Generation and characterization of mature hepatocyte organoids for liver metabolic studies

The liver is an organ with critical metabolic functions including absorbance, production and distribution of nutrients, detoxification, and bile production. These functions are processed in hepatocytes. The high prevalence of metabolic liver diseases has become a significant healthcare burden worldwide (Moon et al., 2020); among such diseases, non-alcoholic fatty liver disease (NAFLD) and alcoholic liver disease (ALD) are the primary causes of chronic liver disorders including liver cancers. However, studies of metabolic liver diseases and drug discoveries targeting liver metabolic functions have been predominantly carried out in genetic, chemically induced or diet-driven animal models (Van Herck et al., 2017). These in vivo approaches, although highly relevant physiologically, are limited by the high demands on resources and technical skills, and they are therefore not suitable for rapid and high-throughput screening for mechanistic or therapeutic discoveries in studies of human metabolic liver diseases. Furthermore, human-specific liver metabolism or disorders, such as hepatitis B virus (HBV) infection, cannot be readily studied in animal models (Hu et al., 2019).

Primary hepatocytes (PHs), as an in vitro model for mechanistic discovery in liver studies, are limited by quick loss of hepatic characteristics in two-dimensional (2D) culture (Thompson and Takebe, 2021; Yang et al., 2023). Therefore, various approaches have been developed for establishing and maintaining long-term hepatocyte cultures in vitro. These approaches include inducing hepatocyte-like differentiation from renewable pluripotent stem cells (PSCs) (Si-Tayeb et al., 2010; Takebe et al., 2013) or fibroblasts (Du et al., 2014; Huang et al., 2014); generating organoids from PHs, hepatoblasts or adult liver progenitor cells (LPCs) (Hendriks et al., 2023; Hu et al., 2018; Huch et al., 2015; Peng et al., 2018); or culturing hepatocytes as spheroids (Landry et al., 1985). These culture systems partially resemble hepatic functions and have been used in cellular, molecular and pharmacological studies of liver functions, regeneration, viral infection and drug screening (Liu et al., 2023; Thompson and Takebe, 2021; Yang et al., 2023). It appears that three-dimensional (3D) cultures of hepatocyte organoids (HOs), hepatocyte spheroids or hepatocyte-like cells induced from human PSCs in defined conditions better resemble mature hepatocytes in postnatal livers (Liu et al., 2023; Thompson and Takebe, 2021; Yang et al., 2023).

Self-organized 3D HO cultures provide a robust culture system that permits formation of advanced tissue structures, such as the bile canaliculi, and hepatocyte regulation in response to distinct stimulation (Hu et al., 2018; Liu et al., 2022; Peng et al., 2018). However, these HOs express low levels of metabolic genes characteristic of mature hepatocytes in postnatal livers. Instead, they express high levels of fetal hepatic or inflammatory genes, indicating that the hepatocytes in the HOs are immature, fetal or injury-induced hepatocyte-like cells. Moreover, human HOs derived from adult hepatocytes remain hard to expand (Hendriks et al., 2021; Hu et al., 2018), hampering their application in mechanistic investigations of mature hepatocyte metabolism in adult liver and therapeutic testing for liver diseases, particularly metabolic disorders (Hu et al., 2018; Peng et al., 2018; Shiota et al., 2021).

Many approaches have been taken to promote in vitro hepatic maturation of hepatocyte cultures, including HOs (Thompson and Takebe, 2021). In HOs, dexamethasone (DEX) has been shown to induce hepatic differentiation of LPCs and enhance the expression of metabolic genes (Huch et al., 2015; Peng et al., 2018). In addition, growth factor reduction and Wnt signaling activation might also be critical in inducing metabolic gene expression in the HOs (Peng et al., 2018). However, how each of these factors contribute to the maturation of HOs has not been systematically investigated. Moreover, long-term maintenance of differentiated HOs is challenging (De Crignis et al., 2021), highlighting the need to develop reliable culture conditions for hepatic maturation in HOs.

In the present study, we systemically and thoroughly tested the function of many factors involved in HO maturation, based on which we developed a new culture condition that contains an appropriate level of triiodothyronine (T3) to allow successful generation of organoids of both mouse and human origin with metabolic gene expression and regulation characteristic of adult livers. T3, the activated form of thyroid hormone, has been demonstrated to control developmental metamorphosis (Gilbert et al., 1996). Moreover, levels of circulating thyroid hormones rise more than 50-fold in newborn mice shortly after birth (Hirose et al., 2019), and thyroid hormone receptor β (THRβ) promotes the maturation of induced PSC (iPSC)-derived hepatocyte-like cells (Ma et al., 2022). However, to date it has remained unclear whether T3 is functionally critical for hepatic maturation of HOs, and its dose-dependent effects have not been investigated previously. Here, by using transcriptomic profiling and various functional assays, we show holistically that, compared to previously reported HOs, the mature hepatocyte organoids (MHOs) cultured in the medium we have developed exhibit hepatic metabolic functions – such as bile acid production, urea generation, detoxification, lipid and glucose metabolism, zonal hepatic gene expression, and expression of sodium taurocholate co-transporting polypeptide (NTCP, also known as SLC10A1), which is a transmembrane protein that is highly expressed in mature human hepatocytes and a mediator of bile acid transport that also serves as the receptor responsible for the cellular entry of both HBV and its satellite hepatitis delta virus (HDV) (Yan et al., 2012) – at robust levels that allow their use as physiologically relevant models for studies of liver biology and diseases. The maturation process of the organoids is reversible, and the organoids can be stored for long-term use. The MHOs also provide a more economic and efficient system that can be easily scaled up for high-throughput screens to identify and test therapeutic interventions for human liver diseases.

HOs can be generated from mature hepatocytes within two weeks of culture. However, these organoids exhibit high Afp expression levels and low metabolic gene expression (Hu et al., 2018), indicating hepatocyte dedifferentiation. These changes limit the use of HOs as a model for liver metabolic studies. We therefore decided to change the culture medium to generate MHOs in vitro. We first examined the various components in the published liver organoid culture media and found that they fell into three groups: Wnt agonists (RSPO1, CHIR-99021), growth factors [hereafter referred to as GFs; hepatocyte growth factor (HGF), FGF7, FGF10 and EGF] and the TGFβ type I receptor inhibitor A83-01. As TGFβ negatively regulates hepatocyte differentiation (Schaub et al., 2018), we retained TGFβ inhibitor in the culture medium. Wnt agonists and GFs are necessary for initial clone formation and expansion of hepatocytes (Hu et al., 2018). As activation of Wnt and/or GF signaling causes hepatocellular carcinoma, in which hepatocytes dedifferentiate (Llovet et al., 2021), we reasoned that these factors might inhibit hepatocyte maturation. We therefore completely removed Wnt agonists and GFs in the maturation culture medium, either individually or in combination (Fig. 1A). Strikingly, expression of the fetal hepatic marker Afp and the LPC marker Opn (also known as Spp1) was reduced after removal of GFs (RGF) (Fig. 1B; Fig. S1A), whereas expression of hepatic metabolic genes that are expressed in adult liver – such as Cyp7a1 and Cyp8b1, which encode critical enzymes in bile acid synthesis, and Arg1, which encodes a key enzyme in the urea cycle – was drastically induced by RGF, and less so by only removal of Wnt agonists (–Wnt activation; Fig. 1C,D; Fig. S1B). Interestingly, expression of Axin2, a transcription readout of Wnt signaling, was upregulated by RGF, suggesting that Wnt–β-catenin (CTNNB1) signaling activity was increased by RGF (Fig. S1C). Importantly, RGF or RGF with Wnt agonist removal did not reduce the expression of the hepatocyte-specific transcription factor Hnf4a (Fig. 1E), indicating that hepatocyte cell fate was maintained. We then determined whether the alteration in hepatic gene expression led to corresponding functional changes by examining bile acid and urea levels in the culture medium. Indeed, consistent with the robust induction of Cyp7a1 and Arg1 expression after RGF, bile acid and urea were produced and secreted to the medium after RGF and less so by only removal of Wnt agonists (–Wnt activation; Fig. 1F,G). In addition, compared to the control, removal of the ROCK1 and ROCK2 (collectively referred to as ROCK) inhibitor Y-27632 from the medium (–ROCKi) did not alter the expression of Afp, Cyp7a1 and Arg1, as well as bile acid and urea production (Fig. 1B–D,F,G). These results indicate that GFs, but not Wnt signaling, critically inhibit hepatocyte maturation in the organoid culture medium and that RGF effectively promotes hepatocyte maturation.

The HOs cultured with RGF expressed genes characteristic of metabolic functions in mature hepatocytes, but the expression of the fetal hepatic gene Afp was still high. We reasoned that addition of maturation-promoting factors was necessary. Although the concentration of T3 in the commercial B27 supplement made by Gibco is undisclosed, the B27 recipe has been reported in the literature (Bayerl et al., 2021), and the T3 concentration is ∼3 nM, which is the physiological T3 concentration in the serum of human and mice (Hulbert, 2000). As T3 concentration is higher in the liver than the serum (Hulbert, 2000), we added more T3 to the RGF culture medium (Fig. 2A). Indeed, additional T3 drastically reduced Afp expression and strongly promoted metabolic gene expression (Fig. 2B–D), and these changes were progressive and time dependent (Fig. S2A–C). In addition, as the glucocorticoid receptor (GR) is known to play critical roles in sensing metabolic changes and regulating hepatocyte metabolic functions (Okun et al., 2015; Rose et al., 2011) and the GR agonist DEX has been used in a recent study to promote HO maturation (Hu et al., 2018; Peng et al., 2018), we also tested DEX in the HO maturation medium (Fig. 2A). Consistent with previous findings (Hu et al., 2018; Peng et al., 2018), DEX inhibited Afp expression in the presence of GFs (Fig. S2D). In addition, we found that GFs showed dose-dependent effects in suppressing hepatocyte maturation, and that complete removal of GFs was required for more complete suppression of Afp expression and for expression of the bile acid metabolic genes Cyp7a1 and Cyp8b1, as well as for bile acid production (Fig. S2D,E). Notably, even low concentrations of GFs inhibited Cyp7a1 and Cyp8b1 expression as well as bile acid production, further confirming the necessity of RGF in HO maturation (Fig. S2D,E). Although T3 and DEX both drastically inhibited Afp expression, bringing it to low levels comparable to those in the freshly isolated PHs, they differed in their regulation of specific hepatic metabolic pathways (Fig. 2B). T3 had a stronger effect promoting the expression of Cyp7a1 to a level comparable to that in PHs, and DEX has a stronger effect promoting Arg1 expression (Fig. 2C,D). Consistent with these differential changes in metabolic gene expression, T3-treated HOs produced higher levels of bile acid, whereas DEX treatment promoted more urea production (Fig. 2E,F). Additionally, bile acid exportation was faster in the HOs cultured with T3 (Fig. S2F). As the liver plays a critical role in glucose homeostasis, we tested gluconeogenesis in the organoid cultures. T3 and DEX treatment individually or in combination improved glucose production in the gluconeogenesis assay (Fig. 2G), suggesting that they have additive roles in regulating glucose metabolism. Moreover, given the crucial role of mature hepatocytes of the adult liver in detoxification and drug metabolism (Thompson and Takebe, 2021; Yang et al., 2023), we tested the activities of representative enzymes involved in these processes. Again, T3 and DEX treatments had different effects. Whereas T3 could not further promote CYP1A2 activity after RGF, DEX treatment could. However, T3 and DEX together synergistically promoted CYP1A2 activity (Fig. 2H). In sharp contrast, 2D PH cultures under the same conditions failed to show similar induction or maintenance of metabolic enzymes (Fig. S2G). These results show that the 3D organoid cultures we have developed are superior to the previous PHs cultured in 2D conditions for studies of metabolic regulation in the liver.

The presence of polyploid hepatocytes is a unique phenomenon observed in the adult liver. To further demonstrate hepatocyte maturation in MHOs, we analyzed the hepatocyte ploidy. Compared with organoids in culture with GFs, organoids in various maturation conditions showed increased ratios of polyploidy. The ratios were also higher than those in PHs isolated from the liver of mice at postnatal day 28 (P28; Fig. 2I). The similarity in hepatocyte ploidy between in vivo hepatocytes and HOs cultured in our new maturation conditions further supports the hepatocyte maturation of HOs in these cultures. Therefore, our organoid culture system can also potentially be employed to investigate the regulation of diploid to polyploid conversion as hepatocytes mature – a phenomenon that is also important in liver regeneration and tumor formation.

It is known that glucocorticoids have the potential to induce lipid accumulation in the liver (Vegiopoulos and Herzig, 2007; Woods et al., 2015), whereas THRβ activation reduces lipid accumulation in the liver (Hatziagelaki et al., 2022). Consistent with previous findings, we found that T3 reduced lipid accumulation, whereas DEX strongly promoted it (Fig. 2J), and T3 attenuated DEX-induced lipid accumulation (Fig. 2J). In addition, albumin secretion in the RGF and T3 (RGF+T3) condition was comparable to that observed in the presence of GFs (+GF) (Fig. 2K), thus we selected the RGF+T3 condition for further investigation.

To further test hepatocyte maturation, we conducted bulk RNA-sequencing (RNA-seq) to profile the transcriptomic changes under different maturation conditions (Fig. 3A). We compared our new RNA-seq data for organoids and PHs to previously published RNA-seq data for mouse fetal hepatoblasts (FBs; GSE90047; Yang et al., 2017). We first established sets of fetal and mature hepatic markers, which functionally represent molecular features of hepatocytes (Fig. S3A, Table S1), by comparing the transcriptomic profiles between FBs and adult PHs. FBs isolated from different embryonic stages showed a progressive maturation process. In contrast, the whole transcriptomic profile of HOs cultured in the +GF condition represented a unique state (Fig. 3B). The transcriptomic profiling of fetal or mature hepatic gene expression showed that HOs in the +GF condition were more similar to FBs at embryonic day (E)18.5 in terms of fetal gene expression, whereas the expression of mature hepatic markers was more similar to that in the FBs at E15.5, compared to the PHs (Fig. S3B), indicating that the cells in PH-derived HOs were more similar to late embryonic stage FBs when cultured in the +GF condition. Therefore, the +GF condition caused PH dedifferentiation. Indeed, HOs with RGF had a transcriptional profile more similar to that of the PHs. The presence of T3 and DEX further enhanced this trend (Fig. 3B,C; Fig. S3B,C), which was further supported by gene set enrichment analysis (GSEA) and functional enrichment analysis (Fig. 3D,E). In addition, HOs in the same maturation medium showed progressive maturation in a time-dependent manner (Fig. 3C,E; Fig. S3B,C). There were no significant differences in transcriptomic profiles between 6-day and 12-day samples, suggesting that organoid maturation can be achieved in 6 days and maintained for a longer time period (Fig. 3C,E; Fig. S3B,C).

We further analyzed the maturation of HOs by comparing expression levels of the known key genes for hepatocyte identity and metabolic maturation (Fig. 3F). Indeed, HOs in all conditions expressed pan-hepatocyte genes. Importantly, expression levels of Afp and other fetal hepatic genes were drastically reduced in the presence of T3 and/or DEX. Notably, the expression of Cdkn1a (which encodes p21) and Cdkn2a (which encodes p16), as well as the inflammatory cytokines and chemokines, such as Cxcl1, Cxcl2, Ccl2 and Tnf, which are involved in the senescence and senescence-associated secretory phenotype (SASP) (Gorgoulis et al., 2019), was drastically reduced in the presence of T3 and DEX, suggesting that hepatocytes in the MHOs did not undergo senescence (Fig. 3G). Consistent with our initial findings, the expression of genes involved in many hepatic metabolic pathways – including bile acid production and secretion, the urea cycle, lipid metabolism, and glucose metabolism – was strongly induced by T3 and DEX treatment, to levels comparable to those in PHs (Fig. 3H–K). These data suggest that at the transcriptomic level, RGF with T3 and DEX addition induces holistic hepatocyte maturation with profound changes in gene expression. T3 and DEX treatments showed distinctive and overlapping functions in promoting hepatocyte maturation in HOs.

The healthy liver has remarkable regenerative ability following partial hepatectomy or acute liver injury due to the proliferative response of mature hepatocytes to external stimuli (Fausto, 2000). We therefore asked whether the MHOs exhibit similar regeneration and proliferation potential (Fig. 4A). We did not observe any Ki67 (MKI67)-positive cells, significant change in cell count or 5-ethynyl-2′-deoxyuridine (EdU) incorporation in the MHOs, indicating that these cells were not proliferative (Fig. 4B; Fig. S4A,B). In addition, compared to the +GF control HOs, no significant difference in cell viability was observed in the MHOs (Fig. S4C), suggesting that the hepatocytes in the MHOs were in a viable and quiescent state. However, upon switching from the maturation medium to a growth medium containing GFs, mature hepatocytes in the MHOs quickly re-entered the cell cycle (Fig. 4B). The ratio of Ki67-positive cells in these organoids was drastically increased to levels similar to those in HOs prior to maturation induction (Fig. 4B). Therefore, when MHOs were switched to the +GF medium, the cells in the HOs were likely to have undergone re-dedifferentiation (re-DeDiff). To further characterize the changes, we performed transcriptomic analyses for these HOs (Fig. 4C,D). Surprisingly, the expression of fetal hepatic markers was higher in the re-DeDiff HOs than in the original HOs in the +GF condition (Fig. 4D), suggesting that MHOs showed a more robust response to GFs. The increased sensitivity to GF treatment could be due to epigenetic changes gained during in vitro culture. Therefore, the cells in the HO gained more plasticity. Although the expression of fetal hepatic genes was higher in the re-DeDiff HOs, principial component analysis (PCA) indicated that the transcriptomic profiles of HOs in +GF and re-DeDiff conditions were very similar (Fig. 4C). This observation was exemplified by the similar expression levels of fetal and hepatic markers, as well as by the results of GSEA analysis (Fig. 4D; Fig. S4D). In the re-DeDiff HOs, expression of metabolic genes found to be expressed in mature hepatocytes, such as Arg1, Cyp7a1 and Cyp8b1, was inhibited, and the expression levels of fetal hepatic genes such as Afp were dramatically elevated again (Fig. 4D). These data indicate that the hepatocyte maturation process is completely reversible in HO culture.

Hepatocytes in the adult livers are tightly regulated by many signaling pathways to maintain homeostasis and physiological functions. Among these, the Hippo signaling pathway plays critical roles in maintaining liver organ size and hepatocyte cell fate (Russell and Camargo, 2022). Activation of YAP (also known as YAP1), the transcription transducer of Hippo signaling, in hepatocytes leads to enhanced cell proliferation and hepatocyte dedifferentiation followed by biliary epithelial cell (BEC)-like cell differentiation (Liu et al., 2022; Yimlamai et al., 2014). We employed the HOs from TetO-YAP mice (Yimlamai et al., 2014) to induce expression of YAP^S127A (which encodes human YAP with an activating mutation) by Cre and doxycycline treatment in growth and maturation medium (Fig. 4E). Indeed, in the original +GF medium, we found that YAP activation strongly induced Krt19 expression and reduced Hnf4a expression (Fig. 4F), suggesting an LPC to BEC-like cell conversion. In addition, in MHOs, YAP activation induced cell proliferation (Fig. 4G), robust expression of the YAP target gene Cyr61 (also known as Ccn1), the fetal liver marker gene Afp, and the LPC and BEC markers Sox9 and Opn, whereas the expression of the BEC marker gene Krt19 was not elevated (Fig. 4F; Fig. S4E). These findings suggest that YAP activation in MHOs first induces hepatocyte dedifferentiation but not BEC gene expression, recapitulating changes that are observed in vivo (Liu et al., 2022).

It is known that expression of hepatic metabolic genes is critically regulated by Wnt–β-catenin signaling (Perugorria et al., 2019). Wnt–β-catenin signaling is required for metabolic zonation by inducing the expression of pericentral metabolic genes and inhibiting periportal gene expression in vivo (Benhamouche et al., 2006). Activation of Wnt signaling is also commonly observed in liver tumors (Perugorria et al., 2019). Overexpressing constitutively activated β-catenin (CTNNB1^ΔN90) via adeno-associated virus (AAV) infection activated Wnt–β-catenin signaling in the HOs in both growth medium and maturation medium, as determined by measuring Axin2 expression (Fig. 4H,I). Interestingly, expression of the pericentral metabolic gene Glul was upregulated upon CTNNB1^ΔN90 overexpression in both HOs in the +GF condition and MHOs, but more robust upregulation was induced in the MHOs, whereas Afp expression was not upregulated (Fig. 4I), suggesting that activation of Wnt–β-catenin signaling in MHOs may induce pericentral hepatocyte fate. To further test this, we performed transcriptomic analysis of HOs overexpressing CTNNB1^ΔN90. We found that pericentral and periportal markers were both induced in the MHOs. Strikingly, pericentral markers were more robustly upregulated and periportal markers were drastically inhibited in the MHOs with CTNNB1^ΔN90 overexpression (Fig. 4J,K; Fig. S4F, Table S1). In contrast to the effects of YAP activation, CTNNB1^ΔN90 overexpression only strongly induced the expression of the fetal hepatic gene Afp and cell proliferation in HOs cultured in the +GF condition but failed to do so in the MHOs (Fig. 4I,L). In line with this finding, overexpression of CTNNB1^ΔN90 only induced a drastic upregulation of fetal hepatic marker genes in HOs cultured in the +GF condition (Fig. S4G). These finding are consistent with previous findings in vivo that Wnt signaling regulates hepatic metabolic gene expression in the adult liver and promotes cell proliferation in hepatoblastoma, which occurs in early childhood.

In the liver, expression of Wnt-induced metabolic genes such as Glul and Cyp2e1 is extremely high in the first few layers of cells juxtaposing the central vein (Benhamouche et al., 2006; Halpern et al., 2017). We therefore asked whether a gradient of Wnt signaling activity regulates the expression of distinct metabolic genes in different metabolic zones in the MHOs. Indeed, high concentrations of the Wnt agonist CHIR-99021 (15 µM) in the MHO culture medium strongly promoted expression of the Wnt–β-catenin target gene Axin2 and the pericentral genes Glul, Cyp7a1, Cyp2e1 and Cyp1a2, whereas it suppressed expression of the mid-zone and periportal genes Cyp8b1 and Arg1. Furthermore, progressive reduction of CHIR-99021 concentration reversed this trend (Fig. 4M). These data indicate that the MHOs can be induced to express metabolic genes characteristic to distinct metabolic zones by fine tuning Wnt signaling activity in the culture medium, and that the MHOs we generated are powerful tools to dissect the regulatory gene circuitry that determines metabolic zonation.

The liver is an organ with critical metabolic functions, and among these functions, bile acid production is uniquely important (Chiang and Ferrell, 2019). It has been very difficult to study bile acid metabolism in PHs as they dedifferentiate quickly (Thompson and Takebe, 2021). Because MHOs produced bile acid and expressed genes involved in bile acid synthesis, we tested whether regulation of bile acid metabolism in the MHOs is similar to that in vivo. It is known that bile acid production in the liver is tightly controlled by a negative feedback loop involving FGF15 (in mouse, ortholog of human FGF19)–FGFR4 signaling and the nuclear receptor FXR (also known as NR1H4; Chiang and Ferrell, 2019). Bile acid acts through activation of FXR in hepatocytes to inhibit Cyp7a1 expression (Goodwin et al., 2000; Lu et al., 2000). We treated the MHOs with human FGF19 or the FXR agonist GW4064 (Fig. 5A). Both FGF19 and GW4064 strongly inhibited Cyp7a1 and Cyp8b1 expression (Fig. 5B,C), as well as bile acid production (Fig. 5D,E). Interestingly, the expression of the FXR target gene Shp (also known as Nr0b2), which mediates FXR-induced Cyp7a1 inhibition (Goodwin et al., 2000; Lu et al., 2000), was strongly induced by both FGF19 and GW4064 treatment (Fig. 5B,C). Loss of Fxr in the MHOs abolished the gene expression changes induced by GW4064, but not those induced by FGF19 (Fig. 5B–E), indicating that FGF19 can act via FXR-independent pathways to regulate bile acid metabolism. Additionally, consistent with previous findings in vivo (Sinal et al., 2000), the Fxr^−/− MHOs, but not the Fxr^−/− fetal-like HOs, showed lipid accumulation and elevated expression of genes involved in lipid synthesis (Fig. 5F,G). These findings show that the MHOs we have generated in vitro could complement or even replace the in vivo mouse models for genetic and pharmacological assays in bile acid metabolism studies.

We next investigated whether MHOs could serve as a reliable in vitro model for studying NAFLD and ALD, two important liver metabolic diseases. To mimic the pathological condition of NAFLD, we treated the MHOs with oleic acid (OA), a common free fatty acid (FFA) found in the diet (Fig. 5H). Importantly, only the MHOs, not the HOs in +GF conditions, developed severe steatosis (Fig. 5I). Peroxisome proliferator-activated receptors (PPARs) are ligand-activated transcription factors that are activated by fatty acids and regulate lipid metabolism (Montaigne et al., 2021). Because PPARα (also known as PPARA) is the predominant PPAR isoform in the liver (Montaigne et al., 2021), we therefore tested PPARα activity in the HOs. Indeed, expression of the PPARα target genes Fabp1 and Cpt1a was only strongly induced in MHOs and was not induced in fetal liver-like organoids (Fig. 5J), indicating a higher intracellular FFA level and lipid metabolic activity in the MHOs. In addition, expression of Fasn and Scd1, which are genes involved in lipid synthesis, only responded significantly to OA treatment in the MHOs (Fig. 5J). Therefore, the MHOs can be used as an in vitro model for NAFLD studies.

We then tested whether MHOs could also be used as an in vitro model for ALD studies (Fig. 5K). Similar to the effects of FFA treatment, ethanol-induced steatosis was more significant in the MHOs (Fig. 5L; Fig. S5A), probably due to increased expression of ethanol metabolic genes in the MHOs (Fig. S5B). In addition, alcohol-induced liver damage is known to be associated with inflammation (Wang et al., 2012). We next tested whether alcohol treatment could directly induce hepatocyte injury in HOs. Because serum alanine aminotransferase (ALT) activity is a biomarker for liver injury, and because expression of Gpt, which encodes an ALT isozyme, was drastically increased in all MHOs (Fig. S5C), we tested ALT activities in the medium as readouts for hepatocyte injury. Interestingly, ALT levels were increased by ethanol treatment more robustly in the MHOs, although low levels of ALT activity could also be detected in the culture medium of untreated MHOs (Fig. 5M). These results show that alcohol treatment directly and robustly induces hepatocyte injury in MHOs, which might trigger and/or shape downstream inflammatory responses.

We next evaluated whether MHOs can be used as an in vitro model to study drug-induced liver injury (DILI). Acetaminophen (APAP), a commonly used antipyretic and analgesic drug, is one of the most common drugs that can cause DILI (Yoon et al., 2016). The hepatotoxicity of APAP mainly depends on CYP2E1, a cytochrome P450 that metabolizes many low-molecular-mass toxins (Lee et al., 1996). CYP2E1 metabolizes ∼10% of APAP to form the toxic electrophilic intermediate N-acetyl-p-benzoquinoneimine (NAPQI), which can be neutralized by conjugation with glutathione (GSH). APAP overdose leads to depletion of GSH and accumulation of NAPQI, leading to mitochondrial damage and cell death. Clinically, N-acetylcysteine (NAC) is used as an antidote for APAP overdose before severe liver injury happens because it can restore cellular GSH (Yoon et al., 2016). In vitro study of DILI is challenging due to a lack of cytochrome P450 activity in hepatoma cell lines. For example, APAP only induces limited cell death in the HepG2 cell line, and the cell death is not dependent on the CYP2E1-induced NAPQI accumulation (Behrends et al., 2019). Interestingly, we found that Cyp2e1 expression was upregulated in the MHOs and could be further increased by DEX and Wnt activation (Figs 3F and 4K,M). Therefore, we compared APAP-induced hepatotoxicity for HOs in the +GF, RGF+T3, and RGF and DEX (RGF+DEX) conditions with Wnt activation (Fig. 5N). In addition, we noticed that NAC, which is a known antidote for APAP overdose, is supplemented in the basal medium and could also be used to test whether APAP induces NAPQI-dependent cell death in the HOs. We tested the APAP-induced hepatotoxicity in medium with or without NAC (Fig. 5N). Indeed, APAP induced cell death with or without NAC in HOs in all conditions (Fig. 5O). However, cell death was only rescued by NAC in MHOs under the RGF+T3 and RGF+DEX conditions (Fig. 5O), suggesting that APAP induced NAPQI-dependent cell death only in MHOs. Notably, APAP induced the most severe cell death, with the best rescue by NAC, in the MHOs in the RGF+DEX medium (Fig. 5O,P), suggesting that MHOs in the RGF+DEX medium are the most suitable for studying APAP-induced hepatotoxicity.

Inherent differences in metabolism between humans and mice might complicate the translation of our findings made in mice. Therefore, we decided to test whether the culture conditions we developed could be used to promote maturation of fetal human HOs (fhHOs). We first established fhHO cultures from three individual donors using previously reported culture conditions (Hendriks et al., 2023), which resulted in fetal-like HOs (Fig. 6A–D). After that, hepatic maturation of fhHOs was initiated (Fig. 6A–C). The hepatic maturation in fhHOs was also induced by both T3 and DEX, as evidenced by decreased AFP expression and increased expression of CYP7A1, CYP8B1 and ARG1 (Fig. S6A). However, unlike mouse HOs, human hepatic maturation was synergistically induced by both T3 and DEX (Fig. 6D; Fig. S6A). As expected, hepatocytes in the human MHOs (hMHOs) did not proliferate in the culture medium after 9 days (Fig. 6E). Concurrently, AFP expression was significantly inhibited, while expression of metabolic genes such as CYP7A1, CYP8B1 and ARG1 was markedly induced (Fig. 6D; Fig. S6A,B). Consistently, production of bile acid and urea in the culture medium exhibited a significant increase (Fig. 6F,G), and the expression and activity of detoxification enzymes such as CYP3A4 and CYP1A2 were dramatically elevated in the maturation medium (Fig. 6H,I; Fig. S6C). Compared to other media recently used for hepatic maturation of fhHOs, which involve the use of T3 or a commercially available serum-free medium, HepatoZYME (Ma et al., 2022; Wesley et al., 2022), we found that our combination of both RGF and a low concentration of T3, which was first adopted by our studies, is required to promote holistic metabolic maturation based on gene expression and functional assays such as bile acid production and urea metabolism (Fig. S6D–I). Similar to mouse MHOs, hMHOs demonstrated a robust response to FGF19 and GW4064, as indicated by strong inhibition of CYP7A1 expression and elevated SHP expression, along with reduced bile acid production in the medium (Fig. 6J–L; Fig. S6J). Additionally, glycogen storage and albumin secretion were enhanced in the hMHOs (Fig. 6M,N). These findings suggest that, in comparison to previously reported culture conditions, the combination of RGF, T3 and DEX synergistically induces formation of hMHOs.

To further assess whether hMHOs serve as a powerful model for human metabolic disorders, we treated the organoids with FFA or ethanol (Fig. 6O). Consistent with previous findings (Hendriks et al., 2023), FFA treatment induced steatosis under the previously reported culture conditions (control). However, lipid accumulation was significantly more severe in the hMHOs (Fig. 6P,Q). Similar to mouse MHOs, the hMHOs exhibited strong induction of CPT1A and FABP1 expression (Fig. S6K). Surprisingly, ethanol treatment induced steatosis only in the hMHOs (Fig. 6R,S). In addition, ALT activity was only elevated after ethanol treatment in the hMHOs (Fig. 6T). This observation could be attributed to the strongly induced expression of ethanol metabolic genes in the hMHOs (Fig. S6L). These findings indicate that hMHOs are more responsive to metabolic stimuli, such as FFA and ethanol, and may serve as a more robust model for human metabolic disorders compared to HOs cultured under previously reported conditions.

HBV infection can cause severe liver diseases ranging from fibrosis to cancer (Llovet et al., 2021). Highly relevant and representative in vitro culture models that can be used to study HBV biology and pathology, particularly long-term virus–host interactions, and to test effective antiviral therapies are still lacking. Current HBV infection assays are primarily based on human PH cultures and NTCP-overexpressing HepG2 cells (HepG2-NTCP cells; Ni et al., 2014; Nkongolo et al., 2014), which cannot accurately mimic human pathophysiological conditions (Li et al., 2020). Recently, HOs derived from adult human LPCs have been shown to be a model for HBV infection (De Crignis et al., 2021). However, HBV infection efficiency in fhHOs is not clear. As mature hepatocytes exhibit greater efficiency in supporting HBV infection, we tested whether hMHOs differentiated from fhHOs are also more susceptible to HBV infection. We found that expression of NTCP, which encodes the receptor for HBV entrance into hepatocytes (Yan et al., 2012), was upregulated in the hMHOs (Fig. 7A). Indeed, when infected by HBV at multiplicity of infection (MOI) ranging from 100 to 1000, hMHOs better supported HBV infection, resulting in a significant increase in levels of HBV DNA, HBV total RNA and HBV 3.5 kb RNA within the cells, compared to levels in the fetal-like HOs cultured in the control medium (Fig. 7B–D; Fig. S7A–H). These results indicate that the hMHOs can effectively support HBV infection, leading to robust HBV covalently closed circular DNA (cccDNA) transcriptional activity.

In this study, we developed a novel culture condition that permitted generation of MHOs from hepatocytes of both mouse and human origins. Compared to previously reported HOs, the transcriptomic profiling and a diverse array of functional assays showed that MHOs exhibit improved metabolic functions that are characteristic of adult mouse or human livers, including bile acid production, urea generation, detoxification, and lipid and glucose metabolism. Our data indicate that the MHOs are powerful in vitro tools for studies of adult liver biology and diseases. Complementary to various animal models, the hMHOs are particularly powerful, with many direct research applications that will help to improve understanding of human liver biology and diseases.

HO cultures have previously been established from both mouse and human cells, and they mimic the physiological architecture and some functions of the liver parenchyma, such as detoxification, lipid transport, bile acid transport, and glycogen storage (Hendriks et al., 2023; Hu et al., 2018; Kimura et al., 2022; Peng et al., 2018; Shinozawa et al., 2021). HOs therefore show potential as a tool to study liver steatosis, cholestasis and even HBV infection, as well as for liver regeneration (De Crignis et al., 2021; Hendriks et al., 2023; Kimura et al., 2022; Peng et al., 2018; Shinozawa et al., 2021). However, a common issue arising from the previously published culture methods is the fetal-like phenotype of the HOs, which have very low expression of hepatic metabolic genes when compared to the expression in PHs isolated from adult livers (Shiota et al., 2021; Thompson and Takebe, 2021; Yang et al., 2023), hampering their use in studying adult liver biology and diseases. In this study, the mouse and human MHOs generated by the novel culture conditions we have developed showed superior performance in various established assays of liver metabolism. As proof of principle, we showed that the human and mouse MHOs can be used to study bile acid production, zonal metabolic gene expression, NAFLD, ALD, HBV infection, oncogenic liver changes and more.

The maturation condition we developed for HO culture allows reproducible, rapid and robust hepatic maturation. RGF is essential for HO maturation. Consistently, hepatic metabolic function is inhibited by GF-induced hepatocyte proliferation during liver regeneration (Fausto, 2000). Previously, fhHOs have been successfully used as a model for mimicking human liver steatosis (Hendriks et al., 2023). However, the reported HOs are likely still fetal-like with low expression levels of hepatic metabolic enzymes, potentially limiting their applications. Compared to the fhHOs cultured under the previously reported conditions (Hendriks et al., 2023), our hMHOs have a higher sensitivity in response to FFA treatment (Fig. 6P,Q). Importantly, considering the many critical metabolic pathways in hepatocytes of adult livers, such as bile acid and urea metabolism, our culture condition that allows holistic upregulation of metabolic genes has a distinct advantage in mimicking normal physiological conditions as well as modeling human cells. For instance, the FGF15 (FGF19 in humans)–FGFR4–ERK signaling pathway is pivotal for suppressing bile acid production by inhibiting Cyp7a1 expression (Inagaki et al., 2005). Other GFs might crosstalk with this regulatory network, as ERK signaling is a common downstream signal transducer. Indeed, we found that RGF strongly upregulated Cyp7a1 expression and bile acid production (Fig. 1C,F). As such, the MHOs can potentially be combined with gut organoid cultures to investigate the gut–liver FGF15/FGF19–bile acid axis in vitro without relying on mouse genetic models. Notably, many genes involved in bile acid metabolism are regulated in hMHOs similarly to the in vivo setting. Both FGF19 and FXR agonist suppressed the expression of CYP7A1 and reduced bile acid production (Figs 5B–E and 6K,L), demonstrating the potential of this system in studies of bile acid metabolism in vitro.

Here, we have shown that a critical factor for the promotion of hepatocyte maturation in MHOs is T3. THRβ is the most abundant thyroid hormone receptor in hepatocytes, and it mediates T3 activities. THRβ is known to be critical for regulating hepatic metabolic functions such as de novo lipogenesis, β-oxidation of fatty acids, mitophagy and cholesterol synthesis (Hatziagelaki et al., 2022). Recently THRβ activation has been found to promote hepatic maturation of iPSC-derived hepatocyte-like cells (Ma et al., 2022). Indeed, we found that T3 induced CYP3A4 expression and reduced AFP expression in hMHOs. However, high concentrations of T3 (3 µM) and HGF, as used in previously reported cultures (Ma et al., 2022), inhibited CYP7A1 and ARG1 expression, resulting in low bile acid and urea production (Fig. S6D–F). DEX, as a GR agonist that suppresses inflammation, is widely used in vitro for hepatic differentiation and maturation (Thompson and Takebe, 2021). DEX has been added to the medium to promote HO differentiation (Hu et al., 2018; Peng et al., 2018). Indeed, we and others have shown that DEX inhibits AFP expression in the presence of GFs (Fig. S2D) (Hu et al., 2018; Peng et al., 2018). However, bile acid metabolic gene expression and bile acid production are still inhibited by GFs, even with greatly reduced concentrations (Fig. S2D,E). Additionally, we found that DEX induced a significant steatosis phenotype in the HOs when combined with RGF, limiting its application in HO cultures that maintain normal hepatocyte phenotypes. Even a low concentration (200 nM) of DEX induced severe lipid accumulation in the organoids (Fig. 2J). It is important to note that we showed that T3 with DEX in the RGF medium greatly attenuated lipid accumulation (Fig. 2J). As the basal B27 supplement already contains cortisone to activate the GR, adding only T3 to the RGF medium is sufficient to induce HO maturation. Although the RGF+T3 condition was used in most of our experiments, the RGF+DEX condition showed better sensitivity in assays of APAP-induced hepatoxicity (Fig. 5N–P), whereas the RGF+T3+DEX condition showed improvement in urea production, glucose production and CYP1A2 activity (Fig. 2F–H). Therefore, different MHO culture conditions may be chosen depending on the assays to be performed. In the hMHOs, T3 and DEX synergistically improved hepatic maturation without significant lipid accumulation (Fig. 6P,Q). We think that T3, DEX and RGF play similar roles in promoting hepatic maturation of both human and mouse HOs. Although mouse adult hepatocytes were used to form HOs, in the +GF condition, the mature hepatocytes were dedifferentiated (Fig. 3C; Fig. S3B). In contrast, the human HO cultures were directly derived from human hepatoblasts, and the presence of GFs prevented hepatocyte maturation. Therefore, both human and mouse HOs cultured in the +GF medium maintained hepatocytes in an immature state, which can be induced to differentiate by DEX, T3 and RGF. The culture medium we developed here performs better than HepatoZYME, a commercially available serum-free medium, in promoting differentiation of fhHOs (Wesley et al., 2022). HepatoZYME induced mild downregulation of AFP expression and upregulated CYP7A1 expression as well as bile acid production, similar to the levels observed in human HOs with RGF medium (Fig. S6G,H). However, ARG1 expression and urea production in the HOs cultured in HepatoZYME were still much lower than that observed in the RGF medium (Fig. S6G,I). Importantly, adding T3 and/or DEX to the HepatoZYME did not improve its activity in promoting hepatocyte maturation (Fig. S6G–I). Therefore, the maturation medium we developed is fundamentally different from HepatoZYME.

MHOs can be used to investigate human liver metabolic regulation and diseases. The hMHOs, compared to the mouse MHOs, displayed a more dramatic increase in lipid accumulation within the organoids following FFA treatment (Fig. 6P,Q). The enhanced sensitivity to FFA treatment and robust changes to the expression profiles of genes involved in lipid metabolism indicate that hMHOs are invaluable in vitro tools for studying human NAFLD. Interestingly, the hMHOs were also more sensitive to ethanol treatment in terms of lipid accumulation (Fig. 6R,S). Therefore, long-term culture of hMHOs represents a complementary and human-relevant model (compared to the current animal models) for investigating human metabolic disorders including human ALD and NAFLD.

It is known that liver metabolic zonation is functionally important (Benhamouche et al., 2006; Halpern et al., 2017). As a proof of principle, we found that β-catenin activation induced pericentral hepatocyte fate in the MHOs (Fig. 4I–K), as found in vivo (Benhamouche et al., 2006). Moreover, the GSK3β inhibitor CHIR-99021 altered expression of zonal metabolic genes in a dose-dependent manner (Fig. 4M), suggesting that a Wnt signaling gradient patterns the liver metabolic zonation. These data suggest that MHOs offer a good in vitro system to dissect the regulation of zonal metabolic gene expression. Importantly, HO maturation is reversible. By switching to a culture medium with GFs, MHOs can be made to go back to a fetal-like HO state with proliferative hepatocytes. As Wnt signaling controls both liver zonal metabolism and tumorigenesis, the effects of Wnt signaling on growth promotion and metabolic regulation can be readily tested in MHOs. Thus, MHOs provide a valuable in vitro tool that previously was not available for the study of metabolic and/or oncogenic roles of critical signaling pathways in the liver. The hMHOs can also be genetically engineered to model human liver diseases and to study the underlying regulatory pathways. The system can also easily be scaled up for use in high-throughput genetic or chemical screening, which are essential for drug discovery and testing.

The identification of human NTCP as the HBV receptor has led to the generation of HepG2-NTCP cell lines, which have significantly contributed to uncovering certain aspects of the HBV life cycle (Guo et al., 2023; Ni et al., 2014; Nkongolo et al., 2014; Yan et al., 2012). However, HepG2-NTCP cells present real challenges in terms of oncogenic transformation. The hMHOs we generated exhibited much increased NTCP expression levels and higher susceptibility to HBV infection, suggesting their potential as a useful model in investigating virus–host interactions and the HBV life cycle, including intercellular virus spreading.

Animal protocols and procedures were approved by the Harvard Medical School Institutional Animal Care and Use Committee. Mice were housed in pathogen-free facility in a 12 h light–dark cycle. Both male and female mice were used for experiments. Rosa26^{lox−stop−lox−rtTA/+}; Col1a1^{Teto−YapS127A/+} (TetO-YAP) mice were obtained from Dr Fernando Camargo's lab (Boston Children's Hospital, Boston, MA, USA; Yimlamai et al., 2014). The Fxr knockout mice (JAX:004144) and B6J (JAX:000664) mice were ordered from The Jackson Laboratory.

Two-step liver perfusion was performed. Briefly, the mouse heart and liver were surgically exposed after anesthesia, a 25G infusion set (Terumo) was immediately inserted into the right atrium, and the portal vein was cut with scissors at the same time. Mouse livers were then perfused sequentially with pre-warmed (37°C) liver perfusion medium (Thermo Fisher Scientific) at a flow rate of 3 ml/min for 2 min, and pre-warmed (37°C) digestion solution (0.75 mg/ml collagenase type I in DMEM; Thermo Fisher Scientific) was perfused at the same flow rate for 3 min. During the perfusion, the portal vein was manually occluded by cotton tips every 30 s for 10 s. After perfusion, the liver was surgically removed and cut into small pieces in ice-cold DMEM, and hepatocytes from the digested tissue were released into the medium by gentle pipetting. Next, hepatocytes were passed through a 70 μm cell strainer (Corning) and centrifuged at 50 g for 3 min. The viable hepatocytes were purified using 40% Percoll (Cytiva) and centrifuged at 150 g for 5 min. Freshly isolated mouse hepatocytes were cultured in collagen I-coated plates (Corning) using the RGF medium for HO cultures (see below).

Human fetal liver was obtained from medically or elective indicated termination of pregnancy through a nonprofit intermediary working with outpatient clinics (Advanced Bioscience Resources, ABV, Alameda, CA, USA). Written informed consent of the maternal donors was obtained in all cases, under regulations governing the clinic. Human fetal hepatocytes were isolated from three donors. Briefly, 19- to 21-week gestation period human fetal livers (Advance Bioscience Resources) were cut into small pieces and digested with liver digest medium (Thermo Fisher Scientific) at 37°C for 30 min. The connective tissue was removed by passing the digested tissue through a 70 μm filter (Corning). Hepatocytes were obtained by two centrifugation steps at 4°C: 250 g for 10 min and 20 g for 5 min. Red blood cells were then removed with red blood cell lysis buffer (Invitrogen), and hepatocytes were washed with complete Iscove's modified Dulbecco's medium (Invitrogen). The project was reviewed by the University of Maryland's Office of Human Research Ethics, which determined that this submission does not constitute human subjects research as defined under federal regulations [45 CFR 46.102(d or f) and 21 CFR 56.102(c), (e) and (l)].

Mouse and human fetal HO culture was performed as previously described (Hendriks et al., 2021; Hu et al., 2018). Hepatocytes isolated from both male and female mice at the age of 6–10 weeks were used generate HOs. Briefly, freshly isolated hepatocytes were counted and mixed with the Cultrex Reduced Growth Factor Basement Membrane Extract (BME), Type R1 (R&D Systems). A total of 100,000 cells was mixed with 50 μl of BME and seeded in a cell culture plate. After the BME was solidified, the HO growth medium (+GF medium) was added to the plate. The +GF medium for mouse HO culture contained Advanced DMEM/F12 (Thermo Fisher Scientific) with B27 minus vitamin A (Thermo Fisher Scientific), 10 mM HEPES (Thermo Fisher Scientific), GlutaMAX (Thermo Fisher Scientific), 50 μg/ml Primocin (InvivoGen), 15% RSPO1-conditioned medium [home-made using Cultrex HA-R-Spondin1-Fc 293T cells (Bio-Techne), according to the manufacturer’s instructions], 3 μM CHIR-99021 (MCE), 50 ng/ml EGF (Biolegend), 25 ng/ml HGF (Biolegend), 50 ng/ml FGF7 (Peprotech), 50 ng/ml FGF10 (Biolegend), 1 μM A83-01 (Sigma), 5 μM ROCK inhibitor Y-27632 (MCE), 10 nM gastrin (Sigma), 10 mM nicotinamide (Sigma), and 1 mM N-acetylcysteine (Sigma). To culture human fetal HOs, cryopreserved human fetal hepatocytes were retrieved, resuspended in the +GF medium, and mixed with the Cultrex BME, Type R1 (R&D Systems) at a volume ratio of 1:5. A total of 30,000 cells was mixed with 50 μl of BME and seeded in the cell culture plate. After the BME was solidified, the +GF medium for human HO culture was added to the plate. The +GF medium for human HO culture contained Advanced DMEM/F12 (Thermo Fisher Scientific) with B27 minus vitamin A (Thermo Fisher Scientific), 10 mM HEPES (Thermo Fisher Scientific), GlutaMAX (Thermo Fisher Scientific), 50 μg/ml Primocin (InvivoGen), 15% RSPO1-conditioned medium (home-made), 3 μM CHIR-99021 (MCE), 50 ng/ml EGF (Biolegend), 50 ng/ml HGF (Biolegend), 50 ng/ml FGF7 (Peprotech), 50 ng/ml FGF10 (Biolegend), 20 ng/ml TGFα (BioLegend), 1 μM A83-01 (Sigma), 5 μM ROCK inhibitor Y-27632 (MCE), 10 nM gastrin (Sigma), 2.5 mM nicotinamide (Sigma) and 1.25 mM N-acetylcysteine (Sigma). Before experimental manipulation, the HO cultures underwent one passage to enrich well-formed HOs. HOs were retrieved from the BME using Cultrex Organoid Harvesting Solution (R&D Systems). Well-formed organoids were picked from the bottom of a well and further washed in Advanced DMEM-F12 and re-embedded into the BME for subsequent culture. The cystic bile duct organoids were hand-picked or further removed by 36% Percoll (Cytiva) by centrifugation at 150 g for 5 min during passage.

To test the culture conditions for mouse HO maturation, several conditions were evaluated in the experiments. The ‘−Wnt activation' medium refers to the removal of the Wnt agonists RSPO1 and CHIR-99021 from the +GF medium. The ‘−ROCKi' medium refers to the removal of the ROCK inhibitor Y-27632 from the +GF medium. The +GF medium was switched to the maturation medium (RGF basal medium) for hepatic maturation of HOs. The basal medium for human and mouse HO maturation contained William's E Medium (Thermo Fisher Scientific) with B27 minus vitamin A (Thermo Fisher Scientific), 10 mM HEPES (Thermo Fisher Scientific), GlutaMAX (Thermo Fisher Scientific), penicillin–streptomycin (Thermo Fisher Scientific), 50 μg/ml Primocin (InvivoGen), 15% RSPO1-conditioned medium (home-made), 3 μM CHIR-99021 (MCE), 1 μM A83-01 (MCE), 5 μM ROCK inhibitor Y-27632 (MCE), 10 nM gastrin (Sigma), 2.5 mM nicotinamide (Sigma) and 1.25 mM N-acetylcysteine (Sigma). Additional components were added to the RGF medium to promote hepatic maturation of HOs. For maturation of mouse HOs, the media used included: RGF with 10 nM T3 (Sigma) (RGF+T3); RGF with 200 nM DEX (Sigma) (RGF+DEX); RGF with 10 nM T3 and 200 nM DEX (RGF+T3+DEX). For maturation of fetal human HOs (fhHOs), the media used included: RGF with 20 nM T3 (RGF+T3); RGF with 500 nM DEX (RFG+DEX); RGF with 20 nM T3 and 500 nM DEX (RGF+T3+DEX). For both HO maintenance and maturation, the media were changed every 3 days. To test the culture conditions for fhHO maturation, several conditions were evaluated in the experiments. In the ‘HGF+T3+DEX’ culture media, 50 ng/ml HGF (Biolegend) and 500 nM DEX were added to the RGF medium. T3 was added at the concentrations indicated in the figures. In the ‘HZ’ media, HepatoZYME (Thermo Fisher Scientific) was supplemented with B27 minus vitamin A (Thermo Fisher Scientific), GlutaMAX (Thermo Fisher Scientific), penicillin–streptomycin (Thermo Fisher Scientific) and 50 μg/ml Primocin (InvivoGen). Additional T3 and DEX were added to the ‘HZ’ medium at concentrations of  20 nM and 500 nM, respectively.

To induce rtTA expression in the HOs from the TetO-YAP mice, AAV-TBG-Cre (Addgene) was mixed with hepatocytes in BME at a concentration of 1×10¹⁰ genome copies (GC)/ml before seeding. For the induction of YAP^S127A expression in the TetO-YAP cells, doxycycline (Sigma) was added to the culture medium at a concentration of 200 ng/ml. To overexpress CTNNB1^ΔN90 in the HOs, AAV-TBG-CTNNB1^ΔN90 (self-made, this paper) was mixed with fragmented HOs in BME at a concentration of 2×10¹⁰ GC/ml during passage. The pAAV-TBG-CTNNB1^ΔN90 plasmid was made by cloning CTNNB1^ΔN90 from the pT3-CTNNB1^ΔN90 plasmid (Addgene, 31785) into the pAAV-TBG-Null plasmid (Addgene, 105536).

In the APAP experiment, HOs were cultured in various culture conditions with 15 μM CHIR-99021 (MCE) before APAP treatment. Then, the HOs were treated with 2.5 mM acetaminophen (MCE) with and without the addition of 1.25 mM N-acetylcysteine in the +GF, RGF+T3 and RGF+DEX media that were free of N-acetylcysteine (Sigma) for 24 h. PBS was used as control treatment for acetaminophen and N-acetylcysteine.

In the ethanol treatment experiment, in a 3-day culture period, HOs were treated with 50 mM ethanol (200 proof, USP, Decon Labs) for 36 h and additional 50 mM ethanol for another 36 h without changing the medium. HOs and the culture media were harvested at the end of the culture period for analysis. PBS treatment was used as control.

In the FGF19 treatment experiment, in a 3-day culture period, HOs were treated with 50 ng/ml FGF19 (PeproTech) every 24 h without changing the medium. HOs and the culture media were harvested the end of the culture period for analysis. PBS treatment was used as control.

In the GW4064 treatment experiment, HOs were treated with 5 μM GW4064 (MCE) for 3 days. HOs and the culture media were harvested the end of the culture period for analysis. DMSO treatment was used as control.

HO culture media were diluted with RIPA buffer containing protease inhibitor (Roche) at a ratio of 1:2. Protein samples were further denatured using NuPAGE LDS Sample Buffer (Thermo Fisher Scientific) and loaded into NuPAGE Bis-Tris Mini Protein Gels, 4–12% (Thermo Fisher Scientific), for electrophoresis according to the manufacturer's instructions. The protein gel was then transferred to a nitrocellulose blotting membrane (Cytiva). Ponceau S (Sigma) staining for the total albumin in the medium at around the 65 kDa site was used as a loading control. Human and mouse albumin were detected with the Immobilon Crescendo Western HRP substrate (Millipore) using a PXi4 Chemiluminescent Imaging System (Syngene). Primary antibodies were goat anti-mouse albumin antibody (BETHYL, A90-134A; 1:5000) and goat anti-human albumin antibody (BETHYL, A80-129A; 1:5000). The secondary antibody was mouse anti-goat IgG–HRP (Santa Cruz Biotechnology, sc-2354: 1:2000).

The EdU assay was performed using a Click-iT Plus EdU Cell Proliferation Kit (Thermo Fisher Scientific) according to the manufacturer's instructions.

HOs were dehydrated, embedded in paraffin and sectioned at 6 μm for Hematoxylin and Eosin (H&E) staining and periodic acid–Schiff (PAS) staining. Paraffin sections were rehydrated before staining. For H&E staining, HO sections were stained with Mayer’s Hematoxylin solution (G-Biosciences) and Eosin staining solution (Thermo Fisher Scientific) subsequently. For PAS staining, HO sections were treated with 0.5% periodic acid solution (Electron Microscopy Sciences) for 5 minutes, rinsed, stained with Schiff’s reagent (Electron Microscopy Sciences) for 15 minutes, and counterstained with Mayer’s Hematoxylin solution (G-Biosciences) subsequently. The stained sections were dehydrated and mounted for imaging. Images were captured using a Keyence BZ-X710 microscope with a 20× objective lens.

Organoids were harvested using Cultrex Organoid Harvesting Solution and dissociated using Accutase solution (Corning). Single-cell suspensions were stained with 0.4% Trypan Blue solution (Sigma). Trypan Blue-positive cells and total cells were counted immediately after staining using a hemocytometer.

RNA was isolated from organoid pellets and directly isolated from human cryopreserved hepatocytes (BioIVT) using the PureLink RNA mini kit (Thermo Fisher Scientific) according to the manufacturer's instructions. For reverse transcription-quantitative PCR (RT-qPCR), cDNA was synthesized from RNA using the SuperScript IV VILO Master Mix (Thermo Fisher Scientific) according to the manufacturer's instructions, with 30–100 ng RNA used per reaction. qPCR was performed using 2× UltraSYBR Mixture (Cwbio) on the One-Step Plus qPCR System (Thermo Fisher Scientific). The gene expression data were normalized to Gapdh (mouse) or GAPDH (human), and presented as mRNA fold change using the 2^−ΔΔCt method. The sequence information of qPCR primers is provided in Table S2.

DNA was removed from RNA using a RapidOut DNA Removal Kit (Thermo Fisher Scientific) according to the manufacturer's instructions. 400–1000 ng RNA per sample reaction was used for mRNA isolation using an NEBNext Poly(A) mRNA Magnetic Isolation Module (New England Biolabs). Libraries were prepared using an xGen RNA Library Prep Kit (IDT) according to the manufacturer's instructions. Barcoded libraries were sequenced on an Illumina NextSeq 550/1000 at the Genomics Technology Laboratory of NCI. Paired-end sequencing mode (2×45 bp) was used in the experiments. All organoid samples and primary hepatocytes were sequenced with three replicates for each condition and time point.

Raw counts were normalized, and differential expression analysis was performed using the edgeR package (Robinson and Oshlack, 2010). Heatmaps were generated using the ‘pheatmap’ function in the R software or the ‘Morpheus’ web tool (https://software.broadinstitute.org/morpheus/). Gene ontology analysis was performed using the ‘ClusterProfiler’ (Yu et al., 2012). GSEA was performed using the GSEA 4.0 software or ‘ClusterProfiler’. KEGG functional analysis was performed using ClusterProfiler’. Bulk RNA-seq data of mouse fetal hepatoblasts (E13.5, E15.5, E18.5; two replicates for each time point) was retrieved from GEO (https://www.ncbi.nlm.nih.gov/geo/) under accession GSE90047 (Yang et al., 2017).

Whole-mount immunohistochemistry for organoids was performed as described previously (Dekkers et al., 2019). Briefly, organoids were retrieved from BME using Cultrex Organoid Harvesting Solution and fixed in 4% paraformaldehyde (PFA) for 4 h on ice. Subsequently, the organoids were washed using PBT (0.1% Tween-20 in PBS) and blocked in organoid wash buffer (OWB; 0.2% BSA, 0.1% Triton-X100) for 30 min. Organoids were incubated overnight with diluted primary and secondary antibodies in the OWB at 4°C. Between incubation with primary and secondary antibodies and after incubation with secondary antibodies, the organoids were washed using OWB for at least 3×60 min. Primary antibodies used were anti-Ki-67 (MA5-14520, Invitrogen; 1:200) and anti-HNF-4α (sc-6556, Santa Cruz Biotechnology; 1:100). Secondary antibodies used were donkey anti-goat IgG Alexa Fluor 647 (A21447, Invitrogen; 1:1000) and donkey anti-rabbit IgG Alexa Fluor 568 (A10042, Invitrogen; 1:1000). Organoids were pre-stained with DAPI (Sigma, 1 µg/ml) and mounted in a fructose–glycerol clearing solution (60% glycerol and 2.5 M fructose) for imaging. Images were captured using a Leica DM6 confocal microscope with a 20× objective lens.

Both Oil Red O (ORO; Sigma) staining and HCS LipidTOX Green Neutral Lipid Stain dye (Invitrogen) staining were performed. HOs were harvested using Cultrex Organoid Harvesting Solution and fixed with 4% PFA on ice for 1 h. For whole-mount ORO staining, fixed HOs were stained with the ORO staining solution for 10 min, counterstained with Mayer's Hematoxylin solution (G-Biosciences) and washed before mounting. For whole-mount LipidTOX staining, the fixed HOs were stained in the LipidTOX staining solution for 30 min, washed and counterstained with DAPI according to the manufacturer's instructions. The stained HOs were mounted in fructose–glycerol clearing solution for imaging. Images were captured using a Keyence BZ-X710 microscope with a 20× objective lens.

Total bile acid in the medium was measured using the Total Bile Acids Assay kit (Diazyme) according to the manufacturer's instructions. Urea levels in the medium were measured using the QuantiChrom Urea Assay Kit (BioAssay Systems) according to the manufacturer's instructions. The CYP3A4 and CYP1A2 activities in the HOs were measured using the P450-Glo CYP3A4 Assay kit (Promega, V9001) and P450-Glo CYP1A2 Assay kit (Promega, V8771), respectively, according to the manufacturer's instructions. To measure the ALT levels, the fresh culture medium from HO cultures was first concentrated using Amicon Ultracel-10 centrifugal filter units (Millipore), followed by measurement using the ALT Activity Assay kit (Sigma-Aldrich), according to the manufacturer's instructions. In all biochemical assays, the +GF and RGF medium without HO were used as blank controls, and total DNA from each test well was used for data normalization.

Glucose concentration in the medium was measured using Autokit Glucose (Fujifilm) according to the manufacturer's instructions. In the glucose production assay, advanced DMEM/F12 (Thermo Fisher Scientific) was used as the basal medium. The HOs were washed with PBS and switched to a mixed medium with glucose-free-DMEM/William's E Medium at a ratio of 3:1, with 10 µM forskolin, and incubated for 12 h for glucose production. Glucose production was calculated by the difference in glucose concentration between samples and the blank medium control. After the experiment, the HOs were dissociated and counted for data normalization.

To analyze the ploidy of hepatocytes, organoids were harvested using Cultrex Organoid Harvesting Solution and dissociated using Accutase solution (Corning). Single-cell suspensions of HOs were fixed with ice-cold 70% ethanol on ice for 1 h and washed with PBS. Next, cells were incubated with RNase A (Sigma-Aldrich) solution in PBT (100 µg/ml) at room temperature for 15 min. Cells were then stained with Hoechst 33342 (Invitrogen) (1 µg/ml) in PBS, washed and analyzed with an Attune NxT Flow Cytometry Analyzer (Thermo Fisher Scientific).

FFA solution was prepared following a previously described protocol (Knight et al., 2021). Briefly, oleic acid (OA) and palmitic acid (PA) (Sigma-Aldrich) were conjugated with BSA (Sigma-Aldrich) in 10% BSA solution. 5 mM oleic acid–BSA stock solution and 5 mM palmitic acid–BSA stock solution was made for the experiments. Organoids were seeded in small Matrigel dome [12 µl/droplet; BME, Type R1 (R&D Systems)] for easier liquid exchange.

The HBV inoculum was HepG2.2.15 cell-culture derived virus obtained as previously reported (Murphy et al., 2016). Organoids were spinoculated with HBV (MOI, 1000; 4% PEG) at 25°C for 1 h followed by an incubation at 37°C, 5% CO₂ for 16 h. Cells were then washed several times with PBS containing trypsin (5-fold dilution). Next, cells were maintained with maturation and control media with a medium change every 2–3 days for ∼2 weeks.

Total RNA and DNA were isolated from organoids with HBV and control using the AllPrep DNA/RNA/Protein Mini Kit (80004, Qiagen). The quality control of the nucleic acids was assessed using a NanoDrop One Microvolume UV-Vis Spectrophotometer (Thermo Fisher Scientific). Absolute HBV DNA quantification was performed by real-time qPCR. The data were normalized with extracted total DNA and presented as HBV copies/ng of total DNA. For HBV RNA analysis, total RNA was reverse transcribed into cDNA using SuperScript III Reverse Transcriptase (Thermo Fisher Scientific). HBV total RNA and 3.5 kb RNA were measured by real-time qPCR using HBV-specific primers. Data were normalized with GAPDH and presented as mRNA fold change using the 2^−ΔΔCt method as previously described (Ahodantin et al., 2019). All the real-time qPCR reactions were performed using Power SYBR Green PCR Master Mix (Thermo Fisher Scientific) with a CFX Opus 384 Real-Time PCR System from Bio-Rad. The sequence information of qPCR primers is provided in Table S2.

Data are presented as mean±s.d. Data quantification and analyses were plotted using GraphPad Prism 8. Unless mentioned otherwise, P-value was determined by one-way ANOVA with Sidak's test for comparison of more than two sample groups, and presented as *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001.

We thank the Yang lab members for stimulating discussion. We thank the Su lab members, including Weirong Yuan, for technical support. We also thank the Institute of Human Virology at the University of Maryland School of Medicine. We thank Praju Vikas Anekal and Paula Montero-Llopis of the Microscopy Resources on the North Quad (MicRoN) core at Harvard Medical School for helpful discussions and training in confocal imaging. The Dana-Farber/Harvard Cancer Center is supported in part by the NCI Cancer Center Support Grant NIH P30CA06516.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261961.reviewer-comments.pdf