Resemblance of the human liver sinusoid in a fluidic device with biomedical and pharmaceutical applications

Abstract Maintenance of the complex phenotype of primary hepatocytes in vitro represents a limitation for developing liver support systems and reliable tools for biomedical research and drug screening. We herein aimed at developing a biosystem able to preserve human and rodent hepatocytes phenotype in vitro based on the main characteristics of the liver sinusoid: unique cellular architecture, endothelial biodynamic stimulation, and parenchymal zonation. Primary hepatocytes and liver sinusoidal endothelial cells (LSEC) were isolated from control and cirrhotic human or control rat livers and cultured in conventional in vitro platforms or within our liver‐resembling device. Hepatocytes phenotype, function, and response to hepatotoxic drugs were analyzed. Results evidenced that mimicking the in vivo sinusoidal environment within our biosystem, primary human and rat hepatocytes cocultured with functional LSEC maintained morphology and showed high albumin and urea production, enhanced cytochrome P450 family 3 subfamily A member 4 (CYP3A4) activity, and maintained expression of hepatocyte nuclear factor 4 alpha (hnf4α) and transporters, showing delayed hepatocyte dedifferentiation. In addition, differentiated hepatocytes cultured within this liver‐resembling device responded to acute treatment with known hepatotoxic drugs significantly different from those seen in conventional culture platforms. In conclusion, this study describes a new bioengineered device that mimics the human sinusoid in vitro, representing a novel method to study liver diseases and toxicology.


| INTRODUCTION
Primary hepatocytes are highly specialized cells used as the main tool for assessing hepatotoxicity, cellular transplantation, biomedical research, and as an essential component of active bioartificial devices to support liver function (Baccarani et al., 2004;Godoy et al., 2013;Nicolas et al., 2017). Nevertheless, specific functions and differentiated phenotype are progressively lost when hepatocytes are cultured in vitro, leading to loss of enzymatic activity and detoxification capacity, changes in cell morphology and function, and deregulation of transporters expression (Elaut et al., 2006;Rowe et al., 2010). Several approaches have been proposed to overcome/delay this dedifferentiation process, including sandwich cultures, spheroid systems, or the development of sinusoidal-mimicking devices known as liver-on-a-chip (Fraczek, Bolleyn, Vanhaecke, Rogiers, & Vinken, 2013;Lauschke, Hendriks, Bell, Andersson, & Ingelman-Sundberg, 2016).
Liver-on-a-chip are usually low-volume miniaturized devices that enable the culture of hepatic cells in different configurations both under flow or static conditions. In a healthy liver, hepatocyte functions are partially maintained by microenvironmental signaling from neighboring cells; for this reason, hepatocytes within these liver-resembling devices are often studied in coculture with nonparenchymal cells. Liver sinusoidal endothelial cells (LSEC), hepatic macrophages, and hepatic stellate cells constitute the major populations of nonparenchymal cells in the liver (Arias et al., 2009;Wisse et al., 1996). They play central roles both in liver physiology and pathology, and therefore cannot be ignored to generate reliable coculture systems (Marrone, Shah, & Gracia-Sancho, 2016;Usta et al., 2015), and to guarantee a greater translational capability in studies using human liver cells.
Considering the above-mentioned background, the design, development, and future applicability of a liver-on-a-chip device requires accurate selection of the hepatic cell type to be cultured, as well as the internal and external environmental stimuli that will modulate the phenotype of hosted cells (Fraczek et al., 2013).
We hypothesized that maintaining a physiological sinusoid-like environment allowing the paracrine communication between hepatocytes and functional LSEC would provide a suitable milieu for maintaining the phenotype and function of these cells, delaying hepatocyte dedifferentiation, and being more sensitive in predicting hepatotoxicity than conventional two-dimensional in vitro cultures.
To test this hypothesis, and mainly focusing on its translational applicability, the primary aim of our study was to cautiously characterize the phenotype and function of primary human hepatocytes cocultured with primary functional human LSEC within a fluidic device that mimics the hepatic sinusoid (Illa et al., 2014) and compare with conventional configurations. In addition, and as a secondary aim, we studied this liver-on-a-chip as a potential tool for preclinical research on the fields of chronic liver disease and hepatotoxicity.
Supplementary experiments using primary rat cells were performed to endorse the model in a non-human experimental scenario.

| Isolation of human and rat hepatocytes and LSEC
Human cells were isolated from remnant tissue approximately weighing 20 g obtained after human partial hepatectomy to excise tumor metastasis from colon carcinoma (for healthy cells; note that obtained peritumoral tissue was confirmed as "normal" by anatomical pathologists) and from the discarded tissue after liver transplantation (chronic ethanol etiology, for cirrhotic cells). Ethics Committee of the Hospital Clínic de Hepatocytes and LSEC were isolated using standardized protocols (Gracia-Sancho et al., 2007;Oie, Snapkov, Elvevold, Sveinbjornsson, & Smedsrod, 2016) and cultured as detailed in Supporting Information Materials. Highly pure and viable cells were used. Cell density under each individual experimental condition was 10 6 hepatocytes and 2.5·10 5 LSEC.
2.2 | Liver-on-a-chip technology and culture of primary cells Our team has recently developed a fluidic device whose detailed fabrication and features were previously described in Illa et al. (2014) and is herein termed Exoliver. Briefly, it consists of a sinusoidalmimicking layered structure that allows coculture of different cell types and fluidic stimulation of the top layer of the device. LSEC were grown in the upper area on a hydrophilic polytetrafluoroethylene microporous membrane with homogeneous and continuous shear stress stimulation, whereas hepatocytes were plated in the lower poly (methyl methacrylate) area of the device. Dynamic Exoliver configurations started with a shear stress stimulus of 0.1 dyn/cm 2 that was gradually increased during the first 2 hr of culture until reaching 1.15 dyn/cm 2 (1.5 ml/min), with a total amount of 43 ml unidirectional recirculating culture media. Exoliver, reservoir, filters, and most of the tubing were placed inside an incubator to maintain physiological conditions (37°C, 5% CO 2 ). Five different experimental configurations were considered for this study (Figure 1).
The day after the isolation, hepatocytes and LSEC were rinsed twice with the Dulbecco phosphate-buffered saline (02-023-1A; Reactiva), and media was changed to Dulbecco modified Eagle medium (DMEMF12; 11320074; Gibco) supplemented with 2.97% dextran (31392; Sigma, Darmstadt, Germany) to simulate blood viscosity, 2% fetal bovine serum Then, transwells and bioreactors were assembled and perfusion of the dynamic conditions started. Human and rat cultures were maintained for 3 or 7 days, and then disassembling of the bioreactor was performed to separately analyze both cell types. Cell supernatant analysis, quantitative polymerase chain reaction, and CYP3A4 assay were performed under all experimental conditions mentioned above, as described in Supporting Information Methods.
Once concluded that there were no significant differences in the studied markers between conventional mono-and coculture configurations, we decided to eliminate the conventional coculture condition in the 7-day human experiments to maximize cell seeding under the other conditions obtained from the scarce liver tissue available after surgery.

| Statistics and data analyses
Statistical analysis was performed with SPSS Statistics19 software for Windows. Results were expressed as mean ± standard error of mean. To assess differences between groups, we performed one-way analysis of variance with least significant difference (LSD) post-hoc tests when variables were parametric and Mann-Whitney test for nonparametric variables. Differences between groups were considered as significant when p value ≤ 0.05. Each experiment was performed in duplicate from at least n = 3 independent isolations.  Conventional configurations showed no significant differences in any of the studied parameters neither after 3 days nor 7 days of culture.

| Exoliver maintains human hepatocyte phenotype and function
Maintenance of hepatocytes phenotype using this liver-on-a-chip device was confirmed in a second species. Supporting Information Figure 2 shows all data regarding coculture of rat primary hepatocytes and LSEC.

| Exoliver prevents hepatocytes morphology deterioration
Primary human or rat hepatocytes cultured in the previously described conditions exhibited different morphology. The characteristic polygonal shape and angular edges from freshly isolated hepatocytes were gradually lost upon culture in conventional platforms. As shown in Figure 3, hepatocytes became flattened with diffuse separation between cells (Day 3), further acquiring myofibroblast-like morphology, finally leading to cell aggregation in clusters (Day 7). Prevention of the in vitro dedifferentiation process in the optimal Exoliver configuration was associated with maintenance of hepatocyte polygonal shape both after 3 and 7 days of culture. Suboptimal Exoliver configurations did not maintain hepatocyte morphology (Supporting Information Figure 3). Considering all the collected data, translational experiments of the device were performed using the optimal configuration of the device and compared with conventional cell culture method.

| Exoliver as a tool to study chronic liver disease
Primary hepatocytes isolated from human cirrhotic livers and cultured in the optimal Exoliver configuration (dynamic coculture with functional LSEC) exhibited significantly better-preserved phenotype in comparison with cells in monoculture using two-dimensional conventional methods ( Figure 4). Indeed, albumin and urea production and secretion to the culture media was significantly higher in Exoliver-cultured hepatocytes. Moreover, lower mRNA expression of the transporter abcc3 and higher mRNA expression of the transporter slc22a1 were found in hepatocytes cultured using the device, suggesting an overall maintenance of hepatocyte phenotype. F I G U R E 2 Evaluation of healthy primary human hepatocytes after 3 days (a-c) or 7 days (d-f) of culture under the experimental conditions described in Figure 1. Synthetic capacity of hepatocytes was measured as albumin and urea secretion, phase I enzymes detoxification capacity as cytochrome P450 family 3 subfamily A member 4 (CYP3A4) activity and cell phenotype markers as gene expression of the transcription factor hepatocyte nuclear factor 4 alpha (hnf4α), solute carrier family 22 member 1 (slc22a1), and ATP-binding cassette subfamily C member 3 (abcc3) transporters. Data derive from n = 4 independent experiments were normalized to conventional monoculture condition (fold change of 1) and expressed as mean ± standard error of the mean.

| DISCUSSION
The current study demonstrates for the first time that it is possible to maintain primary human hepatocytes in vitro when cocultured with functional primary LSEC in a sinusoidal-like milieu. The study has been developed using a liver-resembling device that mimics the F I G U R E 4 Assessment of cirrhotic primary human hepatocytes after 3 days of culture in conventional monoculture or dynamic coculture using Exoliver. Synthetic capacity of hepatocytes was measured as albumin and urea secretion and cell phenotype markers as gene expression of the transcription factor hnf4α, slc22a1, and abcc3. Data derive from n = 4 independent experiments were normalized to conventional monoculture condition (fold change of 1) and expressed as mean ± standard error of the mean. *p value < 0.05 versus conventional culture. ExL: Exoliver F I G U R E 3 Primary healthy human and rat hepatocytes morphology after culture for 3 or 7 days in conventional monoculture or Exoliver optimal configuration. Images were taken at 10× magnification [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 5 Exoliver-cultured hepatocytes response to acute drug-induced injury. Hepatocytes viability was assessed as urea and albumin synthesis and transaminases and LDH release to the culture media. Healthy human hepatocytes were cultured in the optimal Exoliver configuration (with LSEC) or in conventional monoculture. After 24 hr of culture, hepatocytes received acute toxic insult and were cultured for additional 24 hr with 100 μM troglitazone (a), 100 μM tolcapone (b), 1 mM diclofenac (c), or 40 mM acetaminophen (d). Cell morphology (e) and release of cell death markers (soluble keratin 18 and caspase-cleaved keratin 18) (f,g) were analyzed. Images were taken at 10x magnification. Data derived from n = 4 independent experiments were normalized to vehicle concentration (fold change of 1) and expressed as mean ± standard error of the mean. *p value < 0.05 versus its corresponding vehicle. ALT: alanine aminotransferase; AST: aspartate aminotransferase; ExL: Exoliver; K18: keratin 18; LDH: lactate dehydrogenase; LSEC: liver sinusoidal endothelial cells [Color figure can be viewed at wileyonlinelibrary.com] unique architecture of the liver sinusoid allowing layered coculture of multiple cell types with controlled endothelial shear stress stimulation and paracrine communications, as it occurs in the human liver.
We herein demonstrate that the benefits of this coculture system rely on the presence of functional LSEC. Indeed, the device benefits are mainly lost in both suboptimal Exoliver configurations: a perfused monoculture of hepatocytes or coculture of cells without biomechanical stimulation. In the first scenario, and although indirect flow stimulation per se might exert some beneficial effects on hepatocytes (Dash et al., 2013;Kang et al., 2015;Rashidi, Alhaque, Szkolnicka, Flint, & Hay, 2016), we observed that this configuration was inferior to the coculture of hepatocytes with flow-stimulated LSEC. In the second situation, hepatocytes phenotype was lost in the absence of endothelial shear stress probably due to LSEC dedifferentiation upon isolation and in vitro culture (March, Hui, Underhill, Khetani, & Bhatia, 2009).
However, our investigations demonstrate that LSEC functional phenotype can be efficiently maintained under dynamic culture (Marrone et al., 2013;Shah et al., 1997), ultimately leading to hepatocytes maintenance (Dash et al., 2013). Underlying mechanisms of such protection may derive from the fact that LSEC cultured in static configurations become rapidly dysfunctional, driving molecular signaling to hepatocytes that ultimately promote, or at least do not prevent, their dedifferentiation. In addition, functional LSEC might release soluble factors or membrane-embedded entities that contribute to maintain hepatocyte phenotype (Ding et al., 2010;Hu et al., 2014;Koch et al., 2017). In fact, upregulation of hepcidin/hamp in Exoliver cocultured hepatocytes (Supporting Information Figure 1D) supports angiocrine communication from functional LSEC. We cannot discard that future designs of the device, in which direct contact interactions between cells may be allowed as it occurs in the sinusoids, would give superior beneficial results than those herein described.
Interestingly, and most likely due to the singular design of the device, a relative gradient in oxygen along the culture area was observed (Supporting Information Figure 5A). Specialization of liver cells along the portal triad-central vein axis is known as zonation, and major drivers for such compartmentalization include nutrients, hormones, and growth factors, but specially oxygen. Because zonation directly affects macronutrient metabolism, morphology, and xenobiotic transformation in hepatocytes, oxygen gradient could indeed contribute to better reproduce the sinusoidal milieu and therefore to the maintenance of hepatocytes in the device (Kietzmann, 2017). Importantly, Exoliver-cultured hepatocytes at the inflow area were enriched in genes predominantly expressed in periportal areas of the human liver, whereas hepatocytes at the outflow predominantly expressed pericentral typical genes Considering the beneficial effects of this biosystem in preserving the phenotype of healthy human hepatocytes, we subsequently aimed at demonstrating its translational potential in two clinically relevant areas.
Data demonstrating maintenance of the phenotype of human cirrhotic hepatocytes creates a new preclinical stage to test the efficacy of novel therapeutic options for the chronic liver disease. Indeed, the device may offer highly valuable information about the effects of a certain chemical entity in a human liver-like environment just before administering it to humans. As an example, data from our team using the herein described device demonstrate that a caspase inhibitor that is currently at clinical evaluation for the treatment of chronic liver disease improves human cirrhotic hepatocytes without evidence of toxicity (Gracia-Sancho, Contreras, Vila, Garcia-Caldero, Spada, & Bosch, 2016).
Further translational studies focused on the field of drug-induced liver injury. Interestingly, Exoliver-cultured hepatocytes responded significantly different to hepatotoxic drugs than dedifferentiated cells. These data suggest that concentrations of drugs previously proposed to be hepatotoxic in vitro may not truly promote cell death when tested in functional hepatocytes. Vice versa, it is now conceivable that some drugs that were withdrawn due to toxicity in two-dimensional primary cultures could have not been harmful if tested in a more physiological environment. Although primary hepatocytes are considered the current gold standard for short-term in vitro toxicant testing, they are severely hindered by the lack of three-dimensional organization, nonparenchymal cells, nutrient access, and cell-cell interactions, which can be found in liver-on-a-chip devices. For this reason, preclinical research should consider the analysis of toxicity in physiologically resembling devices, which may provide valuable data that would complement results obtained in two-dimensional hepatocytes cultures.
We are aware that our study has limitations; probably the most important is that our device does not entirely recapitulate the diversity of cells found in the liver sinusoid. Adding extra cell layers, with hepatic stellate cells and/or macrophages, would probably increase its biological resemblance. Nevertheless, our data show that adding perfused LSEC per se is able to maintain hepatocytes phenotype, suggesting that LSEC play a major role in hepatocyte homeostasis.
It is true that future perspectives on liver bioengineering research are set in generating improved in vitro culture systems for modeling human diseases and performing valuable assays. The development of in vitro devices that address systemic human biology using liver-resembling devices in combination with other organ biosystems is highly needed (Coppeta et al., 2017;Maschmeyer et al., 2015). The herein described platform may contribute to create these body-on-a-chip structures that will ultimately allow a global understanding of prodrugs and metabolites' effects in various organs.
To sum up, our study describes a novel bioengineered device that resembles the human liver in vitro, currently representing the preclinical setup closest to the bedside. Altogether encourages its applicability for the study of liver diseases and toxicology.

ACKNOWLEDGMENTS
This study was carried out at the Esther Koplowitz Center-IDIBAPS.
The fabrication of Exoliver was performed by the platform of Production of Biomaterials and Biomolecules of the ICTS "NANBIOSIS," more specifically by the U8 Unit of the CIBER in Bioengineering, Biomaterials & Nanomedicine (CIBERBBN) at the IMB-CNM (CSIC).