Calcium/Calmodulin-dependent Protein Kinase II (cid:1) 2 and (cid:2) Isoforms Regulate Potassium Currents of Rat Brain Capillary Endothelial Cells under Hypoxic Conditions*

Endothelial K (cid:3) and Ca 2 (cid:3) homeostasis plays an important role in the regulation of tissue supply and metabo-lism under normal and pathological conditions. How-ever, the exact molecular mechanism of how Ca 2 (cid:3) is involved in the regulation of K (cid:3) homeostasis in capillary endothelial cells, especially under oxidative stress, is not clear. To reveal Ca 2 (cid:3) -triggered pathways, which modulate K (cid:3) homeostasis, Ca 2 (cid:3) /calmodulin-dependent protein kinase II and voltage-gated outward K (cid:3) currents were studied in rat brain capillary endothelial cells under hypoxia. Whole cell voltage-clamp measure-ments showed voltage-gated outward K (cid:3) current with transient and sustained components. mRNA and protein of Ca 2 (cid:3) /calmodulin-dependent protein kinase II (cid:1) 2 and two (cid:2) isoenzymes were identified. Activation of the isoforms (autophosphorylation) was typically achieved by the Ca 2 (cid:3) ionophore ionomycin, which was prevented by the Ca 2 (cid:3) /calmodulin-dependent protein kinase II-spe-cific inhibitor KN-93. Hypoxia resulted in autophosphorylation

Endothelial cells (EC) 1 regulate numerous processes of the vasculature, e.g. blood coagulation, angiogenesis, barrier, and transport functions. They serve both as target and release site of vasoactive agents adjusting circulatory parameters (1). As a primary barrier between circulation and tissue, they play a major role in the regulation of fluid and mineral distribution between extravascular and intravascular space. This is effectively controlled by microvascular EC representing a huge surface.
There is an increasing number of ion channel types revealing different expression patterns when comparing EC of different origin (1). They play a role in the regulation of membrane potential (2), Ca 2ϩ -signaling pathways (3), cell volume (4), vessel permeability (5), and mechanosensor function (6). Although EC are regarded to be nonexcitable, voltage-gated Ca 2ϩ (7) and K ϩ (8) channels have been reported. The channels may be involved in the release of vasoactive compounds regulated mainly via intracellular Ca 2ϩ movements. Ion channels are described in a more detailed manner on macrovascular EC. Little is known about microvascular EC. The latter express voltage-dependent Ca 2ϩ (9), Na ϩ (10), ATP-sensitive (11), and stretch-sensitive nonselective cation (12) channels. Concerning K ϩ channels, Ca 2ϩ -activated (13) and voltage-gated K ϩ channels (14) have been observed in microvascular EC.
Ionic and the closely related osmotic homeostasis partly regulated by ion channels (K ϩ , Na ϩ , Cl Ϫ ) play an important role in preserving EC morphology and EC-EC coupling required for normal vessel permeability (15). Morphological impairment of EC, such as swelling or EC disorganization, may cause blood flow disturbances (16) or increased extravasation (17). The tightness of the blood-brain barrier is greatly affected by hypoxia as indicated by increased paraendothelial permeability (18) and endothelial swelling (19). One possible mechanism in endothelial swelling is an alteration of the K ϩ channel function, which may result in the loss of the K ϩ recycling pathway necessary for a proper Na ϩ ,K ϩ -ATPase function (20).
Members of the diverse superfamily of voltage-activated K ϩ channels have been demonstrated to be modulated by phosphorylation catalyzed by different protein kinases, among them Ca 2ϩ /calmodulin-dependent protein kinase II (CaMKII). Phosphorylation of the K ϩ channel subtype Kv1.4 at the N terminus by CaMKII affects the fast inactivation kinetics and the current amplitude (21). Modulation of K ϩ current inactivation by CaMKII is reported in excitable cells such as atrial myocytes (22) and colonic smooth muscle cells (23).
CaMKII is a ubiquitously expressed Ser/Thr protein kinase and transmits Ca 2ϩ signals to a variety of targets ranging from ion channels to transcription factors. CaMKII is encoded by four separate genes (␣, ␤, ␥, ␦) with partial tissue-specific expression of its subunits and their numerous splicing variants (24). In the macrovascular endothelium CaMKII is shown to be involved in the mechanism of thrombin-induced barrier dysfunction (25), and NO synthase regulation (26,27). As intracellular Ca 2ϩ is increased in EC under pathological conditions, such as hypoxia (28), CaMKII activation is a potential candidate to be involved in the mechanisms of such a cell injury.
Here we hypothesize that CaMKII is activated in EC upon oxidative stress, which in turn modulates K ϩ channels involved in maintaining the endothelial supply and barrier function. Therefore, the aim was to elucidate the effect of hypoxia on CaMKII activity and K ϩ current in brain capillary endothelial cells forming the blood-brain barrier. The presence and activation of various CaMKII isoenzymes and distinct voltage-gated K ϩ current components and their modulation were identified under normal and hypoxic conditions. Both the K ϩ current and the CaMKII activation were sensitive against CaMKII blocker indicating a regulatory relevance of this pathway for the K ϩ homeostasis of the microvascular endothelium.
To detect CaMKII␦, a subclass-specific antibody was used recognizing ␦ isoforms containing the C-terminal second variable domain (29). For detection of autophosphorylated CaMKII, an antibody was raised in rabbit against the phosphopeptide MHRQEpTVDC corresponding to the amino acid sequence surrounding the autophosphorylation site Thr 287 in CaMKII␦. The antibody specific for the ␥-class of CaMKII was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Peroxidase-conjugated anti-rabbit IgG was from Sigma-Aldrich. The anti-goat IgG conjugated with peroxidase was obtained from Dianova (Hamburg, Germany). The enhanced chemiluminescence kit was from Amersham Biosciences. All other reagents were of analytical grade or better.
Cell Culture-Immortalized rat brain capillary endothelial cell line 4 (rBCEC4) (30) was cultured on 10-mm diameter round glass coverslips (Assistent, Berlin, Germany) covered by rat tail collagen for electrophysiology or in 6-well plates (BD PharMingen) for RT-PCR and immunoblotting. Cells were grown in Dulbecco's modified Eagle's medium containing 4.5 g/liter glucose, 10% fetal bovine serum, 1.2 mM L-glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, 2.5 g/ml amphotericin, 100 g/ml sodium heparin, 110 g/ml sodium pyruvate, and 10 g/ml endothelial cell growth factor (ECGF). Cultures were maintained in a humidified atmosphere with 5% CO 2 at 37°C. Medium was changed each second day. Cells were subcultured weekly from passage 20 to 30. For hypoxia, cells were transferred for 5 h to the Concept Plus Anaerobic Workstation (Ruskinn Technology Ltd., Leeds, UK) containing a gas mixture of 90% N 2 , 5% CO 2 , and 5% H 2 at 37°C and 100% humidity (31). For analyzing CaMKII autophosphorylation after hypoxia, cells were subjected to normoxic conditions for 30 min up to the fixation to simulate conditions comparable to those used in the patchclamp experiments. For pretreatment, KN-93 or KN-92 was added to the medium 30 min before starting hypoxia in a quantity to reach a final concentration of 20 M.
Electrophysiology-Patch-clamp recordings were performed in whole cell configuration at room temperature (23°C). Cells were transferred from the culture medium into extracellular solution flowing through a round coverslip recording chamber with a 150-l volume (RC-25, Warner Instruments Co., Hamden, CT). Control and test solutions were perfused through the chamber by gravity with an 8 -10 ml/h perfusion rate. Clearly separated cells were visually selected for patch-clamp recording. A digital camera was used to take images.
Whole cell K ϩ currents were measured by an Axopatch 200A amplifier (Axon Instruments, Inc.). Clampex 8.1 software (Axon Instruments, Inc.) was used for command generation and data acquisition. Currents were digitized at 10 kHz and filtered with low-pass Bessel characteristics at 5 kHz frequency. Total cell capacitance and series resistance were calculated and compensated by adjusting the amplifier's whole cell capacitance and series resistance (0 -85%) compensation controls. Only cells with access resistance of Ͻ10 megohm were accepted for further investigations. Liquid junction potentials were calculated (Ϫ10 to Ϫ13 mV), and data were corrected for each experimental condition.
Voltage-gated outward potassium currents were elicited by step commands with 1 s duration from Ϫ80 to ϩ80 mV in steps of 10 mV preceded by a 2-s prepulse at Ϫ90 mV (Fig. 3, Stimulus). The interstimulus interval was set to 10 s. This stimulus protocol is suitable to elicit the A-type K ϩ current. The amplitude of the sustained current component (I sus ) was determined as the amplitude of the outward current at the end of each trace. The maximal amplitude of the transient outward current (I to ) was calculated by subtracting I sus from the peak outward current (I peak ). The current amplitudes were normalized to cell capacitance in each cell.
RT-PCR-Total RNA from cells was isolated with the Invisorb system according to the manufacturer's instructions and from rat brain using a guanidine thiocyanate-based method (32). After isolation, remaining DNA contaminations were digested and RNA was re-extracted (29). RT-PCR was performed as described (33). PCR conditions for amplification of isoforms of the four known CaMKII classes are described as follows: PCR products were analyzed with ethidium bromide-stained 2% agarose gels or 4% native polyacrylamide gels. For sequencing, amplification products were extracted from gels, and standard cycle sequencing reactions were performed commercially by InViTek (Berlin-Buch, Germany).
Preparation of Cell Lysates and Immunoblotting-Cell cultures were washed with HEPES-buffered Hanks balanced salt solution and incubated for 15 min at 37°C for stabilization. 250 nM ionomycin was used for activation of CaMKII. For inhibition of CaMKII, cells were preincubated with 20 M KN-93 for 30 min before the addition of ionomycin. Cells were fixed by the addition of ice-cold trichloroacetic acid to a final concentration of 5% and immediately placed on ice. They were washed twice with ice-cold 2.5% trichloroacetic acid and harvested. The precipitates were centrifuged at 12,000 ϫ g at 4°C for 10 min. The supernatant was aspirated, and the pellets were dissolved in electrophoresis sample buffer at pH 6.8, neutralized if necessary with 1 M Tris, and heated at 95°C for 5 min. Electrophoretic separation on 10% SDSpolyacrylamide gels and immunoblotting were performed as described in Ref. 29 with 20 g of cell protein per lane. The autophosphorylated CaMKII-specific antibody was used in a final concentration of 1 g/ml. The antibodies obtained from commercial sources were used according to the manufacturer's instructions. Visualization of the immunoreaction was by enhanced chemiluminescence kit and autoradiography on x-ray film. Protein was measured in cell soluble lysates by a modified Lowry method (34).
Data Analysis-Clampfit 8.1 software (Axon Instruments, Inc.) was used for off-line raw data analysis. Further data processing and analysis was performed by Microcal™ Origin ® software (Microcal Software, Inc.). Two-dimensional area measurement of digital images of cultured cells was made by Scion Image for Windows software (Scion Corp., Frederick, MD). The time constant of the fast inactivation phase ( fast ) of the transient current was calculated by biexponential fitting on the decay phase of the outward current. The extent of inactivation was determined as the contribution of the transient outward current to the total peak current (I to /I peak ). Values are presented as mean Ϯ S.E. To reveal the significance of differences, paired or unpaired Student's t tests were used with significance defined as: *, p Ͻ 0.05; **, p Ͻ 0.01. (Fig. 1) show the same cultured cells before and after the hypoxia protocol. Arrows indicate the regions of apparent swelling. The surface area of the cells subjected to 5 h of hypoxia showed a significant increase compared with the initial values measured before the initiation of hypoxia protocol, whereas in the case of the control cell group, which was measured after the same period without hypoxia, no significant increase was seen. (See Fig. 1; initial value, 481 Ϯ 23 m 2 , n ϭ 29; control value, 510 Ϯ 37 m 2 , n ϭ 20; hypoxic value, 617 Ϯ 36 m 2 , n ϭ 21, p Ͻ 0.05 compared with the initial value.) These results suggest that hypoxia caused a significant increase of the cell volume (swelling).

Swelling of Hypoxic Cells-Micrographs
Components of the Voltage-gated Outward Potassium Current and Their Selectivity-To reveal possible pathways involved in the adaptation to swelling and Ca 2ϩ overload caused by hypoxia of rBCEC4, voltage-gated outward potassium currents were investigated. Two different components of the total outward current were identified, a transient outward component (I to ) with fast activation and rapid inactivation and a sustained component (I sus ) that showed no inactivation within the period of stimulus waveform (see "Experimental Procedures").
Different potassium channel blockers were used for pharmacological characterization of the outward current components. 4-AP and TEA showed a dose-dependent inhibitory effect on current amplitudes (Fig. 2, A and B). The inhibition revealed different characteristics as presented by a sigmoidal fit to the plots of normalized current values against the concentration of 4-AP (Fig. 2C) and TEA (Fig. 2D). 4-AP blocked I to with a half-maximal inhibitory concentration (IC 50 ) of 1.1 Ϯ 0.04 mM. The Hill coefficient was 2.9 Ϯ 0.05 (n ϭ 3). 4-AP inhibition reached a maximal effect at 5 mM, which decreased the current amplitude to 9% compared with control values of ϩ70 mV (Fig.  2C). Similar values for the 4-AP effect were reported earlier in EC (13) and myocytes (35). TEA inhibited I sus with an IC 50 value of 2.8 Ϯ 0.5 mM and a Hill coefficient of 1.4 Ϯ 0.3 (n ϭ 3). 10 mM TEA reached maximal blocking effect and reduced I sus to 21% of the control amplitude at ϩ70 mV (Fig. 2D). Comparable values of TEA inhibition were observed in EC (13,14) and neurons (36). 1 M charybdotoxin, an inhibitor of Ca 2ϩ -activated K ϩ current, did not influence I to and I sus significantly (not shown).
Hypoxia Increased the Amplitude of Voltage-gated Outward Potassium Current Components, and KN-93 Impeded the Effect of Hypoxia-To focus on Ca 2ϩ -triggered pathways, which may modulate K ϩ homeostasis, one group of cells was pretreated with KN-93 before the hypoxia protocol. Registrations were made under control conditions (Fig. 3A), after hypoxia (Fig.  3B), and after hypoxia preceded by KN-93 application (Fig. 3C). The current-voltage (I-V) relationship of I to and I sus (Fig. 3, D and E) were calculated in each experimental condition mentioned above. In cells subjected to hypoxia the amplitudes of both I to and I sus were elevated (I to at ϩ70 mV, 11.5 Ϯ 2.3 pA/pF, n ϭ 22 in control group versus 15.9 Ϯ 1 pA/pF, n ϭ 16 in hypoxic group, p Ͻ 0.05; I sus at ϩ70 mV, 11.4 Ϯ 2.4 pA/pF, n ϭ KN-92, the inactive analogue of KN-93, was used by the same application protocol as a negative control to reveal that the effect of KN-93 could be related to CaMKII inhibition and not to a direct modulatory effect on K ϩ channels. KN-92 did not affect the hypoxia-induced elevation of the amplitude of I to (at ϩ70 mV, 15.6 Ϯ 1.5 pA/pF, n ϭ 11) and I sus (at ϩ70 mV, 23.7 Ϯ 8.8 pA/pF, n ϭ 11).
Inactivation of I to Decreases after Hypoxia, and the Effect of Hypoxia Is Prevented by KN-93fast , the time constant of the fast inactivation phase of I to (see "Experimental Procedures") showed voltage dependence in the ϩ10 to ϩ80 mV voltage range under all experimental conditions (Fig. 4B). Traces shown in Fig. 4A were recorded at ϩ70 mV holding command (Fig. 4, Stimulus) under control conditions and 5 h after hypoxia alone and preceded by KN-93 administration. fast elevated significantly after 5 h of hypoxia indicating the slower inactivation of I to (control group at ϩ70 mV, 76.7 Ϯ 2.5 ms, n ϭ 21 versus hypoxic group at 89.7 Ϯ 3.2 ms, n ϭ 16, p Ͻ 0.01). A previous application of KN-93 prevented the effect of hypoxia on fast (at ϩ70 mV, 71.6 Ϯ 2.6 ms, n ϭ 15) (Fig. 4B).
The I to /I peak ratio, which describes the extent of inactivation (see "Experimental Procedures"), decreased significantly after hypoxia, pointing to a decreased contribution of I to to the total current (control group, 0.5 Ϯ 0.04, n ϭ 22 versus hypoxic group, 0.4 Ϯ 0.02, n ϭ 16, p Ͻ 0.01). In the group subjected to KN-93 application before hypoxia, the I to /I peak value was in the control range (at ϩ70 mV, 0.51 Ϯ 0.04, n ϭ 15) (Fig. 4C).
Pretreatment of the cells with KN-92 before hypoxia did not impeded the alteration of the fast value (at ϩ70 mV, 93.4 Ϯ 4.9 ms, n ϭ 11) and the I to /I peak ratio (at ϩ70 mV, 0.4 Ϯ 0.06, n ϭ 11).
mRNA Expression of CaMKII Isoenzymes in rBCEC4 Cells-To assess the expression pattern of CaMKII in rBCEC4, we determined the mRNA transcript pattern by RT-PCR and immunoblotting with class-or isoform-specific primer combinations (Fig. 5). The class-specific primers for ␣, ␤, and ␥ isoforms of CaMKII were derived from the conserved domains of these classes flanking their variable parts. Therefore, all expressed isoforms of these classes should be detected. For ␦-CaMKII, primers specific for individual isoforms of this class were used. Since ␣and ␤-CaMKII are known to be neuronally expressed (37), RNA from rat brain was amplified in parallel as a positive control. As expected, no ␣and ␤-CaMKII-specific signals were obtained in rBCEC4 cells, whereas products of the expected sizes were amplified from the brain RNA fraction (Fig.  5, upper left and upper middle). With the ␥-class-specific combination, two products were detected in rBCEC4-derived preparations that were identified by size or by sequencing as iso- forms ␥ B and ␥ C (38) (Fig. 5, upper right). ␦-CaMKII is characterized by the appearance of a second C-terminal variable domain (39). To assess the transcript pattern of CaMKII␦ in rBCEC4 we concentrated on the non-neuronal members of this subclass (isoforms ␦ 2 , ␦ 3 , ␦ 4 , ␦ 9 ) (29). As shown in the lower part of Fig. 5, rBCEC4 expressed only transcripts for isoform ␦ 2 . The very faint band arising in ␦ 4 -specific amplification reactions is not close to the size of the positive control obtained with the ␦ 4 -containing plasmids (530 bp, positive control). Therefore, it was considered to be an artifact of the PCR.
Protein Expression of CaMKII in rBCEC4 -To analyze the expressed proteins of CaMKII in rBCEC4 based on the mRNA data, CaMKII␦-and CaMKII␥-specific antibodies were employed. The ␦-specific antibody specifically recognizes isoforms with the unique C-terminal extension that comprises a subset of the CaMKII␦ isoforms known at present (29). The ␥-specific antibody is directed against an amino acid sequence conserved in all known ␥ isoforms, and thus is able to detect all forms of CaMKII␥. Fig. 6A shows immunoblots with extracts of rBCEC4. The ␦-specific antibody exclusively detects a signal at about 54 kDa (lane 1). Based on our transcript data and the apparent molecular mass, this protein may be attributed to the CaMKII isoform ␦ 2 . The ␥-specific antibody produces three reactive bands at about 60, 54, and 42 kDa. The prominent protein band at 60 kDa correlates with the deduced molecular mass of CaMKII␥ B , and according to our mRNA data it most likely represents the isoform ␥ B . The very faint 54-kDa band, migrating exactly at the position of ␦ 2 , coincides in size with the deduced molecular mass of isoform ␥ C . However, we cannot exclude the possibility that the signal is produced by a crossreaction of the antibody with CaMKII␦ 2 . The nature of the prominent immunoreactive protein band at about 42 kDa remains to be identified. It is not in the size range of any known CaMKII protein.
Activation of CaMKII Isoforms in rBCEC4 -CaMKII␦ 2 and CaMKII␥ became autophosphorylated when rBCEC4 were subjected to ionomycin. At a concentration of 250 nM ionomycin leads to the detection of two protein bands by the CaMKII autophosphorylation site-specific antibodies, corresponding to the positions of CaMKII␦ 2 (54 kDa) and CaMKII␥ B (60 kDa). The autophosphorylation of the 54-kDa protein was very low in controls and increased quickly and transiently in response to ionomycin with the maximum of 30 s. It was not detectable anymore at 10 min. In contrast, the 60-kDa protein showed an apparent higher basal level of autophosphorylation that increased more slowly after ionomycin application and remained stable for at least 10 min. Interestingly, there was no autophosphorylation-specific reaction at the position of 42 kDa, supporting our assumption that the protein detected by anti-CaMKII␥ does not represent a functional CaMKII isoform. A 30-min pretreatment with 20 M KN-93 reduced the autophosphorylation of both isoenzymes considerably (Fig. 6B).
Effect of Hypoxia on CaMKII Autophosphorylation in rBCEC4 -Hypoxia resulted in activation of CaMKII in rB-CEC4 as demonstrated by autophosphorylation of both CaMKII␦ 2 and CaMKII␥ (Fig. 7A). In controls there was only very low autophosphorylation detectable for ␦ 2 , which increased significantly because of hypoxia (control, 10 Ϯ 2.6 OD, n ϭ 3 versus hypoxia, 85.6 Ϯ 7.4 OD, n ϭ 5, p Ͻ 0.01) (Fig. 7B). In control cells ␥ B elicits a higher autophosphorylation than ␦ 2 resulting in hypoxia-treated cells in a higher maximum. However, the cells show a smaller increment of autophosphorylation compared with that of ␦ 2 . The increase in autophosphorylation of the ␥ B isoenzymes was about 60% and statistically significant (control, 74.3 Ϯ 5.7 OD, n ϭ 3 versus hypoxia, 121.8 Ϯ 10.7 OD, n ϭ 5, p Ͻ 0.05). The CaMKII inhibitor KN-93 completely abolished the response in autophosphorylation to hypoxia for both CaMKII␦ 2 (11.6 Ϯ 1.6 OD, n ϭ 5) and CaMKII␥ (75.6 Ϯ 9.9 OD, n ϭ 5). DISCUSSION The current study for the first time provides evidence for a role of CaMKII in the regulation of voltage-gated K ϩ channels in hypoxic rat brain capillary endothelial cells. A function of the CaMKII has been described in macrovascular endothelial cells but not in the widely differentiated microvasculature that is specialized depending on surrounding tissue. In aortic endothelial cells, CaMKII can be involved in the response to oxidative stress caused by H 2 O 2 (26) or in thrombin-induced cytoskeletal reorganization and endothelial barrier dysfunction (25). The isoenzyme pattern of CaMKII in the endothelium is unknown. In this study the presence of CaMKII␦ 2 and CaMKII␥ is demonstrated in brain capillary endothelial cells, and evidence is presented for their functional coupling to voltage-gated K ϩ currents under hypoxic conditions. Hypoxia increases the intraendothelial free Ca 2ϩ concentration as shown earlier (28). The isoenzymes identified in rBCEC4 are activated (autophosphorylated) in a Ca 2ϩ -dependent manner by hypoxia. The autophosphorylation and the alteration of the K ϩ current characteristics, both caused by hypoxia, are prevented by the selective CaMKII inhibitor KN-93.
The function of the outward voltage-gated K ϩ current measured in the brain capillary endothelial cells used in the current study is probably related to the regulation of membrane potential. K ϩ channels are involved in membrane potential regulation in depolarized cells, and they can counteract further membrane depolarization (14). This may possess a regulatory function during membrane potential oscillations elicited by vasoactive agonists (40) or during membrane depolarization caused by oxidative stress (41). Regulation of membrane potential plays a pivotal role in Ca 2ϩ homeostasis since voltagegated Ca 2ϩ channels serve as one possible pathway for Ca 2ϩ influx (42).
In rBCEC4 we found a transient and a sustained K ϩ current component. The characteristics of the transient current corre- sponds to the A-type, a fast-activating and fast-inactivating outward voltage-gated K ϩ current. To the sustained component, a delayed rectifier current contributes to a great extent as demonstrated by the effect of TEA. Both the A-type and the delayed rectifier current are reported to be present in primary capillary endothelial cells (13,14). This means that rBCEC4 preserves primary properties, making them suitable for further electrophysiological studies. Moreover, the action of the specific K ϩ channel blockers 4-AP and TEA suggests that the majority of the outward current is carried by K ϩ . Finally, the absence of a charybdotoxin effect indicates that Ca 2ϩ -activated K ϩ channels are not directly involved in the modulation of the voltagegated K ϩ current components investigated.
Upon hypoxia, rBCEC4 shows an increased value and a reduced I to /I peak ratio, indicating a decrease of the transient current inactivation that leads to elevation of the transient and sustained K ϩ current components. A similar change in K ϩ currents has been observed in myocytes dialyzed with autophosphorylated CaMKII (23). The hypoxia-related rise in the value, the diminished I to /I peak ratio, and the elevated transient and sustained K ϩ current amplitudes can be prevented by the specific CaMKII inhibitor KN-93. Taken together, the assumption is supported that CaMKII enhances the release of K ϩ from hypoxic rBCEC4 when voltage-gated K ϩ channels are operating.
Immunohistochemistry experiments reveal an expression of Kv1.4 channels in rBCEC4 (data not shown). These channels are rapidly inactivated via their N-terminal inactivation domain (43). In human embryonic kidney (HEK) cells, phosphorylation of the N-terminal part of the transfected Kv1.4 channel by CaMKII decelerates the inactivation of the fast-inactivating A-type current (21). Therefore, it is proposed that CaMKII regulates the inactivation of the K ϩ current in rBCEC4. This conclusion is supported by a considerable deceleration of the transient current inactivation under hypoxic conditions when the detected CaMKII isoenzymes were activated. It is known that voltage-gated K ϩ channels can be modulated by serine/threonine phosphorylation regulated by Ca 2ϩ (44). CaMKII is such a Ca 2ϩ -dependent serine/ threonine protein kinase (24). The assumption of a CaMKIIregulated K ϩ current is further supported by using the CaMKII inhibitor KN-93, which inhibits the activation of CaMKII and prevents the decrease of the inactivation of the K ϩ current in our experiments. Up to now, CaMKII-regulated K ϩ channels have not been reported in nonexcitable cells. In myocytes the regulation of a transient A-type K ϩ current by CaMKII is described under control conditions that are sensitive to 23). To our knowledge, a function of CaMKII has not yet been known for disturbances caused by oxidative stress.
Beside its specific CaMKII inhibitory effect, KN-93 can exert a nonspecific action on K ϩ currents directly blocking K ϩ chan- , and for individual members of the ␦-CaMKII class (lower). As a negative control, RT-PCR of rBCEC4 RNA in the absence of reverse transcriptase was carried out (2). Where indicated, control PCR with plasmids containing the corresponding isoforms was done in parallel (positive control). Sequencing of the corresponding signal revealed the identity of a ␤-derived rBCEC4 product as an artifact (unspecific). For ␣-, ␤-, and ␦-CaMKII 2% agarose gels and for the separation of the ␥-CaMKII products, a 4% native polyacrylamide gel were used. Base pair (bp) values represent the bands of the respective standard samples. Arrows indicate the size (bp) of the detected signal. nels. In rabbit smooth muscle cells KN-93 blocks the voltagedependent K ϩ current significantly and dose dependently in a manner that is similar to its non-CaMKII inhibiting analogue . In our experiments the action of KN-93 was specific; its effect was confined to CaMKII inhibition only. This is confirmed since KN-92 has no effect on and I to /I peak value or the current amplitudes under hypoxic conditions. After hypoxia we have found a significant cell swelling that is a typical sign of hypoxic cell injury (19). Activation of the CaMKII isoenzymes observed in parallel reflects enhanced intracellular Ca 2ϩ concentration. The hypoxic increase of cytosolic Ca 2ϩ and the hypoxic cell swelling are well described in endothelial cells (28). Hypoxia experiments on cultured human endothelial cells give evidence that the Ca 2ϩ uptake is mainly due to the influx of extracellular Ca 2ϩ (46). Cellular Ca 2ϩ accumulation is known to be related to morphological changes. Laser scanning microscopic studies revealed that bleb formation induced by hypoxic stress involves regional elevation of Ca 2ϩ in human umbilical vein endothelial cells (47). In addition, hypoxia impairs the blood-brain barrier function of brain endothelial cells as indicated by increased paraendothelial permeability (18). In our investigation hypoxia conditions (18,28) that are proven to elevate the cytosolic concentration of Ca 2ϩ (28) necessary to intensify the activation of CaMKII were used.
As we show here for the first time, brain capillary endothelial cells express non-neuronal members of the CaMKII␥ and CaMKII␦ classes at both the transcript and protein levels. There are no transcripts of the primarily neuronally expressed ␣ and ␤ subunits of CaMKII (24) that are detectable. From the known CaMKII␦ isoenzymes being prominently expressed in non-neuronal excitable tissue, only ␦ 2 is detected in rBCEC4. ␦ 2 is ubiquitously expressed in cells of peripheral tissues and represents a constitutively expressed isoform in the heart (29,48,49). Transcripts of two isoenzymes of CaMKII␥ are identified in rBCEC4. Immunoblot experiments give evidence that ␥ B protein is expressed, whereas a faint band is consistent in size with the predicted molecular mass of ␥ C but may also result from a cross-reaction with the highly expressed ␦ 2 protein. ␥ B has been demonstrated to elicit higher total as well as autonomous autophosphorylation-dependent catalytic activity than ␥ C (50). Thus, our data suggest that ␥ B is the dominant isoform of the CaMKII␥ class in brain capillary endothelial cells.
In response to the elevation of intracellular Ca 2ϩ by treat- ment with ionomycin, as demonstrated in rBCEC4, ␦ 2 shows a more rapid activation in terms of autophosphorylation compared with CaMKII ␥ B . Autophosphorylation of ␥ B in response to ionomycin occurs more slowly but remains at plateau levels, whereas ␦ 2 is rapidly dephosphorylated. This different time course of activation can be due to differences in the accessibility of activating Ca 2ϩ and/or calmodulin, e.g. by distinct intracellular localization of the isoenzymes, different phosphatase action, or specific properties of individual CaMKII isoforms. It has been shown that ␥ B needs higher concentrations of calmodulin for activation, but ␥ B binds calmodulin more tightly than other CaMKII␥ isoforms (51). However, kinetic properties of members of the CaMKII␦ class have not yet been investigated. After hypoxic treatment of rBCEC4 autophosphorylation of both ␦ 2 and ␥ B is found to be significantly elevated. The response to hypoxia appears to be more pronounced for ␦ 2 . Previous findings on a rat myoblast cell line show that ␦ 2 also localizes to the plasma membrane (29). This supports the notion that ␦ 2 may be involved in the modulation of K ϩ channel properties in rBCEC. The autophosphorylation of CaMKII persisted after hypoxia for a longer time. Thus, CaMKII remains activated during measurement of the K ϩ current, and it is most likely that the hypoxia-induced changes of K ϩ channel properties are caused by CaMKII activation.
In conclusion, one can state that because of hypoxic ATP reduction (19), elevation of cytoplasmic Ca 2ϩ level (28) results in the activation of CaMKII␦ 2 , which is thought to be mainly responsible for decreasing the inactivation of the outward K ϩ current. Our results concerning the increased outward K ϩ current mediated by the activation of CaMKII support the hypothesis that this pathway may augment the counteracting regulatory effect against depolarization caused by hypoxia.