Expression and hepatobiliary transport characteristics of the concentrative and equilibrative nucleoside transporters in sandwich-cultured human hepatocytes

Abstract

We previously reported that both the concentrative (hCNT) and equilibrative (hENT) nucleoside transporters are expressed in the human liver (21). Here we report a study that investigated the expression of these transporters (transcripts and proteins) and their role in the hepatobiliary transport of nucleosides/nucleoside drugs using sandwich-cultured human hepatocytes. In the hepatic tissue, the rank order of the mRNA expression of the transporters was hCNT1 ≈ hENT1 > hENT2 ≈ hCNT2 > hCNT3. In sandwich-cultured hepatocytes, the mRNA expression of hCNT2 and hENT2 was comparable to that in hepatic tissue, whereas the expression of corresponding transporters in the two-dimensional hepatocyte cultures was lower. Colocalization studies demonstrated predominant localization of these transporters at the sinusoidal membrane and of hENT1, hCNT1, and hCNT2 at the canalicular membrane. In the sandwich-cultured hepatocytes, ENTs were the major contributors to the transport of thymidine (hENT1, 63%; hENT2, 23%) or guanosine (hENT1, 53%; hENT2, 24%) into the hepatocytes followed by hCNT1 (10%) for thymidine or hCNT2 (23%) for guanosine. Although ribavirin was predominately transported (89%) into the hepatocytes by hENT1, fialuridine (FIAU) was transported by both hENT1 (30%) and hCNTs (61%). The extensively metabolized natural nucleosides were not effluxed into the bile, whereas significant biliary-efflux was observed of FIAU (19%), ribavirin (30%), and formycin B (35%). We conclude that the hepatic activity of hENT1 and hCNT1/2 transporters will determine the in vivo hepatic distribution and therefore the efficacy and/or toxicity of nucleoside drugs used to treat hepatic diseases.

human equilibrative and concentrative nucleoside transporters (hENT and hCNT) are two families of transmembrane proteins that facilitate transport of hydrophilic nucleosides across cell membranes (2, 22, 27). Although the hENT family members, hENT1, 2, and 3, mediate bidirectional transport of purine and pyrimidine nucleosides down a concentration gradient (2), the hCNT family members, hCNT1, 2, and 3, are cation-dependent symporters that mediate unidirectional transport of nucleosides into cells (22). Unlike the broad specificity of hENTs and hCNT3, hCNT1 (N2 or cit transport system) transports primarily pyrimidine nucleosides and hCNT2 (N1 or cif transport system) transports primarily purine nucleosides (22, 27). Uridine and adenosine are ubiquitous substrates of all the nucleoside transporters.

Nucleoside transporters also transport nucleoside drugs [e.g., ribavirin, fialuridine (FIAU), and gemcitabine] into cells (10, 25, 29, 37). For example, ribavirin (in combination with interferon) is frontline therapy for the treatment of hepatitis C (17, 24). Similarly, FIAU was developed for the treatment of hepatitis B but failed in clinical trials due to significant mitochondrial toxicity resulting in hepatic failure (19, 34). These nucleoside antiviral drugs are hydrophilic and therefore need to be transported into the hepatocytes to produce their efficacy and toxicity. Although the types of nucleoside transporters expressed in the human liver were recently reported (18, 21), their subcellular localization (e.g., sinusoidal or canalicular) and their relative functional activity has never been determined. Such determination is important for understanding the mechanisms of hepatic toxicity of nucleoside drugs (e.g., FIAU) and to better design nucleoside drugs that are preferentially targeted to the liver. For example the hematological toxicity of ribavirin is dose-limiting because of its transport into erythrocytes by hENT1 and its subsequent metabolism and accumulation there (26). Such toxicity could be reduced if ribavirin was highly extracted by the liver (e.g., by nucleoside transporters) during hepatic first pass after oral administration. To design nucleoside drugs that are preferentially transported into certain tissues (to achieve targeted drug delivery), it is important to have a detailed understanding of the type of nucleoside transporters expressed in various tissues, their relative expression, and their subcellular distribution.

With the recent wider availability of human hepatocytes, the above studies are now possible. However, the conventional two-dimensional cultures suffer from loss of architecture present in the human liver. That is, they lack the polarized nature of the hepatocytes, namely the canalicular and sinusoidal membranes. This lack of polarization results in internalization of several membrane transporters normally expressed in the canalicular membrane [e.g., P-glycoprotein and multidrug resistance-associated protein 2 (MRP2); 23, 40]. Since some of the concentrative nucleoside transporters are thought to be present in the canalicular membrane (11, 14), internalization of these transporters could result in underestimation of their role in hepatobiliary transport of nucleoside drugs. This disadvantage of the two-dimensional hepatocyte cultures is overcome by sandwich-cultured hepatocytes (5, 15). This model is widely accepted as a more powerful in vitro tool to determine the magnitude of drug transport into the hepatocytes as well as of biliary secretion of drugs (20, 23). Therefore, we used sandwich-cultured primary human hepatocytes to determine the types of nucleoside transporters present in the liver, their subcellular distribution, and their relative activity in hepatobiliary transport of nucleosides and nucleoside drugs.

MATERIALS AND METHODS

Materials.

³H-adenosine, ³H-guanosine, ³H-cytidine, ³H-thymidine, ³H-uridine, ³H-ribavirin, ³H-formycin B, ¹⁴C-FIAU, FIAU, 1-(2-deoxy-2-fluoro-β-d-arabinofuranosyl) (FAU), FAU monophosphate (FAUMP), ribavirin monophosphate (RMP), and ribavirin triphosphate (RTP) were obtained from Moravek Radiochemicals (Brea, CA). Uridine, thymidine, inosine, nitrobenzylthioinosine, EGTA, dexamethasone, DAPI, Trypan blue and other chemicals were from Sigma (St. Louis, MO). Rat tail collagen (type I), Matrigel [high concentration (HC)], and NU-serum I culture supplement were from BD Biosciences (San Jose, CA). Williams E medium and Hanks balanced salt solution without CaCl₂ and MgCl₂ were from GIBCO, Invitrogen (Carlsbad, CA). BCA protein assay reagent was from Pierce Chemical (Rockford, IL). Mouse anti-human CD26 (dipeptidylpeptidase IV; DPP IV) monoclonal antibody was from BD Pharmingen (San Diego, CA). Mouse anti-human P-glycoprotein antibody (F4) was from Kamiya Biomedicals (Seattle, WA), and mouse anti-E-cadherin antibody was from Zymed (San Francisco, CA). Anti-human rabbit polyclonal antibodies for hENT1, hENT2, hCNT1, and hCNT2 were as described previously (21). Alexa-488- and -594-conjugated goat anti-rabbit, anti-mouse, and anti-rat secondary antibodies, fluorescent anti-fade mounting reagent, insulin-transferrin-selenium, and penicillin-streptomycin were from Molecular Probes, Invitrogen (Eugene, OR). Fetal calf serum was from Hyclone Laboratories (Logan, UT). Calf-intestinal phosphatase was from New England Biolabs (Beverly, MA).

Hepatic culture and tissues.

Human liver samples and primary human hepatocyte cultures used in this study were obtained from multiple human donors (n = 18) of both sexes. A detailed summary on the donor information, hepatocyte viability, and their use is given in Table 1. The collection and use of human tissue for research was approved by the University of Washington Human Subjects Review Board. Human liver samples (n = 3) were obtained from an existing bank maintained by the University of Washington School of Pharmacy (Seattle, WA). Primary human hepatocytes in suspension or in plated formats (two-dimensional or sandwich-culture) were kindly donated by CellzDirect (Pittsboro, NC, n = 18). Hepatocytes in culture were maintained in Williams E medium containing 2 mM l-glutamine, 5 μg/ml each of insulin, selenium, and transferrin, 25 nM dexamethasone, 50 U/ml penicillin, 50 μg/ml of streptomycin, and 10% fetal calf serum (culture medium) in a humidified incubator with 95% atmospheric air and 5% CO₂. Cells were replenished with fresh medium every 24 h.

Table 1.

Demographics of the human liver donors

Donor	Age, yr/Sex	Race	Weight, lbs	Hepatocyte Viability, %	History	Applications
HL8003	58/F	C	113	84	NA	real-time PCR, transport assays
HL430	52/M	C	205	86	NA	real-time PCR
HL449	50/M	C	212	84	NA	real-time PCR
HL472	50/F	H	158	88	NA	transport assays
HL500	56/F	C	170	74	NA	transport assays
HL483	37/F	C	209	89	NA	transport assays
HL484						transport assays
HL573	77/M	H	NA	NA	obese (BMI = 32)	immunolocalization
HL574	68/M	ME	NA	NA	obese (BMI = 33)	transport assays, immunolocalization
HL420	52/F	C	172	86	obese, alcohol use/abuse	transport assays
NA	27/F	C	NA	NA	NA	transport assays
HL557	68/M	C	201	88	NA	transport assays, immunolocalization
NA	71/M	C	NA	NA	obese (BMI = 38)	transport assays
HL155*	21/M	C	185	NA	NA	real-time PCR
HL167*	44/M	C	NA	NA	NA	real-time PCR
HL169*	57/M	C	NA	NA	NA	real-time PCR
HL754	57/F	C	190	95	NA	transport assays, immunolocalization
NA	57/M	C	NA	NA	NA	transport assays

^*Obtained from an existing liver bank maintained by the University Of Washington School of Pharmacy, Seattle, WA. All of the remaining human hepatocytes are obtained from CellzDirect. The subjects had no known history of or exposure to hepatitis B, hepatitis C, or HIV. In addition, they did not have cirrhosis, biliary disease, or known drug or alcohol abuse. M, male; F, female; C, Caucasian; H, Hispanic; ME, Middle-Eastern; BMI, body-mass index; NA, data not available.

Immunofluorescence studies using sandwich-cultured primary human hepatocytes.

For immunolocalization analysis, hepatocytes obtained in suspension from CellzDirect were plated on glass coverslips in a sandwich configuration. Briefly, six-well clusters containing glass coverslips were layered with diluted collagen (45 μg/cm², 2 ml/well) and allowed to bind overnight at 4°C. The excess fluid was removed from the coated surface and allowed to dry overnight. The wells with coverslips were washed with tissue culture grade water before plating with hepatocytes. Suspension hepatocytes in cold preservation medium were centrifuged at 300 g for 4 min at 4°C. The supernatant was removed, and the hepatocytes were resuspended in Williams E medium containing 100 nM dexamethasone and 10% NU-serum (attachment medium). Cell viability was measured by Trypan blue exclusion, and 1 × 10⁶ viable cells were plated in each well. Cells were allowed to attach by incubation at 37°C in a humidified incubator with 5% CO₂ for 3 h. The attachment medium was replaced with Williams E culture medium containing Matrigel at a final concentration of 0.233 mg/ml and returned to the incubator. Cells were replenished with fresh culture medium (for composition, please see above) every 24 h. For conventional two-dimensional culture, the procedure described for sandwich-cultured hepatocytes was followed except that the Matrigel-overlay step was omitted.

Hepatocytes in sandwich-culture were washed once in PBS (37°C) and fixed in 2% paraformaldehyde. Cells were blocked and permeabilized with a solution containing 1% goat serum and 0.25% Triton X-100. Cells were then coimmunostained with rabbit polyclonal antibodies for each of the nucleoside transporters and rat E-cadherin monoclonal antibody, mouse CD-26 monoclonal antibody, or mouse P-glycoprotein monoclonal antibody. Incubation with antibodies was performed in a solution containing 1% goat serum and 0.25% Tween 20 for 1 h at room temperature. Secondary antibodies (rabbit, mouse, or rat) conjugated with Alexa 488 or Alexa 594 (Invitrogen, Molecular Probes) were used as appropriate. Cells were washed three times for 15 min each after primary and secondary antibodies incubation steps. Nuclei were counterstained with DAPI (Molecular Probes, Invitrogen). The cells were finally rinsed briefly with water (to remove salts), and the coverslips were mounted onto glass slides. Images of immunostained cells were acquired with an Olympus inverted IX70 fluorescence microscope fitted with a CCD camera. The images were captured by Softmax-pro software and deconvoluted using a DV Linux image analysis system.

Real-time PCR analysis.

Hepatocytes were cultured in two-dimensional or in sandwich configurations in 6-cm dishes for 72–96 h (for details please see above). Total cellular RNA was isolated from liver tissue (0.5 mg) or two-dimensional or sandwich-cultured hepatocytes with the RNeasy Kit (Qiagen, Valencia, CA) as per manufacturer's instructions. On-column DNA digestion with DNase I (2 U/μl) for 20 min at room temperature was included in the protocol. RNA was quantified spectrophotometrically (BioPhotometer; Eppendorf, Westbury, NY) using the 260/280 nM absorbance ratio (ratio of 1.8 to 2.0), and the integrity was verified by agarose gel electrophoresis. One microgram of total RNA was used to perform the RT reaction using the ABI TaqMan reverse transcription kit and oligo-dT primers (Applied Biosystems, Foster City, CA) in a 20-μl reaction mixture. The cDNAs were semiquantified by fluorescence-based real-time RT-PCR using TaqMan technology with the ABI Prism 7000 sequence detection system (Applied Biosystems). Validated TaqMan probes and primers for hENT1 (Hs00191940_m1), hENT2 (Hs00155426_m1), hCNT1 (Hs00188418_m1), hCNT2 (Hs00188407_m1), and human GusB (hGusB) (Hs99999908_m1), purchased from Applied Biosystems, were used for analysis. Amplification efficiencies of real-time RT-PCR reactions were tested using the validation experiment for relative quantification of gene expression (User Bulletin no. 2); ΔC_t values of serial dilutions were plotted against the log of starting template concentrations to generate standard curves and estimate slopes for each of the genes tested. Template cDNA for serial dilutions was generated by mixing cDNAs prepared from BeWo, JAR, MCF-10A, MBA-MB-231, MCF-7, HepG2, and HeLa human cell lines and human liver and placental tissues. C_t is defined as the cycle at which the reporter fluorescence exceeds 10 times the standard deviation of the mean baseline emission for cycles 3–10.

The PCR reaction was performed in a 20-μl reaction mixture that contained the forward and reverse primers at 300 nM each, the probe at 100 nM, 10 μl of 2× Universal PCR Master Mix (Applied Biosystems), 5 μl of sterile Millipore water, and 4 μl cDNA from a 20-μl RT reaction mixture. PCR was performed under the following conditions: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min, in a MicroAmp optical 96-well plate. hGusB was used as an internal control to normalize each of the mRNA analyzed. Relative levels of various nucleoside transporter genes were calculated according to the equation 2^ΔΔCt, where ΔC_t represents the differences in cycle threshold numbers between the target gene and hGusB [ΔC_t = C_{t(test gene)} − C_t(hGusB)], and ΔΔC_t represents the differences between ΔC_{t(test gene)} and ΔC_{t(test gene with highest CT)}. The relative expression level of each of the nucleoside transporter genes was plotted relative to the lowest expresser within a sample (hCNT3, Fig. 1A) or to message level in a reference sample (human liver, Fig. 1B).

Fig. 1.

Expression of equilibrative nucleoside transporters (ENTs) and concentrative nucleoside transporters (CNTs) in human liver tissue and two-dimensional and sandwich-cultured primary human hepatocytes. The expression of ENTs and CNTs was analyzed by quantitative real-time RT-PCR. A: expression of ENTs and CNTs in three human livers. The mRNA level of each of the nucleoside transporters was expressed relative to that of hCNT3 in that liver. Average C_t values for hCNT3 expression in human liver samples was 37.5, whereas the values in control samples (no reverse transcriptase) were either >41.9 or below the level of detection. The open bars are the mean ± SD of the three livers. Note, the y-axis is a log scale. The interliver variability in the relative expression of the various nucleoside transporters was low. B: comparison of hENTs and CNTs expression in hepatocytes (n = 3) cultured in two-dimensional and sandwich-cultured configuration expressed as percent ± SD of that in a single liver tissue (HL8003). Note, compared with human liver, the expression of hCNT2 (30.4%, **P < 0.005) and hENT2 (75.2%, *P < 0.05), mRNA levels were significantly lower in two-dimensional cultures but not in sandwich-cultured hepatocytes.

Nucleoside transport and biliary efflux in sandwich-cultured primary human hepatocytes.

Cells were plated in 24-well sandwich-cultured format (0.375 × 10⁶ viable cells/well) and cultured for ∼72–96 h. Transport studies in sandwich-cultured human hepatocytes were carried out in sodium-containing buffer (28). Cells were briefly rinsed with PBS and preincubated for 20 min in 0.5 ml of sodium-containing buffer containing the following control or inhibitors: DMSO (vehicle control), 10 nM nitrobenzylmercaptopurine riboside (NBMPR) (to inhibit hENT1) or 10 μM NBMPR (to inhibit hENT1 and hENT2) or 200 mM cold thymidine and 10 μM NBMPR (to inhibit hCNT1, hCNT3, hENT1, and hENT2) or 200 mM cold inosine, and 10 μM NBMPR (to inhibit hCNT2, hCNT3, hENT1, and hENT2). hCNT3 contribution was determined indirectly by comparing the difference in the uptake of ³H-guanosine in the presence of 10 μM NBMPR and 10 μM NBMPR + 200 mM thymidine. After aspiration of the preincubation buffer, the cells were subsequently incubated for 10 min with the same control or inhibitors but this time containing 1 μM of ³H- or ¹⁴C-labeled nucleoside. Transport was arrested at 10 min by immediately washing the cells with ice-cold sodium-free buffer containing 20 mM cold uridine (3 × 0.5 ml). Subsequently, cells were lysed with 0.5 ml of 1 N NaOH, agitated for 20 min in a cell shaker, and neutralized by 0.5 ml of 1 M HCl. An aliquot of the lysate (50 μl) was used to measure the protein content using the BCA reagent (Pierce), and the remaining amount (950 μl) was mixed with 10 ml of liquid scintillation fluid (Ecoscint, Atlanta, GA) and counted using a Packard CR 2000 LS counter (Packard, Downers Grove, IL).

To measure biliary efflux, the hepatocytes were preincubated in Ca²⁺ containing (as described above) or Ca²⁺, Mg²⁺-free Hanks balanced salt solution (31, 32) in the presence of 1 mM EGTA for 5–10 min. Transport experiments were subsequently performed under these two conditions (as described above), and the accumulation of radioactivity was measured at the end of the transport period. Total biliary efflux was calculated by measuring the difference in the accumulated radiolabeled substrate under Ca²⁺-containing (control) and Ca²⁺-free conditions. Transport experiments for biliary efflux measurements were performed on two independent batches of hepatocytes, and each experiment was conducted in triplicate. All transport or efflux rates were expressed as pmol/mg per 10 min transport.

Quantification of intracellular metabolism of nucleosides by high-performance liquid chromatography.

The magnitude of metabolism of guanosine, formycin B, ribavirin, or FIAU at the end of the transport period (10 min) was determined in separate batches of hepatocytes. Briefly, sandwich-cultured hepatocytes incubated in Ca²⁺ buffer were lysed in lysis buffer (10 mM Tris·HCl, pH 8.0, 0.5% non-Idet P-40, 1 mM EDTA, and 2 mM PMSF), and nuclei were pelleted at 500 g for 5 min. To determine the concentration of ribavirin and its metabolites, 50 μl of either PBS alone or PBS containing 1,000 U/ml of calf intestinal alkaline phosphatase (CIP) were incubated with hepatocyte lysates at 37°C for 30 min. Subsequently, 60 μl of 6% perchloroacetic acid were added to 50 μl of both CIP-treated and -untreated hepatocyte lysates and vortexed. Later, 20 μl of 2 M K₂HPO₄ were added, and the sample was vortexed and centrifuged at 20,000 g for 4 min at room temperature. For analyzing guanosine, FIAU, and their metabolites, 500 μl of acetonitrile were added to 500 μl of hepatocyte lysate, and the samples were vortexed and centrifuged at 20,000 g for 4 min at room temperature. The efficiency of CIP dephosphorylation reactions was >95% and was verified with pilot experiments using ribavirin phosphates [RMP, ribavirin diphosphate, and RTP] and various times of incubation with CIP (Endres CJ, Govindarajan R, and Unadkat JD, unpublished observations). A sample (100 μl) of the supernatant (CIP-treated and -untreated) was injected onto the HPLC column [Atlantis dC18; Waters, Milford, MA; 150 × 4.6 mm, 3 μm (ribavirin); Zorbax XDB-C18, 150 × 4.6 mm, 3 μm (guanosine and FIAU)]. Ribavirin and its metabolites were eluted from the column with a mobile phase containing A) 100 mM potassium phosphate (pH 6.2), 0.1% N,N-dimethylhexylamine and B) methanol at a flow rate of 1.0 ml/min. The initial condition was 100% A. Between 4.5 and 6.0 min, the mobile phase composition linearly decreased from 100% A to 70% A and was held at 70% A until the end of the run at 12 min. After 12 min, the column was allowed to immediately return to 100% A and allowed to reequilibrate for 5 min before the next injection. Fractions were collected for every 30 s from 1.5 to 6 min. For analyzing guanosine and FIAU (and their metabolites), the column was eluted with a mobile phase containing A) 50 mM ammonium formate (A: 50 mM potassium formate for FIAU) and B) methanol. The HPLC conditions were as follows: flow rate, 1.0 ml/min; 100% A for 5 min, 98% A and 2% B from 5 to 10 min, 80% A and 20% B from 10 to 12.5 min, 100% B from 12.5 to 19 min, and 100% A from 19 to 25 min. Fractions were collected every 30 s for 12.5 min for guanosine and for 25 min for FIAU. Elution times of ribavirin, ribavirin phosphates (RMP and RTP), and other ribavirin metabolites (TCOOH, RTCOOH, TCONH2; kindly provided by Valeant Pharmaceuticals International) were confirmed by injection of their cold standards and detection at 207 nm (Waters 2996 PDA). Similarly, elution time of guanosine, guanosine phosphates (GMP, GTP), FIAU, FAU, and FAUMP were confirmed by injection of these standards and detection at 254 nm. The radioactivity content in each fraction was counted on a scintillation counter and expressed as percent of total radioactivity injected. The increase in the radioactive content of the fractions containing the parent compound, after incubation with CIP, was quantified and interpreted as the degree of phosphorylation of the nucleoside.

RESULTS

Relative expression of hENTs and hCNTs in human liver tissue, conventional two-dimensional-primary human hepatocytes, and sandwich-cultured primary human hepatocytes.

Since the amplification efficiency of all the transporter-specific primers/probes used in the study was ≥90% (see materials and methods), we were able to compare the relative expression of the transporters in human liver and both types of hepatocyte cultures. Consistent with our earlier findings (21), we found significant expression of hCNT1, hENT1, hENT2, and hCNT2 transcripts in the human liver tissue. The expression of hCNT3 transcripts was very low (Fig. 1A). The expression of the transporters across the three livers was consistent and followed the rank order: hCNT1 ≈ hENT1 > hENT2 ≈ hCNT2 > hCNT3.

To test whether the expression of ENTs and CNTs in hepatocyte changes on culturing, we compared the expression of nucleoside transporter transcripts in primary human hepatocytes (n = 3) cultured on solid supports (two-dimensional or sandwich configurations) with a single human liver tissue as a reference control. Compared with the human liver, the expression of hCNT2 and hENT2 was significantly lower (30.4% and 75.2%; P < 0.05) in conventional two-dimensional hepatocyte cultures, whereas the expression of the remaining transporters was comparable (Fig. 1B). In contrast, in the same batches of the hepatocytes cultured in the sandwich configuration, the expression of ENT and CNT transcripts was consistent and comparable to that observed in the human liver tissue (Fig. 1B).

Spatial localization of ENTs and CNTs in sandwich-cultured primary human hepatocytes.

We next analyzed the spatial localization of the various ENTs and CNTs in sandwich-cultured human hepatocytes. E-cadherin was used as a sinusoidal (lateral) marker (4), and CD-26 (DPP IV) was used as a bile-canalicular (apical) marker (4, 8). Earlier, during culturing (<48 h after plating), a larger proportion of hENT1 in the hepatocytes was found endocytosed in vesicular structures within the cytoplasm (Fig. 2A). hENT1 colocalized with E-cadherin in these intracytoplasmic vesicles (Fig. 2A, left, arrowheads). Partial colocalization of hENT1 was also found with early endosomal antigen-1 and Rab-5, two markers of early endosomal vesicles (data not shown). With increasing duration of culture (48–96 h), the intracellular hENT1 and E-cadherin levels decreased progressively, whereas their expression increased at the sinusoidal membrane (Fig. 2A, right, arrows). At 72 h of culture, distinct canalicular structures were formed (arrowheads, top). In addition, E-cadherin staining was predominantly in the sinusoidal regions of the hepatocytes (arrows; bottom), and CD-26 (arrowheads; middle) staining was predominantly in the bile canaliculi-like regions where there were cell-to-cell contacts (Fig. 2B). Because it takes ∼48–72 h (after plating) for these membrane proteins to congregate into sinusoidal and canalicular domains, all the remaining experiments were carried out between 72 and 96 h of culture.

Fig. 2.

Spatial localization of ENTs and CNTs in sandwich-cultured primary human hepatocytes. A: during initial days after culturing, E-cadherin and hENT1 were predominantly localized intracellularly. Hepatocytes were double immunostained for E-cadherin and hENT1 at 12, 24, and 48 h after plating. Note, a significant fraction of E-cadherin and hENT1 was intracellular at 12- and 24-h time points but not at the 48-h time point. B: sorting of apical (CD-26) and sinusoidal (E-cadherin) markers in the hepatocytes. Hepatocytes cultured in sandwich configuration for 72–96 h showed distinct canalicular structure formation (top, arrowheads). Hepatocytes in this configuration were immunostained for CD-26 (middle, arrowheads) and E-cadherin (bottom, arrows). Note the sinusoidal pattern of E-cadherin staining at the cell periphery and the canalicular pattern of CD-26 staining at the cell-to-cell contact areas. C: polarized localization of ENTs and CNTs in the hepatic membranes. Hepatocytes in sandwich configuration (72–96 h) were double immunostained with transporter-specific rabbit polyclonal antibody and rat E-cadherin monoclonal antibody or mouse CD-26 monoclonal antibody. Note significant colocalization (yellow, merged images) of all of the transporters with E-cadherin in the sinusoidal membrane (arrowheads). D: localization of ENT1, CNT1, and CNT2 in the canalicular membrane. Partial colocalization of hENT1, hCNT1, and hCNT2 and CD-26 was observed (arrows). Nuclei stained with DAPI are blue. Original magnifications are ×60 (A, B, D), ×40 (C), and ×20 (B, top).

Double immunolocalization of hepatocytes (72–96 h of plating; 1 × 10⁶ viable cells/well) with transporter-specific antibody, E-cadherin, or CD-26 antibody showed predominant colocalization of hENT1 and hENT2 with E-cadherin (Fig. 2C, first and second rows, merged images) and partial colocalization of hENT1, but not hENT2 (data not shown), with CD-26 (Fig. 2D, top, arrows). This suggests the presence of hENT1 and hENT2 mainly at the hepatocyte sinusoidal membrane with a fraction of hENT1 also trafficking to the canalicular membrane. hCNT1 and hCNT2 partially colocalized with both E-cadherin (Fig. 2C, third and fourth rows, merged images) and CD-26 (Fig. 2D, middle and bottom, arrows), suggesting incorporation of hCNTs into both the sinusoidal and canalicular membranes. However, compared with hCNT2, only a small fraction of hCNT1 was identified at the canalicular membrane (Fig. 2D, middle, arrows). A similar localization pattern of ENTs and CNTs was also observed at higher cell densities (1.5 × 10⁶ viable cells/well). The localization of these transporters at canalicular membrane was also verified with an anti-P-glycoprotein antibody (data not shown). hCNT3 localization was not tested because an anti-hCNT3 antibody was not available to us. Collectively, these data suggest that, although hENTs and hCNTs were predominantly localized at the sinusoidal membrane, at least a fraction of hENT1, hCNT2, and hCNT1 also traffic to the canalicular membrane of the hepatocytes.

Hepatic transport of natural nucleosides and nucleoside drugs into sandwich-cultured primary human hepatocytes.

In the presence of a Na⁺-gradient, the transport of tritiated natural nucleosides (1 μM; adenosine, uridine, guanosine, thymidine, or cytidine) into sandwich-cultured hepatocytes was linear for at least ∼20 min (data not shown). Therefore, a 10-min time point was chosen for all remaining transport experiments (Table 2) with each substrate (³H-guanosine or ³H-thymidine) tested using three independent batches of sandwich-cultured hepatocytes (Fig. 3A). On the basis of these data, hENT1 and hENT2 were the major contributors to the hepatic transport of both thymidine (hENT1, 63.3%; hENT2, 23.1%) and guanosine (hENT1, 53.6%; hENT2, 23.9%), followed by hCNT1 (9.6%) for thymidine and hCNT2 (23.4%) for guanosine. The contribution of hCNT3 toward thymidine or guanosine transport was negligible (0.02 and 0.1%, respectively). These data suggest that the transport of both purine and pyrimidine nucleosides into sandwich-cultured hepatocytes is predominantly mediated by hENT1 followed by hENT2 ≈ hCNT2 for guanosine and hENT2 ≈ hCNT1 for thymidine. It should be noted here that the absolute rate of uptake of each nucleoside was largest for hENT1 and approximately equal for the remaining transporters.

Table 2.

Hepatobiliary transport of natural nucleosides, nucleoside drugs, and a poorly-metabolized nucleoside, formycin B

Substrate	Total Transport by Nucleoside Transporters, pmol/mg per 10 min	Mediated Transport, pmol/mg per 10 min				Net Biliary Efflux, pmol/mg per 10 min
		hENT1	hENT2	hCNT1	hCNT2
3H-Guanosine	4.62±0.5 (53.6±6.7)	2.5±0.3 (53.6±6.7)	1.1±0.1 (23.9±2.4)	—	1.1±0.2 (23.4±4.1)	0.2±0.1 (3.9±0.7)
3H-Thymidine	8.89±0.7	5.8±0.7 (63.3±7.7)	1.9±0.1 (23.1±1.6)	1.0±0.1 (9.6±0.7)	—	0.7±0.1 (7.0±1.2)
3H-Ribavirin	8.84±0.8	7.8±1.0 (88.6±10.9)	0.6±0.1 (6.1±0.7)	0.1±0.0 (0.7±0.0)	0.4±0.0 (4.49±0.1)	3.3±0.0 (30.2±5.7)
14C-FIAU	28.5±10.4	8.6±1.1 (29.9±3.5)	4.2±0.5 (14.8±1.8)	15.2±2.0* (53.1±7.2)	2.4±0.4* (8.4±1.3)	13.22±0.4 (18.8±3.6)
3H-Formycin B	11.9±0.9	—	3.4±0.2 (35.1±2.1)	0.7±0.0 (6.9±0.1)	5.5±0.5 (57.3±4.7)	4.8±0.4 (35.4±6.1)

Values are means ± SD from 3 independent batches of sandwich-cultured hepatocytes. Hepatic transport and biliary efflux were measured as described in materials and methods. Data were derived from experiments such as those depicted in Fig. 3, A and B. Total transport was calculated by measuring the difference in transport of nucleosides into hepatocytes between Na buffer (solid bar, Fig. 3A) and either Na buffer containing 10 μM nitrobenzylmercaptopurine riboside (NBMPR) + 200 mM inosine (for purine nucleosides, hatched bar, Fig. 3A) or Na buffer containing 10 μM NBMPR + 200 mM thymidine (for pyrimidine nucleosides, stippled bar, Fig. 3A). Numbers in parenthesis represent percent contribution to total transport mediated by nucleoside transporter or biliary efflux.

^*Represents possible added contribution from human concentrative nucleoside transporter (hCNT)3. hENT, human equilibrative nucleoside transporter; FIAU, fialuridine.

Fig. 3.

Transport and biliary efflux of ³H-guanosine, ³H-thymidine, ³H-ribavirin, ¹⁴C-fialuridine (FIAU), or ³H-formycin B into sandwich-cultured human hepatocytes grown to confluence. A: transport of radiolabeled (1 μM) ³H-guanosine, ³H-thymidine, ³H-ribavirin, ¹⁴C-FIAU, or ³H-formycin B in sandwich-cultured hepatocytes in the presence of sodium was determined in the presence or absence of various inhibitors (see materials and methods). The bars represent means ± SD of data from 3 independent batches of hepatocytes, each determined in triplicate. From these data, the transport mediated by nucleoside transporters was calculated and is presented in Table 2. The transport of various compounds except FIAU is represented by the left y-axis, whereas FIAU transport is represented by the right y-axis. NBMPR, nitrobenzylmercaptopurine riboside. B: radioactive contents of ³H-guanosine, ³H-thymidine, ³H-ribavirin, ¹⁴C-FIAU, or ³H-formycin B measured after conducting transport experiments in sandwich-cultured hepatocytes preincubated with Ca²⁺-containing (open bars) or Ca²⁺-free medium (closed bars) (see materials and methods). The bars represent means ± SD of data from 2 independent batches of hepatocytes, each determined in triplicate. The net biliary efflux amount of various nucleoside substrates is provided in Table 2. The cell and the bile content of various compounds except FIAU are represented by the left y-axis, whereas FIAU cell and bile contents are represented by the right y-axis.

We next evaluated the transporters involved in the influx of antiviral nucleoside drugs (1 μM) into sandwich-cultured hepatocytes (Fig. 3A). A purine (guanosine) analog, ribavirin [1-(β-d-ribofuranosyl)-1H-1, 2, 4-triazole-3-carboxamide] and a pyrimidine analog, FIAU, were chosen as model drugs for evaluation. Similar to the natural purine nucleoside (guanosine), ³H-ribavirin transport was mediated predominantly by hENT1, but its percent contribution was larger (88.6%) than that seen for guanosine (53.6%). Human ENT2 and hCNT2 contributed only 6.1% and 4.5%, respectively, to the transport of ³H-ribavirin. ¹⁴C-FIAU transport into sandwich-cultured hepatocytes was mediated by hCNT1 (53.1%) followed by hENT1 (29.9%) and hENT2 (14.8%), respectively. However, it should be noted that the absolute rate of hENT1-mediated transport of FIAU and ribavirin was comparable. ³H-formycin B, a poorly metabolized purine nucleoside, was transported into the hepatocytes (Table 2) by hCNT2 (57.3%), hENT2 (35.1%), and hCNT1 (6.9%).

Biliary efflux of natural nucleosides and nucleoside drugs in sandwich-cultured primary human hepatocytes.

The net biliary efflux of the nucleosides was determined by the difference in the radioactivity content of the hepatocytes in the presence or absence of bile-canaliculi (see materials and methods). The disassembly of tight junctions by the Ca²⁺-free solutions was confirmed by observing the hepatocytes under a phase-contrast microscope for loss of bile-canaliculi-like structures. When a separate set of cells were analyzed by immunolocalization experiments, a clear reduction in both E-cadherin (adherens-junctions component) and ZO-1 (tight-junctions component) staining at the cell surface was observed (data not shown). Figure 3B (open and solid bars) shows the rate of increase in hepatocyte-radioactive content under control and canaliculi-disrupted conditions in the presence of an Na⁺-gradient. These data (Table 2) indicate that ³H-guanosine or ³H-thymidine are not significantly effluxed into the bile. In contrast to these natural nucleosides, 30.2% of ³H-ribavirin and 18.8% of ¹⁴C-FIAU radioactivity were effluxed into the bile. ³H-formycin B was effluxed into the bile to a greater extent (35%) than guanosine, thymidine, ribavirin, or FIAU.

Role of metabolism in the magnitude of biliary efflux of various nucleosides.

The lack of significant biliary efflux of ³H-guanosine or ³H-thymidine suggests that the natural nucleosides are either not transported into the bile or are completely metabolized (e.g., phosphorylated) within the hepatocytes and therefore trapped there. To test the latter hypothesis, we measured the intracellular radioactivity content of the natural nucleoside guanosine in sandwich-cultured hepatocyte at the end of the transport period. About ∼99% of guanosine radioactivity in the hepatocyte lysate was recovered as its phosphorylated metabolites (predominantly as GTP form) with only traces present as the parent (0.6%) (Table 3). Unlike guanosine (0.6%), about 20.3% of ribavirin and 33.7% of FIAU radioactivity were present and recovered as the parent compound at the end of the transport period. Also, unlike the extensive phosphorylation of guanosine (99.3%), only 39.8% of ribavirin and 21.4% of FIAU were converted to their phosphorylated metabolites.

Table 3.

Extent of metabolism of nucleosides and nucleoside drugs in sandwich-cultured primary human hepatocytes

Substrate	Parent, %	Phosphorylated Metabolites of Parent, %	Other Metabolites, %
3H-Guanosine	0.6±0.1	99.3±0.9	none
3H-Ribavirin	20.3±1.8	39.8±2.1	41.2±4.3
14C-FIAU	33.7±3.7	21.4±2.2	43.5±4.2

Values are means ± SD of triple determinations from a single batch of sandwich-cultured hepatocytes. Each hepatocyte lysate obtained from hepatocytes incubated with Ca²⁺ buffer was subsequently incubated at 37°C (for 30 min) with or without calf intestinal alkaline phosphatase. Then, the lysate was injected onto the HPLC, and the percent of radioactivity recovered in various HPLC fractions was determined. The increase in radioactivity associated with the fraction containing the parent nucleoside was interpreted as the degree of phosphorylation of the nucleoside. Radioactivity eluting in other fractions under alkaline phosphatase-treated condition was labeled as other metabolites. Greater than 95% of the radioactivity injected onto the HPLC was recovered in the fractions collected.

DISCUSSION

Our earlier study using in situ hybridization and immunohistochemical analysis qualitatively identified expression of hENTs (hENT1 and hENT2) and hCNTs (hCNT1 and hCNT2) in the human liver (21). In this communication, we have compared with the use of semiquantitative real-time PCR analysis the relative expression of ENT and CNT transcripts in several human livers. We did not include hENT3 in our study because it is localized intracellularly and therefore cannot be involved in cellular uptake of nucleosides (3). The rank order of expression of transcripts of ENTs and CNTs was consistent in all three livers as was the absolute level of their expression (Fig. 1A). The hepatic expression of hCNT1 and hENT1 transcripts was significantly higher (∼200-fold and ∼100-fold, respectively) than that of hCNT2. This high expression was consistent with their abundant protein expression (as seen by immunolocalization analysis) at the hepatic sinusoidal membrane (Fig. 2C). Unlike hCNT1 and hENT1, only moderate levels of hENT2 (∼100-fold < hENT1) and hCNT2 (∼100-fold < hCNT1) transcripts and a low level of hCNT3 transcripts were expressed in the human liver. Despite these low levels of transcript expression, significant hENT2 and hCNT2 protein expression was identified in human liver sections (21) and in sandwich-cultured primary human-hepatocytes (present study) (Fig. 2C). Such discrepancy between transcripts and protein expression has been previously observed in the rat liver (12).

Reprogramming in the expression profile (dedifferentiation) of various liver-specific genes has been clearly demonstrated in the hepatocytes after perfusion of liver tissue and culture of hepatocytes on a solid support (16). The rapid decline in mRNA levels of several liver-specific genes (e.g., P-glycoprotein, MRP2, albumin, α-antitrypsin, ligandin, and transferrin) was shown to be prevented or reverted (redifferentiation) by culturing hepatocytes in the sandwich configuration, that is, between two layers of gelled matrix (5, 15, 23, 40). In this study, we tested whether modulating hepatocyte differentiation states (by culturing them in two-dimensional vs. sandwich configuration) can alter the expression of ENTs and CNTs. Hepatocytes cultured in the conventional two-dimensional configuration not only exhibited a flattened morphology with less pronounced formation of bile-canaliculi-like structures (data not shown) but also exhibited significant lower expression of hCNT2 and hENT2 mRNA with slight elevation in hENT1 mRNA levels. These observations suggest that loss of hepatic architecture influences the mRNA expression of at least some of the nucleoside transporters. While this manuscript was being prepared, Fernandez-Veledo et al. (18), reported a time-dependent reduction in the transcript expression of hCNT2, hENT2, and hCNT1 in isolated primary human hepatocytes. However, they observed a partial recovery of hCNT1 transcripts after ∼96 h of culturing the hepatocytes. These results are strikingly similar to ours, except that we did not observe a decrease in hCNT1 transcripts in two-dimensional or sandwich-cultured hepatocytes, most likely due to the longer time duration of our cultures.

Interestingly, the reduction in the transcript levels of hCNT2 and hENT2 observed in the two-dimensional cultures was not observed in sandwich-cultured hepatocytes. Moreover, the mRNA level of all of the ENTs and CNTs was nearly identical to that in the human liver. These data, along with the altered expression profile in two-dimensional cultures, suggest that differentiated state of the hepatocytes is a prerequisite for sustained expression of hCNT2 and hENT2 transporters. Although the present findings show predominant localization of hCNTs and hENTs at the sinusoidal hepatic membranes, significant amounts of hENT1 and hCNT2 were identified in both the sinusoidal and the canalicular membranes (Fig. 2, C and D). Occasionally, diffusion of canalicular proteins to sinusoidal membranes or mistargeting of canalicular proteins to sinusoidal membranes could occur if the tight-junctions components are improperly assembled and the polarized state is not well established. However, we found that, even in cellular fields where canalicular proteins (like P-glycoprotein and DPP IV) were exclusively localized at the bile-canaliculi regions, hENT1 and hCNT2 were localized to both the sinusoidal and the canalicular membrane (data not shown). Collectively, these data suggest that the observed localization of ENT1 and CNTs, at both the sinusoidal and the canalicular membranes, could not be a mere artifact of variations in culture conditions but represents the physiological localization of these transporters. This pattern of localization is somewhat surprising because previous data in human intestinal tissues (9, 38) and Madin-Darby canine kidney (renal epithelial) cells (7, 28, 33) suggest that CNTs are localized to the apical membrane and the ENTs are localized to the basolateral membrane of the differentiated epithelial cells. Thus our data suggest that this pattern of expression does not appear to extend to the polarized hepatocyte, which has a well-differentiated canalicular membrane (14). The mechanism(s) by which nucleoside transporters are sorted and targeted to the apical vs. basolateral domains in the hepatocytes vs. other epithelial cells remains to be elucidated.

Evidence in the literature indicates that the expression of CNTs could be responsive to differentiation signals (1, 18, 21, 13). As indicated above, we have previously shown that CNTs and ENTs are respectively expressed in the apical and basolateral domains of the differentiated epithelial cells, allowing vectorial flux of nucleosides that contributes to the intestinal absorption (9, 38) and renal tubular secretion (28) of nucleosides or nucleoside drugs. Therefore, we determined whether the spatial localization of ENTs and CNTs on the sinusoidal and canalicular membrane of the hepatocytes mediated hepatic uptake and vectorial transport of nucleosides and nucleoside drugs into the bile. We chose to study purine and pyrimidine nucleosides that are highly metabolized intracellularly (e.g., thymidine) and those that are less efficient (e.g., ribavirin, fialuridine) or poorly metabolized (e.g., formycin B).

Consistent with the immunolocalization data, transport studies identified involvement of hENT1/2 and hCNT1/2 in the hepatic transport of nucleoside and nucleoside drugs. However, as expected, the individual contribution of transporters in the transport process varied and was substrate dependent. Consistent with its predominant presence at the sinusoidal membrane, except for formycin B, hENT1 was a significant (if not the dominant) contributor to the transport of the nucleosides into the hepatocytes. Formycin B, a purine, has poorer affinity for hENT1 than hENT2 (6). This may explain the observation that hENT2 and hCNT2 were the dominant contributors to the transport of this nucleoside into hepatocytes. Consistent with its low expression in the hepatocytes, the contribution of hCNT3 to the transport of nucleosides into the hepatocytes was negligible. It is worth noting here that the percent contribution of the various transporters to the transport of each nucleoside is dependent on the participation of the other transporters in the transport process. For example, the extent of FIAU transport by hENT1 is higher than that of any other nucleosides, yet the percent contribution of hENT1 to the transport of the drug into the hepatocytes is the lowest of all the nucleosides because hCNTs are the dominant contributors to the transport of this drug into the hepatocytes. The estimated values for hCNT1- and hCNT2-mediated transport of FIAU (Table 2) may also include contribution from hCNT3, because uridine is a ubiquitous substrate of all CNTs. However, given the minimal contribution of hCNT3 activity in the transport of guanosine or ribavirin, its contribution to the transport of FIAU is likely to be minimal. We have previously shown that FIAU is transported by hENT1 expressed on both the plasma and the mitochondrial membrane and that such transport enhances the mitochondrial toxicity of FIAU (29). The above findings indicate that hENT1-mediated transport of FIAU into the hepatocytes (and therefore its hepatic and mitochondrial toxicity) is augmented by the presence of hCNTs and hENT2 in the sinusoidal membrane. Furthermore, the net FIAU transport into hepatocytes, mediated by nucleoside transporters, was 10-fold greater than the other nucleosides. This significantly greater flux of FIAU into hepatocytes may additionally explain the significant hepatotoxicity of this drug.

For ribavirin, a purine nucleoside drug, hENT1 is the major contributor (∼89%) of its transport into sandwich-cultured hepatocytes, whereas hCNT2 is a minor contributor (most likely due to its lower expression). In contrast, our previous studies using human jejunal brush-border membrane vesicles suggested that hCNT2 (but not hENT1 or 2) is the major contributor of the intestinal absorption of orally administered ribavirin and that this hCNT2-mediated transport of ribavirin could be saturated at clinical doses of ribavirin (600-mg tablets, bid) (37). However, hENT1 plays a major role in the entry and accumulation of ribavirin in erythrocytes, resulting in the dose-limiting hematological toxicity of the drug in patients with hepatitis C (26). Collectively, these data indicate that hCNT2 facilitates the intestinal absorption of ribavirin, but hENT1 determines its hepatic and erythrocyte concentrations and therefore its antiviral efficacy and toxicity.

Our data on vectorial transport of nucleosides into the biliary compartment indicate that naturally occurring nucleosides are not excreted into the bile in any significant quantity. These observations are consistent with an earlier report that speculated that the canalicular CNTs salvage nucleosides from the bile (11). Alternatively, these natural nucleosides may be so rapidly and extensively metabolized in the hepatocytes that they may not have an opportunity to be excreted into the bile by the nucleoside transporters expressed in the canalicular membrane. We tested the latter hypothesis by studying the biliary excretion of two highly metabolized natural nucleosides, one purine (guanosine) and the other pyrimidine (thymidine), and of a poorly metabolized purine nucleoside (formycin B). As expected, the biliary effluxes of both thymidine and guanosine were lower, whereas that of formycin B was more than 5–10-fold higher and constituted almost 35% of the total content of formycin B in sandwich-cultured hepatocytes. To determine whether this phenomenon extended to purine and pyrimidine nucleoside drugs, we studied the net biliary efflux of ribavirin (a purine) and FIAU (a pyrimidine). Unlike the naturally occurring nucleosides, the biliary efflux rate of these drugs was comparable to that of formycin B and accounted for 30.2% of ³H-ribavirin and 18.8% of ¹⁴C-FIAU radioactivity in sandwich-cultured hepatocytes. In humans, about 40% of the ribavirin dose is renally excreted (30), but a significant fraction of the IV dose is not accounted for (36, 39). Thus it is possible that a significant fraction could be excreted in the bile. The expression of hCNT2 throughout the small intestine (21, 35) could also facilitate the hepatobiliary recirculation and reabsorption of biliary-excreted ribavirin. However, our method does not allow us to distinguish biliary excretion of the unchanged nucleoside drug from its nonphosphorylated metabolites. Using HPLC, we found that ∼40 and 45% of the radioactivity in the hepatocyte lysates (including canalicular contents) were associated with ribavirin and FIAU metabolites (excluding parent and its phosphorylated metabolites). Collectively, these data indicate that intracellular metabolism is indeed an important determinant of the magnitude of biliary efflux of nucleosides in sandwich-cultured hepatocytes.

On the basis of the expression, subcellular distribution, and functional characteristics of ENTs and CNTs, we propose a model (Fig. 4) for hepato-biliary transport of nucleosides. CNTs expressed at the sinusoidal hepatic membrane (SM) facilitate hepatic entry of nucleosides, whereas ENTs expressed at the SM mediate sinusoidal uptake (into hepatocytes) or sinusoidal efflux (into blood) depending on the relative concentrations of the nucleosides in hepatocytes and blood. Although CNTs in canalicular membrane allow hepatic reuptake of nucleosides into the hepatocytes, ENT1 could mediate hepatic uptake or biliary efflux of nucleosides depending on the relative concentrations of the nucleosides in hepatocytes and bile. When the nucleosides are rapidly and significantly metabolized in the hepatocytes, they are unlikely to be effluxed into the blood or bile in a significant quantity by the nucleoside transporters. However, this does not preclude the possibility that the metabolites, if they are substrates of other efflux transporters (e.g., breast cancer resistance protein, MRP2), may be effluxed into the bile by these transporters. However, for those that are not rapidly and significantly metabolized, ENT1 may efflux the nucleosides into the bile. Once there, they may be subject to two competing processes: 1) reuptake by the CNTs and 2) flow-mediated movement down the biliary tree into the bile duct. The reuptake process will be governed by the relative functional activity of these transporters and by the biliary concentrations of the nucleosides (likely to be high), which may saturate the transporters. It is not possible to determine the relative activity of the canalicular membrane transporters in the sandwich-cultured hepatocytes. For such determination, canalicular membrane vesicles will need to be used. In the sandwich-cultured hepatocytes, a cholestatic model, the concentrations of the nucleosides in the bile may have been sufficient to saturate the CNTs. Alternatively, the metabolites of these nucleosides may have been effluxed into the bile by other efflux transporters, resulting in the higher biliary efflux of these drugs.

Fig. 4.

A schematic model of the equilibrative and concentrative nucleoside transporters in human hepatocytes. Transport of nucleosides across sinusoidal membrane (SM) is mediated by hENT1, hENT2, hCNT1, and hCNT2 transporters. While the CNTs in the SM mediate uptake of nucleosides from the blood into the hepatocytes, the ENTs in the SM may mediate both sinusoidal uptake into the hepatocytes (H) and sinusoidal efflux of nucleosides into the blood depending on the direction of the concentration gradient. The canalicular membrane (CM) contains hENT1, hCNT1, and hCNT2 transporters. While hCNT1 and hCNT2 mediates reuptake of nucleosides from the bile into hepatocytes, hENT1 could mediate both biliary efflux of nucleosides as well as reuptake of nucleosides into the hepatocytes from bile depending on the direction of the concentration gradient. Other canalicular efflux transporters (CE) could transport nucleosides or their metabolites into the bile. N, nucleus, BC, bile-canaliculus.

In summary, compared with the two-dimensional human hepatocyte culture, sandwich-cultured hepatocytes more closely represent the expression profile of ENTs and CNTs found in the human liver tissue. In addition, we found significant expression of hENT1/2 and hCNT1/2 proteins in the sinusoidal membrane, whereas the canalicular membrane expressed predominately hENT1, hCNT1, and hCNT2. Transport of nucleosides and nucleoside drugs into the hepatocytes confirmed this pattern of protein expression. On the sinusoidal membrane, hENT1 was a significant (if not the dominant) contributor to the total transport of nucleosides into the hepatocytes, whereas the contribution of the CNTs was substrate dependent. The biliary efflux of nucleosides or nucleoside drugs was dependent on their magnitude of intracellular metabolism. Nucleosides that were extensively metabolized intracellularly were not significantly effluxed into the bile, whereas those that were not metabolized extensively were effluxed into the bile. These data indicate that hepatic nucleoside transporters could be utilized to deliver antiviral nucleoside drugs to the liver in an efficient, tissue-directed manner to treat hepatic diseases such as hepatitis B and C.

GRANTS

This study was supported by NIH grant RO1GM054447. We also acknowledge NIH PO1GM32165 that supports the maintenance of the human liver bank of the School of Pharmacy, University of Washington, Seattle, WA.