In vitro primary hepatocyte systems typically elicit drug induction and toxicity responses at concentrations much higher than corresponding in vivo or clinical plasma Cmax levels, contributing to poor in vitro-in vivo correlations. This may be partly due to the absence of physiological parameters that maintain metabolic phenotype in vivo. We hypothesized that restoring hemodynamics and media transport would improve hepatocyte architecture and metabolic function in vitro compared with nonflow cultures. Rat hepatocytes were cultured for 2 wk either in nonflow collagen gel sandwiches with 48-h media changes or under controlled hemodynamics mimicking sinusoidal circulation within a perfused Transwell device. Phenotypic, functional, and metabolic parameters were assessed at multiple times. Hepatocytes in the devices exhibited polarized morphology, retention of differentiation markers [E-cadherin and hepatocyte nuclear factor-4α (HNF-4α)], the canalicular transporter [multidrug-resistant protein-2 (Mrp-2)], and significantly higher levels of liver function compared with nonflow cultures over 2 wk (albumin ∼4-fold and urea ∼5-fold). Gene expression of cytochrome P450 (CYP) enzymes was significantly higher (fold increase over nonflow: CYP1A1: 53.5 ± 10.3; CYP1A2: 64.0 ± 15.1; CYP2B1: 15.2 ± 2.9; CYP2B2: 2.7 ± 0.8; CYP3A2: 4.0 ± 1.4) and translated to significantly higher basal enzyme activity (device vs. nonflow: CYP1A: 6.26 ± 2.41 vs. 0.42 ± 0.015; CYP1B: 3.47 ± 1.66 vs. 0.4 ± 0.09; CYP3A: 11.65 ± 4.70 vs. 2.43 ± 0.56) while retaining inducibility by 3-methylcholanthrene and dexamethasone (fold increase over DMSO: CYP1A = 27.33 and CYP3A = 4.94). These responses were observed at concentrations closer to plasma levels documented in vivo in rats. The retention of in vivo-like hepatocyte phenotype and metabolic function coupled with drug response at more physiological concentrations emphasizes the importance of restoring in vivo physiological transport parameters in vitro.
hepatotoxicity and bioavailability issues comprise over 60% of drug failures during clinical trials (45) and are a major cause of postmarketing withdrawal (23), pointing to the need to develop more efficient and predictive preclinical test systems. Simple cellular and subcellular assays used to screen compound libraries offer the advantage of higher throughput but are often unable to capture complex biological effects that may require a physiological context for drug interactions with cells. Primary in vitro hepatocyte models widely used to study liver disease, drug metabolism, and toxicity are extensively reviewed in the literature (16, 42). The ability to test the metabolic fate of drugs and their direct or indirect toxic effects on the liver depends on retention and maintenance of hepatocyte-specific structure and function in vitro. These important cellular characteristics are directly related to the successful entry, biotransformation, and excretion of drugs and other xenobiotics from hepatocytes, requiring sequential interactions with phase I and phase II metabolic enzymes as well as transporters. Hepatocyte culture models demonstrating physiologically localized efflux transporters in canalicular surfaces require the cells to be maintained in culture for at least 4–7 days (48). This is confounded by the well-documented dedifferentiation and loss of hepatocyte-specific function and metabolic enzyme activity that occurs over days in nonflow-based culture models (1, 2, 4, 10). As a result, it is rare to find in vitro hepatocyte systems that sustain in vivo morphological and physiological features simultaneously and stably over a prolonged period of time.
Many in vivo microenvironmental factors (Fig. 1A) contributing to the hepatocyte phenotype are lacking in in vitro nonflow cultures of hepatocyte monolayers. These include the polarized morphology of hepatocytes with resultant three-dimensional cell-cell interactions and biliary canalicular formation, extracellular matrix effects such as mechanical signaling and localized cytokine and growth factor concentration buildup, and sinusoidal and interstitial flow-mediated effects such as oxygen and nutrient transport. A common approach adopted by many researchers in the field attempting to prevent dedifferentiation and loss of hepatocyte-specific metabolic activity in vitro is to preserve or restore some aspect of the physiological microenvironment or stimuli that the cells experience in vivo. This ranges from supplementing the culture medium with supraphysiological growth factors or hormones (11, 14), modifying the extracellular matrix environment as in the case of collagen gel sandwiches and overlaid cultures to regain polarization and cell matrix signaling (29, 38), creating three-dimensional cell-cell interactions (5, 6, 28), adding paracrine influences, or coculturing different supportive cell types (17, 18, 27, 43), which may not necessarily be a mature liver cell, e.g., an embryonic fibroblast cell line (3).
While nonflow in vitro systems are used for assessment of metabolic function, the limitations are real. The drug concentration used to demonstrate metabolic responses or toxicity is usually very different from corresponding in vivo plasma concentrations achieving similar effects (22, 26, 33, 41, 47). A drug is administered as a bolus dose at a concentration often orders of magnitude higher than tissue exposure in vivo. Additionally, in nonflow cultures that are closed systems, the metabolites produced by the hepatocytes build up over time in the interval between medium changes to reach nonphysiological levels. As a result, the cellular effects may not be reflective of what happens physiologically in vivo. A continuously perfused system would allow the administration of a drug(s) in a manner that increases concentration profiles over time to reach a steady state that can be determined and tailored for a specific drug class. Simultaneously, an open-circuit, nonrecirculating approach would also prevent metabolite build up in the system as would be expected in vivo. Various groups have begun to incorporate fluid movement (8, 12, 35, 36) and perfusion to overcome these issues as well as to ensure adequate oxygenation to three-dimensional tissue structures.
To recreate a cellular hepatocyte system with sinusoidal and interstitial fluid dynamics and resulting solute transport analogous to in vivo liver circulation, we employed a cone-and-plate viscometer-based technology. This has been extensively used to reestablish in vivo blood vessel cell phenotypes by reproducing the exposure of vascular endothelial cells to human-derived hemodynamic blood flow forces in vitro (19, 20, 44). Herein, we show that applying physiological fluid dynamics in a system analogous to the microstructure of the hepatic sinusoid (a process that we refer to throughout the study as controlled hemodynamis) results in the retention of in vivo-like hepatocyte phenotype and metabolic function coupled with drug responses at in vivo concentrations.
MATERIALS AND METHODS
Animal Surgery and Hepatocyte Isolation
All animals used for the experiments were treated according to protocols approved by HemoShear's Animal Care and Use Committee. Hepatocytes were isolated from male Fischer rats (250–350 g) by a modification of Seglen's two-step collagenase perfusion procedure using a 20 ml/min flow rate (39). Briefly, the rats were anaesthetized with isoflurane, following which the abdominal cavity was exposed and the inferior vena cava was cannulated while making an excision in the portal vein for outflow. The liver was perfused in two steps, first with a Ca2+ free buffer to flush out blood and break up intercellular junctions, followed by collagenase, type IV (Sigma C5138), in a Ca2+-containing buffer to digest the extracellular matrix. After the liver was suitably digested, it was excised and freed of the capsule in a petri dish under a sterile hood. An enriched hepatocyte population (∼95% purity) was obtained by two sequential 65 g centrifugation and washing cycles of 10 min each followed by a 10-min spin with 90% Percoll (Sigma P4937). The viability of hepatocytes was determined by trypan blue exclusion test, and only preparations with viability over 85% were used.
Cell Culture and Device Operating Conditions
Hepatocyte culture medium consisted of DMEM/F-12 supplemented with fetal bovine serum (10% at the time of plating and reduced to 2% for maintenance after 24 h). Additionally, the medium contained gentamycin (50 μg/ml), 1% ITS (Fisher/MediaTech MT-25–800CR), 1% NEAA (Fisher no. SH3023801), 1% glutamax (Invitrogen/Gibco 35050–061), and dexamethasone (Sigma no. D4902, 1 μM at plating and 250 nM for maintenance after 24 h).
The cells were plated in a collagen gel sandwich configuration using a previously described protocol (29) with type I rat tail collagen (BD Biosciences). Initial testing was done comparing “traditional” collagen gel sandwich configurations on 100-mm tissue culture-treated sterile Nunc dishes (Fisher) with a “modified” sandwich configuration that employed the sandwich on the under surface of 75-mm polycarbonate Transwells (Corning). For the Nunc dishes, 10 ml of medium were used. For the modified sandwich configuration, the Transwells were reinverted and 15 ml of medium were added to the Transwell. The maintenance medium was changed every 48 h, and after 7 days of culture, the cells were washed with PBS and scraped and the RNA was isolated for RT-PCR. After noting that the two were not significantly different based on the expression of a panel of metabolic and nonmetabolic genes (data not shown), we chose to use the modified sandwich to rule out configuration variables while assessing the effects of transport and hemodynamics. For the experiments described in the study, the maintenance medium was replaced every 48 h in nonflow cultures. For the rest of the cultures, after 24 h in a standard CO2 incubator, the Transwells were set up in a configuration to allow for control of hemodynamics and transport (Fig. 1B) within devices described previously (20). Maintenance medium was continuously perfused on both sides while shear stress was applied on the top surface based on the calculations described below. The devices were housed in a controlled environment at 37°C with 5% CO2 mixed with air.
To compute the range of magnitudes of relevant physiological shear stress in the liver sinusoids, the shear stress, τ (dyn/cm2), was calculated using the following equation for pressure driven flow of a Newtonian fluid through a cylinder; whereby the reference values for pressure gradient across the sinusoid (ΔP), radius of sinusoids (r), and length of the sinusoids (l) were obtained from the literature (31, 46).
For experimental optimization, we tested an expanded range of shear stress conditions in the device and noted that they resulted in different transport profiles of a reference solute dye (horseradish peroxidase) as measured by transmembrane flux concentrations over time. Based on the expression of a select metabolic gene panel from RNA isolated in these experiments (data not shown), we picked an operational shear stress of 0.6 dyn/cm2 for all the subsequent experiments described in this study.
At the prescribed time points in the experimental design, the nonflow and the device Transwells were removed and washed gently with PBS followed by fixation with 4% paraformaldehyde for 30 min. The samples were first permeabilized with 0.1% Triton-X for 20 min then washed with PBS and blocked with 5% goat serum. The incubation with primary antibodies [hepatocyte nuclear factor-4α (HNF-4α): Santa Cruz sc-8987; E-cadherin: Santa Cruz sc-71009; and anti-multidrug-resistant protein-2 (anti-Mrp2): Abcam ab3373] was at a dilution of 1:100 for 1 h. After three washes with PBS with 1% BSA, the secondary antibody was added at a dilution of 1:500 for 1 h. The samples were then washed with PBS plus 1% BSA and then mounted for imaging on a Nikon C1+ Confocal System microscope.
Transmission electron microscopy.
At the chosen time points device Transwells were removed and washed gently with PBS followed by fixation with a solution containing 2.5% glutaraledhyde and 4% paraformaldehyde for 30 min. The fixed samples were sectioned and imaged at the Advanced Microscopy Facility, University of Virginia.
Urea and albumin.
Medium collected from nonflow cultures and devices at various time points was assayed for albumin using a rat-specific ELISA-based kit (Bethyl Laboratories) as per the manufacturer's protocols. Urea was measured from the medium samples using a standard colorimetric assay (QuantiChrom Urea Assay Kit, DIUR-500; Gentaur). All measurements between the systems were normalized for comparison on a per million cells per day rate based on the volume of medium perfused and the number of initially plated cells.
Cytochrome P450 activity assays.
Hepatocytes were cultured in the devices under controlled hemodynamics or nonflow conditions for 5 days before being treated with the chosen inducers or vehicle control for 48 h and harvested at day 7. Filter segments from the device membranes (∼2 cm2 in area) were excised and transferred to standard 24-well plates alongside corresponding segments from nonflow cultures. The cells were incubated with 500 μl of hepatocyte medium containing substrates from commercially available P450-Glo kits (Promega) at the manufacturer-recommended concentrations. After 4 h, the medium was transferred to 96-well plates and assayed for luminescent metabolites to reflect cytochrome P450 (CYP) activity as per the manufacturer's protocol. The ATP content of the cells in the same filter segments or two-dimensional wells was then measured by using the CellTiter-Glo assay (Promega) according to the manufacturer's protocol. CYP values were normalized to ATP content with the assumption that ATP was linearly related to the number of viable cells.
Total RNA was isolated using a PureLink RNA Mini Kit (Invitrogen) and reverse transcribed to cDNA using the iScript cDNA Synthesis Kit (Bio-Rad). Primers were designed for the genes CYP1A1, CYP1A2, CYP2B1, CYP2B2, CYP3A2, and glutathione S-transferase (GST) pi (Table 1). RNA expression was analyzed by real-time RT-PCR using iQ SYBR Green Supermix (Bio-Rad) and a CFX96 Real-Time System (with C1000 Thermal Cycler; Bio-Rad). RNA data were normalized to endogenous expression of β2-microglobulin and reported as a relative quantity compared with traditional nonflow cultures. The primer sequences used for the genes chosen are shown in Table 1.
At various time points over 2 wk in culture (see Fig. 5D), the plated surface of a Transwell was harvested for protein in 1× RIPA buffer (Millipore) containing dithiothreitol (Thermo) and protease inhibitors. Samples were sonicated on ice and centrifuged at 17,000 g for 10 min in a chilled microcentrifuge. Protein concentration was determined using A660nm Protein Reagent (Pierce). Samples were boiled and run on a 7.5% TGX gel (Bio-Rad) before wet-transferring to 0.2 μm PVDF membrane (Bio-Rad) and blocking in 5% nonfat milk (Carnation) at room temperature for 10 min. Membranes were incubated overnight at 4°C in rabbit anti-UDP glucuronosyltransferase antibody (anti-UGT, 1:500 dilution; Cell Signaling). Secondary antibody (horseradish peroxidase, 1:5,000 dilution; Santa Cruz) incubation was at room temperature for 1 h. Chemiluminescent signal was developed using Super Signal West Pico (Pierce) reagent and captured using an Innotech AlphaEase imaging system. For normalization, gels were probed for mouse anti-β-actin (Sigma A1978, 1:2,000 dilution) followed by secondary goat anti-mouse horseradish peroxidase (Santa Cruz sc-2005, 1:10,000 dilution).
Transporter and Biliary Efflux Activity
For imaging of the biliary activity at canalicular junctions, membrane sections containing live cells were washed with PBS and incubated with media containing 10 μM carboxy-2,7-dichlorofluorescein diacetate (CDFDA) for 10 min. Samples were then washed with PBS and placed on glass slides for confocal imaging as described above.
Hepatocytes were cultured in the devices under controlled hemodynamics or nonflow conditions for 6 days before being treated with different concentrations of dexamethasone or vehicle control for 24 h and harvested at day 7. Filter segments from the device membranes (∼2 cm2 in area) were excised and transferred to standard 24-well plates alongside corresponding segments from nonflow cultures. The cells were incubated with 500 μl of thiazolyl blue tetrazolium bromide (MTT) solution in culture medium at a final concentration of 1 mg/ml. After 1 h, the MTT-containing medium was replaced by 500 μl of dimethyl sulfoxide (DMSO) and the plates were agitated on a shaker for 10 min. The purple supernatants were then transferred to 96-well plates and assayed for absorbance at 595 nm using a spectrophotometer.
To compare albumin and urea production, metabolic gene expression, and CYP activity data, two-tailed unpaired Student's t-test was performed. Basal CYP activity values were compared across conditions and induced values were compared with basal values from the corresponding condition before evaluating for statistical significance at P < 0.05.
Controlled Hemodynamics Maintain Differentiated Hepatocyte Phenotype, Polarized Morphology, and Transporter Localization Relative to Traditional Nonflow Monoculture Conditions
To determine the importance of controlled hemodynamics and biological transport in restoring in vivo phenotype, we exposed primary rat hepatocytes to controlled hemodynamics (0.6 dyn/cm2) relative to nonflow cultures over 2 wk in culture and evaluated the differences in differentiated morphology, localized tight junctional proteins, and transporter expression.
Immunostaining patterns of the tight junctional protein E-cadherin at day 7 in modified nonflow collagen gel sandwich cultures (Fig. 2A) displayed higher cytoplasmic levels and discontinuous peripheral membrane distribution. Under controlled hemodynamics (Fig. 2B), hepatocytes exhibited a more differentiated morphology characterized by clearly delineated peripheral membrane localization and lower cytoplasmic levels of E-cadherin. These results were confirmed by morphometric analysis of the images shown in Fig. 2. Transmission electron microscopy images of day 7 cultures under controlled hemodynamics also confirmed the presence of bile canaliculi and tight junctions (Fig. 2C), along with abundant retention of subcellular components such as rough and smooth endoplasmic reticulum and mitochondria (Fig. 2D).
Canalicular transporters, internalized during hepatocyte isolation, are typically absent in hepatocyte monolayers plated on collagen. They require 4–5 days to localize along canalicular surfaces in hepatocytes within collagen gel sandwiches but are lost over time. We hypothesized that fluid transport resulting from controlled hemodynamics would stably maintain the canalicular expression of these transporters while retaining differentiated morphology. We found that canalicular localization of the transporter Mrp-2 (green) is lost in nonflow cultures by day 14 (Fig. 2E) but the canalicular network patterns are stable and extensive under controlled hemodynamics (Fig. 2F). Day 14 cultures maintained under controlled hemodynamics costained for Mrp-2 and HNF-4α, a hepatocyte nuclear transcription factor whose expression is a marker of a mature hepatoctye (Fig. 2H). Compared with tissue sections from intact rat liver (Fig. 2G), day 14 cultures maintained under controlled hemodynamics (Fig. 2H) show very similar staining patterns of Mrp-2 canalicular localization and HNF-4α nuclear localization.
Controlled Hemodynamics Results in Retention of Hepatocyte-Specific Function in Rat Hepatocytes in a Collagen Gel Configuration Relative to Nonflow Cultures over 14 Days
Hepatocytes were cultured in the Transwells under nonflow conditions or controlled hemodynamics for 2 wk, and the medium was sampled at 4, 7, 10, and 14 days. Assays for urea and albumin were performed on the medium, and the values were normalized to production rates over 24 h per million cells based on the initial number of plated cells. Hepatocyte function reflected by secreted albumin measured from medium samples at various time points over 14 days and expressed as micrograms per 106 plated hepatocytes per day (Fig. 3A) showed significantly higher levels (3- to 4-fold) under controlled hemodynamics compared with nonflow cultures (day 7: 97.96 ± 11.34 vs. 25.84 ± 8.22, P = 0.00001; day 14: 87.80 ± 8.62 vs. 33.93 ± 4.39, P = 0.0001). Urea secretion (Fig. 3B) under controlled hemodynamics, expressed as μg/106 plated hepatocytes/day, was four to five fold levels higher than nonflow cultures over 2 wk in culture (day 7: 622.78 ± 33.96 vs. 139.76 ± 13.37, P = 2.7 × 10−9; day 14: 667.71 ± 84.37 vs. 178.68 ± 6.13, P = 1 × 10−6).
Controlled Hemodynamics Differentially Regulates the Expression of Phase I and Phase II Metabolic Genes and Proteins Compared with Nonflow Cultures
To assess the impact of controlled hemodynamics on various metabolic genes, we selected a gene panel that included some of the crucial phase I CYP genes known to be depressed in nonflow cultures over time (Table 1). We also chose the pi isoform of the phase II enzyme GST known to be overexpressed in older nonflow cultures as the cells dedifferentiate. Hepatocytes were cultured in the Transwells under no flow or controlled hemodynamics for 7 days. QRT-PCR was performed for select metabolic genes (Table 1) on RNA samples collected at day 7 from these conditions. All values were normalized to day 7 nonflow cultures. Hepatocytes cultured under controlled hemodynamics resulted in gene expression levels that are consistently higher than nonflow cultures (n = 11, fold changes relative to nonflow cultures; CYP1A1 = 53.5 ± 10.3, P = 0.0003; CYP1A2 = 64.0 ± 15.1, P = 0.005; CYP2B1 = 15.2 ± 2.9, P = 0.001; CYP2B2 = 2.7 ± 0.8, P = 0.09; CYP3A2 = 4.0 ± 1.4, P = 0.075; Fig. 4, A and B). Interestingly, the expression levels of the gene for the pi subunit of phase II enzyme GST, known to increase in nonflow cultures over time (30), were lower in hepatocytes cultured under controlled hemodynamics (−2.3 fold, P = 0.025) compared with nonflow cultures (Fig. 4C).
Additionally, we chose to assess the protein levels of a phase II metabolic enzyme UGT1, also known to decrease in nonflow cultures (40). Hepatocytes were cultured over 2 wk in the Transwells under no flow or controlled hemodynamics (0.6 dyn/cm2). Cell cultures were collected at 4, 7, 11, and 14 days, and cell lysates were obtained as described in the materials and methods and probed for the phase II enzyme UGT1 and β-actin (for normalization). Western blots (Fig. 4D) demonstrate that UGT1 is upregulated under controlled hemodynamics compared with nonflow conditions at all the time points over 2 wk in culture.
Controlled Hemodynamics Results in a Higher Level of Basal CYP Activity Compared with Nonflow Cultures While Exhibiting Extensive Biliary Efflux Activity Concomitantly
To validate that the increase in metabolic genes and proteins translated to changes in metabolic activity, CYP activity was assessed. Primary rat hepatocytes were cultured in the devices under controlled hemodynamics and in nonflow collagen gel cultures. On day 7, filter segments from the devices containing hepatocytes cultured under controlled hemodynamics were transferred to standard 24-well plates, in parallel with nonflow cultures and treated with substrates for the CYP enzymes. CYP assays were performed on day 7 using commercially available P450-Glo kits (Promega). After 4 h, the medium was transferred to 96-well plates and assayed for luminescent metabolites to reflect CYP activity. Values were normalized to cellular ATP content, measured by CellTiter-Glo assay (Promega), for an accurate representation of live cells and to avoid any confounding effects of the collagen gels on total protein measurements. Hepatocytes cultured in the presence of controlled hemodynamics demonstrated a significantly higher amount of basal CYP activity (Fig. 5A) compared with nonflow cultures grown in modified collagen gel sandwiches (controlled hemodynamics vs. nonflow: CYP1A: ∼15-fold, 6.26 ± 2.41 vs. 0.42 ± 0.015, P = 0.014; CYP1B: ∼9-fold, 3.47 ± 1.66 vs. 0.4 ± 0.09, P = 0.033; CYP3A: ∼5-fold, 11.65 ± 4.70 vs. 2.43 ± 0.56, P = 0.002).
To test for active efflux transport into biliary canalicular spaces concomitantly at a time point when cytochrome activity was retained, hepatocyte cultures were maintained under controlled hemodynamics in the system for 7 days. The filter segments were excised and immediately incubated with the substrate CDFDA. The cells were imaged by confocal microscopy during a 20-min exposure to nonfluorescent substrate CDFDA to allow for the hydrolysis of the substrate to the highly fluorescent Mrp-2 substrate carboxy-2,7-dichlorofluorescein (CDF) and its active secretion into the bile canalicular structures, as shown in Fig. 5B.
Hepatocytes Maintained Under Controlled Hemodynamics Display Toxic and Enzyme Induction Responses at Closer to In Vivo Levels Compared with Nonflow Cultures
To determine if hepatocytes maintained under controlled hemodynamics retained in vivo-like metabolic enzyme responses to standard prototypical inducers, hepatocytes were cultured for 5 days in the system, followed by an additional 48-h treatment with 3-methylcholanthrene (a CYP1A and CYP1B inducer; Ref. 37), dexamethasone (a CYP3A inducer; Ref. 34), or vehicle control 0.1% DMSO. Interestingly, 50 μM dexamethasone, which is used routinely in nonflow hepatocyte cell culture protocols, resulted in a toxic response in the cultures under controlled hemodynamics. With the use of the MTT assay, a dose response with dexamethasone demonstrated that 50 μM resulted in toxicity to over 75% of the hepatocytes (Fig. 6A) under controlled hemodynamics compared with nonflow cultures.
Thus, for the induction experiments comparing nonflow cell cultures to controlled hemodynamics, we adjusted the concentrations of the inducer compounds to account for the observed cytoxicity in Fig. 6A. Hepatocytes were treated with 0.1% DMSO or the inducers 3-methylcholanthrene (3-MC; 1 μM in nonflow and 0.1 μM in controlled hemodynamic cultures) and dexamethasone (50 μM in nonflow and 2.5 μM in controlled hemodynamic cultures) for 48 h. CYP assays were performed on 2-cm2 filter segments as described above. As expected, the vehicle (DMSO)-treated cells under controlled hemodynamics displayed higher levels of CYP1A/1B and CYP3A activity relative to nonflow, similar to previous untreated basal values. Even in the presence of higher levels of basal activity, the hepatocytes under controlled hemodynamics showed induction of CYP1A, CYP1B, and CYP3A to 3-MC and dexamethasone compared with DMSO controls (CYP1A/1B induction response to 3-MC over DMSO control: ∼27-fold, 133.06 ± 47.43 vs. 4.87 ± 2.40; CYP3A response to dexamethasone over DMSO control: ∼5-fold, 57.53 ± 19.12 vs. 11.65 ± 4.70; Fig. 6B).
The major focus of the studies presented in this report was to determine whether controlled hemodynamics could enhance and/or prolong primary hepatocyte function in vitro compared with the modified nonflow collagen gel sandwich culture system, while retaining the benefits of the latter, such as polarized morphology. Herein, we provide a number of novel results. First, we show that controlled hemodynamics results in sustained polarized morphology, establishment of bile canaliculi with a functioning transporter, and retention of key hepatocyte differentiation markers, such as the nuclear localization of HNF-4α. Second, we provide direct evidence that controlled hemodynamics also results in a sustained and stable increase in basal hepatocyte-specific functions, such as albumin and urea production, over 2 wk compared with the modified nonflow collagen gel sandwich. Third, we show that key phase I and phase II genes involved in the metabolism of drugs are expressed and sustained at significantly higher levels than the modified collagen gel sandwich. Fourth, drug induction of these enzymes in hepatocytes under controlled hemodynamics occurs at more physiologically relevant drug concentrations. For example, the dose of dexamethasone, which results in induction in vivo in rats, equates to a plasma Cmax value of 0.15–1.5 μg/ml (∼0.35–3.5 μM; Ref. 13). We were able to achieve dexamethasone induction responses at comparable in vivo concentrations under controlled hemodynamics, while nonflow cultures required concentrations that were at least an order of magnitude higher. Taken together, these results highlight the importance of interfacing in vitro biology with in vivo physiological parameters. Specifically, the retention of in vivo-like hepatocyte phenotype and metabolic function coupled with drug responses at more physiological concentrations requires the restoration of in vivo physiological transport parameters in vitro.
The importance of hemodynamic blood flow and fluid transport for regulating cell biology in vivo and in vitro is not novel. In fact, decades of research have repeatedly shown that endothelial cells that line a blood vessel are highly sensitive to changes in hemodynamic blood flow and that differential molecular and physiological phenotypes are the result of varying hemodynamics throughout the vasculature (9). For example, in the human circulation, the internal carotid sinus (located at the carotid artery bifurcation) is a region of the vasculature that is highly susceptible to inflammation, whereas the common carotid artery, a region just proximal, is not. The hemodynamic forces imposed on the endothelium are also regionally distinct, with the internal carotid sinus experiencing low time-averaged shear stress and the common carotid artery experiencing significantly higher time-averaged shear stress, as determined using MRI in humans (15). Using the same technology herein, Hastings et al. (20) showed that exposing endothelial cells in coculture with smooth muscle cells to these specific hemodynamics (15) resulted in ex vivo endothelial and smooth muscle cells with a molecular and physiological phenotype representative of the respective region in vivo.
In translating these principles to the liver, we focused on the primary liver cell type, the hepatocyte, with the goal of recreating physiochemical cues these cells experience in their microenvironment due to hemodynamics and biochemical gradients brought about by sinusoidal blood flow. We chose rat hepatocytes plated in collagen gel sandwiches as our nonflow comparator because that is an extensively utilized system, demonstrated to preserve aspects of hepatocyte morphology such as cell polarity and canalicular expression of transporters typically lost on regular tissue culture surfaces (21, 29). The collagen gel sandwich system is also widely regarded as a superior alternative to conventional hepatocyte monolayers due to the relatively stable differentiated function, expression of efflux transporters, and polarized morphology. We modified and incorporated the collagen sandwich configuration into the cone and plate technology that has been extensively characterized for its ability to recreate a physiological vascular tissue system (19, 20, 44). During the design of our system, we considered that hepatocytes are physiologically shielded from the direct mechanical aspects of sinusoidal blood flow and experience the interstitial flow patterns that filter through the leaky layer of sinusoidal endothelial cells. Thus, as shown in Fig. 1, we used the membrane of a Transwell to act as a mechanical filter that shields hepatocytes plated in a collagen gel sandwich on its underside from the direct hemodynamics that we applied across the membrane. An initial phase of optimization was carried out where we tested a range of operational parameters such as cone speeds and medium viscosities that resulted in surface shear stresses within the physiological range reported in liver sinusoids in vivo and allowed for effective transport of reference solutes across the membrane.
While adapting the collagen gel sandwich into the cone and plate technology, it was important for us to ensure that the cells in the extracellular matrix remained stable in spite of the flow patterns and achieved the polarized morphology and differentiated phenotype that distinguishes nonflow collagen gel sandwich gel systems from hepatocytes plated by other means. This was evidenced by the distinct peripheral E-cadherin staining pattern and the canalicular localization of transporter proteins such as Mrp-2 that was seen in our 7-day cultures. We also wanted to demonstrate that the canalicular localization of transporters translated into active efflux activity out of the cell into the biliary canaliculi because this is a physiological process that can prevent drug or metabolite build up within the cell while testing toxicity. Previous studies show that transporter proteins are internalized at the time of perfusion (48) making fresh hepatocyte suspensions less than ideal for studying transporter-related activity. In sandwich cultures that are widely used to study transporter activity, these proteins reassemble along the bile canalicular surfaces over 3–7 days, a period that is usually concomitant with decline in the various components responsible for drug metabolism and toxicity, namely the various CYP enzymes and phase II enzymes (Fig. 7A, red vs. blue solid lines). It is well documented that the expression and activity levels of many of these important enzymes in hepatocytes are greatly decreased over time in culture as a result of the dedifferentiation process (1, 2, 4, 10). We ran our experiments in parallel with modified nonflow collagen gel sandwich cultures for 7 days and compared the expression of the CYPs 1A1, 1A2, 2B2, and 3A2. We found that by applying physiological hemodynamics, we could maintain the expression levels of these genes significantly higher than nonflow cultures. Moreover, GST pi expression was low relative to nonflow controls. The GST pi isoform of the enzyme is expressed at practically undetectable levels within mature hepatocytes in vivo but appears in conventional nonflow cultures after 24 h and increases over time in culture in parallel with the process of dedifferentiation (30). The finding that GST pi expression is suppressed in intact liver and hepatocytes under controlled hemodynamics relative to nonflow conditions suggests that controlled hemodynamics not only induce the expression of the metabolic genes lost in nonflow cultures but also prevent the upregulation of genes known to go up in culture over time. To test the differences in overall CYP response between the systems, we carried out experiments in both systems for 5 days before treating them with standard drugs known to induce CYP activity. After 48 h of treatment, the enzymatic activity on substrates were tested side-by-side with nonflow cultures ex situ. Isolating microsomes from hepatocytes is the most specific method for measuring individual CYP activities independent of other biological processes. However, we did not perform this as we had already noted higher level of expression of individual CYP genes. Rather, we used commercial P450 whole cell assays to reflect an outcome dependent on a combination of the inherent CYP activity and intracellular drug levels arising from the transporter status of the cells. Thus we demonstrate herein that our system can maintain hepatocyte stability sufficiently to restore the localization of a transporter for functional transport of metabolites into the canaliculi while simultaneously retaining phase I and phase II enzyme metabolic expression profiles and activity (Fig. 7A, blue dashed line).
The “shift” or “reset” of the in vitro hepatocyte phenotype to a more in vivo-like phenotype was accompanied by CYP induction at in vivo physiological drug concentrations. While measuring the CYP activity to confirm the enhanced gene expression that we noted under controlled hemodynamics and evaluating response to classical inducers, we found that 50 μM dexamethasone (20 μg/ml), the concentration recommended for inducing nonflow cultures, was toxic in our system. As a result, the concentration of the dexamethasone was reduced to 1 μg/ml (2.5 μM) for induction experiments so as to not cause cytotoxicity, a level that correlates to equivalent in vivo plasma concentrations that cause CYP induction in rats, ∼0.35–3.5 μM (13). Similarly, induction responses with 3-MC were also seen at 10-fold lower levels under controlled hemodynamics. Thus we demonstrate that CYP induction and cytotoxicity occur at drug concentrations that are one order of magnitude lower than in existing nonflow models. This suggests that our system has the potential to restore in vitro hepatocyte responsiveness to drug concentrations that are observed preclinically or clinically in vivo (Fig. 7B).
Currently, there are major initiatives to create more meaningful heterocellular organotypic models for basic research and drug discovery (7). The discrepancy between drugs that pass early in vitro screening systems and those that succeed in a clinical setting highlights and motivates the need for developing better in vitro liver systems that accurately reflect the in vivo response and are more predictive. The various approaches adopted by different research groups attempting to bridge this gap have one aspect in common, filling in the missing components of the cell culture systems in an attempt to recreate the physiologic environmental cues, be they growth factors in the medium (11, 14), extracellular matrix modifications (29, 38), increasing three-dimensional cell-cell interactions (5, 6, 28), adding paracrine influences by coculturing different supportive cell types (17, 18, 27, 43), or incorporating flow (8, 12, 35, 36). As briefly discussed above, heterotypic cell communication between vascular endothelial and smooth cells in culture affects the phenotype of both cells. For example, Hastings et al. (20) showed that removal of endothelial cells (primed under hemodynamics) in coculture with smooth muscle cells results in phenotypic modulation of the underlying smooth muscle cells, similar to what occurs in human and experimental angioplasty, e.g., restenosis. Indeed, coculturing multiple cell populations from the liver is not novel, e.g., sinusoidal endothelial cells, Kuppfer cells, and hepatocytes, have been cultured in microfiber meshes and bioreactors (8, 24, 25, 32). However, recreating three-dimensional architecture and cell orientation that more closely mimic the spatial localization of cells in the liver is challenging, as is the ability to separate the cells at the end of an experiment for cell-specific mechanistic studies. A premise of this study was that hepatocytes in a healthy liver are separated by the space of Disse from sinusoidal endothelial cells and not in direct contact with liver blood flow hemodynamics and that movement of fluid, i.e., fluid transport, across the space of Disse can regulate underlying hepatocyte biology (the orientation depicted in Fig. 1B). This supports the hypothesis that physiological forces coupled with cell-cell communication are important in recreating in vivo biology in vitro and further studies by our group are ongoing to determine the broad significance of this.
To summarize, we adapted the technology published by Hastings et al. (20) to culture primary rat hepatocytes in a system where they restore and retain their in vivo-like phenotype over 2 wk of culture. Cells cultured in this system display liver-specific functions, such as albumin and urea synthesis, at significantly higher levels and retain higher levels of nuclear factors and P450 enzymes necessary for drug metabolism and known to be lost in nonflow cultures. Under these conditions, hepatocytes exhibit an enhanced ability to metabolize substrates for specific P450 enzymes compared with nonflow cultures and transport metabolites into canaliculi, while retaining the ability to be induced by drugs known to upregulate these enzymes. It will be necessary to expand the fundamental findings presented herein and determine whether various classes of drugs, which cause toxicity in vivo, are replicated at physiological doses in systems such as this. Moreover, whether these fundamental principles that appear to drive a more meaningful biology in rat hepatocytes in vitro translate to human primary hepatocytes remains to be determined.
This work was funded in part by National Institute of Diabetes and Digestive and Kidney Diseases Grant SBIR R43-DK-091104 to HemoShear.
A. Dash, M. B. Simmers, T. G. Deering, D. J. Berry, R. E. Feaver, N. E. Hastings, B. R. Blackman, and B. R. Wamhoff are employees of HemoShear, a biotechnology company that owns the technology described in the study. T. L. Pruett and E. L. LeCluyse are scientific advisors to HemoShear.
Author contributions: A.D., R.E.F., N.E.H., B.R.B., and B.R.W. conception and design of research; A.D., M.B.S., T.G.D., and D.J.B. performed experiments; A.D., M.B.S., and T.G.D. analyzed data; A.D., T.L.P., E.L.L., and B.R.B. interpreted results of experiments; A.D. and M.B.S. prepared figures; A.D., M.B.S., T.G.D., and B.R.W. drafted manuscript; A.D., M.B.S., T.L.P., E.L.L., B.R.B., and B.R.W. edited and revised manuscript; A.D., B.R.B., and B.R.W. approved final version of manuscript.
We thank Nathan Day and Caroline Languasco for providing technical support, Jan Reddick for preparing and imaging samples for transmission electron microscopy, and Allison Armstrong for proofreading.
- Copyright © 2013 the American Physiological Society