Mucosal eosinophils increase in a number of gastrointestinal diseases that are often associated with altered epithelial barrier function, including food allergic enteropathies and inflammatory bowel diseases. Although eosinophils are known to secrete biologically active mediators including granule proteins, their role in gastrointestinal diseases is uncertain. The aim of this study was to determine the impact of eosinophils on intestinal barrier function. Epithelial barrier function was determined in a coculture of eosinophils and T84 epithelial cells and in a murine model of T helper (Th) type 2-mediated colitis. Coculture conditions resulted in decreased transepithelial resistance (TER) and increased transepithelial flux. Cell-free coculture supernatants contained a ≥5-kDa soluble factor that also diminished TER; these supernatants contained the eosinophil-granule proteins major basic protein (MBP) and eosinophil-derived neurotoxin (EDN). T84 barrier function decreased significantly when basolateral surfaces were exposed to native human MBP but not EDN. Additional studies identified downregulation of the tight junctional molecule occludin as at least one mechanism for MBP action. MBP-null mice were protected from inflammation associated with oxazolone colitis compared with wild-type mice. In conclusion, MBP decreases epithelial barrier function and in this manner contributes to the pathogenesis of inflammatory bowel diseases.
- tight junction
- eosinophilic gastroenteritis
- food allergy
surfaces exposed to the external environment, such as the intestinal mucosa, possess a number of immunologically potent cells that lie adjacent to the epithelium, including lymphocytes, dendritic cells, mast cells and eosinophils. A significant body of work has identified interactions between some of these cells and the epithelium; for instance, mucosal mast cells contribute to epithelial disruption in states of allergic inflammation and stress colitis in rodent models (57). Additionally, morphological evidence demonstrates the close effacement of eosinophils with epithelial cells in tissues affected by inflammatory bowel disease (IBD) and allergic gastroenteropathy (56), two diseases notably associated with increased intestinal permeability. Although microscopic sections have demonstrated that eosinophils exist in an activated functional state (18), little is known about which molecules participate in the inflammatory responses and their exact role in the geometrically confined crypt microenvironment.
A number of products, including eosinophil secondary granule proteins (ESGPs), oxygen metabolites, lipid mediators, and proteases, may contribute to epithelial dysfunction. As part of their granule-bound constituents, eosinophils store a number of unique proteins, including major basic protein (MBP), eosinophil-derived neurotoxin (EDN), eosinophil peroxidase (EPO), and eosinophil cationic protein. Deposition of ESGPs in the tissues affected by IBD, celiac sprue and eosinophilic gastroenteritis, indicate the likelihood that activated eosinophils release these mediators at sites of ongoing inflammation (14, 35, 41, 45, 59). In this setting, ESGPs likely mediate inflammation as evidenced by many previous studies demonstrating their capacity to activate a number of cells including basophils, fibroblasts, myocytes, mast cells, platelets, neutrophils, and epithelia (25, 26, 30, 34, 37, 48, 51, 54, 55). In addition, ESGPs contribute to alterations in barrier function in the urinary bladder through undefined mechanisms (38).
In this report, a coculture model is initially defined that permits a mechanistic study of the interactions of anatomically positioned eosinophils and their granule proteins with intestinal epithelial cells. These results suggest that physiologically positioned eosinophils induce a change in barrier function related to the impact of an ESGP, MBP, on the tight junctional protein occludin. The biological relevance of these findings was subsequently determined using mice deficient in MBP-1. These studies demonstrate that relative to wild-type mice, MBP-1-deficient mice fail to develop pathologies associated with T helper type 2 (Th2)-mediated colitis, suggesting a direct link between eosinophils and intestinal dysfunction.
MATERIALS AND METHODS
AML14.3D10 eosinophil myelocytes (a kind gift from Cassandra Paul, Wright State University, Dayton, OH) were used in the majority of coculture studies. AML14.3D10 eosinophil myelocytes are a stable cell line that possesses biochemical and morphological features consistent with native eosinophils (4). AML14.3D10 cells are a cytokine-independent eosinophil myleocyte cell line able to synthesize and release all ESGPs as well as granulocyte/macrophage colony-stimulating factor (GM-CSF; Fig. 1A) (4). Cells were maintained in RPMI 1640 (Life Technologies; Gaithersburg, MD) with 10% fetal calf serum (Life Technologies), 50 μM 2-mercaptoethanol (Sigma; St. Louis, MO), 1 mM sodium pyruvate (Sigma), and 0.1 mM nonessential amino acids (Life Technologies) and used between passages 7 and 25. T84 cells are human colonic carcinoma lines that, when plated on permeable membrane supports, form a polarized monolayer of columnar intestine-like epithelial cells and serve as excellent models for columnar intestinal epithelial cells and were grown as previously described (8). T84 epithelial cells form a high-resistance monolayer and were used between passages 55 and 70. T84 cells were grown as monolayers in a 1:1 mixture of Dulbecco-Vogt-modified Eagle's medium and Hanks F-12 medium supplemented with 15 mM HEPES buffer (pH 7.5), 14 mM NaHCO3, 40 mg/ml penicillin, 8 mg/ml ampicillin, 90 mg/ml streptomycin, and 5% newborn calf serum. Monolayers were subcultured from flasks every 7–14 days by brief trypsin treatment (0.1% trypsin and 0.9 mM EDTA in Ca2+- and Mg2+-free PBS). T84 cells were split and plated on either 0.3- or 5-cm2 inserts (Costar).
Human eosinophil isolation.
Human leukocytes were harvested according to an Institutional Review Board-approved protocol at the Brigham and Women's Hospital as previously described (8). Briefly, leukocytes were freshly isolated from whole blood obtained by venipuncture and anticoagulated. Plasma and mononuclear cells were removed from the buffy coat after centrifugation (400 g for 20 min) at room temperature. Red blood cells (RBCs) were pelleted using 2% gelatin sedimentation. Residual RBCs were removed by lysis in ice-cold NH4Cl buffer. Leukocytes were isolated in a Ficoll-Hypaque separation medium, erythrocytes and polymorphonuclear neutrophils were lysed in 3.5% NaCl, and the remaining cells were washed. The remaining cells were passaged through an eosinophil enrichment column (R&D Systems; Minneapolis, MN). Viable eosinophils were 95% pure as assessed by fast green (4%).
Eosinophil granule isolation.
Eosinophil-secreted granule proteins were isolated as previously described (25). Briefly, eosinophils were washed with ice-cold 0.25 M sucrose and resuspended in the same medium. Vigorous pipetting through a narrow bore pipette disrupted cell membranes. Cell lysates were centrifuged at 600 g to remove cell debris and unbroken cells and at 13,000 g for 20 min to pellet granules. Eosinophil granules were lysed in 10 mM HCl and sonicated. Granule lysates were centrifuged at 40,000 g, and the resulting supernatant was fractionated on a Sephadex G-50 12 × 100-cm column equilibrated with 0.025 M acetate buffer (pH 4.3) containing 0.15 M NaCl. Acetate buffer was harvested for vehicle control conditions. Eosinophil granule proteins were identified by their distinctive chromatographic elution profile from the G-50 column (3). For example, the third protein peak yields fractions containing pure MBP. The purity of protein isolation was determined by SDS-PAGE. the MBP concentration was determined by spectrophotometric absorbance at 277 nm with an extinction coefficient of 26.3 and the EDN concentration at 280 nm with an extinction coefficient of 15.5. The identity of each protein was confirmed by radioimmunoassay or Western blot analysis. Concentrations of MBP used correlated with that seen in vivo and approximated that found in 106 AML14.3D10 cells (4).
Measurement of intestinal barrier function.
Upon reaching confluence based on stable transepithelial resistance (TER) measurements (27), T84 cells were placed in noncontact coculture with varying concentrations of AML14.3D10 cells or freshly isolated eosinophils (model shown in Fig. 1B). Ratios of eosinophils to epithelial cells ranged from 0.75–3.0 eosinophils/epithelial cell. Each AML14.3D10 cell contains ∼10 pg MBP/cell (Ref. 4 and unpublished data). In other experiments, various concentrations of anti-MBP antibody were added to the basolateral well at the beginning of the coculture, and TER was measured as previously described (27). At defined end points, basolateral supernatants were collected and immediately frozen at −80°C. IL-4 and IFN-γ ELISA measurements (Biosource; Camerillo, CA) and lactate dehydrogenase analysis (Promega; Madison, WI) were performed on supernatants.
Western blot analysis.
Cytoplasmic extracts were isolated from confluent monolayers of epithelia on 10-mm2 monolayers as described previously (27). Samples (20 μg/lane) were resolved by nonreducing SDS-PAGE, transferred to nitrocellulose membranes, and blocked overnight in blocking buffer (250 mM NaCl, 0.02% Tween 20, 5% goat serum, and 3% BSA). For Western blot analysis, rabbit anti-MBP, anti-EDN, or anti-occludin (Zymed; San Francisco, CA) was added, blots were washed, and species-matched peroxidase-conjugated secondary Ab was added. Labeled bands from washed blots were detected by enhance chemiluminescence (Amersham) and analyzed by NIH Imagemaker.
Assessment of gene-specific mRNA.
Transcriptional analysis of epithelial cells exposed to purified human MBP was assessed in RNA derived from control or MBP-exposed epithelia (T84 cells at 4 or 24 h post-MBP). RT-PCR analysis of mRNA levels was performed using DNAse-treated total RNA as previously described (27) using primers specific for occludin (forward primer 5′-CGG CTA TGG AGG CTA TGG CTA TG-3′ and reverse primer 5′-ATG ACC CCA GGA CAA TGG C-3′, 680-bp fragment) or control β-actin (forward primer 5′-ATG ACT TCC AAG CTG GCC GTG GCT-3′ and reverse primer 5′-TCT CAG CCC TCT TCA AAA ACT TCT C-3′, 661-bp fragment). Each primer set was amplified using 25 cycles of PCR at 94°C for 45 s, 60°C for 45 s, and 72°C for 1 min and a final extension of 72°C for 5 min. PCRs were then visualized on a 1.5% agarose gel containing 5 μg/ml ethidium bromide.
Confocal analysis of T84 monolayers.
Immunofluorescent staining of epithelial monolayers was performed as previously described (40). T84 cells were grown to confluence on inserts. Cells were placed in coculture with AML14.3D10 eosinophils for 48 h. Monolayers were washed once in PBS, fixed for 10 min at room temperature in 1% paraformaldehyde in cacodylate buffer [0.1 M sodium cacodylate (pH 7.4) and 0.72% sucrose], and permeabilized for 10 min in PBS containing 0.2% Triton X-100 and 3% BSA. After being washed two times with PBS, cells were stained for 1 h with a monoclonal anti-occludin antibody (1:1,000, Santa Cruz Antibody). Monolayers were washed and incubated with goat anti-mouse Oregon green (1 μg/ml, Molecular Probes; Eugene, OR). Laser Sharp imaging software (Bio-Rad; Hercules, CA) was used to determine protein localization.
Th2-mediated colitis model.
The role of MBP in a Th2-mediated model of colitis, oxazolone colitis, was examined using wild-type and MBP-1-null mice (16) in accordance with the Animal Welfare Committee at Brigham and Women's Hospital. Male and female mice were bred on a 129/SvJ background as previously described and used at 10–12 wk of age. Age- and sex-matched wild-type 129/SvJ mice served as controls (16). Oxazolone colitis was induced as previously described with modifications (5, 50). To presensitize mice, a 2 × 2-cm field of the abdominal skin was shaved, and 200 μl of a 3% (wt/vol) solution of 4-ethoxymethylene-2-phenyl-2-oxazoline-5-one (oxazolone, Sigma) in 100% ethanol was applied. Five days after presensitization, mice were challenged intrarectally with 150 μl of 1% oxazolone in 50% ethanol or only 50% ethanol (i.e., vehicle) under anesthesia. Oxazolone was administered per rectum via a 3.5-Fr catheter and a 1-ml syringe. The catheter was inserted so that the tip was 4 cm proximal to the anal verge. To ensure the distribution of oxazolone within the entire colon and cecum, mice were held in a head-down vertical position for 30 s after the injection.
Mice were weighed before the induction of colitis and daily thereafter. Before death, intestinal permeability was assessed as previously described (49). Briefly, mice received an oral gavage of FITC-dextran 4 kDa (40 mg/kg, Sigma) 4 h before death. Upon death, serum was collected and immediately analyzed for FITC-derived fluorescence with results reported as relative units of fluorescence per milliliter of serum.
Electrophysiological data were compared by two-factor ANOVA or by Student's t-test where appropriate. Values are expressed as means ± SE of n monolayers from two to three separate experiments.
Eosinophils alter epithelial barrier function.
Because eosinophils represent a resident cell type anatomically positioned within the lamina propria (18), long-term cocultures were established in which eosinophils were exposed to basolateral epithelial surfaces without direct contact with basal poles (i.e., noncontact coculture; Fig. 1B). As shown in Fig. 2, A and B, increasing numbers of eosinophils within the coculture model led to a progressive decline in epithelial barrier function. Indeed, the concentration-dependent increase in 4-kDa fluorescent tracer flux (Fig. 2A; P < 0.01 by ANOVA) paralleled the progressive decrease in TER (Fig. 2B; P < 0.01 by ANOVA). Such results were not explained by eosinophil cell death. Viability of eosinophils, as determined by trypan blue exclusion, was not different between monoculture of eosinophils and coculture with epithelia (98 ± 2% vs. 97 ± 3% viability, respectively) or epithelial cell death (see below).
The time course of barrier loss during coculture indicates a time-dependent fall in TER, as shown in Fig. 2C (P < 0.01 by ANOVA). This coculture response was not evident in the first 24 h but was maximal by 72 h with a half-time (t1/2) of 36 h. As shown in Fig. 2C (arrow), the replacement of apical and basolateral wells with cell-free fresh media resulted in the recovery of TER values to precoculture levels within 24 h (P = not significant compared with baseline TER), suggesting the release of a bioactive soluble mediator and maintenance of epithelial viability. Moreover, this fall in barrier function was not evident when AML14.3D10 eosinophils were replaced with the untreated myelomonocytic cell line HL-60 C15 in the basolateral well (data not shown), suggesting specificity for eosinophils. A similar decline in TER was measured when peripheral blood human eosinophils were cocultured with T84 cells (Fig. 2D). These results indicate that eosinophil-epithelial coculture results in the liberation of a soluble, bioactive mediator that alters T84 cell monolayer barrier function.
An eosinophil-derived factor alters TER.
Cell-free coculture supernatants were collected and administered to basolateral surfaces of naive T84 monolayers to determine whether a secreted, soluble mediator could account for this result. As shown in Fig. 3A, TER significantly declined with maximal impact at 24 h (P < 0.05 by ANOVA) and a steady decline in bioactivity over an additional 48-h time period. In addition, serial dilution of supernatants derived from cocultures resulted in a loss of this barrier-altering property (Fig. 3B; P < 0.05 by ANOVA).
In an effort to characterize the molecular mass of the responsible factor(s), cell-free products were size fractionated (<100 and <5 kDa). Filtrates were added to naive monolayers, and TER was examined. A barrier-altering product was evident in 100-kDa filtrates (59 ± 1% fall in TER at 24 h, P < 0.01) and lost in 5-kDa filtrates (6 ± 2% fall in TER at 24 h, P = not significant), suggesting that at least one active component was in the nominal molecular mass range from 5 to 100 kDa. Attempts to identify the barrier-altering cytokines IL-4 and IFN-γ (13, 43) within these supernatants (ELISA analysis) proved negative (level of detection <4 pg/ml).
MBP alters epithelial permeability.
On the basis that this barrier-altering mediator 1) is active in noncontact cocultures, 2) can be trapped in cell-free supernatants, and 3) has a molecular mass between 5 and 100 kDa, we hypothesized that the release of eosinophil-secreted granule proteins may be responsible for the loss of barrier function. As shown in Fig. 4A, Western blot analysis was utilized to define the presence of EDN and MBP in supernatants derived from AML14.3D10 eosinophils alone, epithelia alone, or cocultures. As can be seen, neither AML14.3D10 nor T84 cells alone released detectable levels of EDN or MBP, but coculture revealed the presence of both granule proteins as soluble proteins, suggesting the necessity of coculture for mediator release.
We next determined whether native MBP or EDN directly influences epithelial permeability by two approaches. First, as shown in Fig. 4B, native human-derived MBP but not EDN was shown to significantly decrease epithelial barrier function in a concentration-dependent fashion. This influence of MBP was reversible as demonstrated by the rapid recovery after its removal (Fig. 4C). In addition, rabbit anti-MBP antibody significantly inhibited the decline in epithelial barrier function in a concentration-dependent manner (Fig. 4D). When T84 monolayers were pretreated with anti-MBP antibody and then exposed to coculture supernatants, the decline in TER was significantly reduced compared with untreated monolayers at the 48-h time point (97 ± 6.7 vs. 74 ± 1.5 as a percentage of control, P < 0.02). Taken together, these results provide evidence that MBP significantly alters epithelial permeability.
MBP downregulates occludin expression on epithelia.
As an extension of these findings, we sought to define the potential mechanisms of MBP regulation of epithelial barrier function. RT-PCR analysis was used to determine occludin expression at the RNA level. Results revealed a time-dependent loss of occludin RNA with MBP exposure (Fig. 5A). Similarly, Western blot analysis on lysates derived from epithelial cells in coculture revealed a time-dependent reduction of occludin by as much as 51% in 72-h cocultures (Fig. 5B). Finally, after 72 h of coculture, confocal analysis of exposed T84 monolayers revealed decreased occludin expression in a diffuse pattern (Fig. 5, C and D). These results identify occludin as a potential target for MBP-modulated barrier function.
MBP-1-deficient (null) mice are protected from Th2-mediated colitis.
We utilized MBP-1-null mice (16) in a Th2-mediated colitis model to determine the relevance of the in vitro findings suggesting that MBP contributes to mucosal inflammation. Oxazolone colitis is a predominantly Th2-mediated colitis characterized histologically by a mixed colonic infiltrate consisting of neutrophils and eosinophils (5, 16, 50). After the induction of oxazolone colitis, MBP-1-null mice were significantly protected from colitis-induced weight loss compared with wild-type mice (Fig. 6A). This difference was most prominent at early time points (days 1–3). Whereas both wild-type and MBP-1-null mice developed increased permeability at day 2 compared with mice treated with vehicle, MBP-1-null mice were significantly protected from the increased permeability seen in wild-type mice (Fig. 6B). These findings demonstrate that MBP-1 significantly contributes to inflammation in Th2-mediated colitis.
Intestinal inflammatory diseases such as IBD and food allergic diseases are often associated with altered epithelial barrier function (17, 28, 31, 33). In addition, rodent model systems of mucosal inflammation are often associated with increased permeability (64, 65). Yet, no clear consensus exists as to the basis for this leaky gut. The association of these gastrointestinal inflammatory diseases with mucosal eosinophilia and eosinophil granule deposition (10, 14, 42, 47, 59, 60) led to the hypothesis that eosinophil-derived mediators diminish epithelial barrier function. In support of this hypothesis, we demonstrated that supernatants from eosinophil-epithelial cocultures harbor a bioactive mediator, MBP, capable of diminishing epithelial permeability. Additionally, MBP downregulates the tight junctional molecule occludin, thus providing a potential mechanism by which eosinophils participate in barrier dysfunction.
Traditionally, MBP has been characterized as a cytotoxin leading to disruption of cell membrane lipid bilayers, ultimately causing cell death (2, 22). Yet a number of studies demonstrate additional biological properties of MBP. This highly charged cationic protein can induce epithelial mucus, prostaglandin and ion secretion (34, 36, 38, 62), smooth muscle contraction (61), basophil histamine release (51), neutrophil oxidative burst (29), fibroblast IL-6 expression (54), and eosinophil and fibroblast IL-8 secretion (26, 37). The mechanism by which MBP induces these host responses is not yet certain, but the rapid activation of tyrosine phosphorylation signaling by MBP (29) suggests the presence of a MBP receptor. In our present study, we have shown that MBP can induce transcriptional responses, providing further support for a MBP-related pathway.
In contrast to our findings, Brandt et al (6) recently showed that eosinophils did not appear to play a role in an ovalbumin (OVA)-induced model of intestinal inflammation. Sensitized and challenged mice developed increased eosinophils in the bloodstream as well as in the jejunum. Compared with wild-type Balb/c mice, IL-5 transgenic (increased eosinophils) and IL-5/eotaxin-1 double-deficient mice (diminished eosinophils) did not have differences in the diarrhea. In contrast, Hogan and colleages (31a) identified a specific role for eosinophils in gastric dysmotility in a different model of gastrointestinal eosinphilia. In this model system, mice received intraperitoneal sensitizaton followed by an oral challenge with enterically protected OVA-impregnated beads. These OVA systems do not appear to lead to epithelial disruption with increased permeability as seen in the present study. Possible reasons for these differences include the different strains of mice used and differences in the location of the gastric inflammation. For example, the OVA model system affects the jejunum, and thus compensatory water reabsorption may occur in the colon, leading to less-severe phenotypic responses.
Some insight was gained into the mechanism(s) of MBP alterations in epithelial barrier function. Occludin is one of many recognized proteins associated with the apical tight junction (24). This transmembrane protein possesses two extracellular loops that span the tight junctional space and provide the barrier seal necessary to compartmentalize luminal contents from serosal spaces. The occludin COOH-terminal domain interacts with zona occludens-1 (ZO-1), and deletions of this terminus result in the loss of the occludin/ZO-1 association and increased paracellular flux. Studies of transcriptional regulation of occludin have been elucidated recently by identification and characterization of the human occludin promoter (44), but significant work will be necessary to determine whether MBP regulation of occludin is direct or indirect and whether such pathway(s) are specific for MBP.
The mechanism(s) by which eosinophils are activated within the gastrointestinal microenvironment is not certain, but priming with eosinotropic cytokines such as GM-CSF is likely to be critical (53, 58). Columnar intestinal and tracheal epithelia are able to synthesize GM-CSF, providing a local source for eosinophil priming in situ (11, 15, 19). Moreover, evidence from studies using eosinophils from allergic patients suggests that eosinophil priming also occurs in vivo (21, 23, 63) and that GM-CSF, IL-3, and IL-5 priming of eosinophils results in the induction of an “activated” phenotype (9, 52). Although not addressed in this study, we speculate that eosinophils are activated by the epithelial production of GM-CSF, IL-3, or IL-5.
Previous work has demonstrated that the intestinal mucosa is an enriched source of chemokines, cytokines, and growth factors derived from a diverse number of resident cell types (1). In this regard, intestinal epithelial cells, the predominant barrier to the external environment, are anatomically positioned to coordinate interplay of these paracrine mediators. For example, previous studies have demonstrated that epithelial function is dynamically modulated by paracrine mediators released in the local milieu during inflammatory events, including diminished barrier function (13, 32, 43), increased flux of inert molecules (43), decreased Cl− secretory responses to agonist stimulation (13, 32), enhanced expression of major histocompatability complex class II molecules (12, 46), and modulated interaction with neutrophils (12, 13). Although such data suggest that paracrine mediators have the potential to modulate epithelial phenotype, it is not clear in what fashion these mediators are presented to epithelia or whether local concentrations exist to significantly influence these responses.
To ascertain whether resident immunological cell types substantially influence the epithelial phenotype, we used both in vitro model systems and animal models in which intact cell types, producing their plethora of bioactive compounds, interrelate and do so in an anatomically correct fashion. Our work focused on the impact of one eosinophil granule protein, MBP, as a biologically active mediator. The possibility exists that other eosinophil-related products have barrier-altering properties. In that light, EPO was recently shown to be a critical mediator in the pathogenesis of the dextran sodium sulfate colitis model system (20). Other inflammatory cells that infiltrate the mucosa in oxazolone colitis, such as neutrophils, and associated proinflammatory cytokines, such as TNF-α, could play a role in this barrier altering cascade (7, 39).
In conclusion, our results show that exposure of epithelia to an eosinophil-specific mediator, MBP, results in a loss of intestinal barrier function. Furthermore, the results identify a potential target for MBP, namely, the tight junctional protein occludin. We speculate that the release of MBP along the basolateral surface of epithelial cells impacts barrier function, implicating a role for eosinophils in the pathogenesis of inflammatory mucosal diseases.
This study was supported by the Charles Janeway Scholarship; by National Institutes of Health (NIH) Grants K08 DK-02564, R01 DK-62245 (to G. T. Furuta), AI-25230 (to S. J. Ackerman), DK-44319, DK-51362, and DK-53056 and the Harvard Digestive Disease Center (to R. S. Blumberg and E. E. S. Nieuwenhuis).
The authors acknowledge the excellent technical assistance of Donald Lawrence, Robyn Carey, Kristen Synnestevdt, Punitha Ramalingen, and Lei Lu. We also acknowledge insights by Roderick Bronson and the generous advice and support from Dr. Sean Colgan.
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 © 2005 the American Physiological Society