Peroxisome proliferator-activated receptor (PPAR)γ ligands are known to have anti-inflammatory properties that include the inhibition of cytokine signaling, transcription factor activation, and inflammatory gene expression. We have recently observed that increased expression of heat shock protein (HSP)70 correlates with, but is not required for, the anti-inflammatory actions of PPARγ ligands on cytokine signaling. In this study, we provide evidence that the inhibitory actions of PPARγ ligands on cytokine signaling are associated with endoplasmic reticulum (ER) stress or unfolded protein response (UPR) activation in pancreatic β-cells. 15-Deoxy-Δ12,14-prostaglandin J2, at concentrations that inhibit cytokine signaling, stimulates phosphorylation of eukaryotic initiation factor-2α, and this event is followed by a rapid inhibition of protein translation. Under conditions of impaired translation, PPARγ ligands stimulate the expression of a number of ER stress-responsive genes, such as GADD 153, BiP, and HSP70. Importantly, ER stress activation in response to PPARγ ligands or known UPR activators results in the attenuation of IL-1 and IFN-γ signaling. These findings indicate that PPARγ ligands induce ER stress, that ER stress activation is associated with an attenuation of cytokine signaling in β-cells, and that the attenuation of responsiveness to extracellular stimuli appears to be a novel protective action of the UPR in cells undergoing ER stress.
- unfolded protein response
- peroxisome proliferator-activated receptor-γ
- endoplasmic reticulum
peroxisome proliferator-activated receptor (PPAR) represents a family of nuclear receptor ligand-activated transcription factors that participate in the maintenance of lipid and glucose homeostasis and control of cell growth and differentiation (6, 23, 30). Recently, ligands of the PPARγ isoform have been shown to possess potent anti-inflammatory activities (4, 5, 13, 35). PPARγ ligands such as the cyclopentenone 15-deoxy-Δ12,14-prostaglandin J2 (PGJ2), as well as the synthetic thiazolidinediones (troglitazone, rosiglitazone, and ciglitazone) inhibit signaling cascades that result in nuclear factor (NF)-κB-dependent gene activation in a number of cell types (35, 36). One mechanism of action appears to be the direct inhibition of NF-κB activation. Cyclopentenone prostaglandins have been shown to directly inhibit IκB kinase (IKK) (29, 31, 35). The role of PPARγ in mediating the anti-inflammatory actions of these ligands is less clear. PGJ2 has been shown to inhibit LPS-induced NF-κB activation in RAW 264.7 cells deficient in PPARγ, and transfection of PPARγ increases the sensitivity of these cells to the inhibitory actions of PGJ2 on inducible nitric oxide synthase (iNOS) reporter gene activation (36). PPARγ agonists also prevent IFN-γ-induced iNOS and cyclooxygenase (COX)-2 expression in PPARγ−/− macrophages (4). Although PPARγ expression increases the sensitivity of cells to PGJ2-mediated inhibition of inflammatory gene expression, the inhibitory actions of these ligands on inflammatory gene expression and transcription factor activation require concentrations 10- to 100-fold higher than the concentrations known to activate the nuclear receptor (21, 31, 36). These findings suggest that PPARγ is not required but may increase the sensitivity of the cells to the anti-inflammatory actions of these ligands.
The direct inhibition of IKK provides one mechanism by which these ligands inhibit inflammatory gene expression; however, it does not explain the inhibitory actions of these ligands on IL-1-stimulated c-Jun NH2-terminal kinase (JNK) (21) and activator protein (AP)-1 (37) activation, IFN-γ-stimulated signal transducers of activated transcription (STAT)1 activation (41), or inhibition of IFN-γ-dependent gene expression (4). We have shown that the inhibitory actions of PGJ2 on IL-1 and IFN-γ signal transduction in pancreatic β-cells correlate with a stress response, as evidenced by the increased expression of heat shock protein (HSP)70 (21, 41). However, antisense depletion of HSP70 does not modulate the ability of PGJ2 to inhibit IL-1-induced NF-κB or IFN-γ-induced STAT1 activation (41). Furthermore, inhibition of PPARγ activation, either by overexpression of a dominant negative PPARγ mutant or by treatment with the PPARγ antagonist GW-9622, does not suppress the stimulatory actions of PGJ2 on HSP70 expression (or stress response activation), nor does it attenuate the inhibitory actions of this ligand on IL-1 and IFN-γ signaling in β-cells (42).
Because PPARγ ligands activate a stress response (41) and members of this prostanoid family of ligands localize with the endoplasmic reticulum (ER) (39), we considered the possibility that these ligands may activate an ER stress pathway in β-cells. The unfolded protein response (UPR) is one such conserved cellular response that is activated by the accumulation of unfolded proteins in the ER. The UPR is regulated by at least three ER intermembrane transducers: pancreatic ER stress kinase (PERK), inositol requiring 1 (IRE), and activating transcription factor (ATF)6 (8, 14). Upon activation, PERK directly phosphorylates eukaryotic initiation factor (eIF)2α, thereby inhibiting new protein translation (34, 38). Activation of IRE, ATF6, and PERK results in the increased expression of genes containing ER or UPR response elements, and under conditions of ER stress, these transcripts are selectively translated (9, 19, 43). UPR activation promotes cellular survival and recovery from a stressful insult (8, 14, 27), confers resistance to cellular stress (20) and, if activated for prolonged periods, results in cellular death (8, 15, 34).
In this study, the effects of PPARγ ligands on ER stress induction and the effects of ER stress on cytokine signaling have been examined. We show that the PPARγ ligands PGJ2, troglitazone, MCC-555, and ciglitazone induce the expression of UPR-regulated genes HSP70, BiP, growth arrest and DNA damage-inducible gene (GADD) 153 (CHOP), and ATF4 by RINm5F cells and human islets. Consistent with UPR activation, treatment of RINm5F cells or rat islets with PGJ2 results in the inhibition of protein synthesis by a mechanism that is associated with increased eIF2α phosphorylation. The time- and concentration-dependent stimulatory effects of PPARγ ligands on UPR-regulated gene expression, eIF2α phosphorylation, and inhibition of protein translation correlate with the inhibitory actions of PPARγ ligands on IFN-γ- and IL-1-stimulated signaling (21, 41). In direct support for an inhibitory role of UPR activation on extracellular signaling, we show that IFN-γ fails to stimulate STAT1 activation in RINm5F cells treated with known UPR activators tunicamycin and nitric oxide. These findings support UPR activation as one potential mediator of the anti-inflammatory actions of PPARγ ligands on cytokine signaling, and suggest that the inhibition of cellular responses to external stimuli may be a novel protective response associated with ER stress activation in pancreatic β-cells.
Materials and animals.
Sprague-Dawley rats were purchased from Harlan (Indianapolis, IN). RINm5F cells were obtained from the Washington University Tissue Culture Support Center (St. Louis, MO). Human islets were obtained from Washington University School of Medicine Islet Isolation Laboratory through the Islet Cell Resource Consortium 5 U42 RR 16597 (St. Louis, MO) and the Diabetes Research Institute at the University of Miami (Miami, FL). RPMI 1640 containing 1× l-glutamine, CMRL-1066, and Opti-MEM reduced serum tissue culture medium, methionine-deficient MEM, l-glutamine, penicillin, streptomycin, and rat recombinant IFN-γ were from GIBCO-BRL (Grand Island, NY). Fetal calf serum was obtained from Hyclone Laboratories (Logan, UT). Lipofectamine 2000 and Plus Reagent were obtained from Invitrogen (La Jolla, CA). Human recombinant IL-1 was obtained from Cistron Biotechnology (Pine Brook, NJ). PGJ2, rosiglitazone, ciglitazone, troglitazone, and MCC-555 were from Cayman Chemicals (Ann Arbor, MI). Horseradish peroxidase-conjugated donkey anti-rabbit and donkey anti-mouse IgG were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA). Rabbit anti-phospho-STAT1 was from Upstate Biotechnology (Lake Placid, NY). Rabbit anti-STAT1, rabbit anti-CHOP, and rabbit anti-ATF4 were from Santa Cruz Biotechnology, (Santa Cruz, CA). Rabbit anti-phospho-eIF2α antibody was from Cell Signaling (Beverly, MA). All other reagents were from commercially available sources.
Islets were isolated from 250- to 300-g male Sprague-Dawley rats by collagenase digestion as previously described (16, 22). Islets were cultured overnight in complete CMRL-1066 (containing 2 mM l-glutamine, 10% heat-inactivated fetal calf serum, 100 U/ml penicillin, and 100 μg/ml streptomycin) at 37°C under an atmosphere of 95% air-5% CO2 before experimentation.
Protein synthesis in RINm5F cells (2 × 105) or rat islets (150) treated with PPARγ ligands and ER stress inducers was examined by the incorporation of [35S]methionine (33 μCi/ml) during a 30-min incubation in methionine-deficient DMEM (9 parts DMEM lacking methionine to 1 part DMEM). Experiments were terminated by washing with ice-cold PBS, the cells were lysed, and equal amounts of protein (20–30 μg) from the various treatment groups were separated by SDS gel electrophoresis. The incorporation of [35S]methionine into rat islet and RINm5F cell protein was visualized by autoradiography (20).
β-Galactosidase reporter assays.
RINm5F cells (2 × 105/400 μl complete CMRL) were transiently transfected with 0.5 μg of the pCMV-SPORT-β-galactosidase plasmid with Lipofectamine 2000 and Plus Reagent according to the manufacturer's instructions and as described previously (42). After a 24-h incubation at 37°C, the experiments were initiated by the addition of PGJ2 for the indicated time. The cells were harvested and lysed in Reporter Lysis Buffer (Promega, Madison, WI), and β-galactosidase activity assays were performed as previously described (1).
Electrophoresis and Western blotting.
Cellular proteins were separated by SDS polyacrylamide gel electrophoresis and transferred to Highbond ECL nitrocellulose membranes (Amersham Pharmacia Biotech), as previously described (12). Antibody dilutions of 1:1,000 were used for the following antibodies: rabbit anti-phospho-eIF2α, rabbit anti-phospho-STAT1, rabbit anti-STAT1, rabbit anti-ERK, rabbit anti-CHOP, rabbit anti-ATF4. Horseradish peroxidase-conjugated donkey anti-rabbit and donkey anti-mouse secondary antibodies were used at 1:7,000 and 1:5,000 dilutions, respectively. Antigens were detected by enhanced chemiluminescence according to the manufacturer's specifications (Amersham Pharmacia Biotech).
RT-PCR and real-time PCR.
Total RNA was isolated from cells with the RNeasy kit (Qiagen, Valencia, CA). First-strand cDNA synthesis was performed using oligo(dT) and reverse transcriptase. Standard PCR was performed as previously described (41), or real-time PCR was performed using SYBR Green reagent (Qiagen) and the Research DNA Engine Opticon System with continuous fluorescence detection (MJ Research) per manufacturers' protocols. The fold change in expression of UPR target genes was determined by comparing mRNA accumulation in response to treatment with PPARγ ligands at 1 μM and 30–50 μM. Target mRNA accumulation at each concentration of PPARγ ligand was normalized to the mRNA accumulation of the housekeeping internal control GAPDH. Primer sequences for GAPDH and HSP70 were described previously (11, 41). Primer sequences for UPR genes follow: CHOP forward primer 5′-cag agg tca caa gca cct-3′, reverse primer 5′-tcc ctg gtc agg cgc tc-3′; BiP forward primer 5′-agt aag ttc act gtg gtg gc-3′, reverse primer 5′-gcg ctt ggc gtc gaa gac-3′.
PPARγ ligands inhibit protein synthesis.
Recently, we have shown that the inhibitory actions of PGJ2 on cytokine signaling in rat islets and RINm5F cells do not require the activation of PPARγ (42). In characterizing PPARγ-dependent reporter activity, we observed a reduction in β-galactosidase activity (used to determine transfection efficiency for these studies) in RINm5F cells treated with PGJ2 (Fig. 1A). The inhibition of β-galactosidase reporter activity was associated with a reduction in β-galactosidase expression (Fig. 1B), suggesting that PGJ2, at concentrations that inhibit cytokine signaling (15–30 μM) also inhibits protein translation. To examine this possibility directly, the effects of PGJ2 treatment on [35S]methionine incorporation into RINm5F cell or rat islet protein were examined by SDS gel electrophoresis. Treatment with PGJ2 results in the inhibition of [35S]methionine incorporation into rat islet or RINm5F cell protein that is first detectable and maximal following a 2- to 4-h incubation (Fig 1C). As a positive control, the inhibitory actions of the nitric oxide donor sodium (Z)-1(N,N-diethylamino)diazen-1-ium-1,2-diolate (DEA-NO) on protein synthesis in rat islets and RINm5F cells are also shown (17, 32). These findings suggest that the inhibitory actions of PGJ2 on cytokine signaling in RINm5F cells and rat islets correlate with an inhibition of protein synthesis.
PGJ2 stimulates eIF2α phosphorylation.
Under conditions of ER stress, protein synthesis is attenuated as a result of PERK activation and subsequent phosphorylation of eIF2α. Phosphorylation of eIF2α impairs guanine nucleotide exchange, resulting in an inhibition of translation (34, 38). Because PGJ2 inhibits protein synthesis, and PPARγ ligands have been shown to localize to the ER (26, 39), we hypothesized that PPARγ ligands may inhibit cytokine signaling by activating an ER stress response in β-cells. To determine whether PGJ2 is an activator of ER stress, the effects of this ligand on eIF2α phosphorylation were examined by Western blot analysis. In a concentration-dependent manner, PGJ2 induces the phosphorylation of eIF2α in RINm5F cells that is first detectable at 15–30 μM (Fig 2A). At a concentration of 30 μM, PGJ2 stimulates a three- to fourfold increase in eIF2α phosphorylation (as determined by densitometry). The stimulatory actions of PGJ2 on eIF2α phosphorylation are transient, first appearing after a 30-min incubation and persisting for ∼2–3 h (Fig 2B). The nitric oxide donor compound DEA-NO also stimulates eIF2α phosphorylation at concentrations that inhibit protein synthesis in rat islets and RINm5F cells (Fig 2C).
PPARγ ligands stimulate the expression of UPR-responsive genes.
The phosphorylation of eIF2α and the subsequent inhibition of translation suggest that PGJ2 or PPARγ ligands may induce a conserved process that selectively controls the expression of genes similar to that observed under conditions of UPR activation. To test this hypothesis, RINm5F cells were treated for 3 h with concentrations of PGJ2 that fail to inhibit cytokine signaling or protein synthesis (1 μM) or concentrations that inhibit both protein synthesis and cytokine signaling (30 μM). Total RNA was isolated from the cells, and the expression of HSP70, BiP, and CHOP was examined by real-time PCR. As shown in Fig. 3A, PGJ2 stimulates a 10-fold increase in HSP70, a 2.5-fold increase in BiP, and a 4-fold increase in CHOP mRNA accumulation. In a similar manner, PGJ2 stimulates a 10-fold increase in CHOP and a 40-fold increase in HSP70 mRNA accumulation in human islets treated for 3 h with this PPARγ ligand (Fig. 3B). To confirm that this response is not selective for PGJ2, the effects of PPARγ ligands troglitazone, rosiglitazone, ciglitazone, and MCC-555 on CHOP and HSP70 mRNA accumulation in RINm5F cells were examined. Similar to PGJ2, incubation of RINm5F cells for 3 h with these ligands at a concentration of 1 μM fails to stimulate either HSP70 or CHOP mRNA accumulation; however, at anti-inflammatory concentrations of 30–50 μM, each ligand stimulates the accumulation of CHOP, and three of the four ligands (troglitazone being the exception) stimulate HSP70 mRNA accumulation (Fig. 3C). In human islets, 3 h of incubation with ciglitazone, rosiglitazone, or troglitazone at 50 μM results in a three- to fivefold increase in the accumulation of HSP70 and CHOP mRNA (Fig. 3D). These findings suggest that the PPARγ ligand PGJ2 and members of the thiazolidinedione class of PPARγ agonists all stimulate the expression of known UPR responsive genes HSP70, CHOP, and BiP.
PPARγ ligands induce synthesis of UPR-responsive proteins.
Under conditions of reduced protein synthesis due to UPR activation, a number of UPR-responsive genes have been shown to evade this translational blockage (15). Two such genes are ATF4 and CHOP (8, 14). To confirm that these genes are translated under conditions of ER stress, RINm5F cells were treated with increasing concentrations of PGJ2 (3 or 30 μM) or the UPR activator tunicamycin for 3 and 6 h. Treatment of RINm5F cells for 3 h with 30 μM PGJ2 results in the increased expression of both CHOP and ATF4, and the expression of both proteins accumulates to high levels following 6 h of incubation (Fig. 4A). Importantly, this is a concentration of ligand that inhibits cytokine signaling (21, 41) and stimulates eIF2α phosphorylation (Fig 2). At 10-fold lower concentrations (3 μM), which do not inhibit cytokine signaling (21), PGJ2 fails to stimulate CHOP or ATF4 protein accumulation in RINm5F cells (Fig. 4A). Similar to CHOP and ATF4, PGJ2 also stimulates the synthesis of HSP70 (Fig. 4B) under conditions in which this ligand impairs protein translation (Fig. 2). As a positive control for UPR activation, the stimulatory actions of tunicamycin on CHOP and ATF4 expression are shown (Fig. 4A). These findings indicate that UPR-responsive genes known to escape translational inhibition (CHOP and ATF4) and HSP70 are expressed in response to PGJ2 at concentrations that attenuate cytokine signaling.
UPR activation inhibits cytokine signaling in β-cells.
Because PPARγ ligands have been shown to attenuate cytokine-stimulated signaling events in β-cells, and these ligands appear to activate a UPR-like ER stress response, we sought to determine whether activation of the UPR alone is sufficient to inhibit cytokine signaling. For these studies, the UPR was activated by the PPARγ ligand PGJ2, the classical ER stress activator tunicamycin (which interferes with glycoprotein synthesis) (18), or a second UPR activator, nitric oxide (17, 28). Treatment of rat islets or RINm5F cells with IFN-γ for 30 min results in STAT1 phosphorylation, as determined by Western blot analysis using an antibody specific for phosphorylated STAT1 (Fig. 5, A and B). Importantly, IFN-γ fails to stimulate STAT1 phosphorylation in rat islets or RINm5F cells pretreated for 6 h with PGJ2 or tunicamycin prior to cytokine stimulation (Fig. 5, A and B). Similar to the effects of tunicamycin, IFN-γ-stimulated STAT1 phosphorylation is attenuated in RINm5F cells treated with the nitric oxide donor DEA-NO (Fig. 5C). In these experiments, RINm5F cells were treated with DEA-NO for 10–60 min, the cells were washed and allowed to recover for 4 h, and then the effects of a 30-min incubation with IFN-γ on STAT1 phosphorylation were examined. In addition to the inhibition of IFN-γ signaling, DEA-NO also inhibits IL-1-induced JNK phosphorylation and IκB degradation in RINm5F cells (Fig. 5D). These findings suggest that one consequence of ER stress activation is the attenuation of the cellular response to extracellular stimuli, as indicated by the inhibition of IFN-γ- and IL-1-activated signaling pathways following UPR activation.
Proinflammatory cytokines are believed to participate in the loss of β-cell function and mass during the development of autoimmune diabetes and in the rejection of transplanted islets (10, 24). Cytokines such as IL-1 and IFN-γ mediate β-cell damage, in part, by stimulating the expression of iNOS and production of nitric oxide by β-cells (10). Nitric oxide impairs β-cell function by inhibiting mitochondrial oxidative metabolism (10) and by inducing DNA damage in β-cells (7). Ligands of PPARγ have been shown to attenuate diabetes development in the NOD mouse (2), impair cytokine signaling, and inhibit IL-1− and IL-1+ IFN-γ-induced iNOS expression in rodent islets (21, 41). In addition, these ligands have been shown to prevent iNOS, COX-2, IL-1, and IL-6 expression in lipopolysaccharide-stimulated macrophages (4, 5, 13, 36). The mechanisms by which these ligands inhibit inflammatory gene expression have yet to be fully elucidated. It is clear that cyclopentenone prostaglandins such as PGJ2 directly inhibit IKK activity and the subsequent activation of NF-κB (29, 35); however, this mode of action does not explain the ability of these ligands to inhibit IL-1-induced JNK (21) or AP-1 activation (4, 37) or IFN-γ-stimulated STAT1 activation (41). Furthermore, the anti-inflammatory actions of these ligands do not appear to require PPARγ (42).
Insights into the mechanism by which these ligands impair cytokine signaling in β-cells were derived from our observations that the inhibition of signaling correlated with increased expression of HSP70 in β-cells (21, 41) and that, when β-cells express HSP70 following heat shock, IL-1-induced NF-κB activation and iNOS expression are attenuated (33). Although these findings correlate increased expression of HSP70 with the inhibition of cytokine action, antisense depletion of HSP70 does not modify the ability of PGJ2 to inhibit cytokine signaling in RINm5F cells (41). These findings suggest that PGJ2 stimulates a stress response in β-cells and that this stress response is a likely mediator of the inhibitory actions of PGJ2 on cytokine signaling.
Because PPARγ ligands are known to localize to the ER (26, 39), and HSP70 is expressed during ER stress (3, 40), we hypothesized that PPARγ ligands may active an ER stress response in β-cells and that activation of this stress response may protect β-cells from the damaging actions of cytokines by impairing cytokine signaling. In support of this hypothesis, we show that PGJ2 stimulates eIF2α phosphorylation and that this phosphorylation event correlates with impairment of protein synthesis. Although protein synthesis is attenuated, a number of UPR-responsive genes selectively escape this translational blockade. We show that PGJ2 stimulates BiP, CHOP, and HSP70 mRNA accumulation in RINm5F cells and human islets and the translation of CHOP, ATF4, and HSP70 in RINm5F cells. Importantly, the expression of each of these UPR-responsive genes occurs only at concentrations of PGJ2 that have been shown to impair the ability of IL-1 and IFN-γ to activate downstream targets in β-cells (21, 41). Furthermore, the effects do not appear to be specific for PGJ2 as members of the thiazolidinedione class of insulin sensitizers, including troglitazone, rosiglitazone, ciglitazone, and MCC-555, stimulate CHOP and HSP70 mRNA accumulation in RINm5F cells and human islets. These findings indicate that ligands of the PPARγ nuclear receptor induce ER stress in β-cells, and this activation occurs at concentrations of the ligand that inhibit cytokine signaling and inflammatory gene expression.
The mechanism by which these ligands activate the UPR has yet to be determined. PPARγ activation does not appear to be required for UPR activation, as PGJ2 stimulates HSP70 expression and inhibits cytokine signaling in RINm5F cells expressing dominant negative mutants of PPARγ or treated with the PPARγ antagonist GW-9622 (42). PGJ2 has been shown to localize to the ER (39), and this localization is associated with increased expression of BiP and protein disulfide isomerase (25, 26). In addition, we have shown that PGJ2 stimulates tissue transglutaminase activity in RINm5F cells, and the resulting cross-linking of proteins may be the mechanism by which these ligands stimulate ER stress in β-cells (Weber SM and Corbett JA, unpublished observation).
The UPR is believed to be a protective response that is activated in cells under conditions of stress, and activation of this response allows the cell to recover from the damaging insult. Although UPR activation in response to unfolded proteins has been extensively characterized, additional UPR activators include hypoxia/ischemia (14), nutritional deprivation (15), and free radicals such as nitric oxide (27). Recently, Lu et. al. (20) used a genetic approach, in which eIF2α phosphorylation is controlled independently of cell stress, to show that increased eIF2α phosphorylation protects cells from glutamate toxicity and nitrosative stress, suggesting that the UPR may have a functional role in preconditioning. Preconditioning is a protective physiological response to low levels of stress that confers resistance to the damaging actions of severe stress. Consistent with the findings of Lu et al., we now provide evidence that UPR activation functions to attenuate cellular responses to cytokines. We show that signaling cascades activated by IL-1 and IFN-γ are attenuated in RINm5F cells and rat islets in which the UPR has been activated either by treatment with known UPR activators such as tunicamycin and nitric oxide or in response to PGJ2. These findings suggest that, in addition to the attenuation of oxidative and nitrosative stress-mediated cell damage (20), UPR activation also functions to protect cells by preventing the deleterious actions of cytokines on cellular function by impairing the ability of these cytokines to activate downstream signaling pathways. Elucidation of the mechanisms by which UPR activation attenuates cytokine signaling may identify a number of novel therapeutic targets that could be used to protect islets from graft rejection or recurrence of autoimmunity directed against β-cells in the transplantation setting.
This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases Grant DK-52194 (J. A. Corbett). K. T. Chambers was supported by an American Heart Association predoctoral fellowship.
We thank Colleen Bratcher for expert technical assistance.
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 © 2004 by American Physiological Society