Endocrinology and Metabolism

Hexose transporter mRNAs for GLUT4, GLUT5, and GLUT12 predominate in human muscle

Charles A. Stuart, Deling Yin, Mary E. A. Howell, Rhesa J. Dykes, John J. Laffan, Arny A. Ferrando


In the past few years, 8 additional members of the facilitative hexose transporter family have been identified, giving a total of 14 members of the SLC2A family of membrane-bound hexose transporters. To determine which of the new hexose transporters were expressed in muscle, mRNA concentrations of 11 glucose transporters (GLUTs) were quantified and compared. RNA from muscle from 10 normal volunteers was subjected to RT-PCR. Primers were designed that amplified 78- to 241-base fragments, and cDNA standards were cloned for GLUT1, GLUT2, GLUT3, GLUT4, GLUT5, GLUT6, GLUT8, GLUT9, GLUT10, GLUT11, GLUT12, and GAPDH. Seven of these eleven hexose transporters were detectable in normal human muscle. The rank order was GLUT4, GLUT5, GLUT12, GLUT8, GLUT11, GLUT3, and GLUT1, with corresponding concentrations of 404 ± 49, 131 ± 14, 33 ± 4, 5.5 ± 0.5, 4.1 ± 0.4, 1.2 ± .0.1, and 0.9 ± 0.2 copies/ng RNA (means ± SE), respectively, for the 10 subjects. Concentrations of mRNA for GLUT4, GLUT5, and GLUT12 were much higher than those for the remainder of the GLUTs and together accounted for 98% of the total GLUT isoform mRNA. Immunoblots of muscle homogenates verified that the respective proteins for GLUT4, GLUT5, and GLUT12 were present in normal human muscle. Immunofluorescent studies demonstrated that GLUT4 and GLUT12 were predominantly expressed in type I oxidative fibers; however, GLUT5 was expressed predominantly in type II (white) fibers.

  • glucose transporter 1
  • glucose transporter 3

muscle takes up ∼65% of the glucose produced by the liver in the basal, low-insulin concentration, postabsorptive state and is responsible for 80% of the insulin-stimulated glucose disposal (8). Glucose cannot pass through cell membranes without specialized transmembrane protein structures that facilitate movement into the cell. Of several potential facilitative glucose transporters (GLUTs), the GLUT4 isoform is the predominant one in human skeletal muscle and likely provides the bulk of both basal glucose uptake and insulin- or contraction-related augmentation of glucose uptake into muscle. The Glut4 null mouse (females only) surprisingly augmented muscle glucose uptake in response to insulin (11, 31), suggesting that a backup system of insulin-stimulated translocation of GLUTs was in place. Since the time of the initial Glut4−/− studies, several more GLUTs have been identified, and these could be candidates for additional muscle GLUTs.

We have developed techniques that allow use of a quantitative, real-time, light cycler PCR instrument to quantify the amount of mRNA for GLUT1, GLUT2, GLUT3, GLUT4, GLUT5, GLUT6, GLUT8, GLUT9, GLUT10, GLUT11, and GLUT12 in human skeletal muscle. Of seven mRNA isoforms that were detected in muscle RNA, three (GLUT4, GLUT12, and GLUT5) were expressed at much higher levels than the others. Immunohistochemistry and Western blots confirm robust expression of proteins for the isoforms with the highest mRNA expression. Despite low expression of GLUT1 mRNA, GLUT1 protein was strongly represented in muscle sections by immunohistochemistry. Our results suggest that GLUT12 parallels the expression of GLUT4 in type I muscle fibers, but, surprisingly, GLUT5 is expressed predominantly in type II fibers.



Immunohistochemical studies were performed using a Leica TCS SP2 Laser Scanning confocal microscope (Wetzlar, Germany). An iCycler iQ System (Bio-Rad, Hercules, CA) was used for most of the PCR amplifications. RT and oligo(dT) primers were purchased from Applied Biosystems (Branchberg, NJ) as part of the GeneAmp RNA PCR Core Kit. Random primers were obtained from Integrated DNA Technologies (Coralville, IA). Affinity-purified antibodies against human GLUT1, GLUT4, GLUT5, GLUT8, and GLUT11 were purchased from Alpha Diagnostics (San Antonio, TX). Polyclonal antibodies raised in goat against human GLUT4 were purchased from Chemicon (Temecula, CA). Affinity-purified antibodies against human GLUT12 were purchased from Research Diagnostic (Flanders, NJ). Mouse monoclonal antibody specific for slow myosin heavy chain (MAB1628) was purchased from Chemicon. Sheep anti-von Willebrand factor antibody was purchased from Accurate Chemical and Scientific (Westbury, NY). AlexaFluor-647 donkey anti-mouse, AlexaFluor-555 donkey anti-rabbit, and AlexaFluor-488 donkey anti-goat were purchased from Molecular Probes (Eugene, OR). NADH was purchased from Sigma (St. Louis, MO). SuperSignal West Pico Chemiluminescence substrate was purchased from Pierce (Rockford, IL).


Ten volunteers underwent a percutaneous muscle biopsy. All subjects were nonobese (body mass index 22.7 ± 0.7, range 19.7–25.6 kg/m2), and none had parents or siblings with diabetes. This protocol and the consent documents were approved by the Institutional Review Boards at the University of Texas Medical Branch at Galveston and East Tennessee State University (ETSU). After informed consent was obtained, each subject fasted overnight, and after 2 h of quiet recumbency, a muscle biopsy was obtained using a 5-mm Bergstrom-Stille needle as previously described (3). The specimen was frozen in liquid nitrogen within 30 s and maintained in a freezer at −85°C until processed.

RNA isolation and cDNA production.

Total cellular RNA was isolated from skeletal muscle using the RNAzol B method (Tel-Test, Friendswood, TX) with 1 ml RNAzol B/50 mg muscle. The RT reaction was performed using oligo(dT) primers according to the GeneAmp Kit instructions, except as noted in results.

Generation of GLUT standards.

The standards for GLUT1, GLUT3, and GLUT4 included plasmids containing the full cDNA for each. A second set of standards for GLUT1, GLUT3, and GLUT4 and the standards for GLUT2, GLUT5, GLUT6, GLUT8, GLUT9, GLUT10, GLUT11, GLUT12, and GAPDH were generated using the specific primers and RT-PCR amplification of a reference muscle mRNA template, liver RNA (GLUT2, GLUT9, GLUT10), or peripheral blood leukocytes (GLUT6) and subsequently ligated into a plasmid using TOPO TA cloning vector (Invitrogen, Carlsbad, CA). Each plasmid containing full cDNA or a fragment was sequenced to confirm its structure. Standards were diluted to provide dose response curves across the amounts anticipated to be present in normal muscle. Melt curves were performed as part of each assay to verify the purity and accuracy of the standards.

iCycler technique.

The ETSU Molecular Biology Core Laboratory contains a Bio-Rad iCycler Thermal Cycler with the iQ Real-Time PCR Detection System. This PCR system uses a 96-well plate, fluorophore-containing reporters, and a real-time fluorophore excitation and detection system that can monitor 96 samples simultaneously. Table 1 displays the upstream and downstream primers that we have determined effectively amplify the DNA initially obtained with RT and RNA obtained from muscle, liver, or leukocytes. Buffers were those described in the instrument instructions. Magnesium concentration was optimized at 4.5 mM for all 11 isoforms. Analytic software provided with the instrument was used in most quantifications. The data were consistently monitored in the exponential phase of amplification. GLUT quantification was performed at least two separate times with triplicates of each sample in each experiment.

View this table:
Table 1.

iCycler glucose transporter primers

The coefficients of variance for the iCycler quantification of each of the GLUTs were calculated as the standard deviation of the mean divided by the mean of at least three separate determinations for a reference muscle specimen or for mRNA from peripheral blood leukocytes for GLUT6 or a reference liver specimen for GLUT2, GLUT9, and GLUT10. The coefficients of variation for reference samples, each determined from 3 to 12 separate assays, were 44, 42, 28, 29, 24, 32, 33, 2, 19, 2, and 9%, respectively, for GLUT1, GLUT2, GLUT3, GLUT4, GLUT5, GLUT6, GLUT8, GLUT9, GLUT10, GLUT11, and GLUT12.

Immunofluorescence studies in muscle.

Confocal microscopic assessment of specific fluorescent labeling of GLUT protein in normal human muscle sections was performed with minor modifications of methods previously described for GLUT3 (34). Before sectioning, muscle specimens (stored at −85°C) were warmed slowly to 0°C on ice. Sections were cut at 10-μm thickness using a CM3050 S cryostat (Vashaw Scientific, Norcross, GA). Slides were stored at −85°C until use. Slides were warmed to room temperature for 3 min, after which they were placed in acetone for 20 min at −20°C. Slides were subsequently allowed to air dry for 5 min. Each slide was rinsed in room-temperature PBS (0.1 M Na2HPO4, 0.03 M KH2PO4, 0.072 M NaCl, pH 7.3) four times for 10 min and placed in PBS containing 0.4% Triton X-100 and 0.5% bovine serum albumin (BSA) for another 20 min. The tissue was blocked with PBS containing 0.4% Triton X-100, 1% BSA, and 10% normal donkey serum for 2 h. Next, the specimen was incubated in primary antibodies (affinity-purified rabbit anti-hGLUT1 at 1:500, affinity-purified rabbit anti-hGLUT5 at 1:1,000, affinity-purified rabbit anti-hGLUT12 at 1:100, goat anti-hGLUT4 at 1:500, and mouse monoclonal anti-myosin heavy chain antibody at 1:1,000) diluted in PBS with 0.4% Triton X-100 and 1% BSA overnight at room temperature. The slides were rinsed in room-temperature PBS four times for 10 min and placed in PBS containing 0.4% Triton X-100 and 0.5% BSA for another 20 min. They were then incubated in Alexa fluor 488-conjugate donkey anti-goat IgG (1:200), Alexa fluor 555 donkey anti-rabbit (1:200), and Alexa fluor 647 donkey anti-mouse (1:200) in PBS with 0.4% Triton X-100 and 1% BSA for 2 h. The slides were rinsed again four times in PBS. Citifluor medium (Ted Pella, Redding, CA) was added to the specimen on the slide and the coverslip sealed with fingernail polish. Confocal microscopic images were obtained with a Leica confocal microscope. NADH oxidase tetrazolium staining was performed as described by Scarpelli et al. (30).

Immunoblot technique.

Immunoblotting was performed essentially as previously described (33). For most GLUTs to be evaluated, 20 μg of protein from muscle homogenate were separated on a 10% polyacrylamide gel using the Laemmli system (19), transferred to a nitrocellulose membrane, subjected to blocking with 2.5% nonfat dry milk in PBS, incubated with a validated dilution of one of the anti-GLUT antibodies above including 1.25% milk, and developed with enhanced chemiluminescence reagent and X-ray film. Band intensity of a digitized image was estimated using QuantityOne software from Bio-Rad.


Concentrations of GLUT mRNA in human skeletal muscle.

There was no detectable cDNA for GLUT2, GLUT6, GLUT9, or GLUT10 in the RT-treated RNA from the 10 normal muscle specimens. Used as a positive control in the assay, a single peripheral blood mononuclear cell RNA sample contained 42 ± 3 copies of GLUT6/ng RNA. A single sample of human liver was also used as a positive control and demonstrated 78 ± 13, 18 ± 2, and 45 ± 5 copies/ng RNA of GLUT2, GLUT9, and GLUT10, respectively.

Table 2 shows the individual results for measurement of GLUT1, GLUT3, GLUT4, GLUT5, GLUT8, GLUT11, GLUT12, and GAPDH mRNA in these specimens. The rank order of these seven members of the GLUT (SLC2A) family based on normal muscle concentrations of mRNA is GLUT4, GLUT5, GLUT12, GLUT8, GLUT11, GLUT3, and GLUT1.

View this table:
Table 2.

Hexose transporter mRNA quantification in muscle biopsies from 10 normal volunteers*

GLUT5 and GLUT12 protein expression in normal human skeletal muscle.

Immunofluorescence studies using a confocal microscope demonstrated robust signals for GLUT5 and GLUT12. GLUT8 staining was also positive, but less intense (data not shown). GLUT11 antibody did not detect protein in muscle sections by these methods. Figure 1 compares the pattern of GLUT4 and GLUT5 protein distribution in normal muscle. All four panels are from the same image of a biopsy specimen from a normal subject. Figure 1A shows the distribution of MAB1628 staining, which indicates the type I muscle fibers. Figure 1B displays the distribution of GLUT4 staining with the strongest labeling being in the type I fibers. Figure 1C displays GLUT5 staining, showing a pattern reciprocal to the blue of Fig. 1A and the green of Fig. 1B. Figure 1D is a composite of A–C, demonstrating that MAB1628 and GLUT4 co-localize, and GLUT5 staining is predominantly in different fibers. We performed additional studies using sequential muscle biopsy sections, employing immunofluorescence for GLUT5 labeling and light microscopic evaluation of NADH oxidase staining, confirming that GLUT5 expression was associated with fibers with low NADH oxidase activity (data not shown). These data show that GLUT5 is associated predominantly with white, fast-twitch, type IIb fibers, in contrast to the red fiber-associated pattern seen with GLUT3 (34) and GLUT4 (13, 24). Studies performed with individual antibodies against GLUT4, GLUT5, and GLUT12 and control mixtures of antibody with immunizing peptide demonstrated similar patterns of specific signal that were increased in roughly one-half of the fibers in each section (data not shown).

Fig. 1.

Distribution of GLUT5 protein in human skeletal muscle. Affinity-purified rabbit polyclonal antibodies directed at the carboxy-terminal 12 amino acids of GLUT4 and GLUT5 were used in immunohistochemical studies of vastus lateralis muscle from a normal 26-yr-old female volunteer. A–D are from a single confocal microscopic field showing transverse cut muscle fibers. A: photomicrograph labeled with mouse monoclonal antibody against slow myosin heavy chain (MHC) to indicate type I fibers. B: signal from goat anti-hGLUT4 antibody showing predominant type I fiber labeling. C: labeling with affinity-purified rabbit anti-hGLUT5 antibody. A composite of blue, green, and red images is shown in D. GLUT4 and GLUT5 proteins appear to be present in different fibers. These panels demonstrate that higher-intensity GLUT5 staining is associated with fast-twitch, white, type IIb fibers.

Figure 2 displays the pattern of expression of GLUT12 in normal muscle. Similar to the studies of GLUT5 expression described above, we compared the distribution of GLUT12 with that of GLUT4 and used a monoclonal antibody against slow myosin heavy chain as an indicator of type I muscle fibers. Figure 2, A–D, represents a confocal microscopic evaluation of a single section using three contrasting fluorochromes. In this case, GLUT4 second antibody was green (647 nm), GLUT12 second antibody was red (488 nm), and the myosin heavy chain-1 (MHC-1) antibody was blue (555 nm). As shown here, GLUT12 had a pattern nearly identical to that of GLUT4. All of the intense GLUT4/GLUT12 fibers were type I.

Fig. 2.

Co-localization of GLUT4 and GLUT12 in type I muscle fibers. A–D are from a single confocal microscopic, 3-color study using sections from a needle biopsy of vastus lateralis of a 24-yr-old normal male. Similar to Fig. 1, A displays staining of type I fibers with monoclonal antibody against the slow MHC, and B shows GLUT4 labeling. C: specific labeling with rabbit affinity-purified anti-hGLUT12 antibody. D: composite of all 3 colors. This study demonstrates that GLUT4 and GLUT12 are contained in the same type I slow-twitch fibers in human skeletal muscle.

GLUT1 protein expression in muscle.

Figure 3 shows GLUT1 to be expressed at its highest intensity in the capillary endothelial cells between the muscle fibers. GLUT1 staining shows the same pattern as von Willebrand factor, except that some GLUT1 is expressed inside the type I oxidative (red) fibers.

Fig. 3.

GLUT1 protein distribution in normal muscle. A: transverse sectioned normal muscle with labeling of GLUT1 protein using affinity-purified rabbit anti-human GLUT1 antibody. B: an image from the confocal microscope of the same section stained with sheep anti-human von Willebrand factor antibody as a marker of capillary endothelial cells. C: section stained with antibody against MHC-1, indicating which cells are type I oxidative fibers. D: composite of the colors from A–C. These studies show co-localization of GLUT1 signal with von Willebrand factor for the majority of the GLUT1 protein. Some GLUT1 protein appears within the type I fibers as well.

Immunoblots of human muscle showing GLUT5 and GLUT12 protein.

Immunoblots of normal muscle homogenate demonstrated strong signals for GLUT5 and GLUT12. Figure 4 shows bands specific for GLUT5, GLUT12, and GLUT4 proteins. In contrast to GLUT5 (apparent mobility 51 kDa), GLUT12 has an apparent mobility on SDS-PAGE of 60 kDa, consistent with its higher DNA sequence-predicted molecular weight (54,972 vs. 66,965).

Fig. 4.

Immunoblots of muscle homogenate demonstrating GLUT5 and GLUT12 protein. Left: specifically labeled band from normal human skeletal muscle homogenate probed with anti-hGLUT5 antibody, developed with SuperSignal chemiluminescence and X-ray film. The calculated apparent molecular mass was 51 kDa. Right: an immunoblot of normal human muscle incubated with anti-hGLUT4 and anti-hGLUT12 (1:500 and 1:100 dilution, respectively) simultaneously. Nos. at right are the calculated apparent molecular mass in kDa based on relative mobility.


Seven hexose transporters had mRNA that was detected in normal human skeletal muscle using RT and a real-time light cycler PCR. The concentrations of mRNA for GLUT4, GLUT5, and GLUT12 were much higher than those for GLUT8, GLUT11, GLUT3, and GLUT1. GLUT2, GLUT6, GLUT9, and GLUT10 mRNAs were not detected by this method. Immunoblots and immunohistochemistry studies confirmed robust expression for GLUT4, GLUT5, and GLUT12 (proteins). Although GLUT1 mRNA was expressed at low levels in muscle homogenate, strong GLUT1 protein expression was seen in capillary endothelial cells between the muscle fibers.

GLUT5 expression in muscle has been previously described (1), but the high amount of message relative to that of GLUT4 has not been noted. Glucose is by far the predominant carbohydrate fuel used by mammalian tissues, making it no surprise that 12 members of the 14-member GLUT family of membrane proteins transport glucose. Fructose, on the other hand, is found in blood at as much as 500-fold lower concentrations in humans (16). GLUT5 is exclusively a fructose transporter, with no ability to transport glucose (15). Six other isoforms either have been shown to transport fructose or have been deduced to do so because of their primary structure. GLUT2 is a glucose and fructose transporter that is most likely responsible for fructose uptake in the liver (1). GLUT5, GLUT7, GLUT9, and GLUT11 were deduced to have fructose transport activity because of their structural similarity to GLUT5 (14). These are all members of class II of the GLUT family, grouped together because of their high degree of amino acid identity (40–60%). GLUT7 (20, 23) and GLUT9 (9) subsequently have been confirmed to have fructose-inhibitable glucose uptake. GLUT8 and GLUT12 have also been shown to transport both glucose and fructose (27).

GLUT5 has previously been demonstrated in intestinal epithelium, kidney, fat, skeletal muscle, testes, and sperm (1). GLUT5 is not acutely regulated by insulin (25). Fructose feeding in rats increases the expression in intestinal epithelial cells of both sodium-dependent GLUT1 (SGLT1) and GLUT5 about fourfold, with a similar increase in uptake of fructose by the cells (5, 6). However, there was no fructose feeding-induced change in fat or skeletal muscle GLUT5 expression (7).

Kawasaki et al. (16) demonstrated that blood fructose concentrations were elevated 50% and urine content was more than threefold increased in patients with diabetes. These investigators found that improving glycemic control resulted in normalization of the serum fructose concentrations and a marked decrease in urine excretion (16).

Intestinal epithelial GLUT expression in patients with type 2 diabetes was evaluated by Dyer et al. (10). They found that uptake of d-glucose into brush-border membrane vesicles was increased 3.3-fold compared with controls. There was a fourfold increase in SGLT1 and GLUT5 protein in intestinal epithelium from diabetic subjects (10). Concentrations of mRNA for SGLT1, GLUT5, and GLUT2 were increased threefold in the epithelial cell RNA evaluated by Northern blotting. Whether elevated fructose concentrations or increased GLUT5 expression plays a role in insulin resistance or diabetic complications is unclear (21).

GLUT12 was first reported by Rogers et al. (29) in 2002. Northern blots of human tissues identified GLUT12 mRNA in heart, skeletal muscle, and prostate. Western blots demonstrated specific GLUT12 protein in skeletal muscle, fat, and small intestine (29). Because GLUT12 contains one or more dileucine motifs, it was deduced to be predominantly intracellular in the basal state (14). Our immunohistochemistry studies showed strong staining in skeletal muscle fibers. In our comparison of the distribution of GLUT12 with that of GLUT4, oxidative type I fibers expressed both at higher levels than type II fibers, but GLUT4 appeared to have more signal at the sarcolemma than did GLUT12 (Fig. 2). Regulation of GLUT12 by insulin is not known, other than it appears that insulin does not acutely cause translocation in MCF-7 cells in culture (29). MCF-7 cells showed GLUT12 protein localized to the peri-nuclear region of these cells.

In contrast to our results, Gaster et al. (12) recently reported no GLUT12 signal by immunohistochemistry in normal skeletal muscle. They did not examine whether specific GLUT12 mRNA was present. In our studies, both mRNA and protein were identified in normal human muscle specimens. We suspect that Gaster et al. may have evaluated GLUT12 presence in muscle using a bad lot of the antibody, a problem our laboratory also had with the first commercially available lot but not with later lots from Alpha Diagnostics.

Macheda et al. (22) evaluated the expression of GLUT12 in rat fetal tissues using immunohistochemistry. They found GLUT12 protein in the last few days of gestation in heart, skeletal muscle, brown fat, kidney, lung, and chondrocytes. Two additional studies from this group reported finding GLUT12 in breast cancer and normal breast tissue (28), and GLUT12 was present in prostate cancer but not normal prostate (4).

The concentration of GLUT4 mRNA determined in these studies was about twofold higher than we previously published using a “quantitative ribonuclease protection assay” (32). More recent studies in our laboratory before the current method used RT-PCR with competimers and demonstrated 520 ± 70 copies of GLUT4 mRNA/ng RNA in a group of six normal subjects (unpublished data). The concentration of GLUT3 in the current study was 5-fold lower and the concentration of GLUT1 was 40-fold lower than we reported using the ribonuclease protection assay (32), suggesting that the earlier method overestimated the concentrations of these GLUTs.

Protein expression and mRNA levels frequently do not closely correlate. Kern et al. (17) evaluated both mRNA and protein for GLUT1 and GLUT4 in red and white muscle. They found that GLUT1 mRNA and protein were about the same in red and white muscle, but GLUT4 protein was fivefold higher in red muscle, whereas the mRNA was only twofold higher (17). These data suggest that there are posttranscriptional control points for GLUT4 protein content in muscle. Our current data suggest that there are much lower levels of GLUT1 mRNA in normal muscle, but we have not directly compared the protein levels of GLUT1 and GLUT4 in these samples to correlate with the previously published data cited above (18, 24, 35). The immunohistochemical studies included in this report show that GLUT1 protein is present as a strong signal primarily in capillary endothelial cells between the muscle fibers. The intracellular GLUT1 signal is very likely due to t-tubule content of GLUT1, as has been reported for GLUT4 (26) and GLUT3 (34). The endothelial GLUT1 mRNA should be responsible for the majority of that measured in the real-time RT-PCR data presented here.

The data we report here suggest that the GLUT12 mRNA in skeletal muscle is expressed at high levels in normal subjects. The GLUT12 protein is expressed predominantly in red (type I) muscle fibers. The sugar transport specificity appears to be predominantly glucose (27), but the insulin responsiveness of GLUT12 in normal muscle is not yet known.

The results of these studies to establish the amounts of mRNA for GLUTs in normal human skeletal muscle yielded some surprises. GLUT12, thought to be a constitutive intracellular glucose/fructose transporter, had mRNA that was expressed at concentrations exceeded only by GLUT4 and GLUT5. GLUT5, a fructose transporter that does not transport glucose at all (2), was also expressed at mRNA concentrations much greater than GLUT1, GLUT3, GLUT8, and GLUT11. GLUT1, thought to contribute to the cell surface transporter pool (25), possessed mRNA concentrations among the lowest of the muscle GLUTs despite plentiful protein as judged by a strong signal by immunohistochemistry.

Immunohistochemical studies suggested that muscle fiber type affected the expression of the proteins in different ways. Like GLUT3 and GLUT4, which have been shown to be expressed at higher levels in type I (red) fibers, GLUT12 is also expressed predominantly in type I fibers. In contrast, GLUT5 appears to be expressed more in type II (white) fibers. The transport specificity, tissue and subcellular localization, and acute and chronic regulation of these novel transporters will provide important new information as this complex picture unfolds.


These studies were supported in part by National Institutes of Health (NIH) Grant GM-057295 and NIH National Center for Research Resources General Clinical Research Center Grant M01-RR-00073.


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


View Abstract