Biofabricated 3D in vitro model of fibrosis‐induced abnormal hepatoblast/biliary progenitors' expansion of the developing liver

Abstract Congenital disorders of the biliary tract are the primary reason for pediatric liver failure and ultimately for pediatric liver transplant needs. Not all causes of these disorders are well understood, but it is known that liver fibrosis occurs in many of those afflicted. The goal of this study is to develop a simple yet robust model that recapitulates physico‐mechanical and cellular aspects of fibrosis mediated via hepatic stellate cells (HSCs) and their effects on biliary progenitor cells. Liver organoids were fabricated by embedding various HSCs, with distinctive abilities to generate mild to severe fibrotic environments, together with undifferentiated liver progenitor cell line, HepaRG, within a collagen I hydrogel. The fibrotic state of each organoid was characterized by examination of extracellular matrix (ECM) remodeling through quantitative image analysis, rheometry, and qPCR. In tandem, the phenotype of the liver progenitor cell and cluster formation was assessed through histology. Activated HSCs (aHSCs) created a more severe fibrotic state, exemplified by a more highly contracted and rigid ECM, as well higher relative expression of TGF‐β, TIMP‐1, LOXL2, and COL1A2 as compared to immortalized HSCs (LX‐2). Within the more severe fibrotic environment, generated by the aHSCs, higher Notch signaling was associated with an expansion of CK19+ cells as well as the formation of larger, more densely populated cell biliary like‐clusters as compared to mild and non‐fibrotic controls. The expansion of CK19+ cells, coupled with a severely fibrotic environment, are phenomena found within patients suffering from a variety of congenital liver disorders of the biliary tract. Thus, the model presented here can be utilized as a novel in vitro testing platform to test drugs and identify new targets that could benefit pediatric patients that suffer from the biliary dysgenesis associated with a multitude of congenital liver diseases.


| INTRODUCTION
The principle cause of pediatric liver failure, and thus transplant need, is derived from a small subset of congenital cholangiopathies that can, in some cases, manifest in utero. Physiologically, these disorders result in abnormal cellular development or increased hepatoblast proliferation, resulting in tissue malformations during critical phases of biliary tubulogenesis. 1,2 These ductal plate malformations (DPMs) cause the bile ducts to become dilated or cystic and affect their ability to maintain normal bile flow through the entire biliary tract resulting in cholestasis. 1 Major lifelong comorbidities include portal hypertension, cirrhosis, and an increased likelihood of hepatocarcinoma and cholangiocarcinoma. 3 Examples of congenital liver disease include biliary atresia (BA), congenital hepatic fibrosis (CHF), progressive familial intrahepatic cholestasis (PFIC), Caroli's syndrome (CS), and Alagille syndrome (AGS), which affect 1 in 13,000-70,000. 3,4 There are few treatment options beyond liver transplantation, 2,5 but approximately half of all total pediatric transplant need is due to BA and the consequent effects of fibrosis. 2 A commonality of many biliary tract developmental abnormalities is the prevalence of fibrosis within the liver tissue. 4 Fibrosis is the abnormal deposition of extracellular matrix (ECM) caused by a combination of ECM overproduction and the failure to effectively proteolyze these overabundant proteins ultimately resulting in reduced organ function, cirrhosis, or organ failure. Fibrosis and its mechanisms have been studied extensively in adults, but data is limited on the effects of fibrosis during liver development and congenital liver disorders.
Fibrosis is driven by specific cellular phenotype, ECM protein synthesis, and ECM remodeling. As healthy liver progresses into a fibrotic state, there is a large increase in ECM protein synthesis such as collagens I, III, IV, fibronectin, and hepatic stellate cells (HSCs) contraction, which collectively can increase tissue stiffness. 6 The increase in overall stiffness of the fibrotic liver tissue affects function and growth of resident liver cells, and can induce liver progenitor cells to become susceptible to increased Notch signaling. [6][7][8][9] Animal models of congenital liver diseases are costly, time-consuming, may not accurately depict human cellular responses, and lack ability to vary the level of fibrosis to study the resulting abnormalities. Human congenital disease models are difficult to create in vitro since they rely on multiple cell types and aberrant regulation of multiple factors that control tissue development. Most in vitro models of human disease employ human cells in 2D (monolayer) culture systems and genetic manipulation of human embryonic stem cells (hESCs) in vitro. [10][11][12] Gene editing developments in mature human cells and hESCs yielded several new in vitro models; however, most of these target a single cell type and do not replicate the intricate cell-cell interactions between multiple cell types or cell-matrix interactions with ECM proteins during tissue and organ development. In the current study, we describe a 3D coculture system (tissue organoids) composed of human liver progenitor cells and HSCs as a simple and more robust model for use as a drug discovery platform, which replicates aspects of liver fibrosis and abnormal biliary development associated with various liver developmental disorders. To our knowledge, this study is the first in vitro organoid model that compared the effects of differential fibrotic environments, induced by HSCs, on biliary progenitor cells.

| Cell culture
We isolated human primary fetal HSC according to previous reports 13,14 and passaged on tissue culture plastic in DMEM 4500 mg/L glucose, 2% (v/v) FBS. Passages 3-10 were used for experiments. The primary HSCs become activated through passaging, acquiring a highly fibrotic phenotype and referred to as activated HSCs (aHSCs). 15

| Organoid construction and culture
Three different organoids were manufactured by the incorporation of either HepaRG, HepaRG+aHSC, or HepaRG+LX-2 and encapsulated within a collagen I hydrogel. Cells in two-dimensional culture were dissociated in 0.05% Trypsin-EDTA (0.05%), phenol red (ThermoFisher Scientific ©) and neutralized by the respective maintenance media. Cells were counted via hemocytometer. Cells were dispensed into 15 mL centrifuge tubes, centrifuged at 1200 rpm for 5 min and the supernatant removed. A 1 mL/mg solution of collagen I (rat-tail) (BD Biosciences ©), PBS (ThermoFisher Scientific ©), dH 2 O, and NaOH was prepared beforehand and kept on ice. This solution was added to the cell pellet, and the cells dispersed by repeated pipetting, resulting in a solution with a cellular concentration of 5.0 × 10 6 /mL. One hundred microliters of solution was dispensed into each well of a PDMS mold (used to contain the solution until full polymerization of the hydrogel is achieved 16 ) resulting in 5.0 × 10 5 cells for each monoculture organoid whereas cocultures consisted of 2.5 × 10 5 cells of each cell type, totaling 5.0 × 10 5 cells/organoid. Polymerization was achieved by the neutralization of the acidified collagen I by NaOH and by incubation (37 C) for 25 min within a 6-well plate. One aliquot of HepaRG growth supplement (Biopredic ©) was added to 500 mL of Williams E Media resulting in a HepaRG growth medium. Three milliliters of HepaRG growth medium was then added to the wells. Organoids were incubated (37 C) and maintained by changing the media every other day. Organoids were cultured for up to 3 weeks.

| Contraction assay
Contraction assays were used to measure the HSCs' ability to contract the collagen matrix surrounding the embedded cells (i.e., the smaller the organoid, the higher the degree of contraction). All contraction appeared to take place within the first 48 hours, but out of an abundance of caution, analysis was done after 7 days. An image was taken at day 7, and contraction was calculated by image analysis. The surface area of each organoid was calculated via Adobe Photoshop and ImageJ. The magic wand tool was used to isolate the organoid from the background image. All organoids were copied as layers to a document including a ruler used as a scale bar. Images were converted to an 8-bit black and white. Organoid diameters were measured in triplicate, and the surface area calculated then were normalized to Hep-aRG monoculture in terms of percentage of contraction (N = 8).  Figure S1).

| Visiopharm image analysis
Visiopharm™ is an image analysis platform allowing for the assessment of IHC stained proteins, which can be used to infer cell morphology, counts, protein expression, and colocalization among other analysis.

| RNA isolation and qPCR
Organoids were stored in RNALater (Qiagen ©) at −20 C for RNA extraction. RNA was extracted using Fibrous Tissue RNA kit (Qiagen ©) (N = 4-6). One to two steel beads were used for each organoid in the

| Statistical analysis
Statistical analysis was performed using Matlab, Graphpad and Excel. Means were compared between respective groups. Group means were analyzed by t-test, with a confidence interval of 95%  HepaRG+aHSC organoids contracted significantly more to an average of 5.7% compared to the initial surface area.
Increased tissue stiffness, a hallmark of a fibrotic liver, is potentially due to a combination of crosslinking of collagen fibers and contraction. [23][24][25] In a previous study we found a monoculture of aHSC cells created a stiffer environment than the monoculture of LX-2 cells. 26 Here we assessed coculture organoid stiffness by rheometric analysis (Figure 1d). As expected, we observed a significant increase in tissue stiffness in organoids that contained LX-2 or aHSC cells in coculture compared with organoids containing HepaRG cells alone.

| HSCs actively remodel collagen I in the liver organoids
ECM remodeling is highly regulated and generally associated with development, wound healing, or tissue regeneration. However, when ECM remodeling becomes unregulated leads to excessive collagen production and collagen hyper-bundling, via crosslinking, yielding a denser and stiffer ECM subsequently resulting as tissue fibrosis. 27 The  (Figure 3a). HepaRG+LX-2 organoids did not show longer fibers than HepaRG organoids, but HepaRG+LX-2 organoid fibers were thicker. HepaRG+aHSC organoids showed fibers significantly longer and thicker than within HepaRG+LX-2 organoids (Figure 3b).
In a previous study we showed that a monoculture of aHSC cells created longer and thicker fibers than the monoculture of LX-2 cells. 26 Similar analysis of collagen fibers were carried out in healthy and cirrhotic liver specimens. In comparison, cirrhotic liver tissue contained fibers that were thicker, wider, and more highly aligned  HepaRG+LX-2 organoid (Figure 5b). Furthermore, the average number of cells per cluster in the HepaRG+aHSC organoid was significantly higher than HepaRG+LX-2 organoids (Figure 5c). Additionally, we compared HepaRG+aHSC organoids to HepaRG organoids, and observed larger cluster size in HepaRG+aHSC organoids (Supporting Information Figure S5). 3.6 | Effect of differential fibrotic environment on pathways associated with biliary differentiation

| Analysis of fibrosis-related pathways in HSCs
To assess the differences in HepaRG cell differentiation, we measured Notch activation and CK19 expression in HepaRG+aHSC and Hep-aRG+LX-2 organoids ( Figure 6). Notch signaling is important to biliary development generally committing hepatoblasts to a biliary lineage, which could influence the number of CK19 + cells in the organoids.
Accordingly, we hypothesized that a fibrotic environment would induce Notch signaling and activation and thus contain higher numbers of CK19 + cells. Notch activation was assessed by immunostaining for NICD and analyzed using Visiopharm™ software. NICD + cell numbers significantly decreased from week 1 to 3 in HepaRG alone, but increased in HepaRG+aHSC organoids (Figure 6a,b). In contrast, the percentage of NICD + cells remained largely unchanged in HepaRG +LX-2 organoids. To determine the subsequent effect of Notch signaling in HSCs on hepatoblast expansion, we stained sections for CK19, a marker for hepatoblasts and biliary cells, and analyzed them using Visiopharm™ software. Analysis shows the percentage of CK19 + cells decreased from week 1 to 3 in organoids containing HepaRG alone or HepaRG+LX-2 organoids (Figure 6a,c). In contrast, and similar to Notch activation results, the percentage of CK19 + cells increased from week 1 to 3 in HepaRG+aHSC organoids, and the overall percentage of CK19 + cells in HepaRG+aHSC organoids was significantly higher in these organoids (Figure 6c).

| DISCUSSION
Most models of liver fibrosis have been created to simulate conditions found within adults. 28,29 Less emphasis is placed on liver fibrosis during fetal development resulting in a lack of reliable models that simulate aspects congenital liver diseases. In this study, we used a liver progenitor cell line, HepaRG, to simulate the major cell type of the developing liver within the organoids and aHSCs to simulate the physical and chemical cues found in a fibrotic environment. HepaRGs are a hepatoblast-like cell, which differentiates into hepatocyte-like and cholangiocyte-like cells, 30  and biliary diseases studies, respectively. The Jag1dDSL/+ Notch2del1/+ mouse is a model for Alagille syndrome. 43 These models are not optimal for human-specific disease studies due to differences in liver fetal development between species. Cocultured together in a 3D environment we endeavor to more simply and robustly replicate aspects of fetal liver fibrosis.
Identifying physico-mechanical ECM properties, including stiffness and collagen fiber organization, are important for establishing quantitative measures within models that depict varied levels of fibrosis-like state. We recently published two papers using similar techniques for quantification and analysis of collagen fibers. The first was a 3D organoid colorectal cancer model that included the submucosal tissue with primary colonic SMCs and collagen I and found collagen organization was similar to the colonic ECM in vivo. 16 The second was a 3D model of liver fibrosis, which effectively replicated aspects of liver fibrosis of different severities when comparing monocultures of aHSC and LX-2 utilizing similar metrics found within this study. 26 As fibrosis increases in severity, we generally see an increase of tissue stiffness, collagen production, and TIMP1 and LOXL2 expression, which was found in this study and more extensively illustrated in our previous study where we successfully created two different fibrotic environments generated by different HSC types, one activated by in vitro sub-culture (aHSC) and the other immortalized (LX-2); one created a higher stiffness than the other, along with higher fibrosisrelated gene expression levels of COL1A2, TIMP1, and LOXL2. 26 Additionally, our aHSC constructs more extensively remodeled the ECM, can be found in more severe cases of congenital liver disorders. These results indicate that the presence of aHSCs induced higher contraction of the organoids, increased organoid stiffness, the formation of more highly bundled, longer, and thicker fibers, all of which are often associated with the abnormal anatomy of fibrotic liver tissues. 44,45 Capitalizing on these differences in fibrotic environments we have used liver progenitor cells (HepaRG), HSCs and a highly abundant liver ECM protein, collagen I in order to create 3D liver organoids. The undifferentiated HepaRG cells better resemble the developing liver compared with the popular HepG2 hepatocyte-like cells. 46 Utilizing collagen I to fabricate the liver organoids allows for self-aggregation of cells and remodeling of the ECM microenvironment within the organoids without the presence of growth factors that may influence HepaRG cell differentiation as with Matrigel™.
Recent studies tested various environmental stiffness and cellular combinations on HepaRG differentiation, but used skin fibroblasts, 47 or the effects of individual environmental toxins on luminal formation. 33 Neither of these studies utilized activated HSCs that contribute to both cellular signaling and physico-mechanical ECM properties.
In contrast, the coculture of undifferentiated HepaRG cells with the two types of HSCs described above, each capable of generating distinct fibrotic environments, makes our model more relevant for the use in drug studies of fibrosis-related biliary abnormalities. In a previous study we found that the introduction of anti-fibrotic (Alk-5 inhibitor) and pro-fibrotic (Methotrexate) drugs had significant effect on ECM remodeling, gene expression pathways associated with fibrosis and, in the case for the moderately fibrotic LX-2, reduction of overall tissue stiffness within the organoids. 26 Additionally, as these organoids represent several aspects of the fibrotic environment, specifically ECM remodeling and synthesis of soluble factors (TGF-β) associated with both fibrosis and development (Supporting Information Figure S3), addition of ALK5i effectively reduced TGF-β gene expression. 26 During embryonic development hepatoblasts invade the liver diverticulum and begin to proliferate, differentiate into biliary cells and hepatocytes, as well as begin the formation of the bile ducts orchestrated from cues within the fetal mesenchyme. 48 This high concentration of hepatoblasts diminishes over time as the liver matures due to differentiation, which can be tracked by the reduction of CK19 positive cells (a marker for hepatoblasts and biliary cells). Previously we have shown that albumin producing cell clusters are present next to ck19 positive clusters within liver organoids. 49 HSCs have an integral role in the normal development of the liver wherein they produce factors that direct development 50,51 ; without them, livers develop abnormally and having severely diminished hematopoiesis. 52,53 As HSCs are the main driver of ECM remodeling 54 and liver fibrosis, they may also have a role in abnormal liver development as fibrotic HSCs are found in high abundance in cases of biliary atresia (BA) and other cholangiopathies. 55 HSC and HepaRG coculturing lead to the formation of premature ductular structures similar to those observed with hFLPCs. 49,56,57 Over time, CK19 + structures in the HepaRG+aHSC organoids became larger and had higher number of cells per cluster, which was not found in the HepaRG+LX-2 organoids or in HepaRG alone cultures. In parallel, we also found an increase in the percentage CK19 + cells in the more fibrotic organoid created by the aHSCs. This coincides with a higher level of Notch signaling within the organoid as well. This increased Notch signaling can lead to amplified hepatoblast proliferation, resulting in ductal plate malformations, which has been observed in some cases of BA. 58 A hallmark of BA pathology is portalbased fibrosis and the presence of fibrous expansion of the portal tracks with advanced stages of fibrosis. 59 Accordingly, we propose that the differences in cluster size, expansion of CK19 + cells observed between LX-2 and aHSC containing organoids are in part a result of the more highly fibrotic environment created by the aHSCs in the liver organoids.
Collectively, these results suggest that higher levels of ECM remodeling, collagen bundling, and HSC-secreted paracrine factors, observed in the HepaRG+aHSC organoids, may have directed the HepaRG cells toward progressive expansion of CK19 + immature cells and formation larger clusters induced by higher levels of Notch signaling.

| CONCLUSIONS
Utilizing the 3D liver organoid model, we demonstrated that including aHSC in the organoids resulted in a much stiffer environment with more elongated, highly bundled, and aligned collagen fibers as compared to organoids containing the less activated LX-2 cells. Organoids containing aHSCs have higher relative COL1A2, LOXL2, and TGF-β expression, resulting in a fibrosis-associated phenotype. Either directly resulting from the fibrotic environment or indirectly, organoids containing the more highly activated aHSC cells showed a higher percent- supervision; writing-review and editing.

CONFLICT OF INTEREST
The author declares that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.