Human iPS derived progenitors bioengineered into liver organoids using an inverted colloidal crystal poly (ethylene glycol) scaffold

Generation of human organoids from induced pluripotent stem cells (iPSCs) offers exciting possibilities for developmental biology, disease modelling and cell therapy. Signi ﬁ cant advances towards those goals have been hampered by dependence on animal derived matrices (e.g. Matrigel), immortalized cell lines and resultant structures that are dif ﬁ cult to control or scale. To address these challenges, we aimed to develop a fully de ﬁ ned liver organoid platform using inverted colloid crystal (ICC) whose 3-dimensional mechanical properties could be engineered to recapitulate the extracellular niche sensed by hepatic progenitors during human development. iPSC derived hepatic progenitors (IH) formed organoids most optimally in ICC scaffolds constructed with 140 m m diameter pores coated with type I collagen in a two-step process mimicking liver bud formation. The resultant organoids were closer to adult tissue, compared to 2D and 3D controls, with respect to morphology, gene expression, protein secretion, drug metabolism and viral infection and could integrate, vascularise and function following implantation into livers of immune-de ﬁ cient mice. Preliminary interrogation of the underpinning mechanisms highlighted the importance of TGF b and hedgehog signalling pathways. The combination of functional relevance with tuneable mechanical properties leads us to propose this bioengineered platform to be ideally suited for a range of future mechanistic and clinical organoid related applications.

Human iPS derived progenitors bioengineered into liver organoids using an inverted colloidal crystal poly (ethylene glycol) scaffold 1

. Introduction
Induced pluripotent stem cells offer exciting new possibilities in developmental biology, disease modelling and transplantation. Comprehensive realization of that promise is likely to require consolidation with other emerging technologies such as 3D cell culture. Along those lines, hepatic progenitor cells derived from iPSCs [1] and primary tissue [2] were recently shown to form "Organoids" (self-organising miniaturized 3D structures resembling liver tissue) following 3D culture initiated by Matrigel. Organoid generation using Matrigel has similarly been demonstrated across a broad range of tissue types including intestine, kidney and brain [3e5], underlining the importance of this novel approach across the stem cell field. Downstream applications are however limited by the use of Matrigel which is poorly characterized, highly variable and of mouse origin [6]. A bioengineered substitute is therefore essential and was recently reported for intestinal organoid generation [7]. A similar bottom up engineering approach for liver organoid production is urgently needed. We recently developed a novel hepatocyte culture system composed of a 3D-hexagonally arrayed inverted colloidal crystal (ICC) scaffold [8]. In comparison to other 3D culture systems, the ICC boasts several advantages. These include being made from an FDA approved material, polyethylene glycol (PEG), that can be functionalized with select ECM proteins or varied in mechanical stiffness, a highly uniform architecture which incorporates sizeselectable pores to facilitate tissue interconnection across the whole module whilst retaining uniform nutrient penetration to all cells and finally, being transparent to provide an easy means of viewing cells and intra-cellular fluorescence over time. Exploiting these properties, we were able to facilitate induction of organoid structures with advanced liver function in primary human fetal liver cells cultured in ICC's [9]. Lack of donor material, ethical constraints and biological heterogeneity however make further progress difficult with human fetal cells. We therefore turned to the recently established technology of human iPSC-derived hepatocytes [10,11], as a good biological approximation to human fetal liver cells but not limited by the same constraints cited above, to explore their potential for organoid generation.

Study design
The overall objective of this study was to test whether organoids bioengineered in ICC scaffolds exhibited a more physiologically relevant liver phenotype compared to conventional 2D or 3D (Matrigel and Spheroid) models. iPSC-derived hepatic progenitors (IH) differentiated from the same parental line were therefore matured in four different in vitro models (ICC, 3D Matrigel, Spheroid and 2D) for 14 days and compared. Measurement techniques to compare the three models were designed to characterize two major endpoints: morphology (organoid formation) and function (gene expression, protein expression, liver specific functions and in vivo behavior). In transplant experiments, littermate animals were randomly assigned to experimental groups and analysis of samples was performed blindly. Studies were repeated at least three times.

Cells
All human tissues were collected with informed consent following ethical and institutional guidelines. Freshly isolated hepatocytes were obtained from Triangle Research Labs (TRL) and Human Developmental Biology Resource (HDBR) University College London (UCL). Fetal hepatocytes were obtained from 16 to 20 week old fetuses (HDBR), dissociated as previously described [12]. Three different human iPSC lines were used to generate hepatic progenitors (IH) for these experiments: [1,10] (https:// stemcells.nindsgenetics.org/), [11] the latter of which is considered to be of 'clinical grade' and thereby potentially suitable for future human therapy.

ICC fabrication
Sacrificial crystal lattices were constructed on a rhombus PDMS mold with 17mmx15 mm diagonal lengths and 2 mm depth using polystyrene beads with diameter of 40 mm, 60 mm, 100 mm, or 140 mm (Duke Scientific) and annealed under 140 C for 1.5 h. Lattices were then infiltrated with precursor solution of 50 %w/v polyethylene glycol-diacrylate (PEGDA, Alfa Aesar), 10 %w/v acrylate-PEG-N-hydroxysuccinimide (Laysan Bio Inc, AL) and 1 %w/ v photoinitiator (Irgacure ® 2959, BASF, Switzerland) in DI water. The precursor solution underwent free-radical induced polymerization under 75 W xenon ultraviolet (UV) light source (Oriel Instruments, Mountain View, CA) for 10 min. After polymerization, the polystyrene sacrificial lattice was removed by tetrahydrofuran treatment for 2 h. The resulting ICC hydrogel scaffolds were equilibrated in deionized water and conjugated with appropriate ECM proteins, such as 100 mg/mL Collagen I (Col-I, Sigma), 50 mg/mL Fibronectin (Fn, RnD), and 10 mg/mL Laminin 521 (Ln-521, Biolamina) (Fig. 1A). The final tubular ICC scaffolds for 96 well plate format culture were created using 6 mm biopsy punch.
Approximately 4 ml cell suspension was then pipetted onto top and bottom surfaces of partially dehydrated ICC scaffolds and placed in an incubator for 30 min without media to minimize cell loss and maximize cell attachment. Cell-laden ICC scaffolds were then transferred to a new 96-well plate for culture and media were refreshed at 4 h post-seeding to remove excessive cells that are not attached to surface or established cellular clusters. Cells were left for a further 14 days to mature. Media was refreshed every two days. 2D controls were generated using IH cells seeded into standard 48 well plate tissue culture plastic plates coated with the same ECM proteins and cultured with the same reagents as used for ICC culture.

Generation of IH-spheroids
A single cell suspension of IH was prepared and 0.3 Â 10 6 cells were seeded per well of a 24-well Aggrewell-400 (STEMCELL Technologies, Vancouver, Canada). Aggrewell plates were prepared as recommended by the supplier. Centrifugation at 200G for 3 min was carried out to deposit cells into the microwells of the plate.

Bright field imaging
Bright field imaging was captured using Leica DMIL LED equipped with Leica DFC3000 G camera and images processed by LAS X software.

Cell number quantification
Double stranded DNA was collected, quantified and interpolated from standard curve plotted using a range of known cell number as illustrated by Quant-iT™ PicoGreen™ dsDNA Assay Kit (Thermo-Fisher Scientific).

Image analysis
To analyse the confluency and cluster morphology of cells in ICC, selected images of F-actin staining were reconstructed using maximum intensity projection. The confluency index ¼(Cc/ Cp Â 100%) is introduced to quantify the proportion of the ICC surface covered by cell attachment by measuring the circumference of cell lining (Cc) over the overall circumference of pores in ICC (Cp). The cluster index ¼ (Kc/Kp Â 100%) is introduced to quantify the number of cell cluster in each pore (Kc) against the total number of pores (Kp) in the frame.

Fluorescence-activated cell sorting (FACS) staining and analysis
Cells were dissociated from organoid or 2D culture using TrypLE reagent (Thermo Fisher Scientific) and stained using the same condition as immunofluorescence staining protocol stated above. Flow analysis performed using FACSCanto II (BD) and analyized using FlowJo (LLC).

Quantitative real-time PCR (RT-qPCR) analysis
Total RNA was harvested using TRIZOL reagent (Sigma), treated with DNase (Promega) and phenol/chloroform purified. For each sample 0.5 mg of total RNA was reverse transcribed using Super-Script VILO cDNA Synthesis kit (Thermo Fisher Scientific). A typical RT-PCR reaction contained 10 ng of sample cDNA, 0.0075 ml of individual forward and reverse primer each at 100 mM stock, 5 ml Taqman Universal Master mix (Applied Biosystems), 1 ml Taqman target probe (supplementary Figs. 5) and made up to 10 ml with nuclease-free water. Real time PCR reactions were amplified for 40 cycles on a CFX384 Touch™ Real-Time PCR Detection System (Biorad) in triplicate and normalized to ACTB in the same run.

Heat map generation, gene set enrichment analysis, functional network analysis and top canonical pathway analysis
Heat maps were generated from Bulk RNAseq data collected from three Human Fetal Livers, three Human Adult Livers and three iPSC-derived hepatocytes harvested at definitive endodermal stage from the BOB cell line. The three fetal livers included 14pcw, 16pcw and 20pcw. The three adult Livers were Female 18yrs, Male 43yrs and Male 13yrs. RNA was extracted using TRI reagent. Starting with 1 mg input total RNA, Ribosomal RNA was removed using Ribo-Zero Gold rRNA Removal kit (Illumia). Sequencing libraries were prepared using NEBNext ® Ultra™ Directional RNA Library Prep Kit for Illumina (N.E.B) using 100 ng rRNA depleted sample and sequenced on a HiSeq 2500 system in Rapid run mode (Illumina) following a standard protocol. All libraries generated between 15 and 25 M reads. Reads were mapped to GRCh38 reference genome using Bowtie2. Raw counts and normalized gene expression was generated using HT-Seq and DESEq2 packages respectively. Heat map was generated using R (http://www.R-project.org) (R Development Core Team, 2008). Heat maps represent average DESEq2 normalized gene expression values of three independent biological samples. Gene set enrichment analysis was performed on normalized RNA sequencing gene expression data through GSEA software 27,28 run using the hallmark MSigDB gene set collections. To examine the role of IH-ICC organoid in cell polarity and bile acid synthesis and metabolism, we cross-referenced the differential gene expression with the Gene Ontology (GO) annotations and predict the protein interactions networking using Search Tool for the Retrieval of Interacting Genes (STRING). Top canonical pathways analysis was performed using Ingenuity Pathway Analysis (IPA) (Qiagen).

GLuc HCVcc infection and analysis
GLuc HCVcc is a full length Gaussia luciferase reporter virus Jc1FLAG2 (p7-nsGluc2A)35 infectious Jc1 genome that allows us to monitor viral infection in real time. Knock down HCVcc is the corresponding polymerase-defective mutant that inhibits viral replication. Respective viral stocks were prepared using electroporation on Huh 7.5 cells as described [14]. Infectious titers of HCVcc inocula were determined using Huh 7.5 cells as described. Cultures were inoculated for 3 h with HCVcc then washed extensively with PBS and fed with culture media.

Cholyl-lysyl-fluorescein (CLF) transportation
5 mM CLF (Corning) was incubated with cultures treated with 10 mM Troglitazone (TGZ, Sigma) and 0.02% DMSO vehicle control for 45min. The cultures were then washed five times with PBS and incubated in culture media for 30min before confocal imaging. For CLF quantification, cultures where treated by cell lysis buffer (Sigma) and fluorescence intensity was measured using plate reader.

Animal experiments
All animal experiments were performed in accordance with UK Home Office regulations (UK Home Office Project License number PPL 70/8702). Immunodeficient NOD.Cg-Prkdc scid Il2rg tm1Wjl /SzJ (NSG) mice which lack B, T and NK lymphocytes were bred in-house with food and water available ad libitum pre-and post-procedures. A mix of male and female animals were used, aged approximately 6e8 weeks. An incision was made in the capsule of the caudate lobe of the liver and a 'pocket' was created by raising a flap of capsule on either side of the incision. The scaffold was slid into the pocket and covered by the capsular flaps. The left lobe of the liver was then allowed to cover the site of the transplant.

Statistical analysis
N in the paper represents the number of biological replicates of each batch of liver specific differentiation performed using three iPSC lines. Statistical significance (p < 0.05) was determined by using Student's T-test (assume Gaussian distribution, two-tailed) or One-way ANOVA followed by Tukey's posthoc test. .

iPSC-derived hepatic progenitors (IH) form organoids analogous to fetal liver derived primary cells following culture in ICC scaffolds
Human fetal liver cells seeded into ICC scaffolds with the 140 mm pore size and type I collagen (Col-I) coating attach as a single cell layer (Phase I e days 0e5 post-seeding) before organising into morphologically stable, interconnected clusters (Phase II -days 7 post-seeding onward) (Fig. 1C). The resultant 3D structure resembles a mini-liver or hepatic 'organoid' with demonstrated advanced liver function [9]. We therefore hypothesized that organoid formation was critically dependent upon the translation of specific 'cell-matrix' and 'cell-cell' signals by hepatic progenitors in a manner which could be recapitulated using iPSC-derived progenitors (IH) (Supplementary Fig. 1). To test this hypothesis, we therefore engineered a library of ICC scaffolds to evaluate the effects of selected extracellular matrix protein (ECM) coatings and pore sizes on IH-organoid forming ability (Fig. 2). Of the ECM coatings tested, only cells seeded into Col-I were able to attach and self-organize into interconnected clusters throughout the scaffold as observed with primary fetal progenitors (Fetal) (Fig. 2AeC,  Supplementary Fig. 2). Cells seeded into fibronectin (Fn) attached but failed to establish confluency whilst cells seeded into laminin-521 (Ln-521) attached and remained as such throughout the culture. The observed organoid forming ability was matched by functional parameters of liver tissue such as albumin secretion and hepatic gene expression ( Fig. 2D and Supplementary Fig. 3). Having established the optimal ECM coating, we next went on to assess the effects of cell-cell interaction by testing type I collagen coated ICC's composed from different pore sizes. Accordingly, a 140 mm pore size was found to be optimal for organoid formation based on morphology and again confirmed by hepatic function and gene expression (Fig. 2EeH). Through this extensive series of experiments, we observed IH seeded into type I collagen coated ICC scaffolds of 140 mm pore size best replicated the results seen with primary fetal liver cells with minimal cell lost and death . Importantly, the two-phase process of organogenesis was observed only with IH and primary fetal cells but did not occur with primary adult or cancer-like hepatocytes (Huh 7.5 and HepG2) (Supplementary Fig. 6).

IH-ICC derived organoids demonstrate morphological and transcriptomic characteristics of human liver
To evaluate the anatomical similarity between organoids derived from iPSC-hepatic progenitors cultured in ICC scaffolds (IH-ICC) and human liver, we first used a combination of histochemical and immunofluorescence staining to characterize the organoid tissue across the two phases of organogenesis previously described ( Fig. 3AeB and Supplementary Figs. 7-10). In Phase I, hepatic progenitor markers such as keratin 19 (CK19), epithelial cell adhesion molecule (EPCAM) and alpha-fetoprotein (AFP) proteins were expressed in cells lining the circumference of the scaffold's pores. Following onset of organogenesis in Phase II, these (CK19, EPCAM, AFP, EdU) þve progenitor cells were retained at the periphery of the organoid, with more mature cells, characterized by protein expression of asialoglycoprotein receptor 1 (ASGR1), albumin (ALB), and hepatocyte nuclear factor 4-alpha (HNF4A), forming and remodelled by the cell secreted type I collagen (Col-I) at the centre. These data were then validated quantitatively using flow cytometry (Fig. 3C) and by RT-PCR for gene expression (Fig. 3D), allowing us to conclude that organoids became established with unique cellular arrangement in a structure reminiscent of the hepatic buds seen in embryonic [15] and regenerating human liver tissue [16]. This anatomical distribution was not observed when IH were cultured in Matrigel, the current gold standard in the field, or the conventional 3D spheroid model. Embedding in Matrigel, IH established lumencontaining colonies, expressing mature hepatic markers ( Supplementary Fig. 11). On the contrary, IH formed solid cell clusters in spheroid model with more mature cells segregated on the outer layer ( Supplementary Fig. 12). Both Matrigel and spheroid models demonstrated functional properties (protein secretion and metabolic activity) significantly higher compared to 2D culture but notably not significantly higher than in ICC culture ( Supplementary  Fig. 13).
Having validated the organoid structure in our new bioengineered platform, we next sought to characterize the global transcriptomic changes during the different phases of organogenesis using RNA-seq. To do this, organoids were compared to iPSC derived definitive endoderm (DE), 2D cultured IH, primary adult and primary fetal hepatocytes. Principle component analysis (PCA) on DE, 2D, ICC, fetal and adult samples (Fig. 4A) shows that IH have a distinct transcriptional phenotype when cultured in ICC over 2D conditions. A liver-specific gene-set of 296 genes were selected to perform unbiased hierarchical clustering and visualized by heatmap. The unbiased hierarchical clustering revealed that the liverspecific transcriptional signature of organoids was positioned in the cluster next to adult and fetal liver, with higher similarity to fetal than adult, whereas 2D cells clustered next to DE (Fig. 3E).
Overall 489 genes were found to be significantly upregulated in both organoid and 2D samples, with 1349 and 179 genes uniquely upregulated in organoid and 2D systems respectively (Fig. 4B). Gene set enrichment and Ingenuity Pathway (IPA) analysis of the upregulated genes revealed some liver-specific functions such as xenobiotic and bile acid metabolism to be enriched in both organoid and 2D platforms (Supplementary Fig. 14 and Supplementary Tables 1 and 2) but the majority of liver-specific functions such as cholesterol homeostasis, glycolysis, fatty acid metabolism and protein secretion were found to be uniquely enriched in IH-ICC. The top 18 upregulated organoid specific hepatic genes were validated by qPCR (Supplementary Table 3 and Fig. 4C).
Cumulatively, these data confirm our previous morphological observations suggesting IH derived progenitors seeded into ICC scaffolds form organoid structures whose transcriptomic and protein expression profiles resemble human liver structures.

IH-ICC organoids are functionally closer to human liver
Upon extended interrogation, we discovered that IH-organoids are functionally more advanced than the 2D model in albumin production and basal activities of drug metabolizing cytochrome P450 isoforms CYP3A4 and CYP2C9 ( Fig. 5A and B). Similarly, capability for transporting bile salts also appeared more advanced. Hepatic polarity proteins critical for the structure of bile canaliculi such as multidrug resistance protein 2 (MRP2), tight junction protein 1 (ZO-1), CD26 and pan-cadherin were all expressed ( Fig. 5C and Supplementary Fig. 15). Bile salt efflux was then validated through uptake and accumulation of synthetic bile acid, cholyllysyl-fluorescence (CLF), in bile canaliculi-like regions (Fig. 5D). The ability of organoids to retain CLF was eliminated when treated with the BSEP inhibitor, Troglitazone (TGZ), as indicated by fluorescence measurement (Fig. 5E).
The cell polarity and bile acid secretory function of organoid was further validated by cross-referencing the uniquely upregulated differentially expressed genes in ICC with GO categories (GO 0007163: establishment or maintenance of cell polarity; GO 0000902: cell morphogenesis; GO 0007010: cytoskeleton organization; GO 0006699: bile acid biosynthetic process and GO 0008206: bile acid metabolic process) followed by functional network analysis using STRING on selected enriched gene sets ( Fig. 5F and Supplementary Figs. [16][17][18] [17]. The 28 genes uniquely upregulated in organoids could be assigned to the key KEGG pathways such as regulation of actin cytoskeleton, focal adhesion and tight junction required for cell polarization (Fig. 5G).

IH-ICC organoids are suitable for disease modelling and form vascularised tissue following transplantation
Next, to test the ICC organoid's suitability for disease modelling we investigated whether organoids expressed HCV host factors and were susceptible to HCV infection. Transcriptional analysis revealed expression of HCV host factors were enriched in organoids and more closely resembled the expression signature of primary hepatocytes (Fig. 6A). This included genes responsible for HCV entry such as claudin 1 (CLDN1) occludin (OCLN) scavenger receptor, class B, type 1 (SCARB1), CD81, low-density lipoprotein receptor (LDLR) and heparin sulfate glycoprotein (HSGP). In addition, a wide range of apoliproteins responsible for packaging and interacting with viral particles were highly expressed [18]. Immunofluorescence staining showed claudin 1 and occludin appear to localize at the interface of cell-cell junctions, similar to ZO-1 and MRP2 (Fig. 6B and Supplementary Fig. 19). These results are in agreement with the claudin 1 distribution seen in HepG2 and Huh-7 cell models permissive to HCV entry [19,20]. We next infected organoids with genotype 2a HCV reporter virus expressing secreted Gaussia The STRING functional network predicted the associations between proteins (nodes) from regulated genes involved in cell polarity in IH-ICC. The cluster analysis was performed using KMEANS clustering algorithms. Mean ± sd, *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.0001; ns nonsignificant. luciferase (HCVcc) and knock-down HCVcc (kd-HCVcc) which is incapable of replication and acts as a negative control. Luciferase signal was only detected in organoids inoculated with HCVcc cultures, whilst 2D cells and kd-HCVcc inoculated samples failed to exhibit detectable signal (Fig. 6C).
Having confirmed the organoid's preferential suitability for drug metabolism and disease modelling we next sought to explore the effects of in vivo transplantation. A pocket on the caudate lobe of murine liver was created by making an incision in the liver capsule. Organoids were placed into this pocket and sandwiched in place between the left lobe and the lower lateral lobe in order to achieve a bona fide homeostatic environment (Fig. 6D). After 4 weeks, grafts were retrieved for further analysis. H&E staining revealed implants were well integrated into the host parenchyma, without evidence of significant fibrosis/inflammation whilst neo vascularization had successfully occurred between host and donor tissues ( Fig. 6E and Supplementary Fig. 20AeB). Histochemical staining with human albumin confirmed the implanted structures were of human origin, the organoid structure had remained intact and the presence of human albumin in host serum suggested cells remained functional (Fig. 6F and Supplementary Fig. 20CeE).

TGFb and hedgehog signalling pathways are important for organoid formation
To identify signalling pathways involved in the orchestration of hepatic organoid formation, gene set enrichment analysis was performed as described before. The top 15 gene sets uniquely enriched in the ICC were related to metabolic/biosynthetic and inflammatory/immune related processes (Fig. 7A). The enrichment of bile acid metabolism, xenobiotic metabolism, fatty acid metabolism, heme metabolism and cholesterol homeostasis are encouraging signs of liver-specific organogenesis. Notably, three highly conserved developmental pathways were identified through this analysis e hedgehog, notch and TGFb. To confirm their functional relevance, we treated organoids with small molecule inhibitors of hedgehog (Cyclopamine e CYC, 0.2 mM), notch (DAPT, 10 mM) and TGFbR-1 (RepSox, 12.5 mM) and characterized the resultant effects on organoid formation. Morphological observations were also correlated with RT-qPCR evaluation of direct transcriptional targets for each signalling pathway (Fig. 7B). Cells managed to establish Phase I morphology (where cells lined up the surface of ICC) regardless of the treatment. However, cells treated with RepSox and CYC were unable to form typical organoid structures (Phase II), whilst DAPT treatment appeared to have little effect (Fig. 7C). RepSox treated cells arrested in Phase I of organogenesis resembling the observations seen with adult hepatocyte and liver carcinoma cells (Supplementary Fig. 6). CYC treated cells on the other hand, instead of transitioning into typical organoid structures, formed much smaller clusters that were less uniform in size and with a rougher surface. The regulatory network analysis on TGFB and hedgehog signalling pathways revealed several upstream ligands that are significantly up regulated in IH-ICC over 2D and could potentially serve as the ligand for signalling activation (Fig. 7D). The activation of these two pathways in IH-ICC were further validated by the translocation of phopho-SMAD2/3 and zinc finger GLI1 as shown in immunofluorescence staining ( Fig. 7E and Supplementary Fig. 21). To ensure cytotoxicity was not the causal effect, the selected concentration of each inhibitor used was validated to have had minimal to no significant impact on cell viability (Fig. 7F). Furthermore, the dramatic change in organogenesis observed corresponded to a direct pathway effect (Fig. 7G), reduced hepatic gene expression profile (Fig. 7F), and reduced albumin production rate (Fig. 7H).

Discussion
Human organs exist in an incredible variety of shapes and sizes. Understanding the processes through which these varied forms arise during development, are maintained in homeostasis or become perturbed during disease represent some of the most fundamental questions facing us today in biology. iPSC-derived cells provide an excellent model with which to study such questions through ex vivo formation of mini organs known as organoids. iPSC-derived organoids are also highly appealing because they themselves could be used for patient specific disease modelling and therapy. Here we report a new approach to the generation of iPSCderived hepatic organoids, using state of the art bio-engineering technology that by-passes the need for Matrigel and supporting cell co-culture. Utility of this platform was demonstrated by our findings in organoids of advanced hepatic function and dependence on key developmental signalling pathways (Hedgehog and TGFb).
Ex vivo liver organogenesis is particularly challenging due to the daunting complexity of the organ's structure and function. A pivotal breakthrough in this field was made several years ago by Takebe and colleagues when they generated hepatic organoids by co-culturing iPSC-derived hepatic endoderm, HUVECs and MSCs in Matrigel [1]. Immediate downstream translational applications are potentially limited however by dependence upon non FDAcompliant materials such as mouse-sarcoma-derived Matrigel and the difficulty of scaling up production when using multiple cell types (MSCs & HUVECs) [6]. A more recent and equally exciting organoid generation system, developed by Huch & Clevers [2], is similarly limited by the need for Matrigel but also by its need for primary tissue. In this study, we therefore sought to engineer liver organoids free from Matrigel, MSCs, and HUVECs, using a 3D synthetic hydrogel scaffold and hiPSC-derived hepatic progenitors alone. To that end we developed a 3D scaffold called ICC (inverted colloidal crystal) resembling the precise architectural microenvironment into which the liver diverticulum engages during liver bud formation [15]. Due to the ICC's unique bottom-up fabrication approach, we were able to custom engineer 'cell-matrix' interactions through coating the inner lining of the ICC pores with select ECM proteins [15] as well as custom engineer 'cell-cell' interactions through manipulating pore size [8]. These engineering designs cumulatively sought to recapitulate the extracellular niche sensed by hepatic progenitors during bud formation. We observed that the initial bud morphology, which we describe in this study as Phase I organogenesis/budding, could only be achieved when two critical factors were met. First, presence of a suitable ECM coating to facilitate cell attachment which otherwise fails due to the bioinert property of the PEG from which the ICC is manufactured (Fig. 2AeD). Second, seeding into a suitably sized pore, which otherwise forces cells to clump rather than attach to the scaffold's architecture if pore sizes are too small (Fig. 2EeH). Importantly, cell condensation (as opposed to cell-ECM engagement) was associated with down-regulation in hepatic gene expression, lower albumin production rates and failure to progress to Phase II in which cells lining the ICC self-organize into interconnected clusters to create a homogenously repeating organoid tissue structure. This series of morphogenic movements (Phase I / Phase II) was unique to iPSC (IH) and primary (fetal) hepatic progenitors and did not occur with terminally differentiated adult cells or with cancer cell lines ( Supplementary Fig. 6) [21,22]. We therefore hypothesize our 3D scaffolds leverage the intrinsic ability of liver stem/progenitor cells to form organized structures in a manner which conserves the interaction observed between progenitors and their surrounding niche during organogenesis/budding [23].
As a consequence of this two-phase organogenesis/budding, the ICC engineered organoids produce structures with progenitor cells at the periphery and more mature cells in the middle. Such arrangement is consistent with descriptions of early mammalian liver development, where liver diverticulum progressing from single cell lining to pseudostratified epithelium then threedimensional liver bud [25]. This bioengineering approach may therefore provide a useful, but simplistic model to further our understanding of early liver development (budding) in vivo ( Fig. 3A  and B). This configuration is in direct contrast to findings seen with spheroid/cell condensation culture systems where cells at the periphery are more viable and proliferative than cells at the core due to hypoxia and DNA damage [26e29]. That difference underscores the success of the engineering feat achieved here. The fact that liver specific drug metabolism and disease modelling become upregulated in parallel with these morphogenic changes along with the capacity of the organoids to integrate, neo-vascularise and survive in vivo emphasizes the physiological relevance of the tissue created here (Figs. 4 and 5). How such morphogenesis and zonation is programmed has been a long-standing question in biology. Our initial interrogation suggests both the TGFb and hedgehog signalling pathways to be pivotal in this regard. Hedgehog signalling is particularly interesting in the context of liver zonation because it has been reported to orchestrate the position of hepatocytes along an oxygen gradient [24]. Our treatment of progenitors with a hedgehog signalling pathway inhibitor, Cyclopamine, generated not only smaller but dysmorphic organoids ( Fig. 6C and D) and was accompanied by downregulation of liver function ( Fig. 6F and G). These data suggest Hedgehog may be involved in much more than simple cell positioning therefore.
To our knowledge, this is the first attempt to produce liver organoids using just the combination of iPSC-derived hepatic progenitors and a synthetic hydrogel. The unique modular features of the ICC scaffold will allow study of the complex, combinatorial influences of physical and chemical signals during liver organogenesis in a physiologically relevant, dissectible 3D microenvironment. In addition, the scalable structure and clinically compliant material (PEG) and cell lines [11] used, opens up the possibility of future human therapy. These results highlight the enormous potential of bioengineered organoids for discovery and translational science.

Data availability statement
The raw/processed data required to reproduce these findings cannot be shared at this time due to technical or time limitations.