Human liver cell spheroids in extended perfusion bioreactor culture for repeated-dose drug testing


  • Potential conflict of interest: Dr. Björquist owns stock in Cellartis.

  • This work was supported by FCT fellowships SFRH/BD/35296/2007 (to R.T., under the scope of the MIT-Portugal program) SFRH/BD/37102/2007 (to S.L.), project PTDC/EBB-BIO/112786/2009, and by the European Commission under the scope of the Hyperlab project (High Yield and Performance Stem Cell Lab) (HEALTH-2007-1.4-7).


Primary cultures of human hepatocyte spheroids are a promising in vitro model for long-term studies of hepatic metabolism and cytotoxicity. The lack of robust methodologies to culture cell spheroids, as well as a poor characterization of human hepatocyte spheroid architecture and liver-specific functionality, have hampered a widespread adoption of this three-dimensional culture format. In this work, an automated perfusion bioreactor was used to obtain and maintain human hepatocyte spheroids. These spheroids were cultured for 3-4 weeks in serum-free conditions, sustaining their phase I enzyme expression and permitting repeated induction during long culture times; rate of albumin and urea synthesis, as well as phase I and II drug-metabolizing enzyme gene expression and activity of spheroid hepatocyte cultures, presented reproducible profiles, despite basal interdonor variability (n = 3 donors). Immunofluorescence microscopy of human hepatocyte spheroids after 3-4 weeks of long-term culture confirmed the presence of the liver-specific markers, hepatocyte nuclear factor 4α, albumin, cytokeratin 18, and cytochrome P450 3A. Moreover, immunostaining of the atypical protein kinase C apical marker, as well as the excretion of a fluorescent dye, evidenced that these spheroids spontaneously assemble a functional bile canaliculi network, extending from the surface to the interior of the spheroids, after 3-4 weeks of culture. Conclusion: Perfusion bioreactor cultures of primary human hepatocyte spheroids maintain a liver-specific activity and architecture and are thus suitable for drug testing in a long-term, repeated-dose format. (HEPATOLOGY 2012)

The liver-specific functions of hepatocytes, such as albumin secretion or drug-metabolizing activity, are rapidly down-regulated during in vitro primary cultures, limiting their use for drug development and toxicity tests.1 For such assays, the current gold standard for long-term human hepatocyte culture is the collagen sandwich in vitro model.2 The overlaying collagen layer increases cell-cell and cell-matrix contacts, providing a more three-dimensional (3D)-like architecture than a monolayer culture. For rat hepatocyte spheroids, where cell-cell interactions are maximized, liver-specific functions3, 4 and multicellular architecture5, 6 are increased, when compared to monolayer cultures.

The use of microfluidic devices for primary cultures of hepatocytes is a promising approach to enable high-throughput screening in drug development.7, 8 However, the downscaling enabled by these technologies makes the culture environment harder to be controlled and limits the application of microfluidics for long-term primary cultures of hepatocytes. In fact, the most useful applications of microtechnologies for such cultures couple either microfluidic perfusion or coculture micropatterning to 12-9 or 24-well culture plates,10 respectively; still, these technologies do not enable a physiologically relevant long-term culture of primary hepatocytes.

Bioartificial liver devices, using human hepatocytes, are often built in hollow fiber formats; this configuration is likely among the best options for maintaining cultures of large numbers of hepatocytes for prolonged culture periods11; recently, one such device has also been scaled down and adapted to drug testing, using human liver cells.12 The configuration of these bioreactors enables the formation of a liver-like hepatic mass throughout the culture time, making them long-term endpoint assays; nevertheless, these bioreactors were not designed for high-throughput screening and do not allow sampling of the cellular mass throughout the culture period.

Cytochrome P450 (CYP450) expression is down-regulated during in vitro primary culture of hepatocytes; these enzymes are fundamental to xenobiotic metabolization studies, namely in drug development, and the orphan receptor-mediated induction of their messenger RNA (mRNA) synthesis is one of the most important parameters to be assessed in such tests.13 After these enzymes have oxidized a given xenobiotic, the compound is further conjugated with polar groups by phase II enzymes and secreted in the bile canaliculi by phase III enzymes. Thus, a long-term hepatocyte culture system must not only maintain CYP450 basal expression, but must also enable their de novo mRNA synthesis upon exposure to prototypical CYP450 inducers while maintaining the remaining phase II and III activities. Because phase III activity depends on the transport of phase II metabolites through the apical membrane, the presence of bile canaliculi is also necessary for a thorough assessment of drug-metabolizing activity in primary cultures of hepatocytes.

This work focused on the use of perfusion-stirred tank bioreactors for primary cultures of human hepatocytes as spheroids. When cultured in a bioreactor with essentially convective mass transfer and environmental control, hepatocytes experienced much smaller changes in pH, dissolved oxygen (DO), and culture medium composition than any culture system with a constant atmosphere and discrete medium exchanges. pH and DO are controlled by CO2 and N2 injection through the reactor headspace, respectively, whereas the continuous addition of nutrients and removal of metabolic by-products is ensured by the automated perfusion system14; the good mixing minimizes the gradients of these soluble factors in the culture bulk. The results indicate that this bioreactor system of primary culture of human hepatocyte spheroids enables the robust formation of hepatic-like microtissue units that can be repeatedly induced in long-term periods.


2D, two-dimensional; 3D, three-dimensional; 7-EC, 7-ethoxycoumarin; 7-HC, 7-hydroxycoumarin; ANOVA, analysis of variance; aPKC, atypical protein kinase C; BNF, β-naphtoflavone; CDFDA, 5-(and-6)-carboxy-2′,7′-dichlorofluorescein diacetate; CK18, cytokeratin 18; CYP450, cytochrome P450; DAPI, 4′,6-diamidino-2-phenylindole; dave, average diameter; DO, dissolved oxygen; 7- ECOD, ethoxycoumarin O-deethylation; EDTA, ethylene dimaine tetraacetic acid; ELISA, enzyme-linked immunosorbent assay; HCV, hepatitis C virus; HNF4α, hepatocyte nuclear factor 4 alpha; GSTA-1, glutathione S-transferase alpha 1; mRNA, messenger RNA; PBS, phosphate-buffered saline; PFA, paraformaldehyde; Rif, rifampicin; RT, room temperature; SEM, standard error of the mean; UGT2B7, UDP-glucuronosyltransferase-2B7.

Materials and Methods

Primary Cultures of Human Hepatocytes.

Human liver samples originating from patients undergoing liver resection (Supporting Table 1), were obtained from Sahlgrenska Hospital (Gothenburg, Sweden) upon written consent in agreement with ethical approval and from the patient with a signed informed consent agreement. Human liver samples were isolated by a two-step ethylene diamine tetraacetic acid (EDTA)/collagenase type IV perfusion, followed by slow-speed centrifugation to reduce red blood cell content. Cells were resuspended in Williams' E medium and stored overnight (shipped overnight) at 4°C in Williams' E medium before use. At the beginning of the experiments (i.e., upon arrival), cell viability was higher than 80%, as determined by trypan blue exclusion, using a Fuchs-Rosenthal counting chamber.

Perfusion Bioreactor Culture.

Primary cultures of human hepatocytes were performed in Williams' E medium, supplemented with 1% GlutaMAX, 1% penicillin/streptomycin (all from Gibco/Invitrogen, Grand Island, NY) and the Hepatocyte Culture Medium (HCM) SingleQuots kit (Lonza, Walkersville, MD). Bioreactor cultures were inoculated at a cell concentration of 0.2 × 106 viable hepatocyte/mL. To promote cell aggregation into spheroids, Williams' E complete medium was supplemented with 10% fetal bovine serum; cells were cultured in this medium for 72 hours and in serum-free conditions after that (Fig. 1); hepatocyte culture viability after aggregation was always above 90%, as assessed by the trypan blue exclusion method. Initial bioreactor working volume (V) was 300 mL, and the perfusion rate (F) was set to 60 mL/day; this rate was adjusted throughout the culture time (as the culture volume decreased as a result of sampling) to maintain a dilution rate (D) of 0.2 day−1 (i.e., 20% medium exchange per day; D = F/V). Bioreactor (BIOSTAT Qplus system; Sartorius Stedim Biotech S.A., Aubagne, France) cultures were controlled at 37°C (pH = 7.4) and DO at 30% air saturation (approximately 60 μM of O2 or 6.3% oxygen in a controlled atmosphere incubator, assuming an efficient mass transfer to culture bulk).

Figure 1.

Experimental design of the induction of the CYP450 enzymes in primary cultures of hepatocyte spheroids in the bioreactor. Scale bar = 50 μm.

Hepatocyte CYP450 Enzyme Inductions.

Inductions of the hepatocyte spheroids' CYP450 enzymes were performed (Fig. 1). Briefly, inductions in bioreactors were started by adding rifampicin (Rif) and β-naphtoflavone (BNF) at concentrations of 10 and 25 μM, respectively. Bioreactors were perfused with culture medium, containing the same concentrations of both inducers, at a dilution rate of 0.5 day−1 (2.5 times the dilution rate used for normal perfusion culture), for 72 hours; after this period, culture medium was fully exchanged to ensure a complete removal of the induction medium. This induction procedure was performed twice, at day 3 and 2-4 weeks later, for all donors (Fig. 1).

Cell-Concentration Determination.

Spheroids were digested with 0.05% Trypsin/EDTA (Gibco), and the resulting single-cell suspension viability was assessed by the trypan blue exclusion method. Cell counting was performed using a Fuchs-Rosenthal counting chamber.

Hepatocyte Spheroids Visualization and Measurement.

Hepatocyte spheroids were visualized by bright field microscopy (Leica Microsystems GmbH, Wetzlar, Germany), and their average diameter (dave) was determined by a geometric mean of three diameters per spheroid, using the following equation: dave = (d1 × d2 × d3)1/3; spheroid diameters were measured using ImageJ software. Diameter distribution plots were done using GraphPad Prism software (GraphPad Software, Inc., La Jolla, CA).

Determination of Albumin and Urea Synthesis.

Secretion of albumin from hepatocytes was measured by an enzyme-linked immunosorbent assay (ELISA), using the Exocell Albuwell albumin test kit (Exocell Inc., Philadelphia, PA). The assay was performed according to the manufacturer's description. Urea synthesis rate was determined using a quantitative colorimetric urea kit (QuantiChrom Urea Assay Kit, DIUR-500; BioAssay Systems, Hayward, CA), according to the manufacturer's instructions. Albumin and urea-specific synthesis rates were calculated according to the general mass balance equation for a continuous system: q = (ΔC/Δt – D × (Cin – Cout))/[X]V average, where q is the specific synthesis rate, ΔC/Δt is the rate of change of the metabolite (either urea or albumin) in the supernatant, D is the dilution rate (0.2 day−1), Cin and Cout are the inlet and outlet concentrations of the metabolite, and [X]V average is the average viable cell concentration during Δt. Results are expressed as μg/106 cell/day at the indicated time points.

CYP450 Activity Measurement.

7-Ethoxycoumarin O-deethylation (ECOD) activity was measured using the method previously described,15 with slight modifications. Briefly, salicylamide (1.5 mM) was added to the medium to prevent conjugation of 7-hydroxy metabolites (i.e., 7-hydroxycoumarin; 7-HC) of 7-ethoxycoumarin (7-EC). Activity was measured by the rate of formation of 7-HC (umbelliferone) in nmol/106 cell/day.

Quantitative Reverse-Transcription Polymerase Chain Reaction.

Hepatocyte spheroids were collected from bioreactor cultures at different time points and stored at −20°C with RNAprotect Cell Reagent (Qiagen, Valencia, CA). Later, total RNA was extracted using the RNEasy Plus Mini Kit (Qiagen), according to the manufacturer's instructions. Reverse transcription was performed using 0.6 ug of total RNA in a final volume of a 20-uL reaction mix using the High Capacity cDNA Reverse Transcriptase Kit (Applied Biosystems, Foster City, CA). Real-time polymerase chain reaction was performed using ready-to-use TaqMan Gene Expression Assays (Applied Biosystems), according to the manufacturer's instructions. Glyceraldehyde 3-phosphate dehydrogenase was used as the endogenous control.

Whole Mount Immunofluorescence Microscopy.

Hepatocyte spheroids were fixed in 4% (w/v) paraformaldehyde (PFA) in phosphate-buffered saline (PBS) for 1 hour at room temperature (RT), blocked overnight at 4°C in 0.1% Triton X-100 (Sigma-Aldrich, St. Louis, MO) and 0.2% fish-skin gelatin solution in PBS, and subsequently incubated with primary antibodies diluted (1:100) in 0.125% fish-skin gelatine in PBS for 2 days at 4°C. Cells were washed three times with PBS, and secondary antibodies (diluted 1:500 in 0.125% fish-skin gelatin in PBS) were applied to cells overnight at 4°C in the dark. After three washes with PBS, samples were mounted in ProLong Gold antifade (Invitrogen, Carlsbad, CA), containing 4′,6-diamidino-2-phenylindole (DAPI). Cells were visualized using point-scanning (SP5; Leica) or spinning-disk (Andor Technology PLC, Belfast, Northern Ireland) confocal microscopy.

Cryosection Immunofluorescence Microscopy.

Hepatocyte spheroids were frozen in O.C.T. Tissue Tek (Sakura Finetek Europe B.V., Zoeterwoude, The Netherlands) and sectioned in 10-μm-thick slices onto glass coverslips. These coverslips were blocked for 10 minutes at RT in 0.1% Triton X-100 (Sigma-Aldrich) and 0.2% fish-skin gelatin solution in PBS and were subsequently incubated with primary antibodies diluted (1:100) in 0.125% fish-skin gelatin in PBS for 2 hours at 4°C. Cells were washed three times with PBS, and secondary antibodies (diluted 1:500 in 0.125% fish-skin gelatin in PBS) were applied to cells for 1 hour at RT in the dark. After three washes in PBS, samples were mounted in ProLong gold antifade (Invitrogen), containing DAPI. Cells were visualized using a Leica DMI 6000 epifluorescence microscope.

Antibodies for Confocal Immunofluorescence Microscopy.

Primary antibodies used were goat antialbumin, mouse anti-HNF4α (hepatocyte nuclear factor 4 alpha) (Abcam, Cambridge, UK), goat anti-CYP450 3A, mouse anti-aPKC (atypical protein kinase C) (Santa Cruz Biotechnology, Santa Cruz, CA), fluorescein isothiocyanate–conjugated anti–cytokeratin 18 (CK18; Sigma-Aldrich), and Alexa Fluor 488–conjugated Phalloidin (Molecular Probes, Eugene, OR). Secundary antibodies used were antimouse Alexa Fluor 488, antigoat Alexa Fluor 594 (Molecular Probes), and antimouse Cy5 (Abcam).

Bile Canaliculi Functionality.

Human hepatocyte spheroids were collected from the reactor after 2 weeks of culture, washed with PBS, and incubated for 10 minutes in 2 μg/mL of 5-(and-6)-carboxy-2′,7′-dichlorofluorescein diacetate (CDFDA; Molecular Probes) in Williams' E medium. After, spheroids were washed three times with PBS and imaged by confocal spinning-disk microscopy (Andor).

Statistical Analysis.

Unless otherwise stated, all results were subjected to an analysis of variance (ANOVA) single-factor analysis, with α = 0.05, using Microsoft Excell's data-analysis toolpack (Microsoft Corporation, Redmond, WA); P values are presented for statistically significant results (P < 0.05).


Hepatocyte Spheroid Size Distribution, Viability and Liver-Specific Synthesis Time-Course Profiles in Bioreactor Cultures.

Hepatocyte spheroid diameter is a critical variable for the maintenance of viability16 and hepatic phenotype17 in primary rat hepatocyte cultures. Thus, average spheroid diameters in bioreactor cultures were measured based on phase-contrast images after 1 and 2 weeks of culture (Fig. 2a,b, donor c), yielding a size-distribution plot (Fig. 2c). On average, spheroid diameters were 65 ± 7 μm after 1 week and 81 ± 4 μm (value ± standard error of the mean; SEM) after 2 weeks, for the 3 different donors (Fig. 2d), and viability of primary bioreactor cultures of hepatocyte spheroids was maintained above 70% of inoculated viable hepatocytes without major cell death during culture time (Fig. 2d). Urea and albumin secretion, two liver-specific functions essential for ammonia detoxification and for maintaining blood osmotic pressure, respectively, were analyzed during the hepatocyte spheroid bioreactor culture; time-course profiles for urea (Fig. 3a) and albumin (Fig. 3b) synthesis were comparable and reproducible among the 3 donors (Fig. 3). The specific albumin production rate increased steadily along the 15-day culture time, whereas urea production decreased from the onset of the culture, reaching a steady state after 1 week of bioreactor culture. For both liver-specific activities, as much as 10-fold interdonor variability was observed.

Figure 2.

Hepatocyte spheroid diameter distribution and cell viability. Bright-field pictures of human hepatocyte spheroids in the first (a) and second (b) week of culture show an increase in average diameter, as quantified in (c) (donor C, n = 100 spheroids; bars = 200 μm). Average diameter hepatocyte spheroids and average cell-culture viability (in % of viable inoculated hepatocytes) are depicted in (d). Bars in (d) represent standard deviations of n = 3 donors.

Figure 3.

Urea and albumin synthesis in bioreactor hepatocyte spheroid cultures. Bioreactor cultures of human hepatocyte spheroids show reproducible profiles for urea (a) and albumin (b) synthesis for donors A (○), B (□), and C (▵).

Phase I and II Enzyme Expression and Activity in Bioreactor Cultures.

Hepatic drug metabolization typically involves phase I monooxygenase activity (i.e., CYP450 activity), followed by phase II conjugation activity and the transport of hydrophilic metabolites by phase III transporters. Gene expression for three CYP450 isoforms (e.g., 1A2, 2C9, and 3A4) and two conjugation enzymes (e.g., glutathione S-transferase alpha 1 [GSTA1] and UDP-glucuronosyltransferase-2B7 [UGT2B7]) was measured, for the 3 donors, after spheroid aggregation (i.e., at day 3; Fig. 4a) and during the remaining culture time (Fig. 4b-d); in addition, two prototypical inducers, Rif and BNF, were added to cultures in a repeated dose (Fig. 4b-d, dashed boxes). Differences between donors for gene expression of these drug-metabolizing enzymes may vary by 10- to 100-fold2; this interdonor variability is depicted in Fig. 4a, where gene expression for each enzyme was normalized to donor A values. Thus, normalization of relative gene expression to day 3 is required to compare the temporal evolution of the system between donors (Fig. 4b-d). An increase in mRNA expression of drug-metabolizing enzymes was observed at day 5, after the addition of the inducer cocktail (at day 3), for all donors; these expression levels decreased after the first induction period, for donors A and C, whereas for donor B, not all enzymes had a reduction in expression level. The second induction period was performed 72 hours before the end of cultures (between weeks 2 and 4), whereas for donor A and C cultures there was a general up-regulation of gene expression, and donor B culture only showed an increase in CYP3A4 mRNA after this induction.

Figure 4.

Interdonor variability and time-course profiles of phase I and II enzymes in human hepatocyte spheroid bioreactor cultures. Gene expression of CYP450/phase I enzymes IA2, 2C9, and 3A4 and phase II enzymes GSTA1 and UGT2B7 was measured at day 3 (when spheroids were formed) and normalized to donor A values (a); the same genes' expression was monitored for the remaining culture time for donors A (b), B (c), and C (d). Gene-expression values in (b-d) were normalized to respective gene expression at day 3. CYP4501A2 (□), 2C9 (▵), and 3A4 (▿); phase II enzymes GSTA1 (○) and UGT2B7 (•). Dashed areas represent induction period, when spheroids were exposed to both Rif (10 μm) and BNF (25 μm). Bars represent standard deviations of three different samples from the bioreactor.

Figure 5.

Induction of ECOD activity in bioreactor (see also Fig. 1 for methodology). ECOD activity of bioreactor cultures was assessed after 48 and 72 hours of Rif and BNF exposure. Bars represent SEM of at least three independent cell-based measurements; ANOVA was performed by comparing 0- (basal) and 48- or 72-hour values, for each donor. *P < 0.05; ***P < 0.01.

CYP450 activity of hepatocyte spheroids was measured by the metabolization of 7-EC to 7-HC, a reaction mainly catalyzed by the CYP1A family, even though CYP2s and CYP3s are also involved in this deethylation reaction.18 Samples were collected from the bioreactors at 48 and 72 hours, during the first induction period (Fig. 5); the ECOD activity of all the 3 donors significantly increased either by 48 (donor B) or 72 hours (donor A) or at both time points (donor C). These fold increases ranged from 2.6 for donors A (72 hours) and B (48 hours) to 19 and 15 for donor C (at 48 and 72 hours, respectively).

Liver-Specific Markers and Structural Polarity in Bioreactor Cultured Spheroids.

Human hepatocyte spheroids cultured in fully controlled bioreactors, from donors A and C, were analyzed by immunofluorescence microscopy to assess the presence of HNF4α, CK18, albumin, CYP4503A, and polarity markers inside such spheroids. For donor A, both albumin (Fig. 6b, red) and CYP4503A (Fig. 6c, red) were still detectable at 30 days of culture, as well as CK18 (Fig. 6b,c, green); donor C culture stained positive for HNF4α after 20 days in culture (Fig. 6a). In donor D hepatocyte spheroids, actin (phalloidin) staining (Fig. 7a, green) showed an absence of stress fibers, with most of these filaments localizing at the intercellular borders; furthermore, an actin enrichment could be observed, in some cell-cell contacts, which formed canaliculi-like structures (Fig. 7a), very similar to the in vivo liver-tissue architecture.19 The establishment of de novo polarity is more obvious when donor C hepatocyte spheroids were immunostained for aPKC, a kinase associated with the apical domain of epithelial cells (Fig. 7c, green): A series of 24-μm confocal Z-sections showed a bile canaliculi network, which extended to the inner part of the spheroid (Fig. 7c; Supporting Video 1). The functionality of these channels was assessed by imaging the excretion of CDFDA into the canaliculi after being metabolized by the hepatocytes' intracellular esterases (Fig. 7b), in 2-week-old spheroid cultures of donor D.

Figure 6.

Immunofluorescence microscopy of liver-specific antigens in human hepatocyte spheroids after 2 weeks of bioreactor culture. (a) HNF4α (green) colocalizes with nuclear DAPI staining (blue). (b) Albumin (red) and CK18 (green). (c) CYP450 isoform 3A (red), CK18 (green), and nuclei (DAPI, blue). Samples for (a) and (b) were prepared in whole mount, whereas samples for (c) were prepared as 10-μm-thick cryosections.

Figure 7.

Fluorescence microscopy of structural and polarity markers and bile canaliculi function in human hepatocyte spheroids after 2 weeks of bioreactor culture. (a) F-actin (green) localizes to cell membranes and is enriched in bile-canaliculi–like structures. (b) Transport of CDFDA to the apical (canalicular) domain of hepatocytes shows an extensive, interconnected canaliculi network. (c) Confocal z-sections (numbers represent the distance from the spheroid surface, in μm) of a hepatocyte spheroid stained for aPKC (green) and nuclei (DAPI, blue) show several bile-canaliculi–like channels, which extend to the interior of the spheroid.


In this work, a perfused bioreactor system for long-term maintenance of primary cultures of human hepatocyte spheroids was established and tested. In this system, hepatocyte spheroids reproducibly recapitulated in vivo hepatic functions and structure, despite interdonor variability. We hypothesize that these reproducible time-course profiles were made possible because of the tight control of critical environmental variables at physiological values, such as pH and oxygen levels. mRNA expression of CYP450 (phase I) as well as GSTA1 and UGT2B7 (phase II) was maintained up to 4 weeks and increased when the cultures were exposed to the prototypical CYP450 inducers, Rif and BNF, in repeated doses. Phase I ECOD activity of such cultures also responded to such inducers, showing the system's potential for more informative time-course experiments. The spheroid's inner structure resembled the liver architecture, with functional bile canaliculi-like structures and liver-specific markers. Such a system constitutes an ideal long-term culture platform for analyzing hepatic function for drug development tests.

The formation of hepatocyte spheroids was performed during the first 72 hours of culture; phase-contrast microscopy data of this period (data not shown) suggest an initial 24-hour period of cell clustering, when small aggregates (40-50 μm) are formed; these clusters grow in size during the following 2 days (i.e., until 72 hours), as previously reported for primary cultures of rat hepatocyte spheroids.20

The influence of oxygen concentration in primary cultures of hepatocytes has been the subject of several publications,16, 21-23 and the conclusions from these studies are not easily comparable. This is mainly the result of the different culture systems used (i.e., static systems24 need higher oxygen concentrations for hepatocyte culture because the mass transfer relies on diffusion); on the other hand, stirred systems, such as the bioreactor described herein, have convective mass transfer and nearly homogeneous DO concentrations in the culture bulk. Thus, the bioreactor culture oxygen concentration used in this work (30% of air saturation in culture medium, i.e., 60 μM) is in the interval between the known periportal and pericentral oxygen concentration in the rat liver, 90 and 45 μM, respectively.25

The bulk aggregation of hepatocytes into multicellular spheroids depends on parameters such as agitation type, vessel geometry,3, 26, 27 and cell inoculum.3 Brophy et al.26 have shown that rat hepatocyte spheroids could be obtained from single cells by rocking motion with a higher yield of spheroids (85%), when compared to rotational motion (54%). However, this rotational motion was based on shake-flask cultures, and previous work by Wu et al.20 has shown that spinner vessels could yield rat hepatocyte spheroids with an 80% efficiency in the incorporation of single cells into spheroids, after 72 hours. Differences between fluid dynamics in the orbitally shaken flasks and spinner vessels or the different cell inoculum used (1 and 0.3 million hepatocytes per mL, respectively) could explain the difference in both publications. In the work described herein, the previous knowledge from our group concerning the aggregation of rat hepatocytes3, 14, 23 was adapted to the formation of multicellular spheroids in primary cultures of human hepatocytes. For these cultures, the control of spheroid size becomes critical to avoid putative nutrient diffusion limitations; in the system described herein, hepatocyte spheroids had an average size of 81 μm (Fig. 2d), and the number of spheroids with a diameter larger than 200 μm was less than 0.4% of the total population (n = 3 donors). From previously published studies, it is known that rat hepatocyte spheroids with diameters of 100 μm yield a higher albumin production rate than larger ones,17 whereas spheroids of up to 200 μm in diameter have been shown not to be subject to nutrient (e.g., oxygen) limitations.16 Because the aggregation process described here reproducibly yields spheroids with an average 81-μm diameter, it is not expected that hepatocytes within these spheroids are subject to any significant mass transfer limitations. The data obtained for liver-specific activities (e.g., urea and albumin production, Fig. 3; ECOD activity, Fig. 5) and gene expression (CYP450 and phase II enzymes, Fig. 4) confirm the interdonor variability, which has been thoroughly described for primary human hepatocyte cultures and is a direct reflection of in vivo variability1, 28; however, the tight control of critical variables by the perfusion bioreactor system, coupled to an easy cell-sampling system, allowed reproducible liver-specific profiles to be obtained, despite interdonor variability; the use of serum-free media after aggregation is also a likely cause for such a reproducible behavior, because serum is known to down-regulate both albumin synthesis and CYP450 activity of primary cultures of human hepatocytes.22 The profiles of urea-secretion rate for the 3 donors show a significant decrease from the beginning of the cultures. This reduction in urea productivity has been observed by Zeillinger et al.12, 29 in a perfusion hollow fiber bioreactor and may be related to the lower oxygen concentrations inside the spheroids: Retrograde liver-perfusion experiments in rats have shown that lower oxygen concentrations partially inhibit periportal urea synthesis.30

The primary cultures of human hepatocytes in the perfusion bioreactor were inducible for the entire long-term period. Coadministration of Rif and BNF ensured the increase in mRNA synthesis for CYP3A4 (Rif), CYP2C9 (Rif), and CYP1A2 (BNF), and despite the possibility of positive or negative synergies resulting from the use of both inducers, there was a significant induction in the three CYP450 isozymes; such coadministration studies constitute a unique tool to study long-term drug-drug interactions with easy access to hepatocytes for cell-based assays. In fact, the automated perfusion, as well as the oxygen and pH control in these bioreactors, can be used to expose primary cultures of hepatocyte spheroids to repeated drug dosing, as described herein, or long-term time-varying drug concentrations for chronic toxicity assessment.

The maintenance of these hepatic activities and gene expression is enabled by the cell-cell interactions, which, during the bioreactor culture, evolve to a liver-like phenotype, as shown by the presence of albumin, CK18, CYP4503A, and HNF4α (Fig. 6, which are typical hepatic markers. The structural and functional polarity shown by both actin and aPKC staining and CDFDA metabolization (followed by the multidrug resistance protein 2–mediated transport through the apical membrane of the hepatocytes) (Fig. 7) demonstrate that these human hepatocytes, cultured as spheroids, resume cuboidal geometry, without actin stress fibers, and a polarized liver-like architecture. The presence of a functional bile canaliculi network in the hepatocyte spheroids, which had not been shown to be functional in previous studies,6 ensures an efficient, polarized transport of metabolic by-products in these hepatic microtissues. The similarity to in vivo liver tissue makes this system an interesting tool for fundamental studies of hepatic functions in physiological or bioreactor-simulated pathological conditions, namely for more complex drug transport studies. Further improvements to this system may be achieved by adding heterotypical cell interactions, by coculturing hepatocytes with endothelial cells24 or fibroblasts10, 23 or by encapsulating the hepatocyte spheroids in alginate.14

In conclusion, the perfusion bioreactor system presented herein allows for the culturing of a defined size of human hepatocyte spheroids (80 μm), which maintain hepatic liver-specific protein synthesis, CYP450, phase II and III drug-metabolizing enzyme gene expression and activity, as well as liver-like architecture inside the spheroids for 2-4 weeks.


The authors acknowledge Francisca Monteiro for technical support.