Improving hepatic metabolic function with 3D iPSC-derived human hepatocytes

Abstract

Human induced pluripotent stem cells (iPSCs) have the potential to significantly advance drug discovery and clinical applications, particularly in personalized medicine. Synthetic biocompatible materials improve the reproducibility and efficiency of in vitro protocols and enable three-dimensional (3D) cell culture. 3D spheroids and organoids are known to promote hepatic maturation and improve overall liver function. In this study, 3D hepatocyte-like cells (3D HLCs) were differentiated from human iPSCs using a synthetic polypeptide hydrogel scaffold. Their metabolic capacity, including drug biotransformation and mitochondrial function, was assessed using LC–MS semi-targeted metabolomics, providing functional characterization beyond marker expression levels. Compared with conventional monolayer (2D) cultures, 3D HLCs displayed a more mature hepatic phenotype. They acquired xenobiotic metabolism capacity earlier in differentiation, at day 30 rather than day 40 in 2D HLCs. In addition, 3D HLCs showed increased CYP3A4 activity (> 20-fold) and CYP2D6 activity (2–11-fold), as well as enhanced mitochondrial function, including higher β-oxidation, relative to 2D HLCs. The cytotoxic effects of eight drug-induced liver injury (DILI) compounds—amiodarone, azathioprine, cyclosporine A, diclofenac, nefazodone, omeprazole, rosiglitazone, and troglitazone—were also evaluated in 2D and 3D HLCs and compared with 2D primary human hepatocytes (2D PHHs). Drug exposure caused lower cell viability in 3D than in 2D HLCs, more closely resembling the response observed in 2D PHHs. Overall, this study demonstrates the applicability of 3D HLCs for metabolism and toxicology studies, with potential value for personalized drug testing and disease modeling.

Supplementary information

The online version contains supplementary material available at 10.1007/s10565-026-10203-1.

Introduction

Induced pluripotent stem cells (iPSCs) have gained increasing attention in recent years as a promising tool in personalized cell therapies and disease modelling (Hannoun et al. 2016; Kozyra et al. 2018; Mukhopadhyay et al. 2024). Patient-specific cells can be generated through differentiation protocols, while preserving the patient's genetic background and epigenetic profile (Mukhopadhyay et al. 2024). iPSCs derived from patients with liver conditions, e.g., liver cirrhosis and steatosis (Hannoun et al. 2016; Kozyra et al. 2018; Mukhopadhyay et al. 2024) can be utilized to understand pathologies, molecular processes, drug mechanism-of-action, toxicity, and applied in personalized drug therapies (Hannoun et al. 2016). However, despite their promise, differentiation protocols applied to iPSCs frequently result in immature cells that retain embryonic characteristics (Boon et al. 2020; Choudhury et al. 2017). This challenge is particularly evident in hepatocyte-like cells (HLCs) derived from iPSCs, where current differentiation methods typically produce cells with reduced metabolic activity compared to liver tissue or the in vitro gold-standard primary human hepatocytes (PHHs) (Boon et al. 2020; Choudhury et al. 2017; Gupta et al. 2021).

The liver, as a major metabolic hub of the body, handles crucial endogenous metabolic functions, such as glycolysis, β-oxidation, lipogenesis, bile acid, glycogen, and heme biosynthesis (Morio et al. 2021). Many of these pathways depend on mitochondrial function. Oxidative phosphorylation (OXPHOS) and tricarboxylic acid (TCA) cycle activity are key processes in promoting cellular respiration and bioenergetics, with links to anabolic pathways (Martínez-Reyes and Chandel 2020). Furthermore, the liver is also the key organ in xenobiotic detoxification (Guengerich 2021). To handle a wide variety of chemical insults, the liver is supported by the presence of enzyme superfamilies like the cytochrome P450s (CYPs), with a broad substrate specificity (Guengerich 2021). The complexity and interplay of these hepatic functions make the characterization of metabolic activity—and their potential improvement—essential for the broader use of iPSC-derived HLCs. While most studies assess HLCs’ liver function through a handful of markers, by gene expression or protein levels, metabolic activity readouts would benefit from more comprehensive strategies. Omics approaches, in particular metabolomics, can capture both endogenous and xenobiotic metabolism (Bowen et al. 2023; Moco and Buescher 2023).

We and others have utilized iPSC-derived HLCs, produced through a 40-day differentiation protocol (Boon et al. 2020; Pozo Garcia et al. 2025). The obtained HLCs in monolayer (2D) showed a higher metabolic performance compared to the often-used hepatic cell lines (e.g., HepG2), with enhanced mitochondrial function and matured phase I and phase II drug metabolism (Boon et al. 2020; Pozo Garcia et al. 2025). Activity of 6 different CYP isoforms (CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2D6, and CYP3A4) was identified in these 2D HLCs, potentially handling drugs of diverse chemistry (Pozo Garcia et al. 2025).

Three-dimensional (3D) hepatic organoids have been reported to outperform classical 2D cell systems in their representation of tissue-specific architecture, cell–cell communications, and overall promotion of liver function (Kim et al. 2022; Pampaloni et al. 2007). 3D formation can be generated using both scaffold-based (e.g., hydrogel-based support) or scaffold-free techniques (e.g., hanging drop) (Jensen and Teng 2020). Recently, Kiamehr et al. introduced a novel method for generating 3D organoids, termed “organobodies”, which consist of HLCs encapsulated within a self-assembling peptide hydrogel scaffold (Kiamehr et al. 2026), adapted from the previously tested 2D differentiation procedure (Boon et al. 2020). The use of defined and fully synthetic fibrous polymers and gels of self-assembling oligopeptides is a promising alternative to animal-derived extracellular matrices (ECMs). Synthetic materials can improve culture quality control and batch-to-batch reproducibility, in addition to providing a structural support for cell aggregates (Goudarzi et al. 2024; Pampaloni et al. 2007). Furthermore, the nanometer-sized fibers and pores within these scaffolds offer a suitable 3D environment for the cell, while allowing easy exchange of soluble small molecules and growth factors (Jensen and Teng 2020; Pampaloni et al. 2007).

An increase of hepatic maturation in 3D hepatic organobodies compared to 2D HLCs was reported through transcriptomics, including expression of hepatic markers, showcasing the suitability of 3D HLCs for toxicity screening (Kiamehr et al. 2026). However, a comprehensive characterization of the metabolic function of such 3D HLC structures has yet to be performed, including the development of protocols enabling metabolic analyses in these formats (3D) (Langhans 2018). In this study, we assessed metabolic function, including mitochondrial, central carbon, and xenobiotic metabolism, of 3D hepatic organobodies. We compared their performance to 2D HLCs, in predicting cytotoxicity in response to drug-induced liver injury (DILI) drugs, and in response to mitochondrial inhibitors. With this study, we aimed to highlight the potential of using 3D cell structures in metabolic research, of potential use in disease modeling, drug discovery, and toxicity testing.

Materials and methods

Materials were listed in the Supplementary Information (SI) Tables S1−11.

2D cell culture

All cell systems were cultured in an incubator at 37 °C under a humidified atmosphere with 5% CO₂.

iPSCs culture and differentiation to HLCs (hepatocyte-like cells)

Human induced pluripotent stem cells (iPSCs) used in this study (SBAD2, STBCi321-A, https://www.cellosaurus.org/CVCL_ZX54) were generated during the IMI-funded StemBANCC project using the CytoTune 2.0 (ThermoFisher) Sendai viral reprogramming kit from normal dermal fibroblasts (Morrison et al. 2015). SBAD2-3X iPSC line was generated with a donor vector containing Hepatocyte Nuclear Factor 1 Alpha (HNF1A), Forkhead Box A3 (FOXA3), and Prospero Homeobox Protein (PROX1), through a doxycycline inducible cassette as described elsewhere (Ghosh et al. 2021). The SBAD2-3X cell line was cultured and differentiated to HLCs as previously reported (Pozo Garcia et al. 2025).

HepG2 cells

Human hepatoma HepG2 were purchased from European Collection of Authenticated Cell Cultures (ECACC, UK Health Security Agency). Cells were cultured in DMEM media with low glucose (5.5 mM) and pyruvate (1 mM), 1% (v/v) penicillin/streptomycin, and 9% (v/v) FBS. Cells were detached with trypsin and seeded with a density of 45 k cells/cm². Cells were used 48 h after seeding.

Primary human hepatocytes (PHHs)

A mixed gender human hepatocyte pool from 10 donors was purchased from BioIVT, US (product number: X008001-P, lot: SWY) and handled according to the manufacturer’s specifications. Cells were thawed using the OptiThaw Hepatocyte Medium (BioIVT) on Collagen I coated plates. The cell density was 145 k cells/cm² in 6-well plate format and 220 k cells/cm² in 96-well plate format. Six hours after plating, media was changed from hepatocyte plating media to maintenance media. Cells were treated 24 h after plating and kept in culture for no longer than 48 h.

Preparation of 3D organobodies

3D self-assembled peptide hydrogel HLC organobody formation followed a described protocol (Kiamehr et al. 2026). Cells were cultured and differentiated according to the 2D HLCs protocol (Pozo Garcia et al. 2025) until day 8 of differentiation. At day 8, cells cultured in a 10 cm² dish were washed twice with PBS and detached with 3 mL TrypLE Express for 8 min at 37 °C. Detached cells were added to 10 mL pre-warmed LDM media (HLC culture media) containing 10% (v/v) FBS, then centrifuged (5 min, 300 g). The supernatant was discarded, and the cell pellet was recovered in 3 mL 10% (m/v) sucrose/water. After another centrifugation step, cell pellets were resuspended in 10% (m/v) sucrose/water to a final density of 50 k cells/µL. Cell suspension was then mixed with equal parts (v/v) of a gel emulsion containing 5% rat tail collagen, 10% laminin, 20% of 20% (m/v) sucrose/water, and 65% PuraMatrix™ peptide hydrogel (v/v). Cells suspended in gel were pipetted up and down, avoiding air bubbles, for 2 min. The obtained homogeneous cell-gel mixture was dispensed (3 µL per well) to a CellCarrier Spheroid ULA 96-well Microplate containing LDM media with 0.02 µg/mL FGF1, 0.6% (v/v) DMSO, 5 µg/mL doxycycline, and 1:100 RevitaCell™ supplement (100X). After 24 h, RevitaCell™ was removed, and cells continued to differentiate for at least 22 more days, using the media described in the 2D HLCs culture protocol (Pozo Garcia et al. 2025). Feeding volume was 100 µL per well, and each 3D organobody contained ~ 70 k cells.

To prepare 3D hydrogel organobodies from HepG2 cells, the same procedure was used as for the HLCs, with the following differences: cells were detached with trypsin, and HepG2 culture media (mentioned above) was used for the centrifugation, plating, and culture. HepG2 organobodies (~ 70 k cells) were used 48 h after formation.

Immunofluorescence

2D and 3D HLC differentiation was assessed by measuring three specific liver markers using immunofluorescence at days 30 and 40 of differentiation: CYP3A4, albumin, and phosphoenolpyruvate carboxykinase (PEPCK). Cell fixation and antibody staining were conducted as previously described (Meijer et al. 2024) with some modifications. The primary antibody was incubated overnight for 2D HLCs and for 48 h in the case of 3D HLCs, while the secondary antibody for 35 min and 2 h, respectively. The nucleus was stained with Hoechst 33342, and the secondary antibody for all four primary antibodies was Alexa 647 donkey-anti-rabbit IgG. 2D cells were stained in PhenoPlate 96-well microplates (PerkinElmer, Cat. 6055308), while 3D structures in 96-well High Content Screening microplates (Coning, Cat. 4517). The cells were imaged using a 63 × water immersion objective on the Operetta CLS High-Content Imager (PerkinElmer) with confocal imaging and Z-stacks. Image analysis was performed with the Harmony 5.2.2 software. Autofluorescence was excluded, using only a secondary antibody. List of primary and secondary antibodies can be found in Table S6.

CYP metabolism characterization

3D HLCs at day 30 and 40 of the differentiation were characterized for their xenobiotic metabolism, using independent treatment of 6 known metabolic CYP probes: rosiglitazone (10 µM), dextromethorphan (15 µM), midazolam (15 µM), benzydamine (15 µM), bupropion (25 µM), and coumarin (250 µM) (Pozo Garcia et al. 2025). These drugs were selected for their CYP isoform specificity (Pozo Garcia et al. 2025). Drug solutions were prepared in DMSO, with a final DMSO concentration of 0.2% (v/v) in all incubations. For drug treatment, two organobodies were pooled in a well of a CellCarrier Spheroid ULA 96-well Microplate, exposed to 100 µL. After 24 h incubation, 70 µL of the exposed media (extracellular content) was sampled and quenched in 900 µL 80% (v/v) methanol/water with the internal standard (IS), 2 µM meloxicam (80% MeOH + IS) (Pozo Garcia et al. 2025). Control samples (cells incubated with 0.2% (v/v) DMSO) and blank samples (drug stability without cells) were collected in parallel.

2D and 3D HLCs at day 40 of their differentiation were compared for their xenobiotic metabolism using phenacetin (30 µM), rosiglitazone (10 µM), diclofenac (75 µM), dextromethorphan (15 µM), chlorzoxazone (60 µM), midazolam (15 µM), benzydamine (15 µM), bupropion (25 µM), coumarin (250 µM), and 7-ethoxycoumarin (60 µM). For the 2D HLCs, six-well plates were used (2 mL/well), while for the 3D HLCs, five organobodies were pooled into a single well of a 24-well plate and exposed to 600 µL volume on the day of the experiment. For both 2D and 3D HLCs, aliquots of exposed media (100 µL) were taken at 24 h, 48 h, and 72 h after drug administration (extracellular content) and immediately quenched in 900 µL 80% MeOH + IS (Pozo Garcia et al. 2025).

Extracellular contents, previously quenched in 80% MeOH + IS, were extracted for 30 min under agitation at 4 °C. Samples were freeze-dried and dissolved in 100 µL (3D day 30 and day 40 comparison) or 200 µL (2D and 3D comparison) of 50% (v/v) MeOH/water. The extracts were centrifuged (5 min) and transferred into vials for LC–MS analysis.

Mitochondrial metabolism characterization

2D PHHs, 2D and 3D HepG2, and 2D and 3D HLCs at day 30 of differentiation were incubated for 24 h with known inhibitors of central carbon metabolism: 10 mM 2-deoxy-D-glucose, 30 µM etomoxir, 10 µM UK 5099, 100 µM 3-nitropropionic acid, and 5 nM rotenone. 2D models were treated in 6 well plates (2 mL/well) while for 3D models, 7 organobodies were pooled in one well of a 24 well plate (600 µL/well). All conditions had 0.1% (v/v) DMSO final concentration. After 24 h exposure, the media was discarded, cells were washed with 0.9% (m/v) NaCl, and were immediately quenched in liquid nitrogen. Intracellular contents were extracted by scraping the cells with 1.5 mL of 80% (v/v) MeOH/water with ¹³C internal standard (IS, ¹³C-yeast extract, produced in-house). For 3D samples (3D HepG2 and 3D HLCs), extracts were sonicated for 10 min in an ultrasonic bath with ice and vortexed twice for 1 min. 2D and 3D samples were incubated at 4 °C for 30 min in a shaking block, following a centrifugation step (15 min), before recovering the supernatants for overnight evaporation in a vacuum centrifuge. After reconstituting the lysates in 100 µL 60% (v/v) acetonitrile/water, a final centrifugation was performed (5 min), and the supernatants were immediately taken for LC–MS analysis.

Xenobiotic metabolomics analysis

A previously reported LC–MS method was used, detailed in SI (Pozo Garcia et al. 2025). In short, sample acquisition was performed with a UHPLC (Agilent 1290 UHPLC) coupled to a high-resolution time-of-flight mass spectrometer (Agilent 6230B TOF), using reversed-phase chromatography. Putative metabolite identification was proposed based on accurate mass (< 5 ppm mass accuracy) and relative retention time (based on the analyte’s logP). The metabolic probes (drugs) in this study were annotated with a metabolite identification confidence level 1, since authentic standards were acquired together with all samples, while putative metabolites were annotated with a confidence level 2 (Malinowska and Viant 2019), with the exception of 7-hydroxycoumarin, confirmed with authentic standards. Analytical properties of identified drugs and metabolites can be found in Table S12. Drug and metabolite intensities found extracellularly at the different time points were normalized by IS intensity and divided by the intensity of the respective parent drug in the blank samples (at the corresponding time point), normalized by IS using equation (1). The ratio of metabolite intensities to the parent compound in the blank sample accounts for potential drug degradation and matrix effects.

$$\text{Normalized}\text{Intensity}\frac{(\frac{MetaboliteIntensityinMedia\left(T,i,m,e,,p,o,i,n,t\right)}{\text{Internal}\text{Standard}\left(\text{IS}\right)})}{\frac{\text{Parent}\text{Compound}\text{Intensity}\text{in}\text{Blank}(\text{Time}\text{point})}{\text{Internal}\text{Standard}\left(\text{IS}\right)}}÷\mu gDNA$$ 1

For total DNA quantification, the kit Quant-iT™ PicoGreen™ dsDNA Assay Kit and dsDNA Reagent (Thermo Fisher, Cat. P7589) was used, following the manufacturer’s instructions.

Central carbon metabolome analysis

Sample acquisition was performed using a UHPLC (Agilent 1290 UHPLC), coupled to a SCIEX ZenoTOF 7600 system with a heated electrospray ionization source (ESI), using hydrophilic interaction chromatography, reported in (Pozo Garcia et al. 2026). Details and settings of the LC–MS method can be found in SI. List and analytical properties of the measured metabolites are displayed in Table S13. For the majority of the identified metabolites, authentic standards were available. For the remaining metabolites, identification was performed using accurate mass (with a tolerance of < 5 ppm). Data integration was done using SCIEX OS V.3.1.6 software with the AutoPeak algorithm. Data was represented as ratio to control conditions $\frac{IntensityExperimentalCondition/IS}{IntensityControlCondition/IS}$ for each model.

Cytotoxicity measurements

2D PHHs, 2D HepG2, 2D HLCs, and 3D HLCs were exposed for 24 h to 8 DILI drugs at three different concentrations. 2D and 3D HLCs were used at day 40 of differentiation. Cell exposure and viability tests were performed in 96 well plates (1 organobody per well). The drugs and concentrations, low, medium and high, used for the treatment were as follows: amiodarone (7 µM, 14 µM, 28 µM); azathioprine (2.5 µM, 5 µM, 10 µM); cyclosporine A (5 µM, 10 µM, 20 µM); diclofenac (150 µM, 300 µM, 600 µM); nefazodone (9.5 µM, 19 µM, 38 µM); omeprazole (210 µM, 420 µM, 840 µM), rosiglitazone (117.5 µM, 235 µM, 470 µM) and troglitazone (25 µM, 50 µM, 100 µM). All stocks were prepared in DMSO, and all conditions had a final DMSO concentration of 0.5% (v/v). After 24 h of exposure, cell viability was assessed by three different methods: resazurin, lactate, and calcein AM assays, using a CLARIOstar Plus (BMG LabTech). Toxicity was regarded as a statistically significant reduction (using an unpaired t-test with p-value < 0.05) in cell viability relative to the control condition following drug exposure.

Resazurin assay was performed as described elsewhere (Limonciel et al. 2011). After 24 h of treatment, cells were exposed to a final concentration of 44 µM resazurin in cell media. For 2D cell systems, 1.5–2 h resazurin incubation was performed with a volume of 100 µL/well, while the 3D HLCs were incubated for 4 h with 50 µL/well. Fluorescence resorufin in the extracellular media was measured at 540 nm excitation and 590 nm emission.

Lactate assay was performed as previously described (Limonciel et al. 2011). After 24 h treatment, exposed media was collected, and lactate concentrations were measured using a colorimetric assay, based on the conversion of lactate to pyruvate by lactate dehydrogenase (LDH) activity, measured by reducing p-iodonitrotetrazolium violet (INT) to INTH (Limonciel et al. 2011). Drug exposed media (2D models: 5–10 µL; 3D models: 30 µL) was incubated with 90 µL of assay solution for 5–10 min, at room temperature (Limonciel et al. 2011). Absorbance was measured at 490 nm.

Calcein AM assay was conducted by exposing the cells to 1 µM acetoxymethyl (AM) calcein in cell media (1–2 h for 2D models and 4 h for 3D models), after 24 h of drug treatment. Posterior to dye incubation, cells were washed twice with PBS, and calcein AM uptake and cleavage to calcein was measured intracellularly at 485 nm excitation and 535 nm emission (Sitte et al. 2023).

Statistical analyses

Statistical analyses were conducted in R, the statistical language, version 4.3 for processing and visualizing experimental data. Metabolite intensities in LC–MS data were normalized using min–max scaling, representing values from 0 to 1 (Choudhury et al. 2020). Principal component analyses (PCAs) were done on centered and scaled LC–MS data using the ggplot package. Barplots were generated using GraphPad Prism 8.0.1, and statistics were performed with an unpaired t-test or using a one-sample t-test with a hypothetical mean of 1, considering significance for p-value < 0.05.

Results

In the present study, we prepared 3D HLC organobodies, embedded in scaffolds of self-assembled peptide hydrogel based on a proposed protocol (Kiamehr et al. 2026) (Fig. 1A), for characterising their metabolic function at large. Both xenobiotic and mitochondrial metabolism of such 3D HLC structures was examined.

Fig. 1.

2D and 3D HLCs express markers of hepatic differentiation and xenobiotic metabolism compared to 2D HLCs. A. 3D HLCs formation. B-E. Immunostaining of CYP3A4 (B-C) and albumin (D-E) in 2D HLCs (B, D) and 3D HLCs (C, E) at day 30 and day 40 of differentiation. Cells were imaged using a 63 × water immersion objective with confocal imaging, using a scale of 50 µm (2D) and 200 µm (3D). Albumin and CYP3A4 are represented in red, and nuclear staining using Hoechst 33342 is in blue. 3D images combined several image fields to report one organobody structure. Intensity contrast was matched between the two time points, allowing to compare image intensities within each cell model. F. CYP2A6, CYP2B6, CYP2C8/9, CYP2D6, CYP3A4, and FMO1/3 activity of 3D HLCs day 30 (blue) and day 40 (red) measured by LC–MS. Metabolite LC–MS intensity was normalized by the intensity of the parent compound (drug) in the blank sample and by total DNA (µg), N = 3,4 (individual wells from the same differentiation batch), average ± SEM. Statistics were performed using an unpaired t-test, with significance for p-value < 0.05 (*)

3D HLCs xenobiotic metabolism machinery matures earlier compared to 2D HLCs

2D HLCs generated following a 40-day differentiation protocol exhibited mature hepatic metabolism and activity of CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2D6, and CYP3A4, as previously reported (Boon et al. 2020; Pozo Garcia et al. 2025). However, a differentiation time of 40 days proved to be essential for 2D HLCs to acquire drug metabolic activity, as at 30 days HLCs only negligible CYP3A4 activity was detected (Boon et al. 2020; Pozo Garcia et al. 2025). With the hypothesis that 3D structures accelerate HLCs' maturation, an earlier differentiation time point (day 30) was tested.

To assess hepatic maturation, the expression of the hepatic markers CYP3A4 and albumin was tested by immunofluorescence at days 30 and 40 of differentiation in both 2D and 3D HLCs. In line with previous findings (Pozo Garcia et al. 2025), the expression of both liver markers increased from day 30 to day 40 2D HLCs, demonstrating further maturation at the latter time point (Fig. 1B,D). However, in 3D HLCs, both time points (30 and 40 days) showed equivalent expression of CYP3A4 and albumin, Fig. 1C,E. Corroborating these findings, the hepatic marker phosphoenolpyruvate carboxykinase (PEPCK) was also found to be present in days 30 and 40 in 3D HLCs, Fig. S1.

To complement the immunofluorescence results, we performed LC–MS metabolomics to assess CYP activity of 3D HLCs on days 30 and 40, by incubating the cells with CYP-probes. The activity of the drug-relevant isoforms CYP2A6 (coumarin), CYP2B6 (bupropion), CYP2C8/9 (rosiglitazone), CYP2D6 (dextromethorphan), CYP3A4 (midazolam), and flavin monooxygenases (FMOs, benzydamine) was assessed by the presence of their specific drug metabolites (Guengerich 2021, 2005; Pozo Garcia et al. 2025). Details of the measured putative metabolites can be found in Table S12. Both times points, day 30 and day 40, led to activity of the tested CYP isoenzymes (activities of CYP2A6, CYP2B6, CYP2C8/9, CYP2D6, CYP3A4, and FMOs) in 3D HLCs. Higher activity of CYP2B6, CYP2D6 and FMOs were found at day 40 3D HLCs. A comparable activity between the two time points was found for CYP3A4, demonstrated by dihydroxymidazolam levels, Fig. 1F. Additionally, similar activity of CYP2A6 and CYP2C8/9 was found for day 30 and day 40 3D HLCs.

3D HLCs have higher CYP3A4 and CYP2D6 activity compared to 2D HLCs

We further examined whether the 3D HLC organobodies presented advantages for their use in drug metabolism studies compared to 2D HLCs. For this, the activity of two main CYP isoforms CYP3A4 and CYP2D6 was studied. CYP3A4 is reported to metabolize 52% of drugs of diverse chemistry, while CYP2D6 is known to metabolize 20% of drugs, among these many antidepressants and antipsychotics (Guengerich 2021; Nahid and Johnson 2022). Day 40 2D and 3D HLCs were incubated with midazolam (CYP3A4 metabolism, Fig. 2A) and dextromethorphan (CYP3A4 and CYP2D6 metabolism, Fig. 2B). Enzyme activity was tested over three time points (24, 48, and 72 h). Dihydroxymidazolam and methoxymorphinan, products of midazolam and dextromethorphan, respectively, catalyzed by CYP3A4, were significantly (> 20 fold) higher and increased over time, in 3D HLCs compared to 2D HLCs (Fig. 2C). Surprisingly, hydroxymidazolam was not found in 3D HLCs, indicating a fast CYP3A4/5 conversion to form dihydroxymidazolam, leaving the intermediate metabolite below detection levels. Similarly, dextrorphan, a product of dextromethorphan by CYP2D6, had a higher abundance (2–11 fold) in 3D HLCs compared to 2D (Fig. 2D). Higher CYP3A4 and CYP2D6 activity in 3D HLCs represents a significant benefit over 2D HLCs in drug metabolism studies. This is in line with the higher CYP expression observed in a recent study, using the same 3D protocol (Kiamehr et al. 2026) in 3D HLCs when compared to 2D at both day 20 and day 40.

Fig. 2.

3D HLCs have higher CYP3A4 and CYP2D6 activity compared to 2D HLCs. A. Scheme of midazolam metabolism. B. Scheme of dextromethorphan metabolism. C. CYP3A4 activity of day 40 2D HLCs (blue) and day 40 3D HLCs (red) over 24, 48, and 78 h, by LC–MS measurement of dihydroxymidazolam, product of midazolam, and methoxymorphinan, product of dextromethorphan. D. CYP2D6 activity of day 40 2D HLCs (blue) and day 40 3D HLCs (red) over 24, 48, and 78 h, by LC–MS measurement of dextrorphan, product of dextromethorphan. Metabolite LC–MS intensity was normalized by the intensity of the parent compound (drug) in the blank sample and by total DNA (µg), N = 3,4 (individual wells from the same differentiation batch), average ± SEM. Statistics were performed using an unpaired t-test, with significance for p-value < 0.05*

A more comprehensive comparison of 2D and 3D metabolic activities involved testing other drug-relevant CYP isoforms, CYP1A2 (phenacetin), CYP2A6 (coumarin), CYP2B6 (bupropion), CYP2C8/9 (rosiglitazone), CYP2C9 (diclofenac), CYP2E1 (chlorzoxazone), and enzymes (FMOs, and combined phase I and II metabolism), by LC–MS-based xenobiotic metabolomics (Guengerich 2021, 2005; Pozo Garcia et al. 2025). When considering all possible drug biotransformations, 2D HLCs exhibited higher metabolite coverage, as metabolites of all drugs were detected, except for phenacetin (CYP1A2 activity). Twenty-three phase I metabolites were identified in 2D HLCs (Table S12). In contrast, 11 phase I metabolites were identified in the 3D HLCs, with the absence of CYP1A2 and CYP2E1 activity (Table S12). The lower coverage of phase I activity by the 3D HLCs model is likely due to the use of less biomass in the assay, resulting in metabolites below the LC–MS detection limit.

3D HLCs and 3D HepG2 have improved mitochondrial function

To further examine the metabolic characteristics of HLCs in 3D, semi-targeted metabolomics covering intermediates of central carbon metabolism (CCM) was employed to evaluate the cells’ mitochondrial function. Hepatocytes are highly oxidative cells, active in fatty acid β-oxidation, and comprise complex mitochondrial networks (Morio et al. 2021; Wanet et al. 2014). To characterize mitochondrial activity, cells were exposed to inhibitors that block various pathways of the CCM, thereby affecting mitochondrial function and bioenergetics. Here, 2D PHHs, 2D HepG2, day 30 2D HLCs, 3D HepG2, and day 30 3D HLCs were incubated for 24 h with 2-deoxy-D-glucose (2DG; a glucose analogue and glycolysis inhibitor), etomoxir (fatty acid oxidation blocker), 3-nitropropionic acid (complex II inhibitor), UK-5099 (mitochondrial pyruvate carrier inhibitor), and rotenone (complex I inhibitor) (Brouillet et al. 2005; Li et al. 2003; O’Connor et al. 2018; Pajak et al. 2019; Wang et al. 2023), Fig. 3A. Day 30 of the differentiation was selected for HLCs (2D and 3D), since at that time point hepatic markers were already expressed (Figs. S1−3), and CYP activity was not required. 2D PHHs were used here as a gold standard reference for liver metabolism, while 2D HepG2 cells were used as examples of poor mitochondrial function (Gupta et al. 2021; Moreno-Torres et al. 2022; van der Stel et al. 2020). We hypothesized that both HepG2 and HLCs, when embedded in a 3D polypeptide hydrogel scaffold, would exhibit enhanced mitochondrial function compared to their respective 2D conformations.

Fig. 3.

3D HLCs have improved mitochondrial function, compared to conventional 2D culture. A. Scheme of used inhibitors on pathways of the central carbon metabolism (CCM) involving the mitochondria: 2-deoxy-D-glucose, Etomoxir, 3-Nitropropionic acid, UK-5099 Acid, and Rotenone. B-F. LC–MS normalized intensity (to control condition) of selected CCM metabolites in 2D PHHs (black), 2D HepG2 (purple), 2D HLCs (blue), 3D HepG2 (pink), 3D HLCs (red) incubated for 24 h with B. 2-Deoxy-D-glucose (10 mM), C. Etomoxir (30 µM), D. 3-Nitropropionic acid (100 µM), E. UK-5099 (10 µM), and F. Rotenone (5 nM); N = 3–5 (individual wells from the same culture batch) ± SEM. Statistics were performed using one-sample t-test, with stipulate 1 as the mean, considering significance for p-value < 0.05*. VDAC: Voltage dependent anion channel, MPC: Mitochondrial pyruvate carrier, CPT1: Carnitine palmitoyltransferase 1; CPT2: Carnitine palmitoyltransferase 2, NAD(H): Nicotinamide adenine dinucleotide, FAD(H): Flavin adenine dinucleotide, Q: Coenzyme Q, CytC: Cytochrome C, TCA: Tricarboxylic acid, α-KG: α-ketoglutarate, FFA: Free fatty acid, G6P: Glucose-6-phosphate, F1,6BP: Fructose-1,6-bisphosphate, ATP: adenosine triphosphate, GSSG: Glutathione oxidized, GSH: Glutathione reduced. See Figs. S4-5 for supplementary data

The intracellular metabolome analysis (Table S13) of 3D HLCs and 3D HepG2 revealed, through principal component analysis (PCA), that the overall metabolic response to some inhibitor treatment was closer to that obtained in 2D PHHs, compared to 2D HLCs and 2D HepG2, respectively (Fig. S4). Specifically, this was very clear for UK-5099 treatment. When cells were treated with 2DG (Fig. S4A), the cell model that differed the most from the 2D PHHs metabolome was 2D HepG2, as expected, since it relies heavily on glycolysis for energy production. Interestingly, 3D HepG2 grouped closer to 3D HLCs, than to 2D HepG2 (Fig. S4A). To follow the metabolites responsible for driving these cell-specific differences, the PCA loadings plot (Fig. S5) was further examined.

When focusing on the effects in HLCs, 2DG led to impairment of glycolysis (Fig. 3B), in both 2D and 3D models. Fructose-1,6-bisphosphate (F1,6BP) levels were decreased in both cell models, whilst pyruvate only in 2D HLCs, and glucose-6-phosphate (G6P) in 3D HLCs. A decrease in glycolytic intermediates in both models indicated that HLCs can use alternative routes, other than glycolysis, for energy production. Upon etomoxir treatment (Fig. 3C), 3D HLCs were the only model with a reduction in ATP levels, indicating this cell system may rely on fatty acid oxidation to produce ATP. 3-Nitropropionic acid (Fig. 3D), a complex II inhibitor, led to succinate accumulation in both 2D and 3D HLCs. However, fumarate, the product of the reaction, was not altered in 3D HLCs, like the response of 2D PHHs. The inhibitor of the mitochondrial pyruvate carrier, UK-5099 (Fig. 3E), led to a decrease in TCA cycle activity in 3D HLCs (like 2D PHHs), but not in 2D HLCs, as indicated by decreased malate levels. It may be inferred that 3D HLCs have a higher dependency on pyruvate as a TCA cycle substrate than 2D HLCs. When cells were exposed to rotenone, the abundance of ATP, GSSG (glutathione oxidized), and GSH (glutathione reduced) was notably decreased in HLCs (Fig. 3F). These results highlighted the sensitivity of HLCs (both 2D and 3D) to mitochondrial inhibition and oxidative stress in response to rotenone treatment. In summary, 3D HLCs demonstrated a greater reliance on fatty acid β-oxidation to fulfil bioenergetic requirements and enhanced pyruvate oxidation in the TCA cycle.

When assessing the effect of a 3D cell structure over the classical 2D cell culture in a well-characterized cell system, HepG2, we observed some distinct characteristics compared to HLCs. Interestingly, when HepG2 were treated with 2DG (Fig. 3B), it was apparent that the levels of glycolytic intermediates (F1,6BP and pyruvate) were decreased in 3D HepG2 but not in 2D HepG2. The accumulation of G6P suggested that 2D HepG2 cells were more glycolytic-dependent for energy production compared to 3D HepG2 cells. Etomoxir incubation (Fig. 3C) induced a reduction in citrate levels in 3D HepG2 (like 2D PHHs), suggesting impairment of fatty acid oxidation, which was not observed in 2D HepG2. Both 2D and 3D HepG2 accumulated succinate upon 3-nitropropionic acid exposure (Fig. 3D). Nevertheless, fumarate accumulated in both HepG2 models, with a more pronounced effect in 3D compared to 2D. The only model that accumulated high levels of pyruvate upon UK-5099 inhibition (Fig. 3E) was 2D HepG2, a consequence of its high glycolytic dependency to meet energy requirements, as it lacks alternative routes for energy generation. On the contrary, 3D HepG2 probably increased its use of alternative carbon sources, e.g., intracellular fatty acids, to compensate for the inhibition of pyruvate entering the TCA cycle (Fig. 3E). Cells exposed to rotenone (Fig. 3F) did not show a significant decrease in GSH nor GSSG. Overall, 3D HepG2 decreased glucose dependence (thus relying more on OXPHOS) and increased fatty acid β-oxidation to fulfil bioenergetic purposes, compared to 2D HepG2. This suggests a higher mitochondrial function in HepG2 organobodies compared to cells in 2D.

3D HLCs exhibit higher DILI-drug sensitivity compared to 2D HLCs

We next assessed whether the 3D HLC model would be suitable for toxicity testing. Toxicity through bioactivation is a main mechanism behind DILI, a leading cause of acute liver failure, and the primary reason for drug withdrawal (Fernandez-Checa et al. 2021). Drug bioactivation may lead to the production of reactive metabolites, causing cell DNA damage and mitochondrial oxidative stress (Holt and Ju 2006). Predicting DILI remains a challenge due to the poor metabolic capacity of in vitro liver systems. Thus, toxicity is only perceived at higher drug concentrations or in later stages of drug development. Since 3D HLCs have demonstrated enhanced xenobiotic metabolism and mitochondrial function, this model was here tested for the toxicity assessment of 8 DILI compounds: amiodarone (metabolized through CYP3A4) (Hamilton et al. 2020; Li et al. 2023), azathioprine (non-CYP routes) (Díaz-Villamarín et al. 2023), cyclosporine A (CYP3A4) (Akbulut et al. 2015; Kong et al. 2022), diclofenac (CYP2C9) (Amanullah et al. 2022), nefazodone (CYP3A4) (Silva et al. 2016), omeprazole (CYP2C19) (Mahmoudi et al. 2021; Lo Re et al. 2024), rosiglitazone (CYP2C8) (Bazargan et al. 2017; Xu et al. 2022), and troglitazone (CYP3A4) (S and Vuppu 2020; Yu et al. 2020).

Cell viability was tested in 2D PHHs, 2D HepG2, 2D HLCs day 40, and 3D HLCs day 40 after 24 h incubations with DILI drugs by 3 different readouts: resazurin assay, lactate assay, and calcein AM assay (Fig. 4). These three different readouts were employed to compare method sensitivity and differences in method applicability per cell model. A decrease in resazurin conversion, lactate secretion, or calcein AM cleavage can be associated with a reduction in cell viability. Here, day 40 HLCs were selected because CYP expression is necessary for drug biotransformation and bioactivation, providing a fair comparison between 2D and 3D HLCs.

Fig. 4.

3D HLCs are more sensitive than 2D HLCs towards DILI drugs. Cell viability of 2D PHHs, 2D HepG2, 2D HLCs, and 3D HLCs in response to 8 DILI drugs tested by resazurin, lactate, and calcein AM assays. Under each drug’s name, its main metabolizing enzyme is indicated. Low, medium and high drug concentrations, used for the treatment were as follows: amiodarone (7 µM, 14 µM, 28 µM); azathioprine (2.5 µM, 5 µM, 10 µM); cyclosporine A (5 µM, 10 µM, 20 µM); diclofenac (150 µM, 300 µM, 600 µM); nefazodone (9.5 µM, 19 µM, 38 µM); omeprazole (210 µM, 420 µM, 840 µM); rosiglitazone (117.5 µM, 235 µM, 470 µM) and troglitazone (25 µM, 50 µM, 100 µM). Significantly decreased viability, compared to control (0 µM), is indicated by a dot (red: reduced viability at low drug concentration, orange: medium concentration, and green: high concentration), as calculated by an unpaired t-test (considering significance p-value < 0.05). N = 3–15 (individual wells from the same culture batch). “Overall Toxicity” reports cytotoxicity from the method with the highest sensitivity per cell model and drug. Empty cells denote no toxicity. See Figs. S6–8 for supplementary data

Calcein AM assay was the most sensitive method for detecting toxicity in the tested 2D models, demonstrating that all drugs were toxic in 2D PHHs at low concentrations. However, in 3D HLCs, there was no toxicity detected using this readout, likely due to technical limitations in quantifying calcein inside the organobodies. Lactate assay was the most sensitive method to detect toxicity in 3D HLCs, as it identified 6 out of 8 drugs to be toxic (Fig. 4). Resazurin assay was considered the testing method with the lowest sensitivity, as it only detected toxicity of 2 drugs (at the highest concentration) in 2D PHHs. In terms of overall toxicity (considering the outcome of all three methods), 2D PHHs were the cell model with the highest sensitivity towards DILI drugs, as toxicity was detected for all drugs at all three concentrations, followed by 3D HLCs, as toxicity was detected for 6 drugs at all three concentrations (amiodarone and omeprazole did not lead to a reduction in viability).

In particular, the response towards azathioprine and cyclosporine A in 3D HLCs was highly similar to the one obtained for 2D PHHs, Fig. 4–5. In fact, 3D HLCs was the only cell model able to detect azathioprine toxicity at low concentrations, alike PHHs, measured by lactate and calcein AM assays, respectively (Fig. 5A). Azathioprine is conjugated to glutathione, leading to 6-mercaptopurine formation, an active intermediate that is further metabolized to 6-thioguanine nucleotides. These metabolites may cause hepatotoxicity through DNA damage and inhibition of de novo purine synthesis (Díaz-Villamarín et al. 2023). Another DILI compound showing similar toxicity patterns in 3D HLCs and 2D PHHs, assessed by lactate and calcein AM assays, respectively, was cyclosporine A (Fig. 5B). In this case, the parent compound is responsible for causing the toxic outcome. An overdose of cyclosporine A leads to reactive oxygen species production, which results in oxidative stress and mitochondrial dysfunction (Akbulut et al. 2015). These results support 3D HLCs as a sensitive and appropriate system for assessing the toxicity of model DILI drugs.

Fig. 5.

3D HLCs predict azathioprine and cyclosporine A toxicity, just like 2D PHHs. Cell viability of 2D PHHs (black), 2D HepG2 (purple), 2D HLCs (blue), and 3D HLCs (red) using resazurin, lactate, and calcein AM assays, challenged to A. Azathioprine and B. Cyclosporine A at control (0 µM), low, medium, and high concentrations. Azathioprine (2.5 µM, 5 µM, 10 µM) and cyclosporine A (5 µM, 10 µM, 20 µM). N = 4–15 (individual wells from the same culture batch), average ± SEM. Statistics were performed using an unpaired t-test, considering significant when p value < 0.05*. Data is represented as % to control condition

Discussion

The liver is the principal organ responsible for xenobiotic detoxification through metabolism and nutrient supply (Guengerich 2021; Morio et al. 2021). Although iPSCs differentiated to HLCs show promise in metabolism studies, differentiation protocols still struggle to match metabolic activities at the level of PHHs (Boon et al. 2020; Li et al. 2019). In this study, we assessed whether some of these shortcomings would be overcome by using iPSC-HLCs in 3D organobody culture. 3D structures have been reported to enhance a hepatic phenotype, closer to the one found in vivo. With this premise, we employed LC–MS semi-targeted metabolomics to study the xenobiotic and mitochondrial metabolism of 3D organobodies, in combination with toxicity testing.

3D organobodies expedite the maturation of HLCs’ xenobiotic metabolism. Similar CYP3A4 activity levels measured by LC–MS were found in 3D HLCs at day 30 and day 40. However, for 2D HLCs, acquisition of CYP3A4 activity required at least 40 days of culture, as reported previously (Pozo Garcia et al. 2025). Earlier research has shown that methylation patterns and epigenetic modifications of CYP genes differ between 2D and Matrigel-supported 3D HLCs (Kang et al. 2025). CYP genes in 3D HLCs are less methylated compared to 2D, closer to 2D PHHs’ methylation patterns (Kang et al. 2025). Furthermore, CYP histone acetylation was found to be higher in 2D PHHs and 3D HLCs, compared to 2D HLCs. These epigenetic differences between 2D and 3D HLCs possibly explain changes in enzyme expression and activity (Kang et al. 2025). However, the reason behind an earlier maturation of isoenzyme activity remains elusive, as there isn’t a common regulation pattern between CYP2A6, CYP2C8/9, and CYP3A4 (with a shorter maturation time) and CYP2B6 and CYP2D6 (with a longer maturation time) (Hakkola et al. 2020; Plant 2007). Besides culture time, liver specific functions were reported to persist longer in 3D models compared to 2D HLCs (Ma et al. 2016). This suggests a broader range of testing utilizing 3D HLCs, e.g., repeated-dose exposures to mimic toxicity due to drug accumulation.

In drug metabolism and pharmacological research, 3D HLCs offer advantages over 2D HLCs. The increase in expression of drug-metabolizing enzymes (including CYP3A4) in 3D HLCs, compared to 2D HLCs was previously reported, using an analogous 3D system to the one employed here (Kiamehr et al. 2026). Higher CYP3A4 activity in 3D HLCs at day 40 of differentiation (when compared to 2D), was previously reproduced with the use of a fully defined synthetic matrix for spheroid formation (Kumar et al. 2021). Additionally, Kim and colleagues demonstrated that CYP3A4 and CYP2D6 enzyme expression and activity were higher in HLCs Matrigel-maintained organoids, compared to 2D HLCs, following a 50-day culture protocol (Kim et al. 2022). These findings are in line with this study, as CYP3A4 and CYP2D6 activity were enhanced in defined synthetic self-assembled peptide hydrogel organobodies cultured for 30–40 days compared to 2D HLCs.

Differing metabolic profiles between 3D and 2D have been previously reported (Jensen and Teng 2020). The present work demonstrated that HLCs and HepG2 cells cultured as 3D organobodies, using the reported protocol, showed a mitochondrial function closer to the one found in PHHs. This increase in function may be associated with the formation of higher mitochondrial networks in the 3D structures, thereby stimulating mitochondrial biogenesis. Factors like the cytoskeleton, extracellular matrix (ECM) composition, and extracellular mechanical factors, such as cellular stretch patterns, maintain and influence mitochondrial function (Bartolák-Suki et al. 2017). ECM stiffness modulates mitochondrial cluster size and ATP production (Bartolák-Suki et al. 2017). The used polypeptide hydrogel resemble true 3D cell environments in HepG2 and HLCs (Pampaloni et al. 2007), which may consequently improve overall function. Furthermore, complex ECM and cell–cell interactions present in in vivo niches are not well represented in 2D, potentially masking key metabolic patterns that depend on 3D structural and spatial organization (Perez-Ramirez and Christofk 2021). Thus, cells in 3D may overcome some of the architectural limitations present in 2D models (Perez-Ramirez and Christofk 2021). Lastly, the passive diffusion of oxygen through the organoid generates oxygen gradients within the 3D structure (Perez-Ramirez and Christofk 2021), potentially leading to hypoxic conditions that may influence mitochondrial activity. Metabolite comparative analysis of 2D PHHs, 2D HepG2, 2D HLCs, 3D HepG2, and 3D HLCs suggested that 3D structures increased fatty acid β-oxidation, to cover energy requirements (a characteristic of liver cells (Morio et al. 2021)) and decreased glycolytic dependency through an increase in OXPHOS for ATP production. Transcript evidence of lower glucose dependence was previously observed in Matrigel-embedded HepG2 spheroids cultured for 21 days, compared to cells in monolayer (van der Stel et al. 2020). In line with our findings, Kiamehr et al. demonstrated that 3D HLCs exhibit upregulated fatty acid β-oxidation (indicated by genes ACSL5, HSD17B4, HADHB, ACAA2, and EHHADH) compared to 2D, as revealed by RNA sequencing (Kiamehr et al. 2026). Furthermore, a previous study demonstrated that PGC1α (peroxisome proliferator-activated receptor gamma coactivator 1-alpha), a key regulator of mitochondrial biogenesis and energy metabolism (Anderson et al. 2008), was transcriptionally upregulated in 3D HLCs compared to 2D (Kumar et al. 2021), supporting our data.

After the metabolic characterization of 3D models, this study evaluated the ability of 3D HLC vs 2D HLC, as well as 2D PHHs and 2D HepG2, to predict toxicity when exposed to known DILI compounds. Readouts were performed using three viability assays: lactate, resazurin, and calcein AM to obtain a comparative overview of the models' sensitivity (2D PHHs, 2D HepG2, 2D HLCs, and 3D HLCs) to DILI drugs. Interestingly, responses differed depending on the cell model (2D or 3D) and the method employed for the readout. Calcein AM demonstrated the highest sensitivity for assessing toxicity in 2D models, albeit with a higher replicate variability, than the resazurin and lactate assays. Calcein AM's limit of quantification and detection thresholds were previously compared with the resazurin assay, which found the former more sensitive (Sitte et al. 2023). However, this method was not suitable for 3D structures due to technical limitations while quantifying dye accumulation into the organobodies. The possible formation of a necrotic core inside the organobodies makes using microscopy and immunofluorescence readouts challenging. The resazurin assay was the least sensitive method for detecting toxicity in both 2D and 3D models, except for 2D HepG2. The sensitivity of HepG2 cells to the resazurin assay can be attributed to their low content and poorly developed mitochondria. Thus, small changes in the reduction of resazurin to resorufin can have a higher translation into the results compared to other cell models (Arzumanian et al. 2021; Lavogina et al. 2022). Our results strongly suggest that lactate determination is the better-performing method for toxicity assessment in 3D models, as an alternative to the often-used commercially available kits adapted for spheroids (Kiamehr et al. 2026; Kim et al. 2022).

For toxicity assessment, 3D HLCs showed greater sensitivity to DILI drugs than 2D HLCs. The enhanced metabolic capacity of 3D HLCs may contribute to increased drug bioactivation, thereby enhancing cytotoxicity (as shown here with azathioprine). Similarly, previous studies have demonstrated that scaffold-free 3D PHHs exhibit heightened sensitivity to DILI drugs, particularly acetaminophen and azathioprine (bioactivated drugs), when compared to 2D cultures (Li et al. 2020). Analogously, 3D HepG2, already extensively characterized in previous studies, showed enhanced DILI predictivity compared to 2D cultures (Ramaiahgari et al. 2014; Takayama et al. 2013). Further, the increase in mitochondrial function for bioenergetic purposes may be a consequence of greater 3D sensitivity to drugs that induce mitochondrial damage (as shown here by cyclosporine A exposure). This last was supported by previous studies using HepG2 spheroids (van der Stel et al. 2020). In our study, amiodarone did not exhibit any toxicity in 3D HLCs, unlike in 2D models. A similar amiodarone toxicity pattern was obtained when comparing 2D and 3D HepG2 responses, resulting in a higher IC50 for the latter (Li et al. 2023). Amiodarone is highly lipophilic (Shleghm et al. 2020); thus, higher sensitivity in 2D hepatic models may result from a greater lipid concentration compared to 3D, which promotes amiodarone accumulation, or from more efficient lipid metabolism in the 3D models.

The cytotoxic response to each tested DILI compound should be assessed independently within each model, as distinct mechanisms may not strictly be related to CYP activity and expression (Basharat et al. 2020), including for example phase II detoxification processes. Differences in pharmacological responses between 2D and 3D cell systems have been previously noted, citing some of the following reasons: molecular targets specifically expressed only in 3D structures (Friedrich et al. 2009), differences in local drug availability, and changes in intrinsic cellular repair mechanisms, which alter DNA damage and/or apoptosis (Langhans 2018). In summary, 3D structures better mimic drug cell responses through an improved representation of the tissue microenvironment in which cells proliferate, differentiate, and aggregate (Jensen and Teng 2020). For further enhancement of 3D HLCs' performance, immune components (Kupffer cells) could be integrated into the system to improve DILI predictivity (Proctor et al. 2017). Furthermore, other strategies, such as the previously developed educated spheroids (Roux et al. 2024a; Roux et al. 2024b), incubations with human serum to acquire a personalized phenotype in hepatic organoids, could be applied in conjunction to the proposed 3D HLCs system. This would further enable the identification of common molecular mechanisms of DILI, with the potential of improving the overall performance of in vitro liver models.

This study presents 3D HLC organobodies embedded within a synthetic self-assembled peptide hydrogel as a promising new model for research on drug metabolism, drug effects on mitochondrial function, and toxicity. However, some limitations were also encountered during the study. The yield of 3D cell production is often lower than in 2D. This results in a reduced number of cells in 3D structures, compared to cells in monolayers, which can challenge the detection sensitivity of certain techniques. There is also a greater biological variability observed in 3D systems, due to differences in cell number and size. In addition, organobody generation requires more skilled handling than traditional monolayer culture. Lastly, routine in vitro assays (viability, imaging techniques, and mitochondrial respiration) are often not optimized for 3D cell culture and require refinement for this application. Nonetheless, the stability of 3D HLCs facilitates and opens opportunities for use in complex experimental setups, including co-cultures, liver-on-a-chip applications (Liu et al. 2024), and in vitro spatio-temporal NMR studies with living cells (Knitsch et al. 2021). In conclusion, the incorporation of 3D iPSC models into pharmacological applications remains highly promising, offering a more sustainable and precise approach to toxicology and drug development.

Supplementary information

Below is the link to the electronic supplementary material.

Author contributions

VPG and SM designed the study and all experiments. MK and CV provided experimental input, support, and training in 3D organobody preparation. VPG and RV performed in vitro assays, LC–MS measurements, and data analyses. VPG and EN performed cell culture and conducted immunostaining experiments. PJ acquired the funding and provided resources. JCV contributed to the critical discussion of the results. VPG, SM, and JCV wrote the manuscript, and all other authors reviewed, edited, and accepted the submitted version of the manuscript.

Funding

This work was supported by the project RISK-HUNT3R: RISK assessment of chemicals integrating HUman centric Next generation Testing strategies promoting the 3Rs. RISK-HUNT3R has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 964537 and is part of the ASPIS cluster. This work reflects only the authors’ views, and the European Commission is not responsible for any use that may be made of the information it contains. Additional funding for CV is from FWO-SBO-LiMiC #S001121N.

Data availability

Supporting Information is available in PDF format. Metabolomics processed data have been deposited in BioStudies (https://www.ebi.ac.uk/biostudies/) under the unique permanent identifiers S-RHER292 (Fig. 1), S-RHER291 (Fig. 2), and S-RHER293 (Fig. 3), and are publicly available as of the date of publication. Raw data and/or any additional information will be shared by the corresponding author upon request.

Declarations

Clinical trial number: not applicable.

Conflict of interest

Authors MK and CV are inventors on “Spheroidal self-assembled peptide hydrogels comprising cells, WO2022079272A1”, and CV is one of the inventors on “Cell culture media for differentiation of stem cells into hepatocytes, WO2018229274A1”. These patents are owned/assigned to Katholieke Universiteit Leuven and Tampere University Foundations. All other authors present no conflict of interest.