Human iPSC-derived liver co-culture spheroids to model liver fibrosis

The lack of adequate human in vitro models that recapitulate the cellular composition and response of the human liver to injury hampers the development of anti-fibrotic drugs. The goal of this study was to develop a human spheroid culture model to study liver fibrosis by using induced pluripotent stem cell (iPSC)-derived liver cells. iPSCs were independently differentiated towards hepatoblasts (iHepatoblasts), hepatic stellate cells (iHSCs), endothelial cells (iECs) and macrophages (iMΦ), before assembly into free floating spheroids by culturing cells in 96-well U-bottom plates and orbital shaking for up to 21 days to allow further maturation. Through transcriptome analysis, we show further maturation of iECs and iMΦ, the differentiation of the iHepatoblasts towards hepatocyte-like cells (iHeps) and the inactivation of the iHSCs by the end of the 3D culture. Moreover, these cultures display a similar expression of cell-specific marker genes (CYP3A4, PDGFRβ, CD31 and CD68) and sensitivity to hepatotoxicity as spheroids made using freshly isolated primary human liver cells. Furthermore, we show the functionality of the iHeps and the iHSCs by mimicking liver fibrosis through iHep-induced iHSC activation, using acetaminophen. In conclusion, we have established a reproducible human iPSC-derived liver culture model that can be used to mimic fibrosis in vitro as a replacement of primary human liver derived 3D models. The model can be used to investigate pathways involved in fibrosis development and to identify new targets for chronic liver disease therapy.


Introduction
Chronic liver disease (CLD), caused by long-term liver injury, is characterized by inflammation and fibrosis, and in some cases also steatosis.If left untreated, chronic liver injury can eventually result in the development of liver cirrhosis and hepatocellular carcinoma (Asrani et al 2019).
During chronic liver injury, which includes chronic viral hepatitis, continued alcohol use, drug toxicity and steatohepatitis associated with metabolic syndrome, hepatocytes get damaged.During this process, the non-parenchymal cells of the liver [liver sinusoidal endothelial cells (LSECs), Kupffer cells and hepatic stellate cells (HSCs)] are also affected.The HSCs become activated and play a pivotal role in the wound healing response to injury by depositing large amounts of extracellular matrix (Marrone et al 2016, Tsuchida andFriedman 2017).This ultimately results in the development of liver fibrosis.
Although anti-fibrotic treatments are being evaluated in clinical trials (Zhao et al 2022), there are still no FDA-approved HSC-specific anti-fibrotic drugs available in the clinic, leaving an unmet need for our global healthcare (Schuppan et al 2018a).To investigate the onset and progression of liver disease and to screen for potential therapeutics, 2D mono-cultures are mostly used (Zeilinger et al 2016).Primary human hepatocytes (PHHs) are the gold standard for drug toxicity screening, but they show inter-donor variability, and lose most of their functionality quickly upon culturing in 2D (Godoy et al 2016).To investigate human HSC activation, primary human cells or cell lines can be used that are already in a highly activated state due to cell expansion as monolayer cultures (El Taghdouini et al 2015).More importantly, to model fibrosis, hepatocytes and HSCs need to be present to mimic liver injury induced fibrosis (van Grunsven 2017, Mannaerts et al 2020, Lee et al 2023).To this end, researchers have been developing 3D models consisting of multiple primary human liver cell types.The disadvantages of this model are the limited supply of primary human liver cells and the inter-donor variability (Huch et al 2015).One possible solution for the problems presented by primary cells for modelling liver fibrosis is the use of induced pluripotent stem cell (iPSC)-derived liver cells (Tricot et al 2022).iPSCs are an inexhaustible source of cells that allow for genetic modifications and donor variability can be controlled.
In this paper, we describe the development of iPSC-derived liver spheroid cultures to study liver fibrosis.The model consists of four independently differentiated liver cell types, including hepatoblasts (iHepatoblasts), HSCs (iHSCs), macrophages (iMΦ) and endothelial cells (iECs), that are subsequently assembled into spheroids.Medium compositions were optimized to get the best spheroid formation and the highest hepatocyte maturity and HSC functionality.Spheroids containing different cell ratios were evaluated for their functionality and we demonstrate that cellular ratios affect spheroid formation and iHSC activation but not maturation of the iHepatoblasts into hepatocyte-like cells.Integration of iHSCs into liver spheroid cultures renders them more quiescent, allowing for iHSC activation through hepatocyte damage, thereby partly mimicking liver fibrosis in vitro.

Differentiation of iPSCs towards endothelial cells
For the differentiation of iPSCs towards iECs the protocol of De Smedt et al was used (2021).LDM was supplemented at day 0 with 5 µg ml −1 DOX.From day 2 onwards, cells were refreshed every 2 days with LDM containing 5 µg ml −1 DOX and 2% FBS.At day 4, cells were passaged 1:2 using 1 ml per well (6-well plate) of Accutase in Biolaminin (1/20 diluted) coated plates.They were kept in culture until day 14 and passaged whenever 100% confluent.iECs were used from day 14 onwards for 3D spheroid co-cultures.

Human liver cell isolation and integration into 3D spheroids
Postoperative surgery liver pieces (table S2) were directly placed in IGL-1 organ preservation solution (Institut Georges Lopez, IGU.IGL-1 REV07) and isolation procedure, based on the isolation procedure used for primary mouse liver cells (van Os et al 2022), was started within 2 h.The human liver piece was perfused at a perfusion rate of 7.5 ml min −1 for 5 min with perfusion buffer 1 (table S3) for discoloration of liver and removal of red blood cells followed by a 5 min perfusion with perfusion buffer 2 (table S3) containing 0.5 mg ml −1 of Pronase E (Merck, 1074330005) followed by a 5 min perfusion with perfusion buffer 2, containing 0.25 mg ml −1 Collagenase P (Roche, 11 213873001).The liver piece was afterwards torn into tiny pieces using forceps and incubated in perfusion buffer 2 containing 0.5 mg ml −1 Pronase E and 0.5 mg ml −1 Collagenase P for 20 min at 37 • C.After 5 and 10 min one drop of NaOH (pH 14) was added to the beaker.The single cell suspension was then filtered through a metal sieve (Roth, round sieve ROTILABO, 75, 0.5 mm) and the suspension was centrifuged for 7 min at 50 g to spin down hepatocytes.The hepatocyte pellet was then dissolved in 5 ml of perfusion buffer 2 and kept on ice until the Percoll gradient could be performed.The supernatant with NPF cells was then centrifuged for 8 min at 1750 RPM and the remaining cell pellet was incubated for 3 min at room temperature with red blood cell lysis solution.After a wash step with perfusion buffer 2, the NPF cell pellet was resuspended in PBS containing 1% BSA and stained with FACS antibodies for sorting HSC (based on retinyl esters, ex.: 355 nm, em.: 450 nm), LSEC (CD32-APC) and KC (CD45-FITC and CD163-PE/Cy7).Propidium iodide (Merck, P4864) was used to remove dead cells.Gating strategy is shown in figure S1.Cells were resuspended in seeding medium (Williams E medium (Gibco, A12176-01) supplemented with 10% FBS (Tico Europe, FBSEU500), 50 mg ml −1 kanamycin sulphate, 10 mg ml −1 Sodium ampicillin, 100 U ml −1 PenStrep, 292 mg ml −1 L-glutamine and 7 ng ml −1 Glucagon) and counted for seeding in 3D cultures.For hepatocytes, cells were pipetted in perfusion buffer 2 on a Percoll gradient and centrifuged for 20 min at 50 g.The pellet was then washed with cold PBS and the cell pellet was incubated for 3 min at room temperature with red blood cell lysis.Afterwards the cells were resuspended in seeding medium and counted with trypan blue.Spheroids were formed as described for iPSC-derived cells in the ratio 1333 HSC/666 Hepatocyte/400 LSEC/400 KC and kept in culture for 8 days.From 24 h after seeding onwards, Williams E medium was added, without FBS, but with 50 µg ml −1 kanamycin sulphate, 10 µg ml −1 Sodium ampicillin, 100 U ml −1 PenStrep, 292 mg ml −1 L-glutamine, 7 ng ml −1 Glucagon, 0.5 µg ml −1 insulin (Sigma-Aldrich, I1882) and 25 µg ml −1 Hydrocortisone sodium succinate (Solu-CortedPharmacy).Medium was changed every other day.All research performed using primary human cells was approved by the Human Ethics Committee at UZ Brussel, Brussels, Belgium (B.U.N. 143201525406).

Induction of fibrosis by hepatocyte damage
At day 10 after seeding in 3D, fibrosis was induced in the spheroid cultures by exposure to acetaminophen (APAP, Merck Group, A7085-100 G) for 72 h.For ATP measurements, concentrations between 0 and 100 mM were used.For RNA analysis, 6 spheroids per condition were treated with 20 or 40 mM APAP and collected after 72 h.No medium refreshment was performed in those 72 h.

Cell viability assay
For the measurement of ATP, a luciferase-based ATP measurement kit was used, according to the manufacturer's instructions (CellTiter Glo Luminescent Cell Viability Assay Kit, Promega, G7571).After 72 h of treatment with 0,5,10,15,20,30,40,60,80 and 100 mM of APAP, spheroids were individually transferred into a black bottom 96-well plate in 100 µl of medium.Then, 100 µl of ATP substrate was added to the spheroids.After a 10 min incubation at room temperature on an orbital shaker, spheroids were incubated for another 10 min at room temperature without shaking and luminescence was measured using a GloMax (Promega, GM3000).

mRNA analysis
Per condition 6 spheroids were collected and lysed for RNA extraction by ReliaPrep RNA Cell Miniprep System (Promega, 1074308).mRNA was reverse transcribed into cDNA using random primers and MLV reverse transcriptase (Promega, M1701).For qPCR a GoTaq qPCR Master Mix with BRYTE Green (Promega, A6002) was used and measured using a Quantstudio3 Fast PCR system (Thermo Fisher, A28567).Analysis was performed according to the comparative Ct (∆∆Ct) method where each Ct value was normalized to the mean of the reference genes.Using GeNORM, GAPDH, RPL19 and EAR were determined as most appropriate reference genes.Gene specific primers were designed using the NCBI primer tool and primers were manufactured by Integrated DNA Technologies.A list of used primers can be found in table S4.

Immunostaining
Coverslips of differentiated cells in 2D were washed with PBS and fixed in 10% formalin solution for 10 min at room temperature, and washed 3 times with PBS.Cells were permeabilized with 0.1% Triton X-100 for 3 times 5 min and blocked with 2% bovine serum albumin (BSA) for 30 min.Then cells were incubated overnight at 4 • C with a solution of 0.1% Triton-X100 and 1% BSA with the primary antibody, as listed in table S5.The next day cells were washed 3 times 5 min with 0.1% Triton X-100 in PBS and incubated with the secondary antibody for 2 h in a solution of 0.1% Triton X-100 and 1% BSA (table S5).Lastly cells were washed 3 times 5 min with PBS and rinsed briefly with distilled water, before mounting with DAPI (Merck group, D9564)-containing mounting medium (Dako, S3023).
Spheroids were washed with PBS and fixed in 10% formalin solution for 10 min at room temperature, and washed 3 times with PBS.Then spheroids were either embedded in paraffin and sliced into 5 µm sections with a microtome (Microm, HM340E), or they were stained as whole spheroids.The paraffin sections were first deparaffinized and rehydrated with xyleen and ethanol (100%, 90%, 70%).A citratebased antigen retrieval solution was used (Dako, S1699) before permeabilization with 0.05% Tween in PBS.Sections were blocked with 2% BSA for 30 min after which they were incubated with the primary antibody (table S5) in a solution of 1% BSA in PBS.Sections were incubated overnight at room temperature.The next day slides were rinsed with 0.05% Tween and incubated with the secondary antibody (table S5) in the same solution as the primary, for 1 h at room temperature.Then the sections were washed with PBS and distilled water before mounting, the same way as coverslips.
For whole spheroid stainings, spheroids were first blocked for 2 h with 0.5% Triton-X100 and 0.2% fish skin gelatin (Merck group, c0130) in PBS, incubated overnight with the primary antibody solution of 0.1% Triton and 0.125% fish skin gelatin, washed with 0.05% Tween, incubated with the secondary antibody in the same solution as the primary, washed with PBS, rinsed with distilled water and mounted with DAPIcontaining mowiol (Merck group, 81 381).
Images of coverslips and paraffin sections of spheroids were taken with the EVOS (Thermo Fisher Scientific, AMF7000).Images of whole spheroids were taken with a confocal microscope (Zeiss, LSM800).

Hematoxylin and eosin staining
Spheroids were fixed, embedded in paraffin and sliced into 5 µm sections, deparaffinized and rehydrated as described for immunostainings.The slides were incubated with hematoxylin (Carl Roth, T865.2) for 5 min, rinsed with tap water, acidic water and distilled water, and afterwards incubated with eosin (Merck group, E6003) for 3 min.Then they were rinsed with distilled water and dehydrated again with ethanol (70%, 90%, 100 %) and xyleen, after which they were mounted with mounting medium.Images were taken with the EVOS (Thermo Fisher Scientific, AMF7000).
2.14.RNA sequencing and bioinformatic analysis 30 spheroids were pooled and total RNA was extracted using the ReliaPrep RNA Cell Miniprep System (Promega, 1074308).The same was done for freshly isolated PHHs from four different isolations (table S2), iPSCs and iHepatoblasts (day 8 of differentiation).RNA quality was assessed using a Bioanalyzer 6000 and samples with sufficient quality and a concentration of at least 10 ng µl −1 were used.Library preparation was done with the QuantSeq 3 ′ mRNA-Seq Library Prep Kit of Lexogen (Lexogen, 015.96).Single-end sequencing was run on NovaSeq SP 100×6bp resulting in an average of 7 M reads per sample.Quality control was performed with FastQC followed by trimming using AfterQC (Chen et al 2017) followed by mapping of the reads to the reference genome (Homo Sapiens_GRCm38.p6)using STAR (Pouzat and Chaffiol 2009).Assembly of genes and transcripts was performed using python package StringTie (Pertea et al 2015).Once raw counts were generated, further analysis was performed in R using the DESeq2 package (Love et al 2014).Heatmaps were constructed with the ComplexHeatmap package (Gu et al 2016, Gu 2022).Gene ontology analysis was executed on differentially expressed genes (DEGs) with a p-value lower than 0.05 and a fold change higher than 2 or lower than −2 with the enrichGO function of the clusterProfiler package (Yu et al 2012, Wu et al 2021).
The publicly available human liver Single-cell RNA-Sequencing dataset GSE192742 (Guilliams et al 2022) was analyzed using the Seurat package (Stuart et al 2019).Gene regulatory networks were constructed and scored using the pySCENIC package (Van de Sande et al 2020).Gene regulatory networks were scored on the bulk-Seq RNA-Sequencing datasets generated in this paper with gene set variation analysis using the GSVA package (Hanzelmann et al 2013).

Statistics
Statistical tests were performed with Graphpad Prism 9.0.If the sample size was large enough (n > 3), the Shapiro-Wilk test was used to check for normality.For the time courses of the differentiations in 2D, a one-way ANOVA with Bonferroni's multiple comparisons test was used to compare all time points of a differentiation to day 0. To compare different media or ratios over time, a two-way ANOVA was used with Bonferroni's multiple comparisons test.All media on day 13 were compared to the standard medium with amino acids and 2% L-Serine.All cell ratios at day 21 were compared to the standard ratio with iMΦ.For the TGFβ exposures, a one-way ANOVA with Bonferroni's multiple comparisons test was used, and TGFβ-treated spheroids of all conditions were compared to each other.Wilcoxon matched-pairs signed rank test (one-way) was used when spheroids were treated with APAP.Lastly, a Mann Whitney test was used to compare the gene expression of primary human spheroids at day 8 to iPSC-derived spheroids at day 21.Symbols meaning: * p ⩽ 0.05, * * p ⩽ 0.01, * * * p ⩽ 0.001, * * * * p ⩽ 0.0001.

Differentiation of iPSCs towards hepatoblasts, hepatic stellate cells, endothelial cells and monocytes/macrophages in 2D
iPSCs that allow the doxycycline-inducible overexpression of HNF1α, PROX1 and FOXA3 (iPSC3X) can generate hepatocyte-like cells after a differentiation period of 32 days in the presence of high concentrations of essential and non-essential amino acids and 2% Glycine (Boon et al 2020).Integration of these iHeps in spheroids would be ideal, but these fully mature iHeps at day 32 cannot be dissociated sufficiently without causing considerable cell loss.
We thus dissociated cells at a stage where we could still obtain single cells without too much cell loss and tested whether we could differentiate hepatoblasts further in 3D co-cultures.Differentiation of iPSC3X until day 8 of differentiation (hepatoblast stage) (figure 1 iECs were passaged at day 4 and 8 of differentiation and were used at day 14.After ETV2 induction, mature endothelial cell markers CD31 and VWF increased over time, reaching levels similar to freshly isolated primary human LSECs.At day 14 all iECs were positive for CD31 protein expression and negative for OCT4 protein expression (figure S2(B)).
To obtain iPSC-derived iMCs and iPSC-derived iMΦ we used the method of van Wilgenburg et al (Van Wilgenburg et al 2013).First EBs were generated and iMCs were harvested on a weekly basis.For the differentiation towards iMΦ, iMCs were plated in 2D in the presence of Microglia medium, and considered iMΦ after 7 days of differentiation.At the mRNA level, both iMC and iMΦ expressed macrophage markers CD11b, CD16 and CD68 and on protein level they also both expressed CD11b (figure 1(D)) but not OCT4 (figure S2(D)).

Culture and medium conditions for iPSC-derived liver spheroids
Next, we incorporated the differentiated cells into 3D spheroid cultures.For the immune cells, we decided to continue with the iMCs instead of the iMΦ, because the iMCs showed higher gene expression levels of CD16 and CD68, reaching levels that were closer to those of freshly isolated primary human Kupffer cells (figure 1(D)).We used a ratio of 2: 1: 0.6: 0.6 of respectively iHSC: iHepatoblast: iEC: iMC (figure 2(A)).This ratio was chosen based on previous work done in liver spheroids consisting of iHSCs and HepaRG (Coll et al 2018) and spheroids consisting of four different primary mouse liver cell types (van Os et al 2022).However, since we needed further maturation of the iHepatoblasts in these spheroids, and we did not have functionalized hydrogels (Kumar et al 2021) or matrigel (Ouchi et al 2019) that are beneficial for the differentiation towards hepatocytes, we compared several media conditions that could potentially stimulate further differentiation of the iHepatoblasts while at least maintaining the differentiation status of the other cell types in these free floating spheroids.To this end we tested four different media compositions; (i) the medium used for the iHep differentiation which consists of LDM with added components for the maturation towards iHeps including amino acids and 2% L-Serine (Hep GF + AA) (Boon et al 2020), (ii) the iHep differentiation medium without the addition of AA and 2% L-Serine (Hep GF), (iii) a slightly adapted medium composition from Kumar et al which includes growth factors for all different cell types and the addition of AA and 2% L-Serine (All GF + AA), and (iv) the same medium but excluding AA and 2% L-Serine (All GF) (figure 2(A)).Compared to Boon et al we replaced 2% Glycine for 2% L-Serine because iHSCs (figures S3(A) and (B)) and iECs (Kumar et al 2021) do not support these concentrations of Glycine, and L-Serine had a similar beneficial effect on hepatocyte differentiation (Boon et al 2020).For the All GF medium some changes were made to the medium used by Kumar et al.First of all, we switched bFGF to aFGF, as we used aFGF in the 2D differentiations of iHepatoblasts and iHSCs (figures 1(A) and (B)).Before comparing these four media compositions, we tested the effect of 2% FBS on cell maturation, as Kumar et al used 2% FBS in hydrogel co-cultures.Based on cell-type specific gene expression, we could not detect benefits for the addition of FBS (figure S4(A)).Moreover, iHSCs seemed to be reacting better to TGFβ in the absence of FBS (figure S4(B)).Thus, we decided to exclude FBS from all media to limit the use of animalderived products.Lastly, throughout the entire coculture Kumar et al supplemented with retinol and palmitic acid for the maturation of HSCs.We evaluated the use of these components in spheroids consisting of iHepatoblasts and iHSCs, but observed that the spheroids did not form as well and formed expanding blebs (figures S3(C) and (D)).Therefore, we decided to leave out these components during the co-culture period.
Brightfield images showed that the Hep GF medium led to better spheroid formation (figure 2(B)); the spheroids were more sphere-shaped and showed a more consistent shape overall.As expected, the addition of AA and 2% L-Serine to the medium led to significantly higher expression of the mature hepatocyte marker CYP3A4 (figure 2(C)).The highest levels of CYP3A4, PDGFRβ (iHSC) and CD68 (iMC) were reached with the Hep GF medium, with overall better results in the presence of AA (figure 2(C)).By the end of the 3D culture, CD31 (iEC) expression was the highest with All GF + AA, but the expression of CD31 stayed stable throughout the entire culture with all four media (figure 2(C)).
Since activation of HSCs plays an essential role in fibrogenesis and our aim was to develop iPSC-derived spheroid cultures to study fibrosis induction, we also tested the functionality of the iHSCs in the spheroids to decide which medium was the best one to continue with.To this end, we exposed the spheroid cultures at day 11 to TGFβ, a very potent pro-fibrotic growth factor that can activate HSCs directly through binding to the ALK5 receptor (Dewidar et al 2019) (figure 2(D)).Analysis of the mRNA levels of activation markers ACTA2, COL1A1 and LOXL2 at day 13 (figure 2(E)) showed us that iHSCs activate significantly better in the presence of Hep GF + AA (figure 2(E)).
In conclusion, Hep GF + AA led to good spheroid formation, iHep maturation and better functionality of the iHSCs compared to the other media, so we opted to proceed with this medium.

Optimization of cell ratios in iPSC-derived spheroids
We acknowledge that the 2: 1: 0.6: 0.6 ratio of iHSC: iHep: iEC: iMC we used at the start is not representative for the ratio of these cell types in a human liver.To evaluate whether other, more physiologically relevant ratios would yield more mature hepatocytes and/or overall result in more functional liver spheroids, we evaluated different ratios of iPSC-derived cells for spheroid cultures.We thus tested an in vivolike ratio (0.4: 2.8: 0.4: 0.3 of iHSC: iHep: iEC: iMC) based on the results obtained from single nuclei RNA sequencing analysis and image analysis of mouse livers (Guilliams et al 2022).Furthermore, we evaluated an equal (1: 1: 1: 1 of iHSC: iHep: iEC: iMC) ratio as well as the regular standard (Std) ratio with either iPSC derived iMCs or iMΦ.Spheroids with the Std ratio were the only ones that were sphere-shaped and had smooth edges all around on brightfield images which indicates that cells are in a healthy, functional condition (figure 3(A)), while other ratios resulted in incomplete formation of spheroids.These results indicate that iHSCs play an important role in spheroid formation.Using this Std ratio, it did not matter whether we incorporated iMCs or iMΦ, both of which led to nicely formed spheroids with overall smooth edges.Brightfield images suggested that the fewer iHSCs in the cell mixture, the poorer the aggregation of the spheroids (figure 3(A)).Hematoxylin and eosin (H&E) staining did not show necrotic cores in any of the conditions (figure 3(A)).The in vivo-like ratio setting resulted in poor aggregation and yielded multiple smaller aggregates that were impossible to correctly collect and incorporate for the embedding steps of stainings.Surprisingly, hepatocyte maturation was not influenced by the initial cellular seeding ratios as determined by the mRNA levels of the mature hepatocyte marker CYP3A4 at day 21 after seeding in 3D (figure 3(B)).The gene expression of markers for the other cell types, such as PDGFRβ for the iHSCs, CD68 for the iMC/iMΦ and CD31 for the iECs, showed a small increase over time suggesting that they mature slightly in 3D in the presence of other liver cell types, getting closer to levels reached in freshly isolated primary human liver cells (figure 3(B)).Notably, the spheroids with iMΦs showed a similar expression level of CD68 compared to the spheroids with iMCs.Since iMΦs showed lower expression of CD68 compared to iMCs in 2D differentiations (figure 1(D)), these results indicate that the iMΦs mature further in the spheroids and could be used as an alternative to the iMCs.Taken together, these results suggest that medium composition and not cellular ratio is crucial for iHep differentiation and maintenance, and that only spheroid formation is influenced by cellular seeding ratios.
As we did for the medium optimization, we tested the functionality of the iHSCs in the spheroids with different ratios to decide which ratio was the best one to continue with.To this end, we directly activated HSCs with TGFβ at day 11 for 48 h (figure 2(D)) and evaluated the mRNA of activation markers ACTA2, COL1A1 and LOXL2 at day 13 (figure 3(C)).iHSCs in the Std ratio spheroids with iMΦ activated more than iHSCs in other spheroids, suggesting that the reach the highest level of functionality with this ratio in the presence of iMΦ.We thus proceeded with a spheroid seeding ratio of 2 iHSC: 1 iHep: 0.6 iEC: 0.6 iMΦ since this condition allowed for the best formed sphere-shaped spheroids with smooth edges, no necrotic core and the highest TGFβ-induced HSC activation.

iHeps mature and iHSCs become less activated in 3D spheroid cultures
To evaluate the maturity of the iPSC-liver spheroids with the ratio of 2 iHSC: 1 iHep: 0.6 iEC: 0.6 iMΦ, we performed RNA sequencing on spheroids one day after seeding in 3D (D1) and 21 days after seeding (D21).First, a Gene Ontology (GO) analysis of all biological processes was performed on DEGs that were significantly more than 2-fold higher expressed at D1 and DEGs that were higher expressed at D21. Figure 4(A) shows the top 20 most enriched GOs, one day after spheroid formation, that are primarily related to cell cycle, while at the end of the culture there was an enrichment of genes involved in metabolic processes.The latter is indicative of maturation of the iHeps over time in the spheroids which is confirmed by the clear increase in the expression of mature hepatocyte-specific genes displaying the enzymatic activities of Heps, such as phase I and II metabolism, drug and bile acid metabolism and gluconeogenesis (figure 4(B)).Surprisingly, iEC and iMΦ markers increased over time, suggesting further maturation of these cells in 3D as well (figure 4(C)).Next, we performed a SCENIC analysis to determine transcription factor (TF) activity in mature hepatocytes compared to other liver cell types using single cell/nuclei RNAseq data of human livers (Guilliams et al 2022) (figure S5).The activity score of the identified TF gene set, including TFs like PPARA and FOXN3, was verified in the D1 and D21 spheroids with gene set variance analysis, using freshly isolated PHHs as control.We found that the TF activity in spheroids at day 21 was similar to that in the primary cells (figure 4(D)), further confirming the maturity of the iHeps in the spheroids.These results suggest the transition of the iHeps from an immature proliferative state to a more mature differentiated state, with the maturation being higher in 3D cocultures compared to 2D monolayer cultures (figure S6).We confirmed these findings with a Ki67 staining, for proliferating cells, on D1 and D21, which showed a clear decrease of Ki67 positive cells over time (figure 4(E)).Immunofluorescence staining at day 21 with cell type-specific antibodies clearly showed the presence of iHeps (CYP3A4) and iHSCs (PDGFRβ), while iECs (CD31) and iMΦ (CD45) were less abundant, which is in accordance with the cell ratio used (figure 4(F)).Lastly, we evaluated the activation state of the iHSCs in the spheroids.At day 21, spheroids showed decreased expression of fibrotic markers such as COL1A1, −3A1 and −6A3 and an increased expression of markers specific for quiescent HSCs, such as LRAT and COLEC11, suggesting that iHSCs become less activated over time in these spheroid cultures (figure 4(C)).We thus conclude that the iHeps mature over time in the spheroids and that iHSCs become less activated, which are both advantageous characteristics for an in vitro liver fibrosis model.

Mimicking liver fibrosis in iPSC-derived spheroids and comparison to primary human liver spheroids
As iHepatoblasts further mature to iHeps in iPSC spheroids until day 21 of co-culture and iHSCs are less activated in 3D, we tested whether iHeps would respond to a hepatotoxic compound and consequently activate the iHSCs as was previously shown for HepaRG/HSC cultures (Leite et al 2016).APAP was used in these cultures for 72 h to mimic liver injury (figure 5(A)).Treatment with APAP led to a concentration-dependent ruffling of the spheroid border (figure 5(B) and figure S7(A)) and a concentration-dependent drop in ATP and Albumin content, indicating hepatocyte injury (figures 5 (C) and figure S8).This hepatocyte damage was able to induce iHSC activation measured by a concentrationdependent increase in mRNA levels of fibrotic genes COL1A1 and LOXL2 (figures 5(D) and S7(B)).
We next evaluated how well the iPSC-based spheroid cultures reflect primary human liver cultures.To this end we isolated cells from human liver tissue, obtained from liver resections, using a slightly adapted protocol of previously established mouse liver cell isolations (Goncalves et al 2007, Stradiot et al 2017).After purification and isolation of hepatocytes, HSCs, LSECs and Kupffer cells (figure S1), we aggregated cells in the same manner as iPSCderived cells and obtained spheroids that can be kept in culture for at least 8 days.Comparison of cell-type specific mRNA levels of CYP3A4, PDGFRβ, CD31 and CD68 showed that the iPSC-derived liver spheroids reached maturity levels close to those obtained in 8 day old primary human liver spheroids (figure 5(E)).Although the human liver spheroids were not as uniform as iPSC-derived liver spheroids, they were more sensitive to APAP exposure (figures 5(F) and (G)).

Discussion
The pharmaceutical industry is in need for good in vitro models of CLD for the screening and development of drugs to prevent or treat fibrosis since there is currently no approved anti-fibrotic drug available in the clinic in Europe (Schuppan et al 2018a).The generation of iPSC-derived liver models to mimic liver diseases in vitro has gained a lot of attention the last couple of years (Tricot et al 2022).In this paper we developed an iPSC-derived in vitro liver model incorporating four main liver cell types.The medium and cell ratios were optimized to get smooth-edged, sphere-shaped 3D structures with the highest level of iHep maturation and iHSC functionality possible with the current level of differentiation.RNAseq data showed that 3D cocultures of iPSC-derived liver cells benefit from the spheroid culture conditions; hepatocytes are more mature and iHSCs become more quiescent.Lastly, we demonstrated that the iHep maturity and iHSCs in such spheroid cultures allow mimicking liver fibrosis.
Several approaches have been reported on the development of iPSC-derived liver cells.Most studies focused on the generation and functional maintenance of iPSC-derived hepatocytes or hepatocyte-like cells (HLCs) (Hannan et al 2013, Boon et al 2020).The first reports on co-cultures of iPSC-derived hepatocytes and iPSC-derived non-parenchymal cells used Matrigel embedded foregut-differentiated iPSCs to further mature into multiple liver cell lineages.scRNAseq analysis revealed that the final liver organoids contained hepatocyte-and mesenchymal-like cells with biliary and Kupfferlike cells also present (Ouchi et al 2019).While these co-differentiated cells have very good hepatocyte functionality and the hepatocyte transcriptome is similar to primary hepatocytes, the HSC-like cells in these cultures were not extensively characterized (Ouchi et al 2019, Shinozawa et al 2020).Recently Kumar et al reported on the incorporation of iPSC-derived HLCs, iHSCs, iECs and iMCs into functionalized hydrogels and could use this coculture system to model steatohepatitis (Kumar et al 2021).
In the current study, instead of using Matrigel or hydrogels, we formed free-floating spheroids of separately differentiated liver cell types.Since the cells are first differentiated and then incorporated into 3D spheroids, we have full control over the amount and the type of cells.Through continuous rotating of the cultures we prevent spheroid adherence to the plates and avoid formation of multiple small spheroids.Similar findings were reported by De Souza et al who observed stable and reproducible spheroids from a zebrafish hepatocyte cell line using both hanging drop and orbital shaking (70 RPM) techniques in 96-well plates (De Souza et al 2021).Furthermore, studies by Masiello et al suggest that liquid movement, as in orbital shaking, induces shear stress, mimicking the in vivo environment more closely (Masiello et al 2018).Liver shear stress is known to protect liver tissue and stimulate the activity of various CYP enzymes (Ahn et al 2019).The controlled formation of the spheroids gives the advantage of reduced variability between the different batches; both the spheroid formation and cell maturation over time are remarkably reproducible.An additional advantage is that it allows for the optimization of the differentiation of only one cell type independent of the others, and we can use genetically different iPSC-lines for only one or two cell types.For instance one could integrate the use of stress pathway reporter lines (Wijaya et al 2021), or iPSCs with single nucleotide polymorphisms for genes involved in steatosis or fibrosis such as PNPLA3, HSD17B13 and GCKR (Sookoian andPirola 2017, Motomura et al 2021).Furthermore, compared to hydrogels, the use of spheroids makes it easier to do larger drug (i.e, anti-fibrotic) or toxicity screenings.One could implement the use of 384-well plates or incorporate spheroids in commercially available bioreactor systems for flow induction.As most of the commercial organoid bioreactor systems include primary human liver cells that have high inter-donor variability upon a chemical-induced toxic response (Niemeijer et al 2024), iPSC-derived spheroids could be a better alternative for these bioreactor systems.A logical next step would be the use of iPSC-derived liver spheroids in multi-organ systems.A successful combination of hepatic spheroids with non-hepatic organoids such as pancreatic islets, adipose tissue, intestinal organoids and/or muscle cultures will surely advance our understanding of the inter-organ crosstalk that leads to CLDs including metabolic dysfunction-associated steatotic liver disease (Rinella et al 2023).However, currently, very few studies demonstrate expertise in the differentiation of iPSC-derived hepatic-and non-hepatic tissue, forcing many researchers to combine PSC-derived cells with primary cells or even cell lines (Osonoi and Takebe 2024).
We demonstrate that the maturity of the iPSCderived liver cells can still be improved.We found that cell ratio of the different iPSC-derived cells was not important for further maturation of the cells, but mainly influenced spheroid formation.Medium compositions previously optimized for functionalized hydrogel cultures with similar iPSC-derived liver cells turned out to not have a large impact on maturity of the individual cell types in the spheroid cultures; only the high amino acid and L-Serine concentrations had a significant impact on CYP3A4 levels.Most probably, the hyperosmolar environment created by the L-Serine enforces a protective p53-independent quiescent state in the iPSC-derived hepatocytes, thereby promoting functional maturations (Chui et al 2024).Previous studies have already shown that growth factors can be (partly) replaced by small molecules, leading to successful differentiation of hepatocytes (Siller et al 2015, Pan et al 2022, Vanmarcke et al 2023).This suggests that the current co-culture differentiation method can still be further optimized and become independent of growth factors thereby rendering the cultures more robust and cost efficient.We found that the supplementation with growth factors, commonly added for the maturation and maintenance of iEC and iMΦ was not necessary in the co-culture setting.Most likely we can omit the supplementation of these growth factors in iPSCderived 3D co-cultures as the spheroids themselves show increased gene expression levels of VEGFB (De Smedt et al 2021), TGFA (Grotendorst et al 1989) and CSF (Van Wilgenburg et al 2013) while levels of VEGFA (De Smedt et al 2021) remain constant (figure S9).
Regarding variability, the spheroid cultures presented here have the general advantage of iPSCs, where inter-donor variability can be avoided or introduced in a controlled way, unlike donor-derived liver spheroids (Hurrell et al 2020).We compared the iPSC liver spheroids with primary human spheroids formed from unplated freshly isolated human liver cells and seeded using the same cellular ratios.We could demonstrate comparable mRNA levels of celltype specific genes in the two spheroid cultures.This suggests that iPSC-derived liver spheroids can be an alternative to primary human donor liver-derived spheroid cultures for toxicity screening and liver fibrosis modelling.
Another asset of the spheroid cultures described here is the incorporation of four different liver cell types, whereas many 3D in vitro liver models do not consist of all four main liver cell types.Many of these models do not include immune cells (Wang et al 2017, Lucendo-Villarin et al 2020, Kim et al 2023) or iECs (Wang et al 2017).Although not shown here, the presence of iECs in 3D co-cultures with iHeps improves the maturity of the iHeps (Ardalani et al 2019).
One drawback is the lack of cellular organization in the spheroids and the lack of vasculature in contrast to culture systems developed by others (Velazquez et al 2021, Kim et al 2023), which will be necessary to facilitate oxygen and nutrient supply if larger dense tissue structures or spheroids are needed.So far we have not observed a necrotic core in the spheroids, confirming earlier studies that in spheroids of less than 200-300 µm in diameter oxygen and nutrient supply occurs efficiently in the absence of vascularization (Folkman and Hochberg 1973).

Conclusion
In this study we developed a robust iPSC-derived liver co-culture system that is based on the incorporation of separately differentiated liver iHSCs, iHepatoblasts, iECs and iMΦ into spheroids.After incorporation, cells continue differentiating, reaching gene expression levels similar to spheroids derived from freshly isolated primary human liver cells.Here we show how these iPSC-derived liver spheroids can be used to study hepatocyte-damage mediated HSC activation, but they should also be suitable for drug and toxicity screenings and to model other liver diseases.
(A)) led to a clear increase of hepatoblast markers AFP and HNF1α.On protein level, almost all iHepatoblasts expressed HNF4α, whereas the expression of OCT4 is lost at the hepatoblast stage (figureS2(A)).For the differentiation of iPSCs towards iHSCs, we used a slightly modified version of a previously established protocol(Coll et al 2018), in which iPSCs were differentiated in 2D until day 10 and cells were passaged at least once (passage 1 iHSC) (figure1(B)).Passaged iHSCs can be stored frozen and used at P2 or 3 (Vallverdu et al 2021).During iHSC differentiation the expression of HSC markers PDGFRβ and DCN increased, whereas the expression of the pluripotency gene NANOG diminished over time.Passaged iHSCs expressed both αSMA and PDGFRβ proteins (figure 1(B)) and lost OCT4 expression (figure S2(C)).For the differentiation of iPSC-derived iECs we used ETV2-iPSCs that allow the doxycycline-dependent induction of ETS variant gene 2 (ETV2) (De Smedt et al 2021) (figure 1(C)).

Figure 1 .
Figure 1.Differentiation of iPSCs towards liver cells in 2D.Schematic representation of differentiation of iPSCs towards different liver cells with RNA and protein expression of cell type specific markers.Dashed lines indicate the expression level of the respective gene in freshly isolated unplated primary human cells (n = 3-4).(A) Differentiation of iPSC3X towards iHepatoblasts from day 0 to day 8. iHepatoblasts (day 8) stain positive for HNF4α, and express HNF1α and AFP and lack NANOG (n = 3).(B) iPSC differentiation to iHSCs.Passaged iHSCs express DCN, PDGFRβ and αSMA and lack NANOG (n = 3).(C) Differentiation of ETV2-iPSCs to iECs.iECs at day 14 stain positive for CD31 and express VWF (n = 3).(D) Differentiation of iPSCs towards iMC and iMΦ.Both iMC and iMΦ are positive for CD11b and mRNA levels of CD11b, CD16 and CD68 are increased in both iMC and iMΦ (n = 3).Scale bars in panel A represent 50 µm, while all other scale bars represent 100 µm.

Figure 2 .
Figure 2. Optimization of culture media conditions for 3D iPSC-derived liver spheroids.(A) Schematic representation of the assembly of 3D iPSC-derived liver spheroids and composition of differentiation media until day 13 of spheroid culture.(B) Brightfield images of spheroids kept in culture with different media for 13 days.Scale bar represents 100 µm.(C) mRNA levels of Hep marker gene CYP3A4, HSC marker gene PDGFRβ, MΦ marker gene CD68 and EC marker gene CD31 in spheroids that were kept in culture with different media for 13 days.(D) Schematic representation of iHSC activation in 3D using TGFβ for 48 h until day 13.(E) mRNA expression levels of HSC activation genes ACTA2, COL1A1 and LOXL2 after 48 h exposure to TGFβ until day 13 of culture (n = 3 with 6 spheroids per repeat).

Figure 3 .
Figure 3. Optimization of cell culture ratios of 3D iPSC-derived liver spheroids.(A) Ratios, brightfield images and H&E staining of spheroids with different seeding ratios after 21 days in 3D culture.Scale bars represent 50 µm.(B) mRNA levels of hepatocyte marker gene CYP3A4, HSC marker gene PDGFRβ, MΦ marker gene CD68 and EC marker gene CD31 in different ratios over time until day 21.The dashed lines represent the mRNA levels detected in freshly isolated uncultured human liver cells (n = 3-4).(C) mRNA expression levels of HSC activation genes ACTA2, COL1A1 and LOXL2 after 48 h exposure to TGFβ from day 11 to day 13 of culture (n = 6-9 with 6 spheroids per repeat).

Figure 5 .
Figure 5. Hepatotoxicity induction in iPSC-and primary human-derived spheroids.(A) Experimental setup for the induction of fibrosis using APAP.(B) Brightfield images showing concentration-dependent ruffling of spheroids due to 72 h exposure to APAP.(C) ATP viability assay after 72 h of exposure to different concentrations of APAP.Graph demonstrates amount of ATP relative to control.Dots represent the mean of 9 individual spheroids of a total of 5 independent repeats.(D) mRNA expression of COL1A1 and LOXL2 after 72 h exposure to 20 and 40 mM of APAP (n = 8, with 6 spheroids per individual repeat).(E) mRNA levels of hepatocyte marker gene CYP3A4, HSC marker gene PDGFRβ, MΦ marker gene CD68 and endothelial marker gene CD31 in iPSC-derived liver spheroids at day 21 of culture (n = 8, with 6 spheroids per repeat for the iPSC-derived spheroids; and n = 3 with 6 spheroids per repeat for the primary human liver spheroids).(F) Brightfield images of primary human spheroids showing concentration-dependent ruffling of spheroids after 24 h exposure to APAP.(G) ATP viability assay after 24 h of exposure to different concentrations of APAP.Graph demonstrates the % ATP relative to control.Dots represent the mean of 9 individual spheroids of a total of 3 individual repeats.All scale bars represent 100 µm.