Infiltration of activated neutrophils [polymorphonuclear leukocytes (PMN)] into the lung is an important component of the inflammatory response in acute lung injury. The signals required to direct PMN into the different compartments of the lung have not been fully elucidated. In a murine model of LPS-induced lung injury, we investigated the sequential recruitment of PMN into the pulmonary vasculature, lung interstitium, and alveolar space. Mice were exposed to aerosolized LPS and bronchoalveolar lavage fluid (BAL), and lungs were harvested at different time points. We developed a flow cytometry-based technique to assess in vivo trafficking of PMN in the intravascular and extravascular lung compartments. Aerosolized LPS induced consistent PMN migration into all lung compartments. We found that sequestration in the pulmonary vasculature occurred within the first hour. Transendothelial migration into the interstitial space started 1 h after LPS exposure and increased continuously until a plateau was reached between 12 and 24 h. Transepithelial migration into the alveolar air space was delayed, as the first PMN did not appear until 2 h after LPS, reaching a peak at 24 h. Transendothelial migration and transepithelial migration were inhibited by pertussis toxin, indicating involvement of Gαi-coupled receptors. These findings confirm LPS-induced migration of PMN into the lung. For the first time, distinct transmigration steps into the different lung compartments are characterized in vivo.
- polymorphonuclear leukocytes
- pulmonary circulation
- flow cytometry
- pertussis toxin
acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) are characterized by a disturbance of the alveolar-capillary barrier associated with several clinical disorders. There is no specific therapy, and the mortality of this disease is still high. Our current understanding of the molecular mechanisms of ALI/ARDS has recently been described as “embryonic at best” (27).
Migration of activated polymorphonuclear leukocytes (PMN) plays a key role in development of ALI and ARDS (1). Here, we investigate the sequential migration steps from blood to air space (intravascular sequestration, transendothelial migration, transepithelial migration).
A variety of stimuli induce PMN migration into the lung. Endotoxin of gram-negative bacteria [lipopolysaccharide (LPS)] induces a range of inflammatory responses. Toll-like receptor-4 (TLR4) is the most important cellular receptor for LPS. LPS stimulates the response to chemoattractants and increases PMN migration at sites of inflammation (14). TLR4 is essential for LPS-induced PMN migration into the lung as shown by the absence of a response in TLR4-deficient mice (3). In the lung, the response to LPS is regulated by radioresistant cells, most likely endothelial cells (2) or alveolar macrophages (28).
The administration of LPS alone might not reflect the whole complexity of the human disease, because it does not consider preexisting diseases, fluid resuscitation, or mechanical ventilation (36). However, infections with gram-negative bacteria and exposure to their predominant pathogenic component play a key role in both development and outcome of ARDS (26).
PMN trafficking into the vascular compartment of the lung, also known as “margination” (9), and into the bronchoalveolar space has been studied extensively in various models of ALI. Adhesion molecules and CXC chemokines have been shown to be involved. CXC chemokines, such as CXCL8 (IL-8), promote PMN migration into the alveolar compartment (29), and pertussis toxin (PTx)-dependent chemokine receptors are essential for PMN infiltration in the lung (4, 42). Selectins and integrins are thought to be required for PMN sequestration into the vascular compartment (6). However, results from studies using monoclonal antibodies and mutant mice have yielded conflicting results (10). The importance of investigating each step of PMN migration in the lung has been emphasized recently (8).
Methodological limitations complicate the assessment of PMN trafficking in the lung. Most studies employ indirect parameters to assess PMN trafficking in the lung. For instance, the drop in circulating PMN counts in response to an inflammatory stimulus was used to estimate PMN sequestration in the lung vasculature (6, 31). However, recruitment to other organs might occur at the same time. Ex vivo labeling of murine PMN might result in neutrophil activation that makes results uninterpretable (7). Myeloperoxidase activity in the lung is often measured to estimate PMN infiltration, but this technique is not able to distinguish between intravascular and interstitial PMN. Intravital microscopy of the pulmonary microcirculation has recently become available in mice (35, 39) and will promote insights into the interactions between leukocytes and endothelium. However, this technique is technically challenging because of the respiratory movement, requires mechanical ventilation, and allows observation of only the most superficial lung capillaries, which may not be representative of the whole lung. Morphometric analysis, such as electron microscopy (16), is useful but remains semiquantitative, time consuming, and expensive.
In this study, we developed a flow cytometry-based approach to assess the different steps of PMN trafficking in a murine model of LPS-induced ALI. PMN accumulation in the pulmonary vasculature, transendothelial migration into the interstitium, and transepithelial migration from the interstitium into the air space occurred as a sequential process in a time-dependent manner. Our findings improve the current understanding of neutrophil recruitment into the inflamed lung and airways in a model that mimics some aspects of ALI/ARDS.
Wild-type male C57BL/6 mice were obtained from Jackson Labs (Bar Harbor, ME). All animal experiments were approved by the Animal Care and Use Committee of the University of Virginia. Mice were 8–12 wk of age.
Antibodies for flow cytometry.
Rat anti-mouse antibodies for flow cytometry were obtained from Pharmingen (anti-CD45, clone 30-F11), Caltag (anti-7/4, recognizing the 7/4 antigen on murine neutrophils), and eBioscience (anti-TER-119, recognizing glycophorin A-associated TER-119 on cells of the erythroid lineage). Anti-mouse GR-1 antibody was purified from supernatant of the GR-1 hybridoma (ATCC) by the biomolecular facility of University of Virginia. GR-1 was labeled with a staining kit following the manufacturer's directions (Alexa Fluor 633, Molecular Probes). Appropriate rat anti-mouse IgG2a and IgG2b (Pharmingen) were used as isotype controls.
Murine model of ALI.
Aerosolized LPS was utilized to induce PMN infiltration in the lung (40). Besides PMN migration, LPS inhalation is known to induce the expression of various inflammatory mediators such as chemokines and adhesion molecules. LPS has also been shown to increase airway resistance (24). Up to four mice were exposed simultaneously to aerosolized LPS in a custom-built cylindrical chamber (20 cm in length, 9 cm in diameter) connected to an air nebulizer (MicroAir; Omron Healthcare, Vernon Hills, IL). This system produced particles in the range of 1–5 μm. LPS from Salmonella enteritidis (Sigma, St. Louis, MO) was dissolved in 0.9% saline (500 μg/ml), and mice were allowed to inhale LPS for 30 min. One side of the chamber was connected to a vacuum pump, and a constant flow rate of 15 ml/min was ensured by a flow meter (Gilmont Instruments, Barrington, IL).
PMN counts in bronchoalveolar lavage fluid and lung tissue.
At different time after LPS exposure, mice were anesthetized with an intraperitoneal injection of ketamine (125 mg/kg; Sanofi Winthrop Pharmaceuticals, New York, NY), xylazine (12.5 mg/kg; Phoenix Scientific, St. Joseph, MO), and atropine sulfate (0.025 mg/kg; Fujisawa, Deerfield, IL). The pulmonary circulation was rinsed by injection of 10 ml of PBS at 25 cmH2O into the beating right ventricle after the inferior vena cava had been cut to allow exsanguination. This was done to remove nonadherent PMN from the pulmonary vessels. The trachea was cannulated (22 GA Insyte, Becton Dickinson), and 1 ml of PBS was infused intratracheally and withdrawn. This procedure was repeated six times, resulting in a total volume of 7 ml. Bronchoalveolar lavage (BAL) fluid was centrifuged for 5 min at 300 g. The pellet was resuspended in 1 ml of buffer (1% BSA and 0.1% sodium azide in PBS), and a 10-μl aliquot was used for cell count with a hemocytometer (Trypan blue exclusion).
After performing BAL, we harvested lungs en bloc. Mediastinal tissue was removed, and lungs were minced and digested with 125 U/ml collagenase type XI, 60 U/ml hyaluronidase type I-s, and 60 U/ml DNase (all Sigma) at 37°C for 30 min. Digested lungs were passed through a 70-μm cell strainer (BD Falcon, Bedford, MA), and the resulting cell suspension was centrifuged for 10 min at 300 g. The pellet was lysed with 0.83% NH4Cl to remove erythrocytes and centrifuged again. Pellet was resuspended in buffer, and cells were counted with a hemocytometer.
PMN were identified by 1) their typical appearance in the forward scatter-side scatter (FSC/SSC), 2) their expression of CD45+, and 3) two independent PMN markers, 7/4 and GR-1 (19), and the absolute numbers of leukocytes (CD45+) and PMN were calculated. Appropriate isotype antibodies were used to set the gates. All studies were performed on a FACS Calibur (Becton Dickinson, San Jose, CA), and data were analyzed with FlowJo software (Tree Star, Ashland, OR). To confirm the presence of PMN within the different populations as defined by flow cytometry, we sorted both 7/4+GR-1+ and 7/4+GR-1− cells (FACS Vantage, Becton Dickinson) and characterized them morphologically by cytospin (Diff Quick staining, IMEB).
In vivo trafficking experiments.
Dialyzed Alexa 633-labeled rat anti-mouse GR-1 (10 μg) antibody was injected intravenously and allowed to circulate for 5 min to bind to intravascular PMN. After 5 min, mice were killed. BAL was obtained as described above, and the lungs were homogenized in the presence of excess unlabeled anti-GR-1 to prevent possible binding of excess Alexa 633-GR-1 to extravascular PMN. Cell suspensions from BAL and lung tissue were made, and cells were counted in a hemocytometer.
Nonperfused or occluded vessels in the lung might result in trapped neutrophils, not accessible for the injected antibody. This would lead to an underestimation of PMN counts in the intravascular compartment. Nonperfused/occluded vessels contain white and red blood cells. To assess the significance of this phenomenon in our model, a monoclonal antibody to the TER-119 antigen on erythrocytes was injected intravenously (TER-119; eBioscience, San Diego, CA). This 52-kDa molecule is associated with glycophorin A on cells of the erythroid lineage (22). Five minutes after injection, blood and lungs were harvested. Erythrocytes were defined by their typical appearance in the forward-sideward scatter and the amount of TER-119+ erythrocytes in each organ was expressed as percentage of total erythrocytes by flow cytometry. To assess the effect of intravenous injection of anti-GR-1 on peripheral PMN counts, in some experiments, blood was withdrawn from the tail vein and blood counts were performed before and 10 min after antibody injection with an automatic cell counter (Hemavet; Drew Scientific, Dallas, TX). In all experiments, animals exposed to aerosolized saline served as control.
Cytospin of BAL.
In some experiments, cytospins of the BAL (without LPS treatment and 24 h after LPS inhalation) were performed with a cytocentrifuge (Shandon, Southern Sewickley, PA). The cytospun cells were Giemsa stained, air-dried, and coverslipped.
To confirm lung injury in our LPS inhalation model, we determined microvascular permeability using the Evans blue dye extravasation technique. Evans blue (20 mg/kg iv, Sigma-Aldrich) was injected 30 min before death. Lungs were perfused through the spontaneously beating right ventricle. Lungs were removed, and Evans blue was extracted as described previously (32). The absorption of Evans blue was measured at 620 nm (A620) and corrected for the presence of heme pigments: A620 (corrected) = A620 − (1.426 × A740 + 0.030) (45). Extravasated Evans blue was determined 6 h after LPS or saline inhalation and calculated against a standard curve (micrograms Evans blue dye per gram lung).
Chemokines have been shown to regulate PMN migration in the lung (42). To block chemokine-mediated PMN migration, some mice received tail vein injections of 4 μg of PTx from Bordetella pertussis (lyophilized powder, Sigma) 30 min before LPS exposure. This dose completely inhibits Gα1-mediated signaling (41). PTx was dissolved in physiological saline. PMN in lung and BAL were assessed 14 h after LPS exposure. In addition, a dose-response curve was established for each lung compartment 12 h after LPS exposure (0, 0.04, 0.4, or 4 μg PTx/mouse).
Data were analyzed with the Excel software package (Microsoft). PMN counts were compared by paired Student's t-test. P < 0.05 was considered statistically significant. Data were expressed as means ± SE.
LPS-induced PMN recruitment into lung and BAL.
LPS inhalation induced a time-dependent PMN recruitment into lung and BAL (Fig. 1). In the lung, significant numbers of leukocytes (CD45+ cells) were present even before LPS administration, and their numbers increased only moderately from four million to a maximum of nine million cells at 4 h after LPS. Neutrophils were also present in resting lungs (approximately one million per mouse), consistent with the concept of a physiologically marginated pool in the pulmonary vasculature. Lung neutrophil numbers reached more than six million cells/mouse at 4 h of LPS administration, which is severalfold more than the total number of all circulating neutrophils consistent with a previously described release of PMN from the bone marrow (37, 46). At the peak of the response, neutrophils accounted for 74 ± 7% of all leukocytes in the lung.
No PMN were observed in the BAL at 0 h. PMN recruitment into the air space was delayed, and the first PMN did not appear until 2 h. Between 2 and 4 h, neutrophil recruitment was very pronounced. After 48 h, cell counts in BAL were reduced but did not reach baseline.
In vivo GR-1 labeling.
In the lung homogenate, GR-1 labeling was utilized to distinguish between PMN derived from the pulmonary vasculature (GR-1+7/4+) and PMN derived from interstitial space (GR-1−7/4+). We assessed the PMN labeling 5 min after injection of anti-GR-1 antibody. We found that almost all blood PMN (99.2 ± 0.4%) had been stained with GR-1 5 min after antibody injection (Fig. 2, A and B; Table 1). To test whether GR-1 antibody was leaking into the BAL, GR-1 labeling of PMN in BAL (CD45+7/4+) was assessed at different time points after LPS exposure, 5 min after antibody injection. No GR-1+ cells were found in BAL. When GR-1 antibody was added after BAL harvesting, all PMN were GR-1+ (positive control) (Fig. 2, C and D; Table 1). GR-1+ PMN did not appear in the BAL until 4 h after antibody injection (data not shown).
Although all circulating PMN were shown to be GR-1+, potential non- or poorly perfused areas of the pulmonary vasculature might be inaccessible for an intravenously injected antibody. This would result in an underestimation of the intravascular or an overestimation of the interstitial PMN concentration in the lung, respectively. This effect might occur particularly in the injured lung. We therefore labeled erythrocytes by intravenous injection of an antibody to the TER-119 antigen of red blood cells intravenously 24 h after LPS exposure. Red blood cells are not found in the lung interstitium or BAL (5). The amount of TER-119+ cells among all erythrocytes was then determined in both blood and lung homogenate. Five minutes after injection, we found 98.8% of all blood erythrocytes to be TER-119+. At the same time, 97.7% of all erythrocytes were TER-119+ in the lung homogenate, indicating that the injected antibody is able to bind to almost all red cells in the pulmonary vasculature, suggesting that nearly all vessels were perfused.
Effect of anti-GR-1 antibody on PMN blood counts.
GR-1 antibody can induce severe neutropenia when given at high doses (33). Therefore, we performed peripheral blood counts before and 10 min after antibody injection. GR-1 injection did not affect blood PMN counts at the concentration used in our study (0.89 ± 0.14 × 103/μl before, 0.86 ± 0.16 × 103/μl after injection; P = 0.90).
In vivo trafficking experiments.
We assessed the concentration of PMN in the different lung compartments at 0, 1, 2, 4, 12, and 24 h after LPS exposure. At each time point, anti-GR-1 antibody was injected 5 min before death. PMN were identified by flow cytometry (FSC/SSC gate, CD45+7/4+), and GR-1 was utilized to distinguish between intravascular (GR-1+) and interstitial (GR-1−) PMN (Fig. 3). Both populations predominantly consisted of PMN as confirmed morphologically by cytospins (7/4+/GR-1+, 99% PMN; 7/4+GR-1−, 97% PMN; data not shown). At 0 h, 86% of all PMN were found in the vascular compartment. LPS induced PMN accumulation in the pulmonary vasculature. Both absolute and relative PMN counts increased rapidly until a peak was reached after 4 h. After 4 h, PMN counts in the vasculature decreased and returned to baseline at 24 h after LPS exposure (Fig. 3A).
PMN concentration in both interstitium and BAL was negligible at 0 h. After 4 h, 33% of all pulmonary PMN were found in the interstitium. At the same time, PMN represented 91% of all cells in the BAL (Fig. 3B). After 24 h, the majority of PMN (78%) in the lung were found in the extravascular space. Note that PMN in BAL appear GR-1− as the GR-1 antibody injected 5 min before death remains confined to the vasculature. In control animals, saline inhalation induced a mild PMN accumulation in the pulmonary vasculature. No migration into the interstitium or into the alveolar air space was observed in these mice (data not shown).
Kinetics of transendothelial and transepithelial PMN migration.
Interstitium and air space were free of PMN at 0 h. Transendothelial migration into the interstitial space started 1 h after LPS exposure and increased continuously. After 12 h, the majority of PMN in the lung were found extravascular (interstitium and air space) (Fig. 4). Until 2 h, the intravascular accumulation outpaced the neutrophil accumulation in the extravascular space so that the proportion of extravascular PMN did not change. By contrast, at 4 h the rate of accumulation of intravascular neutrophils slowed, and extravascular PMN accounted for 33% of all neutrophils in the lung.
The kinetics of neutrophil recruitment into the BAL was different in that no cells were found at 1 h and only a very small number (130,000 per mouse) at 2 h, after which time neutrophil numbers increased rapidly and then followed the number of interstitial neutrophils with a delay of ∼2 h.
The appearance of PMN in the BAL was confirmed by cytospin. Cytospins of BAL were analyzed in untreated mice and in mice 24 h after LPS inhalation. The predominant cells in BAL of untreated mice were alveolar macrophages (Fig. 4, left inset). As expected, PMN dominated the cell population 24 h after LPS exposure (Fig. 4, right inset).
Vascular leakage was determined to confirm lung injury in our LPS inhalation model. Six hours after LPS inhalation, Evans blue extravasation was significantly higher compared with saline inhalation (66.2 ± 6.2 vs. 29.3 ± 3.7 μg per g lung; P = 0.002) (Fig. 5).
In some mice, 4 μg of PTx (41) were injected intravenously 30 min before LPS challenge. Lungs and BAL were harvested 14 h after LPS exposure, and anti-GR-1 was injected 5 min before death to distinguish between intravascular and interstitial PMN. LPS induced a significant PMN migration into all three lung compartments. Mice pretreated with PTx exhibited normal PMN sequestration into the pulmonary vasculature but showed reduced PMN migration into the lung interstitium (0.5 ± 0.1 × 106 vs. 2.7 ± 0.4 × 106; P < 0.01) (Fig. 6, A and B). Accordingly, almost no PMN were found in the alveolar air space at 14 h after LPS (0.3 ± 0.01 × 106 vs. 2.8 ± 0.3 × 106, P < 0.01) (Fig. 6B). The inhibitory effect of PTx was dose dependent as shown in Fig. 7. This suggests that Gαi is involved at least in transendothelial migration from blood space to interstitium and possibly also in transepithelial migration into the alveolar air space.
In an LPS-induced model of acute lung injury, the sequential migration of PMN into the different compartments of the lung was explored. Using a new and quantitative flow cytometry-based technique, we show that LPS inhalation induced a rapid PMN sequestration in the pulmonary vasculature. Migration into the lung interstitium was observed within 1 h after LPS exposure, while transepithelial migration was delayed. PTx sharply reduced migration into the interstitium and into the alveolar air space but did not affect vascular accumulation.
When antibody injection is used to identify intravascular neutrophils in the lung, the antibody is required to reach all PMN in this compartment. We found a complete GR-1 labeling of all PMN in the systemic circulation. Additionally, the accessibility of the pulmonary vasculature was successfully tested using a marker for erythrocytes.
Even within 5 min, the GR-1 antibody might leak and stain extravascular PMN. In our studies, GR-1+ PMN did not appear until 4 h after antibody injection in the BAL, indicating that the antibody did not reach alveolar air space. To test for antibody leakage into the interstitium, we injected a higher (10-fold) dose of GR-1 that should increase the amount of leaked antibody and lead to an increased fraction of labeled PMN. No such increase was observed in our experiments (data not shown). Finally, a possible endothelial leakage should increase as the LPS-induced alveolo-capillary damage proceeds over time, leading to a continuously rising overestimation of the intravascular PMN fraction. However, intravascular PMN concentration peaks at 4 h after LPS, while PMN concentration in the interstitium increases at this time (Fig. 4). Together these data suggest that antibody leakage into the interstitium did not play a major role in our study.
Alveolar PMN are removed by BAL. Inaccessible airways might exist, particularly in the inflamed lung, leaving (GR-1−) PMN in the air space. This might result in an overestimation of the PMN counts in the interstitial compartment. However, significant contribution of alveolar PMN to the fraction of GR-1− PMN would be reflected by an either constant or increased ratio between interstitial and total extravascular (interstitial + alveolar) PMN over time. In fact, the fraction of interstitial PMN in all extravascular PMN decreases over time (Fig. 4). In addition, PMN counts in the interstitium reach a plateau after 12 h, while the PMN concentration in the air space still arises. Therefore, this effect does not significantly contribute to the results.
Investigating PMN migration in the lung.
Although the lung offers a unique system to study cell migration, molecular mechanisms are still largely unknown.
Transendothelial migration from the vasculature into the lung interstitium was not measured in earlier studies. Indirect measurements, such as a drop in PMN blood counts or lung MPO activity, are not suitable to study this first important migration step. Current knowledge about the interaction between PMN and pulmonary endothelium derives mostly from in vitro studies using endothelial cell lines (25, 30). Attempts were recently made to mimic more realistically the alveolo-capillary barrier in an in vitro system (20). It remains to be shown whether this method will provide insight into pulmonary PMN migration.
Our flow cytometry-based approach is able to reflect many aspects of PMN trafficking in vivo, including the presence of a physiological marginated pool (9), accumulation after challenge, as well as the different migration steps into the alveolar air space and the LPS-induced release of PMN from the bone marrow (37, 46).
Kinetics of PMN trafficking.
Several studies addressed the kinetics of PMN trafficking in the lung. Most of them focused on the initial retention in the pulmonary capillaries. Mathematical models (18, 21), multiple indicator techniques (38), injection of labeled PMN (17), as well as isolated lung models have been developed to describe PMN retention in the lung. PMN sequestration in the pulmonary vasculature in response to various inflammatory stimuli, such as live bacteria (13), complement fragments (31), macrophage inflammatory protein-2 (16), or LPS (2), occurs rapidly within a few minutes. PMN migration into the lung was detected as early as 1–2 h after injection of C5-fragment or Escherichia coli as assessed by radiolabeled PMN (11, 15). PMN infiltration into the BAL has been shown to be delayed for up to 6 h (12, 13, 43, 44). We found evidence of PMN migration into the alveolar air space starting after 2 h with a peak between 12 and 24 h. At 4 h, vascular PMN accumulation reached its maximum, indicating that PMN recruitment from the peripheral circulation was balanced by migration into the interstitium and the alveolar air space at an equal rate. The interstitial space held a significant number of PMN during the migration process, indicating that this space functions as a discrete compartment after an inflammatory stimulus.
Leukocyte-endothelial interactions are essential for the PMN recruitment to the lung (34). The engagement of molecules required for PMN recruitment, such as adhesion molecules and chemokines, varies among different inflammation models. There is good evidence that distinct signals are required for PMN to migrate through the different barriers, and even a single mediator can affect the migration steps differentially. For instance, nitric oxide induces vascular PMN sequestration in a murine model of sepsis but attenuates migration into the alveolar air space (35). In our study, PTx was able to block both transendothelial and transepithelial migration. However, the vascular accumulation was largely unaffected, indicating that chemokine receptor signaling is not required for neutrophil arrest in the pulmonary circulation. It has been previously suggested that chemokines and adhesion molecules both contribute to PMN arrest in the systemic microcirculation (41). It remains to be shown whether this mechanism applies in the lung as well.
The delay of transepithelial PMN migration in PTx-treated mice supports the hypothesis that a distinct signal is required for PMN to advance from the lung interstitium into the alveolar air space. Interestingly, PMN crossing the epithelial barrier seem to be pivotal for inducing lung damage associated with an increase in mortality (23).
Our data establish the first quantitative method for monitoring neutrophil migration from blood to lung interstitium to alveolar air space. Vascular sequestration occurred immediately after LPS challenge, whereas transendothelial and transepithelial migration into the air space was delayed. In ALI, the lung interstitium holds a significant amount of PMN during the migration process. Distinguishing intravascular and interstitial PMN in vivo facilitates new opportunities to study the regulation of PMN migration in the lung.
This study was supported by Medical Faculty of the University of Tübingen Grant 1099-1-0 (to J. Reutershan) and by National Heart, Lung, and Blood Institute Grant HL-73361 (to J. Linden and K. Ley).
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