Free Radical Biology & Medicine 49 (2010) 1008–1013

Contents lists available at ScienceDirect

Free Radical Biology & Medicine

j ourna l homepage: www.e lsev ie r.com/ locate / f reeradb iomed

Original Contribution

Hypoxia and reoxygenation-induced oxidant production increase in microvascularendothelial cells depends on connexin40

Guo Yu a,b,1, Michael Bolon a,c,1, Dale W. Laird c,d, Karel Tyml a,b,c,⁎a Lawson Health Research Institute, Critical Illness Research, London, ON, Canada N6C 2V5b Department of Medical Biophysics, University of Western Ontario, London, ON, Canadac Department of Physiology and Pharmacology, University of Western Ontario, London, ON, Canadad Department of Anatomy and Cell Biology, University of Western Ontario, London, ON, Canada

⁎ Corresponding author. Lawson Health Research InsLondon, ON, Canada N6C 2V5. Fax: +1 519 685 8341.

E-mail address: kyml@lhsc.on.ca (K. Tyml).1 These authors contributed equally to this study.

0891-5849/$ – see front matter © 2010 Elsevier Inc. Adoi:10.1016/j.freeradbiomed.2010.06.005

a b s t r a c t

a r t i c l e i n f o

Article history:Received 25 February 2010Revised 28 May 2010Accepted 2 June 2010Available online 9 June 2010

Keywords:Hypoxia and reoxygenationReactive oxygen speciesConnexin40NADPH oxidaseIntercellular electrical couplingFree radicals

Connexins (Cx) are recognized as structural constituents of gap-junctional intercellular communication(GJIC). However, their function may extend beyond facilitating the exchange of metabolites and electricalsignals between cells. In this study we asked if increased production of reactive oxygen species (ROS) inmicrovascular endothelial cells challenged by hypoxia/reoxygenation (H/R) requires Cx40, independent ofGJIC. Because we showed that this ROS increase depends on NADPH oxidase, we also asked if Cx40 function(i.e., Cx40-dependent reduction in interendothelial electrical coupling after H/R) requires NADPH oxidase.ROS increase was assessed in confluent monolayers of cultured endothelial cells derived from skeletal muscleblood vessels of wild-type (WT) and Cx40−/− mice and in monolayers of GJIC-deficient SKHep1 cellsoverexpressing GFP-tagged Cx40. Electrical coupling was assessed in WT cells and in cells lacking the NADPHoxidase subunit gp91phox or p47phox. H/R elicited a 70–80% ROS increase in WT but not in Cx40−/− cells.The increase was not affected by the gap junction blocker 18α-glycyrrhetinic acid or by preventing the cellsfrom establishing cell-to-cell contact. H/R increased ROS in SKHep1 cells expressing Cx40–GFP, but not incells expressing the control vector. Finally, H/R reduced electrical coupling in WT and gp91phox−/− but notin p47phox−/− cells. Our data indicate that (i) the H/R-induced ROS increase in microvascular endothelialcells requires Cx40, independent of its role in GJIC, and (ii) p47phox rather than NADPH oxidase-derived ROSaffects modulation of intercellular coupling. Together, the results raise an intriguing possibility that H/R-induced signaling in endothelial cells involves a cross-talk between Cx40 and NADPH oxidase.

titute, Critical Illness Research,

ll rights reserved.

© 2010 Elsevier Inc. All rights reserved.

Connexins (Cx) have been studied as the structural building blocksof intercellular channels responsible for gap-junctional intercellularcommunication (GJIC). In total, 21 members of the connexin familyhave been cloned [1]. In one cell, six connexins oligomerize into aconnexon [2,3]. Connexins can intermix to form a wide array ofchannels [1], and clusters of channels (i.e., gap junctions) permitelectrical and metabolic intercellular communication. GJIC is essentialat all stages of development and in the homeostasis of fullydifferentiated tissues and organs [3,4]. Recent evidence suggests animportant role for hemichannels, which allow the exchange ofmaterials between the cytoplasm and the extracellular environment[5,6].

GJIC between cells of the vascular wall accounts for the arteriolarconducted response. Here, electrical coupling between these cellsmediates the quick spread of locally induced dilation/constrictionalong a 2- to 3-mm arteriolar length [7,8]. Among connexins found in

the vascular wall (i.e., Cx37, Cx40, Cx43, Cx45) [9], Cx40 has beenreported to play the central role in the conducted response [10,11].We reported that Cx40 critically participates in the vascularpathophysiology during inflammation, because reduced electricalcoupling between microvascular endothelial cells caused by lipopoly-saccharide or hypoxia/reoxygenation (H/R) was prevented by Cx40knockout [12,13].

Although connexins have been mainly studied for their role inGJIC, emerging evidence indicates that their function may extendbeyond facilitating the exchange of metabolites and electrical signalsbetween cells [14–16]. For example, Giepmans and coworkers [14]showed that Cx43 could serve as a microtubule-anchoring proteinwithin the cell, and McLachlan and coworkers [15] suggested thatCx26 and Cx43 inhibit malignant properties of breast cancer cells via aGJIC-independentmechanism. To our knowledge, there are no reportsof GJIC-independent function of Cx40 under normal or pathophysi-ological conditions.

NADPH oxidase is the major source of reactive oxygen species(ROS) that participate in the development of cardiovascular disease[17]. In endothelial cells, NADPH oxidase comprises the catalyticsubunit Nox2 (i.e., gp91phox), its homologues Nox1 and Nox4, and

1009G. Yu et al. / Free Radical Biology & Medicine 49 (2010) 1008–1013

the regulatory subunits p22phox, p40phox, p47phox, and p67phox[17]. We reported that, in microvascular endothelial cells, NADPHoxidase is entirely responsible for H/R-induced ROS increase, becausep47phox knockout or gp91phox knockout eliminated this increase[18]. Because a participatory role of Cx40 in cardiovascular disease hasbeen recognized [19], it is possible that NADPH oxidase and Cx40functionally interact.

Given the emerging role of connexins as GJIC-independentsignaling partners [15,16], the main objective of this study was totest the hypothesis that the H/R-initiated ROS production increase inmicrovascular endothelial cells requires Cx40. Further, we also askedif the reported Cx40-dependent reduction in interendothelial electri-cal coupling after H/R [12] requires NADPH oxidase.

Methods

Reagents

The gap junction blocker 18α-glycyrrhetinic acid (AGA), theNADPH oxidase inhibitor apocynin, NADPH, lucigenin, nitrobluetetrazolium (NBT), protease inhibitor cocktail (Cat. No. P2714),Griffonia simplicifolia I (GS-I) lectin, dispase, trypsin, bovine serumalbumin (BSA), collagenase II, dimethyl sulfoxide (DMSO), Hepes,DMEM/F12, DMEM, fetal bovine serum (FBS), dialyzed FBS, andDulbecco's phosphate-buffered saline (PBS) were all purchased fromSigma Chemical (St. Louis, MO, USA). Mn(III) tetrakis(4-benzoic acid)porphyrin chloride (MnTBAP) was obtained from Calbiochem (LaJolla, CA, USA). Endothelial growth supplement was from BDBiosciences (Mississauga, ON, Canada). Magnetic beads and magneticparticle concentrator were purchased from Dynal (Lake Success, NY,USA). Heparin was from Leo Laboratories (Ajax, ON, Canada).L-Glutamine, antibiotic–antimycotic solution, and Trizol reagentwere obtained from Gibco (Mississauga, ON, Canada). Bovine calfserum was purchased from Hyclone (Logan, UT, USA).

Animals, cell isolation, and culture

This investigation conformed to the Guide for the Care and Use ofLaboratory Animals published by the U.S. National Institutes of Health(NIH Publication No. 85-23, revised 1996). All protocols were approvedby the Animal Use Subcommittee at the University of Western Ontario.Weusedmale C57BL/6wild-type (WT), p47phox−/−, and gp91phox−/−

mice (The Jackson Laboratory, Bar Harbor, ME, USA) at 18–25 g of bodyweight. Several Cx40−/− mice (C57BL/6 background) were kindlyprovided by Dr. David Paul (Harvard University, Boston, MA, USA) [20].Additional male Cx40−/− mice were produced by crossing Cx40−/−

males with WT C57BL/6 females and by breeding the heterozygousoffspring. Using a standard genotyping procedure, we confirmed theabsence of Cx40 mRNA in Cx40−/− mice. Mouse microvascularendothelial cells (MMEC) were harvested from the hind-limb skeletalmuscle, as detailed by us [21]. Briefly, the muscle was excised, minced,digested in an enzyme solution, and filtered through a nylon mesh.Collected cells were washed in DMEM/F12, grown to confluence, andthen purified by immunoseparation using GS-I lectin-coated magneticbeads. Cells were cultured in a maintenance medium containingDMEM/F12, FBS (10%), endothelial growth supplement (100 μg/ml),heparin (5 U/ml), L-glutamine (0.2 mM) (BD Biosciences), andantibiotic–antimycotic solution (10 μg/ml) (Gibco) in a standardnormoxic incubator (5% CO2, 19% O2, and 76% N2 at 37 °C) and wereused between passages 4 and 15.MMEC here express all endothelial cellconnexins (Cx37, Cx40, Cx43) [12]. Endothelial phenotype wasdetermined by the presence of von Willebrand factor and GS-I lectinantigens as detailed by us [22]. Sixty minutes before experiments, themaintenancemediumwas replaced by a dialyzed serummedium(DSM;5% dialyzed serum in DMEM/F12, without endothelial growth supple-ment and glutamine).

SKHep1 cells (human hepatoma of endothelial origin) [23] werepurchased from the American Type Culture Collection (Rockville, MD,USA) and cultured in DMEM supplemented with bovine calf serum(10%) and antibiotics. Stable expression of green fluorescent protein(GFP)-tagged human Cx40, or GFP only, in SKHep1 cells was achievedusing a retroviral approach following procedures we described indetail elsewhere [24–26]. Using the Zeiss LSM510 META confocalmicroscope, the expression efficiency of Cx40–GFP and GFP wasdetermined to be ∼90%. Sixty minutes before the H/R experiments,the culturemediumwas replaced by a serum-reducedmedium (SRM)containing DMEM and 5% bovine calf serum. Doses of pharmacologicalagents used in the present experiments were based on the literatureand/or preliminary experiments carried out in our laboratory.

Exposure of cells to hypoxia and reoxygenation

As detailed by us [18], cells were exposed to hypoxia by coveringthemwith hypoxicmedium (DSMor SRM) and then placing them intoa hypoxic incubator for 1 h (5% CO2, 0.1% O2, and 94.9% N2 at 37 °C).Cells were reoxygenated abruptly by replacing the hypoxic mediumwith normoxic medium and by placing them into the normoxicincubator for 10–40 min. Control cells were kept in the normoxicincubator for 1 h, after which the normoxic medium was replaced byfresh normoxic medium.

Measurement of ROS production

The nitroblue tetrazolium assay was used to measure intracellularROS, as described by us [18]. In most cases, cells were seeded at equaldensity on 24-well plates to reach confluence. Cell culture mediumwas replaced with the medium containing 1 mg/ml NBT for 30 minbefore the onset of hypoxia. Some experiments were performed in thepresence of 18α-glycyrrhetinic acid or apocynin applied for 1 h beforeand during H/R. After the experiment, the cells were rinsed with PBSand then with 2 MNaOH, and the resultant formazan was dissolved in1 ml of DMSO. Optical density wasmeasured at 654 nmwith a DU 640spectrophotometer (Beckman, Fullerton, CA, USA). To check that allwells per plate had the same cell confluence/total protein, wemeasured the protein amount in some of the wells selected at random(we found no more than 5% difference in protein/well).

Measurement of NADPH oxidase activity

Cells were cultured in 6-well plates to confluence, washed twicewith ice-cold PBS containing 0.1% BSA, and then collected by scrapingin 0.6 ml DMEM/Hepes buffer containing 100× diluted proteaseinhibitor cocktail. The samples were homogenized by sonication for5 s and incubated on ice for 20 min to prepare them for analysis ofNADPH oxidase activity. The activity was measured using a modifiedlucigenin chemiluminescence assay [27]. Briefly, 10 μl of NADPH(100 μM, final concentration) was added to 100 μl of sample on a 96-well plate (white wells in black matrix). After the mixture was keptfor 10 min at 37 °C, 5 μM lucigenin (final concentration) was added,and NADPH-dependent superoxide generation was immediatelymeasured every 5 s for 12 min using a multilabel PerkinElmer WallacVictor 3 counter. We used the average counts per second (cps) per100 μg of protein as a measure of NADPH oxidase activity.

Electrophysiology

To assess electrical coupling, we determined intercellular electricalresistance between cells of the monolayer (i.e., inverse of coupling) asdescribed by us [12]. Briefly, cells were injected with four or fivehyperpolarizing pulses (25 nA, 100-ms pulse width) and the resultingdeflection from the resting membrane potential in cells was recordedat various distances along the monolayer. The intercellular resistance

Fig. 2. Expression of Cx40–GFP in GJIC-deficient human hepatoma cells SKHep1 enablesH/R-initiated ROS increase. (A) SKHep1 cells expressing GFP or Cx40–GFP were grownto confluency and exposed to H (1 h)/R (40 min). Using the NBT assay, only Cx40–GFP-expressing cells exhibited the H/R-initiated ROS increase. *Pb0.05, compared to therespective control, n=10–15 per group. (B) SKHep1 cells expressing GFP or Cx40–GFPwere grown to confluency and exposed to H/R. Using the lucigenin chemiluminescenceassay, H/R increased the NADPH-dependent superoxide production only in Cx40–GFP-expressing cells. *Pb0.05, compared to the respective control, n=8 or 9 per group.

1010 G. Yu et al. / Free Radical Biology & Medicine 49 (2010) 1008–1013

was then computed, based on these recordings and a Bessel functionmodel of the current spread within the monolayer [21].

Statistics

Data are presented as means± SE; n is the number of independentexperiments per treatment group, based on MMEC isolated from atleast three mice. We used Student's t test, at significance Pb0.05,unless otherwise stated.

Results

H/R-initiated ROS increase is Cx40-dependent

To test whether ROS increase is Cx40-dependent, we used twoapproaches. First, based on NBT and lucigenin chemiluminescenceassays, we determined ROS production in confluent MMEC mono-layers from WT and Cx40−/− mice. In WT cells, H/R caused a ROSproduction increase (Fig. 1), which was inhibited by the superoxidedismutase mimetic MnTBAP (data not shown). Significantly, the ROSincrease was absent from cells from Cx40−/− mice (Fig. 1). Cx40deletion attenuated the baseline superoxide production, as assessedby the lucigenin assay (Fig. 1B).

Second, we determined ROS production in confluent monolayersof the GJIC-deficient human hepatoma cell line SKHep1 [23,28], inwhich cells were engineered to overexpress Cx40–GFP. Here, H/Rincreased ROS production in SKHep1 cells stably expressing Cx40–GFP, but not in SKHep1 cells expressing only GFP (Fig. 2). The H/R-initiated ROS increase in these cells was probably due to NADPHoxidase, because the putative NADPH oxidase inhibitor apocynin(1 mM, applied 1 h before and during H/R) eliminated this increase

Fig. 1. Increase in ROS production in microvascular endothelial cells (MMEC; mousehind-limb skeletal muscle origin) after hypoxia and reoxygenation (H/R) is Cx40-dependent. (A) Confluent MMEC cultures from wild-type and Cx40−/− mice wereexposed to H (1 h, 0.1% O2)/R (40 min, hypoxic medium replaced by normoxicmedium). ROS were determined by the NBT assay (details described under Methods).Genetic deletion of Cx40 prevented the H/R-initiated ROS increase. *Pb0.05, comparedto the respective control, n=6–12 per group. (B) Confluent MMEC cultures from wild-type and Cx40−/− mice were exposed to H/R. NADPH-dependent superoxideproduction was determined by the lucigenin chemiluminescence assay. Geneticdeletion of Cx40 prevented the H/R-initiated superoxide production and alsoattenuated the baseline superoxide production. *Pb0.05 compared to the control WTgroup, n=5–8 per group.

(control, 100±2%, n=13; H/R+vehicle, 163±5%, n=18; H/R+apocynin, 99±2%, n=5, Pb0.05 between H/R+vehicle and H/R+apocynin groups). Together, the two independent approachesindicated the Cx40-dependence of H/R-initiated ROS increase.

H/R-initiated ROS increase is independent of GJIC

To test whether the ROS increase in MMEC is GJIC-independent,we also used two approaches. First, ROS were measured in cellsseeded at low density, such that the cell-to-cell contact and thereforegap junctions were minimal. Equal numbers of cells were concur-rently seeded in 24-well plates and in 6-cm-diameter dishes (i.e., thearea of each dishwas 16× larger than that of eachwell). After 24 h, thecells in the dishes remained dispersedwithin the large area of the dish(cell density was 39±4 cells/mm2, n=32microscopic fields of view).Contacts between cells were rare, occurring mainly at cell extensions(Fig. 3A). Dispersed cells demonstrated their endothelial phenotype,staining positive for von Willebrand factor (data not shown). After24 h, cells seeded in wells reached semiconfluency with numerouscell-to-cell contacts (Fig. 3B; density 414±64 cells/mm2, n=6).When the cells in the wells were grown for another 24 h, they formedconfluent monolayers (Fig. 3C; density 1290±38 cells/mm2, n=6).Importantly, dispersed, semiconfluent, and confluent cells showedcomparable increases in ROS production after H (1 h)/R (30–40 min)(Fig. 3D), indicating that this increase was independent of celldensity/cell-to-cell contact. ROS increase was also absent in Cx40−/−

dispersed cells (control, 100±2%; H/R, 104±6%, n=9, PN0.05).The second approach involved measurement of H/R-initiated ROS

in confluent MMEC monolayers treated with the gap junction blockerAGA. Although this agent may have some nonspecific effects, it is areliable inhibitor of GJIC [29–31]. We found that the ROS increase wasnot affected by AGA (20 μM) (Fig. 4A). In a control experiment usingelectrophysiology, we confirmed that AGA blocked electrical coupling

Fig. 3.H/R-initiated ROS increase in microvascular endothelial cells is independent of cell density and cell-to-cell contact. (A–C_MMECwere seeded in 24-well plates or dispersed in6-cm-diameter dishes. After 24 h, cells in the dishes remained dispersed (A), whereas cells in the wells reached semiconfluency (B). When cells in the wells were grown for another24 h, they formed confluent monolayers (C). Bar, 60 μm. (D) Dispersed, semiconfluent, and confluent cells were subjected to H (1 h)/R (30–40 min) and ROS were measured by NBTassay. ROS measurements were normalized to the appropriate control. H/R increased ROS in dispersed, semiconfluent, and confluent cells. *Pb0.05, compared to the respectivecontrol, n=5–14 per group.

1011G. Yu et al. / Free Radical Biology & Medicine 49 (2010) 1008–1013

along the monolayer (Fig. 4B) (i.e., the electrical current did notspread to distances beyond 150 μm). Thus, each of the two approachesindicated that the H/R-initiated ROS increase was GJIC-independent.

H/R-initiated reduction in electrical coupling is NADPHoxidase-dependent

Data in Figs. 1 and 2 indicate that Cx40 is required for the H/R-initiated ROS increase. Because we showed that this increase dependson NADPH oxidase [18], we asked if the reverse is also possible, suchthat Cx40-dependent intercellular coupling in MMEC [12] requiresNADPH oxidase. To this end, we determined electrical coupling beforeand after H/R in confluent MMEC monolayers fromWT, p47phox−/−,and gp91phox−/− mice. A typical increase in intercellular electricalresistance (i.e., reduced coupling) in theWTmonolayer subjected to H(1 h)/R (10–30 min) is shown in Fig. 5 (first two bars). Geneticdeletion of p47phox prevented this increase but deletion of gp91phoxdid not, indicating that H/R-initiated modulation of intercellularelectrical coupling depends on p47phox rather than on NADPHoxidase-derived ROS production, which requires all NADPH oxidasesubunits [17].

Discussion

We used mouse microvascular endothelial cells and SKHep1 cellsto determine oxidant production after hypoxia/reoxygenation. Usingcell culture conditions identical to those in the present study, wepreviously showed that NADPH oxidase in MMEC is the sole source ofH/R-initiated ROS increase, as this increase is completely eliminatedby genetic deletion of gp91phox or p47phox; the NOS inhibitor L-NAME or XO inhibitor allopurinol does not affect this increase [18].NADPH oxidase may also be an important source of H/R-initiated ROSincrease in SKHep1 cells, because we showed in this study that theputative NADPH oxidase inhibitor apocynin inhibits this increase inSKHep1 cells.

This study addressed the hypothesis that H/R-initiated ROSproduction increase in microvascular endothelial cells requiresCx40. To this end, we used two independent approaches includinggenetic deletion of Cx40 and Cx40 overexpression in GJIC-deficientSKHep1 cells. The deletion of Cx40 prevented the H/R-initiated ROSincrease, whereas overexpression of Cx40 enabled this increase.Although each of these approaches may be associated with nonspe-cific effects, together they strongly indicate that the H/R-initiated ROS

Fig. 4. The H/R-initiated ROS increase is not affected by the gap junction blocker 18α-glycyrrhetinic acid (AGA). (A) Confluent MMEC were treated with or without AGA(20 μM; applied 1 h before and during H/R). *Pb0.05, compared to the respectivecontrol, n=11–14 per group. (B) A representative example (n=3) of a membranepotential recording, showing inhibition of electrical coupling between cells of themonolayer after 20 μM AGA. As detailed elsewhere [13], a current-injection microelec-trode E1 and a recording microelectrode E2 were inserted into two cells of themonolayer. In the vehicle-treated monolayer (0.1% DMSO), the initial deflection ofmembrane potential Em2 from 0 to−15 mV indicates the initial insertion of E2 into thecell. Current injections at increasing interelectrode distances of 50, 150, and 250 μm(i.e., electrode E1 inserted into cells at these distances) caused Em2 deflections of −12,−8, and −5 mV, respectively. The return of Em2 to 0 indicates the withdrawal of E2from the cell at the end of the experiment. For the AGA-treated monolayer, Em2

deflections of −9, 0, and 0 mV were observed at distances of 50, 150, and 250 μm,respectively, indicating inhibition of electrical coupling between cells at distancesbeyond 150 μm. (Note. Inserting electrode E1 into a cell at 75 μm and detecting Em2

deflection of −5 mV confirmed that electrode E2 functioned properly at the precedingdistances of 150 and 250 μm).

1012 G. Yu et al. / Free Radical Biology & Medicine 49 (2010) 1008–1013

increase is Cx40-dependent. Because the ROS increase in MMECoriginates entirely from NADPH oxidase activation [18], the Cx40-dependence suggests that Cx40 presence is required for NADPHoxidase activation by H/R.

NADPH oxidase also contributes, among other sources of ROS [32],to the baseline oxidant production in MMEC, because genetic deletionof p47phox or gp91phox, or treatment with apocynin, attenuatesbaseline ROS here by ∼40% [18,33]. The present data demonstrate a

Fig. 5. Hypoxia (1 h) and reoxygenation (10–30 min) increases electrical resistance inconfluent MMEC from wild-type and gp91phox−/− but not p47phox−/− mice.Intercellular electrical resistance (i.e., inverse of coupling) was measured as describedunder Methods. *Pb0.05, compared to the respective control, n=6–30 per group.

comparable attenuation of baseline ROS level in Cx40−/− MMEC(Fig. 1B). However, because Cx40 knockout did not completelyeliminate baseline ROS production, the role of Cx40 in the functionof NADPH oxidase and in the function of other sources of ROS underthe baseline condition may be complex and difficult to elucidate withthe present data set.

This study also addressed the possibility that the H/R-initiatedROS increase is GJIC-independent. To this end, we used twoindependent approaches in which GJIC was inhibited by preventingcell-to-cell contact or by AGA treatment. Because neither experimen-tal procedure prevented the H/R-initiated ROS increase, together thedata strongly indicate that this increase is GJIC-independent. Thus,under H/R-stimulated conditions, Cx40 may be required directly forNADPH oxidase function (e.g., via its role as a binding protein), ratherthan indirectly via mediating an intercellular exchange of metabo-lites/ions through Cx40-containing gap junctions.

To our knowledge, Cx40 has not been studied in terms of itsrequirement for H/R-initiated ROS production or in terms of its GJIC-independent function. There are several levels at which Cx40 couldaffect ROS production. First, genetic deletion of Cx40 in MMEC couldaffect the expression of other proteins, including NADPH oxidasesubunits. To this end, we showed that the elimination of Cx40 doesnot alter Cx37 and Cx43 expression in MMEC [34]. This finding isconsistent with the observation that Cx37 levels do not change inCx40−/− mouse aortic endothelium [35]. Although the effect of Cx40knockout on the expression of any of the NADPH oxidase subunits isnot known, a downregulation of these subunits in SKHep1 cells wouldbe unlikely to explain the lack of H/R-initiated ROS response, becausethe introduction of Cx40 alone rescued this response (Fig. 2).

A second level at which Cx40 could affect increased ROSproduction is through an interaction between Cx40 and NADPHoxidase. In MMEC, the H/R-initiated ROS increase depends on thepreceding activation of ERK1/2 [18]. This activation has been reportedto lead to phosphorylation of p47phox [36] and to its translocationfrom the cytosol to the plasma membrane to produce ROS [37].Because the H/R-initiated increase in intercellular electrical resistancealso depends on ERK1/2 activation [38], it is possible that signalinginvolving ERK1/2 activation and p47phox translocation to themembrane is common to both the Cx40-dependent ROS increaseand the Cx40-dependent resistance increase. Because genetic deletionof p47phox in MMEC eliminates both the Cx40-dependent ROSincrease [18] and the Cx40-dependent increase in intercellularresistance (Fig. 5), we speculate that Cx40 and p47phox interact atthe membrane when the cell is stimulated with H/R. Clearly, furtherwork is required to go beyond this speculation to determine themechanism of Cx40 involvement in the H/R-initiated ROS increase,including the role of p47phox.

The present results challenge our own conclusion that the H/R-initiated increase in intercellular resistance is oxidant-dependent[12]. This conclusion was based on the fact that a 4-h preloading ofMMEC with the antioxidant ascorbate inhibited this resistanceincrease. The present finding that, under H/R-stimulated conditions,gp91phox deletion did not prevent resistance increase (Fig. 5),whereas gp91phox deletion did prevent ROS increase [18], isinconsistent with the reported inhibitory effect of ascorbate. Toreconcile this apparent inconsistency, we note that ascorbatepreloading, among its many effects, has been shown to downregulatep47phox in MMEC and thus attenuate NADPH oxidase-derived ROSproduction [33]. Therefore, the inhibition by ascorbate could bereconciled in terms of a p47phox-dependent, rather than oxidant-dependent, increase in intercellular resistance after H/R.

Based on immunohistochemical examination of the mouse cre-master muscle in vivo, Looft-Wilson and coworkers [39] showed thatCx40 is abundantly present at the perimeter of endothelial cellswithin the microvasculature. The present results imply that if ROSproduction in the cremaster muscle and other tissues were Cx40-

1013G. Yu et al. / Free Radical Biology & Medicine 49 (2010) 1008–1013

dependent, then H/R could initiate a microcirculatory inflammatoryresponse via ROS production [40] at these specific endothelial sites.Further, Cx40 has been detected in the endothelium of large bloodvessels of high-fat diet-fed mice, in areas of progressive atheroma[41]. Here, Cx40-dependent ROS production could potentially partic-ipate in the development of this pathophysiology [42].

In conclusion, our data point to a novel, GJIC-independent function ofCx40 in that Cx40 is required for increased ROS production inmicrovascular endothelial cells challenged with hypoxia/reoxygenation.Our data also indicate that p47phox affects Cx40-dependent modulationof intercellular coupling by H/R. Together, these observations raise thepossibility that H/R-induced signaling in endothelial cells involves across-talk between Cx40 and NADPH oxidase and call for examination ofthe significance of Cx40 in oxidant production during cardiovasculardisease.

Acknowledgments

We thank Dr. Y. Ouellette, Ms. Q. Shao, and Mr. S. Swarbreck fortechnical help and the Heart and Stroke Foundation of Ontario (GrantNA 5568 to K.T. and D.W.L.) and the Division of Vascular Surgery,London Health Sciences Centre, for financial support (K.T.).

References

[1] Sohl, G.; Willecke, K. Gap junctions and the connexin protein family. Cardiovasc.Res. 62:228–232; 2004.

[2] Laird, D. W. Connexin phosphorylation as a regulatory event linked to gap junctioninternalization and degradation. Biochim. Biophys. Acta 1711:172–182; 2005.

[3] Saez, J. C.; Berthoud, V. M.; Branes, M. C.; Martinez, A. D.; Beyer, E. C. Plasmamembrane channels formed by connexins: Their regulation and functions. Physiol.Rev. 83:1359–1400; 2003.

[4] Dhein, S. Pharmacology of gap junctions in the cardiovascular system. Cardiovasc.Res. 62:287–298; 2004.

[5] Spray, D. C.; Ye, Z. C.; Ransom, B. R. Functional connexin "hemichannels": A criticalappraisal. Glia 54:758–773; 2006.

[6] Wong, C.W.; Christen, T.; Roth, I.; Chadjichristos, C. E.; Derouette, J. P.; Foglia, B. F.;Chanson, M.; Goodenough, D. A.; Kwak, B. R. Connexin37 protects againstatherosclerosis by regulating monocyte adhesion. Nat. Med. 12:950–954; 2006.

[7] Emerson, G. G.; Segal, S. S. Electrical activation of endothelium evokesvasodilation and hyperpolarization along hamster feed arteries. Am. J. Physiol.Heart Circ. Physiol. 280:H160–H167; 2001.

[8] Gustafsson, F.; Holstein-Rathlou, N. Conducted vasomotor responses in arterioles:Characteristics, mechanisms and physiological significance. Acta Physiol. Scand.167:11–21; 1999.

[9] Haefliger, J. A.; Nicod, P.; Meda, P. Contribution of connexins to the function of thevascular wall. Cardiovasc. Res. 62:345–356; 2004.

[10] de Wit, C.; Roos, F.; Bolz, S. S.; Kirchhoff, S.; Kruger, O.; Willecke, K.; Pohl, U.Impaired conduction of vasodilation along arterioles in connexin40-deficientmice. Circ. Res. 86:649–655; 2000.

[11] Figueroa, X. F.; Paul, D. L.; Simon, A. M.; Goodenough, D. A.; Day, K. H.; Damon, D. N.;Duling, B. R. Central role of connexin40 in the propagation of electrically activatedvasodilation in mouse cremasteric arterioles in vivo. Circ. Res. 92:793–800; 2003.

[12] Bolon, M. L.; Ouellette, Y.; Li, F.; Tyml, K. Abrupt reoxygenation following hypoxiareduces electrical coupling between endothelial cells ofwild-type but not connexin40null mice in oxidant- and PKA-dependent manner. FASEB J. 19:1725–1727; 2005.

[13] Bolon,M. L.; Kidder, G.M.; Simon, A.M.; Tyml, K. Lipopolysaccharide reduces electricalcoupling in microvascular endothelial cells by targeting connexin40 in a tyrosine-,ERK1/2-, PKA-, and PKC-dependent manner. J. Cell. Physiol. 211:159–166; 2007.

[14] Giepmans, B. N.; Verlaan, I.; Hengeveld, T.; Janssen, H.; Calafat, J.; Falk, M. M.;Moolenaar, W. H. Gap junction protein connexin-43 interacts directly withmicrotubules. Curr. Biol. 11:1364–1368; 2001.

[15] McLachlan, E.; Shao, Q.; Wang, H. L.; Langlois, S.; Laird, D. W. Connexins act astumor suppressors in three-dimensional mammary cell organoids by regulatingdifferentiation and angiogenesis. Cancer Res. 66:9886–9894; 2006.

[16] Xu, X.; Francis, R.; Wei, C. J.; Linask, K. L.; Lo, C. W. Connexin 43-mediatedmodulation of polarized cell movement and the directional migration of cardiacneural crest cells. Development 133:3629–3639; 2006.

[17] Ray, R.; Shah, A. M. NADPH oxidase and endothelial cell function. Clin. Sci. 109:217–226 (London); 2005.

[18] Yu, G.; Peng, T.; Feng, Q.; Tyml, K. Abrupt reoxygenation of microvascularendothelial cells after hypoxia activates ERK1/2 and JNK1, leading to NADPHoxidase-dependent oxidant production. Microcirculation 14:125–136; 2007.

[19] Chadjichristos, C. E.; Scheckenbach, K. E.; van Veen, T. A.; Richani Sarieddine, M. Z.;de, W. C.; Yang, Z.; Roth, I.; Bacchetta, M.; Viswambharan, H.; Foglia, B.; Dudez, T.;van Kempen, M. J.; Coenjaerts, F. E.; Miquerol, L.; Deutsch, U.; Jongsma, H. J.;Chanson, M.; Kwak, B. R. Endothelial-specific deletion of connexin40 promotesatherosclerosis by increasing CD73-dependent leukocyte adhesion. Circulation121:123–131; 2010.

[20] Simon, A. M.; Goodenough, D. A.; Paul, D. L. Mice lacking connexin40 have cardiacconduction abnormalities characteristic of atrioventricular block and bundlebranch block. Curr. Biol. 8:295–298; 1998.

[21] Lidington, D.; Ouellette, Y.; Tyml, K. Endotoxin increases intercellular resistance inmicrovascular endothelial cells by a tyrosine kinase pathway. J. Cell. Physiol. 185:117–125; 2000.

[22] Wilson, J. X.; Dixon, S. J.; Yu, J.; Nees, S.; Tyml, K. Ascorbate uptake by microvascularendothelial cells of rat skeletal muscle.Microcirculation 3:211–221; 1996.

[23] Heffelfinger, S. C.; Hawkins, H. H.; Barrish, J.; Taylor, L.; Darlington, G. J. SK HEP-1:A human cell line of endothelial origin. In Vitro Cell. Dev. Biol. 28A:136–142; 1992.

[24] Mao, A. J.; Bechberger, J.; Lidington, D.; Galipeau, J.; Laird, D. W.; Naus, C. C.Neuronal differentiation and growth control of neuro-2a cells after retroviral genedelivery of connexin43. J. Biol. Chem. 275:34407–34414; 2000.

[25] Qin, H.; Shao, Q.; Curtis, H.; Galipeau, J.; Belliveau, D. J.; Wang, T.; aoui-Jamali, M. A.;Laird, D. W. Retroviral delivery of connexin genes to human breast tumor cellsinhibits in vivo tumor growth by amechanism that is independent of significant gapjunctional intercellular communication. J. Biol. Chem. 277:29132–29138; 2002.

[26] Thomas, T.; Telford, D.; Laird, D. W. Functional domain mapping and selectivetrans-dominant effects exhibited by Cx26 disease-causing mutations. J. Biol. Chem.279:19157–19168; 2004.

[27] Mohazzab, H.; Kaminski, P. M.; Wolin, M. S. Lactate and PO2 modulate superoxideanion production in bovine cardiac myocytes: Potential role of NADH oxidase.Circulation 96:614–620; 1997.

[28] van Rijen, H. V.; van Veen, T. A.; Hermans, M. M.; Jongsma, H. J. Human connexin40gap junction channels are modulated by cAMP. Cardiovasc. Res. 45:941–951; 2000.

[29] Goldberg, G. S.; Moreno, A. P.; Bechberger, J. F.; Hearn, S. S.; Shivers, R. R.;MacPhee, D. J.; Zhang, Y. C.; Naus, C. C. Evidence that disruption of connexonparticle arrangements in gap junction plaques is associated with inhibition of gapjunctional communication by a glycyrrhetinic acid derivative. Exp. Cell Res. 222:48–53; 1996.

[30] Guo, Y.; Martinez-Williams, C.; Gilbert, K. A.; Rannels, D. E. Inhibition of gapjunction communication in alveolar epithelial cells by 18alpha-glycyrrhetinicacid. Am. J. Physiol. 276:L1018–L1026; 1999.

[31] Hoffmann, A.; Gloe, T.; Pohl, U.; Zahler, S. Nitric oxide enhances de novo formationof endothelial gap junctions. Cardiovasc. Res. 60:421–430; 2003.

[32] Frey, R. S.; Ushio-Fukai, M.; Malik, A. B. NADPH oxidase-dependent signaling inendothelial cells: Role in physiology and pathophysiology. Antioxid. Redox Signal.11:791–810; 2009.

[33] Wu, F.; Schuster, D. P.; Tyml, K.; Wilson, J. X. Ascorbate inhibits NADPH oxidasesubunit p47phox expression in microvascular endothelial cells. Free Radic. Biol.Med. 42:124–131; 2007.

[34] Bolon, M. L.; Peng, T.; Kidder, G. M.; Tyml, K. Lipopolysaccharide plus hypoxia andreoxygenation synergistically reduce electrical coupling between microvascularendothelial cells by dephosphorylating connexin40. J. Cell. Physiol. 217:350–359;2008.

[35] Simon, A. M.; McWhorter, A. R.; Chen, H.; Jackson, C. L.; Ouellette, Y. Decreasedintercellular communication and connexin expression in mouse aortic endotheliumduring lipopolysaccharide-induced inflammation. J. Vasc. Res. 41:323–333; 2004.

[36] Dewas, C.; Fay, M.; Gougerot-Pocidalo, M. A.; El-Benna, J. The mitogen-activatedprotein kinase extracellular signal-regulated kinase 1/2 pathway is involved informyl-methionyl-leucyl-phenylalanine-induced p47phox phosphorylation inhuman neutrophils. J. Immunol. 165:5238–5244; 2000.

[37] Waki,K.; Inanami,O.;Yamamori, T.;Kuwabara,M.Extracellular signal-regulatedkinase1/2 is involved in theactivationofNADPHoxidase inducedbyFMLPreceptorbutnotbycomplement receptor 3 in rat neutrophils. Free Radic. Res. 37:665–671; 2003.

[38] Rose, K.; Ouellette, Y.; Bolon, M.; Tyml, K. Hypoxia/reoxygenation reducesmicrovascular endothelial cell coupling by a tyrosine and MAP kinase dependentpathway. J. Cell. Physiol. 204:131–138; 2005.

[39] Looft-Wilson, R. C.; Payne, G. W.; Segal, S. S. Connexin expression and conductedvasodilation along arteriolar endothelium in mouse skeletal muscle. J. Appl.Physiol. 97:1152–1158; 2004.

[40] Cooper, D.; Stokes, K. Y.; Tailor, A.; Granger, D. N. Oxidative stress promotes bloodcell–endothelial cell interactions in the microcirculation. Cardiovasc. Toxicol. 2:165–180; 2002.

[41] Kwak, B. R.; Mulhaupt, F.; Veillard, N.; Gros, D. B.; Mach, F. Altered pattern ofvascular connexin expression in atherosclerotic plaques. Arterioscler. Thromb.Vasc. Biol. 22:225–230; 2002.

[42] Jo, H.; Song, H.; Mowbray, A. Role of NADPH oxidases in disturbed flow. Antioxid.Redox Signal. 8:1609–1619; 2006.