Reactive oxygen species facilitate the EDH response in arterioles by potentiating intracellular endothelial Ca2+ release

There is abundant evidence that H2O2 can act as an endothelium-derived hyperpolarizing factor in the resistance vasculature. However, whilst scavenging H2O2 can abolish endothelial dependent hyperpolarization (EDH) and the associated vascular relaxation in some arteries, EDH-dependent vasorelaxation can often be mimicked only by using relatively high concentrations of H2O2. We have examined the role of H2O2 in EDH-dependent vasodilatation by simultaneously measuring vascular diameter and changes in endothelial cell (EC) [Ca2+]i during the application of H2O2 or carbachol, which triggers EDH. Carbachol (10 µM) induced dilatation of phenylephrine-preconstricted rat cremaster arterioles was largely (73%) preserved in the presence of indomethacin (3 µM) and l-NAME (300 µM). This residual NO- and prostacyclin-independent dilatation was reduced by 89% upon addition of apamin (0.5 µM) and TRAM-34 (10 µM), and by 74% when an extracellular ROS scavenging mixture of SOD and catalase (S&C; 100 U ml−1 each) was present. S&C also reduced the carbachol-induced EC [Ca2+]i increase by 74%. When applied in Ca2+-free external medium, carbachol caused a transient increase in EC [Ca2+]i. This was reduced by catalase, and was enhanced when 1 µM H2O2 was present in the bath. H2O2 -induced dilatation, which occurred only at concentrations ≥100 µM, was reduced by a blocking antibody to TRPM2, which had no effect on carbachol-induced responses. Similarly, iberotoxin and Rp-8bromo cGMP reduced the vasodilatation induced by H2O2, but not by carbachol. Inhibiting PLC, PLA2 or CYP450 2C9 each greatly reduced the carbachol-induced increase in EC [Ca2+]i and vasodilatation, but adding 10 µM H2O2 during PLA2 or CYP450 2C9 inhibition completely restored both responses. The nature of the effective ROS species was investigated by using Fe2+ chelators to block the formation of ∙OH. A cell permeant chelator was able to inhibit EC Ca2+ store release, but cell impermeant chelators reduced both the vasodilatation and EC Ca2+ influx, implying that ∙OH is required for these responses. The results indicate that rather than mediating EDH by acting directly on smooth muscle, H2O2 promotes EDH by acting within EC to enhance Ca2+ release.


Introduction
Endothelium-dependent hyperpolarization (EDH) plays an important role in the vasodilatation of small arteries and arterioles induced by diverse stimuli including locally released factors, blood flow, and low pH [1]. EDH was initially viewed as resulting from the release of a diffusible endothelial factor, termed EDHF [2], but is now also thought to involve the transmission of endothelial cell hyperpolarization to the underlying vascular smooth muscle cells (VSMCs) via myoendothelial gap junctions [3]; the relative importance of each mechanism is thought to vary between different arteries/arterioles and also depends on the degree of pre-existing vascular excitation [4]. In either case, calcium-activated potassium channels K Ca 2.3 and K Ca 3.1 (SK Ca and IK Ca respectively) have been shown to play a significant part in the EDH response (EDHR).
Intriguingly, although H 2 O 2 dilated endothelium-intact human submucosal intestinal microvessels, it caused constriction following endothelial denudation, a finding which led to the suggestion that H 2 O 2 promotes the release of a separate EDHF instead of itself acting as an EDHF [15]. It was subsequently shown by Griffith and co-workers that the application of 100 μM H 2 O 2 , or the oxidizing agent thimerosal, to rabbit aortic valve endothelial cells (EC) enhanced intracellular Ca 2 þ release evoked by the SERCA inhibitor cyclopiazonic acid (CPA). Moreover, CPA-induced vasodilatation of rabbit ileac arteries was attenuated by application of catalase. They proposed that H 2 O 2 could potentiate EDH-dependent vasodilatation by increasing IP 3 -dependent Ca 2 þ release [19]. Soon afterwards this group demonstrated that applying H 2 O 2 concentrations of 10-30 μM also potentiated the EDH-associated vasodilation caused by acetylcholine in rabbit ileac arteries [20].
These studies suggest strongly that H 2 O 2 acts to enhance rather than mediate the EDHR, and pose a number of important questions which are currently unanswered. As the ability of H 2 O 2 to facilitate the EDHR has not been compared to its direct vasodilator effects in any single preparation, it remains unknown which of these effects are likely to be of predominant importance during the EDHR. Also, the pathways leading to H 2 O 2 production within EC which would facilitate increases in EC [Ca 2 þ ] i during the EDH response have not been defined. In addition, it is not clear whether physiological stimuli which have been shown to increase ROS production in EC lead to increases in the EDHR as would be predicted by this mechanism.
We have investigated these issues in arterioles in the intact rat cremaster circulation, as this preparation exhibits robust, sustained and consistent dilatations to both carbachol and H 2 O 2 , and permits endothelial Ca 2 þ concentration changes to be measured simultaneously. We characterized the effects of anti-oxidants and blockers of potential ROS-producing pathways on increases in EC [Ca 2 þ ] i and the associated NO-and PGI 2 -independent vasodilation evoked by the muscarinic receptor agonist carbachol. In addition, we carried out a comparison of the effects of carbachol and exogenously applied H 2 O 2 in order to determine whether H 2 O 2 might be acting as an EDHF. We also assessed whether EDH and ROS contribute to the vasodilatation induced by the inflammatory mediator cytokine IL-1β. Our results show that under basal conditions, endogenous ROS production involving CYP92C, phospholipase A 2 and hydroxyl radical greatly facilitates EC Ca 2 þ release and EDH-associated vasodilation. A brief pre-incubation with IL-1β further enhances the responses to carbachol. Although H 2 O 2 also dilated these arterioles directly, it only did so at high concentrations, and via mechanisms which differ markedly from those evoked by carbachol.

Methods
This study conforms with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 85-23, revised 1996) and is in accordance with UK Home Office regulations (Animals Scientific Procedures) Act, 1986. 4-6 week old male Wistar rats (Charles River, UK) weighing 80-100 g were killed by exposure to an increasing concentration of CO 2 followed by cervical dislocation. A midline incision was made along the abdomen and the skin and muscle layer were retracted to expose the underlying organs. The inferior vena cava was tied off above the bifurcation into the common iliac veins and below that an incision was made to allow blood to drain from the lower extremities. A 0.61 mm fine bore polythene cannula was inserted down to the bifurcation into the common iliac arteries via the abdominal aorta and secured with thread. The left common iliac and the right femoral and internal iliac arteries were ligated to ensure that perfusion was only to the right external iliac artery supplying the cremaster artery and the cremaster muscle. This was then perfused through the cannula with a stabilizing solution (10 Mg 2 þ , 110 NaCl, 8 KCl, 10 HEPES, 1 CaCl 2 þ all mM) containing heparin (30 U/ml) with isoproterenol 10 mM buffered to pH 7.0 70.05 for 10 min to remove all blood.
The cremaster muscle was prepared by a midline incision along the scrotum to expose the right testes. Skin and connective tissue were carefully removed from the underlying cremaster muscle, which was then cut longitudinally along the anterior surface from the apex to the inguinal canal while keeping the vascular supply and drainage at the base intact. The testis was separated from the cremaster muscle and was pulled through the inguinal canal so as not to obstruct the preparation. The muscle was spread out flat onto a clear Sylgard disc using forceps and secured using microhistology pins, and the cardioplegic solution perfusing the circulation was then replaced for the duration of the experiment with PSS containing 0.1% albumin delivered by a gravity controlled reservoir. The preparation was moved to the stage of a Leitz Intravital microscope equipped with Orthoplan optics, and was superfused with PSS at 37°C at a flow rate of 2.5 ml/min. PSS contained the Na þ channel blocker lidocaine (20 mg l À 1 ) to block neural activity and keep the cremaster muscle from contracting.
Following a 30 min equilibration period an appropriate 2nd order arteriole (diameter 50-100 mm) was constricted by superfusion with 30 mM phenylephrine (PE) in combination with 300 mM N(G)-nitro-L-arginine methyl (L-NAME) and 3 mM indomethacin to inhibit NO and PGI 2 mediated vasodilatation. Constriction was maintained for 20 min to ensure its stability, and the muscarinic receptor agonist carbachol (10 mM) was then added for 2 min to the superfusate to elicit vasodilation. Arterioles were used for experiments only if a vasodilatation of 450% (measured as described below) was observed.
During the experiments PE, L-NAME and indomethacin were present throughout. Drugs used to characterize the EDHR were added to the superfusate for 2-20 min before, and during, the application of carbachol. In experiments examining the effect of IL-1β, the preparation was superfused for 15 min with PSS containing the cytokine at a concentration of 100 pM. IL-1β was then washed out briefly and carbachol was applied.

Diameter measurements and EC [Ca 2 þ ] i measurements
Arterioles were visualized with a x40 water-immersion objective lens (allowing diameter measurements to be taken to an accuracy of 1 mm) and were perfused for 1 h at room temperature with PSS containing 10 μM Fura-PE 3 AM, 1% DMSO, 0.02% pluronic acid and 1% albumin in order to selectively load the endothelium with furaPE3. The preparation was then perfused with PSS and the superfusate temperature was raised to 37°C. Following a 15 min equilibration and washout period, arterioles were illuminated alternately at 360 and 380 nm at 1 Hz using a filter wheel (Cairn Research Ltd, UK) and images were recorded at 4530 nm using an Intensified ISIS camera (Photonic Sciences, UK). Images were analyzed using ImageHopper (Samsara Research: Dorking, Surrey, UK) software. Diameter was measured using the inner margin of the 380 nm signal. A semi-quantitative calculation of the endothelial Ca 2 þ concentration (hereafter termed EC [Ca 2 þ ] i ) was made by subtracting the average background (adjacent to the arterioles) signal from the image, normalizing the signals from the vessel during the 360 and 380 nm excitation intervals, and calculating the 360/380 ratio at each pixel. PE had no effect on the 360/380 ratio, which indicates that FuraPE3 was confined to the endothelium. Furthermore, carbachol and TRAM-34 co-applied with apamin produced changes in the 360/380 ratio that would be expected from selective FuraPE3 endothelium loading.

Presentation of data and statistical analysis
Endothelial cell [Ca 2 þ ] i is represented as the peak % increase in the Fura-PE3 360/380 ratio from the baseline. Vasodilatation is represented as the maximum % reversal of PE-induced constriction. Carbachol responses were recorded under control conditions and then again following an intervention (e.g. application of an inhibitor), so that each arteriole acted as its own control and results could be analyzed using a two-tailed paired t-test. It was possible to do so because the responses evoked by carbachol were stable when repeated (rise in Fura-PE3 360/380 ratio and vasodilation of 20.5 72.4% and 75.3 73.5%, respectively, during first application of carbachol; rise in Fura-PE3 360/380 ratio and vasodilation of 22.1 72.6% and 72.5 73.2%, respectively during second response to carbachol applied 15 min later, n¼ 21, ns). Bar charts represent mean 7S.E.M values.

Role of ROS in EDH-associated carbachol-induced vasodilation of rat cremaster arterioles
PE (30 mM) application to cremaster resulted in a sustained constriction (3673%, n¼5), and subsequent carbachol (10 mM) application resulted in a well-maintained vasodilatation, amounting to a 72.672.4% reversal of the constriction (n¼55). Fig. 1A shows an example of this, and Figs. 1B and 1C illustrate the results of 4 paired experiments in which the vasodilatation of arterioles and their accompanying venules was measured. The vasodilatation was reduced somewhat in the presence of L-NAME (300 mM) and indomethacin (3 mM) (Fig. 1B). Interestingly, this was less in the venules than the arterioles (arteriolar vasodilatation reduction 3071.3%, venular reduction 1971.7%; po0.001 t-test, 8 paired vessels). Addition of apamin (0.5 mM) with Tram-34 (10 mM) to block endothelial SK Ca (K Ca 2.3) and IK Ca (K Ca 3.2) channels respectively, resulted in a much greater reduction: a further 6273.3% in arterioles and 7071.3% in venules (po0.05 t-test, 4 paired vessels). When a free radical scavenging mixture of superoxide dismutase (SOD) and catalase (S&C; both 100 U/ml) was used instead of Tram-34 and apamin (Fig. 1C) there was a similar reduction in carbachol-induced vasodilatation (arterioles 4975.3% and venules 6275.2% 4 paired vessels, ns). The carbachol-induced arteriolar vasodilatation was accompanied by a sustained increase in EC [Ca 2 þ ] i which was not affected by the presence of L-NAME and indomethacin (24.5 7 5.4 vs 24.1 74.6% increase, n¼4, ns). All experiments described hereafter were carried out in the presence of L-NAME and indomethacin in order to allow the EDH-associated vasodilatation to be recorded in isolation.
As shown in Fig. 2, both the rise in EC [Ca 2 þ ] i and the vasodilatation induced by carbachol were greatly diminished in the presence of S&C. Fig. 2A illustrates images of the 360/380 fluorescence emission ratio in a Fura PE3 loaded cremaster arteriole in the presence of PE (upper left), and after 10 μM carbachol was applied (upper right). The arteriole was subsequently washed with PSS, superfused with PE in the presence of S&C (lower left), and again exposed to carbachol (lower right). The effect of S&C on the carbachol-induced increase in EC [Ca 2 þ ] i was sustained, as shown in Fig. 2B. Fig. 2C shows that in six paired vessels that S&C application reduced the carbachol-induced vasodilatation (left) and increase in EC [Ca 2 þ ] i (right) on each occasion.
Carbachol stimulates muscarinic M 1 and M 2 receptors which will result in activating phospholipase C and phospholipase A 2 respectively. Inhibiting PLC with 3 μM U73122 resulted in much reduced vasodilatation and Ca 2 þ responses (to 11.671.1% and 13.871.8%, respectively), whereas the inactive U73122 analogue U73343 (3 μM) had no effect on either response (Fig. 3A). This suggests that store release of Ca 2 þ is a key requirement for the EDH response. Similarly PLA 2 inhibition by 3 μM AACOCF 3 also reduced the vasodilatation and Ca 2 þ responses (Fig. 3B), but to a lesser extent. The arachidonic acid released by PLA 2 can interact with a number of enzymes to generate reactive oxygen species (ROS), notably cyclooxygenases, lipoxygenases, xanthine oxidase, and cytochrome P450 isoforms. We used a number of inhibitors to investigate possible sources, namely allopurinol (xanthine oxidase), nordihydroguaiaretic acid (lipoxygenase), and sulfaphenazole (cytochrome P450), while indomethacin was always present to block cyclooxygenase. Of these only sulfaphenazole, a selective cytochrome P450 2C9 inhibitor [21] reduced the responses to carbachol (Fig. 3C). The less specific inhibitors 2-(2-propynyloxy)benzenehexanoic acid (PPOH) and 17-octadecynoc acid (ODYA) were equally effective ( Supplementary Fig. 1). These effects were not due to a reduction in the production of EETs, since neither the EETs antagonist 4, 15-epoxyeicosa-5(Z)-enoic acid (EEZE) nor the soluble epoxide hydrolase (sEH) blocker 12-(3-adamantan-1-yl-ureido)dodecanoic acid (AUDA) had any effect on the carbachol responses ( Supplementary Fig. 2B). It is interesting to note that the reduction in vasodilatation and the Ca 2 þ response with AACOCF 3 and sulfaphenazole was significantly less than with PLC inhibition (AACOCF 3

H 2 O 2 assists release of Ca 2 þ from intracellular stores
Carbachol-induced release of intracellular Ca 2 þ by the endothelium was recorded in arterioles perfused and superfused with Ca 2 þ free PSS containing 0.2 mM EGTA in order to prevent Ca 2 þ influx. Under these conditions carbachol caused a transient increase in EC [Ca 2 þ ] i , which was significantly attenuated by S&C ( Fig. 4A and  B). Further experiments revealed that the increase in EC [Ca 2 þ ] i was similarly reduced by catalase alone, but was unaffected by SOD (Fig. 4B)  confirmed by the inhibitory action of sulfaphenazole (Fig. 4D) on this response.
The results depicted in Figs. 3 and 4 are consistent with the concept that arachidonic acid released by PLA 2 is metabolized by CYP2C9 to produce ROS, which then potentiate carbachol-induced Ca 2 þ release. To further test this idea we added H 2 O 2 (10 mM, a concentration that produced no vasodilatation on its own (see next section and Fig. 6A) to preparations where either PLA 2 or CYP2C9 were blocked with AACOOCF 3 or sulphaphenazole, respectively. In both cases H 2 O 2 fully restored vasodilatation and Ca 2 þ responses (see Fig. 5A & B).

Properties of the response to H 2 O 2 applied exogenously on its own
Application of H 2 O 2 to PE-preconstricted arterioles caused a steeply concentration-dependent vasodilatation, with an apparent threshold at $30 μM and maximum at 100 μM H 2 O 2 (Fig. 6A).
The possibility that the exogenously applied H 2 O 2 resulted in opening the TRPM2 channel was tested by using a blocking antibody that had been shown to block H 2 O 2 -induced Ca 2 þ influx into cultured endothelial cells [22]. It is interesting to note that the EC 50 for activation of this channel by H 2 O 2 ( $ 60 mM), is consistent with the dose-response relationship we found (Fig. 6A). Application of the TRPM2 antibody reduced both the H 2 O 2 -induced rise in EC [Ca 2 þ ] i and vasodilatation, but left these responses to carbachol unaffected (Fig. 6B).
Although both 100 μM H 2 O 2 and carbachol caused roughly equivalent effects on EC [Ca 2 þ ] i and vascular diameter, they apparently did so through different mechanisms. As shown in Figs. 6C and 6D, the vasodilatation elicited by H 2 O 2 was strongly suppressed by the protein kinase G blocker Rp-8-Br-cyclic GMP (100 μM) and the BK Ca channel inhibitor IBTx, (100 nM), and was to a smaller extent also sensitive to the combination of apamin (500 nM) and TRAM-34 (10 mM). Conversely, Rp 8Br-cyclic GMP and IBTx had no effect on the H 2 O 2 -induced rise in EC [Ca 2 þ ] i , which was strongly supressed by the TRPM2 antibody and TRAM-34þapamin. Inhibiting PKG and the BK Ca channel, however, had no effect on the carbachol-mediated responses (Figs. 6E and 6F). TRAM-34 and apamin reduced vasodilatation mediated by H 2 O 2 ( À377 2.8%), to an extent which was which was significantly less than that observed for the response to carbachol ( À 82 72.8%; Po0.001).
In additional experiments we found that CP94, which inhibited carbachol induced vasodilation (see below), had no affect on dilatation to H 2

The nature of the ROS involved in carbachol-mediated effects
Desferrioxamine and the hydroxypyridinones CP94 and CP85 are selective iron chelators, but differ in that CP94 is cell permeable while the other two are not [49]. As shown in Fig. 7A, CP94 (100 μM) significantly reduced the transient carbachol-induced rise in EC [Ca 2 þ ] i in Ca 2 þ free PSS, while neither desferrioxamine nor CP85 (100 μM) had any effect on this response. This supports the view that •OH, rather than H 2 O 2 itself, is the radical responsible for Ca 2 þ release from stores.
In the presence of extracellular Ca 2 þ , all three iron chelators inhibited the carbachol-induced rise in endothelial [Ca 2 þ ] i (Fig. 7B) and reduced the accompanying vasodilation (Fig. 7C). Notably, none of these substances had any effect on the PE induced contraction (not shown).

IL-1β causes ROS-dependent vasodilatation and elevation of EC [Ca 2 þ ] i
We have previously shown that a brief application of a low concentration of interleukin-1β (30 pM for 10 min) results in endothelial ROS generation via NADPH oxidase [23], and that this is prevented by co-application with a low concentration of apocynin (1 mM; [24]). As shown in Fig. 8, a 15 min pre-incubation of the preparation with IL-1β (30 pM), followed by a subsequent brief wash to remove the cytokine, resulted in increased EC [Ca 2 þ ] i and vasodilatation responses that were much reduced by scavenging ROS. Furthermore, this potentiating effect was abolished when apocynin (1 μM) was co-applied and washed off with IL-1β.
NADPH oxidase assembly requires phosphorylation of p47 phox by protein kinase C [25], and when the broad spectrum PKC inhibitor and (C) the cytochrome P450 2C9 inhibitor sulfaphenazole (10 mM) also reduced the [Ca 2 þ ] i and vasodilatation responses. 5 vessels in 5 animals for each data set; paired t-tests *** po 0.001. The reduction with U73122 was greater than with either AACOCF 3 or sulfaphenazole (see text). Ro31-8220 (3 μM) was applied with IL-1β, the potentiation was also abolished. Neither apocynin nor Ro31-8220 affected the response to carbachol in the absence of IL-1β ( Supplementary Fig. 3).

Discussion
The key finding of this study is that whilst the EDHR triggered by carbachol is largely ROS-dependent, it does not involve an effect on VSMCs of H 2 O 2 released by the endothelium. Instead, the results demonstrate that ROS generated by endothelial cytochrome P450 2C9 make a crucial contribution to the EDHR triggered by muscarinic receptor stimulation in rat cremaster arterioles. They do this by acting on the endothelium to potentiate Ca 2 þ release from intracellular stores and enhancing Ca 2 þ entry across the plasma membrane rather than by acting directly as an EDHF. There have been previous indications that this is an important mechanism: thus catalase suppresses cyclopiazonic acid (CPA) induced Ca 2 þ  release in rabbit aortic valve endothelium, as well as EDH-dependent, acetylcholine-induced vasodilatation of rabbit iliac arteries. Furthermore, raising the H 2 O 2 concentration facilitates acetylcholine and CPA induced vasodilatation [19,20]. CPA induced Ca 2 þ release was also shown to be potentiated by the IP 3 R sensitising agent thimerosal, which implies that H 2 O 2 may be exerting a similar sensitising effect, as described for the IP 3 R in human aortic endothelial cells [26]. Based on these and related findings, the late Tudor Griffith and colleagues proposed that H 2 O 2 was acting to promote the EDH response by facilitating IP 3 R-mediated endothelial Ca 2 þ release and thereby raising [Ca 2 þ ] I [19].

Phospholipase C, Phospholipase A 2 and CYP450 2C9
Carbachol activates both M 1 and M 2 receptors, which are linked to phospholipase C and phospholipase A 2 respectively. The present findings indicate that both are required for the full EDHR. We showed that blocking PLC with U73122 inhibits increases in both endothelial [Ca 2 þ ] i and vasodilatation (see Fig. 3A), which is consistent with the finding that many agonists evoke vasodilatation and [Ca 2 þ ] i increases that are dependent on endothelial PLC activation [27]. Our results are consistent with those of Fukao et al. who showed that PLC is crucial for the EDHR in rat mesenteric arteries, implying that IP 3 R-mediated Ca 2 þ release is pivotal for both events [28].
We have also shown that a pathway involving both cPLA 2 and CYP2C9 is the source for ROS in the carbachol-induced EDHR in these arterioles. Our results (Fig. 3B), demonstrate that the cPLA 2 inhibitor AACOCF 3 suppressed the carbachol-induced EDHR. A role for cPLA 2 in EDH has previously been demonstrated in some [29][30][31] but not all arteries [32]. Although this could reflect a role for EETs generated as CYP450 metabolises arachidonic acid produced by cPLA 2 , it may also depend upon ROS produced the same pathway. This possibility was investigated in porcine coronary arteries, but it was found that this did not apply to the EDHR triggered by bradykinin [5]. In contrast, increases in endothelial permeability evoked by bradykinin in cerebral venular capillaries were found to depend on ROS generated when arachidonic acid produced by cPLA 2 is metabolised by cyclooxygenase and lipoxygenase [33].
CYP450 has been thought to be involved in the EDH response by virtue of its production of EETs, which were initially proposed to act via opening BK Ca channels in vascular smooth muscle cells e.g. [34,35]. More recently however, EETs generated by CYP450 2C9 were shown to act within endothelial cells by increasing [Ca 2 þ ] i by recruiting TRP channels to the plasmalemma [21,36].
On the other hand, CYP450 enzymes are also a source of ROS [37], and furthermore, basal ROS production by endothelial cells is dependent on the CYP450 2C9 isoform, as shown by using antisense oligonucleotides [35] and sulfaphenazole [37], both of which suppress EDH. Nevertheless, the possibility that CYP450 2C9 might promote EDH via ROS generation [38], as indicated by the results depicted in Fig. 3C, seems not to have been previously considered. Sulfaphenazole strongly suppressed carbachol-induced Ca 2 þ release, as well as the rise in EC [Ca 2 þ ] i and the associated vasodilatation in normal Ca 2 þ solution, and neither the EETs antagonist EEZE nor AUDA had any effect on the carbachol responses (Supplementary Fig. 2B). Moreover, although in accordance with previous reports e.g. [39] we observed that BK Ca channel blockade with IBTx inhibited the vasodilatation caused by application of exogenous 11, 12 EET, IBTx had no effect on the carbachol-induced vasodilatation ( Supplementary Fig. 2A).

Exogenously applied H 2 O 2
Whereas the experiments with anti-oxidants and blockers of endothelial ROS-producing pathways indicate that H 2 O 2 acts by potentiating the carbachol-induced EC [Ca 2 þ ] increase, they do not exclude an additional direct effect of H 2 O 2 on arteriolar vascular smooth muscle during EDH. We therefore carried out a systematic comparison of the properties of the vasodilator responses to H 2 O 2 and carbachol-induced EDH. The dose response relationship between H 2 O 2 and vasodilatation (Fig. 6A) is consistent with most others that do not show significant vasodilatation below 50 mM e.g. [6,11]. The vasodilatation caused by H 2 O 2 was strongly attenuated by Rp-8Br-cGMP and IBTx (see Fig. 6C). Neither of these blockers had any effect on the H 2 O 2 induced rise in endothelial [Ca 2 þ ] i (see Fig. 6D), consistent with the concept that H 2 O 2 evokes vasodilatation principally via vascular smooth muscle cells by causing direct oxidation-induced stimulation of G kinase [18], this leading to cGMP-mediated opening of BK Ca channels [40]. In contrast, responses to carbachol were insensitive to these agents, implying that any release of H 2 O 2 on its own by the endothelium was too small to activate this pathway. It appears, however, that the rise in endothelial [Ca 2 þ ] i triggered by 100 μM H 2 O 2 also makes a significant contribution to the vasodilatation, since the combination of TRAM-34 and apamin reduced vasodilation to H 2 O 2 . Moreover, a TRPM2 antibody previously shown to block H 2 O 2 -induced rises in [Ca 2 þ ] i associated with activation of these channels in cultured EC greatly suppressed the H 2 O 2 -induced rise in EC [Ca 2 þ ] i and reduced the accompanying vasodilatation [22]. These results suggest that the endothelium-dependent component of the vasodilatation to 100 μM H 2 O 2 in these arterioles was largely caused by the opening of EC TRPM2 channels, leading to a rise in EC [Ca 2 þ ] i and a consequent EDHR. This would be expected to open endothelial K Ca channels, accounting for the small inhibition of the H 2 O 2 induced rise in endothelial [Ca 2 þ ] i caused by TRAM-34 þapamin (Figs. 6D, 6F). The observation that the TRPM2 antibody had no effect on either the rise in endothelial [Ca 2 þ ] i or the vasodilatation induced by carbachol (Fig. 5B)  It has been argued that a relatively high exogenously applied H 2 O 2 concentration does not accurately reflect the true intracellular H 2 O 2 due to the effects of peroxidases and a hypothesized barrier to H 2 O 2 entry across the plasmalemma [41]. On the other hand, H 2 O 2 has been shown to have a high permeability coefficient (1.6 Â 10 À 3 cm s À 1 ) across the EC membrane [42]. Furthermore, we found that the application of just 1 mM H 2 O 2 resulted in a significant increase in carbachol-induced Ca 2 þ release from intracellular stores (see Fig. 4C). Notably, studies have shown that H 2 O 2 at concentrations r100 μM does not itself cause EC Ca 2 þ release in the absence of a co-stimulus. Hu et al. [26] measured the content of the Ca 2 þ stores in permeabilized cultured human aortic endothelial cells (HAEC) using Mag-indo, and found that this was not affected by 100 μM H 2 O 2 , although H 2 O 2 at concentrations above 3 μM sensitized these cells to IP 3 -induced Ca 2 þ release and 10 μM H 2 O 2 promoted histamine-mediated Ca 2 þ release in intact HAEC [43]. Similarly it was found that 100 µM H 2 O 2 had no effect on EC Ca 2 þ store content, measured using Mag fluo-4, in intact rabbit aortic valve endothelium [19].
Similarly, the effects of inhibiting either PLA 2 with AACOCF 3 , or CYP450 2C9 with sulfaphenazole were completely reversed by adding H 2 O 2 (10 μM), a concentration which on its own had no effect (see Figs. 5A and 5B). Thus, it is clear that the suppression of the carbachol-induced EDHR by sulfaphenazole is best explained by a decrease in endothelial H 2 O 2 rather than the inhibition of EET release it causes. This effect would explain the finding that the EDH-dependent vasodilatation to acetylcholine in guinea-pig coronary arteries was antagonised by CYP450 blockers but not IBTx [39], as well as the observation that sulfaphenazole-induced attenuation of the EDHR triggered by acetylcholine and bradykinin in the human forearm microcirculation was prevented by co-administration of the anti-oxidant ascorbate [44].
These findings indicate that ROS produced when CYP 450 2C9 metabolises arachidonic acid generated by cPLA 2 are responsible for an increase in Ca 2 þ release by the IP 3 R. There is evidence [45] that H 2 O 2 is able to sensitize the IP 3 R to various stimuli in cultured endothelial cells. This view is supported to a certain extent by our finding that inhibiting PLC produced a significantly greater reduction in the response to carbachol than inhibiting elements of the ROS generating pathway (see Fig. 3).

Reactive oxygen species mediated signalling
It is apparent that H 2 O 2 generated from the interaction between arachidonic acid and CYP 450 2C9 is important for potentiating the release of Ca 2 þ from stores, since catalase alone, rather that superoxide dismutase reduces effects of carbachol.
The question of whether the Fenton reaction is required for H 2 O 2 to have its effects was addressed by using Fe 2 þ chelators. The involvement of hydroxyl radical in endothelium-dependent vasodilation was first proposed as a consequence of finding that desferrioxamine (DFO) inhibited bradykinin-induced vasodilatation of mouse cerebral arterioles [46]. Few subsequent studies have addressed this issue, however, and the limited amount of information available on the effects of iron chelation on the EDH as such indicates that hydroxyl radical is not involved [10,30]. Nonetheless, in light of reports that DFO attenuated arachidonateinduced vasodilatation and permeability increases in rat cerebral arterioles [17,47], and of our observation that blocking cPLA 2 suppressed the responses to carbachol in cremaster arterioles, we examined whether these were also affected by DFO and the iron chelating 3-hydroxypyridin-4-ones CP94 and CP85. These have little direct radical scavenging ability [48], moreover a variety of 3-hydroxypyridin-4-ones of varying membrane permeability have been synthesised which allowed us to compare the effects of high (CP94) and low (CP85) cell-permeability analogues [49].
If hydroxyl radical rather than H 2 O 2 is sensitising the IP 3 R, it would be predicted that CP94 and DFO but not CP85 should inhibit carbachol-induced Ca 2 þ release, since hydroxyl radical, which immediately reacts with other molecules in its vicinity as it is formed, is unlikely to cross the cell membrane [50]. This is what we observed. Although the enormous reactivity of hydroxyl radical would appear to prevent it from mediating signal transduction, as this would require spatially discrete and reversible interactions, there are mechanisms by which this could occur. For example, hydroxyl anion is largely scavenged by ascorbate to yield dehydroascorbate (see [50]), which has been proposed to be transported into the lumen of the ER where it acts to promote the oxidation of protein thiols [51]. This might be of particular relevance to Ca 2 þ release. Notably, it was shown that treatment of aortic EC with H 2 O 2 leads to glutathionylation of the IP 3 R1 and an increase in its sensitivity to [Ca 2 þ ] [45], and it was proposed that this might be due to its interaction with ERp44, an ER-resident protein which inhibits the IP 3 R1 by binding to it in a redox-sensitive manner, such that oxidation of cysteine residues on the  intraluminal LV3 domain of the IP 3 R1 prevents binding and therefore enhances channel opening [52]. We speculate that such a mechanism could provide a feasible explanation for how the non-localised production of hydroxyl anion within the cytoplasm might result in a localised and reversible effect on signal transduction, in this case an increase in Ca 2 þ release.
Although CP94 profoundly inhibited carbachol-induced endothelial Ca 2 þ release, the cell impermeant CP85 and DFO had no effect (Fig. 6A). In the presence of extracellular Ca 2 þ , however, all the chelators reduced the response, which indicates that the hydroxyl radical formed outside the cell membrane is involved in the opening of the K Ca channels (Figs. 6B and 6C), which is evidence for H 2 O 2 crossing the plasmalemma. There is evidence that several ROS-sensitive Ca 2 þ -permeable channels, including TRPC3, TRPC4 and TRPM2, can contribute to Ca 2 þ influx in EC during the response to muscarinic agonists [27,53]. Of these, to our knowledge a role for hydroxyl radical as a channel activator has only been reported for TRPM2, although this was shown to involve an action on intracellular cysteine residues [54].

IL-1β and EDH
The ability of IL-1β to promote EDH has not, to our knowledge, been reported previously, although it has been reported that this cytokine caused dilatation of new-born pig pial arterioles via a prostanoid-and cyclic nucleotide-dependent-mechanism [55]. The effects of IL-1β were prevented when it was co-applied with the NADPH-oxidase assembly inhibitor apocynin and the PKC blocker Ro31-8220, consistent with previous evidence that Il-1β activates increases superoxide production and activates NADPH oxidase in EC [23,56,57]. The effect of IL-1β on the carbachol-induced EDHR in cremaster arterioles demonstrated several parallels with its effect on the bradykinin-induced permeability increase of rat pial venules described by Woodfin et al. [23] including a similar rapid induction and a sensitivity to ROS scavenging, apocynin and PKC blockade, and we speculate that activation of NADPH oxidase may constitute a widespread mechanism by which IL-1β can acutely regulate EC [Ca 2 þ ] i in the vascular endothelium, and may be one of the factors that are involved in septic shock.

Conclusions
We speculate that arteries/arterioles may fall into three categories with regard to the involvement of H 2 O 2 in EDH. In one group, as evidenced by the lack of effect of catalase, H 2 O 2 is probably not involved in this response [4], although such a role has not been exhaustively examined for every stimulus which triggers EDH in these preparations. In another group, which notably includes coronary arteries, H 2 O 2 itself seems to act as an important EDHF [6,7,10,11]. The reports of Griffith and colleagues [19,20] pointed to the existence of a third group of arteries in which H 2 O 2 facilitates EDH, and focussed attention on IP 3 Rmediated Ca 2 þ release in EC as its likely target. Our results now provide a more detailed picture of how H 2 O 2 acts to promote EDH in the type of arteriole that falls into this category. We find that very small changes in the EC [H 2 O 2 ] around the basal level can strongly modulate the EDHR via an effect on Ca 2 þ release. The data demonstrate that EDH triggered by muscarinic receptor activation is greatly facilitated by ROS which are produced when arachidonic acid synthesised by cPLA 2 is metabolised by CYP450 2C9. A short pre-treatment with IL-1β is able to further enhance carbachol-induced EDH. An increase in EC Ca 2 þ release, which would be expected to potentiate SOCE, and thereby also facilitate EC Ca 2 þ influx, is implicated in these responses. The effects of CP94 and CP85 indicate that hydroxyl radical plays an important role in this facilitation of Ca 2 þ release, and also demonstrate that ROS also promote EC Ca 2 þ influx during EDH.
A limitation of our results is that we could not record changes in [ROS] in the endothelium of these arterioles, possibly because these were very small. We were therefore unable to determine whether muscarinic receptor activation was increasing EC [ROS], or whether basal ROS levels were sufficient to facilitate Ca 2 þ release. PLA 2 has, however, been shown to be activated by increases in EC [Ca 2 þ ] i , e.g. [58] so that it is plausible that PLA 2 -dependent ROS production would occur upon the application of carbachol.

Author contributions
James Chidgey: Conception and design of experiments; collection, analysis and interpretation of data.
Paul Fraser: Conception and design of experiments; analysis and interpretation of data, critical revision of the article.
Philip Aaronson: Conception and design of experiments; analysis and interpretation of data: drafting and revising the article.