The endothelium controls vascular tone not only by releasing NO and prostacyclin, but also by other pathways causing hyperpolarization of the underlying smooth muscle cells. This characteristic was at the origin of the term ‘endothelium-derived hyperpolarizing factor’ (EDHF). However, this acronym includes different mechanisms. Arachidonic acid metabolites derived from the cyclo-oxygenases, lipoxygenases and cytochrome P450 pathways, H2O2, CO, H2S and various peptides can be released by endothelial cells. These factors activate different families of K+ channels and hyperpolarization of the vascular smooth muscle cells contribute to the mechanisms leading to their relaxation. Additionally, another pathway associated with the hyperpolarization of both endothelial and vascular smooth muscle cells contributes also to endothelium-dependent relaxations (EDHF-mediated responses). These responses involve an increase in the intracellular Ca2+ concentration of the endothelial cells, followed by the opening of SKCa and IKCa channels (small and intermediate conductance Ca2+-activated K+ channels respectively). These channels have a distinct subcellular distribution: SKCa are widely distributed over the plasma membrane, whereas IKCa are preferentially expressed in the endothelial projections toward the smooth muscle cells. Following SKCa activation, smooth muscle hyperpolarization is preferentially evoked by electrical coupling through myoendothelial gap junctions, whereas, following IKCa activation, K+ efflux can activate smooth muscle Kir2.1 and/or Na+/K+-ATPase. EDHF-mediated responses are altered by aging and various pathologies. Therapeutic interventions can restore these responses, suggesting that the improvement in the EDHF pathway contributes to their beneficial effect. A better characterization of EDHF-mediated responses should allow the determination of whether or not new drugable targets can be identified for the treatment of cardiovascular diseases.
ECs (endothelial cells) synthesize and release factors that regulate angiogenesis, inflammatory responses and haemostasis, as well as vascular tone and permeability. The endothelium maintains the balance between inhibition and promotion of the proliferation and migration of SMCs (smooth muscle cells), between prevention and stimulation of the adhesion and aggregation of platelets and between thrombogenesis and fibrinolysis, as well as between vasodilation and vasoconstriction. Upsetting this tightly regulated balance leads to endothelial dysfunction [1,2] (Figure 1).
Endothelium-dependent relaxations and contractions
Endothelium-dependent relaxations/vasodilations in response to neurohumoral mediators and physical forces, such as the shear stress exerted by the flowing blood, are generally attributed to the release of NO and/or PGI2 (prostacyclin) [3–6]. However, in numerous blood vessels from different species, including humans, these responses cannot be totally explained by the release of these two mediators. The relaxations observed in the presence of COX (cyclo-oxygenase) and NOS (NO synthase) inhibitors are often associated with hyperpolarization of VSMCs (vascular SMCs) and were first attributed to a non-characterized endothelial factor termed EDHF (endothelium-derived hyperpolarizing factor). The acronym ‘EDHF’ turned out to be confusing because it implies that a single diffusible substance mediates this type of endothelium-dependent relaxation. In fact NO itself, but also numerous identified putative endothelium-derived factors including, CO (carbon monoxide), H2S (hydrogen sulfide), ROS (reactive oxygen species), peptides and arachidonic acid metabolites derived from the COX, LOX (lipoxygenase) and cytochrome P450 mono-oxygenase pathways, can hyperpolarize the underlying SMCs .
Hyperpolarization decreases Ca2+ influx, either by reducing the open probability of CaV (voltage-dependent Ca2+) channels or the CaV channel-dependent activation of the sarcoplasmic reticulum, which is a powerful means to produce the relaxation of VSMCs [7–9].
Another pathway, which does not involve the synthesis and the release of a factor, is associated with the hyperpolarization of both ECs and VSMCs (EDHF-mediated responses) and also contributes to endothelium-dependent relaxations. These responses involve an increase in the intracellular Ca2+ concentration of ECs, followed by the opening of SKCa and IKCa channels [small and intermediate conductance KCa (Ca2+-activated K+) channels respectively], and the subsequent hyperpolarization of these cells. Then, the endothelium-dependent hyperpolarization of the underlying SMCs can be evoked by direct electrical coupling through myoendothelial junctions and/or the accumulation of K+ ions in the intercellular space between the two cell types . The present review will briefly summarize the hyperpolarizing effects of the various endothelium-derived factors and will focus on this latter mechanism, i.e. EDHF-mediated responses.
ARACHIDONIC ACID METABOLITES
In ECs, PGI2 is the principal COX-derived metabolite of arachidonic acid. When activating its preferential receptor, the IP receptor (PGI2 receptor), PGI2 is a potent antithrombotic and antiplatelet agent, and is generally a vasodilator substance . The deletion of PGI2 synthase generates hypertensive mice with arterial sclerosis, whereas, in response to stress or injury, IP-receptor-knockout animals have enhanced platelet activation, thrombosis, intimal hyperplasia, atherosclerosis and restenosis, and are prone to ischaemia/reperfusion injury .
The vascular relaxation to PGI2, or its synthetic analogues, is often associated with the concomitant hyperpolarization of the SMCs, which, depending on the blood vessels and the species, can involve the opening of various populations of K+ channels [11,12]. Therefore, in numerous vascular beds, PGI2 can act as an endothelium-derived hyperpolarizing substance . Since COX inhibitors abolish the basal and stimulated generation of PGI2, and potent and specific antagonist of the IP receptor block its vasodilator responses [14,15], the contribution of PGI2 in endothelium-dependent responses can reasonably be assessed. This prostaglandin plays a role in flow-mediated vasodilation [16,17]; however, its contribution to acute endothelium-dependent relaxations in response to neurohumoral mediators is often minimal  or can only be observed when the other endothelial pathways have been inhibited [18,19]. The contribution of PGI2 to endothelium-dependent responses is increased in eNOS (endothelial NOS)-knockout mice [20,21]. Similarly, in humans with cardiovascular diseases, COX-2-derived prostaglandins can play a compensatory role in the decreased NO bioavailability [22,23], possibly explaining some of the detrimental cardiovascular effects associated with COX-2 inhibitors .
However, in aging and in the course of some cardiovascular diseases, PGI2, along with other prostaglandins, can also act as an endothelium-derived contracting factor by activating smooth muscle TP receptors (thromboxane/endoperoxide receptors) and, thus, contribute to endothelial dysfunction [14,15,25] (Figure 1).
Cytochrome P450 mono-oxygenases
EETs (epoxyeicosatrienoic acids), derived from the endothelial cytochrome P450 2C or 2J epoxygenases, are generally, but not necessarily, vasodilator agents [26,27], whereas 20-HETE (20-hydroxyeicosatetraenoic acid; a metabolite of the cytochrome P450 4A and 4F family preferentially located in VSMCs) is a potent endogenous vasoconstrictor of renal, cerebral, coronary, mesenteric and skeletal muscle arteries  (Figure 1).
EETs play an important role in endothelium-dependent relaxations either as diffusible factors or as essential endothelial intracellular second messenger(s) . When released, they can act as an autocrine agent eliciting endothelium- and NO-dependent relaxations  or, more generally, diffuse toward the underlying VSMCs and produce their relaxation [31,32]. In the latter case, 11,12-EET and 14,15-EET, the two predominant diffusible isoforms, activate BKCa channels (large conductance KCa channels or KCa1.1) of VSMCs [33–36]. Experiments performed both In vitro and in vivo on human blood vessels suggest a contribution of EETs to endothelium-dependent relaxations/vasodilations in coronary and mammary arteries, as well as in peripheral muscular and subcutaneous arterioles [37–40].
Cytochrome P450 metabolites also play an important role in the regulation of the kidney circulation, and contribute to the long-term regulation of blood pressure and Na+ homoeostasis. EETs regulate not only regional blood flow, but activate KATP channels in cardiomyocytes, limit platelet aggregation, exert anti-inflammatory actions, and improve insulin sensitivity and lipid metabolism. Thus they can protect the kidney vasculature from injury during renal and cardiovascular diseases, exert cardioprotection and prevent/delay metabolic syndrome and atherogenesis [41–43].
EETs are rapidly metabolized by soluble epoxide hydrolase to form the generally less active DHETs (dihydroxyeicosatrienoic acids). In mice, disruption of the soluble epoxide hydrolase gene produces no or minor decreases in basal arterial blood pressure , but protects against myocardial ischaemia/reperfusion, heart failure and ischaemic stroke [45–47]. In rats, soluble epoxide hydrolase plays an essential role in AngII (angiotensin II)-induced cardiac hypertrophy , which, in patients with cardiovascular disease, is the most common cause of heart failure. Polymorphisms of EPHX2 (the gene encoding soluble epoxide hydrolase) have been associated with coronary artery diseases, ischaemic stroke and insulin resistance [43,49,50]. Potent and selective inhibitors of this enzyme have been designed and are currently undergoing clinical trials for the treatment of hypertension. The additional anti-inflammatory properties of soluble epoxide hydrolase inhibitors also make them attractive for the treatment of chronic kidney disease in patients with cardiometabolic syndrome .
In rat and porcine coronary arteries, the 12-LOX metabolite 12-S-HETE [12-(S)-hydroxyeicosatetraenoic acid] can be released by ECs and evoke relaxation of the vascular smooth muscle by activating BKCa channels .
In rabbit arteries, the generation of 11,12,15-THETA (11,12,15-trihydroxyeicosatrienoic acid) by reticulocyte 15-LOX-I contributes to acetylcholine-induced endothelium-dependent relaxations . In this species, the age-related decrease in hypotension and in endothelium-dependent relaxations induced by acetylcholine are mediated by a decreased synthesis of 11,12,15-THETA, associated with a down-regulation of 15-LOX expression and a reduced activity of the enzyme [54,55]. By contrast, short-term hypercholesterolaemia, in the absence of atherosclerotic lesions, and chronic hypoxia increase endothelial 15-LOX expression and 11,12,15-THETA production, as well as acetylcholine-induced endothelium-dependent relaxations and hypotension [56,57] (Figure 1).
However, although a strong case for LOX derivatives acting as endothelium-derived hyperpolarizing substances can be built in specific arteries and especially in those of the rabbit, most of the EDHF-mediated responses, including that in human arteries, do not appear to involve metabolites of arachidonic acid produced by this pathway [1,58].
In isolated blood vessels or in vivo in anaesthetized animals, endogenous and exogenous cannabinoids usually have vasodilator properties and are likely to play a role in cardiovascular homoeostasis . However, the suggestion that anandamide could be an endothelium-derived hyperpolarizing substance has not been substantiated [1,60].
Three different isoforms of an enzyme are able to synthesize NO from L-arginine and molecular oxygen: NOS-I [nNOS (neuronal NOS) or NOS-1], NOS-II [iNOS (inducible NOS) or NOS-2] and NOS-III (eNOS or NOS-3). Under various circumstances, such as the presence of NOS inhibitors, low levels of L-arginine or oxidized BH4 (tetrahydrobiopterin) but also under physiological conditions, NOS can generate O2− (superoxide anion), which is reduced to H2O2 either spontaneously or enzymatically by SOD (superoxide dismutase). H2O2 can act locally close to its site of production or, as it is an uncharged molecule, diffuse through the cell membrane and act on neighbouring cells [5,61–63] (Figure 1).
NO is a potent vasodilator and a powerful inhibitor of platelet adhesion and aggregation. The relaxations elicited by NO are generally associated with the stimulation of the cytosolic soluble guanylate cyclase and the subsequent activation of PKG (cGMP-dependent protein kinase) . NO regulates the activity of various K+ channels and, depending on the vascular beds, the hyperpolarizing effects of NO on VSMCs can substantially contribute to their relaxation . The activation of BKCa channels by NO contributes to the beneficial effects of currently prescribed drugs associated with the L-arginine/NOS/cGMP pathway, for instance NO donors and phosphodiesterase inhibitors such as sildenafil .
In contrast with COX inhibitors, inhibitors of NOS do not necessarily fully inhibit the production of NO. Thus, in their presence, residual NO can still be produced by ECs and contributes to the relaxation and/or hyperpolarization of the underlying vascular smooth muscle . Furthermore, NO can also be stored and released independently of the activation of NOS [66–68]. Therefore it appears likely that the role of NO as an endothelium-derived hyperpolarizing substance may be underestimated by assuming that the presence of an inhibitor of NOS rules out its contribution. A non-NO/non-PGI2-mediated response should be reported not only when a relaxation and/or hyperpolarization is recorded in the combined presence of COX and NOS inhibitors, but when this response is still observed with the additional presence of an NO scavenger . In in vivo studies in humans, the complete blockade of NOS is difficult to achieve (or to demonstrate) and, for obvious ethical reasons, many of the pharmacological tools used to study EDHF-mediated responses cannot be administrated. The limitation of these studies should always be kept in mind when interpreting these findings.
In NOS-3-knockout mice, EDHF-mediated responses play a compensatory role for the absence of endothelial NO  and this adaptation to NOS deletion is gender-specific . Similarly, in resistance arteries from female double knockout mice for NOS-3 and COX-1, endothelium-dependent relaxations are preserved by an EDHF-mediated mechanism, whereas, in arteries from the double knockout males, the endothelium-dependent relaxations are impaired severely. In genetically modified female mice, the double deletion of NOS-3 and COX-1 does not affect mean arterial blood pressure, whereas the corresponding males are hypertensive .
In the mesenteric arteries of mice with disruption of the various NOS isoform genes (single NOS-3-knockout, double NOS-1/NOS-3-knockout and triple NOS-1/NOS-2/NOS-3-knockout), the endothelium-dependent relaxations and hyperpolarizations resistant to NOS and COX inhibitors are progressively reduced as the number of the disrupted NOS genes is increased . The dependency of the responses resistant to NOS and COX inhibitors towards the NOS systems could not be explained by residual NO release, but has been attributed to the production of H2O2  (Figure 1).
Depending on the tissue, the experimental conditions or the concentrations studied, H2O2 possesses dilator or constrictor properties, and can hyperpolarize or depolarize VSMCs . For instance, in the isolated murine mesenteric artery, H2O2, at concentrations lower than 50 μmol/l, produces an endothelium-independent relaxation providing that KCa channels are operational, but, at the same concentrations, elicits a potent contractile response if the activity of these channels is compromised .
In NOS-disrupted murine arteries, relaxations resistant to NOS and COX inhibitors are sensitive to catalase, the enzyme that dismutates H2O2 into water and oxygen, and have been attributed to the NOS-3/Cu,Zn-SOD-dependent formation of H2O2 [76,77] (Figure 1). The maintenance of these endothelium-dependent relaxations/hyperpolarizations in NOS-3-knockout mice was explained by the compensatory endothelial expression of other NOS genes, the production of H2O2 being preserved up to the total disruption of the three NOS genes . Nevertheless, whether or not the decrease in the EDHF-mediated responses is directly associated with the disruption of the NOS genes or is independently associated with the severe phenotype of these mice, especially with the one observed in the triple knockout mice (hypertension, dyslipidaemia, myocardial infarction and nephrogenic diabetes insipidus), remains to be fully assessed [also see the phenotype of mice with a CSE (cystathionine γ-lyase) deletion described below].
The involvement of H2O2 in agonist- and flow-induced endothelium-dependent relaxations/vasodilations has been suggested in other vascular beds , including human mesenteric  and coronary  arteries. In addition, without actually being released by ECs, H2O2 can enhance EDHF-mediated responses by potentiating Ca2+ release from endothelial stores . In various experimental models or in arteries from patients with cardiovascular diseases, H2O2 generation can partially compensate for the decreased NO production, at least in terms of endothelium-dependent relaxations, but in the long term this production of ROS may contribute to vascular oxidative injury [82–84].
Depending on the blood vessels, besides NOSs, several other endothelial enzymes, such as COXs, LOXs, xanthine oxidases, cytochrome P450 epoxygenases, NADPH oxidases and mitochondrial respiratory enzymes, can generate O2− and be at the origin of H2O2 production . In murine arteries, endothelial oxidases other than NOS do not appear to be involved in H2O2-mediated responses ; however, in human coronary arterioles, flow-induced endothelium-dependent dilation is associated with H2O2 generated by the mitochondrial respiratory chain, whereas bradykinin-induced endothelium-dependent relaxation requires NADPH-oxidase-derived H2O2 [86,87].
Therefore, in the presence of NOS and COX inhibitors, residual NO and/or H2O2, both potentially derived from various endothelial NOS isoforms, can act as endothelium-derived hyperpolarizing substances. However, in many arteries, EDHF-mediated responses cannot be attributed to the generation of residual NO or to that of H2O2 [1,74,88].
OTHER GASEOUS MEDIATORS
Besides NO, CO and H2S are also water-soluble low-molecular-mass gases that readily cross lipid membranes and, therefore, diffuse homogeneously and in a non-polarized manner from their production site acting as autocrine and paracrine substances .
The predominant biological source of CO is from haem degradation by HO (haem oxygenase), either from the constitutive (HO-2) or the inducible (HO-1) isoform, both being expressed in VSMCs and ECs [90,91]. In many pathophysiological situations, the HO/CO pathway compensates for the decrease in NO bioavailability . CO is a potent vasodilator in most, but not all, vascular beds. The mechanisms of CO-induced vasodilation involve the stimulation of soluble guanylate cyclase, the inhibition of cytochrome P450-dependent production of 20-HETE and/or the activation of various populations of K+ channels . However, CO is also a tonic inhibitor of NOS, by binding to its prosthetic haem, and can contribute to endothelial dysfunction .
HO-1-knockout mice, although normotensive, when subjected to stress or injury have exacerbated responses. In contrast, overexpression of HO-1 plays a protective role in hypoperfusion and ischaemia/reperfusion injury, and the induction of HO-1 expression by transient haemin administration produces a long-lasting normalization of arterial blood pressure in SHRs (spontaneously hypertensive rats) [93,94]. HO-2-knockout mice are also normotensive, but stroke damage in response to injuries is accentuated in these animals, indicating that HO-2 plays an endogenous neuroprotective role in the brain . CO-releasing molecules have vasodilator, anti-ischaemic and anti-inflammatory effects, and may present some therapeutic interest in cardiovascular diseases .
Two main enzymes are responsible for the production of H2S, CBS (cystathionine β-synthase) and CSE, and both use L-cysteine as a substrate. The physiological cardiovascular effects of H2S, which are generally linked to the activation of the latter enzyme, involve anti-inflammatory and antioxidant properties, vasodilation and a decrease in arterial blood pressure [89,99] (Figure 1).
Mice with the deletion of CSE, an enzyme expressed in multiple tissues and recently identified in ECs, are hypertensive and the endothelium-dependent relaxations of their mesenteric artery in response to methacholine is virtually abolished . H2S is produced and released by ECs, in a Ca2+-dependent manner, following neurohumoral stimulation and evokes relaxation and hyperpolarization of VSMCs by activating KATP (ATP-sensitive K+) channels [100,101]. Thus these results suggest that H2S is an endothelium-derived relaxing and hyperpolarizing factor. However, the precise role of this mediator needs to be substantiated further. In most studies, the endothelium-dependent relaxations of the murine mesenteric artery involve NO release and EDHF-mediated responses, which are not necessarily associated with the activation of KATP channels . The disappearance of both the NO- and EDHF-mediated component of the endothelium-dependent relaxation in CSE-knockout mice is unexplained at present, but could be attributed to the increase in homocysteine levels [102,103]. Again, whether or not the decrease in endothelium-dependent responses (NO-mediated and EDHF-mediated responses) is directly associated with the disruption of the CSE gene or is independently associated with the phenotype of these mice remains to be determined (also see multiple NOS-knockout mice described above).
CNP (C-TYPE NATRIURETIC PEPTIDE)
ECs can theoretically synthesize numerous vasoactive peptides. Among them CNP, a member of the natriuretic peptide family, evokes relaxations and hyperpolarizations of VSMCs, including those of human forearm resistance vessels. The vasodilator effects of CNP are generally attributed to the activation of the NPR-B (natriuretic peptide receptor-B) subtype on the smooth muscle, followed by the stimulation of particulate guanylate cyclase, leading to accumulation of cGMP and the subsequent opening of BKCa and KATP channels [106–108].
In addition, it has been suggested that CNP could contribute to EDHF-mediated responses. CNP would activate the NPR-C receptor subtype and evoke hyperpolarization of SMCs via the cGMP-independent activation of GIRK channels [G-protein-regulated KIR (inward-rectifier K+) channels] [109,110]. However, this hypothesis had not been confirmed [111–113], and in mice deficient for the NPR-C gene EDHF-mediated responses are not altered .
Therefore CNP is unlikely to act as an endothelium-derived relaxing/hyperpolarizing substance and contribute to moment-to-moment endothelium-dependent regulation of vascular tone. Nevertheless, this peptide plays a key role in preventing smooth muscle proliferation, leucocyte recruitment and platelet reactivity. As such, CNP is likely to exert an anti-atherogenic influence on the blood vessel walls [115,116].
EDHF-mediated responses are endothelium-dependent relaxations resistant to NOS and COX inhibitors, which do not involve one of the identified mediators mentioned above (arachidonic acid metabolites, and residual or stored NO, H2O2, CO, H2S or CNP) and which require the activation of endothelial KCa channels.
The three subtypes of KCa channels of large (BKCa), intermediate (IKCa or KCa3.1) and small conductance (SKCa or KCa2.3 isoform) are present in the vascular wall, but with very specific cellular and subcellular localization (Figure 2).
KCa channels in the vascular wall and EDHF-mediated responses
BKCa channels are expressed in virtually all VSMCs , whereas, in most ECs, when freshly isolated, BKCa channels are at best poorly expressed and iberiotoxin-sensitive currents are observed only at very positive potentials [117–119]. In SMCs, BKCa channels, often clustered in groups of 20–100 units, are activated by a general increase in intracellular Ca2+ or by Ca2+ sparks, localized elemental Ca2+ release events from internal Ca2+ stores, which then generate STOCs (spontaneous transient outward currents) [120,121]. BKCa channels are often co-localized in discrete smooth muscle areas with endoplasmic reticulum and form a signal complex physically associated with cationic channels, such TRPC1 (transient receptor potential canonical channel 1)  or indirectly associated with TRPV4 (transient receptor potential vanilloid channel 4). The Ca2+ influx through TRPV4 preferentially stimulates ryanodine receptors located on the endoplasmic reticulum increasing the frequency of Ca2+ sparks. In some arteries, EET-induced hyperpolarization involves the latter Ca2+ signalling complex  (Figure 2). Conversely, there is little evidence for a functional role of SKCa channels in VSMCs, although a non-identified apamin-sensitive conductance has been reported in some arteries [123–125]. Similarly, in healthy and freshly isolated VSMCs, IKCa channels are not or very poorly expressed. However, in proliferating cells, as observed in culture or after vascular injury, the expression of this channel increases dramatically [126,127].
By contrast, the IKCa and SKCa channels (especially the SK3 α subunit) are constitutively expressed in ECs [117,118,128,129], but have a very different spatial distribution. SKCa are diffusely distributed over the plasma membrane with preferential locations at sites of homocellular endothelial gap junctions and caveolin-rich domains, and are associated with various Cxs (connexins). IKCa are localized preferentially within endothelial projections through the internal elastic lamina at the sites of myoendothelial gap junctions [113,130–133] (Figure 2).
Agonists that stimulate GPCRs (G-protein-coupled receptors) and compounds such as the Ca2+ ionophore A23187, thapsigargin and cyclopiazonic acid evoke EDHF-mediated responses. These substances share the property of increasing the endothelial intracellular Ca2+ concentration and activating endothelial SKCa channels (blocked by apamin, scyllatoxin or UCL 1684) and/or IKCa channels (blocked by charybdotoxin or TRAM-34) . 1-EBIO (1-ethyl-2-benzimidazolinone), a non-specific activator of KCa channels , which activates endothelial IKCa and SKCa channels , but not BKCa channels of vascular smooth muscle, hyperpolarizes ECs and produces endothelium-dependent hyperpolarization [135,136]. Similar results were obtained with more potent analogues of 1-EBIO, such as DC-EBIO (5,6-dichloro-EBIO) or NS-309 [137,138], and with derivatives of the neuroprotective agent riluzole such as SK-20 or SK-31 , indicating that activation of endothelial KCa channels and/or EC hyperpolarization elicit(s) EDHF-mediated responses.
The hyperpolarization of the ECs in turn favours the entry of Ca2+ by increasing the driving force for this ion [140,141]. Therefore endothelial KCa channels are not only key players in EDHF-mediated responses, but also contribute to the activation of Ca2+-sensitive enzymes, such as eNOS, and thus to the generation of NO [142,143] (Figure 2).
Hyperpolarization of vascular smooth muscle
The involvement of two populations of endothelial KCa channels in EDHF-mediated responses has been puzzling for a long time, but the discrete role of each channel is now better appreciated.
In the blood vessel wall, gap junctions link SMCs with other SMCs, ECs with other ECs and, in many blood vessels, SMCs with ECs. Cx37, Cx40 and Cx43 are the predominant isoforms of gap-junction proteins expressed in the vascular wall and, in rodents, the Cx37 and Cx40 isoforms are involved preferentially in myoendothelial gap-junction communication [144–146] (Figure 2). The number of myoendothelial gap junctions increases with a reduction in the size of the artery , a phenomenon that parallels the contribution of the EDHF-mediated responses with endothelium-dependent relaxations [148,149]. Endothelium and SMCs can communicate via these myoendothelial gap junctions physically, as Ca2+ and Ins(1,4,5)P3 can diffuse from one cell type to another [150–152], and electrically, as depolarization and hyperpolarization are conducted bi-directionally from one cell type to the other [151–156]. However, endothelium-dependent dilations do not simply propagate electronically, but involve a regenerative mechanism [155,157]. Blockers of gap junctions abolish or partially inhibit EDHF-like responses in many arteries and, in the rat mesenteric artery, antibodies directed against Cx40, when loaded selectively in ECs, block EDHF-mediated responses [144,158–161]. Furthermore, in mice, Cx40 is essential for the acetylcholine-activated regenerative endothelium-dependent vasodilation [157,162]. Activation of either SKCa or IKCa channels leads to endothelium-dependent hyperpolarizations and relaxations of VSMCs, but, in quiescent arteries (in the absence of vasoconstrictor stimulation), EDHF-mediated responses are associated with the preferential activation of SKCa channels and the contribution of myoendothelial gap junctions [113,131,133,163].
Taken in conjunction, the results of these In vitro experiments provide compelling evidence for a major contributing role of myoendothelial gap junctions in EDHF-mediated responses . However, experiments performed in vivo have generally failed to demonstrate such a significant role for myoendothelial gap junctions [165,166]. The origin of this discrepancy is unknown, but may involve the type and size of arteries studied in vivo, the presence of shear stress, sympathetic innervation and circulating hormones, as well as confounding factors such as the use of anaesthetics which inhibit gap junctions [131,164].
Additionally, the efflux of K+ ions associated with the activation of endothelial KCa can contribute to EDHF-mediated responses . The resultant moderate increase in the extracellular K+ concentration (from 1 to 15 mmol/l) can provoke the relaxation of VSMCs  by activating KIR channels  and the Na+/K+ pump . The activation of KIR channels and the Na+/K+ pump overcomes the small depolarizing effects linked to the increase in K+ ions and the net resultant is hyperpolarization and, thus, relaxation of the SMCs. This hypothesis was first demonstrated successfully in the hepatic and mesenteric arteries of the rat , and has been observed in many other blood vessels including human arteries [170–177].
However, this phenomenon is likely to occur only in specialized microdomains situated in the endothelial projections associated with myoendothelial gap junctions. Sections of endoplasmic reticulum densely expressing Ins(1,4,5)P3 receptors, Cx40, IKCa channels and Ca2+-sensing receptors are co-located in these endothelial projections [113,130–133,178]. Repetitive localized Ca2+ events (pulsars), driven by Ins(1,4,5)P3 and/or Ca2+ ions (Ca2+-induced Ca2+ release), originate from these endothelial Ca2+ stores. Ins(1,4,5)P3 can be generated by ECs (for instance following acetylcholine stimulation) or SMCs (for instance following phenylephrine stimulation). In the latter case, Ins(1,4,5)P3 would diffuse toward ECs (possibly with Ca2+ ions) through the myoendothelial gap junctions [113,133,179,180]. The closely situated IKCa channels are activated by these Ca2+ pulsars and the resultant endothelial hyperpolarization can be transmitted to SMCs either via the myoendothelial gap junctions, as described previously, or be elicited by K+ ions accumulating in the restricted extracellular space surrounding these endothelial projections (Figure 3). In the rat mesenteric artery, K+ ion accumulation preferentially activates the Na+/K+ pump [113,181].
The precise role of the Ca2+-sensing receptor at the site of these endothelial projections is not completely understood. Stimulation of the Ca2+-sensing receptor results in selective IKCa channel-dependent endothelial hyperpolarization and endothelium-dependent vascular smooth muscle hyperpolarization [178,182,183]. In quiescent arteries (in the absence of vasoconstrictor stimulation), the Ca2+-sensing receptor would be fully stimulated by the concentration of Ca2+ bathing the ECs and IKCa channel inactivated . Stimulation of smooth muscle (for instance by phenylephrine) opens the CaV channel and, in the small extracellular space surrounding myoendothelial projections, could create a localized Ca2+ sink. The endothelial Ca2+-sensing receptor would detect the changes in extracellular Ca2+ and allow the recruitment of endothelial IKCa channels. The subsequent endothelium-dependent hyperpolarization would restrain excessive activation of the smooth muscle  (Figure 3). Indeed, reducing the extracellular Ca2+ concentration enables IKCa channel activation by acetylcholine [113,163]. Therefore endothelial projections and myoendothelial gap junctions are key structures intrinsically associated with extracellular and intracellular Ca2+ homoeostasis in both ECs and VSMCs [113,131,133].
However, in some blood vessels, K+ does not evoke, or inconsistently produces, relaxations and hyperpolarizations [136,159,160,184,185], indicating that, in these blood vessels, the contribution of K+ ions in EDHF-mediated responses must be, if anything, minimal. The involvement of gap junctions and K+ ions are not necessarily mutually exclusive. The relative proportion of each mechanism almost certainly depends on numerous parameters, including the extracellular concentrations in K+ and Ca2+ ions associated with the state of activation of the underlying VSMCs, the density of myoendothelial gap junctions and the level of the expression of the appropriate isoforms of Na+/K+-ATPase and/or KIR channels [1,131].
Genetically modified animals
In transgenic mice with SK3 gene expression under the control of dietary doxycycline, the suppression of SK3 expression in ECs depolarizes both ECs and VSMCs, reduces the diameter of resistance vessels in situ and increases arterial blood pressure, a reversible phenotype upon restoration of endothelial SK3 expression . Disruption of the IK1 gene reduces the hyperpolarization of ECs and SMCs in response to acetylcholine and decreases the associated vasodilation, because of a substantial reduction in EDHF-mediated responses. Moreover, the IK1 deletion also led to a significant increase in arterial blood pressure and to mild left ventricular hypertrophy . In double knockout mice lacking both SK3 and IK1, an addition of the detrimental effects provoked by the deletion of either gene is observed [188,189]. These results confirm that, in mice, endothelial SKCa and IKCa channels are fundamental determinants of endothelial hyperpolarization and EDHF signalling, and indicate that they actively control vascular tone and contribute to the overall regulation of the circulation.
Cx37 and Cx40 are the predominant gap-junction proteins in murine ECs . Cx40 is involved in endothelial homocellular gap junctions and also in heterocellular gap junctions linking ECs not only to SMCs, but also to renin-producing juxtaglomerular cells. The presence of the latter gap-junction communication is required in order to maintain the Ca2+-dependent inhibitory effects of AngII and that of intrarenal pressure on renin secretion and synthesis, suggesting that the endothelium is strongly involved in the regulation of the renin system. Mice deficient for Cx40 are hypertensive. However, alteration in the control of renin release only partially explained the hypertension observed in Cx40-knockout mice . The arterioles of these animals also had a reduced spread of dilation in response to endothelium-dependent vasodilators and irregular arteriolar vasomotion [192–194].
These results show that the deletion of each key molecular component of EDHF-mediated responses is associated with haemodynamic alterations, suggesting that this endothelial pathway contributes to the overall regulation of arterial blood pressure.
Endothelial dysfunction and therapeutic interventions
Endothelial dysfunction, observed in various cardiovascular diseases, is associated with a decrease in NO synthesis and/or a loss of its biological activities. However, alterations in the EDHF pathway can also contribute to these endothelial dysfunctions or conversely compensate for the loss in NO bioavailability. Alterations of EDHF-mediated responses have been reported with aging and under various pathological conditions (hypertension, atherosclerosis, hypercholesterolaemia, heart failure, ischaemia/reperfusion, angioplasty, eclampsia, diabetes and sepsis) [1,2,195].
No drug is available which has been designed to target EDHF-mediated responses. Nevertheless, therapeutic interventions, with beneficial effects on the cardiovascular system, such as angiotensin-converting enzyme inhibitors, AngII type 1 receptor blockers and phosphodiesterase 3 inhibitors [1,195,196], can restore these responses, suggesting that the improvement in the EDHF pathway contributes to the observed beneficial effect. Similarly, various so-called non-pharmacological therapeutic strategies, including exercise and supplementation with oestrogens, omega-3 polyunsaturated fatty acids, polyphenol derivatives, potassium and/or calcium, help to reverse endothelial dysfunction such as blunted EDHF-mediated responses [1,195].
The improvement or restoration in EDHF responses has not yet been the direct purpose of any pharmaceutical effort. SKA-31, a preferential activator of murine IKCa channels, potentiates EDHF-mediated responses In vitro and lowers mean arterial blood pressure in normotensive and in AngII-hypertensive mice . However, IKCa channels are required for the differentiation of VSMCs, as well as for their proliferation and migration [126,127,197,198]. Selective blockade of IKCa with TRAM-34  prevents phenotypic changes of smooth muscle and coronary artery neointimal formation in two different models of post-angioplasty restenosis and the development of atherosclerosis in ApoE (apolipoprotein E)−/− mice [126,198,200]. IKCa channels are also involved in the proliferation of ECs  and various cancerous cells [202,203]. Therefore activators of IKCa channels may have some unwanted detrimental effects.
Additionally, activation of endothelial TRPCs and SKCa channels, Ca2+-sensing receptors, smooth muscle KIR channels and/or specific isoform(s) of Na+/K+-ATPase, as well as facilitating myoendothelial communication and increasing the expression of appropriate Cxs, channels and receptors, may represent new potential targets. However, the precise role of these various molecular elements is far from being completely understood. For instance, the TRPV4 channel could appear as a promising target in cardiovascular diseases as this cationic channel is involved in Ca2+ entry following endothelial stimulation. . Indeed, the arterial responses to shear stress critically depend on the activation of this endothelial channel, and both the NO- and EDHF-mediated components of acetylcholine-induced vasodilation are attenuated in TRPV4-deficient mice [205,206]. However, GSK1016790A, a specific and potent agonist, which as expected increases endothelial intracellular Ca2+ concentration and produces endothelium-dependent relaxations, also causes endothelial failure, circulatory collapse and death . Rotigaptide (ZP123), an anti-arrhythmic peptide that prevents uncoupling of Cx43-mediated gap-junction communication , has no effect on basal vascular tone and does not enhance endothelium-dependent or -independent vasodilation in the forearm arterial circulation of healthy subjects . Whether or not augmenting Cx43 communication would improve endothelial function in patients with vascular disease and whether or not Cx40 would be a more appropriate target than Cx43 remain to be determined.
CONCLUSION AND PERSPECTIVES
ECs control the tone of the underlying vascular smooth muscle by releasing numerous vasoactive substances, including NO, ROS, K+ ions and metabolites of arachidonic acid (e.g. prostacyclin, EETs and LOX derivatives). Furthermore, the endothelial monolayer behaves as a conductive tissue propagating an electrical signal along the axis of the blood vessel by means of homocellular gap junctions and throughout the vascular wall itself by means of myoendothelial gap junctions. Endothelium-dependent relaxations, independent of the production of NO and PGI2, probably play an important role in cardiovascular physiology in numerous animal species and in humans. They can act as a back up system when NO is inhibited or reduced, but this is not necessarily the case.
However, it is often difficult to reach a conclusion as to the true importance of endothelium-dependent hyperpolarizations because of the use of unspecific pharmacological tools and the lack of electrophysiological measurements. The mechanisms underlying endothelium responses must be carefully dissected in order to be properly identified. The synthesis of more selective compounds, such as non-peptidic inhibitors and activators of SKCa and IKCa channels, may in the future allow the selective blockade/activation of EDHF-mediated responses and, hence, the proper determination of their physiological role in the human circulation. The limited information available suggests that if better (i.e. more potent, more specific and, if possible, orally active) pharmacological tools are developed to modulate the role of the various molecular constituents underlying EDHF-mediated responses, it may be possible to determine whether or not putative cardiovascular targets identified within this pathway are drugable.
The authors' work received no specific grant from any funding agency in the public, commercial or not-forprofit sectors.
- CaV channel
voltage-dependent Ca2+ channel
C-type natriuretic peptide
endothelium-derived hyperpolarizing factor
- KATP channel
ATP-sensitive K+ channel
- KCa channel
Ca2+-activated K+ channel
- BKCa channel
large conductance KCa channel
- IKCa channel
intermediate conductance KCa channel
- IP receptor
- KIR channel
inward rectifier K+ channel
natriuretic peptide receptor
reactive oxygen species
- SKCa channel
small conductance KCa channel
smooth muscle cell
transient receptor potential canonical channel
transient receptor potential vanilloid channel