Tumor necrosis factor (TNF)-α is a key mediator of sepsis-associated multiorgan failure, including the acute respiratory distress syndrome. We examined the role of protein tyrosine phosphorylation in TNF-α-induced pulmonary vascular permeability. Postconfluent human lung microvascular and pulmonary artery endothelial cell (EC) monolayers exposed to human recombinant TNF-α displayed a dose- and time-dependent increase in transendothelial [14C]albumin flux in the absence of EC injury. TNF-α also increased tyrosine phosphorylation of EC proteins, and several substrates were identified as the zonula adherens proteins vascular endothelial (VE)-cadherin, and β-catenin, γ-catenin, and p120 catenin (p120ctn). Prior protein tyrosine kinase (PTK) inhibition protected against the TNF-α effect. TNF-α activated multiple PTKs, including src family PTKs. Prior PTK inhibition with the src-selective agents PP1 and PP2 each protected against ∼60% of the TNF-α-induced increment in [14C]albumin flux. PP2 also blocked TNF-α-induced tyrosine phosphorylation of VE-cadherin, γ-catenin, and p120ctn. To identify which src family kinase(s) was required for TNF-α-induced vascular permeability, small interfering RNA (siRNA) targeting each of the three src family PTKs expressed in human EC, c-src, fyn, and yes, were introduced into the barrier function assay. Only fyn siRNA protected against the TNF-α effect, whereas the c-src and yes siRNAs did not. These combined data suggest that TNF-α regulates the pulmonary vascular endothelial paracellular pathway, in part, through fyn activation.
- acute respiratory distress syndrome
- endothelial barrier function
- zonula adherens
tumor necrosis factor (TNF)-α is a multifunctional cytokine that participates in numerous biological processes and pathological states (47). One disease state that is mediated, in part, by TNF-α is the acute respiratory distress syndrome (ARDS) (49). ARDS involves pulmonary extravasation of fluid, macromolecules, and inflammatory cells into the bronchoalveolar compartment (for review see Ref. 51). In ARDS patients, TNF-α levels can be elevated in both the bloodstream (39) and the bronchoalveolar lavage fluid (28). Infusion of recombinant TNF-α into experimental animals reconstitutes pathological changes seen in these patients (18), whereas baboons pretreated with neutralizing anti-TNF-α antibodies are protected against sepsis-associated lethality (48).
Although TNF-α provokes the biosynthesis and release of multiple endogenous mediators that can directly/indirectly contribute to the endothelial cell (EC) response, it directly influences endothelial barrier function in vitro. TNF-α increases paracellular movement of macromolecules across EC monolayers in a dose- and time-dependent manner (7, 17–19). Although specific dose and time requirements for the TNF-α effect might vary in different EC systems, a stimulus-to-response lag time of ≥2 h is consistently observed (7, 17–19). However, a recent study reported an increase in albumin clearance within 0.5 h of TNF-α treatment (29). Prior protein synthesis inhibition fails to block the TNF-α effect (7, 19). TNF-α-induced endothelial barrier dysfunction is a reversible process and cannot be ascribed solely to EC injury (7, 17). TNF-α-induced loss of endothelial barrier function is temporally coincident with actin reorganization and intercellular gap formation (7). TNF-α increases the G-actin pool at the expense of the F-actin pool, which may be explained through F-actin disassembly or depolymerization (17). Finally, prior F-actin stabilization with phallicidin protects against TNF-α-induced F-actin disassembly and opening of the paracellular pathway (17). The EC-EC adherens junction (AJ), or zonula adherens (ZA), is tethered to the actin cytoskeleton (13), and disruption of this ZA-actin linkage decreases cadherin ectodomain homophilic adhesion (12). Any contribution of ZA reorganization to TNF-α-induced opening of the paracellular pathway is presently unknown.
The state of ZA assembly is regulated, in part, through tyrosine phosphorylation of proteins within the ZA multiprotein complex, including β-catenin, γ-catenin, and p120 catenin (p120ctn) (1, 22, 50) and cadherins themselves (1, 15, 22, 50). In epithelial cells, multiple stimuli increase tyrosine phosphorylation of ZA proteins, including src transformation and mitogenic growth factors (22, 23, 50, 52), thereby provoking ZA disassembly (1). When EC begin to form junctions during the early stages of confluence, vascular endothelial (VE)-cadherin is heavily phosphorylated on tyrosine; on maturation into tight, postconfluent monolayers, the tyrosine phosphorylation state of VE-cadherin is diminished (26). Our laboratory and others have shown that multiple agonists can induce tyrosine phosphorylation of ZA proteins, intracellular gap formation, and opening of the paracellular pathway. These agonists include both endogenous mediators such as thrombospondin (TSP)-1 (20), secreted protein acidic rich in cysteine (SPARC) (55), and histamine (2), as well as, exogenous factors such as bacterial lipopolysaccharide (LPS) (3) and Staphylococcus enterotoxin B (11). Proangiogenic factors such as vascular endothelial growth factor (VEGF) also increase tyrosine phosphorylation of VE-cadherin, β-catenin, γ-catenin, and p120ctn (15).
TNF-α is known to activate multiple downstream signaling events; little is known about its ability to induce protein tyrosine phosphorylation. We asked whether TNF-α might increase vascular endothelial permeability through protein tyrosine kinase (PTK)-driven tyrosine phosphorylation of ZA proteins. TNF-α activates multiple PTKs (33, 46), including members of the src family. The nonreceptor src PTKs (c-src, fyn, yes) have been immunolocalized to subcellular compartments in close proximity to the ZA. c-src is known to tyrosine phosphorylate ZA proteins (e.g., p120ctn) (44). Transformation of cells with v-src has been associated with increased tyrosine phosphorylation of the AJ, reduced cell-cell adhesion, and altered cell morphology (22, 50). Nwariakau et al. (31) recently demonstrated that the src-specific pharmacological inhibitor PP2 protects against TNF-α-induced tyrosine phosphorylation of VE-cadherin and opening of the paracellular pathway. Other studies implicate src family PTKs in VEGF-induced edema formation following stroke (32) and myocardial infarction (53). These studies implicate src family members as candidate PTKs operative during TNF-α-induced opening of the endothelial paracellular pathway.
In the present studies, we used human lung microvascular EC (HMVEC-L) and human pulmonary artery EC (HPAEC) to study the ability of TNF-α to open the pulmonary vascular endothelial paracellular pathway. Prior studies have suggested that protein tyrosine phosphorylation plays a key role in the regulation of endothelial barrier function (3, 11, 15, 20, 42, 54, 55). In the present studies, we demonstrate that TNF-α increases the tyrosine phosphorylation states of EC proteins and identify several substrates for TNF-α-induced tyrosine phosphorylation as ZA proteins. Furthermore, we show that prior PTK inhibition blocks TNF-α-induced tyrosine phosphorylation of ZA proteins and opening of the pulmonary vascular endothelial paracellular pathway. We show that TNF-α activates members of the src family PTKs. However, using a RNA interference strategy, we demonstrate that activation of the src family PTK fyn is required for TNF-α-induced barrier dysfunction.
MATERIALS AND METHODS
Recombinant human TNF-α, with a specific activity of ≥2 × 107 U/mg, was obtained from Endogen (Woburn, MA). Murine anti-human TNF-α antibody and a species- and isotype-matched irrelevant antibody control were purchased from R&D Systems (Minneapolis, MN). Horseradish peroxidase (HRP)-conjugated anti-phosphotyrosine MAb cocktail (PY+-HRP) was purchased from Zymed Laboratories (South San Francisco, CA). Goat anti-human CD31 or platelet endothelial cell adhesion molecule-1 (PECAM-1) IgG, the anti-src rabbit IgG, the anti-fyn mouse IgG, the anti-VE-cadherin goat IgG, and HRP-conjugated donkey anti-goat IgG antibody were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Specific murine IgG antibodies raised against α-, β-, and γ-catenins and p120ctn as well as HRP-conjugated-anti-mouse IgG and HRP-conjugated-anti-rabbit IgG were all purchased from BD Transduction Labs (Lexington, KY). The anti-VE-cadherin (30q8A) murine IgG antibody was provided by ICOS (Bothel, WA). The rabbit anti-human phospho-src (Y416) IgG and the fluorescein isothiocyanate (FITC)-labeled murine anti-phosphotyrosine antibody were obtained from Cell Signaling Technologies (Beverly, MA). The fluorescein-conjugated phalloidin, the FITC-conjugated goat anti-mouse IgG antibody, and the FITC-conjugated donkey anti-goat IgG antibody were purchased from Molecular Probes (Eugene, OR). The PTK inhibitors geldanamycin and PP2 were purchased from Calbiochem (La Jolla, CA). PP1 was purchased from BIOMOL (Plymouth Meeting, PA). The anti-src mouse IgG, the anti-yes rabbit IgG, and the src kinase assay kit were purchased from Upstate Biotechnology (Lake Placid, NY). 14C-labeled bovine serum albumin (BSA), herbimycin A (Herb A), and genistein were purchased from Sigma (St. Louis, MO).
EC tissue culture.
HMVEC-L and HPAEC (Cambrex; Walkersville, MD) were cultured in EC basal medium (EBM-2, Cambrex) supplemented with 5% fetal bovine serum (FBS), human epidermal growth factor (EGF), VEGF, human fibroblast growth factor (with heparin), long R3 insulin-like growth factor I, hydrocortisone, ascorbic acid, gentamicin, and amphotericin B (Cambrex BulletKit, CC-3202) as described previously (42). Only EC from passages 5–10 were studied.
Endothelial barrier function assay.
The endothelial barrier function assay was performed as described previously (17, 19). EC (2 × 105 cells/chamber) were cultured for 72 h on gelatin-impregnated polycarbonate filters (13-mm diameter, 0.4-μm pore size; Nucleopore, Pleasanton, CA) mounted in polystyrene chemotactic chambers (ADAPS, Dedham, MA) inserted into the wells of 24-well plates. The initial baseline barrier function of each monolayer was determined by applying an equivalent and reproducible amount of tracer molecule, 14C-labeled BSA ([14C]BSA; 1.1 pmol/0.5 ml, specific activity 89 μCi/mg protein; Sigma), to each upper compartment for 1 h at 37°C, after which a sample was taken from the lower compartment and counted for 14C activity. Only EC monolayers retaining ≥97% of the tracer were studied. The monolayers were then exposed to TNF-α-enriched medium or medium alone and again assayed for transendothelial [14C]BSA flux. To ensure that the increases in [14C]BSA flux were due solely to TNF-α and not to one or more contaminants, TNF-α was preincubated with a neutralizing anti-TNF-α antibody (IgG). EC monolayers were exposed for 6 h to TNF-α (100 U/ml, 5 ng/ml), TNF-α preincubated for 1 h at room temperature with anti-TNF-α MAb (2 μg/ml), TNF-α preincubated with isotype-matched irrelevant antibody, anti-TNF-α MAb, or irrelevant antibody alone and assayed for [14C]BSA flux as described above. To determine whether TNF-α-induced barrier dysfunction was PTK dependent, HPAEC monolayers were preincubated for varying times with the following broad-spectrum PTK inhibitors: 1.0 μM Herb A (16 h), 185 μM genistein (2 h), or 0.5 μM geldanamycin (2 h). For the more specific PTK inhibitors, EC monolayers were pretreated with 10 μM PP1 (1 h) or 5 μM PP2 (1 h). The preincubation was followed by 6-h exposure to 100 U/ml TNF-α in the presence of identical concentrations of the PTK inhibitors. Equivalent concentrations of dimethyl sulfoxide (0.1%) were present in all samples including the simultaneous inhibitor and medium controls.
Epifluorescence microscopy for F-actin organization and immunolocalization of ZA and phosphotyrosine-containing proteins.
To maintain the same experimental conditions as our permeability assay, HMVEC-L were grown to postconfluence and stained directly on polycarbonate filters as previously described (17, 55). For phosphotyrosine immunolocalization, EC were serum and growth factor starved overnight, after which they were exposed for 6 h to TNF-α (100 U/ml) or medium alone. The monolayers were fixed with 4% paraformaldehyde and permeabilized with 0.5% Triton X-100 in PBS. For F-actin visualization, EC monolayers were stained with fluorescein-phalloidin (1.65 × 10−7 M, 20 min) as previously described (17, 55). For immunolocalization of phosphotyrosine and ZA proteins, EC were blocked with 1% BSA in PBS for 0.5 h, incubated with FITC-labeled anti-phosphotyrosine, anti-VE-cadherin, anti-β-catenin, anti-γ-catenin, or anti-p120ctn antibody for 1.5 h, washed with PBS, and then incubated for 1 h with either FITC-conjugated goat anti-mouse IgG or donkey anti-goat IgG. The filters were mounted cell side up on glass microscope slides and visualized through a Zeiss Axioscop 20 microscope (×100 oil, 1.3 numerical aperture Plan Neofluor objective), and photographs were captured with a Spot-RT digital camera (Diagnostics, Sterling Heights, MI).
EC tube formation assay.
HMVEC-L were cultured at 4 × 105 EC/3 ml in each well of a Matrigel (BD Discovery Labware; Bedford, MA)-coated six-well plate and incubated for 16 h to allow for EC capillary tube formation as described previously (54). The preformed tubes were incubated for 6 h with TNF-α (100 U/ml) or medium alone, after which they were visualized through an inverted Zeiss microscope and photographed with a digital camera.
Detection of apoptosis/EC injury.
To determine whether TNF-α-induced barrier dysfunction might be explained through EC apoptosis, TNF-α-exposed EC were analyzed with DNA laddering and terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling (TUNEL) assays as previously described (5). EC monolayers were cultured to confluence and treated with TNF-α (100 U/ml) or medium alone for 16 h. In other studies, a Trypan blue exclusion assay was used to determine whether TNF-α induced a loss of plasma membrane integrity and cell death.
HMVEC-L monolayers were serum and growth factor starved overnight, after which they were exposed for various times to varying concentrations of TNF-α or to medium alone. During the final 20 min of each treatment period, protein tyrosine phosphatase (PTP) inhibition with sodium orthovanadate (200 μM) and phenylarsine oxide (PAO) (1.0 μM) was introduced as previously described (20). EC were then lysed with ice-cold modified radioimmunoprecipitation assay buffer containing 50 mM Tris·HCl, pH 7.4, 1% Nonidet P-40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EGTA, 100 mg/ml type 1 DNase, 1 mM vanadate, 1 mM NaF, 10 mM pyrophosphate, 1 mM PAO, 1 mg/ml pepstatin A (all purchased from Sigma), and 1 tablet of Complete Protease Inhibitor Cocktail per 20 ml (Roche Molecular Biochemicals, Indianapolis, IN). The lysates were centrifuged, and the supernatants were assayed for protein concentration with a standard Bio-Rad Protein assay kit (Bio-Rad Chemical Division). The samples were resolved by either 6% or 8–16% gradient sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE; Novex, San Diego, CA) and transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore, Bedford, MA). The blot was blocked with 2% BSA-PBS and incubated with an HRP-conjugated anti-phosphotyrosine MAb cocktail. The blot was developed with enhanced chemiluminescence (ECL) and exposed to X-ray film (Kodak, Rochester, NY). To ensure equal protein loading, blots were stripped with 100 mM 2-mercaptoethanol, 2% SDS, and 62.5 mM Tris·HCl, pH 6.7, and reprobed with goat anti-PECAM-1 IgG, followed by HRP-conjugated donkey anti-goat IgG, and again developed with ECL. PECAM-1 is constitutively expressed in EC (30), and its level of expression is not influenced by TNF-α alone (38). Phosphotyrosine-containing bands of interest were quantified by laser densitometry (Molecular Dynamics, Sunnydale, CA) and normalized to PECAM-1 as a housekeeping gene product to ensure equal loading and transfer.
EC lysates were precleared by incubation with anti-murine IgG cross-linked to agarose (Sigma) for 1 h at 4°C and then incubated overnight at 4°C with specific murine MAbs raised against VE-cadherin, α-, β-, and γ-catenins, and p120ctn. The resultant immune complexes were immobilized by incubation with anti-murine IgG cross-linked to agarose, centrifuged, washed, boiled for 5 min in sample buffer, and again centrifuged. The supernatants were processed for phosphotyrosine immunoblotting as described above. To control for protein immunoprecipitation, loading, and transfer, blots were stripped and reprobed with the immunoprecipitating antibody. Bands were quantified by laser densitometry.
In-gel kinase assay.
Postconfluent HPAEC monolayers were serum and growth factor starved overnight and then exposed to TNF-α (100 U/ml) or medium alone for increasing time periods up to 20 min, after which they were lysed in kinase lysis buffer containing 20 mM Tris·HCl (pH 8.0), 5 mM MgCl2, 10 mM EGTA, 1% Triton X-100, 50 mM NaF, 100 μM sodium orthovanadate, 2 nM calyculin A, 10 nM okadaic acid, and protease inhibitors (1 complete tablet per 10 ml) as described previously (40). The EC lysates were resolved by a 10% SDS-PAGE containing 200 μg/ml poly-Glu/Tyr (4:1; Sigma). Once electrophoresis was complete, the SDS was removed by a 1-h wash in 20% isopropanol in 50 mM Tris·HCl (pH 8.0), followed by a second 1-h wash in 50 mM Tris·HCl (pH 8.0) containing 1 mM DTT. The proteins were then denatured with a 1-h wash containing 6 M guanidine-HCl, 20 mM DTT, and 2 mM EDTA in 50 mM Tris·HCl (pH 8.0) and then renatured overnight with a wash containing 1 mM DTT, 2 mM EDTA, and 0.04% Tween 20 in 50 mM Tris·HCl at 4°C. The gel was then incubated for 1 h in kinase buffer containing 40 mM HEPES (pH 8.0), 1 mM DTT, 0.1 mM EGTA, 20 mM MgCl2, and 100 μM sodium orthovanadate, followed by a 1-h incubation in kinase buffer with 30 μM ATP and 10 μCi/ml [γ-32P]ATP and then three to five washes with 5% trichloroacetic acid containing 1% sodium pyrophosphate. The gel was then dried and exposed to X-ray film.
Detection of src kinase activity.
Postconfluent HPAEC monolayers were serum and growth factor starved overnight and then exposed to TNF-α (100 U/ml) or medium alone for increasing time periods up to 6 h. EC were lysed in a lysis buffer containing 50 mM Tris·HCl, pH 7.4, 1% Nonidet-P40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA, 1 mM vanadate, 1 mM NaF, 1 μg/ml pepstatin A, and two Complete Protease Inhibitor tablets. The EC lysates were resolved by SDS-PAGE and transferred to PVDF. The membranes were blocked with 5% milk, washed, and incubated with anti-phospho-src (Y416) overnight at 4°C. The membranes were then incubated with anti-rabbit IgG-HRP and developed with ECL. To ensure equal loading, membranes were stripped and reprobed with the anti-human src polyclonal antibody. The immunoreactive bands were then quantified with laser densitometry. The in vitro src kinase assay was performed according to the manufacturer’s protocol. HPAEC lysates were immunoprecipitated with the anti-human src polyclonal antibody. This anti-src polyclonal is raised against the highly conserved COOH terminus and has been recognized to react with human c-src, fyn, and yes. The immunoprecipitates were then resuspended in kinase assay buffer (50 mM HEPES, pH 7.5, 0.1 mM EDTA, 0.015% polyoxyethylene glycol dodecyl ether). The final reaction occurred in kinase assay buffer along with 15 mM MgCl2, 1 mM vanadate, 150 μM ATP, and 33 μCi of [γ-32P]ATP with and without 150 μM synthetic src substrate (KVEKIGEGTYGVVYK) for 15 min at 30°C. The reaction was terminated with 1% phosphoric acid. The samples were then spotted on P81 filter paper and washed four times with 1% phosphoric acid followed by a wash with methanol. The 32P activity retained by the filter paper was counted in a liquid scintillation counter (Packard Instrument; Downers Grove, IL).
Reverse transcriptase-polymerase chain reaction for c-Src, Fyn, and Yes in HMVEC-L.
Total RNA was isolated from HMVEC-L with Absolutely RNA Miniprep Kit (Stratagene, La Jolla, CA). Complementary DNA was generated from RNA with oligo(dT)12–18 primers and SuperScript III reverse transcriptase (RT) (Invitrogen, Carlsbad, CA). The oligonucleotide primers used for polymerase chain reaction (PCR) were human Fyn: sense 5′-CTAGGGATCCGTACTGGAGAGACAGGTTAC-3′, antisense 5′-CGTGAATTCTGTGGGGTACAACTCGATGC-3′; human c-Src: sense 5′-CTAGGGATCCAGGCTGAGGAGTGGTATT-TTG-3′, antisense 5′-CGTGAATTCTTGGACGTGGGGCACACGG-3′; human Yes: sense 5′-TCCATTCAGGCAGAAGAATGG-3′, antisense 5′-CCAAGCATCTTTTGCTAGACC-3′. PCR was performed in a Thermo Px2 Thermal cycler. After an initial 2 min of denaturing at 94°C, 30 cycles consisting of 30 s at 94°C (denaturation), 1 min at 60°C (annealing), and 1 min at 72°C (extension) were completed, followed by a 7-min final extension at 72°C.
Knockdown of src family kinases through RNA interference.
SMARTPool small interfering RNA (siRNA) duplex products designed to target c-src, fyn, and yes as well as the appropriate control siRNA duplexes were purchased from Dharmacon (Lafayette, CO). The c-src, fyn, yes, and control siRNAs were preincubated with TransMessenger transfection reagent (Qiagen, Valencia, CA) according to the manufacturer’s protocol, and the transfection complexes were presented to HMVEC-L cultured to ∼100% confluence for 3 h in the absence of serum. At 72 h after transfection, EC were lysed and processed for immunoblotting with anti-c-src, anti-fyn, or anti-yes antibodies. To confirm equivalent protein loading, blots were stripped and reprobed with anti-PECAM-1 antibodies and developed with ECL as described above. Blots were scanned by laser densitometry, and the src-, fyn-, and yes-immunoreactive bands were quantified. Once knockdown had been established, the siRNAs were introduced into the barrier function assay. HMVEC-L were cultured for 72 h as stated above. An initial baseline barrier function was taken, and only EC monolayers that retained at least 97% of the tracer molecule were transfected with c-src, fyn, yes, or control siRNA or medium alone as stated above. After 72 h, baselines were reevaluated with the same standards. Tight monolayers were exposed to TNF-α (100 U/ml) or medium alone for 6 h and then assayed for transendothelial [14C]albumin flux.
ANOVA was used to compare the mean responses among experimental and control groups for all experiments. The Dunnett and Scheffé F-test was used to determine between which groups significant differences existed. A P value of <0.05 was considered significant.
TNF-α increases transendothelial [14C]BSA flux.
At 6 h, TNF-α induced a dose-dependent increase in [14C]BSA flux across HMVEC-L monolayers (Fig. 1A). The minimum TNF-α concentration that increased transendothelial flux compared with the medium control was 0.5 U/ml (0.025 ng/ml). The maximal increase in albumin flux was seen with TNF-α at 100 U/ml (5 ng/ml), after which the EC response appeared to plateau. TNF-α also induced a dose-dependent increase in albumin flux across HPAEC monolayers (Fig. 1B). Here, the minimum suprathreshold dose was 20-fold higher than that seen with HMVEC-L. The TNF-α effect was also time dependent (Fig. 1C). HMVEC-L monolayers were exposed to two fixed concentrations of TNF-α (10 and 100 U/ml) for increasing exposure times (0.5–6 h). At 100 and 10 U/ml, TNF-α exposure was associated with a prolonged stimulus-to-response lag time of 3 and 4 h, respectively. To exclude any contribution of endotoxin and/or other contaminants to the increases in [14C]BSA flux, 100 U/ml TNF-α was preincubated with murine anti-human TNF-α neutralizing antibody for 1 h at room temperature. TNF-α (100 U/ml) preincubated with anti-TNF-α antibody failed to increase transendothelial [14C]BSA flux compared with the medium control and was associated with flux that was markedly decreased compared with TNF-α (100 U/ml) alone (Fig. 1D). Preincubation of TNF-α with a species- and isotype-matched irrelevant antibody control did not alter the TNF-α effect. Incubation of the EC monolayer with either the anti-TNF-α antibody or the irrelevant antibody alone did not alter [14C]BSA flux. Therefore, the TNF-α-induced changes in barrier function could be ascribed solely to TNF-α and not to endotoxin or other contaminants.
TNF-α induces intercellular gap formation.
Postconfluent HMVEC-L monolayers cultured under conditions identical to those in the barrier function studies were exposed for 6 h to TNF-α (100 U/ml) or medium alone. The monolayers were then fixed, permeabilized, stained with fluorescein-phalloidin, and examined by epifluorescence microscopy. The medium control monolayers contained continuous transcytoplasmic actin microfilaments and tight cell-to-cell apposition without intercellular gaps (data not shown). TNF-α-exposed EC monolayers displayed actin reorganization and intercellular gaps (data not shown). In the EC capillary tube experiments, HMVEC-L were plated on Matrigel and allowed to attach to the matrix; EC tubes formed within 18 h. The preformed tubes were treated for 6 h with TNF-α (100 U/ml) or medium alone. Medium control tubes were continuous (Fig. 2A), whereas the TNF-α-exposed EC tubes displayed gaps between neighboring cells (indicated by arrows in Fig. 2B), with no loss of attachment to the underlying matrix.
Exclusion of TNF-α-induced EC death.
HMVEC-L monolayers were exposed to TNF-α (100 U/ml), at a concentration that induces the maximal increase in albumin flux, for a period of 16 h, which well exceeds the 6 h used in the permeability assays. TNF-α did not induce apoptosis as seen by DNA laddering or TUNEL staining (Fig. 2, C–F), nor did TNF-α (100 U/ml, 6 h) induce loss of EC viability compared with the simultaneous medium control as demonstrated by Trypan blue exclusion [TNF-α 95.55 ± 0.84% viability (n = 11) vs. medium control 96.35 ± 0.82% viability (n = 11); P = 0.5064]. These results indicate that TNF-α does not induce EC apoptosis or death in HMVEC-L and that TNF-α-induced barrier dysfunction cannot be explained through changes in EC viability.
TNF-α increases tyrosine phosphorylation of ZA proteins.
Postconfluent HMVEC-L monolayers were serum and growth factor starved overnight, after which they were exposed for 6 h to increasing concentrations of TNF-α or medium alone and processed for phosphotyrosine immunoblotting. To optimize the phosphotyrosine signal, PTP inhibitors were added 20 min before EC lysis. TNF-α increased tyrosine phosphorylation of EC proteins in a dose-dependent manner (Fig. 3A). Phosphotyrosine-containing bands were seen with approximate apparent molecular masses ranging from 66 to 260 kDa. More specifically, bands that migrated with gel mobilities indicative of molecular masses of ∼180, 130, 120, 92, 88, and 66 kDa displayed increased phosphotyrosine signal. TNF-α at a concentration of 100 U/ml (5 ng/ml) clearly increased protein tyrosine phosphorylation compared with the medium control. The TNF-α-induced increase in protein tyrosine phosphorylation was also time dependent (Fig. 3B). TNF-α at 100 U/ml (5 ng/ml) increased tyrosine phosphorylation as early as 2 h (mol mass ∼88 kDa), with further increases at 4 h (mol mass ∼130, 122, 92, and 66 kDa) and 6 h (mol mass ∼260 and 180 kDa) (Fig. 3B). A phosphotyrosine-containing band was detected at ∼260 kDa, but its appearance was not consistently reproducible on an 8–16% gradient gel. On a 6% gel, this phosphotyrosine-containing band was clearly resolved (Fig. 3C). Of interest, several of these phosphotyrosine-containing bands migrated with apparent molecular masses compatible with those of ZA proteins (130, 120, 92, and 88 kDa).
To determine subcellular localization of substrates for TNF-α-induced tyrosine phosphorylation, serum- and growth factor-starved postconfluent HMVEC-L monolayers were exposed for 6 h to TNF-α (100 U/ml) or medium alone and processed for phosphotyrosine immunofluorescence microscopy (Fig. 4, A and B). TNF-α-exposed EC monolayers displayed enhanced phosphotyrosine staining at intercellular boundaries (Fig. 4B) compared with the medium control monolayers (Fig. 4A). These findings indicate that TNF-α preferentially stimulates tyrosine phosphorylation of proteins that are either enriched to or upon phosphorylation translocate to cell-cell junctions in postconfluent EC monolayers.
Phosphotyrosine immunoblotting of TNF-α-exposed postconfluent HMVEC-L monolayers displayed increased protein tyrosine phosphorylation compared with simultaneous medium controls (Fig. 3, A–C), and these proteins were enriched to the intercellular boundaries (Fig. 4A). To determine whether substrates for TNF-α-induced tyrosine phosphorylation might include ZA components, we used specific antibodies to screen for the ZA proteins VE-cadherin, α-catenin, β-catenin, γ-catenin, and p120ctn. There was an increase in tyrosine phosphorylation of ZA proteins VE-cadherin, β-catenin, γ-catenin, and p120ctn in TNF-α-treated EC compared with the simultaneous medium controls (Fig. 4C). TNF-α increased the tyrosine phosphorylation of VE-cadherin ∼21.2-fold, while increasing the tyrosine phosphorylation of β-catenin, γ-catenin, and p120ctn ∼6.6-, ∼2.3-, and ∼1.5-fold, respectively. There was no detectable change in α-catenin tyrosine phosphorylation. Immunoblots were stripped and reprobed with the immunoprecipitating antibody to ensure comparable immunoprecipitation efficiency, protein loading, and transfer (Fig. 4C, bottom).
Effect of TNF-α on immunolocalization of ZA proteins.
Postconfluent HMVEC-L monolayers cultured on polycarbonate filters were exposed for 6 h to 100 U/ml TNF-α or medium alone. The monolayers were fixed, permeabilized, and probed with antibodies to the ZA proteins VE-cadherin, β-catenin, γ-catenin, and p120ctn. Medium control EC monolayers displayed continuous staining for VE-cadherin (Fig. 5A), β-catenin (Fig. 5C), γ-catenin (Fig. 5E), and p120ctn (Fig. 5G) restricted to the intercellular boundaries. In contrast, TNF-α-exposed EC displayed discontinuous staining with the complete absence of signal in areas of intercellular gaps indicated by arrows in Fig. 5, B, D, F, and H. Loss of VE-cadherin, β-catenin, γ-catenin, and p120ctn staining from the EC-EC boundary was not coincident with any detectable changes in cytoplasmic signal for these same proteins.
Effect of PTK inhibition on TNF-α induced changes in transendothelial [14C]BSA flux.
TNF-α (100 U/ml, 6 h)-treated HPAEC monolayers exhibited greater than twofold increase in [14C]BSA flux compared with the simultaneous medium controls (Fig. 1B). Preincubation of these EC monolayers with the broad-spectrum PTK inhibitors genistein, Herb A, and geldanamycin each completely blocked the TNF-α-induced increase in albumin flux, reducing it to below basal levels (Fig. 6A). Each of these three PTK inhibitors successfully gained entry into HPAEC, where they diminished protein tyrosine phosphorylation (Fig. 6B). Further analysis using laser densitometry revealed that the broad-spectrum PTK inhibitors genistein, Herb A, and geldanamycin protected against ∼100% of TNF-α-induced tyrosine phosphorylation of EC proteins. Therefore, the TNF-α-induced increase in transendothelial [14C]BSA flux was PTK dependent.
Role of src family PTKs in TNF-α-induced barrier dysfunction.
Postconfluent HPAEC were serum and growth factor starved overnight, exposed to TNF-α (100 U/ml) for increasing time points up to 20 min, and then lysed and processed for the in-gel kinase assay. Although multiple bands appeared, one prominent band appeared with an apparent Mr of ∼60,000 evident as early as 5 min and persisting throughout 20 min post-TNF-α treatment (Fig. 7A).
Since the in-gel kinase assay indicated PTK activity in response to TNF-α that migrated with a Mr of ∼60,000, we examined whether TNF-α activates one or more src family PTKs. Postconfluent HPAEC were serum and growth factor starved overnight, exposed to TNF-α (100 U/ml) or medium alone for increasing time periods, and lysed. EC lysates were resolved by SDS-PAGE and transferred to PVDF, and the blots were probed with anti-phospho-src (Y416) antibodies. Phosphorylation of Y416 indicates src activation (43). Figure 7B is a representative blot of two to four experiments, and laser densitometry was used to quantify the results in Fig. 7C. TNF-α increased src Y416 phosphorylation at 10 and 60 min compared with simultaneous medium controls (Fig. 7C). After demonstrating that TNF-α increases src family PTK Y416 phosphorylation, we asked whether TNF-α also increases src PTK activity for a defined peptide substrate. HPAEC were exposed to TNF-α or medium as stated above, and EC lysates were immunoprecipitated with a polyclonal anti-pan-src antibody raised against the conserved portion that recognizes the three members of the src family that are reportedly expressed in EC, c-src, fyn, and yes (8). The src immunoprecipitates were incubated with [32P]ATP and peptide in an in vitro kinase assay. The src immunoprecipitates from TNF-α-treated HPAEC increased the phosphorylation of the peptide substrate ∼1.4- to 1.5-fold at both 10 and 60 min after TNF-α exposure compared with src immunoprecipitates obtained from the medium control EC (Fig. 7D). The patterns of src activation over 60 min measured in these two experimental systems (Fig. 7, C and D) were superimposable. These combined data demonstrate that, in human EC, TNF-α activates one or more src family kinases in a temporally restricted biphasic manner.
In one previous study TNF-α was shown to activate src family kinases (46), and these same PTKs are known to phosphorylate ZA proteins in epithelia (43, 44). As a first step in examining the potential role of src family kinases in TNF-α-induced barrier dysfunction, two distinct src inhibitors (PP1 and PP2) were introduced into the barrier function assay. Prior src inhibition blocked ∼55–60% of TNF-α-induced increments in transendothelial albumin flux in HPAEC (Fig. 8A). These data indicate that TNF-α-induced loss of barrier function is mediated, in part, through activation of one or more src family PTKs.
TNF-α induces src-dependent phosphorylation of ZA proteins.
TNF-α increases both src PTK activation (Fig. 7) and tyrosine phosphorylation of ZA proteins (Fig. 4C). To determine whether these two TNF-α-induced EC responses might be causally related, EC were exposed to TNF-α in the presence and absence of selective src PTK inhibition. HMVEC-L were serum and growth factor starved overnight and then exposed to medium alone, TNF-α, or TNF-α and PP2 for a period of 6 h, with the last 20 min in the presence of PTP inhibitors. The EC were lysed and immunoprecipitated with antibodies against VE-cadherin, β-catenin, γ-catenin, and p120ctn. The immunoprecipitates were processed for phosphotyrosine immunoblotting. Prior src inhibition blocked TNF-α-induced tyrosine phosphorylation of VE-cadherin, γ-catenin, and p120ctn but failed to block β-catenin tyrosine phosphorylation (Fig. 8B). More specifically, laser densitometry revealed that addition of PP2 reduced TNF-α-induced tyrosine phosphorylation of VE-cadherin by ∼96%, γ-catenin by ∼100%, and p120ctn by ∼67%. These studies suggest that one or more src family kinases is operative in TNF-α-induced tyrosine phosphorylation of VE-cadherin, γ-catenin, and p120ctn, but not β-catenin.
TNF-α increases paracellular permeability and VE-cadherin tyrosine phosphorylation through fyn activation.
Postconfluent HMVEC-L were examined for mRNA expression of the three src family members previously described in EC with RT-PCR (8). mRNA for c-src, fyn, and yes were detected (Fig. 9A). To determine which src family kinase(s) was operative in TNF-α-induced endothelial barrier dysfunction, postconfluent HMVEC-L were transfected with c-src, fyn, yes, or control siRNA. After 72 h, c-src, fyn, and yes proteins each were knocked down >94%, >99%, and >75%, respectively, compared with the simultaneous controls (Fig. 9B). There was no apparent crossover knockdown of nontargeted src family PTKs (Fig. 9B). TNF-α increased albumin flux across HMVEC-L monolayers transfected with control siRNA by ∼3.5-fold (Fig. 9C). This situation was true for both c-src and yes siRNA-transfected EC. Prior transfection with fyn siRNA reduced TNF-α-induced albumin flux by ∼75%. These data indicate that fyn is the operative src family PTK in TNF-α-induced opening of the endothelial paracellular pathway. Pharmacological inhibition of src blocks TNF-α-induced tyrosine phosphorylation of the ZA proteins VE-cadherin, γ-catenin, and p120ctn (Fig. 8B). To examine the role of fyn in TNF-α-induced tyrosine phosphorylation of ZA proteins, we used siRNA directed toward fyn followed by an immunoprecipitation strategy. Knockdown of fyn blocks TNF-α-induced tyrosine phosphorylation of VE-cadherin (Fig. 9D). Quantitation of this band by laser densitometry demonstrates that introduction of fyn siRNA reduces TNF-α-induced tyrosine phosphorylation of VE-cadherin by ∼95%. Experiments involving γ-catenin and p120ctn were inconclusive (data not shown).
In the present report, we have demonstrated that TNF-α activates PTKs in human pulmonary vascular EC, including one or more src family PTKs. TNF-α increased tyrosine phosphorylation of EC proteins in a dose- and time-dependent manner. These phosphotyrosine-containing proteins were immunolocalized to the intercellular boundaries, and several of these substrates were identified as the ZA proteins VE-cadherin, β-catenin, γ-catenin, and p120ctn. We determined that TNF-α reduced cell-cell association leading to intercellular gap formation in both EC monolayers and preformed EC capillary tubes. This reduction in cell-cell association was associated with changes in the actin cytoskeleton and loss of ZA protein staining within intercellular gaps. TNF-α increased transendothelial [14C]BSA flux in a dose- and time-dependent manner in the absence of apparent EC injury and/or apoptosis. Prior broad-spectrum PTK inhibition attenuated TNF-α-induced increments in albumin flux. Prior addition of the src-selective inhibitors PP1 and PP2 each protected against ∼60% of the TNF-α-induced loss of barrier function. Prior src inhibition also diminished TNF-α-induced increases in ZA protein tyrosine phosphorylation. Using a siRNA strategy, we were able to identify the src family member fyn as the PTK essential to TNF-α-induced barrier dysfunction. These combined data indicate that TNF-α opens the pulmonary vascular endothelial paracellular pathway, in part, through fyn-dependent tyrosine phosphorylation of one or more EC proteins, including VE-cadherin.
TNF-α increased [14C]BSA flux across pulmonary vascular endothelia derived from both large (HPAEC)- and small (HMVEC-L)-caliber vessels in a dose-dependent manner. In our barrier function assay system, HMVEC-L were more sensitive to TNF-α than HPAEC. In HMVEC-L TNF-α concentrations as low as 0.5 U/ml (0.025 ng/ml) increased transendothelial albumin flux, whereas in HPAEC a 20-fold higher dose was required (10 U/ml, 0.5 ng/ml). In HMVEC-L increments in albumin flux appeared to plateau at a TNF-α concentration >100 U/ml (5 ng/ml), whereas in HPAEC the albumin flux continued to increase through the highest concentration tested (1000 U/ml, 50 ng/ml). These findings are comparable to those reported for the same EC types exposed to LPS (4). To the best of our knowledge, this is the first report to establish the TNF-α concentration requirements for physiologically relevant HMVEC-L.
TNF-α induces protein tyrosine phosphorylation in EC (31). The exact role that these tyrosine phosphorylation events play in barrier regulation is unclear. In the present study, TNF-α increased tyrosine phosphorylation of EC proteins in a dose- and time-dependent manner (Fig. 3). These tyrosine phosphorylation events occur ∼1 h before loss of barrier function. The increase in protein tyrosine phosphorylation is temporally coincident with the TNF-α-induced actin reorganization and intercellular gap formation described in bovine pulmonary artery endothelial cells (17). The minimum TNF-α dose required for increased tyrosine phosphorylation of HMVEC-L was 100 U/ml, whereas the minimum dose required for altering barrier function was 0.5 U/ml. The dose requirements for increased tyrosine phosphorylation and increased barrier dysfunction may differ because of the differential sensitivities between phosphotyrosine immunoblotting and the barrier function assay. This is further complicated by the requirement of PTP inhibition for the phosphotyrosine immunoblotting studies. In the present study, the phosphotyrosine signal was localized to the intercellular boundaries (Fig. 4B). Several other established mediators of increased endothelial paracellular permeability also increase tyrosine phosphorylation of proteins enriched to the EC-EC boundaries. These agonists include inflammatory mediators (2), proangiogenic factors (15), counteradhesive proteins (20, 55), and bacterial products (3, 11).
Tyrosine phosphorylation of ZA proteins is known to regulate cadherin ectodomain homophilic adhesion; increased ZA protein tyrosine phosphorylation decreases such adhesion (22, 23, 50). TNF-α increases the tyrosine phosphorylation state of multiple EC proteins with an apparent Mr of ∼66,000–260,000 (Fig. 3). Four substrates have been identified as the ZA proteins VE-cadherin, β-catenin, γ-catenin, and p120ctn (Fig. 4C). This is the first demonstration that TNF-α increases tyrosine phosphorylation of ZA proteins in HMVEC-L. Other established mediators of vascular permeability such as histamine (2), VEGF (15), complement-activated neutrophils (45), TSP-1 (20), and SPARC (55) all increase tyrosine phosphorylation of ZA proteins. It is conceivable that tyrosine phosphorylation of ZA proteins may provide a final common pathway through which a number of barrier-altering agents regulate endothelial barrier function. Under basal postconfluent conditions, the tyrosine phosphorylation state of ZA proteins is restrained and these proteins are largely sequestered at the intercellular boundaries (Fig. 5, A, C, E, G). After exposure to TNF-α, ZA protein tyrosine phosphorylation is increased and these same proteins depart from areas of intercellular gap formation (Fig. 5, B, D, F, H). These results are in agreement with a previous study in human umbilical vein endothelial cells (HUVEC) in which TNF-α induced redistribution of VE-cadherin from the cell periphery (31). Several studies have associated tyrosine phosphorylation of ZA proteins with ZA disassembly and disruption of the ZA-actin cytoskeletal linkage. In one epithelial system, the engagement of EGF with its receptor, EGFR, induced tyrosine phosphorylation of ZA proteins and ZA disassembly (23). In our system, after tyrosine hyperphosphorylation of ZA proteins, disassociation of VE-cadherin from its catenin-binding partners could not be detected in either coimmunoprecipitation or glutathione S-transferase-VE-cadherin binding assays (54). It is possible that VE-cadherin and the catenins leave the intercellular boundary in a complex, permitting a more rapid and dynamic response to environmental demands.
Increases in EC protein tyrosine phosphorylation can be explained by increases in PTK activity and/or decreases in PTP activity. In EC TNF-α can modulate PTKs (33, 46) and PTPs (21), and both have been implicated in barrier regulation (31, 42, 54). In the present study, broad-spectrum PTK inhibition reduced both TNF-α-induced protein tyrosine phosphorylation (Fig. 6B) and endothelial barrier dysfunction (Fig. 6A). Actin reorganization, a prerequisite event for TNF-α-induced endothelial barrier dysfunction (17), has also been shown to be PTK dependent in response to other barrier-altering agents (3, 11, 20, 55). These data, together with one previous study (16), implicate the action of one or more PTKs during agonist-driven opening of the endothelial paracellular pathway.
We have determined that TNF-α activates one or more src family PTKs in a time-dependent manner in human pulmonary vascular EC (Fig. 7). Src family PTKs have been implicated in the regulation of vascular permeability in vitro (31, 45) as well as in vivo (32, 53). Pharmacological inhibition of src family PTKs has been associated with reduction of vascular permeability in response to several agonists including VEGF (14), complement-activated neutrophils (45), and hyperoxia (25). Pharmacological inhibition of src reduces edema formation in mouse models for both stroke (32) and myocardial infarction (53). More relevant to the present study, Nwaraiku et al. (31) found that pretreatment of HUVEC monolayers with PP2 (10 μM) blocked TNF-α-induced increases in transendothelial FITC-dextran flux. We found that pharmacological inhibition of src with PP1 (10 μM) and PP2 (5 μM) each blocked ∼60% of TNF-α-induced increases in albumin flux across HPAEC monolayers (Fig. 8A). Since broad-spectrum PTK inhibitors totally blocked TNF-α-induced increments in paracellular permeability (Fig. 6A), whereas src-selective PTK inhibitors only partially blocked the TNF-α effect (Fig. 8A), other PTKs may be involved. src family PTKs may regulate endothelial barrier function through tyrosine phosphorylation of ZA proteins. In HUVEC treated with complement-activated neutrophils, PP2 blocks tyrosine phosphorylation of β-catenin (45). A recent study demonstrated that preincubation of HUVEC with PP2 blocked TNF-α-induced tyrosine phosphorylation of VE-cadherin (31). In the present study, with HMVEC-L, PP2 blocked TNF-α-induced tyrosine phosphorylation of VE-cadherin, γ-catenin, and p120ctn, but not β-catenin. These data implicate src family PTKs in both TNF-α-induced tyrosine phosphorylation of ZA proteins and barrier dysfunction.
We have demonstrated that src family PTKs participate in TNF-α-induced opening of the pulmonary vascular paracellular pathway. Murine knockouts of each of the src PTKs expressed in EC (c-src, fyn, yes) fail to demonstrate a phenotypic abnormality within the lungs or pulmonary vasculature (27). Even fyn−/−/yes−/− double knockouts display no phenotypic changes within the lungs (41). Of course, because these reported knockouts were not challenged with TNF-α, the role of one or more of these PTKs in the host response to TNF-α was not tested. In the current report, knockdown of fyn protected against TNF-α-induced opening of the pulmonary microvascular endothelial paracellular pathway, whereas knockdown of either c-src or yes failed to do so (Fig. 9C). Although fyn knockouts have phenotypic abnormalities unrelated to the pulmonary vasculature, the fact that selective deletion of fyn alone protects against the TNF-α effect indicates a unique function that fyn performs in HMVEC-L that cannot be executed by either of the other two EC-expressed src family PTKs, c-src or yes. We are the first to demonstrate that fyn has a novel role in the human pulmonary microvascular EC response to TNF-α.
The exact role of fyn in TNF-α-induced endothelial barrier dysfunction is presently unknown. fyn might directly regulate tyrosine phosphorylation of ZA proteins. fyn can be targeted to the plasma membrane through myristoylation and palmitoylation of its NH2 terminus (37), placing fyn in close proximity to the ZA. In the present report, we found that knockdown of fyn blocked TNF-α-induced tyrosine phosphorylation of VE-cadherin (Fig. 9C). Whether fyn directly phosphorylates VE-cadherin on tyrosine is unknown. Presently, very little is known about the role of fyn in EC-EC adhesion. Recent evidence has suggested that fyn is involved in the formation of cell-cell adhesions in keratinocytes (10). In the presence of high Ca2+ levels, fyn is activated and will colocalize with E-cadherin in keratinocytes (9, 10). fyn−/− keratinocytes exposed to increased Ca2+ levels displayed reduced levels of β-catenin, γ-catenin, and p120ctn tyrosine phosphorylation compared with wild-type keratinocytes (9). In intestinal epithelial cells, fyn phosphorylates β-catenin on tyrosine and transfection with fyn alters the association between β- and α-catenin (36). In murine neurons, fyn is constitutively bound to the neural cell adhesion molecule NCAM140 (6). fyn also may play a role in TNF-α-induced cytoskeletal reorganization, a requirement for barrier dysfunction. In fibroblasts, fyn is involved in actin pedestal formation following infection with Escherichia coli (35). In human dermal microvascular EC, TSP-1-induced activation of p38 mitogen-activated protein kinase (MAPK) requires fyn (24). This is relevant because p38 MAPK can be activated by TNF-α and it has been implicated in regulation of the EC cytoskeleton (34). The ZA and/or actin cytoskeleton may provide mechanisms through which fyn might regulate TNF-α-induced endothelial barrier dysfunction.
Other src PTKs have been implicated in the regulation of vascular permeability (14, 32, 53). Using a mouse skin model, Eliceiri et al. (14) found that c-src- and yes-knockout mice each failed to respond to VEGF challenge, whereas fyn-knockout mice did respond. In subsequent studies by the same investigators, VEGF-induced edema in mouse models of stroke (32) and myocardial infarction (53) were c-src dependent. The difference between these reports and the present study may be ascribed to lung-specific requirements for fyn and/or specific signaling requisites for TNF-α.
In summary, we have demonstrated that TNF-α increases tyrosine phosphorylation of VE-cadherin and opens the paracellular pathway in human lung endothelia through fyn activation. This mechanism only explains ∼60% of the disruption of barrier integrity in response to the TNF-α stimulus. Additional studies are required to determine whether blockade of this fyn-dependent pathway will improve survival in patients with acute lung injury or ARDS.
This work was supported by National Heart, Lung, and Blood Institute Grants HL-70155 and HL-58064 (S. E. Goldblum). D. J. Angelini was a recipient of an American Heart Association Pre-Doctoral Fellowship (Mid-Atlantic Affiliate, 0215179U).
We thank Seth Crawford for figure preparation, Shirley Taylor and Doreen Bury for manuscript preparation, and Dr. Doug Bannerman for critical review of the manuscript.
Present address of D. J. Angelini: Dept. of Anesthesiology and Critical Care Medicine, Johns Hopkins University, 720 Rutland Ave., Ross 355, Baltimore, MD 21205.
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 © 2006 the American Physiological Society