Dehydroascorbate transport in human chondrocytes is regulated by hypoxia and is a physiologically relevant source of ascorbic acid in the joint




To evaluate the dehydroascorbate (DHA) transport mechanisms in human chondrocytes.


The transport of L-14C-DHA in human chondrocytes was analyzed under various conditions, including the use of RNA interference (RNAi), to determine the role of glucose transporter 1 (GLUT-1) and GLUT-3 in L-14C-DHA transport and to evaluate the effects of physiologically relevant oxygen tensions on L-14C-DHA transport. In order to estimate the contributions of reduced ascorbic acid (AA) and DHA to intracellular ascorbic acid (Asc), the quantities of AA and DHA were measured in synovial fluid samples from osteoarthritis (OA) patients and compared with the reported levels in rheumatoid arthritis (RA) patients.


DHA transport in human chondrocytes was glucose-sensitive, temperature-dependent, cytochalasin B–inhibitable, modestly stereoselective for L-DHA, and up-regulated by low oxygen tension. Based on the RNAi results, GLUT-1 mediated, at least in part, the uptake of DHA, whereas GLUT-3 had a minimal effect on DHA transport. DHA constituted a mean 8% of the total Asc in the synovial fluid of OA joints, in contrast to 80% of the reported total Asc in RA joints.


We provide the first evidence that chondrocytes transport DHA via the GLUTs and that this transport mechanism is modestly selective for L-DHA. In the setting of up-regulated DHA transport at low oxygen tensions, DHA would contribute 26% of the total intracellular Asc in OA chondrocytes and 94% of that in RA chondrocytes. These results demonstrate that DHA is a physiologically relevant source of Asc for chondrocytes, particularly in the setting of an inflammatory arthritis, such as RA.

Ascorbic acid (Asc), more commonly known as vitamin C, is required for the synthesis of the most abundant protein in cartilage, type II collagen. Asc exists in 2 forms: the reduced form (AA) and the oxidized form, dehydroascorbate (DHA). Both AA and DHA are available through the diet from fruits and vegetables and are absorbed throughout the entire length of the small intestine (1). Diet constitutes the sole source of Asc for humans because we lack the enzyme gulonolactone oxidase, which is necessary for the synthesis of Asc from glucose (2). During biosynthetic and antioxidant reactions, AA is reversibly oxidized to DHA. Intracellular DHA can be reduced back to AA by NADPH-dependent thioredoxin reductase (3), glutathione-dependent DHA reductase (3), and glutaredoxin (4). This regenerates AA for utilization as an enzymatic cofactor and an antioxidant. DHA recycling in this manner results in low intracellular concentrations of DHA, which favors the continued cellular uptake of DHA (3). Although DHA can theoretically contribute to levels of intracellular Asc, the mechanism of DHA transport and the physiologic relevance of this transport process in chondrocytes have never been investigated.

Two distinct pathways of Asc transport across cellular membranes have been discovered to date, the sodium-dependent vitamin C transporters (SVCTs) (5) and the glucose transporters (GLUTs) (6, 7). The majority of vitamin C in plasma exists as AA (8), which is transported by the SVCTs (5). We have recently shown that human chondrocytes express SVCT-2, but not SVCT-1, and that this mechanism of transport is capable of achieving high intracellular AA levels, with concentrations elevated up to 960-fold over those in the extracellular milieu (9).

In contrast, the GLUTs are numerous and are ubiquitous in their expression. To date, a total of 13 members of the GLUT family have been described (10, 11). Thus far, several GLUTs have been shown to be capable of transporting DHA, including GLUT-1 (6), GLUT-3 (7), GLUT-4 (7, 12), and possibly, GLUT-2 (6, 7). Several GLUTs have also been reported to be incapable of transporting DHA, including GLUT-5 and SGLT-1, a sodium-dependent D-glucose cotransporter (7). Of the GLUTs currently known to transport DHA, only GLUT-1 and GLUT-3 are expressed by human articular chondrocytes (13–15). Nevertheless, it has been assumed that DHA contributes little to intracellular levels of Asc because of the low circulatory levels of DHA (16), the instability of DHA (17), and a significant inhibition of DHA transport by physiologic concentrations of glucose (∼10 mM). (The 50% inhibition concentration [IC50] of GLUT-1 and GLUT-3 by D-glucose is 10 mM and 4 mM, respectively [7].)

However, not taken into consideration is the fact that articular cartilage is avascular and functions at a lower oxygen tension than most tissues. This may alter the tissue redox state and, thus, the local concentration of DHA, and/or it may regulate the transport pathways for DHA. A gradient of oxygen tensions exists throughout the zones of cartilage, from ∼6% oxygen in the superficial zone to a lower limit of ∼1% oxygen in the deeper zones (18, 19). Hypoxia has been shown to alter gene expression in chondrocytes (20–26); therefore, this unique feature of cartilage was taken into consideration in our experiments in order to more accurately reflect physiologic conditions. Our data represent the first evidence that chondrocytes transport DHA via a mechanism that is up-regulated by low oxygen tension and that this transport mechanism is physiologically relevant in cartilage.


Chondrocyte isolation and cell culture.

Articular cartilage was obtained from surgical waste tissues at the time of knee replacement surgery in 7 patients with osteoarthritis (OA). Cartilage samples were harvested from nonlesional areas, and primary human chondrocytes were isolated. Each cartilage specimen was minced, and chondrocytes were isolated by enzymatic digestion, similar to previously published methods (27). Cells at each passage were tested for the expression of the chondrocyte genes, type II collagen, and aggrecan, using reverse transcription–polymerase chain reaction (RT-PCR).

DHA transport assay in isolated primary human chondrocytes.

DHA transport was measured using a modified version of the uptake assay described by Wilson and Dixon (28). Primary human chondrocytes were seeded at a density of 4.5 × 105 cells/well on a 6-well plate. Prior to the transport assay, cells were plated for 48 hours in Dulbecco's modified Eagle's medium/Ham's F-12 containing 10% fetal calf serum, but no AA or DHA. Cells were washed in transport buffer containing 134 mM LiCl, 5.4 mM KCl, 1.8 mM CaCl2, 0.8 mM MgSO4, 20 mM HEPES, pH 7.3, with 1N KOH. DHA uptake was assessed by incubating the chondrocytes in transport buffer containing 189 μM L-14C-DHA. Stock solutions of L-14C-DHA were generated immediately before use by incubating L-14C-AA (4 mCi/mmole; NEN Life Science Products, Boston, MA) in 0.4 mM homocysteine with an AA oxidase spatula (Roche Diagnostics, Mannheim, Germany) for 15 minutes at room temperature, with stirring. The standard DHA transport assay conditions were a temperature of 37°C, an atmosphere containing 5% CO2 and 21% O2, and a duration of 10 minutes. DHA uptake was terminated by washing the cells 4 times in ice-cold phosphate buffered saline (PBS).

The cells were lysed with Puregene cell lysis solution (Gentra Systems, Minneapolis, MN), and scintillation counting was performed in Uniscint BD (National Diagnostics, Atlanta, GA). Total picomoles of L-14C-DHA transported was calculated using the total disintegrations per minute of L-14C-DHA transported and the specific activity of the radiolabel.

We performed DHA transport assays under various conditions designed to augment or inhibit the function of the DHA transporters, including variations of the following: presence or absence of 134 mM Na+; presence or absence of 10 mM, 50 mM, or 100 mMD-glucose; temperature of 4°C or 37°C; 0.1 mM cytochalasin B (Sigma, St. Louis, MO) and 0.1 mM cytochalasin E (Sigma); 2 mM unlabeled L-DHA, D-DHA, sodium L-DHA, sodium D-DHA, L-AA (Sigma), D-isoascorbic acid (D-AA; Aldrich), sodium L-AA (Sigma), and sodium D-AA (Aldrich); and 1%, 2.5%, 5%, and 21% O2 for 24 or 72 hours. Cytochalasin B inhibits DHA transport via the GLUTs (29), and it also inhibits actin polymerization. Cytochalasin E has no known inhibitory effects on transport mechanisms, but it also inhibits actin polymerization and, thus, is a control for the actin effects of cytochalasin B. The properties of the AA forms used in these experiments have been discussed in detail previously (30). Stock solutions of different unlabeled forms of DHA were generated in the same manner as the L-14C-DHA.

High-performance liquid chromatography (HPLC).

The integrity of the L-14C-DHA stocks utilized in these experiments was determined by measuring the proportion of AA and DHA by HPLC. AA was measured with an electrochemical detector, using the method of Lee et al (31). The amount of DHA in the stocks and samples was determined by reducing the DHA to AA, as previously described (32).

RNA isolation.

Monolayer cells were grown to confluency on 35-mm plates, and the cells were lysed using 1 ml of TRIzol reagent (Gibco, Grand Island, NY). The RNA extraction procedure was performed according to the manufacturer's protocol through the phase separation step. The aqueous phase was transferred to a new tube, and 10 μg of transfer RNA (Sigma) was added. Then, 0.5 ml of isopropanol was added, and the sample was frozen overnight at −80°C. The RNA was pelleted at 13,000 revolutions per minute for 20 minutes at 4°C, the liquid phase was discarded, and the pellet was air dried. The RNA was then processed according to the manufacturer's protocol for the Qiagen RNeasy kit (Qiagen, Valencia, CA).


Total RNA was reverse transcribed into complementary DNA (cDNA), using Superscript II reverse transcriptase (Gibco) and random hexamer primers. The use of random hexamers allowed the amplification of 18S ribosomal RNA (rRNA) as a control. Intron-spanning primers were designed for GLUT-1 (13), GLUT-3 (13), GLUT-6 (5′-TTGCTGCCAACCTGACTCTG-3′ and 5′-GTCCTTCACGCAAGGGAAAG-3′), GLUT-8 (5′-ACATCTCCGAAATCGCCTAC-3′ and 5′-CCGATGATGAAGGGCTTGTA-3′), GLUT-9 (5′-TGCTGAGCCTTCCCTTTCTC-3′ and 5′-CCACTGCAGAAAGAGGCGAT-3′), GLUT-10 (5′-AGGACCAATGAGGACCAAAG-3′ and 5′-AGGAAGGAGAGGCTGATGAA-3′), and GLUT-11 (5′-TCATCAATGCCCCGACCTTG-3′ and 5′-TCATTCCCGCAGAGCTCCAT-3′), which corresponded to the human sequences available in GenBank. Primers specific for α1 type II collagen and aggrecan were generously provided by Dr. Carl Flannery (Genetics Institute, Cambridge, MA). Annealing temperatures were 50°C for GLUT-1, 58°C for GLUT-3, 68°C for GLUT-6 and GLUT-9, 65°C for GLUT-8 and GLUT-10, 71°C for GLUT-11, 55°C for α1 type II collagen, and 61.3°C for aggrecan. Standard PCR procedures were used with AmpliTaq Gold DNA polymerase (Roche Diagnostics).

Functional analysis of GLUT-1 and GLUT-3.

RNA interference (RNAi).

The RNAi transfection procedure was performed using a Nucleofector device from Amaxa (Gaithersburg, MD), the primary human chondrocyte kit (Amaxa), and a total of 3 μg of short interfering RNAs (siRNA). Chondrocytes were transfected on program U24 with the following siRNA: 3 μg of Silencer Negative Control No. 1 siRNA (Ambion, Austin, TX) to control for nonspecific effects, a pool of 3 different human GLUT-1–specific siRNA (1 μg each of siRNA identification nos. 17981, 18074, and 18160; Ambion), or a pool of 2 different human GLUT-3–specific siRNA (2.5 μg each of siRNA identification nos. 18786 and 18882; Ambion). The cells were incubated at 37°C in an atmosphere containing 5% CO2 and 21% O2 for 65–72 hours after the transfection to allow suppression of the GLUT-1 and GLUT-3 messenger RNA (mRNA) expression and turnover of preexisting GLUT proteins. The cells were subsequently analyzed for DHA transport (described above) or for GLUT-1 and GLUT-3 protein concentrations, or they were treated with TRIzol to isolate the RNA for real-time RT-PCR.

Real-time RT-PCR.

The cDNA generated from the RNAi experiments was subjected to real-time RT-PCR to quantify the changes in gene expression that occurred upon suppression of the GLUT-1 and GLUT-3 transcript levels. The ABI Prism 7000 sequence detection system and relative quantification software (Applied Biosystems, Foster City, CA) were used for the real-time analyses. The real-time reactions were each performed in quadruplicate in a final volume of 25 μl, according to the manufacturer's instructions. Expression levels of GLUT-1 and GLUT-3 were compared between chondrocytes from the same specimen that had been transfected with either the negative control siRNA or the GLUT-1 or GLUT-3 siRNA. Transcript levels were determined by real-time RT-PCR, using the following Applied Biosystems primer and probe sets: 18S-PDAR (18S rRNA), Hs00197884_m1 (GLUT-1), and Hs00359840_m1 (GLUT-3).

GLUT-1 and GLUT-3 mRNA levels were normalized to the 18S rRNA levels, and relative quantification was determined by the 2math image formula, where ΔCt represents the difference between the threshold cycle (Ct) values for the negative control and GLUT-1– or GLUT-3–transfected cells (33). We chose to use 18S rRNA for normalization since other housekeeping genes, such as GAPDH and β-actin, have been shown to be regulated in chondrocytes (34). In addition, Grimshaw and Mason (21) have demonstrated that the abundance of 18S rRNA is not altered under a variety of oxygen tensions, including <0.1% oxygen tension for 2 days or 5%, 10%, or 20% oxygen tension for 2 or 7 days. Based on this information, we thought that 18S rRNA was the only housekeeping gene that had been validated in chondrocytes under different oxygen tensions and times and, thus, was the most appropriate normalization control for our experiments. One limitation of using 18S for normalization is the fact that, in order to detect 18S rRNA by real-time PCR, it was necessary to dilute the stock solution of cDNA that was used for the GLUT-1 and GLUT-3 real-time PCRs. While it is possible that a pipetting error could occur during this dilution step, all results obtained for the 18S expression levels were nearly equivalent. The data are expressed as a percentage of the mean fold change in levels of mRNA for the experimental samples (GLUT-1 or GLUT-3 siRNA) as compared with the calibrator (negative control siRNA).

Western blot analyses.

Whole cell lysates (in the Puregene cell lysis solution) from the RNAi experiments were evaluated for GLUT-1 and GLUT-3 protein expression. Protein concentrations were determined using the Detergent Compatible Protein Assay (Bio-Rad, Hercules, CA). For each condition tested, a total of 40 μg of protein was separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis on 12% minigels, under reducing conditions. Proteins were transferred onto Immun-Blot polyvinylidene difluoride (Bio-Rad) membranes and incubated overnight at 4°C in blocking buffer (5% nonfat dry milk in PBS/0.1% Tween 20 [PBS–Tween]). Membranes were washed with PBS–Tween between antibody treatments.

Polyclonal primary antibodies against GLUT-1 (AB1341; Chemicon, Temecula, CA) and GLUT-3 (AB1345; Chemicon) were diluted 1:1,000 in blocking buffer and incubated with the membranes for 1 hour at room temperature. Anti-rabbit IgG-HRP (Sigma) secondary antibody was diluted 1:16,000 and incubated with the membranes for 1 hour at room temperature. Membranes were subsequently washed 3 times in Tris buffered saline, and antigen–antibody complexes were visualized using ECL Plus Western Blotting Detection Reagents (Amersham Biosciences, Piscataway, NJ). The resulting films were scanned using a Hewlett-Packard ScanJet 3c page scanner (Hewlett-Packard, McMinnville, OR), and the images were analyzed using Adobe Photoshop CS (Adobe Systems, San Jose, CA) and image analysis software (NIH Image 1.6; National Institutes of Health, Bethesda, MD; online at to determine the average pixel density for each band. The percentage of GLUT-1 and GLUT-3 protein in each sample was determined relative to the average pixel density of GLUT-1 and GLUT-3 in the siControl samples, which was set at 100%. Coomassie staining of the transferred gels was used to verify equal protein loading of each sample.

Synovial fluid and plasma collection.

Synovial fluid and blood were obtained from a subset of 23 participants in the Prediction of Osteoarthritis Progression (POP) study, with the approval of the Duke University Institutional Review Board. All participants had symptomatic OA of the knee, with grades 1–4 radiographic severity according to the Kellgren/Lawrence scoring system (35). Joint fluid was aspirated from both knees when feasible, and blood was drawn concurrently. Samples were collected into cold perchloric acid (to stabilize AA) (32) or into empty tubes on ice (using fresh synovial fluid or EDTA-treated plasma immediately upon thawing). Within 90 minutes, homocysteine was added to the untreated samples and incubated at room temperature for 30 minutes to reduce DHA to AA for measurement of total Asc (32), followed by the addition of cold perchloric acid. All samples were frozen at −80°C until HPLC was performed (as described above) to determine the amount of AA and DHA in the samples. Synovial fluid samples from a total of 33 knees were included in this study.

Statistical analyses.

Statistical computations were performed using GraphPad Prism version 3.00 (GraphPad Software, San Diego, CA) and the data analysis feature of Microsoft Excel (Microsoft, Redmond, WA). Descriptive statistics, sample means, and standard errors for all values were calculated for the subgroups of interest. For descriptive purposes, pairwise comparisons between subgroups of interest were performed using analysis of variance and the Newman-Keuls post hoc test or the paired t-test. P values less than 0.05 were considered significant. The percentages of AA and DHA in synovial fluid and blood from OA patients in this study were compared with percentages reported in the literature for samples from patients without arthritis and from patients with rheumatoid arthritis (RA) (36).


Glucose transporter expression.

Primary human chondrocytes expressed transcripts for all 7 of the glucose transporters analyzed: GLUT-1, GLUT-3, GLUT-6, GLUT-8, GLUT-9, GLUT-10, and GLUT-11 (data not shown). The chondrocytic phenotype of these cells was confirmed by RT-PCR, showing mRNA expression of the major protein components in cartilage (type II collagen and aggrecan) at all of the cell passages used for these experiments (data not shown).

Glucose suppression, temperature dependence, and cytochalasin B inhibition of DHA transport.

The transport of DHA into primary human chondrocytes was highest in the absence of glucose at 37°C (Figure 1A). DHA transport was suppressed by 66% in the presence of 10 mM glucose at both 37°C (P < 0.001) and 4°C and by up to 80% in the presence of 100 mM glucose (P < 0.001) (Figure 1B). Passive diffusion of DHA into chondrocytes, represented by transport at 4°C in the presence of 10 mM glucose, constituted 11% of transport under optimized conditions (37°C without glucose). There was no statistically significant reduction in DHA transport by sodium in the presence or absence of glucose, although in the absence of glucose, there was a slight inhibitory effect (Figure 1A). Therefore, the sodium independence and the inhibition of DHA transport by glucose at both 37°C and 4°C suggested a passive uptake mechanism compatible with sodium-independent GLUT-mediated transport.

Figure 1.

Glucose suppression, temperature dependence, and cytochalasin B inhibition of dehydroascorbate (DHA) transport. A, L-14C-DHA uptake was measured after 10 minutes at either 37°C or 4°C in the presence and absence of 134 mM sodium and/or 10 mMD-glucose (Glc). Values are the mean and SEM picomoles of L-14C-DHA/4.5 × 105 cells. ∗ = P < 0.001 versus the same treatment at 37°C. B, L-14C-DHA uptake was measured after 10 minutes in the presence of 50 mM or 100 mMD-glucose, 0.1 mM cytochalasin B (Cyto B), or 0.1 mM cytochalasin E (Cyto E). The amount of L-14C-DHA transported in the samples without inhibitor was set at 100%. Values are the mean and SEM percentage of L-14C-DHA transported. ∗ = P < 0.001 versus the sample without inhibitor.

In addition to a dose-dependent inhibition of DHA transport by glucose, the transport of DHA into primary human chondrocytes was inhibited by 95% in the presence of cytochalasin B (P < 0.001), a specific inhibitor of the GLUTs and of actin polymerization. Cytochalasin E, a control for the actin effects of cytochalasin B, suppressed DHA transport by 40% (P < 0.001). Thus, the majority of the inhibition of DHA transport by cytochalasin B was attributable to inhibition of the GLUTs. Treatment with cytochalasin B decreased the transport of L-14C-DHA to a level representing passive diffusion alone. These results taken together suggested that the transport of DHA by human chondrocytes was mediated entirely by GLUTs.

Modest stereoselective transport of the L-forms of DHA.

To determine the stereospecificity of DHA uptake by chondrocytes, the L-14C-DHA uptake assay was performed in the presence of various DHA and AA forms (Figure 2). The following isoforms were able to effectively compete with L-14C-DHA transport: L-DHA, D-DHA, sodium L-DHA, L-AA, D-AA, sodium L-AA, and sodium D-AA (P < 0.05). Only sodium D-DHA was unable to compete with L-14C-DHA for transport into chondrocytes. Additionally, Dixon plot analyses demonstrated that L-AA and D-AA were noncompetitive inhibitors of L-14C-DHA transport (data not shown). The L-forms of both DHA and AA were able to compete more efficiently than the D-forms of both DHA and AA for transport into primary human chondrocytes. Generally, the sodium-containing forms of both DHA and AA were able to compete more efficiently than the sodium-free forms of both DHA and AA for transport into primary human chondrocytes. Overall, these results demonstrated a modest L-form stereospecificity of DHA transport by chondrocytes.

Figure 2.

Modest stereoselective transport of the L-forms of dehydroascorbate (DHA). L-14C-DHA uptake was measured after 10 minutes in the presence of a 2 mM concentration of the following unlabeled competitors: no competitor (None), L-DHA, D-DHA, sodium L-DHA (Na L-DHA), sodium D-DHA (Na D-DHA), L-ascorbic acid (L-AA), D-isoascorbic acid (D-AA), sodium L-AA (Na L-AA), or Na D-AA (Na D-AA). Values are the mean and SEM percentage of L-14C-DHA transported. The amount of L-14C-DHA transported in the samples without competitor was set at 100%. a = P < 0.05 versus the sample with no competitor; b = P < 0.05 versus the sample with the corresponding D-form; c = P < 0.001 versus sodium L-AA.

Suppression of GLUT-1 and GLUT-3 expression by RNAi.

To assess the role of GLUT-1 and GLUT-3 in mediating DHA transport in primary human chondrocytes, we suppressed the expression of GLUT-1 and GLUT-3 with sequence-specific siRNA (Figure 3). GLUT-1 mRNA levels were decreased ∼90% by GLUT-1 siRNA, as compared with the negative control (P < 0.01) (Figure 3A). GLUT-3 transcript levels were suppressed ∼80% by GLUT-3 siRNA (P < 0.01). Under these same conditions, GLUT-1 protein concentrations were reduced 41% (P < 0.05) and GLUT-3 protein concentrations were suppressed 39% (P < 0.05), compared with the protein concentrations in cells treated with negative control siRNA (Figure 3B). As shown in Figure 3C, the decrease in GLUT-1 mRNA and protein levels resulted in a 37% decrease in the transport of DHA (P < 0.01), whereas the suppression of GLUT-3 resulted in only a 12% decrease in DHA transport. Thus, suppression of GLUT-1 accounted for a portion of the DHA transport in human chondrocytes, but GLUT-3 suppression had only a minimal effect on the uptake of DHA in chondrocytes.

Figure 3.

Suppression of glucose transporters (GLUTs) 1 and 3 expression by RNA interference. Primary human chondrocytes were transfected with a negative control small interfering RNA (siRNA) or siRNA specific for GLUT-1 (siGLUT-1) or GLUT-3 (siGLUT-3). GLUT-1 and GLUT-3 steady-state mRNA levels, protein levels, and dehydroascorbate (DHA) uptake were assessed in these cells 65–72 hours posttransfection. A, GLUT-1 and GLUT-3 mRNA levels, as determined by real-time reverse transcription–polymerase chain reaction, in siRNA-transfected cells. Values are the mean and SEM levels of mRNA for GLUT-1 or GLUT-3 expressed as a percentage of the negative control. The mRNA level in cells transfected with the negative control siRNA was set at 100%. ∗ = P < 0.01 versus the negative control (siControl). B. GLUT-1 and GLUT-3 protein expression, as determined by Western blotting, in siRNA-transfected cells. Values are the mean and SEM levels of GLUT-1 or GLUT-3 protein expressed as a percentage of the negative control. The protein level in cells transfected with the negative control siRNA was set at 100%. # = P < 0.05 versus the negative control (siControl). C. L-14C-DHA transport in siRNA-transfected cells. Values are the mean and SEM percentage of L-14C-DHA transported. The amount of DHA transported in cells transfected with the negative control siRNA was set at 100%.

Up-regulation of DHA, but not AA, transport at low oxygen tensions.

After 24 hours under 1% oxygen tension, DHA transport into primary human chondrocytes was increased a mean of 3.5-fold (Figure 4A). This was true in both the presence of glucose (3.8-fold; P < 0.001) and in the absence of glucose (3.2-fold; P < 0.001), as compared with corresponding samples at 21% oxygen tension. However, as expected for transport via the GLUTs, the transport of DHA was significantly inhibited by glucose under all oxygen tensions (P < 0.001). A dose-response was observed, with the greatest increase in L-14C-DHA uptake at 1% oxygen (3.5-fold; P < 0.001), an intermediate increase at 2.5% oxygen (1.6-fold in the absence of glucose; P < 0.001), and no increase at 5% oxygen tension over 24 hours, as compared with 21% oxygen tension.

Figure 4.

Up-regulation of dehydroascorbate (DHA) but not ascorbic acid (AA) transport under low oxygen tensions. Primary human chondrocytes were incubated under different oxygen tensions for 24 or 72 hours. L-14C-DHA or L-14C-AA transport was measured after 10 minutes at 37°C in an atmosphere of 5% CO2 and 21% O2. A, L-14C-DHA uptake was measured after 24 hours under the oxygen tensions indicated. Values are the mean and SEM percentage of L-14C-DHA transported. Solid bars represent uptake in samples containing 10 mMD-glucose; gray bars represent uptake in samples in the absence of D-glucose. ∗ = P < 0.001 versus the corresponding sample in the absence of glucose. # = P < 0.001 versus the corresponding sample at 21% oxygen tension. B, L-14C-DHA or L-14C-AA uptake was measured after 72 hours under the oxygen tensions indicated. Values are the mean and SEM picomoles of L-14C-DHA or L-14C-AA per 4.5 × 105 cells. P < 0.001 comparing AA transport at 5% oxygen tension versus DHA transport at 5% oxygen tension.

Exposure to 5% oxygen tension for a longer period of time (72 hours) led to a 2.7-fold increase in DHA transport (P < 0.001), as compared with transport at 21% oxygen tension (Figure 4B). On the other hand, there was no change in AA transport at either 1% or 5% oxygen tension (Figure 4B, showing data for 5% oxygen after 72 hours). Under physiologic conditions in the joint (in the presence of sodium and glucose at 5% oxygen tension), the amount of DHA transported into primary human chondrocytes was 4-fold greater than the amount of AA transported (P < 0.001).

Table 1 provides a summary of the effects of oxygen tension on GLUT-1 and GLUT-3 mRNA levels, as measured by real-time RT-PCR. There were minimal increases in GLUT-1 and GLUT-3 mRNA levels at 5% oxygen tension after 24 hours, which was consistent with the absence of a measurable change in the uptake of DHA under these conditions. After 24 hours at 1% oxygen tension and after 72 hours at 5% oxygen tension, there were substantial increases in both GLUT-1 and GLUT-3 mRNA levels, and a coincident increase in DHA uptake was also observed.

Table 1. Effects of oxygen tension on DHA uptake and GLUT-1 and GLUT-3 mRNA levels*
O2 tensionTime, hoursL-14C-DHA uptake, fold changeFold change in mRNA levels
  • *

    Levels of mRNA for glucose transporter 1 (GLUT-1) and GLUT-3 were determined by real-time reverse transcription–polymerase chain reaction. Data are expressed in comparison with the mRNA levels at 21% oxygen tension. All data were corrected for 18S ribosomal RNA levels in each sample. DHA = dehydroascorbate.

1%24Increased 3.8+8.6+4.3
5%24No change+1.6+1.8
5%72Increased 2.7+5.2+2.5

AA and DHA in patient samples.

HPLC analyses of synovial fluid from patients with knee OA showed that, on average, the synovial fluid contained 74 μM AA and 6 μM DHA. The paired plasma samples from these patients contained 49 μM AA and 5 μM DHA. Therefore, the vast majority of Asc in both plasma and synovial fluid was in the form of AA in patients with OA (P < 0.001). There was no difference in the proportion of synovial fluid DHA from knees with different levels of OA severity (data not shown). Overall, the synovial fluid contained significantly higher concentrations of AA than did the plasma (P < 0.001), corresponding to 8% of the total Asc in the form of DHA and 92% as AA (Table 2). By comparison, the previously reported percentages of AA and DHA in the blood of patients without arthritis, 86% AA and 14% DHA (34), were similar to those measured in OA patients. In contrast, reported values for synovial fluid and blood from RA patients (20% AA and 80% DHA in synovial fluid and 19% AA and 81% DHA in blood [34]) were roughly the inverse of those in OA patients.

Table 2. Percentages of AA and DHA in blood and synovial fluid from 3 groups of patients*
Patient groupBloodSynovial fluidPredicted physiologic transport pathways
% AA% DHA% AA% DHASVCT-2 (%)GLUTs (%)
  • *

    Data for the normal (nonarthritic samples) and rheumatoid arthritis (RA) groups are from ref. 36. AA = ascorbic acid; DHA = dehydroascorbate; SVCT-2 = sodium-dependent vitamin C transporter 2; GLUTs = glucose transporters; OA = osteoarthritis; ND = not determined.

OA (n = 23)9199287426
Normal (n = 20)8614NDND6139
RA (n = 13)19812080694


The mechanism by which DHA was transported into chondrocytes was glucose-regulated, suggesting that the transport was mediated by the GLUTs. Moreover, the sodium independence of DHA transport indicated that the transport of DHA was not occurring via the SVCTs. These conclusions were further supported by the complete inhibition of the facilitated diffusion of DHA by cytochalasin B.

Although many GLUTs transport glucose, only GLUTs 1, 3, and 4 have been shown to transport DHA in oocyte expression systems (6, 7, 12). However, the transport of DHA by GLUTs has been expected to be rather minimal in vivo. The plasma and synovial fluid, a dialysate of plasma, contain ∼10 mMD-glucose, and levels can be as high as 20 mMD-glucose in patients with diabetes (37). The IC50 values for inhibition of GLUT-3 and GLUT-1 by D-glucose fall within this physiologic range of glucose concentrations (4–10 mM) (7).

However, assessments of the contribution of DHA to the total intracellular Asc have not accounted for the unique hypoxic environment of cartilage. We found that DHA transport was enhanced after just 24 hours at 1% oxygen tension, similar to conditions in the deeper zones of articular cartilage. There was a time-dependent up-regulation of DHA transport under 2.5% and 5% oxygen tensions, replicating conditions found in the middle and superficial zones of cartilage, respectively. The increase in DHA transport at low oxygen tensions occurred in both the presence and absence of glucose. Additionally, GLUT-1 and GLUT-3 transcript levels were also up-regulated under the same oxygen conditions that enhanced DHA transport, further supporting a role for GLUTs in chondrocyte DHA transport. These results are consistent with the up-regulation in response to low oxygen tensions for GLUT-1 (38) in fibroblasts (39), muscle cells (40), transfected glioma cells (41), neurons (42), and astroglia (42) and for GLUT-3 in muscle cells (38), neurons (42), and astroglia (42). In addition, a 3.4-fold increase in glucose consumption in chick chondrocytes has been observed at 1% oxygen tension, as compared with consumption at 21% oxygen (20). These data are consistent with the interpretation that GLUTs mediate the DHA transport in chondrocytes under normoxic, as well as physiologically relevant hypoxic, conditions.

In contrast, the uptake of AA was not altered by low oxygen tension. Under physiologic conditions in the joint, including the presence of glucose and an average 5% oxygen tension, the rate of DHA transport was 4-fold greater than the rate of AA transport. These results demonstrate that the DHA transport mechanism is more robust than the AA transport mechanism in chondrocytes.

DHA accounted for only ∼10% of the total plasma Asc both in our OA patients and in the patients without arthritis described in the literature (36). In contrast, DHA comprised ∼80% of the total Asc in synovial fluid from patients with RA (36). We calculated the relative contributions of DHA and AA to total Asc in chondrocytes, based on the concordance of blood and synovial fluid levels, the percentages of DHA and AA in the synovial fluids from the different patient populations, and the fact that DHA transport was 4 times more robust than AA transport (Table 2). We estimate that DHA would constitute 26% of the total Asc transported in chondrocytes from OA patients, 39% of that transported in chondrocytes from patients without arthritis, and 94% of that transported in chondrocytes from RA patients. Therefore, in OA and nonarthritic chondrocytes, the majority of Asc is expected to be transported through SVCT-2, the AA transporter in chondrocytes (9), whereas in RA chondrocytes, the majority of Asc is expected to be transported as DHA, via the GLUTs. Our values for DHA transport into chondrocytes were likely an underestimate. Detailed HPLC analyses of the L-14C-DHA starting material revealed that, on average, the oxidation process resulted in conversion of all of the L-14C-AA to L-14C-DHA (mean ± SEM 41 ± 6%) or metabolites of DHA (59%). Taking this into account, DHA uptake may be as much as 8 times greater than AA uptake within joint tissues.

Surprisingly, we observed noncompetitive inhibition of DHA uptake by AA (L- and D-forms), although DHA did not inhibit AA transport (9). This is consistent with reports of the inhibitory effects of AA on 2-deoxyglucose transport in hippocampal cultures overexpressing GLUT-1 (43). Likewise, the inhibition was asymmetric, since glucose did not inhibit AA uptake (43). AA may inhibit DHA transport via electrostatic interactions with the transporter. Additionally, the sodium-containing forms of AA and DHA were more effective competitors of L-14C-DHA transport than were the sodium-free forms, perhaps due to an inhibition of intracellular DHA reduction. The extracellular sodium would drive the Na+/H+ exchanger, pumping H+, which is necessary for the reduction of DHA to AA, out of the chondrocytes. This would reduce the efficiency of DHA reduction to AA inside the chondrocytes and, thus, would build up a concentration gradient of DHA that would more quickly equilibrate with the extracellular DHA concentration. Therefore, less DHA would be expected to be transported into the chondrocytes in the presence of sodium, which was statistically apparent in competition experiments with Asc and was modestly apparent in the presence of sodium and the absence of glucose without competitors.

The suppression of GLUT-1 mRNA and protein levels and the resultant decrease in DHA uptake reveal that GLUT-1 mediates, at least in part, the passive transport of DHA in primary human chondrocytes. On the other hand, the suppression of GLUT-3 mRNA and protein levels had only a modest effect on the uptake of DHA, suggesting that this transporter plays a minor role in the transport of DHA in chondrocytes. The knockdown of GLUT-1 protein in our cells was incomplete and thus did not account for all of the DHA transport activity. The half-life of GLUT-1 protein in other cells is ∼6–14 hours (44, 45), whereas the half-life of GLUT-3 protein is 15 hours (45). We therefore expected the duration of our RNAi experiments to be sufficient to allow the turnover of all preexisting GLUTs. However, ∼60% of GLUT-1 and GLUT-3 proteins was still detectable in the cell lysates after 72 hours. Therefore, the half-life of preformed GLUT-1 and GLUT-3 proteins in chondrocytes may be longer than that reported for adipocytes (44) and muscle cells (45). The reduction in GLUT-1 protein concentrations in siRNA experiments was proportional to the reduction in DHA transport, suggesting that GLUT-1 plays a key role in DHA transport in chondrocytes.

Since we were unable to abolish all DHA uptake in these experiments, we cannot rule out the possibility that other transporters may also mediate DHA uptake in chondrocytes. Other GLUTs expressed by chondrocytes (GLUTs 6, 8, 9, 10, and 11) have not yet been assessed for their ability to transport DHA. Most likely, any other transporters that may mediate the uptake of DHA are GLUTs, since the DHA transport activity was suppressed by high concentrations of D-glucose and was completely inhibited by cytochalasin B, a GLUT inhibitor.

In summary, we provide the first evidence that human chondrocytes transport DHA via the GLUTs and that this transport mechanism in chondrocytes is modestly selective for L-DHA and is a physiologically relevant pathway for increasing intracellular Asc levels. This DHA transport mechanism is expected to be extremely important in RA patients, where the inflammatory environment increases the proportion of DHA in the blood and synovial fluid. By virtue of the up-regulation of the DHA pathway in the hypoxic milieu of cartilage, this mechanism of transport is also expected to provide a substantial amount of intracellular Asc to chondrocytes, which is necessary for their function and cartilage extracellular matrix production.


We thank Dr. Beverley Fermor and Benjamin Moeller for advice and assistance with the hypoxia experiments and Drs. Fermor, Farshid Guilak, and Mark Dewhirst for providing low-oxygen gas mixtures for the hypoxia experiments. We also thank Dr. Brice Weinberg for providing the Nucleofector device for the RNAi experiments and Dr. Carl Flannery for providing the type II collagen and aggrecan primers.