Engineering of human hepatic tissue with functional vascular networks

Abstract

Although absolute organ shortage highlights the needs of alternative organ sources for regenerative medicine, the generation of a three-dimensional (3D) and complex vital organ, such as well-vascularized liver, remains a challenge. To this end, tissue engineering holds great promise; however, this approach is significantly limited by the failure of early vascularization in vivo after implantation. Here, we established a stable 3D in vitro pre-vascularization platform to generate human hepatic tissue after implantation in vivo. Human fetal liver cells (hFLCs) were mixed with human umbilical vein endothelial cells (HUVECs) and mesenchymal stem cells (hMSCs) and were implanted into a collagen/fibronectin matrix composite that was used as a 3-D carrier. After a couple of days, the fluorescent HUVECs developed premature vascular networks in vitro, which were stabilized by hMSCs. The establishment of functional vessels inside the pre-vascularized constructs was proven using dextran infusion studies after implantation under a transparency cranial window. Furthermore, dynamic morphological changes during embryonic liver cell maturation were intravitaly quantified with high-resolution confocal microscope analysis. The engineered human hepatic tissue demonstrated multiple liver-specific features, both structural and functional. Our new techniques discussed here can be implemented in future clinical uses and industrial uses, such as drug testing.

Introduction

Orthotopic cell replacement is currently the major way to generate human organs, as evidenced by the transplantation of hematopoietic stem cells or adult hepatocytes.^1,2 Repopulated human organs within immunodeficient animals hold promise for pathological modeling and industrial uses, such as drug testing; however, the uses of these strategies in the clinical setting are severely limited. Successful engraftment relies on the severe and acute damage in the recipients, the degree of which is unpredictable, difficult to reproduce, and dependent on many variable factors.³ More importantly, several clinical trials have reported the disappointing clinical outcomes of cell transplantation methods, as well as potential adverse events.^4,5

Given the limitations of cell replacement strategies, tissue engineering holds a number of possibilities for clinical use. Progress in tissue engineering has led to the creation of avascular tissue constructs, such as skin, bladders, vessels and cartilage.^6,7 However, such constructs consist of relatively simple cellular types and can be implanted without the reconstruction of the vascular supply, as nutrients and oxygen are supplied by diffusion from adjacent tissues. There still remain hurdles to creating a complex, well-vascularized organ, such as a liver. The recent promising approaches using decellularized scaffolds from adult organ are successful for the engineering of transplantable organs, including the heart, liver, and kidney.^8-10 Nevertheless, these approaches require several critical improvements, as the spatial relationships of multiple cellular types along with surrounding matrices shift rapidly from their immature states during the course of natural embryonic development.¹¹

During physiological liver development, immature hepatic cells dynamically interact with stromal supporting populations, including endothelial and mesenchymal cells, in a spatiotemporal manner.^12-14 Early endothelial and mesenchymal interactions provide an organogenic stimulus that initiates hepatogenesis in epithelial cells prior to functional blood vessel establishment.^12,13 Collectively, we hypothesized that the introduction of premature vascular networks reconstructed by endothelial and mesenchymal cells is a rational approach to the engineering of liver tissues. Herein, we report a possible tissue-engineering approach to reconstitute human hepatic tissue from embryonic liver cells cocultivated with endothelial and mesenchymal cells inside a three-dimensional collagen and fibronectin matrix composite.

Results

In vitro pre-vascularization of human fetal liver cells (hFLC)

The addition of mesenchymal stem cells (hMSCs) to the collagen/fibronectin matrix promoted the extensive branching of enhanced green fluorescence protein (EGFP)-labeled human umbilical vein endothelial cell (HUVECs) in vitro (Fig. S1A–D). The in vitro engineered constructs were implanted into immunodeficient mice. EGFP-HUVECs that were co-implanted with hMSCs initially formed long, interconnected tubes, and they subsequently connected with the host blood vessels approximately 14 d after implantation (Fig. S1E); on the other hand, vessels from HUVECs alone disappeared within 1 mo of transplantation (Fig. S1F). Histological analysis showed that endothelial networks were wrapped and stabilized by the αSMA positive perivascular cells that had differentiated from the hMSCs (Fig. S2).

Next, the EGFP-labeled human fetal liver cells (EGFP-hFLCs) were cultivated with Kusabira orange (KO)-HUVECs and hMSCs inside the collagen/fibronectin matrix (Fig. 1A). As indicated by the fluorescence microscopy imaging, after 48 h of culture, the endothelial cells started sprouting in vitro in the presence of the hFLCs when cocultured with hMSCs (Fig. 1B). The use of KO-labeled hMSCs and the non-labeled-hFLCs showed their close proximity to the EGFP-HUVECs, indicating the stabilizing effects of the hMSC-derived perivascular cells (Fig. 1C). Thus, the addition of hMSCs into the culture promoted the pre-vascularization of the human FLCs through endothelial network formation in vitro.

Figure 1. In vitro pre-vascularization of human fetal liver cells. (A) Schematic representation of our tissue engineering strategy. (B) Fluorescence microscopy images of the in vitro cultivated gels at multiple time points. The EGFP-hFLCs were cocultivated with Kusabira orange (KO)-HUVECs and non-labeled-hMSCs. The HUVECs began sprouting in vitro, similar to the HUVECs/hMSCs-derived constructs. Bars, 200 µm. (C) Localization of the human mesenchymal stem cells. Arrowhead shows perivascular distribution of MSCs in vitro. Bars, 100µm.

Implantation of murine fetal liver cells (mFLC) into a cranial window

Engineered constructs were implanted into mice bearing transparent cranial windows, which permitted the continuous observation of the implants (Fig. 1A). To validate the applicability of this model, EGFP-labeled E13.5 fetal liver cells (mFLCs) were mixed with the collagen/fibronectin matrix and were implanted into the cranial window. Intravital fluorescence microscopy imaging showed that the implanted cells successfully engrafted in vivo (Fig. 2 A, upper). Twenty days after implantation, the liver cells had expanded extensively and formed unique round-shaped cellular clusters that incorporated the functional microvascular networks from the host (Fig. 2A, lower). Dextran infusion analysis identified the connection between the recipient and transplant-derived vessel at the edge of the implants (Fig. 2B).

Figure 2. Murine hepatic tissue reconstitution from embryonic liver cells under cranial window. (A) Intravital fluorescence microscopy imaging of EGFP-labeled E13.5 murine fetal liver cell implants mixed with HUVECs and hMSCs. (B) Thirty days after implantation, tetramethylrhodamine dextran was infused via the tail vein to visualize the functional blood vessels. The images show the edge of the implants. (C and D) HE staining of the engineered murine liver tissue. The hepatic cord-like tissues contained sinusoidal endothelial cells. (E) Bile duct-like structures formed inside the clusters. The lower panel shows the expression of cytokeratins by immunostaining. Arrowhead shows the biliary epithelial cells.

The engineered constructs were harvested and histologically examined 30 d after implantation. Surprisingly, HE staining showed hepatic cord-, sinusoid-, and bile duct-like structures that are characteristics of adult liver tissue (Fig. 2C, D, and E). Cytokeratin immunostaining confirmed the bile duct-like structure formation, which generally appeared from E15.5 liver tissues (Fig. 2E).¹⁵ Liver tissue reconstitutions were enhanced by the addition of HGF and EGF, known factors that promote hepatic stem/progenitor cell expansion (data not shown).^16,17 Thus, ectopically engineered constructs shared more histological characteristics with adult liver tissues than with the original E13.5 fetal liver tissues. We concluded that our system allowed the fetal liver cell types to interact three-dimensionally and organize into mature liver tissues, thus mimicking proper developmental courses.

Live imaging analyses during human hepatic tissue formation with functional vasculatures

In vitro pre-vascularized human hepatic constructs were then implanted under the cranial window. Intravital fluorescence microscopy imaging showed robust expansion of the round hFLCs during the first several days (Fig. 3A). After 4 - 7 d, the hFLCs started to interact and tightly adhere with each other, presumably due to the formation of cell-cell junctions. After 14 d, it was hard to distinguish the single liver cells by fluorescence microscopy. High-resolution confocal microscope analysis reflected the changes in the single-liver-cell morphology into hepatocyte-like cells (Fig. 3C). Concomitantly, endothelial cells assembled to form functional vascular networks stabilized by the hMSCs-derived perivascular cells during hepatic maturation (Fig. 3B). Tetramethylrhodamine-dextran infusion studies revealed the presence of functional blood vessels throughout the engineered liver tissues at day 10 (Fig. 3D). Quantification of the functional vascular networks on day 30 revealed that the engineered liver tissue contained a denser vascular network than the HUVEC hMSC only implants, but there was no significant difference in the diameters (~12 µm) of the functional vessels (Fig. 3E and F).

Figure 3. Engineering of human hepatic tissue with functional vascular networks. (A and B) Intravital fluorescence microscopy imaging of the EGFP-hFLC/KO-HUVEC/non-labeled-hMSC (A) or non-labeled-hFLC/EGFP-HUVEC/KO-hMSC (B) derived implants. Bars, 500 (upper) and 100 (lower) µm. (C) High-resolution imaging of single-liver-cell morphology using intravital confocal microscopy. Bars, 75 µm. (D) The formation of a functional vascular network formation inside the formed tissues was confirmed using rhodamine dextran infusion via the tail vein. (E) Quantification of the functional vessel density in the SCID mice implanted with HUVECs and hMSCs only or those implanted with hFLCs, HUVECs and hMSCs in vivo. The results represent the mean ± S.D.; n = 3, *: P < 0.05. The formed human vasculature is much denser in the liver tissue than in the HUVEC/hMSC only implants. (F) The blood vessel diameter was quantified. The results represent the mean ± S.D.; n = 3.

Intravital assessment of hFLC maturation in vivo

During normal liver development, the shape of single liver cells changes from circular-shaped at the embryonic stage to cobble-stone like morphology at the mature, adult stage, as visualized with cytokeratin immunostaining (Fig. 4A, upper left). These observations suggested that intravital monitoring of the cellular morphology could be one indicator to predict the state of differentiation in vivo without tissue harvesting (Fig. 3C). To quantify the time-course dependent changes in morphology, cell circularity (form factors) was calculated using the IN Cell Investigator software (Fig. 4A, bottom).¹⁸ Form factors of the E13.5 murine liver cells were 0.833 ± 0.18 and decreased to 0.568 ± 0.16 after 8 wk in adult liver tissues. Similarly, intravital confocal microscopy images showed that the form factors of day 0 human fetal liver cells were 0.93 ± 0.07, and these had decreased to 0.512 ± 0.13 30 d after implantation, indicating the maturation of the liver cells (Fig. 4B). Consistent with these expectations, an enzyme-linked immunosorbent assay (ELISA) showed that human albumin was detected in the blood serum samples collected on day 30. Furthermore, the functionality of the human fetal liver cell (hFLC)-derived transplants was compared with that of fresh frozen human adult hepatocyte (hAH) implants, in terms of albumin production. Interestingly, albumin could be detected even 5 d after hAH implantation; however, the amount quickly diminished to 0 (zero) after 30 d. On the other hand, the albumin production capacity of the hFLC implants appeared relatively later after implantation, but the amount finally was higher than that of hAH (hFLC; 20.2 ± 7.8 ng/ml on day 30, hAH; 14.2 ± 2.9 ng/ml on day 5), whereas no albumin is detected in hFLC only transplants without HUVEC/MSC (Fig. 4C; Fig.S3). Further analyses of these implants may reveal the superiority of immature hepatic cells or progenitors for liver engineering rather than terminally differentiated mature hepatocytes.

Figure 4. Characterization of the engineered human hepatic tissue. (A and B) Intravital cellular morphological assessment by calculation of the form factors (circularity). The values of the normal liver tissues or the engineered liver tissues were determined using the IN Cell Investigator Software (GE healthcare). The bottom panels show the representative segmentations of each cell. The normal liver tissues were visualized by cytokeratin 8 and 18 immunostaining. (C) The amount of human albumin in the mouse serum. The results represent the mean ± S.D.; n = 3. (D) HE staining and immunohistochemical analyses of the engineered human liver tissue. The immunohistochemistry analysis shows the expression of albumin (hALB) and cytokeratin 8.18 (CK8.18). (E) Ultrastructural images of the hFLC-derived liver tissues at original magnification, × 47 800 and × 18 400 (from left to right). The cell-cell junctions, sinusoid-like structures and bile-canaliculi like structures are shown in each representative image. S, sinusoid; Nuc, nucleus; EC, endothelial cells; BC, bile canaliculus; MV, microvilli.

Formation of functional human hepatic tissue through the pre-vascularized hFLC implantation

Engineered constructs were harvested on day 30 and further characterized ex vivo through histochemical analysis. HE staining of the paraffin-embedded histological sections demonstrated that the hepatic cord-like structures were in line with the sinusoid-like endothelial cells, which is unique to liver histology (Fig. 4D, left). Immunohistochemical analyses showed that these hepatocytes do not express the immature marker α-fetoprotein (AFP), but do express human albumin (hALB) (Fig. 4D, upper right) and cytokeratin 8 and 18 (CK8.18) (Fig. 4D, lower right), which are known hepatocyte markers. To determine the ultrastructures of the engineered tissue, electron microscope analyses were performed. Transmission electron microscopy (TEM) images highlighted the cell-cell junctions between the mature hepatocyte-like cells and the sinusoid-like (S) structures (Fig. 4E, left). The endothelial cells were located adjacent to the hepatocytes derived from the hFLCs (Fig. 4E, middle). The presence of a bile canaliculi-like (BC) structure with microvilli (MV), which is a thin tube that collects the bile secreted by the hepatocytes, was also proven by TEM analyses (Fig. 4E, right). A portion of the transplants harbored tissue with an immature histological feature, as evidenced by the presence of proliferating hepatic cells and unclear sinusoidal structures. Therefore, it is important to uncover the cellular or molecular determinants that instruct each cell type to follow the correct fate to achieve efficient liver reconstitution. Taken together, we concluded that our tissue engineering approach is feasible for the engineering of functional and vascularized human hepatic tissues, the microstructures of which closely resembled the characteristics of adult liver tissue.

Discussion

Cell–extracellular matrix (ECM) interaction plays a fundamental role in regulating cellular processes, such as cell proliferation, migration, differentiation and survival.¹⁹ Among various natural or synthetic ECMs, we selected a collagen and fibronectin matrix composite as a three-dimensional carrier because each component of the matrix has been proven to be effective in controlling both vascular and liver cell migration, adhesion, differentiation and proliferation,^20-23 as well as exhibit relatively good mechanical properties. Combined with these matrices, we have successfully established a stable platform for creating long lasting-blood vessels using human cells consistent with previous report,²³ although engineered blood vessels have often been found to be immature and unstable.²⁴

Tissue engineering holds great promise to meet the increasing demands of organ transplantation. However, it has not yet been successfully achieved for thick and highly vascularized tissues, such as the liver. Most of the conventional approaches have three major limitations. First, the major source for tissue engineering relies on the use of adult type mature hepatocytes, which are known to drastically lose their functionality even after short-term culture. Second, the culture conditions are adjusted mainly to enhance the survival of the parenchymal cells, but not the vascular cells, thus impeding the quick formation of durable vessels after transplantation due to reduced endothelial cell viability. Third, upon transplantation, the tissue reconstitution efficiency largely depends on the degree of tissue injury to provide an optimal proliferative stimulus, most of which are not necessarily relevant to the real clinical scenarios.³

In the present study, we demonstrated the use of murine or human fetal cells cocultured with stromal cell types in engineering mature hepatic tissues after ectopic transplantation without any liver damage. We adopted a collagen and fibronectin matrix composite as a three-dimensional carrier and added mesenchymal stem cells into the culture, as they have been shown to promote the maintenance and differentiation of both immature endothelial and liver cells.^22,23,25 Notably, the human fetal liver cells were viable even after 48 h of culture, and the endothelial cells rigorously sprouted in vitro, suggesting the successful maintenance of the endothelial cells and fetal liver cells. Transplantation of pre-vascularized fetal liver cells finally reconstituted functional hepatic tissue after functional vascular reconstitution in vivo. Although these approaches are effective in engineering muscle²⁶ or bone,²⁷ few studies have been performed in other vital organs, such as liver. Thus, we here demonstrate the efficacy of pre-vascularized embryonic liver cell implantation in the engineering of human vascularized liver tissue rather than adult hepatocyte use. In future studies, it is important to determine the roles of the supporting cell types and how they act as facilitators of the expansion and differentiation of immature hepatic epithelial cells in both spatial and temporal manners. Our approach will contribute to a basis not only for pharmaceutical use but also for future clinical transplant use as an alternative to organ transplantation.

Intravital imaging is a valuable tool for characterizing dynamic cellular processes, including vascularization.^12,25 Cranial window model is the most commonly used live imaging modality in the study of brain physiology, tumor angiogenesis and microcirculation by repetitive intravital fluorescence microscopy. However, it remains unclear whether this model provides a sufficient microenvironment to assess the regenerative approaches of endoderm-derived organs, such as the liver. In this study, the cranial window transplantation approach also provided a unique intravital monitoring system for the evaluation of immature human fetal liver cell maturation and differentiation. Interestingly, we found that we could predict the state of differentiation based on the morphology of the transplanted human hepatic cells by intravital imaging combined with functional analysis, similar to that of the murine natural developmental process. Our cranial window transplant model thereby provides a unique live imaging system to study the precise spatiotemporal relationships among the multiple cellular types in vivo. This will be especially useful to study the developmental process of human embryonic sources, as they have been hardly characterized before. Further understandings in the molecular and/or cellular mechanisms underlying the organogenesis process will allow us to dissect previously uncovered knowledge of human developmental biology and will contribute to the revision of the regenerative strategy by more precisely recapitulating the physiological dynamic cellular process during organogenesis.

Materials and Methods

Murine fetal liver cell (mFLC) isolation

mFLCs were isolated from fetal C57BL/6N or CAG::EGFP(enhanced green fluorescent protein) transgenic mice at embryonic day 13.5 (E13.5) (SLC). mFLCs were mechanically dissociated by pipetting in Dulbecco's Modified Eagle's Medium (DMEM) containing 10% fetal bovine serum (FBS) (JRH Bioscience, USA). The liver cell fraction was separated from the non-parenchymal cells by several rounds of low-speed centrifugation (690 rpm at 4 °C for 1 min). After passing the dissociated cells twice through a 70 μm cell strainer (Falcon, USA), single-cell suspensions were obtained and used in the murine liver engineering studies, as described below.

Cell culture

Human fetal liver cells (hFLCs) were purchased from Applied Cell Biology Research Institute (Cat no. CS-ABI-3716) and were plated on type 4 collagen-coated 6-well plates (BD Bioscience) at a density of 30,000 cells/cm². The cells were cultured in our fresh standard medium, which is a 1:1 mixture of Dulbecco’s modified Eagle medium and F-12 (Sigma Aldrich) supplemented with 10% FBS (Lot. 7219F) (ICN Biochemical), 50 mmol/L HEPES (Wako Pure Chemical Industries), 2 mmol/L L-glutamine (Life Technologies Corporation), 50 mmol/L β-mercaptoethanol (Sigma), 1 × penicillin/streptomycin (Life Technologies), 10 mmol/L nicotinamide (Sigma), 1 × 10⁻⁴ M dexamethone (Sigma), and 1 µg/mL Insulin (Wako). Additionally, 50 ng/ml human recombinant HGF and 20 ng/ml EGF (Sigma) were added before cultivation. Human umbilical vein endothelial cells (HUVECs) and human mesenchymal stem cells (hMSCs) were purchased from Cambrex and were maintained in Endothelial Growth Medium and in MSC Growth Medium (Cambrex), respectively. The cells were maintained at 37 °C in a humidified 5% CO₂ incubator.

Retroviral transduction

For live imaging studies, the genes for enhanced green fluorescent protein (GFP) and kusabira orange (KO) were retrovirally introduced into each cell as previously described.²² Briefly, the retrovirus vector pGCDNsam IRES-EGFP or KO was transfected into the packaging cell strains (293 gp and 293 gpg cells; kindly provided by Dr. Masafumi Onodera) in which the production of the virus can be induced using a tetracycline-inducible system. Virus-bearing supernatant was harvested, passed through a 0.45-µm filter (Whatman, GE Healthcare) and was immediately used for infection.

Tissue engineered vessels and liver

For implantation, 8 x 10⁵ HUVECs with 2 × 10⁵ hMSCs, 3 × 10⁷ mFLCs with 8 x 10⁵ HUVECs and 2 × 10⁵ hMSCs, or 1 × 10⁶ hFLCs with 8 × 10⁵ HUVECs and 2 × 10⁵ hMSCs were suspended in 1 ml of rat-tail type 1 collagen (1.5 mg/ml) (BD) and human plasma fibronectin (90 mg/ml) in 25 mM HEPES (Sigma)-buffered EGM medium at 4 °C. The pH was adjusted to 7.4 using 1 N NaOH (Fisher Science, USA). The cell suspension was pipetted into 12-well plates (BD) and warmed to 37 °C for 30 min to allow the polymerization of the collagen. Each solidified gel construct was covered with 1 ml of warmed EGM medium. After one day of culture in 5% CO₂, a skin puncher was applied to create circular disk-shape pieces of the construct (4-mm diameter), and these pieces were implanted into the cranial windows of 8 wk old non-obese diabetic/severe combined immunodeficient (NOD/SCID) mice (Sankyo Lab. Co.). The mice were bred and maintained in accordance with our institutional guidelines for the use of laboratory animals. The in vivo fate of the fluorescent protein-labeled cells was tracked using intravital imaging with a fluorescence microscope (model BZ-9000; Keyence) or a Leica TCS SP5 confocal microscope (Leica Microsystems). The excitation wavelength or laser power must be adjusted to each fluorescent sample at every experimental setting. The perfused vessels were highlighted by tail vein injection of 1% tetramethyl-rhodamine-labeled dextran (MW 2 000 000; Life Technologies), indicating the formation of functional vessels.

Quantification of the perfused fraction of the engrafted vessels

The fluorescence-conjugated dextran was intravenously injected into the animals, and confocal image stacks were acquired for both the implanted vessels and the dextran staining. Image projections were then processed using the MetaMorph Angiogenesis Module software (Molecular Devices). Total tubule length, tubule % per field, and tube diameter were then automatically logged into an Excel spreadsheet.

Quantification of the engrafted liver cell morphology

Images collected from the intravital confocal microscope were processed using the IN Cell Investigator Software (GE Healthcare); the states of liver cell differentiation were classified using the “form factor,” which is a standard estimate of circularity that relates perimeter length to area. This measurement varies from 0 to 1, with 1 being a perfect circle. Combined with the results from the murine fetal (E13.5) or adult (8 w) liver cell morphologies, a high value (0.8–0.9) was observed in the immature liver cell morphology, whereas a lower value (~0.6) was observed in the mature liver cell population.

Enzyme linked immunosorbent assay (ELISA)

The collected blood sample was allowed to clot in a centrifuge tube (~5 min) at room temperature; the clot was loosened from the sides of the tube and kept at 4 °C (melting ice) for 20 min. The clotted blood was then centrifuged for 10–15 min at 400 g in a refrigerated centrifuge (4 °C), and the serum fraction was removed, taking care to exclude any red blood cells or clotted materials. The human albumin concentration in the mouse serum samples was measured using a Human Albumin ELISA Quantitation Kit (Bethyl Laboratories, Inc.) according to the manufacturer’s instructions.

Tissue processing and immunohistochemical staining

Tissue constructs were fixed overnight at 4 °C in 4% paraformaldehyde, were routinely processed and were embedded in paraffin. Transverse sections (4 µm) were placed on MAS coated slides (MATSUNAMI) for immunohistochemistry or staining with hematoxylin/eosin (HE) and periodic acid-Schiff (PAS). Immunofluorescence staining was performed with an autoclave antigen retrieval in citrate buffer (pH 6.0). The secondary antibodies were conjugated to Alexa Fluor (Life Technologies), and DAPI (Sigma) staining was used to counterstain the nucleus. The following primary antibodies were raised against human proteins and were used in this study: CD31, smooth muscle actin, AFP, CK8.18 (all from Dako), and albumin (BD). The tissue sections were incubated with secondary antibody for 1 h at room temperature. The images were taken using a LSM510 laser-scanning microscope (Carl Zeiss Co.).

Transmission electron microscopy

Samples were fixed in 2.5% glutaraldehyde (Wako Pure Chemical Corp.) and images were taken in Tokai Electron Microscopy Inc.

Statistical Analysis

The data are expressed as the mean ± SD from three independent experiments. Differences between three or four experimental groups were analyzed using the Kruskal–Wallis test by ranks, and post-hoc comparisons were made using the Mann-Whitney U test with Bonferroni’s correction. Two-tailed P values < 0.05 were considered significant.

10.4161/org.27590