemicals and their mixtures (Tsatsakis et al. 2016). Indeed, several groups have demonstrated the value of HepaRG in this context (Gijbels et al. 2020, 2019; Tsatsakis et al. 2016; Vilas-Boas et al. 2019, 2020). Still, the need for additional equipment and lower throughput are considerable limitations of using PhysioMimix LC12 as a first-choice liver in vitro model. 2D multi-well plate studies with either PHH or HepaRG are a sensible first-choice model for concentration- and time-course evaluations. Selected concentrations can then be tested in a flow-based system like PhysioMimix LC12 to confirm the findings and achieve even greater physiological relevance.
The data from studies with cholestatic compounds offer additional practical considerations for the choice between HepaRG and PHH in various culture modalities and decision contexts. A growing body of evidence indicates that a variety of consumer use chemicals and food additives can act as potential causes of cholestasis and can result in liver injury (Gijbels et al. 2021; Vilas-Boas et al. 2019, 2020). In addition, substantial clinical and experimental data demonstrate links between cholestasis and metabolic-associated fatty liver disease (MAFLD), mediated by shared disruptions in bile acid metabolism and signalling, particularly via FXR and other bile acid-regulated pathways (Evangelakos et al. 2021). MAFLD incidence is increasing worldwide and possible associations with diet, lifestyle, and chronic low-level exposure to various chemicals (Fritsche et al. 2023). Thus, in vitro liver models that can be used to study chemicals that may lead to cholestasis are needed (Vinken 2018). It has been suggested that HepaRG are capable of addressing the challenges of modelling drug-induced cholestatic liver injury due to their hepatoblast-like capacity to differentiate into both hepatocyte-like cells and cholangiocyte-like cells depending on the culture conditions applied (Parent et al. 2004). Hepatocyte metabolism/transport is a major driver of bile acid synthesis from cholesterol and their conjugation to various forms for transporter-mediated secretion, while cholangiocytes are critical for recycling bile acids to maintain biliary flow via re-absorption and are known to shift to more proliferative states during cholestatic liver injury in repair damaged biliary epithelia (Chiang and Ferrell 2018).
In this study, we found that both PHH and HepaRG cultured in 2D and PhysioMimix LC12 can be used for detection of cholestasis in short-term cultures (7 days). In addition, HepaRG cultured in PhysioMimix LC12 can be extended for longer-term studies (30 days) of cholestatic hazard. However, we found that PHH showed higher sensitivity than HepaRG when toxicity studies were performed, especially in terms of bile acid secretion. We observed CYP3A4 activity and secreted bile acids were the most informative biomarkers related to cholestatic liver injury that can arise through different mechanisms. PHH and HepaRG cells cultured in 2D or PhysioMimix LC12 responded to treatment with BOS and ANIT with marked increases in CYP3A4 activity. Indeed, these chemicals are known to act by dysregulating bile acid transporters and related nuclear receptor activity (Halilbasic et al. 2013; Wu et al. 2021). By contrast, 2-Oct did not affect CYP3A4, commensurate with its known effects on cholangiocytes, not hepatocytes. Importantly, because HepaRG cultures evaluated in this study contain cholangiocyte-like cells, this hybrid cell model did demonstrate a response to 2-Oct when bile acid content in the media was used as a biomarker. Taken together, these data indicate that in future studies it is important to evaluate both changes in metabolic pathways and secreted bile acids. It has been shown that secreted bile acids exceed that of intracellular bile acids (9:1) in both PHH and HepaRG (Sharanek et al. 2015). Therefore, sampling of cell culture medium for bile acid analysis is a sensitive and practical biomarker where cells can be undisturbed for prolonged cultures, especially in studies using PhysioMimix LC12 and other liver MPS where serial sampling can be performed easily due to much larger medium volume than in 2D plates. Still, it is worth noting that the bile acid profiles of PHH and HepaRG differed considerably, especially for conjugated isoforms that are most abundant in liver cell cultures (de Bruijn et al. 2022; Sharanek et al. 2015). Glycine conjugates were most abundant in PHH cultures, consistent with that in human liver, while HepaRG mainly produced taurine conjugates. Previous studies with HepaRG showed that cofactor supply may impact the conjugation pathway preference (de Bruijn et al. 2022; Sharanek et al. 2015). This difference also may be related to the presence of much higher concentrations of DMSO in HepaRG cultures as compared to PHH, a condition that may inhibit some enzymes. Further optimization of HepaRG culture conditions to balance the proportions of hepatocyte- and cholangiocyte-like cells with bile acid metabolism and secretion profiles is warranted. Still, our data show that changes in the total bile acid content of cell culture medium serves as a robust later-stage biomarker of the potential cholestatic hazard of test chemicals in both PHH and HepaRG.
It is also noteworthy that while 2D cultures of both PHH and HepaRG showed relatively low basal CYP3A4 activity in 7-day experiments, the inducibility of this biomarker by both BOS and ANIT was not only concordant with that in PhysioMimix LC12 cultures, but the fold change was about twice that in PhysioMimix LC12. Long-term experiments in HepaRG yielded the same outcome. This finding demonstrates the higher potential sensitivity of 2D cultures to the chemical-induced CYP induction. With greater magnitude of response, this model may be more appropriate for medium- to high-throughput screening and the lower basal activity should not be interpreted as a flaw of the model, but rather a reflection of zone-2 liver metabolic function that retains further inducibility. Still, effects on the bile acid release from cells into the medium, in both 7 and 30 day-long experiments, were observed only in the PhysioMimix LC12, suggesting that a dynamic flow-enabled MPS is important for detection of indirect-acting cholestatic substances.
While this study offers a detailed comparison of PHH, HepaRG, and iHeps in both traditional 2D cultures and three different liver MPS, several limitations should be noted. First, the reliance on commercially sourced, pre-differentiated cell sources means that inter-individual variability was not represented in our study. The variability can be modeled with PHH from different donors (Niemeijer et al. 2024), but HepaRG will remain a single-individual model. Second, although the study tested both short-term and extended culture durations, some endpoints, such as albumin and bile acid secretion, showed temporal fluctuations that may confound interpretation of long-term toxicity or adaptation. Third, while the comparison of 2D versus MPS cultures was extensive, the evaluation of additional liver cell types or co-culture systems (e.g., inclusion of non-parenchymal cells or immune components) was not performed, potentially limiting the physiological relevance of the tested models, particularly for compounds like 2-Oct that may act via indirect immune-mediated mechanisms. Fourth, the study investigated only a few well-characterized cholestatic agents (bosentan, ANIT, 2-Oct) at a single high concentration, so results should be interpreted with caution and cannot be generalized to all types of cholestatic injury at various stages of progression. Fifth, HepaRG were only cultured under fully differentiating conditions with high (i.e., > 1.6%) DMSO, which may have a baseline effect on cholesterol and bile acid metabolism and transporter functionality that limits detection of earlier-stage cholestatic liver injury under these conditions. Finally, the primary assessment of cholestatic potential relied on biochemical biomarkers (CYP3A4 activity and bile acid secretion) and did not incorporate -omics approaches or functional readouts such as transporter expression or nuclear receptor activity, limiting mechanistic interpretability.
In summary, this study demonstrated that both PHH and HepaRG, when cultured under advanced (PhysioMimix LC12 MPS) and conventional (2D) conditions, are demonstrating comparable basal hepatocellular function that is superior to that of iHeps. Both PHH and HepaRG may be used as in vitro models for detecting and characterizing cholestatic hazard posed by xenobiotics; however, PHH showed more consistent responses as compared to HepaRG. We also show that HepaRG cells exhibited sustained function over 30 days when cultured in PhysioMimix LC12 MPS and were able to respond to cholestatic compounds with a decrease in bile acid secretion, albeit low throughput of the MPS model and high variability resulted in a significant effect only for 2-Oct. Overall, this research shows that total secreted bile acid levels and CYP3A4 activity are sensitive and informative later-stage biomarkers for detection of liver injury potential and advances the utility of next-generation in vitro liver models as predictive tools for regulatory science (Carmichael et al. 2022).
Supplementary Information
Below is the link to the electronic supplementary material.
Supplementary Material 1