an increase in fluid shear stress to endothelial cells induces the generation of nitric oxide (NO) and reactive oxygen species (ROS), enhances intracellular Ca²⁺ concentration, and activates a signaling cascade involving extracellular signal-regulated kinases 1 and 2 (ERK1/2), nuclear factor-κB (NF-κB), and activator protein-1 (AP-1) in endothelial cells (17, 20, 21, 23, 28). Conversely, we have shown that removal of shear stress (abrupt cessation of perfusion) in endothelial cells in situ or after flow adaptation in vitro elicits similar responses including increased generation of ROS, activation of NF-κB and AP-1, and an increase of intracellular Ca²⁺(3, 5, 33, 36). Because our experimental models in vivo or in vitro permit continued oxygenation in the absence of perfusion, we have used the term “oxygenated ischemia” to differentiate it from the usual models that include anoxia as a condition of ischemia. We have recently demonstrated an increase in NO generation under these conditions (7, 18, 22).

How do endothelial cells sense the removal of shear stress that induces generation of NO? Clearly, changes in cellular oxygenation are not involved, suggesting that mechanosensors are primarily responsible for the effect. Several potential candidates such as ion channels, caveolae, integrins, and G proteins may serve to convert mechanical stimulation associated with changes in flow to chemical signals (11, 31, 34). In an isolated and continuously ventilated lung model of oxygenated ischemia with naturally flow-adapted endothelial cells, we have demonstrated that ischemia results in rapid (within seconds) endothelial cell membrane depolarization, possibly associated with K⁺channel inactivation (3-6), and ROS generation via membrane-bound NADPH oxidase (6, 40). These results suggest that an ion channel can sense loss of shear but do not exclude the possible involvement of other sensors. Caveolae have been suggested to be shear sensitive and to represent a possible alternate or additional sensor for loss of shear with ischemia (29). These specialized invaginated microdomains appear on the cell surface of many cell types and are especially abundant in certain vascular endothelia (25, 30). They are enriched with cholesterol, glycosphingolipids, and lipid-anchored signaling molecules and contain caveolin as a principal component (30). Caveolin, a 21- to 24-kDa membrane protein, binds directly to cholesterol and interacts with such signaling molecules as G protein, Ras, Src family kinases, and the endothelial form of NO synthase (eNOS) (25). Caveolae/caveolin-1 can regulate the activation of ERK1/2 and eNOS associated with increased shear stress (26, 28), although its role in decreased shear (ischemia) has not been reported. Mitogen-activated protein (MAP) kinases are important signaling components linking extracellular stimuli to cellular response, and shear induction of eNOS expression depends on ERK1/2 activation (12, 31, 34).

The goal of this work was to investigate the signaling pathway for NO generation with simulated ischemia in flow-adapted endothelial cells. We first reproduced the phenomenon of increased ROS and NO generation with ischemia in endothelial cells that were flow adapted using an artificial capillary system. We then asked the following questions: Is the increase of NO generation with ischemia Ca²⁺/calmodulin dependent? Can membrane cholesterol-sensitive compartments such as caveolae serve as flow sensors? Is the ERK1/2 cascade the pathway that links the flow sensor to NO generation? Are ROS required for ERK1/2 activation?

For flow adaptation, we used an artificial capillary system in which endothelial cells line the inner surface of polypropylene tubes to form tubular structures as in an in vivo capillary system. Pulsatile fluid flow through these tubes was used for adaptation of shear stress (36). Using this artificial capillary system for growing and flow adapting bovine pulmonary artery endothelial cells (BPAEC), we show that abrupt cessation of shear stress leads to NO generatin that is dependent on Ca²⁺/calmodulin and activation of ERK1/2. ERK1/2 activation requires ROS and is sensitive to membrane cholesterol depletion. We speculate that membrane cholesterol-sensitive compartments such as caveolae serve as flow sensors for NO generation with ischemia and that the ERK1/2 cascade mediates signaling from the sensor to the effector enzyme (eNOS).

MATERIALS AND METHODS

Cell culture.

BPAEC (CCL-209) obtained from the American Type Culture Collection (Manassas, VA) were propagated in MEM containing Earle's salts. Cultures were maintained at 37°C in a humidified atmosphere containing 95% air-5% CO₂. Passages 17–25 of BPAECs were used for experiments.

Artificial capillary system and simulated ischemia.

BPAECs were cultured under flow using commercially available artificial capillary technology (CellMax Quad, Cellco; Germantown, MD) as described previously (6). Briefly, each CellMax system has a central pump station capable of accommodating four flow paths with cartridges. Each cartridge consists of 230 semipermeable polypropylene hollow fibers (artificial capillaries) mounted in a hard polycarbonate casing, with ports allowing perfusion via the luminal or the abluminal compartment.

The inner lumen of the fibers was coated with ProNectin F (Protein Polymer Technologies; San Diego, CA). Cells from five confluent T 75-cm² flasks of BPAECs were seeded per cartridge. To prevent the unattached BPAEC from being flushed out of the fibers, the perfusing media was routed to the abluminal side during a 24-h cell attachment period. The perfusion was then rerouted to perfuse the luminal side and cells were cultured under pulsatile flow of MEM culture medium for 3 days at 1 dyn/cm² mean shear stress. These cells were considered to be flow adapted. Ischemia was simulated by abruptly rerouting flow from the luminal to the abluminal compartment. This protocol eliminated endothelial shear stress but allowed continued oxygenation. We have shown previously by analysis of medium samples obtained from the cartridge lumen using an oxygen electrode that Po₂ during abluminal flow (ischemia) is similar to control (36), indicating the adequacy of gaseous diffusion between the luminal and abluminal compartments.

The ischemic time for study of ERK1/2 was varied from 0 to 4 h. For the study of NO or ROS generation, a 1-h period of ischemia was used. Krebs-Ringer bicarbonate (KRB) solution, pH 7.4, was the perfusate for study of NO and ROS generation and MEM was the perfusate for other studies. Each experiment had a corresponding continuously perfused control. In some experiments, cells were pretreated with various inhibitors including 10 μM diphenyleneiodonium (DPI) chloride (ICN Biochemicals; Cleveland, OH), 2.5 mM N-acetylcysteine (NAC), 1,000 U/ml catalase, 1 μg/ml filipin, 10 mM cyclodextrin, 100 μM genistein (all from Sigma), 50 μM PD-98059, 10 μM U-0126, or 10 μM calmidizolium chloride (all from Calbiochem, La Jolla, CA). Pretreatment was 1 h for cyclodextrin and 30 min for other inhibitors.

At the end of the control (constant flow) or ischemic period, cells were removed from the cartridges with trypsin (0.25% for 5 min) and centrifuged (200 g for 5 min). For protein assay, cell pellets were resuspended in a solution of 50 mM potassium phosphate, 0.1 mM EDTA, and 0.1% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate buffer (pH 7.0). The cells were sonicated on ice using a probe sonicator by two bursts of 15 s each with a 10-s interval at 40% of maximal power output. Cell sonicate protein concentration was determined by Coomassie blue assay (BioRad; Richmond, CA) with bovine IgG as the standard.

Immunoblots.

Immunoblotting by the Western blotting technique with commercially available antibodies (Santa Cruz Biotechnology) was used to assess levels of total and phosphorylated ERK1/2. Trypinized and pelleted cells were resuspended in iced-cold buffer consisting of 145 mM NaCl, 0.1 mM MgCl2, 15 mM HEPES, 10 mM EGTA, 1 mM Na₃VO₄, 1% Triton X-100, and protease inhibitor cocktail (0.5 mM phenylmethylsulfonyl fluoride, 50 mg/l leupeptin, 25 mg/l aprotinin, and 25 mg/l pepstatin, all from Sigma). Cell protein extracts were prepared by sonication as above and added to loading buffer, boiled, and electrophoresed on a 12% SDS-PAGE gel. After transfer to a nitrocellulose membrane and blocking for 1 h with Tris-buffered saline (20 mM Tris · HCl, pH 7.5; 500 mM NaCl; 0.05% Tween 20) plus 3% dry milk, the membranes were incubated for 2 h with the first antibody, washed, and then incubated with peroxidase-conjugated affinity-purified goat IgG (Jackson ImmunoResearch Lab; West Grove, PA). The reaction was revealed with the enhanced chemoluminescence detection method (Amersham).

Assessment of NO and ROS generation with ischemia. NO generation was assessed by labeling cells in the flow cartridges by perfusion for 30 min with 5 μM 4,5-diaminofluorescein diacetate (DAF-2 DA, Calbiochem). ROS generation was assessed using a similar protocol with 5 μM dichlorofluorescein diacetate (DCF DA). The diacetate compounds are deesterified intracellularly to DAF-2 or DCF, respectively. NO and its higher oxides such as NO₂ and/or NO₃ provide the third nitrogen to form a triazo ring from the two amino groups of the nonfluorescent DAF-2 and convert it to fluorescent triazolofluorescein (DAF-2T) (24). DCF is oxidized by H₂O₂ or other ROS to form the fluorescent dichlorofluorescein (36). After cells were loaded, they were then washed with 100 ml of dye-free perfusate before the experimental period. Fluorescence was measured in the cell sonicate of harvested cells at 515 nm (excitation 490 nm) for DAF-2T and at 530 nm (excitation 490 nm) for oxidized DCF.

Statistics.

Data represent means ± SE. Statistical analysis was carried out by ANOVA using SigmaStat (Jandel; San Rafael, CA). Differences were considered statistically significant at P < 0.05.

RESULTS

To study the time course of the ischemic effect on the level of phosphorylated and total ERK1/2, flow-adapted BPAECs were subjected to control (continuous luminal flow) or ischemia (abluminal flow) for 0 to 4 h (Fig.1). By immunoblot analysis, ischemia for 10 min resulted in an increase of phosphorylated ERK1 and phosphorylated ERK2 of 158 ± 14 and 143 ± 10%, respectively, above control. ERK2 phosphorylation reached a peak at 20 min of ischemia (203 ± 21% above control). Phosphorylation of ERK1/2 reached a plateau after 30 min of ischemia (increase of 186 ± 7% for ERK1 and 182 ± 28% for ERK2), which represented the peak for phosphorylated ERK1 and slightly less than the peak for phosphorylated ERK2. Therefore, 30 min of ischemia was chosen for subsequent experiments related to ERK1/2 activation. Levels of total ERK1/2 did not change significantly during ischemia for 0 to 4 h (Fig. 1).

Cholesterol or cholesterol-sensitive compartments such as caveolae in the plasma membrane may be critical for shear stress-dependent ERK1/2 activation (27, 29). To investigate the role of the plasma membrane, cholesterol, or a cholesterol-sensitive compartment in ERK1/2 activation with ischemia, filipin and cyclodextrin, membrane-permeable and membrane-impermeable cholesterol-binding drugs, respectively, were used. Flow-adapted BPAECs were preincubated with 1 μg/ml filipin for 30 min or 10 mM cyclodextrin for 1 h and then subjected to 30-min control perfusion or 30-min ischemia (Fig.2). ERK1/2 phosphorylation with ischemia was abolished by filipin and significantly decreased by cyclodextrin.

Using this in vitro model of oxygenated ischemia, we have shown previously that ischemia in flow-adapted endothelial cells leads to an increase of ROS generation (36). This effect was confirmed in the present study by the demonstration of a 64% increase in DCF fluorescence in cells exposed to ischemia compared with control (Table 1). To determine whether an increase in ROS is associated with ERK1/2 activation during ischemia, flow-adapted BPAECs were pretreated for 30 min with a flavoprotein inhibitor (DPI), an antioxidant (NAC), or an H₂O₂ scavenger (catalase) (Fig.3). These agents, which have been shown previously to inhibit ischemia-mediated ROS generation in these cells (22, 36), significantly inhibited ERK1/2 phosphorylation.

Because both a cholesterol-sensitive compartment and ROS were required for ERK activation, we investigated whether these two factors were coupled. Pretreatment of cells with filipin had no effect on DCF oxidation with ischemia (Table 1), indicating that ROS generation with ischemia is not altered by cholesterol depletion.

To elucidate the role of ERK1/2 in the signal transduction pathway with ischemia in flow-adapted BPAECs, we employed a general inhibitor of tyrosine kinases (genistein) or putative-specific inhibitors of mitogen and extracellular signal-regulated kinase kinase (MEK) (PD-98059 and U-0126). Preincubation for 30 min with these inhibitors blocked ERK1/2 activation with ischemia (Fig.4).

DAF-2 DA was used to determine NO generation with ischemia. We have used this dye previously to show NO generation in lung endothelial cells both in situ and in culture and have demonstrated specificity by inhibition with an NOS inhibitor,N^G-nitro-l-arginine methyl ester (7, 22). Flow-adapted endothelial cells subjected to 1-h ischemia exhibited an 80% increase in NO production, as indicated by DAF-2T fluorescence, compared with corresponding continuously perfused controls (Fig. 5). The increase in NO generation with ischemia was inhibited by pretreatment with either cholesterol-depleting reagents (filipin and cyclodextrin) or ROS scavengers (catalase and NAC) (Fig. 5). NO production with ischemia was also inhibited by preincubation with PD-98059, U-0126, or genistein. None of the inhibitors decreased DAF fluorescence in vitro, indicating that the effect was not due to quenching of the fluorophore (data not shown). Finally, depletion of extracellular Ca²⁺ by perfusion with Ca²⁺-free KRB solution containing 1 mM EGTA or inhibition of Ca²⁺/calmodulin formation with calmidazolium chloride also inhibited NO production with ischemia (Fig. 5).

DISCUSSION

Acute lung ischemia can occur as a result of pulmonary embolism or interruption of circulation during surgery, as for example in lung transplantation. This study utilized an isolated cell system as a model for ischemia in the pulmonary circulation. Because endothelial cells in the in situ lung are expected to be naturally adapted to the normal blood flow, cultured cells were flow adapted to 1 dyn/cm² of shear stress for 3 days before the evaluation of ischemia. This level of shear stress was chosen for adaptation because higher levels resulted in increased cell detachment. Although the level of shear stress used for adaptation is low compared with physiological values for large vessels, shear stress of this magnitude may be physiological in capillaries. We have shown previously that 1 dyn/cm² of shear stress does flow adapt endothelial cells so they can generate an ischemic response as manifested by ROS generation and activation of intracellular transcription factors (36). Furthermore, the response of flow-adapted endothelial cells in vitro to abrupt cessation of flow is similar to the responses we have observed in the intact lung (22,36).

MAP kinases (members of the family of serine/threonine kinases) appear to be important mediators of mechanosignal transduction (8, 31,34). ERK1/2 are among the most important of these protein kinases because many of the substrates for ERK1/2 are critical mediators of cell function. The ERK1/2 cascade may provide a link from extracellular mechanical stimulation to cellular regulation of gene expression and protein function. Increased shear stress has been shown clearly to activate ERK1/2 in bovine aortic endothelial cells as well as human umbilical vein endothelial cells maintained in static (no flow) culture (8, 19). This present study with flow-adapted BAPECs demonstrates ERK1/2 activation in response to abrupt removal of shear stress. Thus the mechanotransduction pathway in endothelial cells may respond similarly to increased flow in statically adapted cells and to removal of flow in flow-adapted cells, suggesting that cells respond to any mechanical perturbation from a maintained set point.

Cholesterol or cholesterol-sensitive compartments such as caveolae in the plasma membrane appear to play a key role in shear stress-dependent ERK1/2 activation (27, 29). In this study, we have presented evidence that plasma membrane cholesterol is involved in ERK1/2 activation with simulated ischemia in flow-adapted BPAECs. Treatment with filipin, a membrane-permeable cholesterol-binding agent, or cyclodextrin, a membrane-impermeable cholesterol-binding agent, blocked ERK1/2 activation with removal of shear stress (Fig. 2). The role of membrane cholesterol in mechanotransduction may be related to the regulation of membrane fluidity, which may affect function of membrane proteins (16,27). Considering that cholesterol directly binds to caveolin-1, it could directly influence the function of this protein which interacts with and regulates the activity of signaling molecules such as G proteins, Src, and Ras (25, 30). Thus a possible mechanism for the cholesterol effect is that removal of shear stress results in increased interaction between membrane cholesterol and caveolin-1, or between the scaffolding domain of caveolin-1 and signaling molecules (e.g., G_αi-2, Src, and Ras) that are upstream regulators of the flow-sensitive ERK pathway.

Recent studies have indicated that ROS production may result in ERK activation, possibly through activation of Ras (1, 2). ROS have been shown to regulate the Ras/Raf-1/ERK pathway with cyclic strain (37) and to serve as the mediator for shear-induced tyrosine phosphorylation in endothelial cells (39). Our recent studies in the isolated perfused lung and with the in vitro artificial capillary system indicate that ischemia induces ROS generation through a pathway that is sensitive to DPI, a flavoprotein inhibitor (3, 22, 36). This pathway appears to be membrane-bound NADPH oxidase because extracellular superoxide is the major species produced (22) and ROS production is abolished by “knockout” of gp91^phox (6). To evaluate the role of ROS in ERK1/2 activation by ischemia in BPAEC, we used DPI to inhibit ROS production and catalase and NAC as ROS scavengers. Catalase was an extracellular scavenger, suggesting that ROS generation was into the extracellular space as expected for the membrane-bound oxidase (22). The finding that these inhibitors blocked ERK1/2 activation with ischemia indicates a role for ROS in these events (Fig. 3). The cholesterol-sensitive compartment and ROS are independent stimuli for activation of ERK because cholesterol depletion did not alter ROS production. However, both were necessary because inhibition of either prevented ERK activation.

Ischemia resulted in an increase in NO generation that was suppressed by inhibitors of ERK1/2 activation, with Ca²⁺ depletion, or by inhibition of calmodulin. eNOS contains multiple consensus sites for phosphorylation and its activation requires phosphorylation on both serine and threonine residues. Because ERK1/2 are serine/threonine kinases, they are suitable candidates for regulation of the sustained phase of shear-induced NO release (9, 15, 34, 39). In a recent study, published in abstract form, shear stress-induced eNOS expression was dependent on ERK1/2 activation (12). In the present study, inhibitors of ERK1/2 activation and cholesterol-binding reagents blocked ERK phosphorylation and NO generation, compatible with the hypothesis that activation of NOS during ischemia in flow-adapted endothelial cells occurs via ERK1/2 activation (Fig. 5).

The present study demonstrates that both Ca²⁺-free KRB solution and calmidazolium chloride, an inhibitor of Ca²⁺/calmodulin, suppress NO generation with ischemia, indicating that NO generation is Ca²⁺/calmodulin dependent (Fig. 5). eNOS is classified as a constitutive and strictly Ca²⁺-calmodulin-dependent enzyme, whereas inducible NOS is Ca²⁺ independent (15). Thus eNOS is the isoform most likely to be activated by ERK1/2 during ischemia in flow-adapted BPAECs. Our previous studies (7, 22) of lung endothelial cells in situ (using the isolated rat lung) and in vitro have established that intracellular Ca²⁺ does increase with ischemia as required for eNOS activation. The increase of Ca²⁺ with ischemia is due to an initial release from internal stores followed by influx from the extracellular medium (7). One possible mechanism for Ca²⁺ influx with ischemia may be related to depolarization of the endothelial cell membrane with subsequent activation of voltage-dependent Ca²⁺ channels (VDCC) (4, 6, 33). Although functional VDCC have not been reported for passaged endothelial cells, their activity has been shown in confluent primary endothelial cell culture. (10, 32). It is possible that VDCC are lost during static cell culture but can be upregulated during flow adaptation. As alternative possibilities, Ca²⁺ influx may be due to Ca²⁺release-activated Ca²⁺ flux or receptor-mediated cation channels (14, 38).

In summary, we have shown that simulated ischemia in flow-adapted endothelial cells induces NO generation dependent on membrane cholesterol, ROS, and Ca²⁺/calmodulin (Fig. 6). Cholesterol-sensitive compartments such as caveolae in the plasma membrane may serve as flow sensors for NO generation with ischemia, whereas the ERK1/2 cascade, activated by ROS, may mediate signaling from the sensor to the effector enzyme (eNOS).

FOOTNOTES

• The study research was supported by National Heart, Lung, and Blood Institute Grant P50-HL-60290 (to A. B. Fisher).

• Address for reprint requests and other correspondence: A. B. Fisher, Institute for Environmental Medicine, Univ. of Pennsylvania Medical Center, 1 John Morgan Bldg., 3620 Hamilton Walk, Philadelphia, PA 19104-6068 (E-mail: [email protected]med.upenn.edu).

• 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.