| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
,1
* Department of Biological Engineering,
Center for Environmental Health Sciences, and
Department of Mechanical Engineering, MIT, Cambridge, Massachusetts, USA; and
Center for Biologic Imaging, University of Pittsburgh, Pittsburgh, Pennsylvania, USA
1Correspondence: Biological Engineering Division and Department of Mechanical Engineering, MIT, 16–429, 77 Mass. Ave., Cambridge, MA 02139, USA. E-mail: griff{at}mit.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: tissue engineering • liver • microvascular
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
20% of all liver cells and play a key role in hepatic physiology and pathophysiology (1
2
3)
. SECs scavenge modified molecules from blood, regulate facets of liver lipid uptake and virus binding, and protect liver during stress (2
, 4
5
6)
. Toxic injury of SECs can result in liver failure (7)
and is associated with tumor formation (8)
.
SECs have a distinctive phenotype characterized in part by fenestrations, low or absent surface expression of the characteristic endothelial marker CD31 (PECAM-1), and in rats, expression of the specific surface marker SE-1, which has an as yet unknown function (1
, 9
10
11)
. Although liver SECs express very low (and by some methods undetectable) levels of CD31 on the cell surface under normal conditions, their surface expression by SEC is observed in disease and can be induced in culture (11
, 12)
. Relative to endothelial cells from many other tissues, liver-derived SECs do not adapt well to in vitro culture. Their survival in culture is generally poor, and their distinctive phenotype is labile: characteristics that present challenges for studies of SEC biology (12
13
14
15
16
17
18)
. SEC lines have been established (19)
but as yet have not been shown to exhibit complete phenotypic responses in vitro and are not reported to express the SE-1marker.
Although a complete picture of the cues that regulate SEC behavior has not yet emerged, many of the factors associated with angiogenesis or with homeostasis of endothelial cells from other tissues have also been implicated in survival and function of SEC in vivo and in vitro, including vascular endothelial cell growth factor (VEGF), hepatocyte growth factor (HGF), platelet-derived growth factor (PDGF), and epidermal growth factor receptor (EGFR) ligands (12
, 18
, 20
21
22)
. Many of these factors are believed to operate in paracrine fashion. For example, VEGF is expressed by hepatocytes, stellate cells, and Kupffer cells, which are in close proximity to SEC (22
, 23)
, and SEC constitutively express VEGF receptor 1 (VEGFR-1, flt-1) and up-regulate VEGFR-1 and VEGFR-2 (flk1, KDR) during liver regeneration (22)
and under laminar flow shear stress in vitro (20)
. Autocrine stimulation has been less well explored for SEC, but EGFR ligands are known to act in autocrine fashion in some endothelial cells (24)
.
In serum-free, hormonally defined media, purified rat SECs plated on extracellular matrix survive less than a week, even when plated at high cell density, and lose fenestrae within a few days (13
, 14)
. Significant apoptosis of purified SEC is observed soon after plating in vitro even in the presence of serum and VEGF (17)
. Coculture with primary hepatocytes, an apparent source of VEGF and other paracrine factors important for SEC survival, improves retention of some phenotypic behaviors (12
, 25)
, but SECs still fail to persist much longer than a week in coculture (16)
. Hepatocytes lose many features of liver-specific gene expression in culture (26)
, and it is possible that factors regulating SEC behavior may be among them; alternatively, the lack of certain nonparenchymal cells in primary hepatocyte cultures, or alteration of their phenotypes, may contribute to paracrine environments that are not conducive to SEC survival and function.
The composition and physical properties of extracellular matrix (ECM) work in concert with soluble factors to direct cell phenotype. In general, most cells show increased proliferation and decreased differentiation on stiff 2D matrices and behave more physiologically on or in 3D gels or in 3D cell aggregates (27)
, and both hepatocytes and stellate cell behavior in vitro are influenced by matrix stiffness (28
, 29)
. Under normal physiological conditions, only a wispy basement membrane exists between endothelial cells and hepatocytes. Many chronic liver diseases cause a significant increase in the amount and stiffness of ECM within the liver parenchyma and alter its composition, changes that in turn are associated with loss of sinusoidal endothelial fenestrae and development of capillary-like features (12
, 28
, 30
, 31)
.
In culture formats that foster 3D tissue formation from primary liver cell isolates, a condition that might be expected to mimic certain features of both the paracrine signaling and ECM environment, cells that stain positive for general endothelial markers are observed up to several weeks after cultures are initiated, even from cultures that are enriched for hepatocytes (16
, 32
33
34
35)
. However, the presence of SE-1-positive cells in long-term cultures has rarely been reported. Supporting the idea that the combination of paracrine factors and mechanical/matrix properties in 3D may be conducive to SEC survival and function, Harada and coworkers (32)
observed SE-1-positive cells when colonies of small hepatocytes and nonparenchymal cells (NPCs), obtained after 3 wk of 2D culture in medium containing serum and DMSO, were cultured for an additional 3 wk in collagen sponges, a condition that resulted in enhanced expression of hepatocyte functions and appearance of multiple cell types. DMSO in general fosters retention of hepatic function in primary hepatocyte cultures (36)
and differentiated function in cultures of other cell types (37)
, and improvements in hepatocyte function in vitro are also often reported in various 3D culture configurations initiated with primary hepatocyte isolates (26
, 38
39
40
41
42
43)
. These factors together may foster local environments favorable for maintaining SE-1-positive cells.
An open question is whether 3D cocultures of freshly isolated SECs with standard hepatocyte isolates (which contain a small fraction of associated NPCs) can foster survival and phenotypic behavior (as characterized by expression of SE-1) in long-term culture in the absence of serum and DMSO. To address this question, we employed a microreactor culture system designed to foster organization of primary liver cells into tissue-like units (300 µm on each side) and to perfuse each tissue with microscale flow, while allowing in situ imaging (Fig. 1
; refs. 26
, 44
). Local controlled perfusion in 3D cultures is an important means to distribute oxygen, nutrients, and growth factors through the tissue, and flow is becoming increasingly appreciated as a regulator of microvascular endothelial function (24
, 45)
. We have previously found that many liver-enriched programs of gene expression are maintained at approximately physiological values when the hepatocyte-rich fraction is cultured in this system (26)
. We find that the 3D coculture format appears to be permissive for long-term (12 days) survival of SE-1-positive cells in the absence of serum and DMSO.
|
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
was used with the following modifications (26
, 47)
: 0.305 g/l niacinamide (Sigma-Aldrich, St. Louis, MO, USA); 2.25 g/l glucose (Sigma-Aldrich); 1 mM L-glutamine (Sigma-Aldrich); 0.0544 mg/l ZnCl2 (Sigma-Aldrich); 0.0750 mg/l ZnSO47H2O (Sigma-Aldrich); 0.020 mg/l CuSO45H2O (Sigma-Aldrich); 0.025 mg/l MnSO4 (Sigma-Aldrich); 20 ng/ml EGF (Collaborative, Bedford, MA); and 0 ng/ml HGF. Endothelial growth medium (EGM-2 bullet-kit, Cambrex, Walkersville, MD, USA) was mixed with HGM 1:1 (v/v) to create "HEGM" for the initial 6 days of culture.
Cell isolation and characterization
Cells were isolated from 150 to 230 g male Fischer rats by a two-step collagenase perfusion (26)
. A 25 ml/min flow rate was used for isolation of hepatocytes, and separate isolations with a 15 ml/min flow rate were used for isolation of the endothelial cell fraction. Hepatocyte viability was 90–95% (trypan blue, Life Technologies, Carlsbad, CA, USA). The supernatant material from the first two centrifugation steps of the cell isolate was used to isolate endothelial cells at room temperature (13)
. The small hepatocytes were eliminated from the supernatant with a 5 min 100 g centrifugation step. All nonparenchymal cells in the remaining supernatant were obtained after a 7 min 350 g centrifugation. These cells were resuspended in 20 ml of PBS at room temperature and loaded onto 25%/50% Percoll (Sigma-Aldrich)/PBS step-gradient columns in 50 ml conical tubes (BD Biosciences, San Jose, CA, USA). After 20 min of 900 g centrifugation, cells located at the interface between 25 and 50% Percoll/PBS were collected and washed with fresh PBS. The cell pellet after a 10 min 350 g centrifugation step was used as the endothelial cell fraction. The enhanced green fluorescent protein (EGFP)-endothelial cell fraction was obtained from separate cell isolations from 150–250 g female EGFP-positive Sprague-Dawley rats, which were bred from EGFP-positive males that were a generous gift of M. Okabe (48
, 49)
. The relative numbers of SECs, stellate cells, and Kupffer cells in the endothelial cell isolates were assessed by analyzing aliquots by FACS and by immunostaining cells plated for 2 h on collagen I (Cohesion, Palo Alto, CA, USA).
For immunostaining, adherent cells were fixed in 2% paraformaldehyde (EMS, Hatfield, PA, USA) and stained with the following antibodies: anti-GFAP (Chemicon, Temecula, CA, USA), anti-CD163/ED-2 (Serotec, Oxford, UK), SE-1 (IBL America, Minneapolis, MN, USA), and anti-CD31/PECAM-1 (Chemicon), followed by Cy3 secondary goat antibodies (Jackson ImmunoResearch, West Grove, PA, USA). Nuclei of plated cells were marked with Hoechst 33342 (Molecular Probes, Carlsbad, CA, USA), and images were taken in four randomly selected areas per well and imported into Metamorph (Universal Imaging, Downingtown, PA, USA). For FACS analysis, cells from liver isolates were pelleted at 3000 rpm, washed once with PBS + 0.1% Tween-20 (PBS-T), pelleted again, and fixed in PBS containing 2% paraformaldehyde for 20 min at 25°C. Cells were then washed in PBS-T, pelleted, and fixed and permeabilized in methanol at –20°C. Fixed cells were labeled using the following primary antibodies: rabbit anti-albumin (Cappel, Solon, OH, USA), mouse anti-cytokeratin 18 (Sigma-Aldrich), mouse anti-CD163/ED2 (Serotec), mouse anti-GFAP (BD), anti-SE-1 (IBL America), or anti-CD31/PECAM-1 (Chemicon) for 1 h at –25°C in PBS-T and then goat anti-rabbit AlexaFluor488-conjugate (Invitrogen, Carlsbad, CA, USA) and goat anti-mouse AlexaFluor647-conjugate (Invitrogen) secondary antibodies for 1 h at –25°C in PBS-T. Labeled cells were analyzed using a FACS Calibur flow cytometer (BD Biosciences) and FlowJo software (TreeStar, Ashland, OR, USA).
For four different isolations, adherent Cy3-positive cells were counted and divided by total nuclei to yield percent positive of each cell type: 9–12% CD31-positive (endothelial cells other than sinusoidal), 3–5% GFAP-positive (stellate cells), and 10% ED2-positive (Kupffer cells). FACS analysis, which included anti-SE-1 (SECs), anti-CK18 (hepatocytes), and anti-rat albumin, gave similar results: 74% SE-1 positive, 2–10% GFAP positive, 10% ED-2 positive, 9% CK18 positive, and 9% albumin positive. FACS analysis of the isolated hepatocyte population comprised the following distributions: 95–99% CK18 /albumin positive, 0.5–0.9% ED-2 positive, 0.2–1.5% GFAP positive, 0.3–0.5% SE-1-positive, and 0.1% CD31 positive (n=1 for CD31).
Microreactor seeding and maintenance
Spheroids were formed by seeding 500 ml spinner flasks (Bellco Glass, Vineland, NJ, USA) containing 100 ml of HEGM with 20 x 106 hepatocytes (mono-culture) or 20 x 106 hepatocytes plus 60 x 106 total cells from the liver endothelial cell fraction (coculture) and stirring at 85 rpm for 24 h at 8.5% CO2 and 37°C. Spheroids were filtered sequentially through 300 and 100 µm nylon meshes (SEFAR, Depew, NY, USA), resuspended in 25 ml HEGM, centrifuged at 40 g for 2 min, and resuspended in 30 ml HEGM for seeding.
Microreactors (Fig. 1)
were operated as described previously (44
, 47)
using either silicon (for 2-photon imaging) or polycarbonate scaffolds (for histology) of identical geometries. Scaffolds were coated with collagen I (Cohesion) as described previously (44)
, and comparable tissue morphologies were observed with either scaffold material. The system was primed with HEGM (1 h, 37°C). Spheroids were introduced by syringe, and then axial flow was maintained at 0.5 ml/min. Crossflow through the scaffold (40 µl/min) was directed toward the filter for the first 24 h, then the culture medium was changed, an inline double filter of 0.8/0.2 µm size (Pall, East Hills, NY, USA) was installed on the crossflow line, and the flow direction was permanently reversed to flush debris. The medium and the inline filter were changed every 3 days thereafter. For all cultures, HEGM was used for the first 6 days and HGM thereafter (i.e., additional 7 days).
2D collagen gel sandwich
Collagen gel sandwich cultures ("2D" cultures) were prepared in 6-well plates (BD Biosciences) using 600 µl collagen solution (Cohesion) for the lower layer (gelled overnight at 37°C) and 300 µl for the upper layer, which was added 4 h after cell seeding and allowed to gel for 1 h before addition of culture medium (26)
. Cells were seeded at 50,000 total cells/cm2 using the same cell population as the microreactor cultures: mono-cultures were initiated with only the hepatocyte fraction; cocultures comprised a 3:1 mix of endothelial cell fraction: hepatocyte fraction. Medium was changed every other day. HEGM was used for the first 6 days, followed by HGM. Phase-contrast and green fluorescence images of the cultures were taken with the software Openlab (Improvision, Lexington, MA, USA). These images were imported into Metamorph (Universal Imaging), and their intensity values were extracted and overlapped to produce the final composite images.
Scanning electron micrograph and Image Analysis
Microreactor tissue was processed for acanning electron micrograph (SEM) imaging according to previously published protocols (50
, 51)
and viewed under a JEOL JSM-5600 LV scanning electron microscope. Features comprising pores of 5–10 µm in diameter in each image of a single channel were quantified in Metamorph (Universal Imaging). Images were taken from experiments using three separate perfusions with two microreactors in each group per perfusion. A total of 56 images was counted.
In situ Two-photon Microscopy and Image Analysis
A protocol similar to that described previously was used (44)
. The distance between the objective and the microreactor was regulated with a MIPO 500 piezoelectric driver (Piezosystem Jena, Hopedale, MA, USA) with a range of 400 µm. A Zeiss Water Achroplan 20x objective was used to accommodate long working distance between the viewing window and the top of the tissue mass (Fig. 1)
to maximize the viewing area without compromising resolution. A microscope environmental chamber was constructed to maintain cultures in humidified air with 8.5% CO2 and at 37°C during imaging.
Two microreactors were imaged per experiment. Four channels were randomly selected on day 1 postseeding from each microreactor and observed on days 2, 6, and 11 postperfusion. For each channel imaged, the piezo was initially brought to its zero position and the objective was allowed to focus on the microreactor filter. Between scanning each image, the piezo was instructed to move at 3 µm per step in the z-direction, for a total of 61 images per channel. For consistency, all samples were excited by the same laser power at 100 mW before entering the microscope. These images were imported into Metamorph for low-pass filtering noise reduction and image analysis.
In Metamorph, raw image data without filtering were first subjected to a universal threshold to define EGFP-positive objects before morphometry analysis. These data were read by a Matlab v6.5 script file that counted the total number of pixels that were above threshold value. This number was divided by the total number of pixels in each image to yield percent fluorescent pixels. All images for one channel of one microreactor on one time point were averaged to yield percentage of fluorescent pixels.
For 3D reconstruction, low-pass filtered images of 13 day postperfusion coculture microreactors were imported into Imaris (Bitplane, St. Paul, MN, USA). Avi movies and 3D stereoview image snapshots were taken.
RNA and DNA isolation
One milliliter of Trizol (Invitrogen) was added directly to each well in 2D cultures and each microreactor sample and kept at –80°C until ready for RNA and DNA isolation, which was performed according to the manufacturer’s protocols. RNA from the aqueous phase was purified using the RNeasy mini kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. The concentration and quality of purified RNA were determined by assessing the ratio of absorbance at 260–280 nm, and only samples with a ratio within the range 1.7–2.1 were used. The RNA was stored at –80°C. The interfacial and chloroform phases remaining after RNA extraction into the aqueous phase were used to isolate genomic DNA using the manufacturer’s protocol.
Quantification of EGFP Cell Percentages in Co-Cultures
The fraction of EGFP cells present when the female EGFP liver endothelial cell fraction was cocultured with nonfluorescent male Fischer rat-derived hepatocytes was determined by the ratio of genomic GAPDH to Y-chromosome (Sry). A calibration curve was constructed by mixing a defined number of cells from the EGFP liver endothelial cell fraction (cultured on tissue culture treated 6-well plates with EGM for 1 day) with freshly isolated Fischer hepatocytes to achieve a range of ratios between 0 and 90% female cells. Genomic DNA was isolated from these samples and quantitative PCR was performed to obtain ratio of Sry:GAPDH, using the SYBR Green kit (Qiagen) according to the manufacturer’s instructions. With the use of this calibration curve, the percentage of female cells was determined on samples of 1-day old co- and mono-culture spheroids, 13 day postperfusion co- and mono-culture microreactors, and 13-day postperfusion co- and mono-culture 2D collagen gel sandwich cultures. Primers (salt-free purity from Qiagen Operon) were Sry: (forward) 5'-GCCTCCTGGAAAAGGGCC-3', (reverse) 5'-GAGAGAGGCACAAGTTGGC-3'; GAPDH: (forward) 5'-GTGGTGCAGGATGCATTGCTGA-3', (reverse) 5'-ATGCTGGTGCTGAGTATGTCG-3'.
Immunohistochemistry
On day 13, microreactors were perfused with 2% paraformaldehyde in PBS for 1 h, washed in 6.8% sucrose in PBS overnight and dried in 100% ethanol for 1 h before embedding in Technovit8100 (EMS) according to manufacturer’s instructions. Scaffold pieces were removed under a microtome, and embedded tissue samples were re-embedded in Technovit8100. Samples were cut into 4-micron sections and stained (52)
. Primary antibodies included mouse-antirat SE-1 (IBL America), anti-CD31 (Chemicon), ED2 (Serotec), anti-GFAP (Serotec), and anti-SMA (Sigma). Goat-anti-mouse biotin-conjugated antibodies and strepavidin-conjugated Cy3 were used as secondary and tertiary antibodies (Jackson Immunoresearch).
Global Transcriptional Profiling and Analysis of Expression Data
cRNA preparation
Total RNA was isolated with RNeasy kit (Qiagen). cRNA was synthesized at the MIT BioMicro Center as follows. Poly(A)+ mRNA (2 µg) was converted into single-stranded cDNA using a modified oligo(dT) primer with a 5' T7 RNA polymerase promoter sequence and reverse transcriptase. Double-stranded cDNA was generated using DNA polymerase and DNA ligase. Biotin-labeled cRNA was generated using in vitro transcription with T7 RNA polymerase. cRNA was quantified at UV260, and 15 µg of RNA were fragmented randomly using (200 mM Tris-acetate, 500 mM KOAc, and 150 mM MgOAc) at 94°C for 35 min.
GenChip hybridizations
Five biological replicates of 3D mono-culture microreactor, four biological replicates of 2D mono-culture collagen-gel sandwich, and two biological replicates of in vivo liver samples were hybridized to Rat GeneChip arrays (Affymetrix, Santa Clara, CA, USA). Hybridizations and scanning were carried out at the MIT BioMicro Center using the following procedure: fragmented cRNA at a concentration of 0.05 µg/µl was hybridized to GeneChip in 200 L of Affy buffer (100 mM MES, 1 M NaCl, 20 mM EDTA, and 0.01% Tween 20) with GeneChip eukaryotic hybridization controls (GeneChip eukaryotic hybridization controls kit) in the presence of 0.1 mg/ml herring sperm DNA and 0.5 mg/ml acetylated BSA at 40°C for 16 h with constant rotation. Arrays were rinsed after hybridization with 200 µl of stringent wash buffer (100 mM MES, 0.1 M NaCl, and 0.01% Tween 20) followed by a nonstringent wash (6x SSPE and 0.01% Tween 20); 20x SSPE had the following composition: 3 M NaCl, 0.2 M NaH2PO4, and 0.02 M EDTA. Staining was done with 2 µg/ml streptavidin-phycoerytherin and 1 mg/ml acetylated BSA in 6x SSPE-T. Arrays were scanned using an Affymetrix GeneChip Scanner 3000.
Image and data analysis
To enable direct GeneChip comparisons, data were normalized using Robust Multi-Chip Average (RMA; ref 53
). Fold changes of 3D mono-culture to 2D mono-culture reactor were calculated by dividing the average intensity values of the experimental samples (3D mono culture reactor) by the average of the reference control samples (2D mono-culture collagen gel sandwich). To identify biologically significant and statistically significant differential gene expression, an ANOVA statistic was applied (P<0.05) between groups with the additional metric of significant difference in the expression of a gene where fold change between 3D and 2D were 
1.5 or –1.5. Biological network and pathway analysis were carried out using Ingenuity software tool (http://www.ingenuity.com/). For Ingenuity network and pathway analysis, the identifiers of the significantly up- or down-regulated genes with their corresponding fold changes were integrated into the Ingenuity software database. The differentially expressed genes were mapped onto the corresponding gene objects in the Ingenuity database, and P values of enrichment of differentially expressed genes within protein interactomes were defined.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
10% of the cell population. The level of green fluorescence expressed by this SE-1cell-enriched population from EGFP rats was much brighter than the autofluorescence from wild-type (Fischer rat) hepatocytes (Fig. 2
), thus allowing a high signal-to-noise ratio in 3D imaging experiments.
|
For live, noninvasive imaging, we used multiphoton microscopy, which uses long-wavelength excitation light that penetrates deeply and excites only a tiny volume, preserving tissue viability (44
, 54)
. Penetration depth is particularly critical here, as the continuous perfusion of the tissue requires a distance of several hundred microns between the objective and the tissue (Fig. 1)
.
Sequential 3D images of EGFP-expressing cells in microreactor culture revealed an evolution of organization from an initial dispersed state to networks over approximately a week (Fig. 3
). On day 2 postperfusion, the EGFP-expressing cells had a mostly rounded morphology and were distributed seemingly randomly among the cell aggregates. By day 6, a substantial fraction of the EGFP cells was more elongated and EGFP cells appeared to be more numerous. Both these trends were accentuated by day 11 (Fig. 3)
. The EGFP cell shape was quantitatively determined by morphometry analysis to be elongating over time (see Supplemental Data File 1). Furthermore, 3D reconstruction images showed connected networks of EGFP cells spanning the depth of the coculture microreactor channels on day 13 postperfusion (static stereo images are shown in Fig. 4
; rotational movies are shown in Supplemental Data Files 2–6).
|
|
Parallel to the two-photon experiment, 2D collagen gel sandwich cultures were established using identical culture media and the same starting cell populations: mono-cultures were established with only the hepatocyte fraction without any deliberate addition of nonparenchymal cells; cocultures were initiated with a 1:3 mix of hepatocyte fraction and SE-1-positive cell enriched cell fraction. These cultures were observed with an epifluorescence microscope on the same days as the microreactor multiphoton experiment (Fig. 5
). On day 2, there were many EGFP cells scattered around hepatocyte islands (Fig. 5A
). Many of these cells attached to the outer perimeter of hepatocyte islands, similar to what was reported in Matrigel cocultures (16)
. By day 6, fewer EGFP cells were seen (Fig. 5B-C
) and the EGFP cells were almost absent by day 11 (Fig. 5D
). Some of the surviving EGFP cells showed dendritic cell processes suggestive of stellate cell morphology (Fig. 5C
). The morphology of hepatocytes in 2D collagen gel sandwich mono-cultures (Fig. 5E-H
) was similar to the hepatocyte morphology in 2D endothelial cocultures. Cells with the morphology of stellate cells were also observed in mono-cultures at longer times (Fig. 5G
). Overall the EGFP cell population was not well maintained in 2D cocultures.
|
SEM images of mono-cultures maintained 3 days in perfusion culture (Fig. 6
) showed cell-cell contact and large (>10 µm) fluid duct structures, corresponding to previous findings (44)
. Both mono-cultures and cocultures showed cell surface microvilli typical of hepatocytes, and both cultures exhibited numerous small pore structures ranging 5–10 µm in diameter (Fig. 6)
. These tiny pore structures appearing on the surface were 3 times more numerous in cocultures than in mono-cultures (nmicrovessel=19.5±2.1 per channel for cocultures and nmicrovessel=7.3±5.4 per channel for mono-cultures; mean±SD). Occasionally (i.e., about one in 10 channels), a smoothly wrinkled cell, distinctly different in appearance from hepatocytes, is observed around the opening of one of these tiny vessel-like structures (Fig. 6D
).
|
Because only the EGFP cells emitted green fluorescence above a threshold level, the percentage of fluorescent pixels per image was used as a surrogate measure of cell number. The pixel count confirmed that the number of EGFP cells increased over 10 days of perfused microreactor culture (Fig. 7
A). We confirmed this finding using quantitative PCR analysis of genomic DNA (ratio of Sry to genomic GADPH). The percentage of EGFP cells in 3D microreactor culture increased from an initial level of 7.5% (i.e., 7.5% of cells present in spheroids after the 24 h aggregation process were EGFP-derived) to 12% over 13 days in culture while control mono-cultures had no EGFP (female) cells present at day 13 (Fig. 7B
) Thus, 3D perfused microreactor culture fosters retention and proliferation of the EGFP cells, while the 2D collagen gel sandwich culture does not support these behaviors. Further, the EGFP cell population enriched in SE-1-positive cells did not survive without the hepatocytes in the microreactors (data not shown), although microreactor culture fostered attachment and survival of rat lung microvessel endothelial cells on the scaffold (data not shown). Finally, cells did not survive in microreactor culture in the absence of perfusion flow (data not shown).
|
SE-1-positive cells persist in microreactor culture
The change in culture medium at day 6 from a medium containing serum, VEGF, and FGF to a medium lacking these factors might lead to loss of SECs, which have a strict requirement for VEGF. Because several cell types were initially present in the EGFP cell population used in cocultures, we used immunostaining to investigate the cell composition in the tissue structures on day 13. Using the SE-1 antibody against SECs, we found them present throughout the interior of the tissue in the coculture but largely absent in mono-culture (Fig. 8
A–B). The SE-1 stain does not appear strongly colocalized with EGFP expression in the in vitro cultures; however, a similar lack of colocalization of SE-1 and EGFP is seen in vivo, where SE-1 stains sinusoidal surfaces (Fig. 8C
).
|
Cells positive for CD31 staining were present in the coculture, mainly at tissue-fluid interface, but were not observed in mono-culture (Fig. 8D-E
). In vivo, CD31 is found strongly expressed on large vessel endothelium in liver (Fig. 8F
). Thus, both sinusoidal and large vessel endothelial cells were present in the cocultures and located in physiologically appropriate regions of the tissue (i.e., small vessel within the tissue and large vessel at the tissue interface with the fluid). Kupffer cells were present in both groups but appeared to be more numerous in the coculture (Fig. 8G-H
). Quiescent stellate cells were marked by GFAP staining in both groups at fluid-tissue interface (Fig. 9
A–B). Activated stellate cells were seen occasionally along the walls of the channels but were not prominent in the tissue (Fig. 9C-D
).
|
Global transcriptional profiling and pathway mapping identified differential expression of endothelial-related cytokine networks and metabolism networks in 3D compared to 2D mono-culture
To illuminate possible molecular mechanisms that foster persistence of SE-1-positive cells in 3D, we used comparative expression profiling to establish genome-wide differences between the culture modalities (see Materials and Methods). We compared the behavior of 2D and 3D mono-cultures to assess whether intrinsic differences between the host environments for SE-1-positive cells might exist. After eliminating probesets that are unknown or poorly characterized, we found 950 probesets representing 563 unique genes that are differentially expressed between the two systems (full list is in Supplemental Data File 7). Most of the canonical growth factors associated with angiogenesis or homeostasis of endothelial cell function, including VEGF isoforms, FGF isoforms, angiopoietins, and EGFR ligands, did not show significant expression differences between 3D and 2D mono-culture. Most of the genes that were differentially regulated between culture systems are not known to be related to endothelial cell function in ways that have been reported in the literature. However, at least 14 differentially expressed genes/probesets with some relevance to endothelial cell behavior can be identified, and these genes along with short descriptions of their relevance, with appropriate literature references are provided in Table 1
. Notably, several transcripts are related to transforming growth factor-(TGF-ß) signaling, including increased expression in 3D cultures of TGF-ß3 and increased expression in 3D cultures of 
2 macroglobulin, which binds TGF-ß to maintain latency. The three TGF-ß family members and their cognate receptors are implicated in multiple facets of angiogenesis, exerting context-dependent effects on endothelial cells that may be either proangiogenic or antiangiogenic (55
, 56)
. 3D cultures also showed a greatly enhanced expression of transferrin, an iron-carrying protein that has been implicated in endothelial cell migration and angiogenesis. Enhanced expression of metallothonein in 3D (where it is up-regulated 1.7-fold compared to in vivo) compared to 2D (where it is strongly down-regulated compared to in vivo) may also create a local protection against oxidative stress.
|
In an effort to understand molecular pathways that are differentially modulated between the systems, we integrated the genes that were significantly changed between culture systems with their known gene products and looked for an enrichment of molecular interactions (see Materials and Methods). Network analysis identified significant (P<10–30) interactions among 12 of the 14 endothelial related proteins described in Table 1
(Fig. 10
).
|
Analysis of the complete list of transcripts modulated between 3D and 2D mono-culture revealed several significant interactomes (See Supplemental Data File 8 for a full list). Notably, a network involving drug metabolism, lipid metabolism, and molecular transport was identified as containing highly significant interactions (P<10–17), suggesting that general maintenance of hepatocyte function may also contribute to survival of endothelial cells (see Fig. 11
). We have previously found that many of the important transcription factors that regulate programs of liver-enriched genes, including HNF4, as well as many drug metabolizing enzymes (e.g., CYP3A), are maintained at physiological levels in hepatocyte mono-cultures in the microreactor, whereas they are down-regulated in 2D culture as assessed by RT-PCR, Western blots, and functional assays for metabolism (26)
. These previous findings were confirmed and extended by our studies here that show that CYPs, and phase II metabolism genes are more highly expressed in 3D that in 2D (Fig. 11)
. Additionally, one of the three most highly significant networks (P<10–51) arising from analysis of the complete list of transcripts involved Fos-related proteins (see Supplemental Data File 9), including metallothionein. Both Fos and metallothionein also participated in the significant endothelial-related network (Fig. 10)
.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
VEGF is an essential survival, growth, and differentiation factor for endothelial cells in general and for primary liver SECs in culture in particular (2
, 17)
. Hepatocytes are a major paracrine source of VEGF in vivo (5
, 22
, 23)
, and conditioned medium from primary hepatocyte cultures supports survival and function of SEC in a manner that is abrogated by antibodies to VEGF (12)
. Differences in the average values of VEGF expression, as assessed by global transcriptional profiling, do not explain the differences in SE-1-positive cell survival in 2D and 3D, as we found no significant changes in the expression levels of VEGF-A or VEGF-C gene probes between 2D and 3D cultures, and expression levels were comparable to in vivo (data not shown). However, our global gene expression analysis revealed several unexpected molecules (Table 1)
that may have provided protection for and stimulated the survival of the SECs in the 3D microreactor. In particular, metallothionein and its regulator Fos are both implicated in one of the most highly significant molecular interaction networks.
Additional environmental cues that work synergistically with VEGF to maintain SEC survival and phenotype are not well understood. Cultures initiated with purified SECs are reported to die or be overgrown by contaminating cell types within 8–10 days, even in the presence of VEGF (18)
. Addition of hepatocyte-conditioned medium to SEC culture modestly improves survival but stimulates overgrowth by stellate cells after about a week of culture (14)
. The immediate decline of highly pure cell populations prepared using the SE-1 antibody can be inhibited by addition of orthovandate (17)
, a general inhibitor of protein phosphotyrosyl phosphatases. This treatment results in an increase in intracellular phosphotyrosine levels and thus presumably mimics stimulation by unknown prosurvival signals, but this treatment does not prolong survival beyond about a week in vitro.
Although many of the known paracrine factors that regulate SEC survival emanate from hepatocytes, coculture with hepatocytes in the presence of VEGF is still not sufficient to ensure survival of primary SECs isolated with a protocol similar to that described herein (16)
. Primary liver SECs, as well as primary endothelial cells derived from dermal microvessels or umbilical vein, form tubes on matrigel. Hepatocytes added to endothelial tube cultures migrate toward vessels regardless of the tissue of origin, but only dermal endothelial cells exhibit prolonged (weeks) survival in coculture; the HUVEC and liver endothelial cell microvessels in these cocultures collapse and disappear within about a week (16)
. Endothelial cells from many tissues can be induced to form stable microvessel networks in culture the presence of fibroblasts (73
, 74)
and indeed the survival of dermal endothelial cells in the matrigel coculture model was attributed in part to the presence of supporting fibroblast-like cells contaminating the dermal microvessel endothelial cell population (16)
. Because matrigel fosters a high degree of differentiated function in hepatocytes, it seems probable that paracrine factors from cells other than mature hepatocytes might be needed to maintain liver SECs in culture. It is possible that stellate cells in our cultures may contribute to stabilization of SE-1-positive endothelial cells. Stellate cells were present as a minor contaminant of both hepatocyte and liver endothelial cell isolates used here (at least 0.2–1.5 and 2–10% respectively in this work, based on the FACS analysis using anti-GFAP antibodies), and quiescent stellate cells were observed throughout the tissue structures formed after 13 days in either 3D mono-culture or coculture.
However, other nonhepatocyte cells may contribute as well. In the EGFP-positive endothelial cell population, we used to initiate our cocultures, 25% did not stain positive for the sinusoidal marker SE-1; furthermore, a significant fraction of the EGFP-positive cells observed in histological sections after 13 days in culture were not stained with SE-1-postitive (or CD31). Although we can likely attribute some of the EGFP-positive, SE-1-negative cell population to Kupffer cells and stellate cells, additional cell types including hepatocyte progenitors, which can be activated during certain liver injury processes may be present (75)
.
Liver-derived endothelial cells have been observed in cultures of rat and porcine primary hepatocytes maintained for several weeks in 3D culture in medium without added serum or VEGF, although whether these cells represent phenotypically normal SECs or endogenous large vessel endothelial cells that are also present in the initial cell isolate is not known, as the pan-endothelial marker CD31, vimentin, and other morphological characteristics rather than SE-1 were used to delineate populations (33
34
35)
. Interestingly, when a cell population relatively depleted in mature hepatocytes and enriched in small hepatocytes (34%) and SE-1-positive cells (20%) is cultured for several weeks in medium containing serum and DMSO, most SE-1 positive cells die within 10 days (15)
, but SE-1-positive cells are observed when cells are cultured in 3D collagen matrices (32)
.
Our SEM studies were limited to observations of the tissue surfaces, and although we observed an increase in 2–5 µm vessel-like openings on the tissue surface in cocultures (Fig. 6)
, we did not observe fenestrated endothelium on the tissue surface. This result is not surprising, since the immunohistochemistry revealed that SE-1-positive cells were located within the tissue, in regions inaccessible to observation by SEM, and CD31+ endothelial cells were located on the tissue surface.
Our culture method differs from previous methods in at least two ways that may be important for stabilization of the connective tissue components: localized perfusion flow and 3D mechanical support of the tissue structure under flow. Flow across pulmonary microvessel endothelial cells cultured on collagen gels is associated with enhanced tube formation of cells growing into the gel (76)
, and slow interstitial perfusion has been reported to induce organization of endothelial cells into microvessel networks in a VEGF-depended manner (24
, 45)
. It is unclear whether some of the network-like structures observed by two-photon microscopy represent endothelial networks representative of liver sinusoids. To facilitate this direction of inquiry, as well as to assess the effects of other culture parameters such as the ratio of endothelial cells to hepatocytes, local flow rates, and oxygen tension on SEC survival and function, we are adapting the perfusion reactor to a 12-well format driven by microfluidic pumping system (26
, 77)
.
In conclusion, we have found that SE-1 positive liver-derived SECs can persist for 13 days in culture in the absence of VEGF and serum in the presence of primary adult hepatocytes when cells are cultured in a 3D tissue format. These results provide the foundation for future work defining factors that support longer term survival and vessel formation.
|
ACKNOWLEDGMENTS |
|---|
Received for publication October 10, 2006. Accepted for publication February 22, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
