10.1152/ajpgi.00108. 2002.— Iron exacerbates various types of liver injury in which nuclear factor (NF)-κB-driven genes are implicated. This study tested a hypothesis that iron directly elicits the signaling required for activation of NF-κB and stimulation of tumor necrosis factor (TNF)-α gene expression in Kupffer cells. Addition of Fe2+ but not Fe3+ (∼5–50 μM) to cultured rat Kupffer cells increased TNF-α release and TNF-α promoter activity in a NF-κB-dependent manner. Cu+ but not Cu2+ stimulated TNF-α protein release and promoter activity but with less potency. Fe2+ caused a disappearance of the cytosolic inhibitor κBα, a concomitant increase in nuclear p65 protein, and increased DNA binding of p50/p50 and p65/p50 without affecting activator protein-1 binding. Addition of Fe2+ to the cells resulted in an increase in electron paramagnetic resonance-detectable ·OH peaking at 15 min, preceding activation of NF-κB but coinciding with activation of inhibitor κB kinase (IKK) but not c-Jun NH2-terminal kinase. In conclusion, Fe2+ serves as a direct agonist to activate IKK, NF-κB, and TNF-α promoter activity and to induce the release of TNF-α protein by cultured Kupffer cells in a redox status-dependent manner. We propose that this finding offers a molecular basis for iron-mediated accentuation of TNF-α-dependent liver injury.
- tumor necrosis factor-α
- free radical
- inhibitor κB kinase
- electron paramagnetic resonance
- nuclear factor-κB
iron potentiates various forms of liver injury (4, 19, 28, 41), and chelation of iron or decreasing iron content conversely ameliorates the injury (9, 22, 30, 32). The most accepted explanation for iron's effects is an iron-catalyzed Fenton pathway resulting in the generation of ·OH and consequent oxidative tissue injury. In particular, if the generation of reactive oxygen species (ROS) is already enhanced by underlying disease processes, a slight increase in hepatic iron content may suffice for robust production of ·OH and accentuation of oxidative damage, as exemplified in experimental alcoholic liver injury (41). This accentuation of liver injury is accompanied by enhanced nuclear factor (NF)-κB activation and expression of proinflammatory mediators (43). The latter events may merely reflect a consequence of enhanced hepatocellular necrosis or may also be considered as causal processes. In fact, at nontoxic concentrations, iron is known to promote macrophage functions, including antimicrobial effects (18) and tumor necrosis factor (TNF)-mediated cytotoxicity (46). More specifically, recent evidence suggests the role of iron in promoting cytokine expression (7, 14) and NF-κB activation (42) by hepatic macrophages.
Even though a catalytically active pool of iron is estimated to be extremely small in normal tissues, the pathological conditions may cause a transient release of iron from the intracellular compartments into the microenvironment. For instance, oxidative stress is known to release iron from ferritin through either reduction of Fe3+by O · or oxidative destruction of ferritin proteins (6, 39). Alternatively, · NO may cause mobilization of intracellular iron (11, 13, 21) by targeting iron-sulfur groups contained in several key enzymes (12, 17). Thus it is conceivable that in liver diseases in which mild iron accumulation, oxidative stress, and TNF-α induction commonly coexist, the transient release of catalytically active iron may serve to facilitate oxidative signaling for proinflammatory NF-κB activation.
The present study tested whether direct addition of ionic iron to cultured Kupffer cells leads to activation of NF-κB and induction of TNF-α expression. Our results demonstrate that Fe2+ but not Fe3+ at concentrations as low as 5 μM stimulates TNF-α release. It also induces TNF-α promoter activity in an NF-κB-dependent manner, and this effect is associated with time-dependent activation of inhibitor κB (IκB) kinase (IKK) and NF-κB without affecting activator protein (AP)-1 binding. Collectively, these results support a notion that iron can serve as a direct agonist to induce intracellular signaling for NF-κB activation in Kupffer cells in a redox status-dependent manner.
MATERIALS AND METHODS
Kupffer cell isolation and culture.
Kupffer cells were isolated from normal Wistar rats by in situ sequential digestion of the liver with pronase and collagenase and arabinogalactan gradient ultracentrifugation as previously described (22, 42). The adherence purification method was performed to raise the purity of Kupffer cells cultured onto a 100-mm dish to >96% as determined by phagocytosis of 1-μm latex beads. The viability was tested by the trypan blue exclusion test and always exceeded 97%. The cells were incubated with DMEM containing 5% fetal calf serum for 2 days, following the adherence method for in vitro experiments. For iron or copper treatment, the cells were washed twice with PBS, incubated in serum-free DMEM, and exposed to ferrous sulfate, ferric ammonium sulfate, cuprous chloride, or cupric sulfate (∼1–50 μM) for 4 h to assess their effects on the release of TNF-α and TNF-α promoter activity. For activation of IKK and NF-κB, as well as electron paramagnetic resonance (EPR) detection of radicals, the cells were incubated for shorter periods (from ∼5 min to 4 h) as specified below and in the figure legends. As a positive control, the cells were treated with lipopolysaccharide (LPS;Escherichia coli 055:B5, 500 ng/ml, Sigma, St. Louis, MO).
Nuclear protein extraction and EMSA.
To examine the effects of Fe2+ on DNA binding by NF-κB and AP-1, nuclear proteins were extracted from cultured Kupffer cells by using the method of Schreiber et al. (35). The extracts (5 μg) were incubated in a reaction mixture [20 mM HEPES, pH 7.6, 100 mM KCl, 0.2 mM EDTA, 2 mM dithiothreitol (DTT), 20% glycerol, and 200 μg/ml poly(dI-dC)] on ice with the double-strand κB consensus sequence (3), the κB site from TNF-α promoter (8), or the AP-1 binding site (2) labeled with 32P. After a 20-min incubation, the reaction mixture was resolved on a 6% nondenaturing polyacrylamide gel and the gel was dried for subsequent autoradiography. Densitometric analysis of the intensity of shifted bands was performed by using the Kodak Electrophoresis Documentation and Analysis System and imaging analysis software (Eastman Kodak, Rochester, NY). For the supershift assays, antibodies against p50 and p65 (Santa Cruz Biotechnology, Santa Cruz, CA) were added to the reaction mixture for an additional 30 min.
IκBα and p65 immunoblot analysis.
Cytoplasmic and nuclear extracts of iron-stimulated, cultured Kupffer cells were examined for IκBα and p65 levels by immunoblot analysis, respectively. Cytoplasmic or nuclear proteins (10 μg) were mixed with 2× sample buffer (100 mM Tris · HCl, pH 6.8, 4% SDS, 20% glycerol, and 10% β-mercaptoethanol) and separated by 10% PAGE under reducing conditions. The proteins were transferred to nitrocellulose filters (Bio-Rad, Hercules, CA) and treated overnight at 4°C with 5% BLOTTO [5% nonfat milk with (in mM) 50 Tris · HCl, pH 7.5, 50 NaCl, 1 EDTA, and 1 DTT]. The filters were then incubated with rabbit polyclonal anti-human p65 (Biomol, Plymouth Meeting, PA) or anti-human IκBα (Santa Cruz Biotechnology) at 1:1,000 dilution in TBST (10 mM Tris · HCl, pH 8.0, 150 mM NaCl, and 0.05% Tween 20) with 1% BSA at room temperature for 2 h, followed by three washes with TBS and 0.2% Tween 20. The filters were then incubated with horseradish peroxidase-conjugated goat anti-rabbit IgG (Sigma) at 1:2,000 dilution at room temperature for 2 h. The immobilized p65 and IκBα antibody complexes were detected by chemiluminescence by using an enhanced chemiluminescence kit (Amersham, Arlington Heights, IL).
EPR spectra of iron-treated Kupffer cells.
To determine time-dependent changes in the generation of free radicals by iron-treated Kupffer cells, the cells (107 cells/ml) were suspended in PBS containing 5–10 mM glucose with or without ferrous sulfate (50 μM). At different time points (0, 5, 10, 20, and 30 min), aliquots of the samples were withdrawn from the reaction mixtures, mixed with 50 mM α-(4-pyridyl-1-oxide)-N-t-butylnitrone (POBN) and 0.1% (vol/vol) DMSO, and immediately transferred to bottom-sealed Pasteur pipettes. The EPR spectra were recorded at room temperature in a Bruker ECS 106 spectrometer operating at 9.8 GHz. Instrument conditions were as follows: modulation frequency, 100 kHz; time constant, 1.3s; sweep scan, 18 G/min; modulation amplitude, 0.9 G; and microwave power, 20 mW. The spectra were compared with simulated ones obtained by using the published hyperfine splitting constants and the simulation program from Oklahoma Research Center.
IKK and JNK assays.
To assay the activity of IKK, Kupffer cells cultured in 100-mm dishes were treated with ferrous sulfate for ∼0–45 min or LPS (500 ng/ml) for 15 min, washed with PBS once, and lysed with a lysis buffer (in mM: 20 Tris · HCl, pH 7.5, 20 NaF, 20 β-glycerophosphate, 0.5 Na3VO4, 2.5 metabisulfite, 5 benzamidine, 1 EDTA, 0.5 EGTA, and 300 NaCl, with 10% glycerol and protease inhibitors and 1.5% Triton X-100). The lysates were immediately frozen in liquid nitrogen and stored at −80°C until assay. IKK activity was determined as previously described (29). Briefly, IKK was immunoprecipitated by IKKα antibodies and protein G-Sepharose. The assay was performed at 30°C for 1 h in buffer containing 20 mM Tris · HCl, pH 7.5, 20 mM MgCl2, 2 mM DTT, 20 μM ATP, 2 μg/30 μl glutathione-S-transferase (GST)-IκBα, and [γ-32P]ATP (0.5 μCi). The reaction was stopped by addition of Laemmli buffer and was resolved by 10% SDS-PAGE followed by a transfer onto a nitrocellulose membrane. Phosphate incorporated into GST-IκBα was visualized by exposing the membrane to a PhosphorImager. The c-Jun NH2-terminal kinase (JNK) assay was performed similarly, except that antibodies against JNK-1 (Santa Cruz Biotechnology) and protein G-Sepharose were used to immunoprecipitate JNK-1 and that GST-c-Jun (Santa Cruz Biotechnology) was used as a substrate. For both IKK and JNK, total protein levels were assessed by immunoblot analysis of the cell lysates.
Transfection and TNF-α promoter analysis.
To assess the effects of ionic iron and copper on TNF-α promoter activity, cultured Kupffer cells were transiently transfected with a TNF-α promoter-luciferase construct using Targefect F-2 (Targeting System, San Diego, CA). The construct was created by ligating a 1.4-kb mouse TNF-α promoter (a KpnI and HindIII fragment) (15) into the pGL3-Basic plasmid (Promega, Madison, WI). For determination of transfection efficiency, Renilla phRL-TK vector was used. For transfection, 3-day-cultured Kupffer cells in six-well plates were treated with 2 μg of the reporter gene, 0.02 μg Renilla phRL-TK, and 2 μl of F-2 reagent in 1 ml serum-free RPMI for 2 h. Then 1 ml of RPMI with 10% FCS was added to achieve the final FCS concentration of 5% for overnight incubation. On the next day, the medium was changed to new DMEM with 10% FCS and the cells were incubated for 24 h. During the last 14 h of the incubation, the medium was changed to serum-free RPMI with or without ferrous sulfate, ferric ammonium sulfate, cuprous chloride, or cupric sulfate (10 or 50 μM), and the cell lysate was collected for luciferase assay by using the Dual-Luciferase Reporter assay system (Promega). Four experiments were performed independently, and all results were normalized for transfection efficiency as determined by Renilla luciferase activity. To determine the dependence of iron's effects on NF-κB, the cells were also cotransfected with the IκBα super repressor plasmid, which expresses IκBα with S32A/S36A mutations (16), or the empty vector. These plasmids were kindly provided by Dr. Richard Rippe (University of North Carolina at Chapel Hill).
For RT-PCR analysis for TNF-α, 3 μg of total RNA was reverse transcribed into cDNA by a Moloney murine leukemia virus reverse transcriptase and oligo(dT)15 at 37°C for 60 min. Synthesized cDNA was amplified by denaturation at 94°C for 4 min, followed by multiple (25 for β-actin and 43 for TNF-α) cycles of denaturation (95°C, 30 s), annealing (58°C, 30 s), and extension (72°C, 60 s). Primers used for TNF-α were sense, 5′-ATGAGCACAGAAAGCATGATG and antisense, 5′-TACAGGCTTGTCACTCGAATT, and for β-actin they were sense, 5′-CACGGCATTGTAACCAACTG and antisense, 5′-AGGGCAACATAGCACAGCTT.
The effects of iron and copper on the release of TNF-α by cultured Kupffer cells were examined by analyzing the TNF-α protein in the media with a commercially available mouse TNF-α immunoassay kit (R&D Systems, Minneapolis, MN).
The numerical data were expressed as means ± SD, and comparison between treated and control groups was performed by Student'st-test.
Fe2+ but not Fe3+ stimulates release of TNF-α.
We first tested whether iron stimulates the release of TNF-α by cultured Kupffer cells. As shown in Fig.1, the addition of Fe2+ but not Fe3+ increased TNF-α release by twofold at 5 μM and eightfold at 10 and 50 μM during the 4-h treatment period. Interestingly, Cu+ but not Cu2+ also stimulated TNF-α release at 10 and 50 μM, but its effect seemed less potent compared with Fe2+. Thus these results demonstrate direct stimulation of Kupffer cell TNF-α release by iron and copper in a redox status-dependent manner. It should also be noted that no toxicity was observed in Kupffer cells exposed to ∼1–50 μM of iron or copper as assessed by lactate dehydrogenase release or Sytox green nucleic acid staining (Molecular Probes, Eugene, OR).
Iron stimulates TNF-α promoter activity.
We then tested whether Fe2+ stimulates the TNF-α promoter in cultured Kupffer cells. The promoter activity was indeed increased ∼2–3 fold with 10 and 50 μM Fe2+ (Fig.2 A). Cu+ (50 μM) also slightly increased TNF-α promoter activity, but Cu2+and Fe3+ did not (Fig. 2 A). Cotransfection of a super repressor IκBα vector completely abrogated the stimulation with 50 μM Fe2+, whereas cotransfection with a LacZ vector did not (Fig. 2 B). Stimulation of the promoter activity by 50 μM Fe2+ was about half of the maximal response achieved by LPS (500 ng/ml) in a serum-free condition (Fig.2 B). These results establish that Fe2+ activates TNF-α promoter in a NF-κB-dependent manner.
Fe2+ increases TNF-α mRNA levels.
We then examined whether TNF-α promoter activity induced by treatment with Fe2+ is associated with increased mRNA levels for this cytokine. As shown in RT-PCR data in Fig. 2 C, the iron treatment increased TNF-α message. Densitometric analysis and standardization with β-actin data showed 2.3- and 2.0-fold increases in TNF-α message by 10 and 50 μM Fe2+, respectively.
Fe2+ activates NF-κB in cultured Kupffer cells.
Next, we examined whether Fe2+ increases the binding of nuclear proteins to the κB site in cultured rat Kupffer cells. At 10 and 50 μM, there was increased DNA binding regardless of whether we used the consensus sequence (Fig. 3) or the κB site from the TNF-α promoter (data not shown). Figure3 A shows the representative EMSA results obtained with 50 μM Fe2+. Increased binding was noted from 30 min following the iron addition and lasted for 2–4 h. Densitometric analysis of three sets of EMSA results demonstrated 3.4 ± 1.0-fold and 2.1 ± 0.8-fold increases (n = 3,P < 0.05) in p65/p50 and p50/p50 binding at 30 min after the treatment with Fe2+, respectively. At 2 h, the intensities of both bands were only moderately increased by 67% for p65/p50 and 86% for p50/p50. AP-1 binding was analyzed by using the same nuclear extracts, but no changes were noted (Fig.3 A). Similar results were observed with 10 μM Fe2+ (data not shown). The supershift assay was performed to identify the proteins encompassing the two sizes of the DNA-protein complexes detected by NF-κB EMSA. This assay revealed that they were a p50/p50 homodimer and a p65/p50 heterodimer (Fig. 3 B). To confirm that iron-induced enhancement in NF-κB DNA binding was due to activation of the transcription factor, we performed Western blot analysis for cytosolic IκBα and nuclear p65. As shown in the representative blots in Fig. 4, the cytosolic level of IκBα was transiently reduced at 30 min–1 h while the nuclear p65 level increased from 30 min to ∼2–4 h after the iron addition. Loading of cytosolic or nuclear proteins was equal, as shown by the staining of the proteins on the filters (Fig.4). These results were confirmed in three independent experiments. These results support an interpretation that the iron treatment caused IκBα degradation, NF-κB activation, and nuclear translocation of the RelA protein, resulting in increased DNA binding by NF-κB, all commencing at 30 min. In addition, the lack of the AP-1 response suggests that the effect of Fe2+ on NF-κB is rather selective.
Direct addition of Fe2+ to nuclear proteins does not increase RelA binding.
Even though our Western blot results strongly supported that activation of NF-κB was most likely responsible for iron-induced enhancement in DNA binding of this transcription factor, it was still possible that iron directly increased the association of the nuclear NF-κB to the κB site in the nucleus. To test this possibility, Fe2+was added to the nuclear extracts prepared from the resting cultured Kupffer cells at 0.1, 1, 10, and 50 μM and the effects were analyzed by EMSA. The results demonstrated that the binding of p50/p50 but not of p65/p50 was apparently increased by the treatment (Fig.5), and densitometric analysis of three sets of data showed 25 ± 7, 46 ± 11, 97 ± 18, and 121 ± 21% increases in p50/p50 binding at 0.1, 1, 10, and 50 μM, respectively, and confirmed no increase in p65/p50 binding. These data suggested that this direct effect of iron on the nuclear extracts could not explain the increased binding of p65/p50 observed in the iron-treated cells.
Iron activates IKK.
To investigate the mechanisms of iron-mediated activation of NF-κB, we examined the effect of Fe2+ on IKK activity in cultured Kupffer cells at different time points. As shown in Fig.6, IKK activity, as assessed by phosphorylation of GST-IκBα, was increased at 15 min, whereas the total IKK level was unchanged. As a positive control, LPS-stimulated IKK activity is shown. The timing of IKK activation preceded the disappearance of cytosolic IκBα at 30 min after addition of iron (Fig. 4). In contrast, iron did not induce JNK activity (Fig. 6), and this result corroborated unchanged AP-1 binding by iron (Fig.3 A). Another stress-activated mitogen-activated protein kinase (MAPK), p38, was also assessed. The level of phosphorylated p38 was also unaffected by the iron treatment, suggesting that Fe2+ did not activate this MAPK (H. She, unpublished observations). The results on IKK and JNK were confirmed in at least three independent experiments. Thus these results demonstrate for the first time that Fe2+ activates IKK and support a notion that Fe2+ serves as an agonist to stimulate signal transduction, which is rather selective for activation of NF-κB.
Iron increases EPR-detectable radicals before NF-κB activation.
NF-κB is a redox-sensitive transcription factor, and ROS are implicated in its activation (1, 34, 36, 38). Thus we postulated that Fe2+ stimulates ROS production in Kupffer cells preceding activation of NF-κB. In fact, Fe2+ can react with oxygen in aqueous solution to produce Fe3+ and O ·, and this ROS may be responsible for the observed effect. Fe2+ may also catalyze the formation of ·OH from H2O2, which is generated from basal NADPH oxidase activity of cultured Kupffer cells. To address these possibilities, the cells were treated with Fe2+ for ∼0–30 min, ROS was trapped with POBN, and EPR spectra were analyzed. Kupffer cells without iron treatment exhibited an EPR spectrum constituted by an equal mixture of three spin adducts: methyl, hydroxyl, and O · (Fig.7 B). Addition of 50 μM Fe2+ to these cells resulted in an enhancement of the hydroxyl and methyl-POBN adduct signals (Fig. 7 A). The formation of these adducts must have relied on the production of hydroxyl radical from an iron-catalyzed Fenton reaction. The methyl adduct was likely produced at the attack of the ·OH on the methyl moiety of DMSO and the subsequent trap of this methyl radical by POBN. Both signals increased with incubation time (Fig. 7 C) up to a maximum at ∼15–20 min, regaining the initial values after 30 min. These increases in the steady-state concentration of these radicals indicate that the transient increases probably occurred as part of a response mechanism or signal transduction pathway on stimulus of exogenous iron. In particular, the fact that the peak of the radical generation at 15–20 min coincided with IKK activation and preceded activation of NF-κB at 30 min suggests the signaling role of the former in the latter events. In fact, this notion was developed in previous studies (37) that demonstrated activation of NF-κB by ·OH-gener- ating systems and a reversal of this effect by ·OH scavengers or metal chelators in Jurkat cells.
Biological and mechanistic implications.
The results presented by the current study demonstrate a direct stimulatory effect of Fe2+ on signal transduction for NF-κB activation in cultured Kupffer cells. The effect is seen at least at the level of IKK activation and extended to the most downstream level of TNF-α protein expression. These results suggest a possibility that iron may serve as an independent agonist for activation of NF-κB and induction of NF-κB-responsive genes in Kupffer cells in vivo. In fact, iron supplementation aggravates liver injury induced by alcohol (41) or hepatitis viral infection (4) in experimental animals. In a clinical setting, the increased hepatic iron content frequently accompanies many different types of liver disease, such as alcoholic liver disease (28), viral hepatitis (10), and nonalcoholic steatohepatitis (5, 25), and iron reduction modalities often ameliorate such liver damage (10). Acute iron loading to the isolated perfused rat liver results in early increases in Kupffer cell-dependent respiratory activity (40), and iron directly enhances interleukin-1 secretion by macrophages stimulated by interferon-γ and LPS (7). We have previously demonstrated that the treatment of cultured Kupffer cells with an iron chelator effectively suppressed activation of NF-κB (22). Therefore, the evidence presented by the current study offers the pivotal molecular basis for the link between iron and NF-κB activation suggested by the earlier studies. Indeed, in pathological livers, iron that is compartmentalized into protein-bound forms may be released transiently into the microenvironment due to oxidative (6, 39) or nitrosative (11, 13, 21) stress. This catalytically active pool of iron may directly activate NF-κB in Kupffer cells in vivo.
It is also known that iron overload inhibits functions of macrophages, including expression of proinflammatory cytokines (24, 26,45). These effects are likely due to cytotoxicity of the cells exposed to either high or chronic iron loading. Indeed, acute iron overload via phagocytosis of erythrocytes is shown to cause cell toxicity in cultured Kupffer cells (20). In our study, Kupffer cells exposed to Fe2+ iron at different concentrations up to 50 μM did not show signs of cytotoxicity, and under such conditions, the direct agonistic effect on NF-κB was evident.
Our results also demonstrate that the peak of ·OH generation coincides with activation of IKK in iron-treated Kupffer cells, suggesting that either this most potent radical or downstream molecules may be the potential effectors for IKK activation. It is presumed that this radical is generated by Fe2+ via a Fenton pathway catalyzing one electron reduction of H2O2. The role of metal-catalyzed generation of ·OH in NF-κB activation has previously been proposed (22, 37, 42), and our present data further support the notion. However, it remains to be determined whether and how ·OH indeed activates IKK. It may exert direct effects on IKK, such as oxidation of cysteine residues within the activation loop of IKKα and -β and a tighter conformation of the complex for phosphorylation of IκBα via disulfide bond formation (33). It may also mediate IKK activation via its effects on upstream kinases. For instance, thioredoxin can be oxidized by ·OH, and this may cause a release of apoptosis signal-regulating kinase 1 (ASK1), which is usually bound to thioredoxin as an inactive form (31). Released ASK1 can then be oligomerized for activation of p38, which may in turn lead to activation of NF-κB (23). However, since ·OH is extremely reactive, it is difficult to conceive that such selective oxidation of target molecules can be achieved with this radical without additional regulatory mechanisms. ·OH may also target other unknown inhibitors of IKK. Alternatively, ·OH may induce intracellular lipid peroxidation, and lipid peroxides or their end products, such as aldehydes, may regulate signaling for IKK activation. Indeed, 4-hydroxynonenal, one such aldehydic product, has been shown to activate JNK (27, 44) and p38 MAPKs (44). However, in our study, Fe2+ activated IKK independently of JNK (Fig. 6) or p38 (unpublished data) MAPK activities. Obviously, future studies are needed to better delineate the molecular steps connecting iron and IKK activation.
This work was supported by National Institutes of Health grants R37-AA-06603, P50-AA-11999 (USC-UCLA Research Center for Alcoholic Liver and Pancreatic Diseases), P30-DK-48522 (USC Research Center for Liver Diseases), R24-AA-12885 (Non-Parenchymal Liver Cell Core), and the Medical Research Service of the Department of Veterans Affairs. S. Xiong was supported by a Cooley's Anemia Foundation Postdoctoral Award.
Address for reprint requests and other correspondence: H. Tsukamoto, Keck School of Medicine, Univ. of Southern California, 1333 San Pablo St., MMR-402, Los Angeles, CA 90033-9141 (E-mail:).
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.
- Copyright © 2002 the American Physiological Society