Leveraging IGF signaling to improve the spatial organization and regenerative potential of iPSC-derived vascularized liver organoids

Abstract

Liver tissue engineering offers a promising alternative for end-stage liver disease, yet the recreation of functional vasculature remains a major bottleneck to clinical translation. Here, we developed vascularized liver organoids by integrating human induced pluripotent stem cell (iPSC)–derived hepatoblasts and endothelial cells into decellularized scaffolds functionalized with an anti-CD31 aptamer–based vascular coating agent (VCA). This facilitated spatially coordinated organization of vasculature and parenchyma. Spatial transcriptomic profiling and subsequent functional perturbation demonstrated IGF2-IGF1R-AKT/MAPK signaling as a key axis governing spatial organization and functional maturation of the liver organoids. Furthermore, exogenous IGF2 synergized with the VCA to augment the structural and functional refinement of liver organoids, which translated into markedly improved therapeutic outcomes following transplantation into a chronic liver failure mouse model. Collectively, these findings establish a comprehensive framework for generating physiologically relevant liver tissues from iPSCs and demonstrate the utility of spatial transcriptomics for uncovering regenerative mechanisms. This approach advances the feasibility of autologous, transplantable liver grafts for personalized regenerative therapy.

IGF signaling and aptamer-guided vasculature shape regenerative iPSC-derived liver organoids for chronic liver failure repair.

INTRODUCTION

End-stage liver failure remains a notable global health challenge, with the number of affected patients continuously increasing (1). Although liver transplantation is the most effective treatment option, the severe shortage of donor organs has driven the demand for alternative therapeutic strategies. In response, several approaches, such as three-dimensional (3D) printing and scaffold-based tissue engineering, have been extensively explored to develop transplantable liver tissues (2, 3). During fabrication of such tissues, biocompatible scaffolds are essential not only to replicate the 3D microenvironment of the cells but also to support cell survival, migration, and differentiation. Thus, the significance of the proper scaffolds has emerged in the context of constructing histologically organized, biomimetic tissues. Among various biomaterials, decellularized extracellular matrix (dECM) scaffolds, which are naturally derived from animal organs, inherently retain tissue-specific architecture and cell-binding motifs that promote cellular attachment and tissue-specific functions (4, 5). Notably, whole-organ dECM scaffolds offer the structural complexity needed to reconstruct liver tissue in a manner that closely mimics both the morphology and function of the native liver. Thus, whole-organ dECM scaffolds have increasingly been investigated as a platform for developing transplantable bioengineered grafts (6).

Despite these advances, the generation of clinically relevant liver grafts still faces major hurdles. First, the use of allogeneic cells can trigger immune responses, thereby limiting graft acceptance. The application of patient-derived induced pluripotent stem cells (iPSCs), differentiated into hepatic lineages, provides a compelling strategy to mitigate immune rejection and produce personalized bioengineered liver tissues (7, 8). Nevertheless, even with autologous cell sources, insufficient vascularization within bioengineered tissues remains a critical bottleneck, which often results in ischemia or hyperacute rejection postimplantation (9). To address this challenge, numerous studies have explored reendothelialization strategies in bioengineered constructs by modifying the vascular luminal surface to enhance cell-ECM interactions (10–12). In our previous work, we developed an effective strategy using an anti-CD31 aptamer as a vascular coating agent (VCA), which significantly enhanced the adhesion and vessel formation of human umbilical cord endothelial cells (ECs) within the dECM scaffolds (13). This approach holds promise for promoting functional vasculature and improving the engraftment potential of engineered liver tissues.

Given the considerable size and architectural complexity and heterogeneity of bioengineered tissues, the functional identity of each cell type is deeply linked to its precise spatial localization within the tissue microenvironment. For instance, hepatocytes residing in the parenchymal regions must perform liver-specific metabolic functions, while ECs are expected to localize along the vascular lumens to maintain vascular integrity and perfusion. Therefore, bulk-level analyses that average gene expression across the entire tissue may obscure these critical spatially defined roles. It is thus imperative to consider the spatial localization of cells within the tissue to accurately capture the functional heterogeneity and compartmentalization of cells within bioengineered tissues. Although single-cell RNA sequencing, such as Smart-seq2, has provided unprecedented insights into cellular heterogeneity at a high resolution, it inevitably loses spatial context, which is crucial for understanding how tissue architecture influences cellular behavior and function (14–16). In this light, integrative spatial transcriptomic analysis, combining spatial coordinates with gene expression data, rather than relying solely on single-cell resolution, offers a powerful strategy to investigate how cells interact with their local environment and how regional gene expression patterns contribute to tissue-level function and organization. State-of-the-art spatial transcriptomic platforms are generally classified into two categories: counting-based and probe-based approaches. Counting-based platforms, such as 10X Visium (16) or Slide-seq (17), capture polyadenylated transcripts on spatially barcoded arrays and quantify them by next-generation sequencing, thereby enabling unbiased transcriptome-wide profiling with spatial context. However, they are limited by relatively low spatial resolution and spot level signal averaging. Probe-based approaches including 10X Xenium (18) or MERFISH (19) rely on targeted in situ hybridization or imaging-based detection of predefined gene panels, enabling high spatial resolution down to the single-cell or even subcellular level; nonetheless, they are constrained by limited transcript coverage and the need for prior target selection.

Within this spectrum, phenotype-based high-throughput laser-aided isolation (PHLI) represents a complementary approach that focuses on region of interest (ROI)–based transcriptomic analysis. PHLI sequencing (PHLI-seq) integrates imaging with laser-mediated pressure catapulting to isolate single cells or small cell clusters on the basis of their histological phenotype and precise anatomical location within tissue sections, as demonstrated in our previous work (20, 21). Given that the isolated cells are subjected to high-throughput RNA sequencing, PHLI-seq enables capturing full-length transcriptomic analysis in spatially defined regions within large heterogeneous samples. Unlike other spatial transcriptomic platforms with limited spatial resolution or transcript coverage, PHLI-seq provides granular resolution with broad coverage across tissues, making it valuable for dissecting spatial heterogeneity in structurally complex engineered tissues. As such, PHLI-seq serves as a valuable complement to existing platforms, bridging high-resolution phenotypic selection with unbiased transcriptomic readout.

In this study, our primary objective is to fabricate iPSC-derived bioengineered liver organoids with functional vascular networks by using dECM liver scaffolds loaded with an anti-CD31 aptamer. To determine the mechanisms by which ECs recruited via the aptamer efficiently integrate and organize within the tissue, we performed in-depth spatial transcriptomic analysis of the vascularized liver organoids. PHLI-seq was used to isolate specific regions, including liver parenchyma and vasculature, from tissue sections, followed by transcriptomic profiling of each compartment. Through this integrated analysis of spatially resolved gene expression and cellular organization, we identified insulin-like growth factor 2 (IGF2)-IGF1 receptor (IGF1R)-AKT/mitogen-activated protein kinase (MAPK) signaling as a central regulatory axis that coordinates vascular remodeling and hepatocyte maturation within the engineered liver tissues. Furthermore, we applied exogenous IGF2 during construct maturation to enhance liver tissue organization and functionality and demonstrated that transplantation of these IGF2-primed liver organoids into a murine model of chronic liver failure significantly promoted in vivo hepatic regeneration.

RESULTS

Production of whole-liver scaffolds enriched with ECM and growth factors

An extracellular matrix serves as a mechanical and biochemical niche for cells during organogenesis. To this end, whole-liver scaffolds were produced by decellularization of rat livers. Characterization of the dECM liver scaffolds involved several histological analyses, including hematoxylin and eosin (H&E) staining, picrosirius red staining, and scanning electron microscope (SEM) analysis. These analyses indicated that all cells from the original organs were removed, while the ECM structures remained intact in the dECM liver scaffolds (Fig. 1, A and B). In particular, immunostaining for α-galactosidase (α-GAL), an epitope known to trigger immune rejection in humans (22), confirmed the elimination of antigenic molecules from resident animal cells (Fig. 1C). Furthermore, concentrations of ECM components, such as collagen, elastin, and glycosaminoglycans (GAGs), remained largely unchanged in the dECM scaffolds, indicating the preservation of the ECM (Fig. 1D). Given that ECM proteins have been shown to play a pivotal role in directing organogenesis from human iPSCs (23), the maintained composition of the dECM scaffolds underscores their potential to provide an instructive microenvironment for liver tissue development. Furthermore, liquid chromatography–tandem mass spectrometry (LC-MS/MS) proteomic analysis of the native livers and the dECM liver scaffolds was conducted to quantify the composition of proteins enriched in each liver. The results revealed that the matrisomes, referring to proteins associated with ECM, were significantly enriched in the dECM livers compared to native livers (Fig. 1E). Among matrisomes, both groups were mainly composed of core matrisomes including collagens, ECM glycoproteins, and proteoglycans rather than matrisome-associated proteins including ECM regulators, ECM-affiliated proteins, and secreted factors (Fig. 1F) (24). In particular, the ECM glycoproteins were the primary constituents of core matrisomes in the native livers, whereas collagens were predominant in the dECM livers. ECM regulators were the main component of matrisome-associated proteins in both groups. To quantitatively compare the proportional composition of ECM components between native liver and dECM liver scaffolds, chi-square tests were performed. A test of homogeneity for the core matrisome proteins revealed a highly significant difference between native and dECM livers (χ² = 109.91, P = 2.76 × 10⁻²⁴), indicating that the relative abundance of core ECM components was markedly altered following decellularization. For the matrisome-associated proteins, a chi-square test also indicated a significant difference in composition (χ² = 35.22, P = 7.53 × 10⁻⁸), suggesting selective enrichment of certain ECM categories during the decellularization process. When comparing the overall ratio of core matrisome versus matrisome-associated proteins, we found a significant shift in the proportion (χ² = 13.24, P = 0.00024), with dECM liver scaffolds exhibiting a relatively higher abundance of core matrisomal proteins. Furthermore, proteins related to growth factors, vascular development and angiogenesis, and liver development were enriched in the dECM liver scaffolds (Fig. 1, G to I). Given these results, the dECM liver scaffolds provide a biocompatible ECM niche for liver-composing stem cells and can be further used for the creation of vascularized liver organoids.

Fig. 1. Characterization of the dECM scaffold as a niche for the liver and vasculature.

(A) Representative images of the native livers and the dECM liver scaffolds after H&E staining (left) and picrosirius red staining (right). Scale bars, 100 μm. (B) SEM analysis of the native livers and the dECM liver scaffolds. Scale bars, 50 μm (left) and 10 μm (right). (C) Representative confocal images of the livers stained with α-GAL (red) and nuclei (blue, DAPI staining) and quantification of α-GAL MFI in each group (n = 4). Scale bars, 100 μm. (D) Quantification of collagen (left), elastin (middle), and GAGs (right) in the native livers and the dECM livers. Each content was normalized to the dry weight of the livers (n = 4). (E) Ratio of matrisomal and nonmatrisomal proteins in each liver, revealed by LC-MS/MS analysis. (F) Pie charts showing the relative proportions of core matrisome and matrisome-associated proteins in the native liver and dECM liver. The upper stacked bar graphs represent the subclass composition within core matrisome (ECM glycoproteins, collagens, and proteoglycans) and matrisome-associated proteins (ECM-affiliated proteins, ECM regulators, and secreted factors), while the lower bar graphs present the corresponding percentage values for each subclass. (G to I) Heatmap showing the normalized log₂ expression values of proteins associated with growth factors (G), vascular development and angiogenesis (H), and liver development (I) in the native livers and the dECM livers. Quantitative data are presented as the means ± SD. Statistical differences between the groups were determined by unpaired Student’s t test (*P < 0.05 and ***P < 0.001 versus native liver; ns, not statistically significant).

Differentiation of liver-composing cells from a single line of human iPSCs

We first aimed to create vascularized liver tissues with hepatocytes and ECs, both of which can be obtained through differentiation from a single line of iPSCs (fig. S1A). For large-scale production, we induced highly proliferative progenitor cells, hepatoblasts, followed by further differentiation into mature hepatocytes. During hepatocyte commitment, quantitative real-time polymerase chain reaction (qRT-PCR) data revealed distinct gene expression patterns at each developmental stage, with a decrease in pluripotency marker expression over time (fig. S1B). AFP⁺ CK19⁺ hepatoblasts were efficiently differentiated into ALB⁺ CYP3A4⁺ hepatocytes, as confirmed by flow cytometry analysis (fig. S1, C and D). The differentiated hepatocytes also expressed hepatic markers, such as cytokeratin 18 (CK18), zonula occludens–1 (ZO-1), and hepatocyte nuclear factor 4α (HNF4α), and displayed the typical binucleated morphology of mature hepatocytes (fig. S1E). Human iPSCs also differentiated into endothelial progenitor cells (EPCs) and mature ECs (fig. S2A). During endothelial commitment, the expression of specific markers indicative of the mesoderm, progenitor phase, and maturation phase was observed, while the expression of pluripotency markers within the cells gradually decreased (fig. S2B). After the maturation of CD34⁺ VEGFR2⁺ EPCs, ECs expressed angiogenic markers, including CD31, CD144, von Willebrand Factor (vWF), and Claudin-5 (CLDN5) (fig. S2, C to E). Together, we successfully established iPSC-derived liver progenitor cells and EPCs with robust potential for further differentiation and maturation.

Reconstruction of vascularized liver tissues with VCA-loaded dECM scaffolds

Given that fully differentiated hepatocytes and ECs have limited proliferation capacity in vitro (25), we bioengineered liver constructs by assembling iPSC-derived progenitor cells into the dECM liver scaffolds (Fig. 2A). Initially, induced hepatoblasts were engrafted into the dECM scaffolds and maintained for 2 weeks to allow for sufficient expansion. To enhance vascularization in the dECM liver scaffolds, an anti-CD31 aptamer, which is a short oligonucleotide that specifically binds to CD31 proteins, was used as a VCA. It has been previously reported to promote EC adhesion to the vascular lumen during recellularization (13). Following EPC repopulation, hepatoblasts and EPCs were subsequently differentiated into hepatocytes and ECs, respectively, to facilitate the formation of functional liver structures. Given that the two types of cells require different medium conditions for maturation, a transwell system was used to optimize a codifferentiation medium. (4,5-Dimethyl thiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assays demonstrated that a 1:1 mixture of hepatocyte media and EC media provided the optimal condition for sustaining the viability of both cell types simultaneously (fig. S3, A and B). Furthermore, the expression of lineage-specific markers, as confirmed by immunostaining, was preserved under this condition, indicating that the iPSC-derived liver tissues allowed to undergo self-organization in this optimized coculture medium (fig. S3, C to F). Histological analysis revealed well-defined liver organoid structures, including liver parenchyma and vasculature (Fig. 2B). In particular, the VCA-loaded liver tissues displayed increased alignment of endothelium expressing CD31 and CD144 compared to the phosphate-buffered saline (PBS) control group (Fig. 2C). We also investigated the presence of vascular smooth muscle cells given the bipotency of EPCs. Notably, vascular smooth muscle cell–like cells were observed sporadically in the VCA group and appeared to colocalize with ECs, suggesting the partial formation of vascular wall structures within the organoids (Fig. 2D). To evaluate the 3D luminal structure as well as barrier functions, each organoid was perfused with fluorescein isothiocyanate-dextran and Bandeiraea simplicifolia lectin I (BSL-I). The VCA-loaded group displayed a 3D luminal structure with a continuous and tight endothelial barrier, showing no detectable dextran leakage, in contrast to the control group, which exhibited extensive dextran leakage (fig. S4A and movies S1 and S2). Along with reendothelialized vasculatures, a considerable albumin (ALB) expression was observed in the VCA-loaded liver tissues (Fig. 2E). In contrast, the liver without VCA featured incomplete structural organization and maturation of both parenchyma and blood vessels. To investigate whether the vascularized liver organoids recapitulate hepatic zonation, we performed immunostaining of canonical zonation markers that were selected on the basis of recent high-throughput liver transcriptomic datasets (26–28). The results revealed a predominant expression of periportal (zone 1) markers such as E-cadherin and TIMP1 throughout the organoids, whereas pericentral (zone 3) markers CYP2E1 and ITGA5 were either scarcely detected (fig. S4, B and C). These findings suggested that the hepatocyte-like cells in our model largely exhibit a zone 1–like identity, with limited zonation-related spatial organization. Moreover, to explore the extent of cellular diversity in the organoids, we assessed the presence of cholangiocytes (CK19), liver sinusoidal ECs (LSECs; F8), and hepatic stellate cells (Cytoglobin). CK19⁺ cells were detected in both groups; however, their distribution did not correspond to duct-like structures and showed no apparent differences between PBS and VCA groups (fig. S4D). Cytoglobin⁺ cells and F8⁺ cells were sporadically observed, suggesting the potential emergence of stellate-like cells and LSEC-like cells, respectively; however, no significant differences were noted between the groups (fig. S4E). Collectively, our liver organoids primarily consist of periportal hepatocytes and vascular ECs with a limited presence of nonparenchymal cells. In terms of functionalities, the VCA group also showed enhanced capacity of glycogen synthesis, indicated by increased magenta granules, and significantly elevated secretion of ALB and urea (Fig. 2, F to H). When assessing vascular-specific functionality, the VCA group exhibited an increased secretion of vascular-related factors, including vascular endothelial growth factor (VEGF) and nitric oxide, compared to the PBS control group (Fig. 2, I and J). Notably, terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) assay results showed markedly reduced apoptotic cell death in the VCA-loaded livers, which may be attributed to enhanced cell-cell interactions within the bioengineered livers (Fig. 2K) (13). Together, the introduction of VCA to dECM liver scaffolds contributed to the fabrication of advanced liver tissue constructs, with improved structural similarity and cellular functions.

Fig. 2. Generation of human iPSC–derived vascularized liver organoids using dECM scaffolds.

(A) Schematic illustration of reconstructing vascularized liver organoids by using iPSC-derived hepatoblasts and EPCs. (B) Representative images of liver organoids with dECM liver scaffolds loaded with PBS control (PBS) or anti-CD31 aptamer (VCA) after H&E staining (top) and picrosirius red staining (bottom). The PBS group indicated a nonendothelialized vascular wall, characterized by the absence of cells (green) inside the vascular collagen layer (red) (red arrow). The VCA showed a reendothelialized layer with ECs attached on the luminal side group (purple arrows). (C) Representative confocal images of the liver organoids stained with CD31 (green), CD144 (red), and nuclei (blue, DAPI staining) and quantification of reendothelialized area in each group (n = 4). (D) Representative confocal images of the liver organoids stained with CD31 (green), PDGFRβ (red), and nuclei (blue, DAPI staining) and quantification of PDGFRβ MFI in each group (n = 4). (E) Representative confocal images of the liver organoids stained with CD31 (green), ALB (red), and nuclei (blue, DAPI staining) and quantification of ALB expression in each group (n = 4). (F) PAS-stained images showing the glycogen synthesis functions in the liver organoids in each group. (G to J) Secretion levels of ALB (G), urea (H), VEGF (I), and nitric oxide (J) from the liver organoids in each group (n = 4). (K) Representative images of TUNEL staining in liver organoids in each group, showing apoptotic cells (red) and nuclei (blue, DAPI staining) and its quantification (n = 4). All scale bars, 100 μm. Quantitative data are presented as the means ± SD. Statistical differences between the groups were determined by unpaired Student’s t test (*P < 0.05, **P < 0.01, and ***P < 0.001 versus PBS).

Spatial isolation of regionally specific cells from iPSC-derived vascularized liver organoids

To elucidate how VCA enhances vascularization within liver tissues, spatial transcriptomic analysis was conducted using PHLI technology, which enabled the isolation of ∼10 cells from each ROI (Fig. 3A). More than 2300 genes were successfully identified from vascular ROIs, and more than 1600 genes were identified from parenchymal ROIs in both groups (Fig. 3B). From bright-field images of the whole tissues, we designated ROIs distinguishing the liver parenchyma or vasculatures (Fig. 3C). To ensure that each ROI encompassed structurally organized regions, immunostaining was performed, revealing ALB expression in the parenchymal areas and CD31 expression in the vascular regions despite differences in fluorescence intensity between the PBS and VCA groups (Fig. 3D). Cells from each targeted ROI were then individually collected, followed by RNA extraction and full-length RNA sequencing. Heatmap analysis of differentially expressed genes (DEGs) revealed clear differences in hierarchical clustering between the PBS and VCA groups (Fig. 3E). In the PBS group, the gene expression profiles of parenchymal hepatocytes and ECs did not cluster separately but remained intermixed, suggesting that these cells failed to engraft in their original locations because of insufficient vascular development within the organoids. By contrast, in the VCA group, these populations isolated from each compartment formed distinct clusters, indicating improved structural compartmentalization and regional identity. Together, these results underscore the role of vascularization in guiding tissue architecture and validate the robustness of our methodology for isolating region-specific cells and transcriptomic profiling in iPSC-derived liver organoids.

Fig. 3. PHLI-seq for spatial analysis of parenchymal and nonparenchymal regions in iPSC-derived liver organoids.

(A) Schematic of the PHLI strategy applied to iPSC-derived liver organoids treated with PBS or VCA. The organoids were sectioned and processed for spatial phenotype-guided separation of parenchymal and vascular zones, followed by transcriptomic analysis. (B) Each bar represents the total number of genes identified from individual ROIs isolated from vascular or parenchymal regions of iPSC-derived liver organoids in group (vascular, n = 3; parenchymal, n = 2). (C) Bright-field images of each liver organoid. Blue-painted boxes indicate ROIs for spatial transcriptomic analysis. Areas designated with colored dashed lines are subjected to validation of endothelialization. (D) Confocal images of the ROIs designated in (B) stained with CD31 (green), ALB (red), and nuclei (blue, DAPI staining). For staining, serially sectioned samples in each group were used. Colored bars placed above each image match the dashed color box in (B), enabling spatial correlation between transcriptomic sampling sites and their histological structure. Note that distinct separation between CD31⁺ endothelial zones and ALB⁺ hepatic parenchyma is evident in the VCA group. Scale bars, 100 μm. (E) Heatmaps showing transcriptomic profiles of spatially separated parenchymal (left) and vascular (right) regions in the PBS and VCA groups. Hierarchical clustering demonstrates clear segregation of gene expression patterns between the two anatomical compartments in the VCA-loaded vascularized liver organoids, while the PBS group shows no such separation resulting from insufficient vascularization. Quantitative data are presented as the means ± SD. Statistical differences between the groups were determined by unpaired Student’s t test (ns, not statistically significant).

Spatial transcriptomic analysis of vascular regions in the liver organoids

To elucidate how VCA-loaded dECM scaffolds contributed to remodeling of the vascular microenvironment of liver organoids, we then conducted the comparative transcriptomic analysis of each vascular or parenchymal region isolated from the PBS control group and the VCA group. Following spatial RNA sequencing, the distribution of aligned reads across genomic regions in the vascular transcriptomes was examined, showing high alignment quality (Fig. 4A). The gene body coverage profile demonstrated moderate 3′ bias and indicated preserved RNA integrity across the transcriptomes (Fig. 4B). DEG analysis in vascular regions revealed distinct transcriptional signatures between the PBS and VCA groups, with hierarchical clustering showing that vascular ECs from each group clustered tightly together, reflecting high intragroup transcriptomic concordance (Fig. 4C). Gene ontology (GO) enrichment analysis of vascular DEGs clearly indicated that terms related to angiogenesis, such as fluid shear stress, EC migration, and endothelial tube morphogenesis, as well as cellular junctions, such as adherens junction, tight junction, and focal adhesion, were top ranked (Fig. 4D). We also found the enrichment of protein phosphorylation, MAPK cascade, and IGF receptor binding, suggesting an involvement of several signaling pathways during vascularization in the liver tissues (Fig. 4D). This is consistent with previous reports suggesting the pivotal role of the IGF system in vascular remodeling (29, 30). In particular, IGF1 and IGF2 have been shown to activate intracellular signaling cascades, such as phosphatidylinositol 3-kinase (PI3K)/AKT and MAPK/ERK pathways in ECs, which are well-established mediators of cell proliferation, migration, and survival during angiogenesis (31). In this context, we also noted that gene set enrichment analysis (GSEA) revealed coordinated enrichment of not only kinase binding and growth factor binding but also signaling pathways including PI3K, MAPK, and insulin signaling in the VCA group (Fig. 4E and fig. S5A). The genes implicated in these pathways, such as IGF2, MAP7D2, RHOB, BMP4, SRC, and SDC4, were significantly up-regulated in the VCA group compared to the PBS control group (Fig. 4E and fig. S5B). Furthermore, the down-regulation of apoptosis-inducing genes (CALR and VMP1) and the up-regulation of antiapoptotic genes (JUN and BRD4) were observed in the VCA group (Fig. 4F), which was consistent with Fig. 2K. Given that the IRS (insulin receptor substrate)-PI3K-AKT axis plays a critical role in promoting EC survival primarily by inhibiting apoptosis (32), these results point to a potential role for IGF-driven downstream signaling in promoting EC viability. Together, coating the vascular lumen of dECM scaffolds with anti-CD31 aptamers improved angiogenesis and cell survival by modulating IGF-mediated signaling and its downstream pathways in the vascular cells, thereby supporting vascular remodeling within the engineered liver tissues.

Fig. 4. Spatial transcriptomic analysis of vascular compartment in liver organoids.

(A) Pie chart illustrating the distribution of aligned reads across genomic features including exonic, intronic, and intergenic regions in vascular cells. (B) Gene body coverage plot showing the 5′ to 3′ read distribution of the full-length transcriptomes in vascular cells. (C) Heatmap showing DEGs in vascular regions from the PBS- and VCA-loaded liver organoids. Hierarchical clustering reveals tight clustering within each group, indicating intragroup transcriptomic similarity (n = 3). (D) GO enrichment analysis of genes up-regulated in vascular regions of the VCA-loaded liver organoids. (E) GSEA plots showing positive enrichment of gene sets in the VCA group, including the insulin signaling pathway, MAPK signaling pathway, and PI3K signal transduction. The apoptotic signaling pathway was negatively enriched in the VCA group. (F) Violin plots showing the expression levels (log₂[DESeq2-normalized count + 1]) of representative genes involved in IGF, MAPK, PI3K/AKT, and apoptotic signaling pathways in PBS- and VCA-loaded liver organoids (n = 3). In each plot, the thick solid line indicates the median value, and the thin dashed lines represent the 25th and 75th percentiles. Statistical significance was assessed using the Wald test as implemented in the DESeq2 package, and P values for each gene comparison are indicated above the plots.

IGF2 drives parenchymal phenotype in spatially organized VCA-loaded liver organoids

It has been highlighted that the hepatic endothelium contributed to establishing hepatic functions and regenerative potential, underscoring the essential role of endothelial-parenchymal cross-talk in liver homeostasis and repair (33, 34). Thus, we also performed transcriptomic profiling of isolated parenchymal cells to investigate the effects of vascularization on liver organoids. Reads were predominantly mapped to exonic regions with uniform coverage along the gene body, indicating high alignment quality and minimal 3′ bias (fig. S6, A and B). Hierarchical clustering based on parenchymal DEGs revealed a clear separation between the two groups, with tight intragroup clustering, suggesting that the transcriptional differences in the parenchyma were likely driven by the enhanced reendothelialization achieved through VCA application (fig. S6C). GO enrichment analysis of these parenchymal DEGs showed significant enrichment of MAPK-related pathways, indicating that kinase-mediated signaling was stimulated not only within vascular compartments but also across the parenchymal compartment in the VCA group (fig. S6D). To verify the shared transcriptomic alterations between the vessels and parenchyma in the PBS- and VCA-loaded liver organoids, we identified the DEGs that were commonly altered (Fig. 5A). Among the overlapping DEGs, IGF2 was concordantly and significantly altered in both compartments, underscoring its central role as a shared regulator in the structural maturation of the VCA-loaded liver organoids (Fig. 5B). GO analysis of the shared DEGs revealed strong associations with terms including “IGF receptor binding” and “protein kinase binding” (Fig. 5C). Representative genes from IGF signaling including IGF2 and GNAS were consistently up-regulated, whereas INSR, a receptor known to counteract IGF signaling, was down-regulated in the whole VCA-loaded liver organoids (Fig. 5D). In addition, genes involved in protein kinase binding were also significantly altered in the whole organoids (Fig. 5E). These combined results suggested that VCA-mediated spatial modulation extended beyond the vasculature and influenced parenchymal transcriptomic states, with IGF2 coordinating kinase activation in an angiocrine and paracrine manner. This interpretation is further supported by previous studies demonstrating that IGF2, predominantly expressed in the liver, plays a pivotal role in hepatic development, functional maturation, and tissue homeostasis (35–37). In VCA-loaded liver organoids, IGF2 expression was greatly up-regulated in both parenchymal and vascular regions (Fig. 5, F and G). We also observed increased phosphorylation of IGF1R, AKT, and p44/42 MAPK pathways (Fig. 5, H and I). Collectively, these results provided compelling evidence that IGF and its downstream pathways play a crucial role in orchestrating spatial organization of iPSC-derived liver organoids.

Fig. 5. Enhanced IGF signaling in both vessels and parenchymal regions of liver organoids.

(A) Venn diagram illustrating the overlap of DEGs identified between PBS- and VCA-treated groups within the vascular and parenchymal compartments. (B) Scatterplot comparing statistical significance of overlapping DEGs between vascular (x axis) and parenchymal (y axis) compartments. Note that IGF2 showed high significance in both compartments. (C) GO enrichment analysis of overlapping DEGs in the liver organoids. (D and E) Violin plots showing the expression of significant shared DEGs associated with IGF receptor binding (D) and protein kinase binding (E), showing up-regulation in the VCA-loaded liver organoids. In each plot, the thick solid line indicates the median value, and the thin dashed lines represent the 25th and 75th percentiles. (F) Representative confocal images of liver organoids stained with IGF2 (red) and nuclei (blue, DAPI staining). Scale bars, 200 μm. (G) Quantification of IGF2 expression in each group (n = 5). (H) Western blotting analysis of phosphorylated and total forms of IGF1R, p44/42 MAPK, and AKT in the PBS- or VCA-loaded liver organoids. (I) Quantification of the expression of each protein relative to β-actin as a loading control (n = 4). Quantitative data are presented as the means ± SD. Statistical differences between the groups were determined by unpaired Student’s t test (**P < 0.01 and ***P < 0.001 versus PBS).

IGF signaling is essential for maintaining viability and functionality in iPSC-ECs

On the basis of spatial transcriptomic analysis, we investigated the impact of IGF signaling, a key upstream regulator, on the endothelial lineage specification from iPSCs. Previous studies have identified IGF2 and its receptor IGF1R as essential mediators of endothelial survival, sprouting, migration, and tip cell maintenance in primary ECs, with knockdown studies demonstrating impaired angiogenesis both in vitro and in vivo (38–40). Hence, after IGF2 was silenced in ECs differentiated from iPSCs using small interfering RNA (siRNA), a notable suppression of IGF1R activation was observed (Fig. 6A). We next examined the activation status of downstream signaling pathways, including AKT and p44/42 MAPK signaling. Western blot analysis revealed significant inactivation of both pathways in siIGF2-treated ECs (Fig. 6B). These alterations led to elevated apoptosis, as shown by increased cleaved caspase-3 expression and annexin V staining, as well as reduced cell viability as measured by MTT assay, in siIGF2-treated ECs (fig. S7, A and B). Then, we attempted to determine whether IGF-mediated signaling influences endothelial identity as well as the efficiency of endothelial commitment. As a result, IGF2 knockdown significantly decreased mRNA levels of endothelial-specific genes and the CD31⁺CD144⁺ populations compared to control ECs (Fig. 6C and fig. S7C). Immunofluorescence analysis also demonstrated that IGF2-supressed ECs lost their distinctive characteristics, such as UEA-1 and vWF expression (Fig. 6D). Functional features were evaluated by acetylated low-density lipoprotein (Ac-LDL) uptake and tube formation capacity in Matrigel. Compared to control ECs, siIGF2-treated ECs exhibited impaired Ac-LDL uptake and formed fewer capillary-like branching networks (Fig. 6, E and F). Consistent with these results, IGF2 knockdown resulted in a marked reduction in phosphorylation levels of endothelial nitric oxide synthase, indicative of defective endothelial maturation (fig. S7D). In addition, secretion of nitric oxide and VEGF, two hallmarks of angiogenic and vasodilatory capacity, was significantly diminished in IGF2-silenced ECs (Fig. 6, G and H). These results collectively demonstrate that IGF2 signaling is critical for orchestrating endothelial differentiation and functional maturation, primarily through the activation of IGF1R-AKT/MAPK pathways, thereby promoting vascular phenotype establishment at both molecular and phenotypic levels.

Fig. 6. IGF2 knockdown reduces differentiation and angiogenic potential in iPSC-ECs.

(A) Western blog analysis of IGF2, total IGF1R, and p-IGF1R in iPSC-ECs transfected with siControl or siIGF2. Bar graphs indicate quantification of IGF2 normalized to β-actin and p-IGF1R normalized to total IGF1R (n = 3). (B) Western blot analysis of total p44/42 MAPK, p-p44/42 MAPK, total AKT, p-AKT, and cleaved caspase-3 in iPSC-ECs transfected with siControl or siIGF2. Bar graphs showing quantification of p-p44/42 MAPK normalized to total p44/42 MAPK, p-AKT normalized to total AKT, and cleaved caspase-3 normalized to β-actin (n = 3). (C) Flow cytometry analysis showing the proportion of CD31⁺VE-cadherin⁺ cells in each group (n = 5). (D) Immunofluorescence staining for endothelial markers, UEA-1 (green) and vWF (red), and quantification of UEA-1⁺vWF⁺ cells in iPSC-ECs transfected with siControl or siIGF2 (n = 5). (E) Representative confocal images and quantification of Ac-LDL (DiI-AcLDL, red) uptake in each group (n = 5). (F) Tube formation assay using iPSC-ECs cultured on Matrigel with or without siIGF2. Representative images show a decrease of capillary-like structures upon siIGF2 treatment, and the branching points were quantified (n = 7). (G and H) Secretion of nitric oxide (G) and VEGF (H) from iPSC-ECs in each group (n = 4). All scale bars, 100 μm. Quantitative data are presented as the means ± SD. Statistical differences between the groups were determined by unpaired Student’s t test (**P < 0.01 and ***P < 0.001 versus siControl).

IGF signaling is essential for maintaining viability and functionality in iPSC-hepatocytes

To elucidate the role of IGF2-IGF1R signaling in hepatocyte differentiation and maturation, we silenced IGF2 gene in iPSC-derived hepatoblasts, which were subsequently matured into hepatocytes. Following IGF2 knockdown, the phosphorylation of IGF1R, AKT, and MAPK was markedly reduced in differentiated hepatocytes, indicating the global attenuation of IGF-driven intracellular signaling cascades (Fig. 7, A and B). In addition to the lower efficiency of hepatocyte differentiation resulting from IGF inactivation (41), this was also associated with a rise in apoptotic activity, as evidenced by elevated cleaved caspase-3 expression and a higher proportion of annexin V⁺ cells detected by flow cytometry (Fig. 7C and fig. S8A). In line with this, a significant decline in cell viability in the IGF2-knockdown group was observed by MTT assay, suggesting a supporting role of IGF signaling in hepatocyte survival (fig. S8B). Expression of key hepatic transcription factors and functional proteins was substantially down-regulated at both transcript and protein levels (Fig. 7D and fig. S8, C and D). Immunofluorescence analysis demonstrated that IGF2 silencing disrupted typical binucleated morphology with diminished expression of hepatocyte-specific markers (ALB, CK18, and HNF4α) as well as the tight junction protein ZO-1 compared to control iPSC-hepatocytes (Fig. 7E). Further assessments revealed impaired hepatic performance with significant reductions in glycogen synthesis, urea, and albumin production in the IGF2-silenced group, confirming functional immaturity (Fig. 7, F to H). These combined results underscored the importance of IGF signaling in regulating hepatocyte survival and maturation, primarily through the IGF1R-AKT/MAPK axis to ensure proper structural integrity and metabolic functionality in iPSC-derived hepatocytes.

Fig. 7. Silencing IGF2 signaling disrupts the characteristics of iPSC-hepatocytes.

(A) Western blog analysis and quantification of IGF2 normalized to GAPDH and p-IGF1R normalized to total IGF1R in iPSC-hepatocytes transfected with siControl or siIGF2 (n = 3). (B) Western blot analysis and quantification of p-p44/42 MAPK normalized to total p44/42 MAPK and p-AKT normalized to total AKT, normalized to GAPDH in iPSC-hepatocytes transfected with siControl or siIGF2 (n = 3). (C) Western blot analysis and quantification of cleaved caspase-3 expression normalized to β-actin in iPSC-hepatocytes transfected with siControl or siIGF2 (n = 3). (D) Flow cytometry analysis and quantification of the proportion of ALB⁺CYP3A4⁺ cells in each group (n = 4). (E) Immunofluorescence staining for hepatic markers in iPSC-hepatocytes transfected with siControl or siIGF2. Top: ALB (green) and CK18 (red); bottom: ZO-1 (green) and HNF4α (red). Nuclei were stained with DAPI. Double-positive cells were quantified (n = 5). Scale bars, 100 μm. (F) Representative PAS-stained images of iPSC-hepatocytes transfected with siControl or siIGF2. Scale bars, 40 μm. (G and H) Secretion of albumin (G) and urea (H) from iPSC-hepatocytes in each group (n = 4). Quantitative data are presented as the means ± SD. Statistical differences between the groups were determined by unpaired Student’s t test (*P < 0.05, **P < 0.01, and ***P < 0.001 versus siControl).

Exogenous IGF2 synergistically facilitates structural and functional maturation of vascularized liver organoids

To evaluate whether exogenous IGF2 enhances spatial organization in the bioengineered liver tissues, human recombinant IGF2 (rIGF2) was supplemented during the maturation phase. Upon IGF2 treatment, we first examined the activation of IGF-downstream signaling pathways including IGF, AKT, and p44/42 MAPK in each liver organoid. Extending previous findings that VCA augmented the activation of these pathways, we confirmed that fully vascularized liver organoids with VCA exhibited the highest levels of IGF1R, AKT, and MAPK phosphorylation upon IGF treatment (Fig. 8, A and B). The activation of IGF2-IGF1R-AKT/MAPK signaling within the organoids enhanced their regenerative performance, aligning with previous findings that implicated IGF2 expressed by hepatocytes in driving hepatic proliferation and repair mechanisms (42). In particular, AKT activation also contributed to a reduction in apoptosis in the VCA + rIGF2 group, as demonstrated by a marked decrease in cleaved caspase-3 expression (Fig. 8, C and D). To determine whether these molecular changes translated into improved tissue organization, we assessed hepatic and vascular compartmentalization upon exogenous IGF2 treatment across groups. Previously, we have reported that the AKT pathway is involved in efficient vascular reconstruction as well as survival of ECs within anti-CD31 aptamer–coated dECM scaffolds (13). In accordance with these results, in organoids produced with the PBS-loaded dECM scaffolds, the expression of ALB as well as reendothelialized rate was elevated by IGF-mediated signaling activation (Fig. 8, E and F). IGF2 supplement further increased reendothelialization in the VCA-loaded scaffolds but did not significantly alter the maturity of liver parenchyma. This suggested that the paracrine effects derived from ECs lining the vascular lumen in the VCA-loaded tissues may already elicit a maximal effect on liver maturation (13). Furthermore, VCA + rIGF2 showed significantly higher mRNA expression of not only hepatic markers including ALB, TTR, CYP3A4, and HNF4α but also endothelial markers, such as VEGFR2, CLDN5, and VEGFA, relative to other groups (Fig. 8, G and H). The PBS + rIGF2 group also showed up-regulation of these genes relative to PBS control but to a lesser extent than the VCA + rIGF2 group, indicating that both vascular and IGF-mediated cues contribute to maturation of liver organoids, with synergistic enhancement when combined. In accordance with these results, measurements of secreted albumin and urea demonstrated significantly higher production in the VCA + rIGF2 group, reflecting robust liver-specific functions (Fig. 8, I and J). Moreover, production of VEGF and nitric oxide, key indicators of endothelial function, was markedly elevated in the VCA + rIGF2 organoids, underscoring the synergistic enhancement of vascular maturity by IGF activation and VCA (Fig. 8, K and L). Collectively, these results demonstrated that IGF-downstream pathways synergized with VCA-mediated vascularization to facilitate spatial organization within the bioengineered tissues, primarily through angiocrine and paracrine mechanisms, thereby promoting both structural and functional maturation.

Fig. 8. Exogenous IGF2 promotes spatial reconstruction of bioengineered liver tissues.

(A and B) Western blot analysis (A) of phosphorylated and total forms of IGF1R, p-44/42 MAPK, and AKT in each liver tissue and quantification (B) of the expression of each protein relative to β-actin as a loading control (n = 3). (C and D) Western blot analysis of cleaved caspase-3 (C) and quantification of expression levels normalized to β-actin (D) in each liver tissue (n = 3). (E) Representative immunofluorescence staining of iPSC-derived vascularized liver organoids with CD31 (green), ALB (red), and nuclei (blue, DAPI staining). PBS (with PBS-loaded dECM scaffold and vehicle), VCA (with VCA-loaded dECM scaffold and vehicle), PBS + IGF2 (with PBS-loaded dECM scaffold and rIGF2), and VCA + IGF2 (with VCA-loaded dECM scaffold and rIGF2). Scale bars, 100 μm. (F) Quantification of reendothelialized area (left) and ALB expression (right) in each group (n = 6). (G) qRT-PCR analysis of liver-specific genes, including TTR, CYP1A1, PROX1, and HNF1A, in each group (n = 3). (H) qRT-PCR analysis of vascular genes, including VEGFR2, CLDN5, and VEGFA, in each group (n = 3). (I to L) Quantification of secreted levels of ALB (I), urea (J), VEGF (K), and nitric oxide (L) from the liver organoids in each group. Quantitative data are presented as the means ± SD. Statistical differences between the groups were determined by a one-way ANOVA with Tukey’s post hoc test (*P < 0.05, **P < 0.01, and ***P < 0.001; ns, not statistically significant versus PBS, ##P < 0.01 and ###P < 0.001).

IGF2-preconditioned liver organoids promote in vivo liver regeneration

To assess the in vivo therapeutic potential of the bioengineered liver organoids in each group, we established a murine model of chronic liver failure induced by thioacetamide (TAA) administration (Fig. 9A). The liver organoids produced with PBS- or VCA-loaded dECM scaffolds with or without rIGF2 treatment were transplanted between hepatic lobules and maintained for 4 weeks under TAA exposure. To investigate the regenerative behavior of transplanted human hepatocytes, we performed coimmunostaining of human nuclei (HuNu) with annexin A2 (ANXA2), a marker associated with hepatocyte motility and regenerative migration (43). Notably, a greater number of HuNu⁺ ANXA2⁺ human hepatocytes were detected in the recipient livers of the VCA + rIGF2 group, particularly localized around perivascular regions (Fig. 9, B and C). These findings indicated that vascularization may enhance not only the survival and engraftment of the transplanted cells but also their acquisition of a motile, regeneration-associated phenotype in vivo. To further investigate the aspects of tissue repair upon organoid transplantation, we conducted immunofluorescence staining for the liver stem cell marker leucine-rich G repeat–containing protein–coupled receptor 5 (LGR5) in recipient mouse livers. Quantitative analysis showed that LGR5⁺ cells were predominantly localized in perivascular regions in the VCA + rIGF2 group, rather than distributed in the liver parenchymal regions during regeneration (Fig. 9, D and E). In accordance with the previous studies reporting that liver regeneration is particularly active near the portal triads and is dependent on signals from ECs (44, 45), this spatial enrichment of LGR5⁺ cells reflects the critical role of the vascular niche in activating stem cell populations in the injured liver. This regenerative response was notably augmented by transplanted VCA + rIGF2 liver organoids containing fully endothelialized vasculature. Picrosirius red staining of mouse liver sections showed an extensive collagen deposition in the PBS + veh group, while transplantation of VCA + rIGF2 organoids significantly mitigated fibrotic burden compared to all other groups, indicating superior regenerative potential (Fig. 9, F and G). Consistently, fibrotic genes were markedly suppressed in the VCA + rIGF2 group, further highlighting the synergistic role of vascularization and IGF2 signaling in liver organoids for alleviating fibrotic progression (Fig. 9H). Moreover, mice receiving VCA + rIGF2 organoids displayed markedly reduced serum levels of alanine aminotransferase (ALT) and aspartate transaminase (AST), key biochemical markers of liver injury, which returned to values within the normal physiological range (Fig. 9, I and J). This indicated the highest regenerative performance of VCA + rIGF2 organoids. Although IGF2 priming also confers partial regenerative outcomes, its combination with VCA-loaded vascularized organoids is essential for maximizing the therapeutic efficacy. This underscores the critical importance of efficient vascularization through a synergistic combination of VCA application and IGF-mediated proregenerative signaling activation, which yields the most structurally and functionally advanced bioengineered liver organoids. Ultimately, liver organoids creating a robust vascular and proregenerative niche can potentiate liver tissue repair following transplantation by promoting cell engraftment and stem cell activation.

Fig. 9. IGF-treated vascularized liver organoids effectively ameliorate liver fibrosis in the chronic liver failure mouse model.

(A) Schematic illustration of the experimental timeline. TAA was administered in drinking water for 8 weeks to induce chronic liver failure in mice, followed by orthotopic transplantation of engineered liver organoids. The groups included (i) PBS + vehicle, (ii) VCA + vehicle, (iii) PBS + rIGF2, and (iv) VCA + rIGF2. Mice were then maintained on TAA for an additional 4 weeks before harvest. (B and C) Representative confocal images (B) of harvested mouse livers probed with HuNu (green) and ANXA2 (red) and quantification of engrafted human ANXA2⁺ cells (C) in each group. Scale bars, 100 μm (n = 3). (D and E) Representative confocal images (D) of harvested mouse livers probed with a liver stem cell marker, LGR5 (green), and quantification of engrafted human cells (E) in each group. Scale bars, 100 μm (n = 3). Note the distinct localization patterns of liver stem cells in the perivascular versus parenchymal regions during liver regeneration. (F and G) Representative images (F) of harvested livers stained with picrosirius red and quantification of fibrotic lesions (G) in each group. Scale bars, 40 μm. (H) qRT-PCR analysis of liver fibrosis–related genes, including Timp2, αSMA, Vimentin, and TGFb1, in the mouse liver harvested from each group (n = 3). (I and J) Serum levels of ALT (I) and AST (J), indicators of liver injury, were quantified in mice from each group (n = 3). Each red dashed line represents the maximum value of the normal range for healthy mice. Quantitative data are presented as the means ± SD. Statistical differences between the groups were determined by a one-way ANOVA with Tukey’s post hoc test (*P < 0.05, **P < 0.01, and ***P < 0.001; ns, not statistically significant versus Sham, #P < 0.05, ##P < 0.01, and ###P < 0.001 versus PBS + veh).

DISCUSSION

The dECM scaffolds derived from animals have gained notable attention in the field of organ engineering because of their native 3D architecture, vasculature-like conduits, and retention of essential ECM components (46, 47). Among the various forms of dECM-based scaffolds, solubilized dECM hydrogels have often been used to fabricate microscale organoid platforms, particularly for high-throughput drug screening applications (48). However, for transplantation as a regenerative therapy, the use of whole-organ dECM scaffolds in liver organoid production is essential not only to support therapeutically relevant cell loads but also to enable the reconstitution of functional hepatic microanatomy. One of the key challenges in whole-organ liver engineering lies in the vascular lumen, which in its decellularized scaffold lacks an endothelial lining. If left unmodified, a denuded luminal surface can provoke thrombus formation and hyperacute rejection upon perfusion or transplantation of engineered organs. Therefore, efficient reendothelialization of the vascular compartment is considered critical for enabling successful engraftment and long-term functionality. To address this, various approaches have been developed, including VEGF supplement (49) and heparin-gelatin coating (10). In addition, affinity-based strategies, such as coating with peptides or aptamers, have emerged to enhance the targeted recruitment and alignment of ECs within the vascular channels (10–12). Here, we developed a robust platform for reconstructing vascularized human liver tissues from a single line of iPSCs. An anti-CD31 aptamer facilitated EC adhesion and vessel formation within dECM liver scaffolds, thereby contributing to exquisite structural organization of liver organoids that recapitulate both parenchyma and vascular networks.

While such structural advancements mark a major step in tissue reconstruction, a deeper understanding of the underlying molecular dynamics remains essential, particularly given that complex engineered tissues often exhibit regional heterogeneity in gene expression that is masked by bulk RNA sequencing. In this context, spatial transcriptomics have emerged as a powerful tool to address this limitation. Unlike spatial transcriptomic platforms, which rely on predefined barcoded spots or typically capture only partial transcriptomes with limited spatial resolution, PHLI-seq enables phenotype-informed, laser-guided isolation of cells with near-single-cell precision while preserving RNA integrity and capturing full-length transcripts (20). This capability distinguishes PHLI-seq from other spatial transcriptomic analyses and allows a finer resolution of transcriptomic profiling across organoid compartments. This, in turn, allows for more accurate assessments of tissue maturation, vascular integration, and the overall fidelity of engineered tissue constructs. However, PHLI-seq poses certain technical limitations of RNA input and potential biases because of its mode of ROI-based isolation. In particular, because PHLI-seq relies on laser-mediated microdissection of small, histologically defined regions, the amount of RNA obtained from each ROI is inherently limited, which may increase variability across replicates and restrict the detection of low-abundance transcripts. Furthermore, the selection of ROIs based on histological features introduces a degree of user-defined bias and may exclude transcriptomic signals from surrounding microenvironments.

PHLI-seq revealed that spatially distinct IGF2-IGF1R-AKT/MAPK signaling networks act as a central regulatory axis coordinating the maturation of both parenchymal and vascular compartments in bioengineered liver tissues. Through a series of loss-of-function experiments, we also confirmed that IGF2 is indispensable for the survival, lineage commitment, and maturation of both hepatocytes and ECs derived from human iPSCs. Notably, IGF signaling is highly expressed during liver development and regulates the differentiation of hepatoblasts through activation of PI3K-AKT and MAPK signaling cascades (41, 50). IGF2 also exerts profound effects on endothelial biology, enhancing angiogenesis, migration, and tube formation by activating IGF1R-mediated MAPK and AKT phosphorylation in ECs (51). This role is not confined to the lineage commitment of each type of cell. In the fetal liver, IGF2 also serves as a critical niche-derived growth factor that supports hematopoietic stem cell expansion within the hepatic microenvironment (36). This dual role of IGF2 in promoting both hepatic and endothelial-associated stem cell populations supports our hypothesis that IGF2 serves as an angiocrine and paracrine coordinator of tissue-level organization. Beyond development, IGF2 has been implicated in liver regeneration, contributing to hepatic proliferation and tissue repair following injury via AKT and MAPK activation (42). Together, these findings position IGF2 as a multifunctional morphogen with the potential to synchronize vascular and hepatic maturation in bioengineered liver organoids. In our study, the combination of anti-CD31 aptamer–guided vascularization and IGF2 stimulation synergistically enhanced spatial organization of liver structures including reendothelialization, tissue viability, hepatic function, and endothelial integrity through MAPK and AKT pathways. Although IGF2 knockdown effectively revealed its function in 2D iPSC-derived hepatocytes and ECs, extending this approach to 3D liver organoids posed major technical challenges. IGF2 suppression caused severe loss of viability, proliferation, and maturation, preventing cells from withstanding the dissociation and repopulation process, and moreover, the inherently low proliferative capacity of iPSC-derived hepatocytes further limited their recovery after knockdown, resulting in insufficient cell density for organoid development (52). Alternative delivery methods such as direct in-scaffold transfection were also hindered by the large size and complex geometry of the dECM liver scaffolds. Despite these limitations, the gain-of-function experiments using exogenous IGF2 provided valuable insights into its role in promoting spatial organization and functional integrity within liver organoids. These improvements translated into the in vivo efficacy of liver organoids, as IGF2-treated liver organoids significantly reduced fibrosis and restored liver function in a murine model of chronic liver injury. Although we observed a clear effect of IGF signaling, the optimal concentration and timing of IGF2 supplementation remain to be determined. Fine-tuning the activation level of the IGF-IGF1R-MAPK/AKT axis would be essential to ensure both efficacy and safety in future translational efforts. Furthermore, other signaling pathways may act cooperatively with IGF signaling to regulate maturation and functions of reprogrammed cells. The Epac-Rap1 pathway enhances barrier function in induced ECs (53, 54), while Wnt signaling is widely involved in hepatic lineage specification (33, 55). For instance, Ebrahim et al. (56) demonstrated that recellularization of dECM liver scaffolds with iPSCs up-regulated multiple Wnt ligands, which were associated with enhanced expression of hepatic specification markers and restoration of liver-specific functions. Thus, dissecting the roles of various molecular cues will be critical to comprehensively understand the regulatory landscape and optimize bioengineered liver organoids. Together, our findings highlighted the importance of integrating developmental and proregenerative signaling with biomaterial-based spatial guidance to engineer functional, transplantable liver organoids.

Despite the promising structural and functional outcomes achieved with our liver organoids, there are several limitations to be addressed for clinical applications. The xenogeneic dECM scaffolds used in this study were derived from rats, raising concerns about interspecies anatomical differences and ECM compatibility with human cells. While dECM scaffolds are often regarded as minimally immunogenic and highly biocompatible after proper decellularization, the immunological safety of such scaffolds remains a key consideration for translational applications. Insufficient decellularization may leave behind immunogenic cellular remnants, whereas overly harsh decellularization can lead to undesirable ECM alterations, both of which may compromise immune tolerance (57). Therefore, achieving an optimal balance between complete cell removal and ECM preservation is critical for minimizing scaffold immunogenicity. Thus, we attempted to thoroughly validate such immunogenicity of the dECM liver scaffolds through multiple approaches. First, our decellularization protocol was carefully optimized and demonstrated effective removal of cellular components and preserved ECM fibers, as demonstrated in Fig. 1A (13). Second, we specifically verified the elimination of the immunogenic α-GAL epitope, a known xenogeneic antigen via immunofluorescence staining, which revealed no detectable signal in the dECM liver scaffolds (Fig. 1C). Last, we conducted LC-MS/MS of the dECM liver, which confirmed the significant enrichment of normal ECM proteins. Together, these results validate that our dECM scaffolds are minimally immunogenic with a rich repertoire of native ECM proteins. Nevertheless, to support future clinical translation, strategies to ensure the immunological safety of the xenogeneic dECM liver scaffold should be profoundly investigated. Despite the greater biochemical homology of human liver dECM, its scarcity and ethical constraints render it impractical for routine use in preclinical research and therapeutic applications, making dECM scaffolds derived from large animals a more feasible clinical analog. Next, our organoids, which predominantly express periportal (zone 1)–specific markers, did not fully recapitulate the spatially zonated architecture of the native liver. These unclear zonation patterns in our vascularized liver organoids may be attributed to multiple factors: (i) Recent studies have emphasized that the establishment and maintenance of hepatic zonation are governed predominantly by LSECs, rather than vascular ECs (33, 58, 59). However, the endothelial differentiation protocol used in this study primarily directed cells toward an arterial endothelial phenotype (60). These arterial ECs differ markedly from the venous-like, fenestrated phenotype characteristic of LSEC and likely contributed to the emergence of only a limited number of LSEC-like cells in our organoids. Consequently, environmental gradients necessary for zonation may not have been sufficiently established. (ii) Emerging evidence has suggested that mature cholangiocytes also play an instructive role in portal-central zonation by secreting Wnt and FGF19 signals, which modulate the metabolic phenotype of neighboring hepatocytes in a zone-specific manner (55, 61). Although cholangiocyte-like cells have been detected in our model, they lacked structural and functional maturity, including organized bile duct–like structures. Moreover, given that ECM microstructure and biochemical cues critically govern cell patterning and tissue-specific organization (62), the rat-derived scaffolds without a gallbladder and distinct bile drainage system may inherently restrict the reconstruction of the hepatobiliary structure and functions. (iii) In addition, other nonparenchymal liver cell types, such as Kupffer cells and hepatic stellate cells, are also known to contribute to the vascular niche and functional parenchymal architecture (33, 63, 64). However, their abundance was not sufficient to fully recapitulate the complex architecture of the native liver. Thus, future efforts aimed at enhancing the cellular heterogeneity of the organoids including the integration of liver-resident cell types and applying biochemical or flow-based cues will be essential to reconstruct more physiologically relevant zonated liver organoid models. Last, future studies evaluating long-term engraftment and immune compatibility in large animal transplantation models will be needed to advance clinical translation.

In conclusion, we present a strategy for reconstructing highly organized, vascularized liver tissues by combining anti-CD31 aptamer–loaded dECM scaffolds with IGF signaling activation. This synergistic approach not only enhanced spatial organization but also promoted functional maturation across both vascular and parenchymal compartments. PHLI-based spatial transcriptomics enabled compartment-specific profiling, revealing how IGF2-mediated signaling networks coordinate tissue architecture within bioengineered tissue constructs. In vivo transplantation into a murine chronic liver failure model further demonstrated the regenerative efficacy of vascularized liver organoids. Collectively, this work highlights a comprehensive framework for generating human iPSC–derived transplantable liver grafts by integrating engineered biomaterials and spatial transcriptomic technology.

MATERIALS AND METHODS

Human iPSC culture and differentiation

All human resources were approved by Seoul National University Institutional Reviewer Board (IRB No. 23-03-013). The human iPSC line no. CMC-003 was gifted from the National Stem Cell Bank of Korea. As previously described (65, 66), human iPSCs were maintained in StemFlex Medium (A3349401, Thermo Fisher Scientific, US) on vitronectin (Gibco, US)–coated dishes and passaged with ReLeSR (STEMCELL Technologies, Canada). On the basis of a previous report (50), iPSCs were differentiated into hepatoblasts and mature hepatocytes. Differentiated hepatoblasts were expanded in Dulbecco’s modified Eagle’s medium/Nutrient Mixture F12 (Gibco) supplemented with 10% fetal bovine serum (Gibco), 1× Insulin-Transferrin-Selenium (Gibco), human hepatocyte growth factor (20 ng/ml; Peprotech), epidermal growth factor (10 ng/ml; Peprotech), 5 μM A83-01 (Peprotech), and 5 μM Y-27632 (Medchem Express, US). To differentiate into hepatocytes, hepatoblasts were maintained in RPMI 1640 supplemented with 1× B-27, 10 μM dexamethasone (Sigma-Aldrich, US), and oncostatin M (10 ng/ml; Peprotech) for 10 days. Next, the iPSCs were also committed into ECs, as previously reported (67). Then, the EPCs were subjected to magnetic cell separation with CD31 microbeads (no. 130-091-935, Miltenyi Biotec, Germany). The positively sorted CD31⁺ induced ECs were seeded on fibronectin (Corning)–coated dishes and maintained in EGM-2 (Lonza, Switzerland) with 10% fetal bovine serum and 10 μM SB431542 (Peprotech). All cells were incubated in a humidified CO₂ (carbon dioxide) incubator at 37°C.

qRT-PCR

Total RNA was extracted from cells or organoids using NucleoZOL reagent (Macherey-Nagel, Germany). Extracted RNA was transcribed into complementary DNA using the Superscript III First-Strand Synthesis system (BRL-18080-051, Invitrogen) under the manufacturer’s instructions. Relative mRNA expression of genes was determined by using the ABI 7300 Realtime PCR system (Applied Biosystems, US) and SYBR Green PCR Master Mix (Applied Biosystems), which was normalized to that of the housekeeping gene GAPDH. The list of primer information was provided (table S1, Supplementary Materials).

Immunofluorescence assay

The cells or cryosectioned liver organoids were fixed with 4% paraformaldehyde and permeabilized with 0.2% Triton X-100. After washing with PBS, the samples were blocked with 5% normal goat serum and incubated with the primary antibodies overnight at 4°C (table S2, Supplementary Materials). After PBS washing, fluorescence-conjugated secondary antibodies (A11008 and A11001; Invitrogen, US) were applied for an hour at room temperature. For nucleus staining, 4,6-diamidino-2-phenylindole dihydrochloride (DAPI; D21490, Invitrogen) was applied for 10 min at room temperature. The stained samples were mounted with fluorescent mounting medium (S302380, DAKO, Denmark) and visualized by confocal microscopy (Eclipse TE2000, Nikon, Japan) and the EZ-C1 3.8 program (Nikon). The mean fluorescence intensity (MFI) or the number of cells expressing the protein of interest was quantified by ImageJ software.

Flow cytometry

The cells differentiated from iPSCs were harvested and permeabilized with a transcription factor buffer set (no. 562574, BD Pharmingen), followed by probing with the fluorescently conjugated antibodies for 15 min at 4°C (table S3, Supplementary Materials). After washing, stained cells were subjected to the MACSQuant Analyzer 10 Flow cytometer and analyzed by MACSQuantify (Miltenyi Biotec) and FlowJo version 9.0.

Production of the dECM liver scaffold

As previously described (6), whole livers were isolated from 6-week-old Sprague-Dawley rats for producing dECM liver scaffolds. For decellularization, the portal vein and bile duct of rat livers were catheterized and perfused with 0.1% sodium dodecyl sulfate (SDS) overnight at 4°C. After washing with PBS, hepatic lobes except the median lobe were removed, followed by sterilization with 0.1% peracetic acid (Sigma-Aldrich). Sterilized dECM liver scaffolds were stored at 4°C before experiments. All experiments were carried out under the guidelines of Institutional Animal Care and Use Committee in Seoul National University (SNU-220814-1-2).

Histological analysis

For H&E staining, the slides were incubated with hematoxylin for 5 min, soaked in 0.5% HCl in ethanol, and subsequently stained with eosin for 1 min. For picrosirius red staining, 0.1% Direct red 80 and 0.1% Fast green FCF (both purchased from Sigma-Aldrich) were mixed in picric acid. The sectioned samples were incubated with the mixed solution for an hour at room temperature, followed by washing with 0.5% acetic acid for two times. For assessing the glycogen synthesis function of liver organoids, periodic acid Schiff (PAS) staining was performed according to the manufacturer’s instructions (ab150680, Abcam, UK). Last, all stained samples were subjected to an increasing sequential ethanol series and mounted in Canada Balsam. The samples were visualized with a microscope and ProgRes CapturePro software (Olympus, Japan).

SEM

The scaffolds were fixed with a mixture of 4% glutaraldehyde and 1% paraformaldehyde in 0.1 M cacodylate buffer, followed by critical point drying. Then, the samples were subjected to sputter coating and then visualized by the Variable Pressure Emission Scanning Electron Microscope (SUPRA55VP, Carl Zeiss, Germany).

Analysis of ECM composition in dECM scaffolds

The contents of ECM components in the dECM liver scaffolds were measured using the Sircol soluble collagen assay (S1000, Biocolor, UK), Fastin Elastin assay (F2000, Biocolor), and Blyscan sulfated GAG assay (B1000, Biocolor). Further proteomic analysis of native liver and dECM liver scaffolds was conducted by LC (Eksigent Ekspert NanoLC 425, Canada)-MS/MS (SCIEX TripleTOF 5600, Japan). The Rattus norvegicus database, containing 31,571 sequences and 17,286,506 residues, was obtained from UniProt (release date: 20 December 2017). The peptide-spectrum match step was conducted with a peptide mass tolerance of 10 ppm (parts per million) and a fragment mass tolerance of 0.8 Da. Peptides and proteins with a false discovery rate below 1% were selected for each group. MaxQuant was used to determine the iBAQ values for each protein. The relative expression of proteins enriched in terms related to growth factors, vascular development and angiogenesis, and liver development in the dECM liver was visualized using a heatmap created with GraphPad Prism version 9.0. (US). Matrisome proteins in both the native liver and dECM liver identified from the LC-MS data using MatrisomeDB 2.0 (https://sites.google.com/uic.edu/matrisome/home) were classified into core matrisomes and matrisome-associated proteins. Proteomics data were provided as a supplementary file. Chi-square tests for homogeneity and independence were performed to compare ECM component proportions between native and dECM livers using R software.

Dextran perfusion assay

To evaluate vascular barrier integrity within the reendothelialized liver organoids, scaffolds were perfused with fluorescein isothiocyanate-dextran (1 mg/ml; 500 kDa; FD500S, Sigma-Aldrich) through the portal vein at a flow rate of 1 to 2 ml/min for 20 min at room temperature. After perfusion, the excess tracer was removed by rinsing with PBS for 5 min, and the samples were fixed by perfusion with 4% paraformaldehyde for 20 min at room temperature and washed with PBS. For high-resolution three-dimensional imaging, optical clearing was performed using the CytoVista tissue clearing reagent (Thermo Fisher Scientific). Briefly, fixed samples were perfused with penetration buffer for 30 min at room temperature and blocked in blocking buffer for 30 min at 37°C. Endothelial networks were then labeled by perfusing rhodamine-conjugated BSL-I (10 μg/ml in blocking buffer, RL1102, Vector Laboratories) for 1 hour at 37°C. After washing, the organoids were dehydrated through a graded ethanol series and incubated in a clearing solution overnight at 4°C. 3D fluorescence imaging was performed using a Thunder 3D imaging system (Leica microsystems, Germany), and reconstructed z-stack fluorescence images were computationally processed by Thunder_LasX software.

TUNEL assay

To detect apoptotic cells in vascularized liver organoids, cryosectioned tissue samples were subjected to TUNEL assay using the ApopTag Red In Situ Apoptosis Detection kit (S7165, Merck Millipore). Then, the samples were subjected to nucleus staining, followed by visualization using confocal microscopy.

Cell viability test

To optimize the coculture condition of cells differentiated toward each lineage, cell viability was examined with MTT reagent (M6494, Invitrogen). The cells were plated on 24-well plates, and a different cell type was plated on transwell inserts (SPL, Republic of Korea). After coculturing for a week, the cells were incubated with 10% MTT reagent for 4 hours at 37°C. The resulting purple formazan was dissolved in dimethyl sulfoxide. The absorbance at 570 nm was measured using a TECAN Infinite M200 Pro (Switzerland).

Generation of iPSC-derived vascularized liver organoids

To fabricate vascularized liver organoids with dECM liver scaffolds, hepatoblasts and ECs were differentiated from iPSCs. Before recellularization, characteristics of each cell type were confirmed by flow cytometry (hepatoblast: ALB⁺ CK18⁺; EC: CD31⁺ CD144⁺). First, 1 × 10⁷ iPSC-derived hepatoblasts were repopulated into sterilized dECM liver scaffolds via bile duct injection and continuously perfused with hepatoblast expansion medium supplemented with Matrigel-growth factor reduced (MAT-GFR; Corning) using a bioreactor and peristaltic pump. After 3 days, an additional 1 × 10⁷ iPSC-derived hepatoblasts were injected into the bile duct of liver organoids and maintained in the same medium for 3 days. Then, the medium was changed to hepatocyte maturation medium supplemented with MAT-GFR. To assess the efficacy of IGF signaling activation, rIGF2 (25 ng/ml; Peprotech) was supplemented from this period. After 6 days of hepatic maturation, the vasculature in liver organoids was coated with PBS or a 600 nM anti-CD31 aptamer (no. 2196-18-01, Aptamer Sciences, Republic of Korea) as a VCA for an hour at 4°C, as described in a previous study (BM ref). Subsequently, the organoids were repopulated with 8 × 10⁶ iPSC-derived ECs through the portal vein and maintained in a 1:1 mixture of hepatocyte maturation medium and endothelial medium. After an additional 7 days of coculture, the vascularized liver organoids were harvested and subjected to further experiments.

Enzyme-linked immunosorbent assay (ELISA)

To analyze the secretome of vascularized liver organoids, the supernatant of conditioned media was harvested, concentrated with Amicon ultra-15 centrifugal filters (UFC9030, Millipore, US), and subjected to enzyme-linked immunosorbent assay (ELISA). To assess EC-specific functions of organoids, the human VEGF Quantikine ELISA kit (DVE00, R&D Systems, US) and Nitric Oxide detection kit (21023, iNtRON Biotechnology, Republic of Korea) were used respectively to determine the concentration of human VEGF and nitric oxide secreted from liver organoids. To evaluate hepatic-specific functions, the concentrations of secreted albumin and urea from organoids were respectively measured by the Human Albumin ELISA Kit (ab108788, Abcam) and QuantiChrom Urea Assay Kit (DIUR-100, Bioassay System, US), according to the manufacturer’s suggestion. Serum samples harvested from mice were subjected to the ALT activity assay (K752-00, Biovision, China) and AST activity assay (753-100, Biovision).

PHLI-seq

Vascularized liver organoids in each group were cryosectioned at a 15-μm thickness. Each tissue section was mounted on indium tin oxide–coated glass slide and stored −80°C. For whole-slide imaging, an automated inverted microscope (Eclipse Ti-E, Nikon) was used to scan the slides that were washed with PBS, and the scanned images were stitched using an internal algorithm (NIS-Elements AR Auto Research, Nikon). After designating the ROIs on the basis of whole-slide images, the samples were subjected to a PHLI instrument for isolating the cells from each ROI. The customized instrument features motorized stages (ACS Motion Control, Israel), a charge-coupled device camera (Jenoptik, Germany), Nd:yttrium-aluminum-garnet nanosecond laser (Minilite Series ML II; Continuum, US) and light sources, a pulse slit, and objective lenses (Mitutoyo) to automate the isolation of cells with precise laser targeting and adjustable laser pulse size. Each isolated region contained ∼5 to 20 cells. The retrieved cells were lysed with proteinase K, and then poly-T capture was conducted with probes, as previously described (20). Reverse transcription was carried out on the cell lysate, followed by PCR to amplify the entire cDNA library. By using Tn5 transposases, the full-length transcriptome library was prepared for next-generation sequencing and subsequently sequenced on an NextSeq (Illumina, US) using 150–base pair paired-end sequencing. To process the index barcodes, Cutadapt was used for demultiplexing, while BBmap filtered out reads with quality scores under 20 or lengths shorter than 25 base pairs. The remaining reads were aligned to the human genome Ensembl GRCh38 using STAR Aligner. For transcriptome-wide quality assessment, the Read Distribution module was used to quantify reads mapped to coding DNA sequence exons, untranslated regions, introns, and intergenic regions using a GENCODE version 38 BED12-formatted annotation. In addition, the Gene Body Coverage module was used to examine the 5′ to 3′ coverage profile across transcripts. The featureCounts was applied with its default parameters for counting uniquely mapped reads per sample, which were then normalized using the DESeq2 R package. Using the DESeq2 package, DEGs with a P value below 0.05 and a log₂ fold change greater than 1 or less than −1 were acquired. Heatmap analysis and hierarchical clustering were visualized using the pheatmap (version 1.0.12) and ComplexHeatmap (version 2.14.0) R packages. GO term analysis was conducted with DEGs by using DAVID Bioinformatics resources (https://davidbioinformatics.nih.gov/), and the enriched GO terms were prioritized according to their P values. GSEA was performed using GSEA 4.3.3 software (Broad Institute, US). Enrichment was evaluated against the following gene sets in terms of molecular function, biological process, and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway. The permutation type was set to gene set (n = 1000 permutations), and significance was determined on the basis of the normalized enrichment score (NES).

RNA interference

Human iPSC–derived hepatocytes (5 × 10⁵) and ECs (5 × 10⁵) were plated on 35-mm dishes. After 24 hours, the cells were transfected with nontargeting control siRNA (siControl, no. 4390844, Thermo Fisher Scientific) or siRNA targeting IGF2 (siIGF2; s7216, no. 4392420, Thermo Fisher Scientific). Briefly, 1 μg of siRNA oligomers was diluted in Opti-MEM I reduced serum medium (Invitrogen) and reacted with a transfection reagent, Lipofectamine RNAiMAX (Invitrogen), for 20 min. The complexes were treated to the cells for 24 hours. The cells were subjected to further analysis 24 to 48 hours after transfection.

Western blot

Tissue lysates were extracted by homogenization with PRO-PREP (iNtRON Biotechnology, Republic of Korea), followed by sonication. Ten micrograms of proteins was loaded on tris-glycine SDS-polyacrylamide gels, separated by electrophoresis, and transferred to a 0.2-μm BioTrace NT nitrocellulose transfer membrane (Cytiva, US). The membrane was blocked with 3% bovine serum albumin and probed with the following primary antibodies overnight at 4°C (table S4, Supplementary Materials). After washing, the membranes were incubated with horseradish peroxidase–conjugated secondary antibodies (G21040 and G21234; both purchased from Invitrogen). Proteins of interest were developed using the Clarity Western ECL Substrate (Bio-Rad, US) and ChemiDoc MP Imaging System with Image Lab Software (Bio-Rad).

Tube formation assay

Seventy-five microliters of MAT-GFR was added to each well of a 96-well plate and solidified at 37°C for an hour. iPSC-derived ECs (2 × 10⁴) transfected with siControl or siIGF2 were seeded on solidified MAT-GFR. The tube-like structures were observed after 18 hours, and the branching points were quantified by using ImageJ software.

Transplantation of liver organoids into the chronic failure mouse model

All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Seoul National University, in accordance with the guidelines for the care and use of laboratory animals (IACUC-211130-2). Chronic liver failure was induced in 6-week-old male BALB/c mice by oral administration of TAA dissolved in sterile drinking water for a total of 12 weeks. After 8 weeks of TAA administration, vascularized liver organoids from four experimental groups were orthotopically transplanted into the liver of the mice: PBS + veh, VCA + veh, PBS + IGF2, and VCA + IGF2. During the additional 4 weeks following transplantation, the mice received continued TAA administration along with prednisolone sodium phosphate (Supelco, PHR2816; 5 mg/kg per day) in sterile drinking water until euthanasia. Blood samples and the liver from each mouse were harvested and subjected to further experiments.

Statistical analysis

Data are presented as the means ± standard deviation (SD). At least three independent experiments were performed unless otherwise stated. As indicated in each figure legend, an unpaired two-tailed t test, one-way analysis of variance (ANOVA), or two-way ANOVA with post hoc Tukey’s test for multiple comparisons was performed with GraphPad Prism software version 9.0. The P value under 0.05 was considered statistically significant and stated in the graph as follows: *P < 0.05, **P < 0.01, and ***P < 0.001.

Supplementary Materials

The PDF file includes:

Figs. S1 to S8

Tables S1 to S4

Legends for movies S1 and S2

Other Supplementary Material for this manuscript includes the following:

Movies S1 and S2