Vitamin C Transport in Oxidized Form across the Rat Blood–Retinal Barrier

Abstract

purpose. To elucidate the mechanisms of vitamin C transport across the blood–retinal barrier (BRB) in vivo and in vitro.

methods. [¹⁴C]Dehydroascorbic acid (DHA) and [¹⁴C]ascorbic acid (AA) transport in the retina across the BRB were examined using in vivo integration plot analysis in rats, and the transport mechanism was characterized using a conditionally immortalized rat retinal capillary endothelial cell line (TR-iBRB2) as an in vitro model of the inner BRB.

results. The apparent influx permeability clearance (K _in) per gram of retina of [¹⁴C]DHA and [¹⁴C]AA was found to be 2.44 × 10³ μL/(min · g retina) and 65.4 μL/(min · g retina), respectively. In the retina and brain, the K _in of [¹⁴C]DHA was approximately 38 times greater than that of [¹⁴C]AA, whereas there was no major difference in the heart. The K _in of [¹⁴C]DHA in the retina was eight times greater than that in the brain. HPLC analysis revealed that most of the vitamin C accumulated in AA form in the retina. These results suggest that vitamin C is mainly transported in DHA form across the BRB and accumulates in AA form in the rat retina. In an in vitro uptake study in TR-iBRB2 cells, the initial uptake rate of [¹⁴C]DHA was 37 times greater than that of [¹⁴C]AA, which is in agreement with the results of the in vivo study. [¹⁴C]DHA uptake by TR-iBRB2 cells took place in an Na⁺-independent and concentration-dependent manner with a K _m of 93.4 μM. This process was inhibited by substrates and inhibitors of glucose transporters. [¹⁴C]DHA uptake was inhibited by d-glucose in a concentration-dependent manner with a 50% inhibition concentration of 5.56 mM. Quantitative real-time PCR and immunostaining analyses revealed that expression of GLUT1 and -3 was greater than that of the Na⁺-dependent l-ascorbic acid transporter (SVCT)-2 in TR-iBRB2 cells.

conclusions. Vitamin C is mainly transported across the BRB as DHA mediated through facilitative glucose transporters and accumulates as AA in the rat retina.

Vitamin C, which is an essential substance in humans, acts as a cofactor in the enzymatic biosynthesis of collagen, catecholamine, and peptide neurohormones and as an antioxidant and/or free radical scavenger to detoxify free radicals in body tissues. ¹ The retina is the only tissue in which light is focused on a group of cells. It is necessary to protect the retina against oxidative stress because light causes free radical oxidation. ² Vitamin C is present in the retina at a high concentration compared with its presence in other organs in humans. ^{1 3} The concentration of l-ascorbic acid (AA) is approximately 1.6 mM in the rat and guinea pig retina, although the plasma concentration in most mammals is 50 to 100 μM, ^{3 4} suggesting that AA is transported from the circulating blood to the retina across the blood–retinal barrier (BRB) through a specific transport process. Possible sources of oxidative stress in diabetes and retinal diseases include increased generation of reactive oxygen species by autooxidation of d-glucose and a reduced tissue concentration of antioxidants. ^{5 6} It is important to elucidate the transport mechanisms for vitamin C as far as the supply of antioxidants in the neural retina is concerned. 

The BRB, which is composed of retinal capillary endothelial cells (inner BRB) and retinal pigmented epithelial cells (RPE, outer BRB), plays a key role in restricting the nonspecific transport of hydrophilic compounds and facilitating the influx and efflux transport of essential molecules and xenobiotics, respectively, from the circulating blood to the retina and vice versa. ^{7 8} Na⁺-dependent l-ascorbic acid transporter (SVCT)-1 and -2 have been cloned and mediate concentrated, high-affinity AA transport that is driven by the Na⁺ electrochemical gradient. ⁹ Although SVCT2 mRNA is present in the retina, ⁹ its localization and transport functions are not fully understood. An Na⁺-dependent l-ascorbic acid transport process and SVCT2 are present in RPE ¹⁰ and lens epithelial cells, ¹¹ respectively. However, the Na⁺-dependent l-ascorbic acid transport process in RPE is inhibited by d-glucose, suggesting that it may not be SVCT1 or -2. ¹⁰ The facilitative glucose transporters, GLUT1 and -3 mediate equilibrative and relatively low-affinity dehydroascorbic acid (DHA) transport. ^{12 13} DHA is an oxidized form of AA, and its plasma concentration is reported to be approximately 10 μM in the rat ¹⁴ and human. ¹⁵ Agus et al. ¹⁶ have reported that DHA crosses the blood–brain barrier (BBB) through GLUT1 at the luminal and abluminal side of BBB and accumulates as reduced AA in the brain. Although GLUT1 is expressed at both the inner BRB and RPE and plays an essential role in supplying d-glucose as an energy source in the neural retina, ^{17 18} our knowledge of vitamin C transport mechanism across the BRB is incomplete. 

The purpose of this study was to elucidate the mechanisms of vitamin C transport across the BRB. To be able to understand the physiological and pathophysiological roles of BRB transport, we wanted to discover whether SVCT2 or glucose transporters make the greatest contribution to the supply of vitamin C to the retina. 

Materials and Methods

Animals

Male Wistar rats, weighing 250 to 300 g, were purchased from SLC (Shizuoka, Japan). The investigations using rats described in this report conformed to the provisions of the Animal Care Committee, Toyama Medical and Pharmaceutical University (2001-190) and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Reagents

l-[1-¹⁴C]Ascorbic acid ([¹⁴C]AA, 13 mCi/mmol) and d-[1-³H(N)]mannitol ([³H]d-mannitol, 17 Ci/mmol) were purchased from PerkinElmer Life Sciences (Boston, MA). l-[¹⁴C]Dehydroascorbic acid ([¹⁴C]DHA, purity more than 90%; see Fig. 2B ) was generated in all experiments by incubating [¹⁴C]AA (1 μM in saline) with ascorbate oxidase (1 unit/1 mmol AA in saline; Sigma-Aldrich, St. Louis, MO) at 37°C for 90 seconds, according to a reported method. ¹⁶ All other chemicals were of reagent grade and available commercially. 

Blood-to-Retina Transport Studies

A Wistar rat was anesthetized with an intramuscular injection of ketamine-xylazine (1.22 mg/kg xylazine and 125 mg/kg ketamine). The femoral artery was cannulated with polyethylene tubing (SP-31, inner diameter 0.5 mm, outer diameter 0.8 mm; Natsume, Tokyo, Japan) containing 100 IU heparin/mL in extracellular fluid buffer (122 mM NaCl, 25 mM NaHCO₃, 3 mM KCl, 1.4 mM CaCl₂, 1.2 mM MgSO₄, 0.4 mM K₂HPO₄, 10 mM d-glucose, and 10 mM HEPES [pH 7.4]) to collect blood samples. Then, [¹⁴C]DHA (5 μCi/rat), [¹⁴C]AA (10 μCi/rat), or [³H]d-mannitol (10 μCi/rat) was injected into the femoral vein. After collection of blood samples, all rats were decapitated, and the retinas, cerebrum, and heart were removed. All samples were dissolved in 2 N NaOH at 50°C for 3 hours. After the samples were dissolved, 50 μL H₂O₂ was added to the blood and heart samples to decolorize them. All samples were neutralized and mixed with liquid scintillation cocktail (ACS II; Amersham, Buckinghamshire, UK) and then the radioactivity was measured in a liquid scintillation counter (LS6500; Beckman-Coulter, Fullerton, CA). 

Determination of Influx Permeability Clearance

The apparent influx permeability clearance (K _in) of [¹⁴C]DHA, [¹⁴C]AA, or [³H]d-mannitol in tissues was determined by integration plot analysis, ^{19 20} by modification of a reported method. ²⁰ Briefly, the tissue uptake of each compound can be expressed as  

\[dX(t)/dt\ {=}\ K_{\mathrm{in}}\ {\times}\ C_{\mathrm{p}}(t)\ {-}\ K_{\mathrm{eff}}\ {\times}\ X(t)\]

where X(t) (dpm/g tissue) and C _p(t) (dpm/mL) are the tissue concentration and plasma concentration at time t, respectively, and K _in and K _eff represent the influx and efflux permeability clearance in tissue, respectively. Integrating equation 1 and solving the apparent tissue concentration in the early-phase gives  

\[X_{\mathrm{app}}(t)\ {=}\ K_{\mathrm{in}}\ {\times}\ \mathrm{AUC}(t)\ {+}\ V_{\mathrm{i}}\ {\times}\ C_{\mathrm{p}}(t)\]

where X _app(t) (dpm/g tissue) is the apparent tissue concentration in the sample, AUC(t) (dpm · min/mL) is the area under the plasma concentration time curve of each compound from time 0 to t and V _i (mL/g tissue) is the volume of interstitial space in the tissue. Division of both sides of equation 2 by C _p(t) gives  

\[X_{\mathrm{app}}(t)/C_{\mathrm{p}}(t)\ {=}\ K_{\mathrm{in}}\ {\times}\ \mathrm{AUC}(t)/C_{\mathrm{p}}(t)\ {+}\ V_{\mathrm{i}}\]

The apparent tissue-to-plasma concentration ratio [V _d(t)] (mL/g tissue) is defined as X _app(t)/C _p(t)  

\[V_{\mathrm{d}}(t)\ {=}\ K_{\mathrm{in}}\ {\times}\ \mathrm{AUC}(t)/C_{\mathrm{p}}(t)\ {+}\ V_{\mathrm{i}}\]

 

when AUC(t)/C _p(t) (minute) is plotted versus V _d(t), as shown in Figure 1 , the early-phase slope represents the K _in in the tissue (μL/(min · g tissue)). The apparent influx permeability clearances of [¹⁴C]DHA, [¹⁴C]AA, or [³H]d-mannitol in retina ( k _{in, retina}), brain (K _{in, brain}), and heart (K _{in, heart}) were determined. 

High-Performance Liquid Chromatography Analysis

The purity of [¹⁴C]DHA prepared in each experiment, and the metabolism of [¹⁴C]DHA in plasma and retina were determined by high-performance liquid chromatography (HPLC). Thirty seconds or 5 minutes after intravenous injection, plasma and retinas were collected and frozen with dry ice. The frozen sample of retina was homogenized in 70% MeOH, and plasma was mixed with 70% MeOH. After centrifugation at 12,550g for 5 minutes, an aliquot of the samples was subjected to HPLC, using a system equipped with an anion exchange column (TSK-Gel NH₂-60; Tosoh, Tokyo, Japan). The mobile phase consisted of 35% 0.05 M KH₂PO₄-65% CH₃CN at a flow rate of 1.0 mL/min. The eluent was collected in vials, and the radioactivity in each fraction was determined by liquid scintillation counting. 

[¹⁴C]DHA and [¹⁴C]AA Uptake by TR-iBRB2 Cells

The conditionally immortalized rat retinal capillary endothelial cell line (TR-iBRB2), which had been established and characterized, ^{21 22 23} was used as an in vitro inner BRB model to characterize DHA and AA transport. TR-iBRB2 cells express GLUT1 protein and have functional 3-O-methyl-d-glucose (3-OMG) transport, with a K _m of 5.56 mM. ²¹ TR-iBRB2 cells (passages 27–38) were cultured at 33°C in Dulbecco’s modified Eagle’s medium (Nissui Pharmaceutical Co., Tokyo, Japan) in 5% CO₂-air, as described previously. ²¹ For the uptake study, cells (5 × 10⁴ cells/cm²) were cultured at 33°C for 2 days on a rat tail collagen-type I–coated 24-well plate (BD Biosciences, San Jose, CA) and washed with 1 mL uptake buffer (134 mM NaCl, 5.2 mM KCl, 0.8 mM MgSO₄, 1.8 mM CaCl₂, and 20 mM HEPES [pH 7.5]) at 37°C. Uptake was initiated by applying 200 μL uptake buffer containing 0.2 μCi [¹⁴C]DHA (69.2 μM) or [¹⁴C]AA (76.9 μM) at 37°C in the presence or absence of inhibitors. For the concentration-dependent study, a concentration range of DHA from 1.7 to 180 μM was prepared using [¹⁴C]DHA from 4.9 nCi to 4.4 μCi. Na⁺-free uptake buffers were prepared in two different ways. The choline- and Li⁺-buffers were prepared by equimolar replacement of NaCl with choline chloride and LiCl, respectively. After a predetermined time period, uptake was terminated by removing the solution, and cells were immersed in ice-cold uptake buffer and solubilized. [¹⁴C]DHA in the uptake buffer was stable over the uptake study period of 3 minutes. Radioactivity was measured by liquid scintillation counting, and the protein content was determined with a kit (DC; Bio-Rad, Hercules, CA) with bovine serum albumin (BSA) as a standard. 

RT-PCR Analysis

Total cellular RNA was prepared from phosphate-buffered saline (PBS)-washed cells using a kit (RNeasy Mini Kit; Qiagen, Hilden, Germany). Single-strand cDNA was made from 1 μg total RNA by reverse transcription (RT), using oligo dT primer. The polymerase chain reaction (PCR) was performed with a gene amplification system (GeneAmp PCR system 9700; Applied Biosystems, Foster City, CA) with GLUT1-, GLUT3-, SVCT1-, SVCT2-, and β-actin–specific primers through 25 cycles of 94°C for 30 seconds, 60°C for 1 minute, and 72°C for 1 minute. The sequences of the specific primers were as follows: sense, 5′-GAT GAT GAA CCT GTT GGC CT-3′; antisense, 5′-AGC GGA CAG CTC CAA GAT G-3′ for rat GLUT1 (Slc2a1, GenBank accession number NM_138827; http://www.ncbi.nlm.nih.gov/Genbank; provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD); sense, 5′-GAC GAG AGT ATC AGG ATG TCA CAG-3′; antisense, 5′-AGG CCA CGT AGA CCA AGA TAG CC-3′ for rat GLUT3 (Slc2a3, GenBank accession number NM_017102); sense, 5′-CCT GTT TAC CGA TGG GGC AAG G-3′ antisense, 5′-ACT CGA TGA TGC CCG CCA GTG T-3′ for rat SVCT1 (Slc23a2, GenBank accession number NM_017316); sense, 5′-ACA CCA CAG AGA TCA CAG TTG CC-3′; antisense, 5′-TGT AAC TTG TAG GCC GTC CAT CC-3′ for rat SVCT2 (Slc23a1, GenBank accession number NM_017315); sense, 5′-TCA TGA AGT GTG ACG TTG ACA TCC GT-3′, antisense, 5′-CCT AGA AGC ATT TGC GGT GCA CGA TG-3′ for the β-actin (GenBank accession number NM_031144). The PCR products were separated by electrophoresis on an agarose gel in the presence of ethidium bromide and visualized under ultraviolet light. The molecular identity of the resultant product was confirmed by restriction analysis with two different restriction enzymes or sequence analysis using a DNA sequencer (Prism 310; Applied Biosystems). 

Quantitative Real-Time PCR

Quantitative real-time PCR was performed on a sequence detection system (Prism 7700 with 2× SYBR Green PCR Master Mix; Applied Biosystems) according to the manufacturer’s protocol. To quantify the amount of specific mRNA in the samples, a standard curve was generated for each run using a plasmid (pGEM-T Easy Vector; Promega, Madison, WI) containing the gene of interest. This enabled standardization of the initial mRNA content of cells relative to the amount of β-actin. The PCR was performed using rat GLUT1-, GLUT3-, SVCT2-, or β-actin–specific primers, and the cycling parameters were those given for RT-PCR analysis. 

Immunostaining Analysis

Cells were cultured on a rat tail collagen-type I–coated coverslip (BD Biosciences, Lincoln Park, NJ) at 33°C for 48 hours. After removal of medium, cells were washed with PBS and fixed in 4% formaldehyde-PBS for 10 minutes at room temperature. Cells were permeated with 0.2% Triton X-100 in PBS for 15 minutes and incubated with blocking agent solution (Block Ace; Dainihon Pharmaceutical Co., Osaka, Japan) for 60 minutes. After washing with PBS, cells were further incubated with rabbit anti-GLUT1 antibody (Chemicon, Temecula, CA), rabbit anti-GLUT3 antibody (Chemicon) or goat anti-SVCT2 antibody (1:200 dilution; Santa Cruz Biotechnology, Santa Cruz, CA) as a primary antibody with 1% BSA for 3 hours at room temperature. Cells were washed with PBS and incubated for 1 hour at room temperature with FITC-conjugated anti-rabbit IgG (Chemicon) or FITC-conjugated anti-goat IgG (Chemicon) (1:50 dilution) as a secondary antibody. Cells were subsequently stained with propidium iodide and viewed by confocal laser scanning microscope (LSM 510; Carl Zeiss Meditec, Oberkochen, Germany). The control experiments were performed in parallel, with normal rabbit or goat IgG used instead of the primary antibody. 

Data Analysis

The uptake of [¹⁴C]DHA and [¹⁴C]AA by TR-iBRB2 cells was expressed as the cell-to-medium (cell/medium) ratio  

\[\mathrm{Cell/medium\ ratio}\ {=}\ ({[}^{\mathrm{14}}\mathrm{C}{]}\ \mathrm{dpm\ per\ mg\ cell\ protein})/({[}^{\mathrm{14}}\mathrm{C}{]}\ \mathrm{dpm\ per}\ {\mu}\mathrm{L\ medium})\]

 

The [³H]d-mannitol uptake study was performed to estimate the volume of adhering water. The resultant cell/medium ratio was 0.1 to 0.2 μL/mg protein, more than 10 times lower than that of [¹⁴C]DHA. Therefore, adhering water was ignored when calculating the cell/medium ratio. 

For kinetic studies, the K _m and the maximum uptake rate (J _max) of DHA were calculated, using the nonlinear least-squares regression analysis program, MULTI ²⁴  

\[J\ {=}\ J_{\mathrm{max}}\ {\times}\ {[}S{]}/(K_{\mathrm{m}}\ {+}\ {[}S{]})\]

where J and [S] are the uptake rate of DHA at 1 minute and the concentration of DHA, respectively. 

The 50% inhibition concentration (IC₅₀) of d-glucose for [¹⁴C]DHA uptake by TR-iBRB2 cells was calculated by fitting the data to a sigmoidal inhibition model ²⁵ using MULTI ²⁴  

\[V\ {=}\ V_{0}/{[}1\ {+}\ ({[}I{]}/\mathrm{IC}_{50})^{n}{]}\]

where V and V ₀ are the uptake of [¹⁴C]DHA in the presence and absence of d-glucose, respectively, and [I] and n is the concentration of d-glucose and the Hill coefficient, respectively. 

Unless otherwise indicated, all data are expressed as the mean ± SEM. Statistical significance of differences among means of several groups was determined by one-way analysis of variance (ANOVA) followed by the modified Fisher least-squares difference method. 

Results

Blood-to-Retina Transport of Vitamin C across the BRB

The in vivo blood-to-retina influx transport of DHA, AA, and d-mannitol across the BRB was evaluated and compared with other tissues by means of the integration plot analysis after intravenous administration of each of the radio-labeled compounds to rats (Fig. 1) . The K _{in, retina} of [¹⁴C]DHA was determined to be 2.44 × 10³ μL/(min · g retina) from the slope representing the apparent influx permeability clearance across the BRB, using equation 4 , whereas the K _{in, retina} of [¹⁴C]AA was 65.4 μL/(min · g retina) (Fig. 1A ; Table 1 ). K _{in, retina} for [¹⁴C]DHA was 37.3-fold greater than that for [¹⁴C]AA. A similar difference was observed in the brain: 38.1 for the ratio between the K _{in, brain} of [¹⁴C]DHA and [¹⁴C]AA, whereas it was only 3.79 in the heart, which does not have a barrier between blood and organ (Figs. 1B 1C ; Table 1 ). In the comparison between retina and brain, the K _{in, retina} of [¹⁴C]DHA and [¹⁴C]AA was approximately eight times greater than the K _{in, brain}, indicating that vitamin C transport across the BRB is greater than that across the BBB. To examine whether blood cells in capillaries contribute to concentrative uptake in the retina, the apparent concentration ratio between blood and plasma (R _B) was measured. The R _B of [¹⁴C]DHA were constant over 10 minutes and not much different from that of [³H]d-mannitol at 3 minutes after administration (Table 1) . These results support the hypothesis that the blood-to-retina DHA transport is much greater than that of AA. 

Figure 2 shows the HPLC chromatograms of [¹⁴C]DHA and [¹⁴C]AA present in the retina and plasma after intravenous administration of 5 μCi [¹⁴C]DHA. [¹⁴C]DHA in plasma was rapidly changed into AA after administration and almost all the [¹⁴C]DHA was converted in 5 minutes (Figs. 2E 2F) . [¹⁴C]DHA accumulated in the retina as AA in a time-dependent manner (Figs. 2C 2D) . 

[¹⁴C]DHA and [¹⁴C]AA Uptake by TR-iBRB2 Cells

To elucidate the transport mechanism of DHA, TR-iBRB2 cells were used as an in vitro rat inner BRB model. ^{21 22 23} The [¹⁴C]DHA and [¹⁴C]AA uptakes by TR-iBRB2 cells exhibited time-dependent increases for at least 3 minutes, with an initial uptake rate of 25.6 μL/(min · mg protein) and 0.700 μL/(min · mg protein), respectively (Fig. 3A) . The uptake clearance of [¹⁴C]DHA in TR-iBRB2 cells was 36.6 times greater than that of [¹⁴C]AA. The [¹⁴C]DHA uptake by TR-iBRB2 cells under Na⁺-free conditions, using choline- or Li⁺-uptake buffer, was not significantly different from the control (Na⁺-uptake buffer; Fig. 3B ). The [¹⁴C]DHA uptake by TR-iBRB2 cells took place in a concentration-dependent manner with a K _m of 93.4 ± 18.8 μM and a J _max of 10.7 ± 1.2 nmol/(min · mg protein) (mean ± SD; Fig. 4 ), indicating that DHA is transported through an Na⁺-independent carrier-mediated transport system in TR-iBRB2 cells. 

Inhibitory Effect of Several Compounds on [¹⁴C]DHA Uptake

The effect of glucose transporters’ substrates and inhibitors on [¹⁴C]DHA uptake by TR-iBRB2 cells is summarized in Table 2 . d-Glucose, 3-OMG, and 2-deoxyglucose (2-DG), glucose transporters’ substrates, at 30 mM, caused marked inhibition (86.4%, 81.1%, and 90.9%, respectively). l-Glucose at 30 mM caused no significant inhibition. Phloretin and cytochalasin B, glucose transporter inhibitors, at 10 μM, produced an inhibition of 60.8% and 83.4%, respectively, whereas the same concentration of phlorizin and cytochalasin E as a control for phloretin and cytochalasin B, respectively, did not have any significant effect. This inhibition of [¹⁴C]DHA uptake supports the hypothesis that facilitative glucose transporters are involved in the uptake process by TR-iBRB2 cells. Moreover, [¹⁴C]DHA uptake was inhibited by d-glucose in a concentration-dependent manner with an IC₅₀ of 5.56 ± 0.57 mM (mean ± SD; Fig. 5 ). 

Expression of GLUT1, GLUT3, SVCT1, and SVCT2 in TR-iBRB2 Cells

To determine vitamin C transporter expression in TR-iBRB2 cells, we performed RT-PCR analysis, using total RNA isolated from TR-iBRB2 cells, brain, and kidney and specific primers of rat GLUT1 and -3 and rat SVCT1 and -2 (Fig. 6A) . β-Actin was used as a housekeeping gene. GLUT1, GLUT3, and SVCT2 mRNA were amplified at 503, 398, and 358 bp, respectively, in TR-iBRB2 cells and brain, whereas SVCT1 mRNA was not. The mRNA expression levels of GLUT1, GLUT3, and SVCT2 were determined in TR-iBRB2 cells by quantitative real-time PCR analysis. The quantity of expression mRNA, compensated with β-actin, for GLUT1, GLUT3, and SVCT2 was 7.39 × 10⁻³ ± 0.91 × 10⁻³, 1.98 × 10⁻⁴ ± 0.25 × 10⁻⁴, and 1.84 × 10⁻⁵ ± 0.26 × 10⁻⁵, respectively (Fig. 6B) . Accordingly, the expression of GLUT1 mRNA was 37.3 and 402 times greater, respectively, than that of GLUT3 and SVCT2 mRNA in TR-iBRB2 cells. 

The expression and localization of GLUT1, GLUT3, and SVCT2 protein in TR-iBRB2 cells were examined by confocal laser scanning microscope (Fig. 7) . The immunostaining by anti-GLUT1 (Figs. 7A 7D) and anti-GLUT3 antibody (Figs. 7B 7E) was observed in TR-iBRB2 cells. No significant fluorescence was observed in TR-iBRB2 cells stained with anti-SVCT2 antibody (Figs. 7C 7F) and normal IgG (data not shown). An x-z section in GLUT1 and GLUT3 staining (Figs. 7D 7E) showed that fluorescence (green) was located over the cell nucleus (red), providing supporting evidence that GLUT1 and GLUT3 are mainly localized on the cell surface of TR-iBRB2 cells. 

Discussion

The present study produces, for the first time, in vivo evidence that vitamin C is mainly transported as DHA across the BRB and accumulates as AA in the retina (Figs. 1 and 2) . DHA is transported by a facilitative glucose transporter, most likely GLUT1, which is expressed at the luminal (blood) and abluminal (retina) side of the inner BRB and RPE (outer BRB), ¹⁷ although the additional contribution of GLUT3 cannot be ruled out at the present time. ²⁶ GLUT1 mRNA expression in TR-iBRB2 cells used as an in vitro model of inner BRB was 37 times greater than that of GLUT3. Nevertheless, GLUT1 and -3 proteins were substantially expressed on the cell surface of TR-iBRB2 cells (Figs. 6 7) . [¹⁴C]DHA uptake by TR-iBRB2 cells took place in an Na⁺-independent and concentration-dependent manner, with a K _m of 93.4 μM (Figs. 3B 4) . This K _m is similar to that obtained for DHA uptake (K _m = 60 μM) in Xenopus oocytes expressing GLUT1. ¹² Moreover, [¹⁴C]DHA uptake by TR-iBRB2 cells was strongly inhibited by glucose transporter substrates, such as d-glucose, 3-OMG, and 2-DG and inhibitors, such as phloretin and cytochalasin B (Table 2) . ²⁷  

The K _{in, retina} of [¹⁴C]DHA (2.44 × 10³ μL/(min · g retina)) was 37 times greater than that of [¹⁴C]AA (65.4 μL/(min · g retina); Fig. 1A ). This result agrees with the ratio between [¹⁴C]DHA and [¹⁴C]AA uptake clearance in TR-iBRB2 cells (Fig. 3A) and supports the hypothesis that DHA is predominantly transported through facilitative glucose transporters at the BRB rather than by AA. However, AA transport into the retina cannot be fully ruled out at the present time, because the K _{in, retina} of [¹⁴C]AA was significantly greater than that of [³H]d-mannitol, which was used as a nonpermeable paracellular marker (Fig. 1A) . In addition, an Na⁺-dependent AA transport process seems to be present in the bovine RPE. ¹⁰ In the brain, a similar difference in the K _{in, brain} between [¹⁴C]DHA and [¹⁴C]AA supports the hypothesis that glucose transporters at the blood–organ barriers such as the BRB and BBB facilitate transport of DHA, but not of AA, as reported at the BBB, ¹⁶ since, in the heart, there was not a great difference in the K _{in, heart} between [¹⁴C]DHA and [¹⁴C]AA (Table 1) . The K _{in, retina} of [¹⁴C]DHA and [¹⁴C]AA was approximately eight times greater than the corresponding values in brain (Table 1) . Possible reasons for this are that the amounts of retinal and brain capillary endothelial cells represent a small percentage of the weight of the entire retina and 0.1% to 0.2% of the weight of the entire brain, respectively, ^{28 29} and RPE (outer BRB) contributes to the supply of essential molecules in the outer segment of the retina, whereas choroid plexus epithelial cells (blood–cerebrospinal fluid barrier) do not play a major role in supplying them for the entire brain. ^{7 30} Moreover, Root-Bernstein et al. ³¹ reported the evidence that DHA, rather than AA, is taken up by human RPE, and its uptake is inhibited by d-glucose in a concentration-dependent manner. ³¹ GLUT1 expression in rat retinal capillary endothelial cells is greater than that in rat brain capillary endothelial cells. ^{32 33}  

The innate vitamin C regulatory mechanism in the retina most likely involves GLUT1 supplying DHA to the retina at the luminal and abluminal sides of the inner BRB and RPE (outer BRB), and the transported DHA is reduced to AA and accumulates in the retina as an antioxidant. Even though GLUT1 is not a concentrative transporter, DHA is rapidly reduced to AA and thus is trapped within the retina (Fig. 2) . The conversion of [¹⁴C]DHA to [¹⁴C]AA in plasma is very rapid compared with the initial uptake of [¹⁴C]DHA (Figs. 1 2) . The level of [¹⁴C]DHA remaining in plasma seems to be underestimated, because it cannot be ignored that [¹⁴C]DHA converts to [¹⁴C]AA during manipulation of the assay. ³⁴ Notably, blood-to-retina influx transport of [¹⁴C]DHA takes place, since it was much greater than that of [¹⁴C]AA (Fig. 1 , Table 1 ). Although the affinity of DHA for facilitative glucose transporters (K _m = 93.4 μM, Fig 4 ) is greater than that of d-glucose, (the K _m estimated for d-glucose uptake by the retina across the rat BRB was 7.81 mM), ³⁵ DHA uptake through facilitative glucose transporters is competitively inhibited by d-glucose, and the normal plasma d-glucose concentration in most mammals is approximately 5 mM. d-Glucose inhibited [¹⁴C]DHA uptake by TR-iBRB2 cells, with an IC₅₀ of 5.56 mM (Fig. 5) . Therefore, DHA transport by facilitative glucose transporters across the BRB does not exhibit complete inhibition (i.e., approximately 50%), under normal conditions. The DHA plasma concentration has been recently determined to be approximately 10 μM (10%–20% of total plasma ascorbate concentration) in the rat and human. ^{14 15} Moreover, DHA is produced by metal-binding proteins, such as serum albumin and by superoxide anions in endothelial cells. ^{36 37} However, hyperglycemia (i.e., diabetic mellitus) increases the blood d-glucose concentration to 20 mM or higher, leading to the inhibition of DHA transport at the BRB. ³¹ Indeed, [¹⁴C]DHA uptake by TR-iBRB2 cells was inhibited by 79% at a concentration of 20 mM d-glucose (Fig. 5) . Although there is contradictory evidence showing regulation of GLUT1 expression in retinal capillary endothelial cells under diabetic conditions, ^{33 38} rats with streptozotocin-induced diabetes show downregulation of GLUT1 expression by 50% in the retina. ³³ In light of these findings, diabetic patients may experience enhanced oxidative stress in the retina because of reduced influx of DHA, leading to the hypothesis that diabetic retinopathy involves dysfunction of DHA influx at the BRB. 

From a pharmacological viewpoint, intravenous administration of DHA may be effective for retinal ischemia-reperfusion to protect the neural retina against oxidative stress. Huang et al. ³⁹ reported that intravenous administration of DHA mediates cerebroprotection after reperfused and nonreperfused cerebral ischemia, but this did not happen with AA. This suggests that DHA can use a transportable prodrug of AA across the BRB and BBB to exert its neuroprotective effects. However, DHA has membrane-disruptive effects in erythrocytes and renal brush border membrane and may destroy the pancreatic beta cells. ^{40 41} Further studies are needed to elucidate the relationship between pharmacologically effective and toxicologically adverse concentrations of DHA in plasma. 

In conclusion, vitamin C is predominantly transported as DHA through facilitative glucose transporters at the BRB and accumulates as AA in the retina. The physiological role of facilitative glucose transporters at the BRB appears to involve the supply of vitamin C from the circulating blood to the retina to protect the neural retina against oxidative stress. These findings provide important information to help us understand the physiological and pathophysiological roles of facilitative glucose transporters at the BRB and assist in the design of a suitable DHA dosage regimen for pharmacological therapies. 

 Supported in part by a Grant-in-Aid for Scientific Research from Japan Society for the Promotion of Science, the Nakatomi Foundation, the Suzuken Memorial Foundation, and the Mochida Memorial Foundation for Medical and Pharmaceutical Research.

 Submitted for publication May 22, 2003; revised August 9 and December 25, 2003; accepted January 5, 2004.

 Disclosure: K.-i. Hosoya, None; A. Minamizono, None; K. Katayama, None; T. Terasaki, None; M. Tomi, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Integration plot of the initial uptake of [¹⁴C]DHA, [¹⁴C]AA, and [³H]d-mannitol by the retina (A), brain (B), and heart (C) after intravenous administration. [¹⁴C]DHA (○, 5 μCi/rat), [¹⁴C]AA (▵, 10 μCi/rat), or [³H]d-mannitol (▪, 10 μCi/rat) was injected into the femoral vein. The counts of [³H]d-mannitol uptake by the retina were close to background. Each point represents the mean ± SEM (n = 3–5).

Table 1.

 

The Apparent Influx Permeability Clearance (K _in) per Gram Rat Tissue and Blood/Plasma Ratio (R _B) of [¹⁴C]DHA, [¹⁴C]AA, and [³H]d-Mannitol

Groups	Kin, retina (μL/(min · g retina))	Kin, brain (μL/(min · g brain))	Kin, heart (μL/(min · g heart))	R B
[14C]DHA	2.44×103± 0.05×103	309 ± 53	49.7 ± 15.1	0.65 ± 0.01
[14C]AA	65.4 ± 10.5	8.12 ± 1.34	13.1 ± 11.0	ND
[3H]d-Mannitol	ND	1.65 ± 0.38	10.8 ± 5.25	0.58 ± 0.03

  K _in was estimated from the initial slope in Figure 1 and is expressed as the mean ± SD. R _B was determined at 3 minutes after intravenous administration of [¹⁴C]DHA or [³H]d-mannitol and is expressed as the mean ± SEM (n = 3). ND, not determined.

Figure 2.

 

Typical HPLC chromatogram of samples of retina (C, D) and plasma (E, F) after intravenous administration of [¹⁴C]DHA (B). [¹⁴C]DHA was generated by incubating [¹⁴C]AA (A) with ascorbate oxidase (1 U/1 mmol AA). [¹⁴C]DHA (5 μCi/rat) was injected into the femoral vein, and retinas and plasma were collected at 30 seconds (C, E) and 5 minutes (D, F).

Figure 3.

 

Time-course of [¹⁴C]DHA and [¹⁴C]AA uptake (A) and Na⁺ independence of [¹⁴C]DHA uptake by TR-iBRB2 cells (B). (A) [¹⁴C]DHA (○, 0.2 μCi) and [¹⁴C]AA (▵, 0.2 μCi) uptake were performed at 37°C. (B) [¹⁴C]DHA (0.2 μCi) uptake was performed in the presence or absence of Na⁺ (Na⁺ was replaced with equimolar choline or Li⁺) at 1 minute and at 37°C. The uptake was expressed as the cell/medium ratio according to equation 5 . Each point represents the mean ± SEM (n = 4).

Figure 4.

 

Concentration-dependence of DHA uptake by TR-iBRB2 cells. [¹⁴C]DHA uptake was performed at 2 minutes at 37°C, over the concentration range from 1.7 μM (4.9 nCi) to 180 μM (4.4 μCi). Data are the mean ± SEM (n = 4). The K _m and J _max were 93.4 ± 18.8 μM and 10.7 ± 1.2 nmol/(min · mg protein), respectively (mean ± SD).

Table 2.

 

Inhibitory Effect of Several Compounds on [¹⁴C]DHA Uptake by TR-iBRB2 Cells

Inhibitors	Percentage of Control
Control	100 ± 12
30 mM d-Glucose	13.6 ± 3.1*
30 mM l-Glucose	81.3 ± 2.7
30 mM 3-OMG	18.9 ± 1.3*
30 mM 2-DG	9.10 ± 1.8*
10 μM Phloretin	39.2 ± 8.3*
10 μM Phloridzin	91.7 ± 8.9
10 μM Cytochalasin B	16.6 ± 7.5*
10 μM Cytochalasin E	100.3 ± 4.6

 [¹⁴C]DHA uptake was performed for 2 minutes at 37°C in the presence or absence of inhibitors. Data are expressed as the mean ± SEM (n = 4–12). 3-OMG: 3-O-methyl-d-glucose, 2-DG: 2-deoxyglucose.

*   P < 0.01, significantly different from control.

Figure 5.

 

Inhibitory effect of d-glucose on [¹⁴C]DHA uptake by TR-iBRB2 cells. [¹⁴C]DHA uptake was performed in the presence (0.5–50 mM) or absence (control) of d-glucose at 2 minutes and at 37°C. Each point represents the mean ± SEM (n = 4). The IC₅₀ is 5.56 ± 0.57 mM (mean ± SD).

Figure 6.

 

RT-PCR analysis of GLUT1, GLUT3, SVCT1, SVCT2, and β-actin (A) and the amount of GLUT1, GLUT3, and SVCT2 mRNA (B) in TR-iBRB2 cells. (A) Lane 1: rat brain; lane 2: rat kidney; lane 3: TR-iBRB2 cells; lane 4: in the absence of reverse transcriptase for TR-iBRB2 cells. Rat brain was used as a positive control for GLUT1, GLUT3, and SVCT2, and rat kidney was used as a positive control for SVCT1. (B) The amount of GLUT1, GLUT3, and SVCT2 mRNA in TR-iBRB2 cells was determined by quantitative real-time PCR analysis. Data are the mean ± SEM (n = 4). The GLUT1, GLUT3, and SVCT2 mRNA content relative to β-actin mRNA (GLUT1/β-actin, GLUT3/β-actin, and SVCT2/β-actin) was 7.39 × 10⁻³ ± 0.91 × 10⁻³, 1.98 × 10⁻⁴ ± 0.25 × 10⁻⁴, and 1.84 × 10⁻⁵ ± 0.26 × 10⁻⁵, respectively.

Figure 7.

 

Immunostaining analysis of GLUT1, GLUT3, and SVCT2 in TR-iBRB2 cells. TR-iBRB2 cells were immunostained by anti-GLUT1 (A, D), anti-GLUT3 (B, E), and anti-SVCT2 (C, F) antibodies. (A, B, C) x-y sections; (D, E, F) x-z sections. Green: immunoreactivity by each antibody; red: nuclei stained by propidium iodide. Scale bar, 10 μm.

The authors thank Hisashi Iizasa and Masanori Tachikawa for valuable discussion.