Recapitulation of bile acid metabolism in hepatobiliary organoids derived from hiPSC

Abstract

Background & Aims

Bile acids are crucial mediators of cholesterol homeostasis, lipid digestion, and detoxification, and reliable in vitro synthesis systems are essential for liver disease research and precision medicine. Although hiPSC-derived hepatic organoids model key human liver functions, fully recapitulating bile acid biosynthesis remains challenging. Here, we generated bile acid-hepatobiliary organoids (BA-HBOs) that capture major aspects of bile acid metabolism, thereby providing a more physiologically and pathologically relevant platform for metabolic studies and drug screening.

Methods

Maturation protocols were optimized with saikosaponin A to enhance bile acid biosynthesis in BA-HBOs. Fibrotic BA-HBOs (FiBA-HBOs) were further established by transforming growth factor-β treatment. Cellular identity and functional markers were evaluated by immunofluorescence, flow cytometry, and quantitative PCR. Organoid heterogeneity was characterized by single-cell RNA sequencing, and bile acid composition and diversity were quantified by liquid chromatography-mass spectrometry/mass spectrometry metabolomics.

Results

BA-HBOs demonstrated hepatic and biliary functions, including organized lineage segregation and robust synthetic and metabolic capacities. Compared with controls, BA-HBOs showed increased bile acid synthesis, improved bile duct structure, and higher transport protein expression (n ≥3, p <0.05). Targeted metabolomics identified a complex spectrum of 33 bile acid species, predominated by glycine-conjugated forms, consistent with human physiology. Single-cell RNA sequencing revealed that BA-HBOs recapitulated the transcriptional landscape of adult liver tissue, and revealed hepatocyte subpopulation restructuring associated with enhanced bile acid metabolism. Moreover, modeling fibrosis-associated dysregulation generated FiBA-HBOs, which exhibited cholestasis-like changes and disease-relevant metabolomic profiles (n ≥3, p <0.05).

Conclusions

BA-HBOs recapitulate key aspects of human liver bile acid metabolism, hepatocyte zonation, and core metabolic processes in vitro, providing a physiologically relevant platform for mechanistic studies of liver disease and screening therapeutic candidates.

Impact and implications

Our hiPSC-derived BA-HBOs synthesize and secrete diverse bile acid species, capturing key transcriptional and metabolic profiles of the human liver. They enable modeling of liver fibrosis accompanied by disruptions in bile acid metabolism, offering a tool to dissect disease mechanisms. These robust organoids create new opportunities for basic liver research, therapeutic development, and precision medicine.

Graphical abstract

Highlights

• •Saikosaponin A activates PPARα/LXR to enhance bile acid metabolism in BA-HBOs.
• •Targeted metabolomics defines a physiologically relevant bile acid profile in BA-HBOs.
• •BA-HBOs capture key multicellular transcriptional and metabolic features of human liver.
• •BA-HBOs show spontaneous zonation and enrichment of metabolic hepatocyte subsets.
• •TGF-β–treated BA-HBOs model fibrosis-related bile acid dysregulation.

Introduction

Bile acids are key products of liver metabolism, with critical roles in cholesterol homeostasis, lipid digestion, and metabolic signaling pathways in the human body.^1,2 Dysregulation of bile acid metabolism underlies a wide range of liver diseases, including cholestasis, non-alcoholic fatty liver disease, and liver fibrosis.[3], [4], [5] However, our understanding of human bile acid homeostasis has been largely constrained by the limitations of animal models and the restricted availability of primary human hepatocytes and cholangiocytes.[6], [7], [8] Such difficulties highlight the urgent need for physiologically and pathologically relevant in vitro models that can faithfully recapitulate the complexity and cellular heterogeneity of human hepatic tissues.

More recently, advances in human induced pluripotent stem cell (hiPSC) technology have enabled the generation of hepatic organoids, providing a promising platform for in vitro disease modeling, mechanism studies, and drug development.^9,10 Despite the use of diverse methodologies to construct hiPSC-derived hepatic organoids,[11], [12], [13], [14] establishing models that faithfully recapitulate the liver’s complex metabolic functions—especially bile acid metabolism—remains a major challenge. Recent studies have utilized hepatic organoids with fluorescent bile acid analogs to model bile flow and analyze the fluid dynamics of these analogs in vitro.[15], [16], [17], [18], [19] However, the content and types of bile acid metabolism in these hepatic organoids are far from those observed in vivo.

Thus, there is an urgent need to develop functional hepatic organoids capable of recapitulating in vivo bile acid metabolism and performing bile acid profiling. In our previous work, we co-differentiated hepatocyte-like cells, cholangio-like cells and various non-parenchymal cells to generate functional hiPSC-derived hepatobiliary organoids (HBOs).^20,21 Notably, we detected three bile acids in the supernatant of maturated HBOs. However, these still have limitations in replicating a complex bile acid profile in vitro, and the underlying mechanisms of bile acids formation in our HBOs are not yet fully understood.

In this study, we introduced saikosaponin A (SSA) as a microenvironmental intervention during the maturation stage of hiPSC-derived hepatobiliary organoids to potentiate bile acid metabolism. This approach led to the generation of an organoid model comprising 33 different types of bile acids, which we have named bile acid-HBOs (BA-HBOs). By integrating molecular characterization, functional assays, and single-cell analyses, we delineated the characteristic of BA-HBOs on hepatobiliary lineage specification, metabolic gene expression, and hepatocyte zonation. Moreover, based on the BA-HBOs, we constructed a liver fibrosis model. Through this model, we identified and analyzed the disruption in bile acid metabolism profiles resulting from liver fibrosis. Our findings support the use of this system as a physiologically and pathologically relevant in vitro model that captures key aspects of human bile acid metabolism.

Materials and methods

Cell culture

The human iPSC line used in this study is UC, which was obtained from Guangzhou Institutes of Biomedicine and Health, Chinese Academy of Science. The cells were maintained in mTeSRplus medium (STEMCELL Technologies, Vancouver, Canada, 100-0276) on plates coated with Matrigel (Corning, NY, USA, 354277) at 37 °C in a humidified atmosphere with 5% CO₂. The medium was refreshed daily, and the cells were passaged every 3–4 days using EDTA (Shownin, Shanghai, China, RP01007), and the ratio was between 1:6 and 1:8.

Differentiation of BA-HBOs

To generate complex BA-HBOs from single iPSCs, we followed the protocol proposed by the Wu group (Table S1).^20,22 A minor proportion of mesodermal cells were retained during the definitive endoderm (DE) stage, whereas mTeSRplus was incorporated as an additional component in the prometaphase differentiation protocol. Before the differentiation, hiPSCs were dissociated into single cells using Accutase (Sigma-Aldrich, A6964) at 37 °C for 6 min. The cells were resuspended in mTeSRplus medium supplemented with 50 μM Y-27632 (Sigma-Aldrich, St. Louis, MO, USA, Y0503) and then plated onto growth factor-reduced Matrigel (Corning, NY, USA, 356231). After 24 h, the medium was changed to fresh mTeSRplus without Y-27632. When the cells had reached a confluence of 100–110%, the medium was replaced with RPMI1640 (GIBCO, MA, USA, 11879020) supplemented with 1% B27 minus insulin (GIBCO, MA, USA, A1895601), 100 ng/ml activin A (NovoProtein, Shanghai, China, C687) and 20 ng/ml BMP4 (Peprotech, Cranbury, NJ, USA, BMP-4) for 3 h (depending on the iPSC line). Then, medium was replaced by RPMI1640 supplemented with 10% mTeSRplus, 1% B27 minus insulin, 100 ng/ml activin A and 20 ng/ml BMP4 for 2 days. Day 3–4: the culture medium was changed to RPMI1640 supplemented with 10% mTeSRplus, 1% B27 minus insulin and 100 ng/ml activin A for another 2 days. Day 5–9: the cultures were placed in RPMI1640 supplemented with 10% mTeSRplus, 1% B27 (GIBCO, MA, USA, 17504-044), 20 ng/ml BMP2 (Peprotech, Cranbury, NJ, USA, BMP-2), and 30 ng/ml FGF4 (Peprotech, Cranbury, NJ, USA, FGF-4) for 5 days. Day 10–15: the medium was replaced by RPMI1640 containing 10% mTeSRplus, 1% B27, 20 ng/ml HGF (NovoProtein, Shanghai, China, CP63-50) and 20 ng/ml KGF (NovoProtein, Shanghai, China, CH73-500) for a 6-day culture. Day 16–25: the cells were cultured in HepatoZYME-SFM (GIBCO, MA, USA, 17705021) medium supplemented with 10% cholesterol⁺ MIX (Chinese patent number: ZL201810211144.X), 10 ng/ml Oncostatin M (OSM) (NovoProtein, Shanghai, China, C099) and 1% Glutamine (GIBCO, MA, USA, A2916801) for 10 days. At this stage, both hepatic and biliary lineages were initially mature. Subsequently, the BA-HBOs were cultured in SFM containing 10% cholesterol⁺ MIX, 10 ng/ml OSM, 0.5 μM dexamethasone (Dex) (MedChemExpress, MCE, Monmouth Junction, NJ, USA, HY-14648) and 1% glutamine, with or without SSA for another 10 days. The medium was changed every 24 h until Day 16, then every 48 h. Table S2 presents a comprehensive overview of the reagents utilized in cell culture and differentiation.

Quantitative gene expression analysis

Total RNA was extracted from the samples using TRIzol™ Reagent (Invitrogen, Waltham, MA, USA, 15596-026). cDNA was synthesized from 500 ng of total RNA using the PrimeScript RT Reagent Kit with gDNA Eraser (Takara, Kusatsu, Shiga, Japan, RR037A). Quantitative PCR was performed with TB Green Premix Ex Taq II (Takara, Kusatsu, Shiga, Japan, RR820A) on a LightCycler 480 II (Roche). Data were quantified by comparing them to a standard curve and subsequently normalized to the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase. The primers used are listed in Table S3.

Flow cytometry and data analysis

Flow cytometry analysis was performed on Days 0, 4, 9, 15, and 25 of the differentiation. The cultures were dissociated enzymatically into single cells with Accutase before Day 15 and TrypLE on Day 25 at 37 °C. The single cells were washed with sterile DPBS and collected by centrifugation at 350 × g for 3 min. Thoroughly resuspend cells and add 250 μl fixation/permeabilization solution (BD, 554714) for 20 min at 4 °C (skip this step when staining cell surface antigens). Then wash the cells two times in Perm/Wash™ buffer (BD, 554714) (1 ml/wash for staining in tubes). Thoroughly resuspend the cells in 100 μl of BD Perm/Wash™ buffer containing a pre-determined optimal concentration of a fluorochrome-conjugated antibody or appropriate negative control. Incubate at 4 °C for 30 min in the dark. Wash the cells two times in Perm/Wash™ buffer (1 ml/wash for staining in tubes) and resuspend in Dulbecco’s phosphate-buffered saline (DPBS) before flow cytometric analysis. Flow cytometry measurements were performed using a BD FACS LSR Fortessa cytometer. FlowJo software (v.10.8.1) was used to process flow cytometry results.

Immunofluorescence

Immunostaining was performed on Day 0, 4, 9, 15, 25, and 35 of the differentiation. The cultures were fixed with 4% paraformaldehyde (Solarbio, Beijing, China, P1110) for 30 min at room temperature, then permeabilized with 0.3% Triton X-100 (Sigma-Aldrich, St. Louis, MO, USA, T9284) in DPBS for 1 h. Wash the cells three times with DPBS, soaking for 5 min each time. After incubating in a blocking solution (5% donkey serum [Solarbio, Beijing, China, SL050] in DPBS) for at least 1 h, the cells were incubated with primary antibodies in 3% donkey serum at room temperature for 1 h or at 4 °C overnight, and subsequently with the secondary antibodies in 1% donkey serum for 30 min at room temperature. If there was a need for nuclear staining, the DAPI would be applied for 3 min. The cultures were washed four times with DPBS for 5 min at room temperature after primary and secondary antibodies and DAPI. All immunofluorescence images were performed using a confocal laser scanning microscope (LSM800, Germany, Carl Zeiss). At least three samples per group were stained, with three non-overlapping fields of view captured per sample. The positive area was quantified using ImageJ (v.ij154) and expressed as a ratio to the total DAPI-stained area. The list of primary and secondary antibodies used in this study were provided in Table S4.

Periodic acid-schiff (PAS) and Oil Red-O (ORO) staining

The PAS staining kit (Solarbio, Beijing, China, G1280) and ORO staining kit (Solarbio, Beijing, China, G1262) were utilized for the experiments in accordance with the manufacturer's instructions.

Indocyanine green (ICG) uptake and release assay

Organoids were incubated with 1 mg/ml ICG (Sigma-Aldrich, St. Louis, MO, USA, 3599-32-4) at 37 °C for 30 min. After medium removal and PBS washing, intracellular ICG accumulation was visually documented using brightfield microscopy. To assess ICG release patterns, organoids were maintained in fresh medium for 12 h followed by microscopic evaluation.

Rhodamine 123 transport assay

The BA-HBOs were pretreated with or without 10 μM verapamil (MCE, Monmouth Junction, NJ, USA, HY-14275) at 37 °C for 30 min. After they were incubated with 10 μg/ml rhodamine 123 (Sigma-Aldrich, St. Louis, MO, USA, R8004) at 37 °C for 1 h, the samples were rinsed three times with DPBS for 5 min and captured with a fluorescence microscopy system (Axio Vert.A1, Germany, Carl Zeiss).

Tunnel assay

The cultures were fixed with 4% paraformaldehyde for 30 min at room temperature, then permeabilized with 0.3% Triton X-100 in DPBS for 10 min. The samples were washed two times with DPBS, then incubated with 100 μl of TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labeling) detection solution (Beyotime, Shanghai, China, C1088) at 4 °C for 30 min in the dark. Finally, the samples were rinsed two times with DPBS for 5 min and captured with a confocal laser scanning microscope (LSM800, Germany, Carl Zeiss).

Single-cell RNA sequencing of BA-HBOs

BA-HBOs and control HBOs were harvested and washed two times with DPBS. The organoids were enzymatically dissociated into single cells using a solution containing type I collagenase (Stemcell, Vancouver, Canada, 07415) and trypsin (GIBCO, MA, USA, 15050057) mixed in a 1:1 ratio, then incubated for 30 min at 37 °C with shaking. The dissociated cells were filtered through a 40 μm cell mesh to achieve a single-cell suspension. Cell viability was assessed using trypan blue staining, with results indicating viability greater than 90%.

The cell suspension was used immediately for single-cell RNA sequencing (scRNA-seq) library preparation using the Singleron Matrix NEO® automated single-cell library preparation system (Singleron Biotechnologies), following the manufacturer’s instructions. For each sample, >5 × 10⁴ viable cells (viability >90%) were loaded to obtain ∼10,000 high-quality cells after quality control filtering. Following cDNA synthesis, single-cell cDNA libraries were prepared according to the Singleron protocol. Quality control was performed after cDNA amplification and library construction using Bioanalyzer analysis and Qubit measurement. Libraries were pooled and sequenced on an Illumina NovaSeq 6000 system with a target read depth of approximately 30,000 reads per cell.

Preprocessing of scRNA-seq data

Sequencing data were aligned and quantified using the Cell Ranger Single-Cell Software Suite against the GRCh38 human reference genome (Ensembl release v.106). For the purpose of quality control and for downstream analysis, the following criteria were applied to barcodes: (1) >1,500 unique molecular identifiers (UMIs), (2) >1,000 genes detected, (3) <20% mitochondrial-encoded DNA, and (4) genes detected in more than three barcodes. Scrublet v.0.2.3 was used to identify potential doublet barcodes, and those with a doublet score above 0.2 were removed from the dataset.

Cross-dataset integration of scRNA-seq data

Public transcriptome profiles were downloaded from the NCBI Gene Expression Omnibus (GEO) database. The single-cell transcriptome of the liver organoid from Gary’s study (GSE154883), Hans’s study (GSE199239), and the human liver atlas (GSE124395) were merged with our scRNA-seq data and plotted into the shared UMAP (uniform manifold approximation and projection) space by Seurat as described above. Cell types were annotated with reference to marker genes from published human datasets.

Single-cell transcriptomic data were retrieved from the NCBI GEO database under accession codes GSE154883 (hiPSC-derived multi-lineage hepatic organoid), GSE199239 (human fetal liver organoid), and GSE124395 (human hepatic reference atlas). These datasets were integrated with in-house generated scRNA-seq profiles through Seurat's harmonized analytical framework (v4.3.0), using reciprocal PCA-based integration followed by joint UMAP dimensionality reduction. Cell-type annotation was performed through systematic interrogation of canonical marker genes derived from established human hepatic cell signatures in peer-reviewed literature.

UCell-based metabolic gene set enrichment analysis

Gene set enrichment analysis was performed on the integrated scRNA-seq datasets using the UCell computational framework (v2.4.0), implementing a rank-based methodology grounded in Mann–Whitney U statistics. Quality-controlled single-cell count matrices were processed through a computational pipeline for signature scoring of predefined metabolic pathways (‘Bile acid metabolism’ and ‘Lipid and cholesterol homeostasis regulation’, Table S5). Enrichment scores were spatially resolved through UMAP manifold projection algorithms, enabling systematic identification of high-metabolic-index cellular domains. Comparative analysis was performed across integrated datasets using non-parametric statistical approaches.

CellChat analysis

Cell–cell communication networks were systematically deciphered using the CellChat computational framework (v1.6.1) through rigorous ligand–receptor interaction analysis. Processed single-cell transcriptomes from BA-HBOs and control HBOs underwent multi-stage computational interrogation: (1) curated CellChatDB (v2023.02) ligand–receptor pairs were probabilistically matched with normalized gene expression matrices; (2) communication probability matrices were derived through mass action principle-based modeling incorporating spatial proximity weights; (3) differential network architectures were resolved via permutation testing comparing SSA-treated vs. untreated cohorts.

Functional enrichment analysis

GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed on differentially expressed genes (q value <0.05, |log₂FC| >1) using clusterProfiler. Significant terms were visualized through bar plots and lollipop charts to highlight enrichment patterns. Redundant GO terms were consolidated using semantic similarity analysis. KEGG pathway mappings were displayed with bar diagrams showing enrichment significance and gene ratios.

Induction of fibrosis in BA-HBOs

On Day 35 of culture, BA-HBOs were treated with transforming growth factor-beta (TGF-β, 10 ng/ml; NovoProtein, CA72) for 5 days consecutively, to induce fibrosis. The control group received an equivalent volume of PBS. On Day 40, conditioned media from the preceding 48 h and organoid samples were harvested for subsequent pathological and phenotypic analyses.

Bile acids metabolomics analyses using LC/MS-MS

Bile acid profiling was systematically performed on BA-HBOs and fibrosis HBOs (FiBA-HBOs), and their conditioned media through a standardized biosampling protocol. To mitigate autocrine feedback effects on bile acid homeostasis, organoid cultures underwent scheduled medium renewal every 48 h from Day 25 onward. Following terminal medium replacement at Day 38, the 48-h conditioned media were collected as extracellular bile acid samples through sterile aspiration.

Organoids were enzymatically dissociated using our established single-cell isolation protocol after triple PBS washing. Cell counting data were used to normalize bile acid concentration calculations. The cell suspension underwent cold centrifugation (100 × g, 5 min, 4 °C) to generate intracellular bile acid samples through complete supernatant removal. Both cellular and extracellular fractions underwent immediate vitrification in liquid nitrogen followed by cryopreservation at -80 °C until liquid chromatography-tandem mass spectrometry (LC-MS) analysis.

Targeted analysis of 33 bile acid species was conducted using LC-MS with multiple reaction monitoring. Chromatographic separation was performed on a reverse-phase C18 column employing gradient elution, coupled with mass spectrometric detection in dual polarity mode using optimized collision energies. A standardized reference panel encompassing all bile acids was established through serial dilution of chemically defined standards to create calibration curves across physiologically relevant concentration ranges (Table S6).

Sample preparation involved protein precipitation followed by vacuum-assisted solvent evaporation. The quantification methodology incorporated isotope-corrected calibration with inter-batch verification protocols, whereas method validation confirmed measurement precision across three concentration thresholds and demonstrated 85–115% accuracy in analyte recovery. Continuous system suitability monitoring ensured analytical consistency throughout the experimental sequence.

Statistical analysis

Statistical analyses were performed by GraphPad Prism 9.5.1 (GraphPad Software, San Diego, CA, USA). Intergroup differences were assessed with Student's t test (two-group comparisons) or one-way ANOVA (multi-group comparisons). Significance thresholds were defined as ∗p <0.05, ∗∗p <0.01 and ∗∗∗p <0.001. Error bars represent standard deviation. Complete statistical parameters are annotated in corresponding figure captions.

Results

Generation and functional maturation of multi-lineage bile acid-HBOs from hiPSCs

To evaluate bile acid metabolism in hiPSC-derived hepatobiliary organoids, we optimized our previously established differentiation protocol.[20], [21], [22] SSA, the main bioactive compound in Radix bupleuri (an herb traditionally used to treat liver diseases),²³ acts as a potent peroxisome proliferator-activated receptor (PPAR) agonist and upregulates liver X receptor (LXR) and retinoid X receptor (RXR), thereby promoting hepatic cholesterol-to-bile acid conversion.[24], [25], [26], [27], [28], [29] We supplemented the maturation medium with SSA to enhance bile acid metabolism in HBOs (Fig. S1A) and defined an optimal concentration that increased metabolic gene expression without cytotoxicity (Fig. S1B–E).

An overview of the protocol for generating bile acid-enhanced HBOs (BA-HBOs) is presented in Fig. 1A. We assessed hiPSC pluripotency before differentiation, with ≥99% of cells co-expressing NANOG and OCT4, as confirmed by immunofluorescence and flow cytometry (Fig. S2A). Multiple assays consistently verified the reliability of each developmental stage (Fig. 1B and Fig. S2). At Day 4, ∼75% of cells co-expressed definitive endoderm markers (FOXA2 and SOX17), while ∼20% were positive for the mesodermal marker (TBXT) (Fig. S2B and C). Hepatic specification became evident by Day 9, as indicated by the co-localization of HNF4α and AFP in progenitor cells. At Day 15, hepatoblasts retained bipotency, with concurrent expression of SOX9 and AFP. By Day 35, organoid maturation yielded spatially organized lineage segregation, with CYP3A4+ hepatocytes adjacent to CK19+ cholangiocytes, thus establishing structured hepatobiliary architecture (Fig. 1B). Sequential qPCR profiled developmental genes expression during HBO maturation (Fig. S2D and E), capturing progression from early specification to terminal differentiation. Immunofluorescence further confirmed that BA-HBOs synthesize abundant albumin (ALB) and express the metabolic enzyme CYP3A4 (Fig. 1C). Meanwhile, cholangiocytes exhibited a columnar epithelioid morphology, with AC-α-tubulin staining revealing polarized luminal cilia and activation of NOTCH2 signaling (Fig. 1D). Functional assays confirmed mature hepatic and biliary activities in BA-HBOs, including glycogen synthesis, lipid metabolism, capacity of ICG clearance, and calcium-dependent transporter function (Fig. 1E–H). Notably, in maturation-stage BA-HBOs, hepatocytes formed layers surrounding cholangio-like structures (Fig. S2F), recapitulating hepatic cord-like organization and expanding the functional surface area for metabolism.

Fig. 1.

Generation and functional assessment of reproducible bile acid-hepatobiliary organoids (BA-HBOs) from hiPSCs.

(A) Schematic illustrating the differentiation process for complex BA-HBOs. (B) Fluorescence micrographs depicting lineage-specific protein markers in BA-HBOs at distinct developmental stages (Days 4, 9, 15, and 25). Immunofluorescence staining: SOX17/FOXA2 (endoderm cells), HNF4α/AFP (HPs), SOX9/AFP (HBs), and CK19/CYP3A4 (mature cholangiocytes/hepatocytes). (C) Immunofluorescence reveals HNF4α-positive hepatocytes in mature BA-HBOs actively secreting ablumin (ALB) and CYP3A4. (D) Immunofluorescence analysis demonstrates cholangiocytes in mature BA-HBOs exhibit both abundant acetylated α-tubulin-labeled primary cilia and substantial Notch2 expression critical for biliary lumen morphogenesis. (E) PAS (Periodic acid-Schiff) staining demonstrates glycogen storage capacity in BA-HBO-derived hepatocytes. (F) ORO (Oil Red-O) staining reveals lipid droplet accumulation in differentiated hepatocytes. (G) Functional assays confirm ICG (indocyanine green) uptake and subsequent biliary excretion in BA-HBO-cultured hepatocytes. (H) Immunofluorescent images demonstrate uptake of Rhodamine 123 into the lumen of organoids without or with Verapamil to inhibit uptake. All scale bars = 50 μm. HBs, hepatoblasts; hiPSC, human induced pluripotent stem cell; HPs, hepatic progenitors.

BA-HBOs recapitulate bile acid synthesis and secretion

To determine the enhancement of hepatic bile acid synthesis and cholangiocyte transport functions in BA-HBOs, we used our 2019 HBO protocol as control.²⁰ Sustained upregulation of rate-limiting bile acid synthases (CYP7A1, CYP7B1, CYP27A1, CYP8B1) from Day 25 to 35 indicated enhanced hepatic biosynthesis of BA-HBOs (Fig. 2A). And immunofluorescence staining showed that HNF4α+ hepatocytes in BA-HBOs expressed higher levels of the major bile acid transporters BSEP and NTCP than those in control organoids (Fig. S3A and B). Furthermore, brightfield microscopy–based morphometric analysis showed increased ductal wall thickness in BA-HBOs (Fig. S3C and D). In parallel, cholangiocyte bile acid transporters SLC51A and SLC51B (OSTα/β) and hepatocyte bile acid metabolic enzyme SULT2A1 (sulfotransferase) and efflux transporter ABCB11 (BSEP) were significantly upregulated (Fig. S3E). Immunofluorescence analysis further confirmed elevated expression of CFTR, AQP1, ASBT, and ZO-1 in bile ductules (Fig. 2B and C), demonstrating improved structural integrity and transport functionality in BA-HBOs.

Fig. 2.

BA-HBOs exhibit enhanced bile acid metabolic function and increased biliary maturation.

(A) The expression of rate-limiting enzymes involved in both the classical (CYP7A1, CYP8B1) and alternative (CYP7B1, CYP27A1) bile acid synthesis pathways was upregulated in BA-HBOs. (B) Immunofluorescence staining and (C) quantitative analysis showed increased expression of ductal functional markers (CFTR, AQP1, ASBT, ZO-1) in BA-HBOs compared with control HBOs, as assessed by the ratio of marker-positive area to CK7-positive biliary epithelial area. (D) Bar graphs show the bile acids profile, including their composition and concentrations, in BA-HBOs (top) and supernatants (bottom). (E) Donut charts display the proportions of tauro-conjugated (TCBA), glyco-conjugated (GCBA), and unconjugated (UCBA). All scale bars = 50 μm. Statistical significance was determined using unpaired t test (, ∗p <0.05, ∗∗p <0.01, ∗∗∗p <0.001), n = 3 biological replicates. BA-HBOs, bile acid-HBOs; ns, not significant.

Based on these findings, we next examined whether this potentiation is mediated by nuclear receptor activation. We focused on the PPAR and LXR pathways and applied their respective agonists or antagonists for 48 h at the late culture stage to minimize non-specific toxicity and disruption of cellular homeostasis associated with prolonged ligand exposure. ELISA analysis showed that both SSA and the PPAR agonist GW7647 (200 nM, MCE, HY-13861) increased CYP7A1 and CYP27A1 expression, whereas the effect of the LXR agonist GW3965 (1 μM, MCE, HY-10627) on these enzymes was comparatively modest (Fig. S4A). Consistently, the PPAR antagonist GW6471 (5 μM, Proteintech, CM06124) more strongly attenuated the SSA-induced upregulation of these key bile acid–synthetic enzymes than the LXR antagonist GSK2033 (0.5 μM, MCE, HY-108688), and total bile acid levels followed the same pattern (Fig. S4A and B). Interestingly, the opposite trend was observed for bile acid transporters: PPAR and LXR agonists both enhanced ASBT and MDR3 expression to some extent, but LXR blockade more effectively blunted the SSA-induced increase in these transporters than PPAR blockade (Fig. S4C). These data suggest that SSA primarily promotes bile acid synthesis through activation of PPAR, whereas it mainly enhances bile acid transport and excretion through activation of LXR.

To directly access the physiological metabolic state of BA-HBOs, we conducted LC/MS-MS for bile acids metabolomics. Remarkably, 33 distinct bile acids were detected in BA-HBOs, an unprecedented finding compared with previous studies (Fig. 2D and Table S6). These bile acids, present in both BA-HBOs and supernatants, were categorized as tauro-conjugated (TCBA), glyco-conjugated (GCBA), and unconjugated (UCBA) forms, with GCBA predominating (Fig. 2E), consistent with human physiology.³⁰

Together, these findings highlight the advanced hepatocytic and biliary epithelial functionality of BA-HBOs in vitro and underscore their value as a model for human bile acid metabolism.

BA-HBOs resemble the transcriptional landscape and core metabolic programs of the human liver

To define the transcriptional and metabolic profiles of BA-HBOs, we performed scRNA-seq on BA-HBOs and control HBOs (2019 protocol)²⁰ at Day 35, and integrated these with published datasets from different samples, including iPSC-derived multi-lineage hepatic organoids (Gary’s study),³¹ fetal liver organoids (Hans’s study),³² and adult hepatic tissues.³³ After quality control, 36,103 cells were retained for downstream analysis (Fig. 3A). Unsupervised clustering identified eight transcriptionally distinct populations corresponding to major hepatic lineages, with cell-type annotations supported by canonical marker gene expression (Fig. 3B). Both BA-HBOs and HBOs captured key transcriptional features of several principal hepatic lineages observed in adult human liver, including hepatocytes, cholangiocytes, hepatic stellate cells, and others, although endothelial cells—one of the most abundant populations in adult liver—were only sparsely represented in BA-HBOs. Overall, the HBO system more closely recapitulated the cellular diversity and lineage composition of adult human liver than previously reported organoid models (Fig. 3C), although differences in cell-type proportions relative to in vivo tissue remain.

Fig. 3.

Single-cell transcriptome analysis of the liver organoids and adult liver data.

(A) UMAP plot of single cells derived from our, Gary’s and Hans’s liver organoids and human liver. (B) Expression of marker genes for annotation of the eight main populations. (C) Barplots showing cell cluster proportions as fractions of total cells across five biological samples. (D) Box plot analysis of Ucell-derived activity scores for bile acid metabolism and lipid and cholesterol metabolism across hepatocytes of five biological samples. Statistical significance was determined using ANOVA test (∗∗∗p <0.001). BA, bile acid; CHs, cholangiocytes; CON, control; ECs, endothelial cells; HCs, hepatocytes; HPCs, hepatic progenitor cells; HSCs, hepatic stellate cells; KCs, Kupffer cells; MeCs, mesenchymal cells; TLs, T lymphocytes.

Comparative analysis of major cell types showed that, in hepatocytes, our organoids displayed the highest enrichment of key metabolic pathways—including drug metabolism–cytochrome P450, primary bile acid biosynthesis, retinol metabolism, and taurine/hypotaurine metabolism—most closely resembling adult liver (Fig. S5A). In cholangiocytes, our model upregulated genes associated with Hippo signaling, gap junctions, and mineral absorption, indicating preserved epithelial barrier and transport functions (Fig. S5A). Hepatic progenitors maintained active Wnt and Hedgehog signaling and cell-cycle programs, consistent with a pro-regenerative, stem-like state (Fig. S5A). Our organoids showed higher expression of representative metabolic markers (APOA1, TTR, GLUL) than control and fetal liver organoids. Expression of key synthetic and matrix-related genes (FGA, SERPINA1, RBP4, FN1) was closer to adult liver levels (Fig. S5B), consistent with a relatively mature hepatic parenchymal state.

To further dissect bile acid and lipid/cholesterol metabolism across models, we performed UCell scoring using curated pan-metabolic gene sets (Table S5) across five single-cell datasets. Quantitative profiling of hepatocytes revealed distinct metabolic maturation states among experimental groups (Fig. 3D and Fig. S6). Notably, BA-HBOs showed significantly higher enrichment scores for bile acid biosynthesis and lipid/cholesterol metabolism than fetal liver organoids and control HBOs (p <0.001), and greater similarity to adult liver (Fig. 3D).

Restructuring of the hepatocyte population in BA-HBOs

Given the spatial and functional zonation of hepatocytes in bile acid and cholesterol metabolism,^34,35 we performed iterative subclustering of single-cell transcriptomic data to identify metabolically specialized hepatocyte subpopulations and quantify changes in their proportions. Unsupervised clustering resolved five transcriptionally distinct hepatocyte subpopulations, labeled Hep1 to Hep5 (Hep1–Hep5; Fig. 4A), each defined by a set of signature genes (Fig. 4B). BA-HBOs showed increased proportions of Hep1, Hep3, and Hep5, accompanied by reduced Hep2 and Hep4 (Fig. 4A and C).

Fig. 4.

Hepatocyte subclustering reveals transcriptional heterogeneity with distinct metabolic signatures.

(A) Hepatocyte subclustering via UMAP analysis in organoid identifies five distinct subpopulations (Hep1–5) and delineates cohort-specific architectures in BA-HBO vs. control groups. (B) Differential gene dot plot delineates top-ranked cluster-specific markers across five hepatocyte subpopulations. (C) Donut chart visualization delineates distinct proportional allocation of hepatocyte subpopulations between BA-HBO and CON-HBO. Lollipop panel series (D–H) systematically maps Hep1–5 subpopulation-specific GO enrichment profiles, with individual subpanels revealing distinct pathway signatures. BA-HBO, bile acid-HBO; CON-HBO, control HBO; Hep, hepatocyte; GO, Gene Ontology.

In mammalian hepatic lobules, hepatocytes display functional zonation based on their location along the portocentral axis.^35,36 To relate these clusters to liver zonation, we compared their transcriptional profiles with established periportal, interzonal, and central venous markers.³⁷ Hep1 and Hep2 were predominantly interzonal (Fig. S7A and D), consistent with roles in hepatic regeneration during homeostasis and injury. Hep3 displayed a hybrid signature, co-expressing central venous and interzonal markers (Fig. S7A, C, and D), suggesting a transitional state during functional repolarization. Hep4 did not show a clear zonal affiliation, whereas Hep5 was mainly periportal (Fig. S7A and B), consistent with enhanced metabolic activity.

To characterize functional specialization, we performed GO and KEGG analyses on highly expressed, cluster-specific genes. Hep1 and Hep2 displayed distinct functional profiles (Fig. 4D and E and Fig. S8). Hep1 was enriched for protein synthesis pathways, particularly cytoplasmic translation and ribonucleoprotein assembly, consistent with a proliferative or growth-supportive state. By contrast, Hep2 was enriched for stress-adaptive pathways, including endoplasmic reticulum stress responses and macroautophagy. Hep3 and Hep5, which were expanded in BA-HBOs, showed upregulation of genes involved in bile acid, cholesterol, and fatty acid metabolism (Fig. 4F and H and Fig. S8). Conversely, hep4 displayed significant downregulation of these metabolic pathways (Fig. 4G and Fig. S8).

Together, these data indicate that hepatocyte subpopulations are restructured in BA-HBOs, with expansion of metabolically active, zonally distinct clusters that contribute to organoid metabolic function.

Modeling liver fibrosis-associated bile acid dysregulation with multi-lineage BA-HBOs

Multi-lineage liver organoids serve as ideal models for liver fibrosis, as they require both hepatocytes and hepatic stellate cells (HSCs).⁷ We confirmed the presence of these cell types in BA-HBOs (Fig. 3B), enabling us to establish a fibrosis model to study bile acid dysregulation. TGF-β, a central driver of fibrosis, induces collagen deposition and inflammatory responses in the liver.³⁸ We treated BA-HBOs with TGF-β to generate fibrotic BA-HBOs (FiBA-HBOs; Fig. 5A). Brightfield imaging showed disrupted hepatocyte architecture, bile duct disorganization, cellular debris, and extracellular matrix deposition (Fig. S9A). As expected, FiBA-HBOs displayed increased expression of collagen production and HSC activation markers (Fig. 5B–D and Fig. S9B and C). They also upregulated inflammation-related factors (TNF-α, IL-2, IL-6, and IL-8), and showed elevated alanine aminotransferase/aspartate aminotransferase (ALT/AST) and γ-glutamyl transferase (GGT), indicating impaired hepatic and biliary function (Fig. 5B and E).

Fig. 5.

Construction of a fibrosis model Using BA-HBOs.

(A) Schematic diagram of the process for inducing FiBA-HBOs using TGF-β and performing pathological phenotype analysis. (B) Quantitative PCR analysis showed increased expression of hepatic stellate cell (HSC) activation-related genes (α-SMA, MMP2, COL3A1, COL1A1) and inflammatory cytokine-related genes (TNF-α, IL-2, IL-6, IL-8) in FiBA-HBOs. (C) Immunofluorescence and (D) fluorescence intensity quantification demonstrated that FiBA-HBOs exhibited increased numbers of activated α-SMA-positive HSC and greater COL3A1 deposition than BA-HBOs. (E) ELISA assays of the supernatant revealed that FiBA-HBOs exhibited increased levels of GGT, AST, and ALT. (F) The total bile acid content was significantly increased within FiBA-HBOs cultures, whereas no significant difference was observed in the 48-h culture supernatant. Statistical significance was determined using unpaired t test (∗p <0.05, ∗∗p <0.01, ∗∗∗p <0.001), n = 3 biological replicates. ALT, alanine aminotransferase; AST, aspartate aminotransferase; BA-HBOs, bile acid-HBOs; FiBA-HBOs, fibrosis BA-HBOs; GGT, γ-glutamyl transferase; ns, not significant; TGF-β, transforming growth factor-β.

These results indicate that FiBA-HBOs can be used to model key features of liver fibrosis. We next explored the alterations in bile acid metabolism. Using LC/MS-MS, we quantified bile acids in both cells and culture supernatants from BA-HBOs and FiBA-HBOs. Although bile acid levels in the supernatant were not significantly increased, intracellular bile acids were markedly elevated in FiBA-HBOs, suggesting cholestasis-like accumulation (Fig. 5F). Principal component analysis (PCA) demonstrated a clear separation between BA-HBOs and FiBA-HBOs (Fig. 6A), and clustering analysis revealed distinct patterns of individual bile acid species (Fig. 6B).The bile acids that increased in FiBA-HBOs largely overlapped between intracellular and supernatant fractions (Fig. 6C and D). Most significantly elevated bile acids were unconjugated (UCBA; Fig. 6E), which are generally more hydrophobic and cytotoxic than their glycine- or taurine-conjugated forms. Among the elevated UCBAs, hydrophobic species included deoxycholic acid (DCA), lithocholic acid (LCA), chenodeoxycholic acid (CDCA), 7-ketodeoxycholic acid (7-KDCA), and 6,7-diketolithocholic acid (6,7-DKLCA) (Fig. 6F). Notably, this pattern of UCBA accumulation is consistent with prior reports of bile acid profile changes in liver fibrosis.[39], [40], [41]

Fig. 6.

Bile acid-targeted metabolomic profiling revealed fibrosis-related bile acid metabolic dysregulation in FiBA-HBOs.

(A) Comparative principal component analysis (PCA) of bile acid profiles between FiBA-HBOs and BA-HBOs in organoid (left) and supernatant (right). (B) Heatmap visualization comparing FiBA-HBOs vs. BA-HBOs in organoid (left) and supernatants (right) displaying mean-normalized concentrations of 33 targeted BAs. (C) Volcano plot representing the levels of significantly changed BAs up- or downregulated in organoid (left) and supernatants (right). (D) Venn diagrams of up- (left) and downregulated (right) BA species between organoid and paired supernatants. (E) Donut charts comparing distribution patterns of three bile acid classes, FiBA-HBOs vs. BA-HBOs in organoid (top) and supernatant (bottom) conditions. (F) Violin plots compare concentrations of DCA, LCA, CDCA, UDCA, 7-KDCA, and 6,7-DKLCA in organoid (μg per 10⁶ cells) and in supernatant (μg/ml/48 h) between FiBA-HBOs and BA-HBOs. Statistical significance was determined using unpaired t test (∗p <0.05, ∗∗p <0.01, ∗∗∗p <0.001), n = 3 biological replicates. 6,7-DKLCA, 6,7-diketolithocholic acid; 7-KDCA, 7-ketodeoxycholic acid; BA-HBOs, bile acid-HBOs; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; FiBA-HBOs, fibrosis BA-HBOs; LCA, lithocholic acid; ns, not significant; UDCA, ursodeoxycholic acid.

These findings suggest that FiBA-HBOs recapitulate key pathological features of liver fibrosis in vivo, including disruptions in bile acid metabolism. They may provide a useful platform for evaluating drug-induced hepatotoxicity and exploring anti-fibrotic therapies.

Discussion

Liver organoids derived from hiPSCs have enabled significant advances in developmental studies, disease modeling, and drug discovery.^9,10,[42], [43], [44] These achievements highlight their promise as novel in vitro models. A critical challenge, however, remains the faithful reconstitution of hepatic metabolic competence in vitro, particularly endogenous bile acid synthesis and canalicular excretion, which require coordinated interaction between hepatocytes and cholangiocytes. Several recent hiPSC-based studies have begun to tackle specific facets of this challenge: systems that promote the formation of extended functional bile canaliculi and visualize bile flow using fluorescent substrates;[15], [16], [17] 3D co-culture approaches that generate functional human bile ducts by mimicking liver epithelial–vascular interactions;¹⁹ and long-term iPSC-derived organoids with sustained hepatocyte-like functions driven by bile acid–FXR signaling.⁴⁵ However, these constructs predominantly function as bile acid transport simulators rather than as biosynthetic units, and consequently fail to restore physiological bile acid profiles.[15], [16], [17], [18], [19] Here, we describe multi-lineage HBOs (BA-HBOs) derived from hiPSCs that approximate human hepatic cellular composition and exhibit key features of mature metabolic function. Notably, this model supports the in vitro biosynthesis of 33 bile acid species, substantially extending the bile acid complexity reported for iPSC-derived hepatic organoids.

In BA-HBOs, SSA–a potent PPAR agonist that has also been linked to increased LXR activity^28,29,46,47—modulates bile acid metabolism through differential engagement of nuclear receptor pathways. Pharmacological perturbation of PPARα and LXR signaling at the late culture stage indicated that PPARα signaling is mainly responsible for the SSA-induced upregulation of CYP7A1 and CYP27A1 and the associated increase in total bile acid levels, whereas LXR signaling is more closely related to the induction of bile acid transporters such as ASBT and MDR3 (Fig. S4). This enzyme activation pattern is consistent with findings from other studies.^25,27 Notably, although earlier studies reported that PPARα suppresses bile acid synthesis via inhibition of CYP7A1, its effect appears to be context-dependent; the promotive effect observed here may therefore reflect the metabolic context of our organoid system. Our results confirm a significant increase in both the concentration and diversity of bile acids in BA-HBOs, encompassing 33 distinct bile acid species (Fig. 2D). The total bile acid levels in BA-HBOs (56.5–61.8 μg per 10⁶ cells, Fig. 5F) exceed those of primary human hepatocytes in sandwich culture (∼40 μg per 10⁶ cells) reported previously.⁴⁸

To systematically evaluate the cellular composition and metabolic discrepancies between BA-HBOs, HBOs, and both human liver tissues and established multicellular hepatic organoids, we integrated our scRNA-seq data with the established datasets atlases for joint cell-type annotation and population analysis (Fig. 3). We confirmed the co-differentiation of non-parenchymal cell lineages—including mesenchymal cells, HSCs, endothelial cells, and Kupffer cells of mesodermal origins—alongside hepatoblasts, hepatocytes, and cholangiocytes of endodermal origin. Although multicellular hepatic organoids have been established, persistent deficits in cellular diversity and population proportions remain.^12,31 BA-HBOs approximated major hepatic cell populations and certain features of adult liver architecture, thereby reducing lineage biases observed in current organoid platforms and providing an improved in vitro model. And BA-HBOs showed higher metabolic capacity than other hepatic organoids and fetal liver organoids, with metabolic profiles that were generally closer to adult liver tissue (Fig. 3D and Figs S5 and S6). Nonetheless, differences in sample processing, sequencing depth, and analysis pipelines across studies may still contribute to some of the observed variation in pathway scores and marker expression, so we interpret these comparisons as indicative trends rather than strictly quantitative benchmarks.

BA-HBOs exhibited a proportional reduction of hepatic progenitor cells (HPCs) accompanied by an expansion of mature hepatocytes (Fig. 3C). Subclustering analysis identified five distinct hepatocyte subpopulations with altered in proportions in BA-HBOs (Fig. 4A and C). Hep1 and Hep2 showed transcriptional similarity to human interzonal hepatocytes but appeared to occupy distinct functional states. BA-HBOs had a higher proportion of Hep1, enriched for protein synthesis and regeneration-related pathways, and fewer Hep2, enriched for ER stress responses and autophagic clearance (Fig. 4D and E). These differences may reflect culture-induced stress and immune activation in organoids,[49], [50], [51] whereas SSA, through antioxidant-mediated hepatoprotection, may mitigate such stress and help maintain BA-HBO integrity.^26,52 Of particular interest are Hep3 and Hep5, whose transcriptional signatures align with periportal and pericentral zonation markers respectively (Fig. S7). BA-HBOs showed a significant expansion of both subsets, which were enriched in core hepatic metabolic pathways, including cholesterol homeostasis, lipid remodeling, and ethanol detoxification. Hep5 exhibited a stronger periportal-like spatial identity with concomitant prominence in bile acid handling machinery.

Numerous studies have demonstrated that hepatic fibrosis is frequently associated with disturbances in bile acid metabolism, with aberrant bile acids acting as critical modulators and potential promoters of fibrosis progression.^41,53,54 However, in vitro modeling of liver fibrosis with disordered bile acid metabolism is particularly challenging, as it necessitates a sophisticated multicellular system—incorporating hepatocytes, cholangiocytes, and stellate cells—that can closely recapitulate relevant bile acid metabolic pathways. Our FiBA-HBO model addresses this need by integrating these key cell lineages, thereby faithfully mimicking the pathological process (Fig. 5, Fig. 6). During fibrosis progression in this system, we observed significant elevations of unconjugated BAs, particularly DCA, LCA, CDCA, and ursodeoxycholic acid (UDCA), consistent with clinical findings in metabolic dysfunction-associated steatotic liver disease/metabolic dysfunction-associated steatohepatitis and viral hepatitis–related fibrosis.[55], [56], [57], [58] Notably, elevated levels of 7-KDCA and 6,7-KLDCA, where hydroxy groups are replaced by keto groups, were detected. This substitution increases hydrophobicity and toxicity, suggesting a potential contribution to fibrosis progression.^59,60

In summary, our hiPSC-derived BA-HBOs synthesize and secrete multiple bile acid species and recapitulate some key transcriptional and metabolic characteristics of human liver tissue. In addition, BA-HBOs can be used to establish an in vitro liver fibrosis model that captures bile acid–related disturbances. Although further refinement will be required to more fully approximate adult liver biology, these models offer a practical platform for basic research and early-stage translational studies.

Abbreviations

6,7-DKLCA, 6,7-diketolithocholic acid; 7-KDCA, 7-ketodeoxycholic acid; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BA-HBOs, bile acid-HBOs; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; FiBA-HBOs, fibrosis BA-HBOs; GCBA, glyco-conjugated BA; GGT, γ-glutamyl transferase; GO, Gene Ontology; HBOs, hepatobiliary organoids; hiPSC, human induced pluripotent stem cell; HPCs, hepatic progenitor cells; HSCs, hepatic stellate cells; ICG, indocyanine green; KEGG, Kyoto Encyclopedia of Genes and Genomes; LCA, lithocholic acid; LC-MS, liquid chromatography-tandem mass spectrometry; PCA, principal component analysis; qPCR, quantitative PCR; scRNA-seq, single-cell RNA sequencing; SSA, saikosaponin A; TCBA, tauro-conjugated BA; TGF-β, transforming growth factor-β; UCBA, unconjugated bile acid; UDCA, ursodeoxycholic acid.

Authors’ contributions

Conceived the study: JuX, JL, FW. Performed the experiments: JuX, JK, JiX, XC, RC. Supervised the work: XS, FW. Contributed to the discussion of the results: XS, SL. Wrote the manuscript: JuX, XS, FW.

Data availability

All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary material. Requests for data can be sent to wufenfang19@126.com.

Financial support

This work was supported by grants from the National Key R&D Program of China (2025YFC3408900), the Shenzhen Medical Research Fund (SMRF.D2301015), the Shenzhen Science and Technology Program (JCYJ20220818103407016), the National Natural Science Foundation of China (82172107), the Guangdong Basic and Applied Basic Research Foundation (2024A1515011222), and the Special Funds for Strategic Emerging Industry of Shenzhen (Grant No. F-2022-Z99-502266), and the Shenzhen Longgang District Science and Technology Innovation Special Fund (LGKCYLWS2022007).

Conflicts of interest

The authors declare no conflicts of interest.

Please refer to the accompanying ICMJE disclosure forms for further details.

Supplementary data

The following are the Supplementary data to this article: