Modeling metabolic-associated steatohepatitis with human pluripotent stem cell-derived liver organoids Authors: Xiaoshan Wu, Dacheng Jiang, Yuchen Wang, Xin Li, Chenyu Liu, Yanhao Chen, Wei Sun, Ruikun He, Yi Yang, Xiaosong Gu, Chunping Jiang, Qiurong Ding Published: 2024 ABSTRACT AbstractBackground:Metabolic-associated steatohepatitis (MASH) is one of the most prevalent liver diseases worldwide, with a global prevalence estimated between 3% and 5%, posing a significant health burden. Human liver organoids (HLOs) have previously been generated to model steatohepatitis, offering a potential cellular disease model for studying MASH. However, the current HLO model lacks detailed molecular characterizations and requires further improvement.Methods:HLOs derived from human pluripotent stem cells were treated with oleic acid and TGFβ to mimic the MASH progression. Treated HLOs were then analyzed using both bulk and single-cell RNA sequencing. Functional characterization was performed through staining with BODIPY, TMRM, CellROX, and Collagen I, as well as terminal deoxynucleotidyl transferase dUTP nick end labeling and ELISA assays. In addition, a test using the MASH HLO model to validate the hepatoprotective effects of several herb extracts was also conducted.Results:Both RNA-seq and single-cell RNA sequencing demonstrated a close resemblance of multiple molecular signatures and key intercellular communications in and between hepatocyte-like cells and stellate-like cells in the MASH HLO model, compared to human MASH. Furthermore, functional characterizations revealed progressive features of human MASH in the MASH HLO model, including severe steatosis, oxidative stress, mitochondrial dysfunction, inflammation, and fibrosis. In addition, the Schisandra extracts have been demonstrated to have significant antioxidative, anti-inflammatory, and antifibrotic properties in the context of MASH.Conclusions:This study offers an improved HLO disease model of human MASH, which can be potentially applied to facilitate the understanding of the MASH pathogenesis and the discovery of effective treatments.Keywords: disease modeling, herb extracts, human liver organoids, human pluripotent stem cells, metabolic-associated steatohepatitis INTRODUCTION MAFLD encompasses a spectrum of liver disorders progressing from early metabolic-associated fatty liver to more severe stages such as metabolic-associated steatohepatitis (MASH).1 Among patients with MASH, some may develop liver fibrosis, culminating in cirrhosis, liver failure, and HCC. Globally, MAFLD affects ~25% of adults,2,3,4 presenting a significant health burden. Recently, resmetirom has received approval from the US Food and Drug Administration as the first drug to treat MAFLD, marking a significant milestone in addressing this highly unmet medical need.5 Despite this progress, the development of effective treatments for MASH remains challenging. Establishing a suitable MASH model is crucial for studying pathogenesis and drug screening. The commonly used MASH models, including animal and in vitro cell models, have certain limitations. For example, the metabolic characteristics of animal models are not completely consistent with humans,6 and in vitro cell models lack the 3D structure and multicellular interactions of the liver.7 Recent advances in human cell culture have highlighted the potentiality to generate 3D human liver organoids (HLOs) derived either from human pluripotent stem cells (hPSCs) or from human primary hepatocytes, which have been applied to investigate DILI,8,9 and steatohepatitis.10 For example, McCarron et al11 established the culture of primary hepatocyte organoids from liver explants of patients with and without MASH, offering valuable contributions to probe metabolic features of steatohepatitis. However, primary hepatocyte organoids lack immune cells and HSCs, thus missing the inflammatory and fibrotic environment that are important in MASH development. To generate HLOs with multilineage components, Kumar et al12 established an organoid system by coculturing PSC-derived hepatoblast-, HSC-, endothelial-, and macrophage-like cells. While this system provides multiple liver cell types that can mimic cell-to-cell interactions during MASH development, this method is laborious and requires further optimization of each nonparenchymal fraction. To improve the multilineage HLO system, Ouchi et al10 developed a codifferentiation protocol to generate HLOs with hepatocyte-like, Kupffer-like, and HSC-like cells. These organoids showed dose-dependent lipid accumulation, increased inflammatory cytokine expression, and HSC activation upon oleic acid (OA) exposure. However, in a different study, Hess et al13 reported that although OA treatment led to lipid accumulation in HLOs, it ameliorated fibrotic signatures and reduced the HSCs population. Overall, while these studies provide valuable insights, further studies are still needed to establish a solid cellular model of steatohepatitis with detailed characterization of the molecular signatures of this model. To this end, we have adjusted the existing protocols and established an improved MASH HLO (M-HLO) model derived from hPSCs in this study. We have further characterized this model and compared its similarities and differences with human MASH through genome-wide bulk RNA-seq, single-cell RNA sequencing (scRNA-seq), and functional assays. Our results demonstrated a close resemblance of multiple molecular signatures, as well as intercellular communications in hepatocyte-like cells and stellate-like cells in the M-HLO model to human MASH. In addition, we also conducted a test using the M-HLO model to validate the hepatoprotective effects of several herb extracts, and the results demonstrated Schisandra extracts with antioxidative, anti-inflammatory, and antifibrotic properties in the context of MASH. Taken together, our study offers an improved in vitro organoid model for MASH derived from hPSCs, which has the potential to advance our understanding of MASH pathogenesis and facilitate the development of novel therapeutic agents. METHODS Generation of HLOs The 1016 human induced pluripotent stem cell clone used in this study was kindly provided by the Department of Stem Cell and Regenerative Biology at Harvard University and the Harvard Stem Cell Institute. The human induced pluripotent stem cells were differentiated into foregut using the method previously described.9,10 Briefly, human induced pluripotent stem cells were cultured in a humidified incubator at 37°C with 5% CO2. The differentiation process was initiated when the cell density reached ~90%. HLOs differentiation followed the process outlined below: Day 1: Medium (RPMI 1640 [Gibco] with 100 ng/mL activin A [Peprotech] and 50 ng/mL bone morphogenetic protein 4 [Peprotech]). Day 2: Medium (RPMI 1640 with 100 ng/mL activin A and 0.2% fetal bovine serum [Gibco]). Day 3: Medium (RPMI 1640 with 100 ng/mL activin A and 2% fetal bovine serum). Days 4–6: Medium (advanced DMEM/F12 [Gibco] with 500 ng/mL FGF4 [Peprotech] and 3 μM CHIR99021 [Sigma]) to obtain foregut cells. Day 7: Foregut cells dissociated using Accutase, resuspended in Matrigel (Corning), with ~1 × 105 cells embedded in 50 μL droplets of Matrigel. Days 7–10: 5F medium (advanced DMEM/F12 with 5 ng/mL FGF2 [Peprotech], 10 ng/mL VEGF [Peprotech], 20 ng/mL EGF [Peprotech], 3 μM CHIR99021, 0.5 μM A83-01 [Tocris], and 50 μg/mL ascorbic acid [Sigma]). Days 10–14: Retinoic acid medium (advanced DMEM/F12 with 1× B27 [Gibco], 1× N2 [Gibco], 1× GlutaMAX [Gibco], and 2 μM retinoic acid [Sigma]) to differentiate parenchymal and nonparenchymal cells and to promote the formation of bile ducts. Day 15: Organoids harvested, resuspended in complete hepatocyte maturation medium ([Lonza] supplemented with 100 nM dexamethasone [Sigma], 20 ng/mL oncostatin M [Peprotech], 10 ng/mL HGF [PeproTech], and 10% Matrigel), cultured in ultra-low attachment plate. Days 16–20: Medium changed every 2 days until harvesting. Establishment of HLO MASH model (M-HLOs) At day 20 of HLO differentiation, ~100 HLOs were seeded per well in a 96-well ultra-low attachment plate. The HLOs were cultured in a complete mature liver culture medium and treated with 500 μM OA (Sigma) for 7 days (OA group) or treated with vehicle (0.375% bovine serum albumin) for 7 days (vehicle group). Alternatively, HLOs were treated with 500 μM OA for 4 days, followed by 3 days of treatment with 500 μM OA and 25 ng/mL TGFβ (M-HLO group), before harvesting and functional analysis. For efficacy evaluation of the treatments with herbal extracts, 100 μg/mL of Schisandra extracts, Salvia P.E, Pueraria lobata extracts (By-Health Co., Ltd.), or 1 μL/mL DMSO were added along with OA and TGFβ treatments for 7 days before harvesting (M-HLOs + Schisandra extract, M-HLOs + Salvia P.E, M-HLOs + Pueraria lobata, and M-HLOs). As a control, 1 μL/mL of DMSO was added to the vehicle group (control group). RNA isolation and RT‑qPCR Total RNA was isolated from HLOs by Trizol reagent (Thermo Fisher Scientific) according to the manufacturer’s instructions. One microgram RNA was reverse transcribed into cDNA with the PrimeScript reverse transcription kit (Takara). Quantitative real-time PCR was performed using the SYBR Green Supermix (Applied Biosystems). The primer sequences used in the RT-qPCR experiments are listed in Supplemental Table S1, http://links.lww.com/HC9/B93. ELISA After differentiation or drug treatments, the supernatant from HLOs was collected. The secretion levels of human albumin, MCP-1, and IL-6 were determined using the Human Serum Albumin ELISA Kit (R&D Systems), Human MCP-1 (CCL2) mini TMB ELISA Kit (Peprotech), and Human IL-6 mini TMB ELISA Kit (Peprotech), respectively. The ELISA procedure was performed according to the protocol provided by the manufacturer. Immunofluorescence staining The organoids were fixed with 4% paraformaldehyde for 30 minutes. Subsequently, organoids were permeabilized with 0.2% Triton X-100 for 1 hour, followed by blocking with 1% bovine serum albumin for 1 hour at 4°C. The organoids were then incubated with the Collagen I antibody (1:500, Abcam) at 4°C for 24 hours, followed by treatment with the secondary antibodies such as Alexa 488 donkey anti-rabbit (1:200, Invitrogen) for 3 hours at 4°C, and stained with DAPI (1:2000, Beyotime). Imaging was performed using a confocal microscope (Zeiss LSM 880 Carl Zeiss AG), and images were analyzed and quantified using the ImageJ-Fiji software. BODIPY/ TMRM/CellROX/terminal deoxynucleotidyl transferase dUTP nick end labeling assays To visualize lipid droplet accumulation, organoids were treated with a 25 µM BODIPY (GLPBIO) solution at 37°C for 30 minutes, fixed with 4% paraformaldehyde, and stained with DAPI. For mitochondrial membrane potential assessment, organoids were incubated in a medium with 100 nM TMRM (Thermo Fisher) and 1 μg/mL DAPI at 37°C for 30 minutes, followed by PBS wash twice. For the oxidative stress evaluation, 5 μM CellROX (Thermo Fisher) was added to hepatocyte maturation medium. Organoids were then incubated with the CellROX solution at 37°C for 30 minutes, followed by fixation and DAPI staining. To detect cell apoptosis, the terminal deoxynucleotidyl transferase dUTP nick end labeling assay was used according to the manufacturer’s instructions (Vazyme). For all assays mentioned above, imaging was performed with a confocal microscope, and fluorescence intensity measurements were conducted using ImageJ-Fiji software. RNA-seq analysis RNA extraction was performed on M-HLO samples and vehicle samples, following the standard procedures of Majorbio Company (Shanghai Majorbio Bio-Pharm Technology Co., Ltd.). Each group consisted of 2 samples, and each sample contained RNA pooled from 3 replicate wells of equal quality. For comparing the similarity of perturbations at the biological pathway level between the M-HLO model and public available data from patients with MAFLD or MASH, a previously described approach was applied.14,15 The human metabolic-associated fatty liver transcriptome data sets GSE49541 and GSE48452 were applied to analyze the differences between patients with advanced and mild NAFLD, and between patients with MASH and healthy individuals, respectively.16,17 The RNA-seq data have been deposited in the Bio-med Big Data Center under the accession NO OEP00005861. Single-cell RNA sequencing analysis The HLO samples were dissociated into single cells using the trypsin-EDTA (0.05%) treatment at 37°C for 10 minutes. Subsequently, about 30,000 cells were loaded into a 96-well plate, including beads with adapters containing different barcodes, for scRNA-seq library preparation using the Singleron Matrix (Singleron Biotechnologies) platform. A total of 22,298 cells were recovered from both HLO groups. The sequencing reads were processed using GEXSCOPE (Singleron Biotechnologies) to analyze the mkfastq and count values for every single cell. The barcode and read data were demultiplexed and converted into fastq files. The raw reads were then aligned to the Hg19 genome, filtered, and used to create a gene-barcode matrix. The Seurat v5 toolkit (https://github.com/satijalab/seurat) was employed for cell and gene filtering, data integration, data scaling, clustering, and visualization. Cell data with unique feature counts <1500 or mitochondrial counts >50% were excluded from further analysis. The normalized HLOs scRNA-seq data were then mapped to the scRNA-seq data from HLOs in the GSE141183 data set. A total of 14,251 cells without clear cell-type bias (Ro/e<2) were excluded, leaving 5677 cells for subsequent analysis. The extracted data set of 5677 cells was used for graph-based clustering analysis and visualization using t-SNE to identify cell populations. Single-cell transcriptomic data (GSE136103) from 5 healthy livers and 5 cirrhotic livers were retrieved for analyzing differentially expressed genes (DEGs) of mesenchymal cells. Intercellular cross-talk was analyzed using CellChat V1.1.3. The pseudotime trajectory analysis was performed by the Monocle 3 v1.3.1.8. The scRNA-seq data have been deposited in the Bio-med Big Data Center under the accession NO OEP00005861. Statistical analysis Statistical analysis for the data from 2 experimental groups was performed using the unpaired 2-tailed t tests. One-way ANOVA with the Dunnett multiple comparisons test was used for multiple parametric data comparisons, while the Kruskal-Wallis test was employed for multiple nonparametric data comparisons. All statistical analyses were conducted using Prism 9 (GraphPad). The error bars in the graphs denote the SEM as defined in the legend, representing the variability of the mean values. RESULTS Presence of MASH-related basal characteristics in hPSC-derived liver organoids after treatment with OA and TGFβ To generate liver organoids from hPSCs, we followed an established retinoic acid–based protocol (Figure 1A).8,9 Initially, we treated the HLOs with 500 μM OA for 7 days, starting at day 20 to model MASH progression. However, subsequent analysis of gene expression showed that OA treatment only slightly induced the expression of several inflammation markers, such as CCL20, with no significant changes observed in fibrosis marker expression (Figure 1B). Previous studies reported that monocyte-derived macrophages recruited from the bloodstream are the primary source of TGFβ, a crucial factor in liver fibrogenesis during MASH progression.18,19 Considering that HLOs lack these recruited immune cell types from the bloodstream, we supplemented TGFβ on top of OA treatment at the later stage of MASH induction for 3 days at a concentration of 25 ng/mL. Gene expression analysis showed that compared to OA treatment alone, OA + TGFβ treatment significantly upregulated the mRNA expression levels of both inflammatory markers (MCP-1, ICAM1, IL-6, and IL-8) and fibrotic markers (COL1α1, TGFβ1, PDGFRβ, TIMP1, and SMAD7) (Figure 1B). Development of an HLO model for MASH. (A) Schematic representation of establishing an organoid model for MASH using hPSC-derived HLOs. DE represents definitive endoderm. (B) The OA group was treated with 500 μM OA for 7 days, while the M-HLO group was first treated with 500 μM OA for 4 days, followed by simultaneous treatment with 500 μM OA and 25 ng/mL TGFβ for 3 days. The gene expression levels of inflammatory marker genes (CCL20, MCP-1, ICAM1, IL-6, and IL-8) and fibrogenic marker genes (COL1α1, TGFβ, PDGFRβ, TIMP1, and SMAD7) were determined by qPCR. n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Values represent means with SEM. p values were assessed by 1-way ANOVA with the Dunnett multiple comparisons test. (C) The KEGG pathway enrichment analysis of DEGs between the vehicle group and the M-HLO group. The size of the dots represents the number of DEGs, and the colors correspond to the adjusted p value (p. adjust). (D) Comparison of the transcriptomic alterations between MASH patients’ data and M-HLO models. The heatmap showed the aberrant pathways in each cohort, with orange color representing the induction of the corresponding pathway and blue color indicating the suppression of the pathway. The shade of colors represents the significance of the enrichment. Abbreviations: DEG, differentially expressed gene; HLOs, human liver organoids; hPSC, human pluripotent stem cell; MASH, metabolic-associated steatohepatitis; M-HLO, MASH HLO; OA, oleic acid. We then performed RNA-seq analysis on the OA + TGF-treated HLOs and the vehicle-treated HLOs. Pathway analysis showed that OA + TGFβ treatment induced dysregulations in complement and coagulation cascades, cholesterol metabolism, and focal adhesion, etc. (Figure 1C). To get a general idea of to what extent this model mimics human MASH, we compared dysregulated pathways from KEGG gene sets in 2 human MASH series16,17 to those in OA + TGFβ-treated HLOs. The results showed clear similarities in dysregulated pathways, including cell focal adhesion, small cell lung cancer, extracellular matrix receptor interaction, and leukocyte transendothelial migration, especially between the MASH versus healthy liver and the OA + TGFβ group versus vehicle group (Figure 1D). This suggests a resemblance of these transcriptomic dysregulations during human MASH progression within the HLO model. Further functional analysis of OA + TGFβ-treated HLOs exhibited the basal characteristics of MASH, including significant lipid droplet accumulation (BODIPY staining), decreased mitochondrial membrane potential (TMRM staining), enhanced oxidative stress (CellROX staining), increased cell apoptosis (terminal deoxynucleotidyl transferase dUTP nick end labeling assay), and activated fibrogenesis (Collagen I staining) (Figure 2A). In addition, a dramatic decrease in the albumin secretion and the upregulation of MCP-1 and IL-6 secretion also indicated impaired liver function and elevated inflammation in the OA + TGFβ group (Figure 2B). Taken together, through cotreatment of OA and TGFβ, HLOs exhibited similarities in functional and transcriptional deficiencies during human MASH progression. We are thus referred to the OA + TGFβ-treated HLOs as “MASH HLOs (M-HLOs)” and vehicle-treated HLOs as “vehicle.” Functional characterizations of HLOs after treatment with OA and TGFβ. (A) Representative confocal images of the vehicle group and M-HLO group stained with BODIPY (green), TMRM (red), CellRox (red), Collagen I (green), and TUNEL (red), respectively, with the cell nuclear dye DAPI in blue. Right: Quantification of mean fluorescence intensity. n = 15, *p < 0.05, **p < 0.01, ****p < 0.0001. Scale bar, 50 μm. (B) Secretion levels of albumin, MCP-1, and IL-6 in the supernatants of the vehicle group and M-HLO group were measured by ELISA. n = 3, **p < 0.01, ***p < 0.001. Values represent means with SEM. p values were assessed by unpaired 2-tailed Student t test (A, B). Abbreviations: HLOs, human liver organoids; MASH, metabolic-associated steatohepatitis; M-HLO, MASH HLO; OA, oleic acid; TUNEL, terminal deoxynucleotidyl transferase dUTP nick end labeling. Single-cell characterization of the MASH HLOs To further elucidate the hepatic lineage alterations in the MASH HLO model, we performed scRNA-seq analysis on the cell populations of both M-HLOs and vehicle HLOs. Data from a total of 5677 single cells were analyzed after quality control. Initial clustering and marker gene annotation identified 4 major cell populations (Figure 3A). The major cell populations identified in HLOs include hepatocyte-like (HNF4A, ALB, and RBP4), stellate-like (COL1a1, PDGFRβ, and DCN), biliary-like (KRT7, KRT17, and SPP1), and biliary tree stem cells (FOXA2 and PROX1) (Figures 3A, C).20 The M-HLO group exhibited substantial changes in the gene expression profiles of each cell population (Figure 3B). However, we failed to identify Kupffer-like cells as seen in the previous studies.10 Although the macrophage marker CD68 was detected,8,9 scRNA-seq analysis failed to identify a distinct subset of cells with similar transcriptomic profiles to Kupffer-like cells. Single-cell analysis on the MASH HLOs. (A, B) The 5677 cells were clustered and analyzed using t-SNE, and the cells were annotated for clustering according to the mapping results. Dots of different colors indicate different clustering subgroups (A). The clustering results are shown according to the different treatments for HLOs (B). (C) The bubble diagram represents the cell types annotated with marker genes for HSCs (COL1A1, PDGFRβ, and DCN), hepatocytes (HNF4A, ALB, and RBP4), biliary dendritic cells (FOXA2 and PROX1), and bile duct cells (KRT7, KRT17, and SPP1). (D) A bar graph shows the relative proportions of the different cell types identified in the scRNA-seq data from the vehicle-treated and M-HLO groups. The total number of cells analyzed was 5677, with 3346 cells from the vehicle group and 2231 cells from the M-HLO group. (E) REACTOME pathway enrichment analysis was used to analyze DEGs in hepatocyte-like cells from M-HLOs. Horizontal coordinates represent fold enrichment, and colors represent p. adjust. (F) REACTOME pathway enrichment analysis of M-HLO hepatocyte-like cells yielded differentially expressed genes enriched in the mitochondrial ETC pathway. Colors represent mean log 2-fold change values. p value <0.05. (G) Relative expression of nuclear genes (TAFM, NRF1, and NRF2) associated with regulating mtRNA transcription in M-HLOs. n = 3, *p < 0.05,**p < 0.01. Values represent means with SEM. p values were assessed by unpaired 2-tailed Student t test. Abbreviations: DEG, differentially expressed gene; HLOs, human liver organoids; MASH, metabolic-associated steatohepatitis; M-HLO, MASH HLO; mtRNA, mitochondrial RNA; scRNA-seq, single-cell RNA sequencing. The proportions of different cell types in HLO scRNA-seq data were next analyzed following the M-HLO group compared to the vehicle group. The analysis revealed notable changes: Hepatocyte-like cells decreased from 32.06% in the vehicle group to 25.25% in the M-HLO group, biliary-like cells showed a substantial reduction, dropping from 11.73% to 2.73%, and biliary tree stem cells remained stable (4.95%–5.06%). Conversely, stellate-like cells exhibited an increase in proportion, rising from 56.21% to 66.97% (Figure 3D). These results suggest that OA + TGFβ treatment significantly impacts the cellular composition of HLOs, notably by increasing the proportion of stellate-like cells while reducing the proportions of hepatocyte-like and biliary-like cells. To further compare with the cell composition features of human MASH, we retrieved human MASH scRNA-seq data under GSE136103, which contains scRNA-seq data from 5 healthy individuals and 5 patients with cirrhosis. We mapped this data set to the single-cell data from our HLOs and compared the cell-type proportions. The results showed similarities in the cellular composition between the M-HLO and the human MASH data (Supplemental Figures S1A, B, http://links.lww.com/HC9/B93), indicating that the M-HLO can, to some extent, reflect the conditions observed in MASH disease. We next re-analyzed the hepatocyte-like cells, which could be further subclustered into a total of 12 distinct subpopulations (Supplemental Figures S1C, D, http://links.lww.com/HC9/B93). The hepatic lobule is the fundamental structural unit of the human liver, displaying significant heterogeneity along the portal-central axis due to variations in metabolism and gene expression across distinct zones.21,22 Interestingly, a comparative expression analysis of genes associated with different liver zones showed that cluster 7 in hepatocyte-like cells in HLOs exhibited enriched expression of genes involved in lipid and cholesterol synthesis (SCD, HMGCS1, ACSS2, and CPS1), indicating similarity to human periportal hepatocytes (Supplemental Figure S2B, http://links.lww.com/HC9/B93). Moreover, the top DEGs in cluster 7, including TM7SF2, SLPI, ACAT2, TM4SF5, and MSMO1 (Supplemental Figure S2B, http://links.lww.com/HC9/B93), coincided with the top DEGs observed in the human periportal hepatocytes.22 In humans, pericentral hepatocytes play a role in drug metabolism and detoxification, constitutively expressing high levels of cytochrome P450 enzymes. Cluster 11 showed enrichment of genes associated with this function, including BCHE, LINC00261, and GOLT1A (Supplemental Figure S2B, http://links.lww.com/HC9/B93). Among them, BCHE is specifically involved in drug and lipid metabolism.23 In addition, GLUL, LINC00261, and GOLT1A were found to be highly expressed in the most central venous sinusoid layer in humans.22,24 In addition, cluster 3 exhibited high expression of ALDH6A1, RPP25L, and TKFC (Supplemental Figure S2B, http://links.lww.com/HC9/B93). Notably, ALDH6A1 functions as a mitochondrial methylmalonate semialdehyde dehydrogenase and plays a crucial role in the metabolic pathways of drugs, valine, and pyrimidine.25,26 These 2 clusters demonstrated similarity to the highly expressed genes found in the human pericentral hepatocytes (Supplemental Figure S2B, http://links.lww.com/HC9/B93).22 In addition, previous research27 indicated that a small cluster of human hepatoblasts expressed ID3 from week 5 and week 6 human fetal livers, coexpressing mesoderm-related genes. In HLOs, ID3+ hepatocyte-like cell clusters (clusters 3, 6, 7, 10, and 11) coexpressed mesoderm-related genes (NR2F2, SOX4, and NCAM1) (Supplemental Figure S2B, http://links.lww.com/HC9/B93), demonstrating similarity to progenitors in the early stages of liver bud formation. Taken together, although still far from mimicking the physiological cell composition of liver tissues, HLOs encompassed several major liver cell types and reflected some key characteristics of liver zonation. Hepatocyte-like cells exhibited deficiencies in mitochondrial complexes in MASH HLOs We next investigated the changes in each cell type in the M-HLOs. Interestingly, we observed that clusters 1, 2, and 6 of hepatocyte-like cells are predominantly composed of cells from the M-HLO group (Supplemental Figure S2A, http://links.lww.com/HC9/B93). In cluster 1, the top DEGs include LIPK, CLEC18C, ULBP2, AQP1, and CLVS2. LIPK plays a role in lipid metabolism and digestion.28 CLEC18C 29 and CLVS2 30 are associated with maintaining the normal morphology of late endosomes or lysosomes. ULBP2 is primarily located on the cell surface, binding to the NKG2D receptor on natural killer cells and activating them.31 In cluster 2, the top DEGs included TAGLN, ITGA2, LTB, LAMC2, and CCL2. Among these, TAGLN,32 ITGA2,33 and LAMC2 34 are associated with the extracellular matrix remodeling, while LTB 35 and CCL2 36 are associated with the inflammatory responses. Cluster 6 exhibited top DEGs including ACTA1, RRP9, MLLT3, WNT6, and WARS2 (Supplemental Figure S2C, http://links.lww.com/HC9/B93). Among these, RRP9 is essential for 18S rRNA processing during ribosome synthesis.37 MLLT3 is a component of the super elongation complex,38 and WARS2 is a mitochondrial aminoacyl-tRNA synthetase involved in protein synthesis.39 These findings suggest that cluster 1 is associated with changes in cell surface membrane function. Cluster 2 is related to active immune response and fibrosis formation, and cluster 6 is primarily involved in the protein synthesis process. The underlying indications of these clusters, specifically from the M-HLOs, require further investigation. In addition, pathway analysis revealed that the DEGs between M-HLOs and vehicle HLOs were enriched in pathways related to eukaryotic translation elongation, citric acid (TCA) cycle and respiratory electron transport, metabolism of amino acids and derivatives, steroid metabolism, and mitochondrial protein import, etc. (Figure 3E), which also usually show deficiencies during human MASH progression.40 The nuclear respiration factors (NRF1 and NRF2)-mitochondrial transcription factor A (TFAM) axis has been proven to be crucial in mediating the function of peroxisome proliferator–activated receptor gamma coactivator-1 alpha in promoting mitochondrial biogenesis.41,42 In the scRNA-seq analysis, a significant downregulation of TFAM expression was observed in hepatocyte-like cells from the M-HLO group, which was further confirmed by qPCR analysis (Figure 3G). In addition, NRF2 expression in the M-HLO group also significantly decreased (Figure 3G), which may contribute to the general downregulation of genes in mitochondrial complexes in hepatocyte-like cells from the MASH group (Figure 3F). Hepatic stellate-like cells in the MASH HLO group revealed molecular characteristics closely associated with fibrogenesis Further clustering analysis of hepatic stellate-like cells identified 15 subpopulations (Figure 4A). Notably, clusters 1 and 14 specifically emerged in the M-HLO group (Figure 4A). Both clusters are characterized by high expression of the PDPN gene, which encodes podoplanin, a type-I integral membrane glycoprotein (Figure 4A). Interestingly, PDPN high subpopulation has also been previously identified in patients with liver cirrhosis43 (Supplemental Figure S3A, http://links.lww.com/HC9/B93), suggesting some similarities between the M-HLOs and human liver cirrhosis regarding stellate-like cells transitions during disease progression. Hepatic stellate-like cells in the M-HLOs group displayed molecular characteristics closely associated with fibrogenesis. (A) UMAP clustering reveals distinct cell populations within the vehicle group and M-HLO groups. Clusters 1 and 14 show high expression of the PDPN gene. (B) GO pathway enrichment analysis was used to analyze DEGs in stellate-like cells from M-HLOs. The size of the dots represents the number of DEGs, and the color represents p. adjust. (C) GSEA pathway enrichment analysis shows enrichment of the collagen formation pathway in stellate-like cells from M-HLOs. (D) A circular diagram illustrates the top 100 receptor-ligand pairs involved in cell-cell communication within the M-HLOs. Abbreviations: DEG, differentially expressed gene; GO, gene ontology; GSEA, gene set enrichment analysis; HLOs, human liver organoids; MASH, metabolic-associated steatohepatitis; M-HLO, MASH HLO. Transcriptomic features of stellate-like cells were assessed in the M-HLO and vehicle groups using the scRNA-seq data set. Gene ontology enrichment analysis of DEGs in stellate-like cells from the M-HLOs revealed upregulation of pathways related to extracellular matrix composition, extracellular structure organization, external encapsulating structure, actin filament organization, cell-matrix adhesion, and actin filament-based processes (Figure 4B). In addition, gene set enrichment analysis showed significant enrichment of pathways related to collagen formation in stellate-like cells of the M-HLOs (Figure 4C), resembling those observed during human MASH progression. To investigate the origin of hepatic stellate-like cells, we performed a trajectory analysis of all cell types in HLOs using Monocle 3. Our analysis revealed 2 sources of cells with high expression of activated HSC markers in HLOs. First, during the early stages of liver bud formation, the mesoderm-derived septum transversum mesenchyme surrounds and invades the nascent liver bud. These septum transversum mesenchyme differentiate into various vessels and mesenchymal cells and further into endothelial cells, mesothelial cells, and HSCs.27,44 In HLOs, stellate-like cells were clustered and analyzed by UMAP, which revealed a high expression level of septum transversum mesenchyme marker genes (ID3, NR2F2, SOX4, and HAND2) in clusters 3, 4, and 11 (Supplemental Figure S3C, http://links.lww.com/HC9/B93). We designated clusters 3, 4, and 11 as the root nodes of the pseudotime trajectory. Along the trajectory, clusters 4 and 11 transition into cluster 0, which shows high expression of HSC markers (PDGFRB, LOX, SPARC, COL1A1, and COL3A1). In the M-HLO group, stellate-like cells (cluster 0) were activated, forming clusters 12, 1, and 14, which exhibited high expression of activated HSC markers (ACTA2, TAGLN, TGFBR1, TIMP1, PDGFRB, S100A11, ITGB1, ITGAV, and PDPN), reaching the endpoint of the pseudotime trajectory (Supplemental Figures S3C, S4A, B, http://links.lww.com/HC9/B93). Interestingly, we identified another group of cells within hepatocyte-like clusters 1, 2, and 6, which also showed high expression of activated HSC markers (S100A16, ITGB1, ITGAV, ACTA2, TGFBR1, TIMP1, TAGLN, IL32, VIM, SPARC, COL1A1, and COL3A1; Supplemental Figure S4D, http://links.lww.com/HC9/B93). This cluster seems to originate from cluster 5, which is characterized by high expression of hepatoblast markers (EPCAM, CEBPA, AFP, APOA1, CYP2C9, ALB, and HNF4A). According to the pseudotime trajectory analysis, in the M-HLO group, cluster 5 transitions through clusters 1 and 2, ultimately reaching the endpoint of cluster 6 upon OA + TGFβ treatment. In contrast, in the vehicle group, cluster 5 differentiates into clusters 3, 7, and 11, which express hepatocyte genes related to pericentral and periportal regions (Supplemental Figures S2B, S4B, C, http://links.lww.com/HC9/B93). In addition, we also identified a similar cluster (cluster 4) in epithelial cell subpopulations from scRNA-seq data of patients with liver cirrhosis. This cluster exhibited high expression of activated HSC markers (S100A16, ITGB1, ITGAV, TGFBR1, and IL32, VIM; Supplemental Figure S4E, http://links.lww.com/HC9/B93).43 In summary, our trajectory analysis elucidates 2 distinct origins for cells with high expression of activated HSC markers in HLOs, highlighting the complex lineage dynamics during liver development and the progression of MASH. A previous study reported that hepatocytes exhibited greater communication with HSCs than with macrophages during MASH progression, with the top-ranked signals between HSCs and hepatocytes including FN1, Collagen, and bone morphogenetic protein 4.24 To investigate whether similar cellular communications could be found between stellate-like cells and hepatocyte-like cells in the M-HLOs, we used CellChat V1.1.3 to analyze the putative ligand-receptor interactions in M-HLOs (Figure 4D). Interestingly, the communication pathways between hepatocyte-like cells and stellate-like cells in the M-HLOs also exhibited significant enrichment of related signaling pathways, including FN1, Collagen, and bone morphogenetic protein 4 (Figure 4D). Taken together, these findings suggest that the M-HLO model resembles the major communication pathways between hepatocyte-like cells and stellate-like cells as observed in human MASH. Furthermore, the Spearman rank correlation coefficient analysis comparing the DEGs between the vehicle and M-HLOs groups and those in human healthy and cirrhosis groups (GSE136103) revealed a significant correlation of DEGs in hepatocyte-like cells (Spearman rho of 0.28, p < 0.001), and stellate-like cells (Spearman rho of 0.395, p < 0.001) (Supplemental Figure S3B, http://links.lww.com/HC9/B93). These findings underscore the potential of the M-HLOs model for studying intercellular communication and the molecular mechanisms involved in MASH progression. Schisandra extracts displayed significant protective effects against disease progression in MASH HLOs We further evaluated the efficacy of Schisandra extracts, Salvia P.E, and Pueraria lobata extracts in M-HLO models, which have been reported to have liver-protective functions and thus have been applied in health products. Take M-HLOs treated with 100 μg/mL Schisandra extracts, Salvia P.E, Pueraria lobata extracts, or 1 μL/mL DMSO for 7 days as M-HLO+ herbal extracts group and the M-HLOs group. Take the HLOs treated with the 1 μL/mL DMSO as a control group. Interestingly, treatment with Schisandra extracts resulted in a significant attenuation of MASH-related phenotypes in M-HLOs, including alleviated lipid accumulation, oxidative stress, fibrogenesis, and cell apoptosis, as well as decreased expression of inflammation and fibrosis-related genes (Figures 5A–D). However, treatment with Salvia P.E or Pueraria lobata extracts did not show significant protective effects in M-HLOs (Supplemental Figures S5A, B, http://links.lww.com/HC9/B93). Schisandra extracts displayed significant protective effects against disease progression in M-HLOs. (A) Representative bright field images of the control, M-HLOs, and M-HLOs + Schisandra extract group. HLOs were treated with 500 μM OA for 4 days before treating them with 500 μM OA + 25 ng/mL TGFβ for 3 days and were simultaneously administrated with Schisandra extracts (100 μg/mL) (M-HLOs + Schisandra extract) or 1 μL/mL DMSO (M-HLOs), respectively. In the control group, HLOs were only treated with 1 μL/mL DMSO. Scale bar, 200 μm. (B) The fluorescent intensity of HLOs stained with CellROX (red)/BODIPY (green)/Collagen I (green) or TUNEL (red) were detected in the M-HLO group and M-HLOs + Schisandra extracts group. Right: quantification of the mean fluorescence intensity. Scale bar, 50 μm. n = 15, *p < 0.05, **p < 0.01. (C, D) The expression levels of the genes associated with the hepatic function, inflammation, and fibrosis markers in the control group, M-HLO group, and M-HLOs + Schisandra extract group were determined by qPCR. n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (E) The level of MCP-1 secretion was measured in the control group, M-HLO group, and M-HLOs + Schisandra extract group, respectively. n = 3, **p < 0.01, ****p < 0.0001 Values represent means with SEM. p values were assessed by 1-way ANOVA with Kruskal-Wallis tests (B) and the Dunnett multiple comparisons test (C– E). Abbreviations: HLOs, human liver organoids; MASH, metabolic-associated steatohepatitis; M-HLO, MASH HLO; OA, oleic acid; TUNEL, terminal deoxynucleotidyl transferase dUTP nick end labeling. DISCUSSION In this study, we presented an improved approach to develop an in vitro multilineage MASH HLO model from hPSCs. By treating with OA + TGFβ, HLOs presented features associated with human MASH, including lipid accumulation, mitochondrial damage, oxidative stress, hepatocyte apoptosis, and fibrogenesis. The in-depth RNA-seq and scRNA-seq analysis also demonstrated at least partial similarities at the transcriptomic level and the cellular communication patterns between the M-HLO model and human MASH. Furthermore, using the M-HLO model, the hepatoprotective effect of the Schisandra extracts was demonstrated. Previously, OA treatment has been used to generate MASH HLOs by treating with varying concentrations of OA for 5 days, resulting in a dose-dependent increase in lipid accumulation, inflammation, and fibrogenesis.10 Unfortunately, we were unable to recapitulate these results in gene expression level, as treatment with OA only slightly increased the expression of some inflammation markers, with no effect on fibrosis-related markers. This aligns with the discoveries from another recent study, where OA treatment actually reduced fibrotic signatures.13 The underlying reasons await more investigation. Nonetheless, considering the pathophysiological mechanism of MASH progression, where the primary source of TGFβ is from monocyte-derived macrophages recruited from the bloodstream,18,19 it is reasonable to add TGFβ to promote fibrogenesis before introducing specific macrophage types into the organoid system. In the M-HLO model, we observed a specific upregulation of the CCL2 and PDPN genes. This aligns with previous studies showing significant CCL2 upregulation in MASH patients compared to patients with NAFL and controls.36 In addition, a distinct PDPN high mesenchymal subpopulation was also identified in individuals with liver cirrhosis.43 Moreover, recent findings suggest that PDPN expression is also evident in cancer-associated fibroblasts,45 suggesting a potential role for PDPN and PDPN high cells in liver fibrogenesis that merits further exploration. We also further determined the origin of activated HSCs in M-HLOs. Interestingly, 2 potential sources of activated hepatic stellate-like cells were identified: (1) from existing stellate-like cells and (2) from subpopulations of hepatocyte-like cells that express hepatoblast cell marker genes after OA + TGFβ treatment. A similar cluster was found in epithelial cells from patients with liver cirrhosis, exhibiting highly activated HSC marker expression (Supplemental Figure S4E, http://links.lww.com/HC9/B93). These findings underscore the remarkable plasticity of early embryonic hepatoblast cells, demonstrating their capacity to differentiate into cholangiocytes and hepatocytes, as well as to adopt activated HSC characteristics in response to OA + TGFβ stimulation. However, whether these discoveries accurately reflect true pathophysiological events during human MASH development needs more investigation. Although the current M-HLOs represent an improvement over single-cell culture systems, significant gaps remain between these organoid models and the actual disease. For example, the RNA-seq comparison between the human MASH and the M-HLOs showed a lack of key immune signals in the M-HLOs (Figure 1C), which are crucial for MASH progression. Significant efforts have now also been made to add immune cells to the human organoid systems, such as by transplanting intestinal organoids into animals with a humanized immune system46 or co-developing hemogenic endothelium–like cells and erythromyeloid progenitors during human colonic organoid differentiation.47 Similar approaches may also be applied to introduce the immune components to HLOs. Recent scRNA-seq analyses of human MASH have highlighted the complex, dynamic changes in the immune system during disease progression,48 but faithfully recapitulating these changes in the dish remains challenging. The other significant challenge is the batch-to-batch variability in the proportion of mesenchymal-like cells within HLOs. The Takebe group reported similar issues.49 They recommended passing organoid suspension through a 100 μm cell strainer to remove larger mesenchymal-like aggregates on days 17–18 of differentiation. In addition, the proportion of mesenchyme-like cells can depend on the cell line and differentiation batch, necessitating experimental determination for each batch. Hess et al13 also reported the proportion of mesenchymal-like cells varied significantly in HLOs, ranging from 5.4% to 48% under control conditions. These findings highlight the need for further optimization to achieve stable cell-type proportions in each HLO batch. In addition, in this study, we used the M-HLO models to evaluate the efficacies of 3 herb extracts, which were all considered to be liver-protective. Interestingly, the results demonstrated a significant alleviating effect from the Schisandra extracts but not from the Salvia P.E or the Pueraria lobata extracts. It is possible that the functional targets of Salvia P.E or Pueraria lobata extracts are immune cell types that are lacking in the M-HLOs or that their liver-protective functions occur through systematic effects by targeting other metabolic tissues, for example, adipose tissue or gut microbiota. CONCLUSIONS In our study, we have established an in vitro 3D multilineage MASH HLOs model that mimics key molecular features of human MASH pathology. By using M-HLOs, we were also able to validate the liver-protective effects of Schisandra extracts in terms of their antioxidative, anti-inflammatory, and antifibrotic properties. While substantial additional investigations and developments are still needed before we can really recapitulate human MASH progression in the dish, this study shown here has demonstrated a more improved human cell system for modeling MASH. We, therefore, offer this human MASH model as a valuable tool for studying MASH disease mechanisms and exploring novel drug candidates. Supplementary Material DATA AVAILABILITY STATEMENT All data obtained and analyzed during this study are available from the corresponding authors once reasonably requested. AUTHOR CONTRIBUTIONS Qiurong Ding, Chunping Jiang, Yi Yang, and Xiaosong Gu conceived, designed, and supervised the project; Xiaoshan Wu, Dacheng Jiang, Yuchen Wang, Xin Li, Chenyu Liu, and Yanhao Chen designed and performed the experiments and analyzed data; Wei Sun and Ruikun He provided experimental materials. Xiaoshan Wu and Qiurong Ding wrote the manuscript. ACKNOWLEDGMENTS The authors are grateful to the Institutional Center for Shared Technologies and Facilities of SINH, CAS, for technical assistance. FUNDING INFORMATION This work was supported by grants from, the National Key R&D Program of China (2023YFA1801100), the National Natural Science Foundation of China (92357306, 82425013, 82330028 and 32300999), Shandong Provincial Laboratory Project (SYS202202), the Research Project of Jinan Microecological Biomedicine Shandong Laboratory (JNL202204A and JNL202219B), Shanghai Science and Technology Commission (21140901700), and By-Health Corp. Ltd. CONFLICTS OF INTEREST The authors have no competing conflicts to report. Footnotes Abbreviations: DEG, differentially expressed gene; hPSC, human pluripotent stem cell; HLOs, human liver organoids; M-HLO, MASH HLO; MASH, metabolic-associated steatohepatitis; OA, oleic acid; sc-RNA-seq, single-cell RNA sequencing. Xiaosong Gu is a senior author. Chunping Jiang and Qiurong Ding contributed equally to this work. Supplemental Digital Content is available for this article. Direct URL citations are provided in the HTML and PDF versions of this article on the journal's website, www.hepcommjournal.com. Contributor Information Xiaoshan Wu, Email: wuxiaoshan2018@sibs.ac.cn. Dacheng Jiang, Email: jiangdacheng2019@sibs.ac.cn. Yuchen Wang, Email: ycwang@sibs.ac.cn. Xin Li, Email: lixin@jnl.ac.cn. Chenyu Liu, Email: liuchenyu2023@sinh.ac.cn. Yanhao Chen, Email: chenyanhao@sibs.ac.cn. Wei Sun, Email: sunw@by-health.com. Ruikun He, Email: herk@by-health.com. Yi Yang, Email: yiyang@ecust.edu.cn. Xiaosong Gu, Email: nervegu@ntu.edu.cn. Chunping Jiang, Email: chunpingjiang@nju.edu.cn. Qiurong Ding, Email: qrding@sinh.ac.cn. REFERENCES