Vascular Smooth Muscle Modulates Endothelial Control of Vasoreactivity via Reactive Oxygen Species Production through Myoendothelial Communications

Background Endothelial control of vascular smooth muscle plays a major role in the resulting vasoreactivity implicated in physiological or pathological circulatory processes. However, a comprehensive understanding of endothelial (EC)/smooth muscle cells (SMC) crosstalk is far from complete. Here, we have examined the role of gap junctions and reactive oxygen species (ROS) in this crosstalk and we demonstrate an active contribution of SMC to endothelial control of vasomotor tone. Methodology/Principal Findings In small intrapulmonary arteries, quantitative RT-PCR, Western Blot analyses and immunofluorescent labeling evidenced connexin (Cx) 37, 40 and 43 in EC and/or SMC. Functional experiments showed that the Cx-mimetic peptide targeted against Cx 37 and Cx 43 (37,43Gap27) (1) reduced contractile and calcium responses to serotonin (5-HT) simultaneously recorded in pulmonary arteries and (2) abolished the diffusion in SMC of carboxyfluorescein-AM loaded in EC. Similarly, contractile and calcium responses to 5-HT were decreased by superoxide dismutase and catalase which, catabolise superoxide anion and H2O2, respectively. Both Cx- and ROS-mediated effects on the responses to 5-HT were reversed by L-NAME, a NO synthase inhibitor or endothelium removal. Electronic paramagnetic resonance directly demonstrated that 5-HT-induced superoxide anion production originated from the SMC. Finally, whereas 5-HT increased NO production, it also decreased cyclic GMP content in isolated intact arteries. Conclusions/Significance These data demonstrate that agonist-induced ROS production in SMC targeting EC via myoendothelial gap junctions reduces endothelial NO-dependent control of pulmonary vasoreactivity. Such SMC modulation of endothelial control may represent a signaling pathway controlling vasoreactivity under not only physiological but also pathological conditions that often implicate excessive ROS production.


Introduction
In vessels, smooth muscle cell (SMC) responses are controlled by endothelial cells (EC) via complex intercellular signaling processes. Regarding vascular tone, the best described interaction is the endothelial-dependent vascular relaxation through the release of nitric oxide (NO) and vasoactive prostanoids [1]. Another pathway is associated with the hyperpolarization of both the EC and SMC and mediated by an endothelium-derived hyperpolarizing factor (EDHF). Whereas EDHF identity is still under debate and may differ among species and vascular segments, there is evidence that EDHF-mediated responses involve epoxyeicosatrienoic acids (EETs) [2], potassium ions and channels [3], reactive oxygen species (ROS) such as hydrogen peroxide (H 2 O 2 ) [4] and myoendothelial junctions [5].
In the pulmonary circulation, endothelial control of smooth muscle tone is of critical importance to maintain low pressure and low resistance. In this connection, most of the medical treatments used in pulmonary hypertension (e.g., NO, prostacyclin…) mimic and amplify the physiological control of EC on SMC. For instance, serotonin (5-HT), a potent vasoconstrictor whose concentration is increased in pulmonary arterial hypertension (PAH) [6,7], acts on both SMC and EC. On one hand, 5-HT can generate NO in pulmonary arterial EC [8]. On the other hand, 5-HT produces ROS in bovine and mice pulmonary arterial SMC and the production of superoxide anion (O 2 N -) facilitates 5-HTinduced pulmonary vasoconstriction [9,10]. O 2 N can interact with NO to produce peroxynitrite which decreases endothelial NO synthase (eNOS) expression and/or loss of eNOS substrate Larginine or cofactors tetrahydrobiopterin (BH4) [11]. We thus hypothesized an EC/SMC crosstalk through NO and O 2 N -. Gap junctions mediate intercellular communication between EC and SMC (myoendothelial gap junctions) and allow direct exchange of ions and small molecules in various tissues including pulmonary arteries [5,12]. Whereas gap junctions are expressed in pulmonary artery [13,14], participate to vascular tone in general [12] and are stimulated by 5-HT in coronary and mesentery SMC [15], no study has been performed on the role of gap junctions in the pulmonary arterial vasoreactivity to 5-HT. Such communication may well participate to the EC/SMC crosstalk. Interestingly, resistance vessels are very important for blood flow regulation and the incidence of myoendothelial gap junctions is higher in resistance than in conduit arteries [16].
In the present study, we have therefore addressed, in small intrapulmonary arteries (IPA), the issue of expression, localization and function of the connexins (Cx) 37, 40 and 43 which are the Cx usually expressed in small vessels. Peptides homologous to the gap 26 and 27 domains of the extracellular loops of Cx 43 ( 43 Gap 26) or Cx 37 and 43 ( [37][38][39][40][41][42][43] Gap 27) respectively, interrupt intercellular communications in a Cx-specific manner and attenuate the calcium and contractile responses to 5-HT. Such effect was reversed by (1) the removal of the endothelium-dependent NO function or (2) the decrease of ROS production. Electronic paramagnetic studies demonstrated that ROS and more specifically O 2 N was produced by the smooth muscle in response to 5-HT. In the same way, the NO and cyclic GMP content were assessed in response to 5-HT.
Altogether, the findings suggest that 5-HT produces O 2 N in the smooth muscle and NO in the endothelium. O 2 N passes through the myoendothelial junctions to decrease endothelial NO production and thus strengthen pulmonary vasoreactivity. This is the first time that a negative control of the endothelial NO function by the smooth muscle is demonstrated. Since ROS production is increased in many cardiovascular diseases, such process could be of great importance under pathological conditions.

Ethics Statement
All animal work has been conducted according to relevant international and national guidelines in accordance with the recommendations of the Weatherall report, ''The use of nonhuman primates in research''. All the protocols used were approved by our local ethics committee named Comité d'éthique régional d'Aquitaine. The protocol number is AP 2/11/2005.

Intracellular Calcium and External Diameter Recording
IPA were cannulated as previously described [17] in an arteriograph (DMT, Denmark). Residual blood was removed by perfusing the vessels with Krebs-HEPES solution, and then vessels were maintained in a no-flow state and held at a constant transmural pressure of 10 mmHg and at a constant temperature of 37uC. Vessels were incubated 1 h in 2 mM Fura-PE3-AM at 37uC and rinsed for 10 min with Krebs-HEPES solution. The arteriograph was then placed on the stage of an inverted epifluorescence microscope (Olympus IX70) equipped with a x10, UPlanApo/0.40 W water-immersion objective (Olympus). The source of excitation light was a xenon arc lamp (175 W), and excitation wavelengths were selected by a wheelfilter (Sutter Instrument Co.) composed of 345 nm and 380 nm filters. Digital images were sampled at 12-bit resolution by a fast-scan, cooled charge-coupled device camera (CoolSNAP fx Monochrome, Photometrics). Preparation was then alternately excited at 345 and 380 nm and emission fluorescence at 510 nm of both excitation wavelengths was imaged every 20 seconds. As previously described, regions of interest were drawn on the vascular wall to determine calcium signal from smooth muscle cells and background was substracted using MetaFluor (Universal Imaging Software) [17]. External diameter of the vessels was measured on these images using MetaMorph (Universal Imaging Software). Contraction was expressed as a percentage related to the percentage of decrease of the initial external diameter.
Either 10 mM or cumulative concentrations of 5-HT (Sigma) were added to the bath, which was bubbled with air. After recording of this first contraction, vessels were washed and pharmacological substances were added both to the bath and to the perfusion solution inside the vessel for one hour. We inhibited gap junction communications through Cx 37 and Cx 43, Cx 40, or Cx 43 with synthetic connexin-mimetic peptides: 37,43 Gap 27 (SRPTEKTIFII), 40 Gap 27 (SRPTEKNVFIV) and 43 Gap 26 (VCYDKSFPISHVR) respectively (Genscript). Specificity of these peptides was checked with an inactive homologous peptide: 43 Gap 20 (EIKKFKYGC). The role of superoxide anion was investigated with 300 U/ml poly-ethylene glycol-superoxide dismutase (PEG-SOD, Sigma) combined to 600 U/ml poly-ethylene glycol-catalase (PEG-catalase; Sigma). Once substances added, a second contraction with either 10 mM or cumulative concentrations of 5-HT was performed. We performed time-matched controls to check that we can produce two similar cumulative concentration-response curves (CCRC) to 5-HT with one hour of delay between the two CCRC ( Figure S1). When indicated, vessels were perfused before contraction protocol with 100 ml of 0.3% 3-((3-cholamidopropyl)diethylammonio)-1-propane sulphonate (CHAPS, Sigma) in order to remove endothelium, as previously described [18].
We checked that such protocol suppressed endothelial function by testing the absence of a relaxation with 10 mM carbamylcholine of 40 mM KCl-induced preconstricted pulmonary arteries ( Figure  S2). CHAPS on its own had no deleterious effect on the contraction and calcium signal induced by high potassium solution (KCl 40 mM) ( Figure S3).

Isometric Contraction Measurements
Isometric contraction was recorded in intrapulmonary arterial rings with an external diameter of 1.5-2 mm as reported previously [19,20]. Briefly, mechanical properties were assessed using organ bath and transducer systems, coupled to IOX software (EMKA Technologie). As determined in preliminary experiments, tissues were set at optimal length by equilibration against a passive load of 0.8 g. At the outset of each experiment, K + -rich (80 mM) solution, obtained by substituting an equimolar amount of KCl for NaCl from Krebs-HEPES solution, was applied in order to obtain a reference contraction used to normalize subsequent contractile responses. Contractile properties were tested by constructing a CCRC to 5-HT (10 nM to 100 mM) for each ring. When indicated, 37,43 Gap 27 was incubated during 1 hour before CCRC to 5-HT.
Endothelial function was tested by relaxation with 10 mM carbamylcholine of 0.3 mM phenylephrine-induced preconstricted pulmonary arterial rings. All experiments were performed at 37uC.
Quantitative RT-PCR RNA Extraction. IPA from one rat was homogenized using 600 ml of Trizol (Invitrogen), then, 120 ml of chloroform (Sigma) was added. The RNA was extracted from the aqueous phase after centrifugation at 15,000 g for 15 min. RNA was precipitated in the presence of isopropanol (Sigma) at 220uC overnight. The pure RNA was obtained by centrifugation at 15,000 g for 15 min and was washed with 80% ethanol (Sigma). The concentration of RNA was measured spectrophotometrically by GeneQuant RNA/DNA calculator (Amersham Pharmacia). The total RNA (1 mg) was reverse transcribed into cDNA by using AMV reverse transcriptase (Promega), RNase inhibitor, and oligo d(T) as a primer at 42uC for 60 min followed by heating at 94uC for 3 min.
Real-time Quantitative Polymerase Chain Reaction (PCR). Real-time quantitative PCR was performed with a Rotor-Gene 2000 (Corbett Research). Triplicate PCR reactions were assembled in 0.1-ml strip tubes containing cDNA from 10 ng of total RNA, 0.2 ml of 506Titanium Taq DNA Polymerase combined to its buffer (Clontech Laboratories), 1 mM dNTP, each of the appropriate primer (Sigma Genosys; see table 1 for concentrations and sequences), and 0.56 SYBR Green (Molecular Probes).
The PCR was performed under the following conditions: denaturation at 95uC for 15 s, annealing temperature (68-70uC) depending on specific primers for 15 s, and extension at 70uC for 15 s, these steps were repeated during 40 cycles. Data collection was performed after each extension step, at a temperature of at least 3.5uC lower than the melting temperature of the amplicon (generally between 80-85uC) to eliminate non-specific fluorescence signal. PCR negative controls were systematically made by using water instead of cDNA. All specific primers were designed by using the primer analysis software (Oligo 6.6, Molecular Biology Insights). The efficiency of the PCR reactions was always more than 90%. Specificity of the amplified PCR products was checked with melting curve analysis and by electrophoresis analysis on a 2% agarose gel containing SYBR Green.

Western Blot
IPA from 4 rats were homogenized on ice in lyses buffer containing 1% IGEPAL, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM amino-ethyl-benzenesulfonyl fluoride hydrochloride (AEBSF), 1.5 mM aprotinin and 0.1 mM leupeptin (Sigma-Aldrich). After 10 minutes of centrifugation at 15,000 g, supernatant is reduced in Laemmli buffer and heated at 90uC for 6 min. Protein extract were subjected to electrophoresis on a 10% acrylamid reducing gel, and transferred to polyvinylidene fluoride (PVDF) membranes (Immobilon-P, Millipore). The immunoblots were then incubated using either rabbit anti-Cx 37, rabbit anti-Cx 40 or mouse anti-Cx 43 (Zymed) overnight at 4uC. After incubation with appropriate secondary antibodies coupled to horseradish peroxydase (HRP, Santa Cruz) for 2 h at room temperature, immunoblots were then revealed by enhanced chemiluminescence acquired using Kodak Image Station 4000 MM. Band densities were quantified using GeneTool software (SynGene). Immunoblots were then stripped and revealed with mouse anti-b-actin for 1 h at room temperature.
Immunofluorescence IPA were fixed in Formalin (Sigma) for 10 min and then embedded in OCT-Compound, frozen at -20uC and cut in 10 mm sections with a cryostat. Sections were first incubated in 0.1% Triton X-100 (Sigma) and 1% bovine serum albumin in PBS for 1 h at room temperature, and then in the primary antibodies (same antibodies as the ones used for Western Blot) overnight at 4uC. Sections were then incubated with the appropriate secondary antibodies coupled to Alexa 546 (Molecular Probes) for 2 h at room temperature. Nuclei were labeled with 45 mM Hoechst 33342 (Molecular Probes). Sections were then observed with a laser scanning confocal microscope TE2000 (Nikon) with a x 60, 1.40 NA plan apochromat oil-immersion objective. Excitation was obtained with a diode laser at 408 nm to observe nuclei, an argon laser at 488 nm to observe autofluorescence of the internal elastic lamina and a helium-neon laser at 543 nm to observe Cx labeling. The emitted light was filtered as appropriate: 450635 nm for nuclei (blue), 515630 nm for internal elastic lamina (green), and 605675 nm for Cx labeling (red).

Dye Transfer
IPA were mounted in an arteriograph as previously described (see Intracellular calcium and external diameter recording). The vessels were allowed to equilibrate and pressurized at 10 mmHg for 30 min, and then lumen was loaded with 10 mM carboxyfluorescein-AM (Molecular Probes) for 30 min before washout with Krebs-HEPES for 30 min. Where indicated, vessels were perfused with 300 mM 37,43 Gap 27 for 1 h before carboxyfluorescein loading. Vessels were subsequently removed from the Table 1. Sequences of the primer pairs (S: sense; AS: antisense) for housekeeping genes (GAPDH, HPRT, PLRPO and YWHAZ) and genes of interest (Cx 37, Cx 40 and Cx 43) are shown as well as GenBank accession number, product length, product Tm and concentrations. arteriograph and fixed in Formalin for 20 min at room temperature before cryopreservation in OCT compound at 220uC. Cryosections (10 mm thick) were prepared and mounted in Fluorescent mounting medium (Dako) and imaged with laser scanning confocal microscope TE2000 (Nikon) with a x 60, 1.40 NA plan apochromat oil-immersion objective. Preparation was excited with an argon laser at 488 nm, and fluorescence emission was filtered at 515 nm630.

Electron Paramagnetic Resonance (EPR) Recordings
Intrapulmonary arteries with an external diameter of 0.3-2 mm were used for EPR studies and cGMP measurements.
Superoxide Anion Spin Trapping. Pulmonary arteries were incubated in the spin trap solution containing 500 mM 1-hydroxy-3methoxycarbonyl-2,2,5,5-tetramethylpyrrolidin (CMH, Noxygen), 25 mM deferoxamine (Sigma) and 5 mM N,N-diethyldithiocarbamate (DETC, Sigma) in Krebs-HEPES at 37uC for 45 min. 5-HT 0.1 mM was added during the spin trap incubation; PEG-SOD 300 U/mL was incubated during 45 min before spin trap incubation. Reaction was stopped by freezing the sample in liquid nitrogen. Samples were then analyzed by EPR spectrometry on a tabletop x-band spectrometer miniscope (MS200, Magnettech). Spectra of the oxidized product of CMH (CMN) were recorded at 77uK using a flask Dewar. Acquisition parameters were as followed: Bo Field: 33416150 G, microwave power: 10 dB, amplitude modulation: 5 G, sweep time: 60 sec, gain: 300 and 3 scans [21]. Signals were quantified by measuring the total amplitude, after correction of baseline and normalized to the protein quantity of the sample in mg/ml. NO Spin Trapping. Detection of NO production was performed using the technique with Fe 2+ -DETC as spin trap. IPA were placed in 0.5 ml of Krebs-HEPES and then treated with 0.5 ml of colloid Fe(DETC) 2 as previously described and incubated for 45 min at 37uC [22]. When indicated, 0.1 mM 5-HT was added during Fe(DETC) 2 incubation. NO measurements were performed at 77uK using EPR spectrometry. Instrument settings were: Bo Field: 3285680 G, microwave power: 10 dB, amplitude modulation: 7 G, sweep time: 150 sec, gain: 900 and 3 numbers of scan. Signals were quantified by measuring the total amplitude, after correction of baseline and normalized to the protein quantity of the sample in mg/ml.

cGMP Measurements
Intrapulmonary arteries were incubated in Krebs-HEPES with or without 5-HT 0.1 mM and with or without PEG-SOD + PEGcatalase (300 U/ml and 600 U/ml respectively) at 37uC for 15 min, cooled in liquid nitrogen and then stored at 280uC. Arteries were homogenized in ice-cold trichloroacetic acid (5%, Sigma) to extract cGMP. cGMP content was assayed as described in the procedure of an ELISA kit obtained from Cayman Chemical & Co. cGMP level was normalized to tissue protein content in mg/ml.

Statistical Analysis
All results are expressed as means6SEM, n indicates the number of IPA for the calcium and contractile studies, dye transfer and immunofluorescence, the number of rats for quantitative RT-PCR, EPR and cGMP measurements, and the number of experiments from 3 pools of rats (4 rats per pool) for Western Blot. Cumulative concentration response curves to 5-HT were fitted to the logistic equation with Origin 6.0 software to determine the maximum effect of 5-HT for each experiment, and a non parametric test for paired samples (Wilcoxon test) was performed on the maximum effect of 5-HT with or without drug.
Quantitative RT-PCR results were analyzed with the GeNorm method [23]. Statistical analyses were performed on all other data using a non parametric test for unpaired samples (Mann-Witney test). Values of P,0.05 were considered significant.

Results
Connexins 37, 40 and 43 are expressed and functional in the pulmonary arterial wall Although very few studies have been performed with controversial results in pulmonary arteries, the three Cx 37, 40 and 43 have been observed in pulmonary arteries [13,14]. Quantitative RT-PCR and Western Blot experiments evidenced the presence of the mRNA and proteins for the three Cx 37, 40 and 43 in small intrapulmonary arteries (IPA) (Figure 1A and B). Immunofluorescent labeling studies confirmed the presence of the Cx 37, 40 and 43 proteins ( Figure 1C). The autofluorescence of the external and internal elastic lamina in green has been previously observed in various vessels [24,25] and is useful to delimit the smooth muscle layers from the endothelial cells and the adventitia. Some punctate labeling was observed for the Cx 37, 40 and 43 in the endothelium and similar labeling was observed in the smooth muscle for the Cx 37 and 40 ( Figure 1C). Punctate staining is characteristic of connexins labeling and has been previously shown in various tissues including vessels [4,24,25]. Since (1) Cx 37 and 43 are expressed in IPA, (2) they have a functional role in both the calcium and contractile signal and (3) they are localized in the endothelium, we then addressed the functional role of the Cx 37 and 43 in the myoendothelial communications and/or the communications between endothelial cells. After loading endothelial cells with a fluorescent dye (carboxyfluorescein), the dye diffused to the smooth muscle cells in control conditions ( Figure 1D left) whereas it stayed in the endothelium when the vessels were pretreated with 37,43 Gap 27 300 mM (Figure 1D right).
Altogether, it can be concluded that the myoendothelial communications are functional and involve the Cx 43 and/or Cx 37.

Vasoreactivity to 5-HT is controlled by gap junctions and endothelium
We then examined if gap junctions had any influence on (1) the intracellular calcium concentration in smooth muscle and (2) the contraction in IPA. Consequently, we checked if any of these connexins had any role in the simultaneously recorded calcium and contractile signals in response to the agonist 5-HT in IPA. Incubation of the vessels with 40 Gap 27 300 mM, the Cx-mimetic peptide used as a specific blocker of the Cx 40, had no effect on both calcium and contractile signals ( Figure S4). In contrast, incubation with 37,43 Gap 27 300 mM, the Cx-mimetic peptide used as a specific blocker of the both Cx 37 and 43, strongly decreased both calcium and contractile signals to 5-HT (Figure 2A and B left). Incubation with 43 Gap26 300 mM, the specific blocker of the Cx 43 alone, had similar effects on calcium and contractile signals to those of the 37,43 Gap 27 ( Figure S5) suggesting that the Cx 43 is involved in the response to 5-HT. Since 43 Gap 20 is a peptide homologous to the intracellular loop of the Cx 43, it can be used as an inactive analog of the Gap 27 specific of the Cx 43 [26]. We observed that 43 Gap 20 300 mM had no effect on the calcium and contractile signals to 5-HT confirming the specificity of these Cx-mimetic peptides ( Figure S6).
Since gap junctions are localized in between smooth muscle cells and/or in between endothelial cells and/or in between smooth muscle and endothelial cells (myoendothelial junctions), we removed the endothelium of the vessels to discriminate the role of these different types of communications. The blocking effect of the 37,43 Gap 27 300 mM was absent in endothelium-denuded vessels ( Figure 2B right) suggesting that the myoendothelial and/or the endothelial junctions are important for the calcium and contractile signals to 5-HT.
Isometric tension recordings showed that whereas 37,43 Gap 27 300 mM blocked the CCRC to 5-HT (0.01-100 mM), it had no effect on the CCRC to phenylephrine (0.1 nM-3 mM), to endothelin-1 (0.1-5000 nM) and to high potassium solutions (4.7-100 mM, Figure S7). Moreover, in systemic vessels such as renal arteries, the CCRC to 5-HT were identical in the presence or in the absence of 43 Gap 26 300 mM ( Figure S8A) whereas in aorta, 43 Gap 26 300 mM partially inhibited the CCRC to 5-HT ( Figure S8B).

NO and O 2
N are involved and interconnected in the calcium and contractile responses to 5-HT Since NO is a major element contributing to the endothelial control of vasoreactivity, we checked the effect of L-NAME, a NO synthase inhibitor, on the blocking effect of 37,43 Gap 27 previously observed on the simultaneously recorded calcium and contractile signals to 5-HT. L-NAME 100 mM reversed the blocking effect of 37,43 Gap 27, 300 mM, on both the intracellular calcium concentration and contraction to 5-HT ( Figure 3A) whereas L-NAME 100 mM alone had no effect ( Figure S9) confirming the importance of NO in the endothelial control of the vasoreactivity through gap junctions. We hypothesized that O 2 N production by 5-HT could account for a reduction of the endothelial NO production leading to an increased contraction since O 2 N is a small element which likely passes the membranes through gap junctions and is able to scavenge NO and/or reduce its production thus decreasing its relaxant effect [11,27]. Indeed, poly-ethylene glycol-superoxide dismutase (PEG-SOD) 300 U/ml and poly-ethylene glycolcatalase (PEG-catalase) 600 U/ml, two substances which catabolise O 2 N -, reduced calcium and contractile responses to 5-HT and such effect was reversed by L-NAME 100 mM ( Figure 3B). Electron paramagnetic resonance (EPR) recordings demonstrated that 5-HT produced O 2 N -( Figure 4A). This O 2 N production was prevented by PEG-SOD 300 U/ml and unchanged when the endothelium was removed by CHAPS 0.3% ( Figure 4A). We checked that CHAPS 0.3% had no effect on its own on basal O 2 N levels ( Figure S10). Consequently, 5-HT produced O 2 N originating from smooth muscle. Moreover, we also directly detected the production of NO by 5-HT with EPR experiments ( Figure 4B). When the vessels were endothelium-denuded by CHAPS or treated with L-NAME 100 mM, the production of NO by 5-HT was abolished ( Figure 4B) showing respectively that 1) NO produced by 5-HT originates from the endothelium and 2) the product detected by EPR is linked to NO synthases.
Finally, since, on one hand, 5-HT produced NO that induced relaxation via cGMP production in SMC and, on the other hand, it also produced O 2 N that antagonized NO, we measured the cGMP content in order to determine if the overall result was in favor of a relaxant or a contractile effect. Whereas 5-HT significantly reduced the basal cGMP content, in the additional presence of PEG-SOD 300 U/ml and PEG-catalase 600 U/ml, it did increase the cGMP content ( Figure 4C). This result indicates that O 2 N produced by 5-HT in SMC decreases the NO produced by 5-HT in EC, thus decreasing cGMP level leading to an increased vascular contraction ( Figure 5A). To the best of our knowledge, this is the first description of a negative feedback of smooth muscle on endothelial relaxant control of vasoreactivity.

Discussion
Our findings provide the first direct evidence of a smooth muscle negative feedback on NO-dependent endothelial vasodilatation thus maintaining some contraction in vessels under physiological pressures. In particular, we showed that the Cx 43 is expressed and functional in IPA and is necessary to develop a sustained agonist-mediated contraction in an integrated vascular model. Our data also support that such process involves agonistinduced production of O 2 N in the smooth muscle and NO in the endothelium. O 2 N -, via myoendothelial junctions, reduces NOinduced cGMP production in smooth muscle and the resulting vessel relaxation ( Figure 5).
In small vessels, gap junctions are usually mainly involved in spontaneous oscillations of membrane potential, intracellular calcium concentration and contraction [24,28,29]. Gap junctions have also been shown to be largely implicated in the mediatorsand endothelium-related relaxation [5,30]. Few studies have  ). B, PEG-SOD (300 U/ml) and PEG-catalase (600 U/ml) also inhibited the calcium and contractile responses to 5-HT (black circles, n = 7 vessels). Both 37-43 Gap 27 and PEG-SOD plus PEG-catalase effects were reversed by L-NAME 100 mM (A and B respectively, black triangles, n = 6 and 9 vessels respectively). Black squares are control dose-response curves to 5-HT (n = 13-16 vessels). * indicates a significant difference when P,0.05. doi:10.1371/journal.pone.0006432.g003 addressed the role of gap junctions in agonist-induced contraction and the results vary according to studies. Indeed, gap junction blockers either increase or decrease the contraction and intracellular calcium to phenylephrine in rat mesenteric arteries [24,29]. Interestingly, arteries that express Cx 43 to a high degree, such as the rat aorta, are more sensitive to adrenergic agonists than those expressing Cx 43 at a low level such as the caudal artery [30,31]. Altogether, these results are consistent with our study namely stimulation of gap junctions, and especially Cx 43, is important to strengthen the agonist-related vascular contraction. However, the mechanisms involved in this regulation of the vascular tone by gap junctions remain elusive.
In vessels, Ca 2+ and inositol 1, 4, 5 trisphosphate (IP 3 ) are intercellular messengers that pass through the intercellular communications and especially the myoendothelial communications [28,33]. In a model of cocultured endothelial and smooth muscle cells, both IP 3 and calcium originating in the vascular SMC appear to cross myoendothelial communications to induce a secondary calcium increase in endothelial cells whereas only calcium, and not IP 3 , originating from EC crosses myoendothelial junctions to generate a calcium dependent response in the vascular SMC [33]. In an integrated model of pressurized mesenteric arteries, such regulation by surrounding SMC via myoendothelial gap junctions, even under basal conditions, has also been described [28]. However, under those conditions, IP 3 from SMC increases calcium in EC and, consequently, enhances endothelium-related relaxation rather than contraction. In our study, the negative control of this secondary endothelial calcium signal linked to relaxation allows the smooth muscle to produce a sustained rather than a transient contraction.
There are some lines of evidence from the literature that O 2 N -(1) is produced by 5-HT in mice and bovine pulmonary arterial SMC and (2) participates to the contraction or proliferation of SMC [10,34]. Although, Liu and Folz [10] used an integrated model of mouse intrapulmonary arteries, they did not address the role of gap junctions in the contraction to 5-HT. In pulmonary artery, 5-HT is well known to induce contraction via stimulation of 5-HT receptors and SMC proliferation via stimulation of 5-HT transporter and/or 5-HT receptors [6]. O 2 N production in mice intrapulmonary arteries is insensitive to GR 127935, a selective inhibitor of 5-HT1B/D receptor and it is sensitive to 5-HT transporter in bovine pulmonary arterial SMC [10,35]. We did not address the role of 5-HT transporter and/or receptor in our model as this issue deserves future study. In the same way, Liu and Folz suggested that O 2 N was produced via a NADPH oxidase pathway [10] and this hypothesis could apply to our study. O 2 N is increased in several models of hypertension including pulmonary arterial hypertension [36,37]. Moreover, it has been observed that increased contraction to 5-HT in aortic strips of hypertensive rats is inhibited by the blockade of the gap junctions [38]. Finally, pulmonary arterial hypertension is linked to an endothelial dysfunction [39]. These data suggest that the process we propose, namely the negative regulation of the endothelial relaxation by the smooth muscle, could also be of great interest in vascular pathophysiological conditions.
In conclusion, we evidenced vascular smooth muscle can negatively control the endothelial relaxation in order to produce a sustained contraction in small vessels at physiological pressures. This control involves O 2 N -, myoendothelial junctions and regulation of endothelial NO function. In that respect, modifications of this process could potentially be involved in vascular pathologies and need to be further investigated. N production was measured by EPR spectrometry in the absence or in the presence of 5-HT 100 mM (black column and white column respectively). The production of O 2 N by 5-HT was reduced by PEG-SOD (A, grey column) whereas it was unchanged when vessels were endotheliumdenuded by CHAPS (A, light grey column). B, EPR spectrometry also showed an increase in NO production by 5-HT 100 mM (white column) and an absence of NO production by 5-HT 100 mM when vessels were endothelium-denuded by CHAPS (light grey column) or treated with L-NAME 100 mM (grey column). C, as measured by Elisa kit, basal cyclic GMP content (black column) was decreased by 5-HT (white column) whereas it was increased when PEG-SOD and PEG-catalase were added to 5-HT (light grey column). Data from EPR spectrometry are expressed as a ratio of the amplitude of the pic (A) out of the protein concentration of each pool of vessels in mg/ml. cGMP contents are expressed as a ratio of the quantity of cGMP in picomoles out of the protein concentration of each pool of vessels in mg/ml. Values are means6S.E.M. and n indicates the number of rats tested. * indicates a significant difference when P,0.05.   Figure S9 Effect of L-NAME on the simultaneously recorded calcium and contractile signals in response to 5-HT. Cumulative concentration-response curves to 5-HT (0.1-100 mM) were performed in the absence or in the presence of L-NAME 100 mM, a NO synthase inhibitor (black squares and circles respectively). Data are means6S.E.M. for 6 vessels and are expressed as a delta ratio (345 nm/380 nm) for calcium signal and a percentage of contraction (top and bottom respectively). Found at: doi:10.1371/journal.pone.0006432.s009 (0.29 MB TIF) Figure S10 Effect of endothelium removal and xanthine plus xanthine oxidase treatment on superoxide anion production. O2N was measured by using CMH spin trapping method and EPR spectrometry on control vessels (black column), vessels whose endothelium was denuded with CHAPS 0.3% (white column, top panel) and vessels treated with xanthine 50 mM plus xanthine oxidase 0.02 U/ml (white column, bottom panel). Data are means6S.E.M. and are expressed as a ratio of the amplitude of the pic (A) out of the protein concentration of each pool of vessels in mg/ml. n indicates the number of vessels tested. Found at: doi:10.1371/journal.pone.0006432.s010 (0.24 MB TIF)