Acute NADPH oxidase activation potentiates cerebrovascular permeability response to bradykinin in ischemia-reperfusion.

Free radical generation is a key event in cerebral reperfusion injury. Bradykinin (Bk) and interleukin-1β (IL-1β) have both been implicated in edema formation after stroke, although acute Bk application itself results in only a modest permeability increase. We have investigated the molecular mechanism by assessing the permeability of single pial venules in a stroke model. Increased permeability on reperfusion was dependent on the duration of ischemia and was prevented by applying the B(2) receptor antagonist HOE 140. Postreperfusion permeability increases were mimicked by applying Bk (5μM) for 10 min and blocked by coapplying the IL-1 receptor antagonist with Bk. Furthermore, 10 min pretreatment with IL-1β resulted in a 3 orders of magnitude leftward shift of the acutely applied Bk concentration-response curve. The left shift was abolished by scavenging free radicals with superoxide dismutase and catalase. Apocynin coapplied with IL-1β completely blocked the potentiation, implying that NADPH oxidase assembly is the immediate target of IL-1β. In conclusion, this is first demonstration that bradykinin, released during cerebral ischemia, leads to IL-1β release, which in turn activates NADPH oxidase leading to blood-brain barrier breakdown.

The processes underlying the life-threatening edema following disruption of the blood–brain barrier after stroke and trauma are still obscure. Interleukin-1β (IL-1β), which can be released rapidly after a noxious stimulus to a number of cells, is one of the proinflammatory cytokines that play a role in cerebral ischemia [1] and has been implicated in subsequent destruction of brain cells [2]. These responses mainly involve gene transcription and new protein synthesis and are therefore relatively slow in onset, but are triggered by rapid signaling from the receptor to the nucleus. The kinase signaling cascades are still only partially characterized, but are increasingly recognized as involving the controlled generation of reactive oxygen species (ROS) acting as signaling intermediates, in contrast to their traditional cytotoxic role [3-6]. There is accumulating evidence that reperfusion after ischemia results in a surge of free radical generation [7] that sets in train a sequence of events that lead to highly destructive cerebral edema [8].

Inflammatory mediators, such as bradykinin, histamine, angiotensin II, and substance P, act via endothelial G-protein-coupled receptors (GPCR) and have been characterized in terms of rapid signaling responses, including elevated Ca2+ or cGMP; secretion of vasoactive mediators such as NO, PGI2, and EDHF within seconds or minutes; and transient increased microvascular permeability. It is clear, however, that at least some GPCR agonists (e.g., AT II and thrombin) can in addition cause longer term changes in endothelial cell phenotype and can stimulate the generation of reactive oxygen species [9-11]. Bradykinin, in particular, is one substance that has been implicated in the development of the damage that follows cerebral ischemia [12], including cerebral edema, but as acute bradykinin administration results in only a small permeability change in the cerebral endothelium [13], the mechanisms underlying subsequent longer term damage remain to be elucidated.

Little is known about the potential interactions between the IL-1β and the GPCR signaling pathways in the vasculature. This study provides evidence for bradykinin (Bk) and IL-1β being released after ischemia–reperfusion injury into the rat cerebral microcirculation to activate independent signaling pathways that interact, resulting in enhanced cerebrovascular permeability. This study, to our knowledge, documents the first report of novel interactions between these inflammatory mediators to potentiate free radical generation.

Methods

Animals and experiments on single cerebral microvessels

The methods used in this study, and its theoretical basis, have been described previously [14]. Experiments were performed on Wistar rats (age 20–30 days, of either sex) within guidelines directed by the UK Home Office Animals (Scientific Procedures) Act, 1986, and local ethics committee guidelines, which conform with the Guide for the Care and Use of Laboratory Animals published by the U.S. National Institutes of Health (NIH Publication No. 85-23, revised 1996). Animals were anesthetized by an intraperitoneal injection of 60 mg kg− 1 body wt sodium pentobarbital diluted in water (25% w/v) and maintained by supplementary injection of 10% of the original dose when necessary. At the end of the experiment the animals were killed by administration of an overdose of the anesthetic.

Single-vessel experiments

The pial microcirculation of the surface of the brain was viewed through a Zeiss ACM fluorescence microscope under 525 to 535 nm illumination, and sulforhodamine B (580 Da) was given via a bolus injection into the carotid artery. The fluorescent signal was analyzed using a video-densitometer. Permeability was measured in a single pial venular capillary (diameter between 8 and 28 μm) from the rate of loss of dye trapped by a glass-occluding probe [14]. Transient ischemia was induced by the injection of degradable starch microspheres [15].

Analysis of bradykinin in brain superfusate

Samples were collected from a specially made cup, which was sealed to the skull after the dura and arachnoid were removed. The cup volume was about 100 μl, and 50-μl aliquots of artificial cerebrospinal fluid, containing captopril (1 μM) and phosphoramidon (200 μM) to block kinin-destroying enzymes (without which there was too little Bk to measure), were removed and replenished at 5-min intervals before and after administration of the starch microspheres for a total of 70 min in all cases. Possible contamination by plasma was monitored from a prior intravenous injection of rhodamine–albumin and examination of aliquots of sample fluid in a microfluorimeter. Samples that gave a reading higher than background were discarded. The fluid samples were stored at − 20 °C until analyzed. Kinins were extracted by using Sep-Pak Vac C18 cartridges and bradykinin-like immunoreactivity was measured [16]. The radioimmunoassay range was from 8 to 4000 fmol per tube. The Bk antibody shows > 80% cross-reactivity with kallidin and about 80% with Ile-Ser-Bk, which is one of the kinins found in rat brain [17], but none with des(Arg9)-Bk.

Measurement of ROS generation in brain tissue

Wistar rats of weight between 200 and 250 g were killed by exposure to CO2 and the brains were removed. The cerebellum was discarded and the cerebral hemispheres were cut into pieces of approximately 0.2 g and placed in the wells of a 96-well plate in 200 μl balanced salt solution. Some tissues were pretreated with 30 pM IL-1β in balanced salt solution (BSS; NaCl 192.5 mM, KCl 5 mM, MgCl2 1 mM, Hepes 10 mM, glucose 10 mM, CaCl2 1 mM, and 1% BSA) for 10 min. The BSS was removed and replaced with 50 μl Amplex red (100 μM and 0.2 U/ml horseradish peroxidase) solution [18]. The agonist was added to the wells and the fluorescence measured at 15 min in a plate reader (Titertek Fluoroskan II) with 544 nm excitation and 590 nm emission wavelengths. Each piece of tissue was weighed and the readings were normalized to the weight.

All chemicals were purchased from Sigma, apart from the Amplex red, which was purchased from Molecular Probes.

Statistics

Unless otherwise stated, the results are expressed as the mean ± standard error of the mean. Sigmoidal concentration–response curves were fitted using Prism version 4.03 for Windows (GraphPad Software, San Diego, CA, USA).

Results

Microsphere-induced ischemia

Starch microspheres were infused into the internal carotid artery and the permeability of a selected venular capillary, measured before the infusion, was measured again as soon as flow recommenced. The blockage lasted from 5 to 60 min and the permeability increased with the duration of the ischemia at the rate of 24.0 ± 1.9 × 10− 6 cm s− 1 h− 1 (Fig. 1A). This rate of increase was much reduced (11.9 ± 0.6 × 10− 6 cm s− 1 h− 1) when a free radical scavenging mixture of superoxide dismutase and catalase (100 U ml− 1 each) was included in the superfusing solution during the blockage and reduced much more (to 4.0 ± 1.0 × 10− 6 cm s− 1 h− 1) with the bradykinin B2 receptor antagonist HOE 140 (1 μM). HOE 140 application during the reperfusion phase reversibly reduced the increased permeability (Figs. 1B and C) after ischemia–reperfusion: the mean permeability of 10.8 ± 2.4 × 10− 6 cm s− 1 being reduced to 2.7 ± 0.43 × 10− 6 cm s− 1. This evidence for bradykinin being continuously generated after ischemia–reperfusion was confirmed by collecting fluid from a cup placed over the cranial window: the total immunoreactive bradykinin formed in the 70 min after ischemia was also proportional to the duration of the ischemia (Fig. 1D).

Fig. 1

The role of bradykinin in cerebrovascular permeability after cerebral ischemia. (A) The permeability of pial venules increased following reperfusion after a defined period of ischemia. Including SOD and catalase in the superfusate reduced this permeability increase, whereas the bradykinin B2 antagonist HOE 140 resulted in a greater reduction (regression coefficients are all significantly different from each other at p < 0.001, analysis of covariance; one vessel from one animal for each data point). (B) Permeability of a single vessel followed before and after ischemia–reperfusion showing that HOE 140 application reversibly reduces the permeability increase. The open symbol indicates the permeability before ischemia. (C) Summary of HOE 140 post-ischemia–reperfusion permeability reduction similar to (B); in which the permeability immediately before and after HOE 140 application is shown from eight venules from eight animals (p < 0.005, paired t test). (D) Immunoreactive bradykinin increase in fluid collected from the brain surface over 70 min from the beginning of ischemia increased significantly with the duration of ischemia (p < 0.05, each point from a single animal).

We have previously shown that the response to acutely applied bradykinin is stable in this preparation for up to 2 h for concentrations up to 50 μM [13], and the foregoing data illustrate that bradykinin is responsible for a considerable proportion of the raised permeability during reperfusion, which is much greater than the maximum response to acutely applied bradykinin [13].

Bradykinin-dependent permeability change

The idea that bradykinin plays a key role in the initial disruption of the blood–brain barrier after ischemia–reperfusion was examined by applying it at a high concentration (5 μM) for 10 min. The permeability response was similar in magnitude to an acute application [13], but did not reverse when the bradykinin was removed; permeability remained raised, with a further increase that commenced 30 min after the bradykinin application (Fig. 2A). This pattern was altered when interleukin-1 receptor antagonist (IL-1ra) was coapplied with bradykinin, and permeability returned to close to its initial value 10 min after the bradykinin + IL-1ra mixture was removed, which indicates that bradykinin may be a factor in the rapid release of IL-1β.

Fig. 2

Role of IL-β in permeability response to bradykinin. (A) High concentration of Bk (5 μM) applied for 10 min resulted in raised permeability that did not reverse after its removal, but remained raised. Permeability increased, without further Bk addition, 30 min after its first application. IL-1ra (5 nM, open symbols) coapplied with Bk resulted in a similar permeability increase and similar sustained increase, but then fell. The differences beyond the 35-min point were significantly different (p < 0.001, t test; 4 vessels from 4 animals in each group). (B) Permeability response to a brief low-concentration Bk application (50 nM) that was just detectible was much enhanced after 10 min IL-1β (30 pM) application (4 vessels from 4 animals). This potentiated response to Bk was stable for 30 min after IL-1β was applied and thereafter increased further. Cycloheximide (CHX; 100 μM; 5 vessels from 5 animals) treatment did not affect the initial IL-1β potentiation of the response to Bk, but did block the subsequent increase. All the Bk and BK + CHX responses after 20 min were significantly greater than the controls (p < 0.05, paired t test), and those at 50 and 90 min were significantly different (p < 0.001). (C) Concentration–response curves for bradykinin before and after treatment with IL-1β. The log EC50 was significantly left-shifted (p < 0.05, F = 11.38 at 1,7 df; data from 10 to 12 vessels from 12 animals).

The possible role of IL-1β in potentiating the permeability response to Bk was further explored by applying it (30 pM) to the brain surface for 10 min and testing the response to acute application of a low Bk concentration (50 nM) that gave a just discernible stable response (see Fig. 1 in Ref. [13]). Fig. 2B shows that IL-1β application itself resulted in a small permeability increase, but after its removal the response to bradykinin was potentiated. There are apparently two phases in this response: an initial stable increase that lasts until 30 min after the start of IL-1β application and a second phase that increases steadily to reach 4.0 ± 0.6 × 10− 6 cm s− 1 at 90 min. The likelihood that this secondary increase was due to the formation of new protein was confirmed by repeating the experiment in the presence of cycloheximide (100 μM), when the permeability response to acutely applied bradykinin (50 nM) remained stable at 1.1 ± 0.05 × 10− 6 cm s− 1 for 90 min (Fig. 2B). All subsequent experiments on single microvessels described below were carried out in the presence of cycloheximide (100 μM) to enable concentration–response curves to be generated without the time-dependent changes. In these experiments that effect of 10 min treatment with IL-1β shifted the bradykinin permeability concentration–response curve to the left by about 3 orders of magnitude (log EC50 from − 6.7 ± 0.28 before IL-1β to − 9.7 ± 0.25 after; Fig. 2C), without affecting the maximal permeability response.

Signaling pathways

We have previously shown that the permeability response to acute Bk application depends on the formation of ROS, as it was inhibited by a combination of superoxide dismutase (SOD) and catalase (100 U ml− 1 each) [13] and furthermore was independent of nitric oxide formation and soluble guanylyl cyclase (sGC) activation [19]. In these experiments SOD and catalase also reduced the response after treatment with IL-1β (Fig. 3A). The possibility that IL-1β treatment activated the nitric oxide–sGC pathway was examined by inhibiting endothelial NO synthase with L-N(G)-Nitroarginine methyl ester (L-NAME), but this did not alter the potentiated permeability response, nor did the sGC inhibitor 1 H-[1,2,4]-oxadiazolo-[4,3]-quinoxalin-1-one (ODQ) alter the post-potentiated response to Bk (Fig. 3B). It is therefore not surprising that the response to histamine, which uses this NO-dependent pathway [20], was not affected by IL-1β treatment (Fig. 3B). Fig. 3C shows that the acute response to a high concentration of Bk (5 μM) was not affected by the protein kinase C (PKC) antagonist calphostin C, and the possibility that IL-1β treatment resulted in activation of such a PKC-dependent pathway was examined by coapplying calphostin C with IL-1β. The potentiated response to a low concentration of Bk (50 nM) after IL-1β was blocked. On the other hand, calphostin C had no effect on the response to Bk when it was applied after IL-β treatment. This indicates that PKC was required to initiate the element that was responsible for the potentiation, but not for its operation.

Fig. 3

Possible signaling pathways in the potentiated response. (A) The permeability response to bradykinin (50 nM), after pretreatment with IL-1β, was significantly attenuated in the presence of SOD and catalase (paired data from eight vessels from eight rats; p < 0.001, t test). (B) IL-1β treatment affected neither the baseline nor the response to histamine (1 μM). The IL-1β-potentiated permeability response to Bk was unaffected by the soluble guanylyl cyclase inhibitor ODQ (1 μM; four vessels from four rats). (C) No change in the acute response to Bk when coapplied with the PKC inhibitor calphostin C, but when calphostin C was coapplied with IL-1β the potentiated response to Bk was blocked. Calphostin C had no effect on the potentiated Bk response when it was applied with Bk after IL-1β treatment (six vessels from six animals). (D) Free radical generation from brain tissue measured by Amplex red. The tissue pretreated with IL-1β produced significantly more fluorescence with bradykinin incubation for 15 min (eight brains; p < 0.001, ANOVA with Bonferroni's multiple comparison test).

Direct measurement of free radical formation in single microvessels proved not to be possible, but as a proof of principle we established that IL-1β treatment enhances free radical generation in small pieces of fresh rat brain tissue (Fig. 3D).

As we reported previously, the acute permeability response to Bk depends on arachidonic acid formation via phospholipase A2 (PLA2) activation and subsequent ROS generation via cyclooxygenase and lipoxygenase [13]. To examine whether IL-1β potentiates free radical formation by making additional PLA2 isoforms available we tested the effects of PLA2 inhibitors. Bromoenol lactone (BEL; a calcium-independent PLA2 inhibitor when used at 0.4 μM) had no effect (100 nM Bk post-IL-1β alone, 0.8 ± 0.09 × 10− 6 cm s− 1, and Bk with BEL, 0.8 ± 0.2 × 10− 6 cm s− 1, four vessels from four rats for each), whereas palmitoyl trifluoromethyl ketone (PACOCF3), which at 100 μM inhibits all PLA2 isoforms [21] and was previously shown to completely block the response to Bk when there was no IL-1β pretreatment [13], reduced the response to Bk (100 nM) post-IL-1β from 0.6 ± 0.1 × 10− 6 to 0.3 ± 0.08 × 10− 6 cm s− 1 (four vessels from four rats for each; see Fig 4A). Thus it seems that IL-1β treatment resulted in an additional source of free radical generation by Bk. This idea was tested by examining the effects of the flavoprotein inhibitor diphenylene iodonium (DPI; 100 μM), which had no effect on the response to Bk before IL-1β treatment (Bk 1 μM, 0.8 ± 0.07, n = 5; with DPI 100 μM, 0.8 ± 0.10, n = 4), but did significantly reduce the response to Bk (100 nM) after IL-1β (0.8 ± 0.04 × 10− 6 cm s− 1, n = 3, to 0.4 ± 0.04 × 10− 6 cm s− 1, n = 4). The combination of PACOCF3 with DPI almost abolished the response to 10 nM Bk after IL-1β from 0.7 ± 0.08 × 10− 6 to 0.1 ± 0.01 × 10− 6 cm s− 1 (n = 4 for each). That the combination of these high doses of PACOCF3 and DPI reduced the permeability response to bradykinin after IL-1β by a significantly greater degree than either antagonist alone (p < 0.01, ANOVA) is evidence that pretreating with IL-1β for just 10 min produced an additional source of bradykinin-stimulated ROS formation.

Fig. 4

Possible sources of ROS in the potentiated response. (A) The IL-1β-potentiated response to Bk was reduced when either the PLA2 antagonist PACOCF3 or the flavoprotein inhibitor DPI was applied separately and was totally blocked when they were applied together. All experiments were paired; four vessels from four animals. (B) The leftward shift of the concentration–response curve for permeability was completely blocked when apocynin (100 μM) was coapplied with IL-1β (30 pM; five vessels from five animals).

Application of apocynin, a more specific inhibitor of NADPH oxidase than DPI, which has been suggested to act by preventing its assembly, blocked the IL-1β potentiation (the log EC50 of the bradykinin permeability responses being − 6.52 ± 0.32, − 9.63 ± 0.27, and − 6.48 ± 0.34 for bradykinin alone, after IL-1β treatment, and after IL-1β coapplied with apocynin 100 μM, respectively; Fig. 4B).

It is possible that because substances were applied to the brain surface the responses to IL-1β were due to cells other than endothelial cells. This was addressed by applying IL-1β to the luminal side via bolus injections of dye solution containing IL-1β (60 pM, as the final solution was mixed with blood) and trapping the solution in a selected vessel for a total of 10 min (two or three applications). Permeability responses to brain-side Bk (0.5 μM) increased on each occasion (Fig. 5B). Similarly, the effects of free radical scavenging by SOD and catalase were effective when applied to the brain side while Bk was applied to the vessel lumen (Fig. 5A), which shows that this mixture of scavengers is effective whether the free radicals are generated inside or outside the cell membrane.

Fig. 5

Localization of IL-1β effects to the endothelium. (A) The permeability response to luminal bradykinin was attenuated by abluminal free radical scavenging (paired experiments; four vessels from four animals; p < 0.01, t test). (B) The permeability response to abluminal bradykinin (0.5 μM) was increased after IL-1β application to the vessel lumen (paired experiments; four vessels from four animals; p < 0.01, t test).

Discussion

Both bradykinin and IL-1β have been implicated in edema formation after stroke and experimental cerebral ischemia, and antagonizing each has been suggested as a therapeutic target for the treatment of stroke [22,23]. The experiments described here provide further evidence for a role for bradykinin and IL-1β in the early burst of reactive oxygen species after cerebral ischemia–reperfusion. We have also shown that a high concentration of bradykinin applied over a number of minutes results in effects that are blocked by interleukin-1 receptor antagonist. This implies that IL-1β is rapidly released from brain tissue, which then activates NADPH oxidase to rapidly potentiate the permeability response to bradykinin.

Starch microspheres were used to produce the reversible ischemia because they produce a total block of a vascular territory by virtue of occluding several small arterioles in parallel. This contrasts with the middle cerebral artery occlusion technique, which generates a reduction in blood flow, but not necessarily a total cessation in the pial vessels studied here. The permeability increase measured within 2 or 3 min of reperfusion after the severe ischemia produced by the starch microsphere infusion was reduced by scavenging free radicals, which is consistent with the reported surge in free radical production that occurs on reperfusion after 1 h of middle cerebral artery occlusion [23]. There are several lines of evidence that indicate bradykinin is the probable source of these free radicals. We have previously shown that bradykinin application results in a free radical-dependent permeability increase in normal rats [13], and in the present experiments the B2 receptor antagonist, HOE 140, applied after reperfusion produced a fall in permeability that reversed as soon as it was removed (Fig. 1B). This indicates that bradykinin formation continued after reperfusion and was not limited to the ischemic period. The longer the blockage, the greater the cerebrovascular permeability increase and bradykinin formation (Fig. 1D). Bradykinin has been associated with free radical generation and associated tissue damage on reperfusion in other preparations, too. Thus early on in reperfusion bradykinin B2 receptor activation in the ischemic kidney enhances malondialdehyde and H2O2 formation and reduces the GSH/GSSG ratio as well as increasing tubular damage [24]. Furthermore, a study [12] designed to examine changes in the kallikrein–kinin system after middle cerebral artery occlusion (MCAO) and reperfusion showed sustained bradykinin release that peaked at 12 h postreperfusion. This was accompanied by an increased B2 receptor density possibly requiring protein synthesis underlying the later phase of permeability increase (see Figs. 2A and B).

The concept that bradykinin formation itself is a key trigger for cerebrovascular dysfunction was first suggested by Unterberg and Baethemann [25]. This hypothesis, however, remains controversial because although a number of studies have shown that inhibiting the B2 receptor reduces cerebral damage [12,26], another study reported that B2−/− mice are more severely damaged after 90 min of MCAO [27]. In the present model we have clear evidence that topical application of a high concentration of bradykinin to the brain surface results in rapid release of IL-1β (Figs. 2A and B). It is possible that bradykinin acts on astrocytes to release ATP [28], which in turn rapidly releases IL-1β from microglia via vesicle shedding [29]. On the other hand, knocking out or blocking the P2X7 receptors for ATP had no effect on ameliorating experimental cerebral ischemia [30]. Interestingly, as bradykinin has recently been shown to reduce lipopolysaccharide-stimulated release of IL-1β and TNF-α from isolated microglia [31], it is possible that cytokine release after bradykinin production is different in different preparations and depends on the cellular distribution of B2 receptors.

There were two phases in the potentiation of cerebrovascular permeability mediated by bradykinin produced by 10 min IL-1β application. The second phase depended on new protein formation involving possibly the formation of new kinin receptors [32]. The large leftward displacement of the bradykinin concentration–response curve is a novel finding (Fig. 2D).

The response to histamine was unaffected by IL-1β pretreatment, indicating that the endothelium had not become generally more responsive to all agonists (Fig. 3B) and is consistent with the observation that IL-4 pretreatment but not IL-1β treatment increases subsequent endothelial responses to histamine [33]. The permeability response to acutely applied bradykinin is mediated via free radical generation and particulate, not soluble, guanylyl cyclase, and conversely the response to histamine is via soluble, not particulate, guanylyl cyclase [19]. The possibility that IL-1β treatment led to bradykinin recruiting soluble guanylyl cyclase for cGMP production was excluded by our finding that the potentiated response was insensitive to the soluble guanylyl cyclase inhibitor ODQ. Scavenging free radicals inhibited the IL-1β-potentiated response as well as the response before IL-1β [13], which points to an additional source of free radicals. The possibility that this was due to additional activation of PLA2 by IL-1β treatment, as observed in fibroblasts [34] and after cerebral ischemia [35], was tested by using PACOCF3 at a sufficiently high concentration as to block all PLA2 isoforms [21]. We had previously shown that this in the absence of IL-1β pretreatment resulted in a complete abolition of the permeability response [13], but here, after IL-1β, about 50% of the response remained. The flavoprotein inhibitor diphenylene iodonium also reduced the IL-1β-potentiated response, but had no effect on the permeability increase with no IL-1β pretreatment. The response was blocked when the two inhibitors were given together, thus supporting the concept that an additional ROS-generating pathway had been recruited.

It is likely that NADPH oxidase is a component of this additional ROS pathway, as both apocynin application and blocking of PKC during IL-1β treatment abolished the effect. The rapidity of the inhibition by apocynin is probably due to the presence of peroxidases that result in the formation of the active apocynin dimer [36]. PKC is required for p47phox phosphorylation, a key element of the preparation of NADPH oxidase for its activation [21], and it is not required for the acute response to bradykinin alone [37]; blocking it after NADPH oxidase assembly did not affect the potentiated response once it had been established.

Bradykinin, IL-1β, and other substances were usually applied on the brain surface to the abluminal side of the microvessels, which exposed other cells to their effects. Some experiments were carried out to investigate whether the effects of IL-1β were due to endothelial cells alone by applying it luminally by injecting it with a dye bolus and so limiting its access to the apical surface of the endothelium. The potentiation of the permeability response was not altered (Fig. 5B), and furthermore, applying the free radical-scavenging mixture of SOD and catalase to the abluminal surface was effective in reducing the permeability response to bradykinin applied to the lumen (Fig. 5A). It is likely that the effects of superoxide are not as tightly localized as once believed, as it has been shown that superoxide itself permeates cell membranes via anionic channels [38], and it is also possible that intracellular SOD generates H2O2 that will permeate cell membranes and so be subject to the extremely effective free radical-scavenging system provided by the mixture of SOD and catalase.

The direct measurement of free radical generation in brain tissue shows that the response to bradykinin became detectable only after the tissue had been exposed to IL-1β for 10 min. This time course is consistent with the finding that TNF-α applied to endothelial cells results in p47phox phosphorylation, and 5 min exposure leads to about 70% of the maximal phosphorylation [10]. IL-1β has also been shown to increase the permeability response to a low concentration of bradykinin in the hamster cheek pouch [39].

IL-1β alone rapidly (within 10 to 15 min of its application) increases superoxide formation in both cultured endothelial cells [40] and retinal pigment epithelial cells, the latter with evidence that NADPH oxidase activation was involved [41]. These data are consistent with the present observation that IL-1β itself results in a small permeability increase (see Figs. 2C and 3A). The potentiation of the response to bradykinin after just a brief exposure to a modest concentration of IL-1β is a novel finding. The indication that NADPH oxidase assembly is the key step in this process has important implications for understanding the development of cerebral edema as a consequence of the cascade of events that follow ischemia–reperfusion.

In conclusion, we have investigated the mechanisms involved in cerebral edema formation after ischemia and reperfusion. It has been clear for some time that bradykinin and IL-1β, with other substances, are involved in a cascade of events that leads to a large disruption of the blood–brain barrier [42,43], which is the basis of vasogenic edema. Here we have shown for the first time that bradykinin, released during ischemia, leads to IL-1β release, which in turn results in rapid potentiation of the permeability response to bradykinin via assembly of NADPH oxidase.