An Ex Vivo Perfusion System Emulating In Vivo Conditions in Noncirrhotic and Cirrhotic Human Liver

Abstract

Various models are used for investigating human liver diseases and testing new drugs. However, data generated in such models have only limited relevance for in vivo conditions in humans. We present here an ex vivo perfusion system using human liver samples that enables the characterization of parameters in a functionally intact tissue context. Resected samples of noncirrhotic liver (NC; n = 10) and cirrhotic liver (CL; n = 12) were perfused for 6-h periods. General and liver-specific parameters (glucose, lactate, oxygen, albumin, urea, and bile acids), liver enzymes (aspartate aminotransferase, alanine aminotransferase, lactate dehydrogenase, glutamate dehydrogenase, and γ-glutamyl transferase), overall (M65) and apoptotic (M30) cell-death markers, and indicators of phase-I/phase-II biotransformations were analyzed. The measurement readings closely resembled (patho)physiological characteristics in patients with NC and CL. Mean courses of glucose levels reflected the CLs' reduced glycogen storage capability. Furthermore, CL samples exhibited significantly stronger increases in lactate, bile acids, and the M30/M65 ratio than NC specimens. Likewise, NC samples exhibited more rapid phase-I transformations of phenacetin, midazolam, and diclofenac and phase-I to phase-II turnover rates of the respective intermediates than CL tissue. Collectively, these findings reveal the better hepatic functionality in NC. Perfusion of human liver tissue with this system emulates in vivo conditions and clearly discriminates between noncirrhotic and cirrhotic tissue. This highly reliable device for investigating basic hepatic functionality and testing safety/toxicity, pharmacokinetics/pharmacodynamics and efficacies of novel therapeutic modalities promises to generate superior data compared with those obtained via existing economic perfusion systems.

Introduction

Unpredicted drug-induced liver injury accounts for termination of ∼22% of drugs in clinical trials and ∼32% of withdrawals of approved drugs (Watkins, 2011). Although animal models have long been used for mimicking physiological and pathophysiological conditions in human liver, data gathered from such models do not always match humans. Ethical considerations limit the use of human test subjects for such studies (Cheng et al., 2011). There is therefore a key need for models that better predict drug-induced liver injury in humans. Furthermore, models that also more reliably predict drug metabolization, pharmacodynamics, pharmacokinetics, and pharmacogenomics are a cornerstone of pharmaceutical development (Liew et al., 2011; Watkins, 2011).

Several established in vitro models use primary human hepatocytes and isolated hepatic tissue to evaluate physiological, pathophysiological, and pharmacological characteristics (Gebhardt et al., 2003; Hewitt et al., 2007; Gómez-Lechón et al., 2008). Primary hepatocytes grown in suspension or two-dimensional culture conditions rapidly lose important phenotypes found in vivo (e.g., phase-I and -II metabolism). Although some culture systems provide three-dimensional scaffolds that maintain hepatocyte function and metabolism for longer periods of time (Funatsu et al., 2001; Gerlach, 2006; Balmert et al., 2011; Zeilinger et al., 2011), the complexity of the intact organ with multiple cell types is not recapitulated. Cultured hepatic tissue (e.g., liver slices) addresses some of the concerns of cell culture, but also has its limitations. Specifically, the oxygen/nutrient gradient normally found in the liver lobule as an effect of blood flow is not present in cultured slices.

An ex vivo model that bypasses some of the limitations raised above involves machine-perfused organs. In such a system, the hepatic architecture is maintained, as well as the lobular flow of oxygen and nutrients. Isolated perfused livers from research animals (e.g., pigs, sheep, or rats) are often used for studying hepatotoxicity and metabolism (Ali et al., 2000; Grosse-Siestrup et al., 2002; Bessems et al., 2006; Thewes et al., 2007). Machine perfusion of human tissue is generally performed to maintain an organ's quality until transplantation and to minimize reperfusion damage (Dutkowski et al., 2008; Monbaliu and Brassil, 2010; Vogel et al., 2010), but few, if any, studies use such a system to characterize hepatic responses.

The first goal of this work was to develop an ex vivo perfusion system of human liver tissue. We present an economic closed-circuit perfusion system that maintains a piece of human liver up to a weight of 55 g as viable for >6 h. We compared the responses in normal and cirrhotic liver (CL) tissue to develop a paradigm in which the characteristics of healthy and diseased livers can be compared. The functional parameters obtained throughout the perfusion period indicate that this setup will enable (patho-)physiological, pharmacodynamic, pharmacokinetic, pharmacogenomic, and toxicological analyses in an environment mimicking the hepatic in vivo situation as closely as possible.

Materials and Methods

Patients and Liver Samples.

All patients provided written informed consent according to the Helsinki declaration of 1995, and the study protocol conformed to the guidelines of the ethics committee of the University Hospital Essen (file number 09-4252). Noncirrhotic liver (NC) tissue (n = 10) was obtained primarily from liver portions that had been partially resected because of different entities of liver-metastasized neoplasms (Table 1) with an ample circumference of surrounding healthy tissue. While the diseased tissue was forwarded to the pathologist, adjacent unaffected tissue was transferred to the laboratory, stripped of any remnants of macroscopically visible pathologic areas, and then connected to the perfusion system (see below). During this procedure tissue pieces were collected for paraffin embedding (4% Roti-Histofix; Carl Roth GmbH and Co. KG, Karlsruhe, Germany) and cryopreservation for subsequent histological analysis. A second sampling was performed at the end of the perfusion period. Cirrhotic liver tissue (n = 12) was derived mainly from the left lobe of whole liver explants in the course of complete orthotopic liver transplantation in patients with hepatitis C, primary biliary cirrhosis, or ethanol toxicity (Table 1).

TABLE 1.

General characteristics of patients and derived liver samples

Experimental intervention indicates the replenishment of glucose when the perfusate value went below 40 mg/dl and how much buffer was totally necessary during 6 h to maintain a pH of 7.35 in a perfusate volume of 200 ml.

Number	Surgical Intervention	Underlying Disease	Gender	Age	Sample Weight	Experimental Intervention
				years	g
NC-1	Right-side hemihepatectomy	Klatskin tumor	Female	61	55	15 mM NaHCO3
NC-2	Left-side hemihepatectomy	Hemangioma	Female	38	18	3 mM NaH2PO4
NC-3	Right-side hemihepatectomy	Klatskin tumor	Female	60	70	15 mM NaHCO3
NC-4	Left-side partial resection	Metastasized colon carcinoma	Female	75	19	2 mM NaH2PO4, 12.5 mM NaHCO3
NC-5	Right-side hemihepatectomy	Metastasized colon carcinoma	Male	43	45	17.5 mM NaHCO3
NC-6	Right-side hemihepatectomy	Metastasized rectal carcinoma	Male	52	55	35 mM NaHCO3
NC-7	Right-side hemihepatectomy	Metastasized soft palate carcinoma	Male	55	54	20 mM NaHCO3
NC-8	Explantation	Hepatocellular carcinoma	Female	59	56	30 mM NaHCO3
NC-9	Right-side hemihepatectomy	Cholangiocellular carcinoma	Male	70	55	22.5 mM NaHCO3
NC-10	Explantation	Hepatocellular carcinoma	Male	59	50	25 mM NaHCO3
CL-1	Explantation	Hepatitis C	Female	61	40	30 mM NaHCO3
CL-2	Explantation	Hepatitis C	Male	50	25	15 mM NaHCO3
CL-3	Explantation	Hepatitis C	Male	51	48	30 mM HaHCO3, 150 mg/dl glucose
CL-4	Explantation	Ethanol toxicity	Male	48	42	30 mM NaHCO3
CL-5	Explantation	Hepatitis C	Male	61	55	55 mM NaHCO3
CL-6	Explantation	Hepatitis C	Male	66	55	40 mM NaHCO3
CL-7	Explantation	Primary biliary cirrhosis	Female	51	32	35 mM NaHCO3
CL-8	Explantation	Hepatitis C	Male	61	38	70 mM NaHCO3, 2× 200 mg/dl glucose
CL-9	Explantation	Hepatitis C	Male	41	31	10 mM NaHCO3
CL-10	Explantation	Hepatitis C	Male	57	57	42.5 mM NaHCO3
CL-11	Explantation	Ethanol toxicity	Male	58	57	52.5 mM NaHCO3
CL-12	Explantation	Ethanol toxicity	Female	34	55	50 mM NaHCO3, 200 mg/dl glucose

Liver Perfusion System.

The perfusion system was designed as a closed circuit providing a constant flow rate, being supplied with nutrients and oxygen, to liver samples of 20 to 55 g (for a schematic drawing, see Fig. 1; for a photograph of the actual setup see Supplemental Fig. 1). A pump drive with two peristaltic heads (Masterflex; Novodirect, Kehl, Germany) was used for the bidirectional transport of the perfusion medium. Medium was oxygenated by a commercially available aquarium pump (Hagen, Holm, Germany) and routed through a custom-built glass heating coil flowed with water kept at 40°C via an external heating outlet (Julabo, Seelbach, Germany). The fluid's pressure was measured manometrically (VBM Medizintechnik, Sulz, Germany), and values ranged from 40 to 100 mm Hg. Before entering the liver sample, the perfusion fluid was degassed by means of a bubble trap (Stem Cell Systems, Berlin, Germany). A three-way valve was installed thereafter for collecting perfusion medium before passing the liver sample. The specimen was connected via four branches of the circuit tubing, ending in venous catheters, and the perfusate efflux was collected in a bowl and reconveyed to the medium reservoir. This connection was furnished with another three-way valve for sampling after liver passage and applying any agents. Liver specimens were maintained at 37°C in a water bath.

Fig. 1.

Schematic diagram of the perfusion assembly.

Experimental Perfusion.

Onset of the perfusion procedure ranged from 1 to 19 h (5.6 ± 1.5) after tissue retrieval. First, the liver piece was connected to the perfusion circuit via two to four cannulas, depending on available vessels, adhered into portal or central veins (macroscopically indiscernible) by the tissue adhesive Histoacryl (Braun, Melsungen, Germany). Second, the surgical cutting area was sealed with tissue adhesive. The liver piece was then rinsed with approximately 500 ml of Hanks' balanced salt solution (PAA Laboratories GmbH, Cölbe, Germany) supplemented with 20 mM HEPES (PAA Laboratories GmbH) and 2 U/ml heparin (Ratiopharm, Ulm, Germany) for removing residual blood, then exchanging the perfusion medium for 250 ml of Williams' medium E (Biochrom, Berlin, Germany) containing 20 mM HEPES. The first 50-ml volume of the perfusate was discarded to remove as much residual blood as possible and establish a standardized zero point with fresh perfusion medium. The circuit was closed by placing the effluent tubing into the medium reservoir (Fig. 1). After the initial sampling, the time measurement was started, and flow rates were adjusted to obtain an appropriate pressure of approximately 50 mm Hg (i.e., 40 or 50 ml/min). The perfusion was run for 6 h, and perfusate samples were collected hourly. For determination of cytochrome P450 activity, additional samples were taken at 0.25 and 0.5 h. Glucose and lactate concentrations as well as pH and pO₂ (general metabolism) were measured instantly on a blood gas analyzer (ABL715; Radiometer, Willich, Germany), and the pH was adjusted to 7.3 to 7.4 by adding 1 to 2 ml of 8.4% sodium bicarbonate solution (Braun) to the perfusate as needed. When small (<20 g) pieces of liver were used, the initial pH turned alkaline; in these cases 2 to 3 mM sodium dihydrogen phosphate (Sigma-Aldrich, Munich, Germany) was added. In case the glucose concentration decreased below 40 mg/dl, the perfusate was supplemented with 0.75 to 1.0 ml of a 40% glucose solution (Braun) corresponding to 150 to 200 mg/dl glucose. At the termination of the experiment, 2 ml of trypan blue (Sigma-Aldrich) was added and perfused for 10 min to allow for the evaluation of perfusion efficiency and identification of the perfused tissue areas. Liver specimens were cut into 1-cm slices, and areas of interest were stored in either 4% paraformaldehyde for paraffin embedding with subsequent histochemical staining or liquid nitrogen before cryosectioning.

Biochemical Parameters and Cell Death Markers.

For determining biochemical parameters, perfusate samples were stored at 4°C overnight and then transferred to the Department of Clinical Chemistry and Laboratory Medicine at University Hospital Essen. The activities of ALT, AST, γGT, GLDH, and LDH were determined on the ADVIA 1800 Chemistry System (Siemens Healthcare Diagnostics, Eschborn, Germany) by using the respective assay cassettes. Human albumin was quantified with the AssayPro Human Albumin ELISA kit (AssayPro, St. Charles, MO).

For comparability, measurements were normalized to the per-gram weight of the different liver specimens. Glucose turnover was calculated by subtracting the initial glucose contents of the culture medium from the perfusates' concentrations at times of sampling. Positive values were defined as production or release of a compound, whereas negative values were assumed to reflect consumption, except for oxygen, whose consumption was calculated by determining the partial pressure deviation in the perfusate before entering, and after exiting, the liver specimen (in mm Hg), multiplied by the flow rate (in dl/min) and the Henry constant (0.3 ml × dl⁻¹ × mm Hg⁻¹).

Overall cell death and apoptosis rates were determined with the M65 and M30 ELISA kits (Peviva, Bromma, Sweden) according to the manufacturer's instructions.

Activity of Cytochrome P450.

Cytochrome P450 activity was assessed by adding a cytochrome P450 substrate cocktail consisting of phenacetin (CYP1A1), midazolam (CYP3A4), and diclofenac (CYP2C9) to the medium reservoir directly before starting the time measurement of each experiment. Phenacetin and diclofenac (both from Sigma-Aldrich) were prepared as stock solutions of 80 and 40 mM, respectively, in dimethyl sulfoxide and diluted in culture medium to final concentrations of 26 or 9 μM, respectively, in the perfusion circuit. Midazolam was provided as an aqueous solution (Dormicum; Roche Pharma, Grenzach-Wyhlen, Germany) at 13.8 mM (= 5 mg/ml) and diluted to a perfusion concentration of 3 μM. Samples of 200 μl were taken from below the bubble trap at the indicated time points in the experimental procedure and stored at −20°C until further processing.

The metabolites N-(4-hydroxyphenyl)acetamide (acetaminophen, paracetamol), 1′-hydroxymidazolam, and 4′-hydroxydiclofenac were analyzed by liquid chromatography/mass spectrometry. Before analysis, 50 μl of the sample and standard were precipitated in 150 μl of ice-cold acetonitrile. The samples were placed in the freezer (−20°C) for 20 min and then centrifuged for 20 min at 4000g and 4 °C. Forty microliters of supernatant was mixed with 60 μl of internal standard or diluted 10 times before addition of the internal standard. The internal standard consisted of 200 nM 1′-hydroxymidazolam-¹³C3 (Toronto Research Chemicals Inc., North York, ON, Canada), paracetamol-d4, and 4′-hydroxydiclofenac-¹³C6 (BD Gentest, Woburn, MA). The liquid chromatography system consisted of a HTS PAL injector (CTC Analytics, Zwingen, Switzerland) combined with an HP 1100 LC binary pump and column oven (Agilent Technologies, Waldbronn, Germany). The separation was performed on a reversed 20-phase HyPurity C18 analytical column (50 × 2.1 mm, 5 μm; Thermo Scientific, Runcorn, UK) at 40°C and with a flow rate of 750 μl/min. The mobile phases consisted of 0.1% (v/v) formic acid in 5% acetonitrile (A) and 0.1% (v/v) formic acid in 95% acetonitrile (B). Detection was performed with a triple quadrupole mass spectrometer (API4000) equipped with an electrospray interface (Applied Biosystems/MDS Sciex, Concorde, Canada). Instrument control, data acquisition, and data evaluation were performed by using Applied Biosystems/MDS Sciex Analyst 1.4 software.

Immunohistochemical Staining of Cytochrome P450 Isoenzymes.

Sections of embedded liver tissue were subjected to deparaffinization and antigen retrieval by using citric buffer with heating in a microwave oven for 20 min. Antibodies raised in rabbit against CYP1A1, CYP3A4, and CYP2C9 (Abcam plc, Cambridge, UK) were diluted in phosphate-buffered saline (PAA Laboratories GmbH; 1:10, 1:1000 and 1:100, respectively), and sections were incubated for 1 h. Immunohistochemistry was performed with the HISTAR detection kit (AbD Serotec, Düsseldorf, Germany), which includes 3,3′-diaminobenzidine as chromogenic dye. Cell nuclei were counterstained with hematoxylin (Sigma-Aldrich), and digital images were captured on the Axioplan microscope with the camera Axiocam HRc (Carl Zeiss GmbH, Jena, Germany).

Statistical Analysis.

Parameters, areas under the curve (AUCs), Pearson correlations, and two-way analysis of variance statistical analyses were calculated and graphically displayed by GraphPad Prism 5 (GraphPad Software, San Diego, CA). Fold changes during perfusion were calculated by setting the 1-h parameter values to one. Maximal changes were determined between the lowest and highest value during perfusion, regardless of the time point. All data are given as means ± S.E.M. Statistical significance was assumed at p ≤ 0.05. Grubbs test was performed in Excel 2007 (Microsoft, Redmond, WA).

Results

General Patients' Characteristics in NC and CL Tissue.

NC liver specimens were obtained from resections on 50% female and 50% male patients (mean age 57.2 ± 3.5 years). Liver explants of CL were derived from 75% male and 25% female patients (mean age 53.3 ± 2.7 years; not significant). The underlying diseases varied strongly between the groups (Table 1).

Histochemical Tissue Characterization before and after Perfusion.

Figure 2 shows representative photomicrographs (hematoxylin and eosin stains) of livers from NC (a-c) and CL (d-f) at explant (a and d) and after 6 h of perfusion (b and e). There was no detectable ex vivo cell death in the samples immediately after explant in either group (Fig. 2, a and d). Sections of NC liver tissue (Fig. 2b) and CL liver tissue (Fig. 2e) showed no major cell damage in the areas that were perfused for 6 h, but showed large areas of necrosis in the regions that were not machine-perfused (Fig. 2, c and f).

Fig. 2.

Hematoxylin and eosin stains of NC tissue (a-c) and CL tissue (d-f) upon retrieval (a and d) and after 6-h perfusion of a perfused area (b and e) versus a nonperfused area (c and f). Scale bar, 1 mm.

General Metabolic Characteristics in NC and CL Specimens.

Time courses of glucose metabolism (production versus consumption), lactate production, and oxygen consumption were recorded (Fig. 3). NC tissues provided a rather stable amount of glucose throughout the perfusion process with a maximum of 1.48 ± 0.48 mg × dl⁻¹ × g⁻¹ at 2 h (Fig. 3A). In contrast, CL tissues tended to switch to glucose consumption 1 to 4 h after the initiation of perfusion; the minimum value in this group was at the end of the perfusion time, −1.83 ± 1.87 mg × dl⁻¹ × g⁻¹ (Fig. 3A). Thus, glucose production was approximately 4.5 times higher in NC as derived from the AUCs (Table 2). Although lactate production increased linearly in both groups (Fig. 3B), the slope in CL was steeper compared with NC and differed significantly (p < 0.001; Table 3), indicating a higher production rate in CL tissue. Lactate values increased by 3.4 ± 0.2-fold for NC and 4.6 ± 0.2-fold for CL during the course from 1 to 6 h. Oxygen consumption did not differ between the groups and remained largely constant throughout the observation period, with a maximum of 1.6 ± 0.3 μl × min⁻¹ × g⁻¹ at 4 h for NC tissue (Fig. 3C).

TABLE 2.

Calculation of AUCs from NC and CL liver samples

AUCs indicate overall production or consumption during the experimentation period. AUC ratios demonstrate the factor by which the values differed between NC and CL samples.

Parameter	AUC		AUC Ratio
	NC	CL	NC/CL
Glucose, mg × h × dl−1 × g−1	8.03	1.76	4.56
Lactate, mmol × h × liter−1 × g−1	1.31	2.45	0.53
Oxygen, μl × h × min−1 × g−1	7.55	6.59	1.14
Urea, mg × h × dl−1 × g−1	0.55	0.52	1.05
Albumin, μg × h × ml−1 × g−1	14	14.69	0.95
Bile acids, μmol × h × liter−1 × g−1	0.41	0.85	0.49
LDH, U × h × liter−1 × g−1	374.3	224.5	1.67
GLDH, U × h × liter−1 × g−1	31.37	8.35	3.76
AST, U × h × liter−1 × g−1	276.5	222.9	1.24
ALT, U × h × liter−1 × g−1	191	67.54	2.83
γGT, U × h × liter−1 × g−1	2.78	1.17	2.37
M65, U × h × liter−1 × g−1	6278	11,869	0.53
M30, U × h × liter−1 × g−1	146.3	316.2	0.46

TABLE 3.

Calculation of slope values from NC and CL liver samples

A slope was indicated when at least one r value of the Pearsson correlation analysis was >0.9 (not shown), although the correlation in the related graph did not appear linear. Differences of the slopes are given as p values.

Parameter	Slope		Difference of Slopes
	NC	CL
	average production/h		p
Glucose, mg × dl−1 × g−1 × h−1	r < 0.9	r < 0.9
Lactate, mmol × liter−1 × g−1 × h−1	0.056 ± 0.005	0.122 ± 0.008	< 0.0001
Oxygen, μl × min−1 × g−1 × h−1	r < 0.9	r < 0.9
Urea, mg × dl−1 × g−1 × h−1	0.022 ± 0.003	0.016 ± 0.001	0.0609
Albumin, μg × ml−1 × g−1 × h−1	0.367 ± 0.076	0.154 ± 0.126	0.1705
Bile acids, μmol × liter−1 × g−1 × h−1	0.015 ± 0.003	0.017 ± 0.003	0.7225
LDH, U × liter−1 × g−1 × h−1	22.79 ± 3.19	7.696 ± 1.41	< 0.0001
GLDH, U × liter−1 × g−1 × h−1	2.98 ± 0.52	0.576 ± 0.12	< 0.0001
AST, U × liter−1 × g−1 × h−1	17.28 ± 2.65	8.656 ± 1.52	0.0038
ALT, U × liter−1 × g−1 × h−1	12.48 ± 2.16	2.002 ± 0.54	< 0.0001
γGT, U × liter−1 × g−1 × h−1	0.097 ± 0.022	0.023 ± 0.005	0.0004
M65, U × liter−1 × g−1 × h−1	207.5 ± 66.6	399.4 ± 194.3	0.3876
M30, U × liter−1 × g−1 × h−1	4.83 ± 2.15	13.86 ± 3.99	0.0617

Fig. 3.

General hepatic metabolic parameters. Glucose (A), lactate (B), and oxygen (C) consumption in noncirrhotic (n = 10) and cirrhotic (n = 12) liver specimens were determined in the perfusates. Values were normalized per gram of tissue and are displayed as means ± S.E.M. Missing data sets were caused by mandatory maintenance of the device.

Urea, Albumin, and Bile Acids in NC and CL Specimens.

The rate of urea production increased in both NC and CL tissues over the course of the perfusion (Fig. 4A) with a fold change of 3.2 ± 0.3 and 2.6 ± 0.3 between 1 and 6 h for NC and CL, respectively. The mean concentrations rose by 0.134 ± 0.028 mg × dl⁻¹ × g⁻¹ for NC and 0.092 ± 0.012 mg × dl⁻¹ × g⁻¹ for CL tissue within 6 h, corresponding to comparable slope values (Table 3). The pattern of albumin production was also similar for NC and CL tissue (Fig. 4B), presenting fold changes of 2.0 ± 0.1 and 1.5 ± 0.1, respectively. The maximal gains in albumin concentrations were 2.35 ± 0.5 μg × ml⁻¹ × g⁻¹ for NC and 0.932 ± 0.279 μg × ml⁻¹ × g⁻¹ for CL within 6 h. The secretion of bile acids increased during the perfusion time with nearly matching rates (Table 3), although the initial concentration of bile acids was higher in the CL group and remained significantly elevated at all times (p = 0.0119) (Fig. 4C). Bile acids increased by 5.9 ± 1.8-fold for NC and 1.9 ± 0.2 for CL between 1 and 6 h. Because one noncirrhotic sample presented more than four times higher values than the means of all others, we removed the bile acid data of this sample from the analysis (the graph and the p value representing the complete data set, including this patient, are given in Supplemental Fig. 2).

Fig. 4.

Hepatic synthesis parameters. Production of urea (A), bile acids (B), and albumin (C) by noncirrhotic (n = 10) and cirrhotic liver (n = 12) specimens was measured. Data were normalized per gram of tissue and are shown as means ± S.E.M. Grubb analysis identified the missing data set of bile acids as an outlier (Supplemental Fig. 2).

Liver Enzymes Are Elevated in NC Versus CL.

To assess the maintenance of liver cell integrity during perfusion, established markers of liver damage were determined in the perfusate (Fig. 5). In CL tissues, slight and consistent increases in the activities of all liver enzymes tested were observed over the perfusion time. The elevations of values in CL were 3.8 ± 0.4-fold for AST, 2.6 ± 0.4-fold for ALT, 3.4 ± 0.3-fold for LDH, 10.6 ± 2.6-fold for GLDH, and 1.9 ± 0.4-fold for γGT from 1 to 6 h. In contrast, beginning at 2 to 3 h after the onset of perfusion, NC livers exhibited a stronger increase in these parameters, leading to significant differences for ALT, GLDH, and γGT between the groups. Between 1 and 6 h of perfusion, the values in NC rose by 12.4 ± 4.6-fold for AST, 18.8 ± 8.9-fold for ALT, 10.6 ± 3.5-fold for LDH, and 17.8 ± 4.4-fold for GLDH. In NC specimens, the release of γGT plateaued ∼4 to 5 h after the initiation of perfusion and slightly dropped during the last hour. Thus, the maximal elevation of 3.4 ± 0.7-fold for γGT was found after 5-h perfusion.

Fig. 5.

Liver enzymes released during perfusion. Release of the cell integrity markers LDH (A), GLDH (B), AST (C), ALT (D) and γGT (E) by noncirrhotic (n = 10) and cirrhotic liver (n = 12) specimens. Data were normalized per gram of tissue and are shown as means ± S.E.M.

Ratio of Apoptosis to Necrosis Is More Pronounced in CL.

As a consequence of hepatocellular cell death, intracellular proteins such as cytokeratin 18 are released into the extracellular matrix. Only during apoptosis a neoepitope of CK18 (M30) is generated by caspase cleavage. In the perfusate, the M65 ELISA detects the total amount of cell death, whereas the M30 ELISA specifically accounts for apoptosis (Fig. 6). M65 mean values ranged from 786 ± 229 U × liter⁻¹ × g⁻¹ in NC liver tissue after 1 h to 1611 ± 489 U × liter⁻¹ × g⁻¹ after 6 h of perfusion, resulting in a fold change of 2.8 ± 0.9 (Fig. 6A). Mean values for CL tissue started with 1351 ± 633 U × liter⁻¹ × g⁻¹ after 1 h and plateaued after 4 h with a maximum of 2706 ± 1487 U × liter⁻¹ × g⁻¹ at 6 h of perfusion, which is equivalent to a 1.7 ± 0.2-fold change. The courses of M30 indicated a stronger slope for CL with fold changes of 2.9 ± 0.8 for NC and 3.0 ± 0.5 for CL tissue (Fig. 6B). Although the courses of M65 or M30 did not differ significantly, the M30/M65 ratio was significantly higher in the CL specimens (p = 0.0066; Fig. 6C). This ratio remained fairly stable at 2% for NC throughout perfusion and increased for CL from 3.2% (1 h) to 5.4% (6 h).

Fig. 6.

Cell death parameters throughout tissue perfusion. Display of marker M65 (A), indicating cell death by both necrosis and apoptosis; marker M30 (B), selectively indicating apoptosis, and ratio of the M30/M65 levels (C). Values were normalized per gram of tissue and are displayed as means ± S.E.M.

Metabolic Activity of Cytochrome P450 Enzymes Differ between NC and CL.

As a more detailed measure of hepatic function, we investigated phase-I conversions of three model cytochrome P450 substrates. In the initial 30-min interval, OH-midazolam and OH-diclofenac were formed quite rapidly, but dropped continuously thereafter (CYP3A4, Fig. 7, B and E; CYP2C9, Fig. 7, C and F), caused by formation of the phase-II products. The glucuronides of OH-midazolam and OH-diclofenac were identified by qualitative quadrupole-time-of-flight analysis (data not shown). The formation rates of OH-midazolam, as calculated from the linear sections at the beginning of the curves, were 15.9 ± 5.2 and 6.1 ± 3.7 nM × g⁻¹ × h⁻¹ for NC or CL tissue, respectively; the accordant formation rates of OH-diclofenac resulted in 0.127 ± 0.044 or 0.062 ± 0.019 μM × g⁻¹ × h⁻¹, respectively. The phase-I metabolite of phenacetin was built within the first 2 h and then plateaued (CYP1A1/1A2; Fig. 7, A and D). The formation rates of paracetamol were 0.306 ± 0.09 versus 0.064 ± 0.033 μM × g⁻¹ × h⁻¹ for NC or CL tissue, respectively. Furthermore, the slopes of CYP1A1/1A2 and CYP3A4 within the linear range were not significantly different from zero for CL tissue. As a result of overlapping phase-I and phase-II reactions, the courses of NC and CL tissue intersected for CYP3A4 and CYP2C9 activity once the phase-I reactions were completed in NC. To quantify the extent of both reactions for NC and CL, the AUCs were calculated before and after this intersection (Fig. 7, G–I). For CYP1A1/1A2 activity a corresponding AUC was defined when the curve reached the plateau in NC. In CL tissue, phase-I products of all three cytochrome P450 substrates were generated much slower than in NC tissue, thus leading to deviations in the AUCs from −69.3 to −24.2% (Fig. 7, G–I). In addition, the protracted presence of phase-I products for CL tissue, with higher AUCs up to 53.8%, suggests a much slower formation of phase-II products in CL. None of the observed differences for cytochrome P450 activities in NC and CL tissue reached statistical significance.

Fig. 7.

Concentrations of cytochrome P450 metabolites (phase-I products) in liver tissue perfusate. A to C, the activities of CYP1A1/1A2 (A), CYP3A4 (B), and CYP2C9 (C) were assessed by adding the respective model substrates to the perfusion medium. D to F, to better illustrate the short time periods during which phase-I reactions occurred (early time periods), the respective curve sections, CYP1A1/1A2 (D), CYP3A4 (E), and CYP2C9 (F), are displayed separately. Hence, late time periods correlate with the occurrence of phase-II reactions. G, H, and I, the corresponding areas under the curves are given for CYP1A1/1A2 (G), CYP3A4 (H), and CYP2C9 (I). Lower AUC values in the phase-I diagrams indicate higher phase-II reactivities.

Immunohistochemical stains of the three cytochrome P450 isoenzymes (Fig. 8) revealed stronger enzyme presence in NC compared with CL tissue before the experiment and slightly reduced amounts after 6 h of perfusion in both types of tissue. Although the intensity within positive hepatocytes was approximately the same for NC and CL, the positive tissue area was clearly smaller in CL. Thus, low cytochrome P450 activity in the patients with CL probably can be attributed to diminished intact parenchymal cell mass in cirrhotic liver tissue.

Fig. 8.

Immunohistochemical stains of CYP1A1, CYP3A4, and CYP2C9 in NC and CL tissue. Depicted are representative photomicrographs of NC and CL tissue for each of the three isoenzymes before and after the perfusion experiment. Positive areas were more restricted in CL than in NC samples. After 6 h of perfusion the positive stained regions were reduced in both NC and CL compared with the preperfusion status. Scale bar, 100 μm.

Discussion

Model systems for evaluating drugs generally use single-type cell cultures or animal models (Groneberg et al., 2002; Gómez-Lechón et al., 2003; Elaut et al., 2006). Although valuable, they have their limitations. Specifically, cell lines sometimes critically differ from their healthy counterparts and, on the opposite side of the spectrum, even humanized (chimeric) animal models do not entirely reflect the actual situation in patients. They may differ in their (patho)physiologies, modes of infection, pharmacological metabolization, and toxicity characteristics and are, moreover, especially fragile (Tateno et al., 2004; Dandri et al., 2005; Vanwolleghem et al., 2010; Lütgehetmann et al., 2012). Such limitations may be remedied with a suitable human-based perfusion system, even more so if such a system can maintain certain pathological conditions. This demand can be met with the presented perfusion system.

Comparison with Other Ex Vivo Systems.

Our system for ex vivo human liver perfusion offers a host of research applications combined with a strong economic advantage compared with perfusion machines that are usually used for preserving organ transplants (Taylor and Baicu, 2010). Generally, such systems cool down the tissue to avoid degeneration and cell death, use expensive protective solutions to limit cold ischemia/reperfusion injury, and require a sterile transportation device. In contrast, our system avoids expensive factors and components that are required for prolonged graft storage but are not mandatory for short-term tissue survival.

To compare the tissue's synthesis performance with other hepatocyte culture systems, reported 1-day production rates were converted into hourly values. Corresponding parameters obtained in NC liver tissue were calculated by using the slopes of the curves (urea, albumin, and lactate) or the AUC (glucose). Our values were normalized per gram of tissue, so that our calculations are based on 5 × 10⁷ cells/cm³ of liver (as determined in a biopsy punch; data not shown). In vivo hepatic production rates of urea (65 μg/h per 10⁷ cells) and albumin (20–30 μg/h per 10⁷ cells) were reported previously (Bhatia et al., 1999). Another important factor is that hepatocytes account for ∼80% of the cellular volume and ∼60% of the hepatic cell number (Kmieć, 2001), whereas liver cells cultured in two-dimensional or three-dimensional environments consist mainly of hepatocytes. Our perfusion system revealed a mean urea production of 7.1 μg/h per 10⁷ cells and a mean albumin production of 11.7 μg/h per 10⁷ cells. When considering the perfused areas, these albumin values match with in vivo data, whereas the low urea production probably reflects reduced protein degradation. Bioreactor cultures produced 44.8 μg/h per 10⁷ cells of urea and 2 μg/h per 10⁷ cells of albumin on day 1 (Zeilinger et al., 2011) or under serum-free conditions 41.8 μg/h per 10⁷ cells of urea and 0.3 μg/h per 10⁷ cells of albumin (Mueller et al., 2011). The discrepancy in albumin production may be explained by the fact that the detectable amount of albumin (and other proteins) under serum-free conditions is reduced by coating the capillaries. On day 1, comparable values of urea (7.5 μg/h per 10⁷ cells) and albumin (3.8 μg/h per 10⁷ cells) were actually reported in two-dimensional cultures (Bhogal et al., 2011). Glucose release was calculated as 92 μg/h per 10⁷ cells in the bioreactors (with high intersample variances), as opposed to 428 μg/h per 10⁷ cells in perfused liver specimens. Lactate production in bioreactors was 60.2 μg/h per 10⁷ cells, in contrast to 160 μg/h per 10⁷ in our system. Whether this discrepancy indicates elevated hepatocyte activity or insufficient oxygen supply will be subject to future investigation.

Comparison with the In Vivo Situation.

Using the system described herein, the parameters measured in non cirrhotic and cirrhotic tissue closely resembled the (patho-) physiological characteristics in vivo.

Increased lactate and bile acid production rates paralleled the situation in patients with CL (Pennington et al., 1977; Akashi et al., 1987; Almenoff et al., 1989; Bernal et al., 2002). The fact that cirrhotic samples initially showed increased glucose production, which later switched to consumption (as opposed to NC samples with a constant production rate), reflects the reduced glycogen storage capacity of cirrhotic livers (Krähenbühl et al., 2003).

In NC, parameters of liver damage (AST, ALT, LDH, GLDH, and γGT) rose much more clearly than in CL. Likewise, chronic liver damage clinically leads to a continuous decline in cell numbers. This is paralleled by a decreasing release of liver enzymes (Canbay et al., 2009), which is sharply contrasted by the sudden and extensive organ damage in acute liver failure (Bechmann et al., 2010). Consequently, high ALT and AST values are associated with better outcome upon acute liver failure (Canbay et al., 2009).

Limitations.

Applications of the novel system may be limited by the availability of appropriate tissue samples. Our perfusion studies were performed at a major hepatological center with a highly experienced liver transplantation department and well attuned personnel. In contrast, smaller sites might not have access to sufficiently large specimens of patients' livers and the desirable infrastructural preconditions, which is a bottleneck situation that also applies to the provision of human samples for isolating primary liver cells. It is noteworthy, however, that this limitation does not argue against using the novel system to improve the conditions for basic and applied hepatologic research in suitable clinical settings.

Absence of other organs or organ systems as, for example, the immune system, confine the available information to liver-related processes. Other in vivo models suffer from the same limitation.

Another current limitation arises from incomplete tissue perfusion under the conditions used; according to trypan blue staining, 50 to 80% of the tissue was adequately perfused. Nevertheless, data normalized to the liver samples' weights were consistent within the NC and CL groups with only small S.D., so that comparable tissue fractions were perfused in each sample. During the perfusion period, liver samples showed a trend toward cell damage at the later points in time. Both this drawback and the overall ex vivo perfusion time shall be further improved by enhancing oxygen transport and/or delivery and continuously providing glucose and additional nutrients.

Benefits.

The novel perfusion system offers some decisive benefits, e.g., for preclinical drug testing in an original, patient-derived setting. It will thus be especially valuable for investigations requiring intact hepatocyte polarity and/or the hepatocytes' interplay with nonparenchymal cell species. Rapid phase-I conversions of three human-relevant cytochrome P450 substrates clearly proved that the hepatocytes were functionally active and performed detoxification reactions. Moreover, rapid consumption of the midazolam and diclofenac phase-I metabolites confirmed the hepatocytes' intact phase-II metabolism. Both types of reactions were more pronounced in NC than in CL specimens. These results are in line with the in vivo situation in such patients and are a considerable improvement over most culture systems. For example, a similar cytochrome P450 testing setup with bioreactor-cultured primary human hepatocytes exhibited comparable phase-I formation clearances, but significantly lower phase-II reactions (Mueller et al., 2011; Zeilinger et al., 2011).

A comparable perfusion system published previously (Melgert et al., 2001) varied in some aspects: First, that system differed in oxygenation and usage of perfusion solution (Krebs-Henseleit buffer), and tissue pieces were stored in cold UW solution for 6 to 39 h before perfusion. In contrast, our procedure used a cold-storage period between tissue retrieval and onset of perfusion of 1 to 19 h (5.6 ± 1.5). Second, in the other system, functional assessment of liver tissue was performed with fewer samples (seven, NC; four, CL) for 1 h as opposed to our perfusion of 10 or 12 specimens, respectively, for 6-h periods, thus allowing for measurements or treatments that require prolonged incubation or reaction times.

The system introduced herein thus offers substantial advantages for the basic characterization of liver function. It opens up options for the ex vivo evaluation of metabolic conditioning and the investigation of innovative treatments for nonalcoholic steatohepatitis, the viral hepatitides, and hepatocellular carcinoma as increasingly relevant disease entities in countries with Westernized lifestyle habits as well as, partly, in Middle Eastern and North African countries (Angulo, 2007; Ertle et al., 2011; Nelson et al., 2011; Rahbari et al., 2011). Moreover, the system can also be used for evaluating investigational new drugs at close-to-human conditions with higher reliability.

Conclusions

By introducing this perfusion system and the baseline characterization of liver specimens derived from different disease entities, we hope to open up new possibilities not only for hepatologic investigations. Continued improvement as to ease of handling, oxygen/nutrient supply, and complementary data on liver tissue parameters may enable a wider range of applications. The informative power of this ex vivo perfusion system may also be enhanced by integrating an existing cutting-edge multigene expression signature that predicts hepatotoxicity (Cheng et al., 2011). Such a follow-up system may pinpoint undesired or dangerous phase-I and/or phase-II metabolites of investigational new drugs with high reliability. Both the present-stage system and its successors are anticipated to provide excellent models for investigating the overall performances and adverse effects of promising new drug candidates in the context of liver disease.

Authorship Contributions

Participated in research design: Schreiter, Marquitan, Sowa, and Gieseler.

Conducted experiments: Schreiter and Marquitan.

Contributed new reagents or analytic tools: Darnell, Bröcker-Preuss, Andersson, Baba, Mathe, and Treckmann.

Performed data analysis: Schreiter, Marquitan, Darnell, and Sowa.

Wrote or contributed to the writing of the manuscript: Schreiter, Darnell, Sowa, Andersson, Furch, Arteel, Gerken, Gieseler, and Canbay.