Opening of connexin 43 hemichannels is increasedby lowering intracellular redox potentialMauricio A. Retamal*†, Kurt A. Schalper*†, Kenji F. Shoji*†, Michael V. L. Bennett‡§, and Juan C. Saez*†§

*Nucleo Milenio Inmunologıa e Inmunoterapia, †Departamento de Ciencias Fisiologicas, Pontificia Universidad Catolica de Chile, Santiago 4860, Chile;and ‡Department of Neuroscience, Albert Einstein College of Medicine, Bronx, NY 10461

Contributed by Michael V. L. Bennett, March 24, 2007 (sent for review February 26, 2007)

Nonjunctional membrane in many cells contains connexin gapjunction hemichannels (or connexons) that can open to allowpermeation of small molecules. Opening of Cx43 hemichannels isinfrequent in normal extracellular Ca2� and enhanced by low Ca2�,positive membrane potentials, and dephosphorylation of criticalresidues. Here we report that lowering intracellular redox poten-tial increases Cx43 hemichannel open probability under otherwisenormal conditions. We studied dye uptake and single-channelactivity in HeLa cells transfected with wild-type Cx43, Cx43 withenhanced GFP attached to its C terminus (Cx43-EGFP), and Cx43with enhanced GFP attached to its N terminus (EGFP-Cx43). Dithio-threitol [(DTT) 10 mM], a membrane permeant-reducing agent,increased the rate of dye uptake by cells expressing Cx43 andCx43-EGFP, but not by parental cells or cells expressing EGFP-Cx43.Induced dye uptake was blocked by La3�, by a peptide gap junctionand hemichannel blocker (gap 26), and by flufenamic acid. DTTincreased Cx43-EGFP hemichannel opening at positive voltages.Bath application of reduced glutathione, a membrane impermeant-reducing agent, did not increase dye uptake, but glutathione in therecording pipette increased hemichannel opening at positive volt-ages, suggesting that it acted intracellularly. DTT caused littlechange in levels of surface Cx43 or Cx43-EGFP, or in intracellular pH.These findings suggest that lowering intracellular redox potentialincreases the opening of Cx43 and Cx43-EGFP hemichannels,possibly by action on cytoplasmic cysteine residues in the connexinC terminus.

cysteine � dithiothreitol � permeation � connexon

Gap junction channels are formed by the union of twohemichannels or connexons, one from each of the apposed

cells, and connecting the cytoplasm of adjacent cells, allowing aflux of small ions and molecules such as ATP, glucose, gluta-thione (GSH), cAMP, and IP3 (1). Thus, gap junction channelsallow electrical coupling, metabolic cooperation, and coordina-tion of cell activities. Although some connexins have limitedexpression, Cx43 is expressed in many cell types, includingastrocytes, fibroblasts, cardiomyocytes, and dendritic cells (2).Hemichannels are assembled in the endoplasmic reticulum andGolgi or post-Golgi compartments and are trafficked to thesurface membrane. Hemichannels have been demonstrated inmany cell types by immunocytochemistry, freeze-fracture EM,electrophysiological recording, dye uptake measurement, and/orbiotinylation of cell surface proteins (3). Cx43 expressed in HeLacells forms hemichannels with a small open probability (4).Hemichannel openings are increased by low extracellular Ca2�

(4, 5), which enhances release of small molecules such as ATP,glutamate, prostaglandin E2, and NAD� (6), which may beparacrine signals. Opening of hemichannels may subserve avariety of functions in physiological and pathological conditions(3, 7), but their role is still controversial.

Metabolic stress induces activation of a large nonselectivecationic channel in cardiomyocytes (8) and increases permeabil-ity to small molecules such as ethidium bromide (EtdBr) andLucifer yellow in astrocytes (9, 10). These actions are ascribableto the opening of Cx43 hemichannels (8–10). In astrocytes, the

increase in dye uptake can be inhibited with nordihydroguai-aretic acid, a blocker of lipoxygenases, trolox, melatonin (9, 11),reduced GSH ethyl ester, or dithiothreitol (DTT) (10). Theseagents are membrane permeant; they reduce the generation offree radicals, act as free radical scavengers, or reduce oxidizedcysteines. The action of reducing agents suggests a role for redoxpotential in controlling hemichannel opening in cells undermetabolic stress. Redox potential is known to modify the activityof different types of ion channels, including ryanodine receptors(12), P2X receptors (13), transient receptor potential (TRP)channels (14), potassium channels (15), and nonselective cationchannels (16). At physiological cytosolic oxygen pressure, nitricoxide (NO) activates ryanodine receptors through S-nitrosylation of free cysteines of the channel subunit, but atatmospheric oxygen pressure, cysteine S-nitrosylation does notoccur because cysteine residues are oxidized (12). In general, invitro studies on functional gap junction channels and hemichan-nels have been performed at atmospheric oxygen pressure, anoxidizing condition that can increase the intracellular redoxpotential. Thus, regulation of connexin-based channels bychange in redox potential of cells in vitro may have beenoverlooked.

In the present work, we used HeLa cells transfected withmouse Cx43, Cx43 with enhanced GFP attached to the Cterminus (Cx43-EGFP), or Cx43 with enhanced GFP attached tothe N terminus (EGFP-Cx43) to evaluate the effect of reducingagents on dye uptake and hemichannel opening. We found thatDTT increases the rate of dye uptake at the resting redoxpotential in cells expressing Cx43 and Cx43-EGFP, but not inparental cells or cells expressing EGFP-Cx43. In whole-cellvoltage clamp studies, DTT increased the open probability ofCx43-EGFP hemichannels at positive potentials. Additionally,GSH, a membrane impermeant, physiological-reducing agent,did not increase the rate of dye uptake when externally applied,but when included in the recording pipette increased openingsat positive potentials, as did bath-applied DTT. DTT had littleor no effect on the level or phosphorylation state of Cx43 in thesurface membrane or on intracellular pH. These data indicatethat the opening of Cx43 hemichannels is regulated by theintracellular redox potential, which may act through the cysteineresidues in the C terminus of Cx43 or in an associated molecule.

ResultsThe Reducing Agent DTT Increases EtdBr Uptake in HeLa Cells Express-ing Cx43-EGFP. Because increased Cx43 hemichannel openingduring metabolic inhibition is reversed by reducing agents (9,

Author contributions: M.A.R., K.A.S., K.F.S., M.V.L.B., and J.C.S. designed research; M.A.R.,K.A.S., K.F.S., and J.C.S. performed research; M.A.R., K.A.S., K.F.S., M.V.L.B., and J.C.S.analyzed data; and M.A.R., K.A.S., K.F.S., M.V.L.B., and J.C.S. wrote the paper.

The authors declare no conflict of interest.

Abbreviations: DTT, dithiothreitol; Cx43-EGFP, Cx43 with enhanced GFP attached to itsC terminus; EGFP-Cx43, Cx43 with enhanced GFP attached to its N terminus; EtdBr, ethidiumbromide; GSH, glutathione; TRP, transient receptor potential.

§To whom correspondence may be addressed. E-mail: jsaez@bio.puc.cl or mbennett@aecom.yu.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0702456104/DC1.

© 2007 by The National Academy of Sciences of the USA

8322–8327 � PNAS � May 15, 2007 � vol. 104 � no. 20 www.pnas.org�cgi�doi�10.1073�pnas.0702456104

10), we studied whether changes in redox potential affecthemichannel opening under normoxic conditions. We usedHeLa cells transfected with Cx43-EGFP; these cells take upEtdBr uptake at the resting potential and exhibit unitary eventsat positive potentials, with conductance and pharmacologicalsensitivity predicted for Cx43 hemichannels (4). As reported formixed cultures of parental and transfected cells (4), cells ex-pressing more Cx43-EGFP showed faster EtdBr uptake (Fig. 1).Application of DTT (10 mM) rapidly increased the rate of dyeuptake in cells showing Cx43-EGFP fluorescence (�3 min,records 1–3; Fig. 1B, bar at top; see also Fig. 2), but there waslittle increase in nonfluorescent cells (records 4–8). In Cx43-EGFP cells, the rate of EtdBr uptake after DTT application wasincreased from 0.29 � 0.01 to 0.48 � 0.02 a.u. per min (P � 0.001;�20 cells expressing high and moderate levels of Cx43-EGFPfrom 13 independent experiments). The rate of dye uptake waslinearly related to the amount of Cx43-EGFP expressed (mea-sured as fluorescence intensity) and was increased proportion-ally by DTT at each level of expression; the ratio of the slopeswas �3 (Fig. 1C; correlation coefficients for control and afterDTT were r2 � 0.84 and 0.92, respectively; P � 0.001 for slopesbefore vs. after DTT). The increase in dye uptake was unex-pected because DTT decreases dye uptake after dye uptake hasbeen increased by metabolic inhibition (10).

DTT at �1 mM had no effect on dye uptake [supportinginformation (SI) Fig. 7 A and D]; DTT at 10 and 30 mMincreased the rate of dye uptake near maximally (by 67 � 5% and82 � 6%; SI Fig. 7 B–D; n � 4, �20 cells per experiment; ***,P � 0.001 vs. control; P � 0.5 for 10 vs. 30 mM).

DTT-Induced Increase in EtdBr Uptake Is Sensitive to HemichannelBlockers. If DTT increases uptake of EtdBr by opening Cx43-EGFP hemichannels, this effect should be inhibited by

hemichannel blockers such as La3� (4, 9, 17) and the gap 26peptide, which has the same sequence as a region of the firstextracellular loop of Cx43 (6). We measured basal uptake for 10min and added DTT, which increased uptake. After �10 min, weadded La�3 (200 �M; Fig. 2A, upper bar) or gap 26 (300 �M;Fig. 2B, upper bar), which rapidly reduced uptake to below thatbefore DTT treatment, in La3� to 65 � 7% of the initial rate(P � 0.001; n � 9 experiments, 145 cells analyzed), and in gap26 to 80 � 5% of the initial rate (P � 0.01; n � 3, 60 cellsanalyzed). Thus, the DTT-induced increase in uptake and�20–30% of basal uptake were mediated by hemichannels. Asshown in Fig. 1C, uptake depended on expression and variedamong cells. Action of La3� and gap 26 did not differ signifi-cantly (P � 0.05). Flufenamic acid (100 �M) also blocked dyeuptake (data not shown).

To exclude uptake through a P2X receptor-dependent path-way or TRPV1 channels, we used specific blockers. oATP (300�M), an irreversible P2X receptor blocker, applied 40 minbefore DTT (Fig. 2C; n � 3, 60 cells) or 10 min after DTT (datanot shown) had no obvious effect on DTT-induced or basaluptake (data not shown). Capsazepine [(CZP) 10 �M], a TRPV1channel blocker (18), did not affect DTT-induced dye uptake(Fig. 2D; n � 3, 60 cells).

Fig. 1. DTT increases the rate of EtdBr uptake by Cx43-EGFP but not byparental cells; uptake is proportional to Cx43-EGFP expression. In mixedcultures of Cx43-EGFP and parental cells EtdBr (5 �M), uptake was measuredevery 100 sec as fluorescence emission of EtdBr binding to DNA (518 nm, AUof intensity). (A) Fluorescence micrograph of the measured cells showing EGFPexpression. (B) Time course of EtdBr uptake in Cx43-EGFP cells (1–3 in A) andparental HeLa cells (4–8 in A) before and after application of 10 mM DTT (barat top). (C) Rate of EtdBr uptake was proportional to Cx43-EGFP fluorescencefor eight cells before (open circles) and after (filled circles) DTT treatment.

15

Time (min)

Dye

upt

ake

(AU

)

La3+

DTT

10

5

00 10 20 30

15

Time (min)

Dye

upt

ake

(AU

)

oATP

DTT

10

5

00 10 20 30

15

Time (min)D

ye u

ptak

e (A

U)

EGFP-Cx43Parental

DTT

10

5

00 5 10 15 20 25

15

Time (min)

Dye

upt

ake

(AU

)

gap 26

DTT

10

5

00 10 20 30

15

Time (min)

Dye

upt

ake

(AU

)

CZPDTT

10

5

00 10 20 30

6

Time (min)

Dye

upt

ake

(AU

)

Cx43

DTT

4

2

00 5 10 15 20

A B

C D

E F

Fig. 2. DTT-induced EtdBr uptake by Cx43-EGFP cells is prevented byhemichannel blockers, but not by blockers of P2X or TRPV1 channels. EGFP-Cx43 cells take up EtdBr at a rate like that of parental cells, but wild-type Cx43does mediate uptake. EtdBr uptake was measured as for Fig. 1 in controlconditions, after addition of 10 mM DTT (bar), and in A–D after later additionof a blocker (second bar). (A) In Cx43-EGFP cells, an increase in the rate of dyeuptake caused by DTT was reversed by 200 �M La3�. (B) Similarly, 300-�M gap26 reversed a DTT-induced increase in uptake. (C) Uptake induced by DTT wasnot affected by oATP, a P2X receptor blocker (300 �M was applied 30 minbefore uptake measurement and had no significant effect on basal uptake).(D) Uptake induced by DTT was not affected by capsazepine (CZP), a specificTRPV1 channel blocker (10 �M). (E) DTT had little effect on EtdBr uptake byEGFP-Cx43 cells (filled circles) or parental cells (open circles largely obscured byfilled circles). (F) DTT increased uptake by cells expressing wild-type Cx43.

Retamal et al. PNAS � May 15, 2007 � vol. 104 � no. 20 � 8323

CELL

BIO

LOG

Y

To examine the possibility that Cx43-EGFP induces expres-sion of other channels or transporters capable of mediatingEtdBr uptake in HeLa cells, we used HeLa cells expressingEGFP-Cx43, which does not form functional hemichannels orchannels, although it has an unmodified C terminus and doesform plaques between cells (4). Basal uptake by EGFP-Cx43cells (Fig. 2E, filled circles) did not differ from that of parentalcells (open circles), and application of 10 mM DTT (bar) did notenhance EtdBr uptake by either type of cell (Fig. 2E; P � 0.05;four comparisons: transfected vs. parental cells, uptake afterDTT vs. basal uptake; n � 3 experiments, 7 EGFP-Cx43 cells, 15parental cells).

DTT Increases EtdBr Uptake by HeLa Cells Expressing Cx43. Becausethe permeation through Cx43-EGFP hemichannels might beaffected by the EGFP, we tested whether DTT affects EtdBruptake in HeLa cells transfected with wild-type Cx43. Immu-nolabeling indicates that �100% of these cells express Cx43(data not shown). Bath application of 10 mM DTT rapidlyincreased dye uptake (Fig. 2F; 0.22 � 0.01 a.u. per min beforeDTT, 0.30 � 0.01 a.u. per min after DTT; P � 0.001; n � 13, 251cells analyzed). This increase is comparable to that in Cx43-EGFP cells (Fig. 1) and presumably depends on expression level.

DTT Increases Opening of Cx43-EGFP Hemichannels at Positive Volt-ages. The opening of Cx43 hemichannels is increased at positivevoltages (4), so we determined the effect of DTT on openingunder these conditions. Polarizing a cell to �20 mV under

whole-cell patch clamp did not induce opening of hemichannelswith or without 10 mM DTT (Fig. 3A Left and Right). Voltagesteps to �40 mV induced openings of a few hemichannels in cellsunder control conditions (1.5 � 0.7 hemichannels per 40-secpulse, the average of the maximum number of channels opensimultaneously during the pulse; Fig. 3B Left) and after treat-ment with DTT (2.0 � 0.9 hemichannels; P � 0.05; n � 5 cells;Fig. 3B Right). Polarizing to �60 mV further increased thenumber of openings, which were more frequent in cells treatedwith DTT (Fig. 3C Left and Right). The maximal number ofhemichannels opening simultaneously was 3.0 � 0.6 in controlmedium (3 in Fig. 3C Left) and 9.0 � 0.8 after DTT (7 in Fig.3C Right; n � 5 cells for each condition, four steps at eachvoltage, two before and two after DTT in each cell; P � 0.01).Hemichannel activity after DTT was blocked by 200 �M La3�

(Fig. 3D; n � 3) or 100 �M flufenamic acid (n � 3, data notshown). No channel activity was seen with parental cells at �60mV after application of DTT (Fig. 3E; n � 5 cells).

DTT decreased the latency to the first opening at positivevoltages. Before DTT, the latency was 11.3 � 3.5 sec at �40 mVand 3.3 � 1.6 sec at �60 mV (Fig. 3F; representative records in Fig.3C Left and Right). After DTT application, the latency was 2.2 �0.4 sec at �40 mV and 1.5 � 0.6 sec at �60 mV; both weresignificantly shorter than before DTT (Fig. 3F, **, P � 0.01; n �5). DTT increased the mean number of open hemichannels duringthe last 30 sec of the response to positive voltage; at �40 mV, theincrease was from 0.11 � 0.01 to 0.26 � 0.01 (Fig. 3G, *, P � 0.05;n � 5), and at �60 mV, the increase was from 0.30 � 0.05 to 2.03 �0.22 (Fig. 3G, ***, P � 0.001; n � 5).

The DTT-induced increase in the number of open channels islikely due to an increase in open probability at the given voltagesbecause there was little change in surface expression (see belowand Fig. 5). We do not know the extent to which the openprobability of ‘‘active’’ hemichannels was increased or newhemichannels were recruited to the active population. Theclosure of channels on return to negative potentials appears toorapid to be due to internalization (data not shown).

DTT in the bath solution slightly but significantly changed theunitary conductance of hemichannels from 249 � 3 ps (at �60mV, n � 99 transitions) under control conditions (Fig. 3C Left)to 228 � 3 ps (at �60 mV, n � 141 transitions) after DTT (n �5 independent experiments; P � 0.001, nonparametric Student’st test). The effect of DTT in reducing unitary conductance isunexplained. The difference between �250 ps and the 220 psvalue reported previously (4) is ascribable to minor differencesin the media used (150 mM vs. 140 mM NaCl, 4 mM vs. 5.4 mMKCl, 1.2 mM vs. 1.8 mM CaCl2).

Reducing Agents Act on Cx43-EGFP Hemichannels at an IntracellularSite. To characterize the redox site(s) as intra- or extracellular,we bath applied GSH, a membrane impermeant-reducing agent.Extracellular application of 10 mM GSH at pH 7.4 (Fig. 4A, bar)had little effect on EtdBr uptake by Cx43-EGFP cells (P � 0.05;n � 6, 75 cells analyzed). We then performed whole-cellrecording with a pipette solution containing either DTT or GSH.At �60 mV with 10 mM DTT in the pipette, there was arelatively high frequency of hemichannel openings similar to theeffect of bath-applied DTT (Fig. 4B Left, n � 3; cf. Fig. 3C Right).At �60 mV, the effect of 10 mM GSH in the pipette was similarto that of DTT (Fig. 4B Right, n � 3). The reducing agents actedmore rapidly than we could resolve. From the data in Fig. 2, thereducing agent might act in course of a single 60-sec positive step.

In conclusion, lowering the intracellular redox potential fa-cilitates voltage-induced opening of Cx43-EGFP hemichannels.Lack of effect of bath-applied GSH on dye uptake indicates thatthe action of reducing agents on accessible, extracellular cys-teines is not sufficient to increase hemichannel opening at theresting potential.

Fig. 3. DTT increases opening of Cx43-EGFP hemichannels at positive po-tentials. Unitary events were recorded by whole-cell voltage clamp. (A) At �20mV, there were no unitary events in the absence (Left) or presence (Right) of10 mM DTT in the bathing medium. (B) At �40 mV, there were a few openingsbefore DTT (Left) and increased openings after DTT (Right). (C) At �60 mV,there were multiple openings before DTT (Left) and markedly increasedopenings after DTT (Right). (Left) Up to three hemichannels were opensimultaneously. (Right) At least seven channels were open simultaneouslylater in the response. (D) La3� (0.2 mM) blocked unitary activity at �60 mV inthe presence of DTT. (E) No channel activity was seen in parental cells at �60mV in the presence of DTT. (F) DTT decreased the latency of first opening at�40 mV and �60 mV (four pulses for each voltage before and after DTT in eachof five cells (**, P � 0.01 vs. control at each potential). (G) DTT increased themean number of hemichannels open during the last 30 sec of pulses at �40 mVand �60 mV (*, P � 0.05; ***, P � 0.001).

8324 � www.pnas.org�cgi�doi�10.1073�pnas.0702456104 Retamal et al.

DTT Has Little Effect on Distribution of Cx43-EGFP or the Phosphory-lation State of Cx43-EGFP or Cx43. Fluorescence at the cell surface(rectangles in Fig. 5A) was monitored in four cells under controlconditions and after DTT application. A total of nine cells wereanalyzed in three experiments. Cytoplasmic inclusions, presum-ably vesicles, exhibited some movement in control and in DTT,as did gap junction plaques (data not shown). Bath applicationof 10 mM DTT (bar) resulted in no consistent change influorescence of Cx43-EGFP at or near the surface (integratedfluorescence in the rectangles in Fig. 5A is plotted in Fig. 5B).This result suggests that DTT did not affect the amount ofCx43-EGFP in the surface membrane or acidify the cytoplasmbecause Cx43-EGFP fluorescence is pH-sensitive (19).

To confirm that DTT does not change the amount of Cx43 orCx43-EGFP in the surface membrane, cell surface proteins werebiotinylated, isolated with NeutrAvidin beads, and resolved byWestern blotting with a C-terminal antibody (9). Western blotsshowed the typical forms of Cx43 present in rat heart homoge-nate (Fig. 5C, lane 1; the p2-p3 bands are phosphorylated and theNP band is not phosphorylated, or if it is phosphorylated, itsmobility is not affected by dephosphorylation) (20). A homog-enate of Cx43-EGFP cells showed an immunoreactive doublet of�60 and 70 kDa (Fig. 5C, lane 2); these bands probablycorrespond to two phosphorylated forms of Cx43-EGFP.Weaker bands were also present with electrophoretic mobilityclose to that of Cx43 in the rat heart. These bands mightcorrespond to Cx43 released by proteolysis of Cx43-EGFP. Theamount of Cx43-EGFP present on the cell surface was notsignificantly different under control conditions and after DTTtreatment (Fig. 5C; 10 mM for 10 min; to 112 � 2% of control,lanes 3 and 4; P � 0.05; n � 4). Similarly, DTT did notsignificantly increase the surface expression of Cx43 (Fig. 5C; to115 � 7% of control, lanes 5 and 6; P � 0.05; n � 3). Thus, theincrease in the number of hemichannels in the surface mem-brane made little, if any, contribution to the increase inhemichannel opening induced by DTT.

Buffering Intracellular Ca2� Does Not Alter DTT Action. Because anincrease in intracellular-free Ca2� concentration may induce theopening of Cx32 hemichannels (21), we tested whether loadingwith the chelator BAPTA alters DTT’s effects on EtdBr uptake

by Cx43-EGFP cells. Addition of 10 mM DTT to cells loadedwith BAPTA increased the rate of dye uptake from 0.32 �0.02 a.u. per min to 0.55 � 0.04 a.u. per min (see SI Fig. 8 A andB; P � 0.001; n � 3, 60 cells analyzed), similar to the effect incells without BAPTA.

DTT Action to Increase Dye Uptake in Normoxic Conditions GraduallyChanges to a Decrease in Dye Uptake During Metabolic Inhibition. Theaction of DTT on dye uptake is opposite in normoxic conditionsand after prolonged metabolic inhibition (9, 10). We examinedthe transition from normoxic to the inhibited condition and

Fig. 4. Intracellular DTT or GSH increases openings of Cx43-EGFP hemichan-nels at �60 mV. (A) In Cx43-EGFP cells, EtdBr uptake was not affected by bathapplication of 10 mM GSH. (B) (Left) In whole-cell patch clamp recording with10 mM DTT in the pipette solution, four to five hemichannels opened inCx43-EGFP cells at �60 mV. (Right) Ten millimolar GSH in the pipette solutionhad about the same effect as 10 mM DTT; four to five hemichannels openedin Cx43-EGFP cells held at �60 mV.

Fig. 5. Distribution and phosphorylation state of Cx43-EGFP and phosphor-ylation state of Cx43 are not affected by DTT. (A) Fluorescence micrographs ofthree Cx43-EGFP cells before and 60 min after 10 mM DTT (horizontal bar).Large intensely fluorescent gap junctions occupied intercellular appositions.Vesicles and nonjunctional surface also showed fluorescence. DTT had noapparent effect on Cx43-EGFP distribution. (B) Fluorescence intensity mea-sured in the four boxes at cell margins in A showed no meaningful changesover time. (C) In Western blots from rat heart (lane 1), three phosphorylatedbands (p2-p3 and a more slowly migrating band) and one more rapidlymigrating, unphosphorylated band (np) were observed. In total homogenatefrom Cx43-EGFP HeLa cultures (50 �g of protein, lane 2), two bands of �65 and70 kDa were detected, corresponding to Cx43-EGFP in different states ofphosphorylation. Surface Cx43-EGFP isolated by biotinylation in control con-ditions (lane 3) and after 10-min treatment with 10 mM DTT (lane 4) showedthe same two bands, although fainter, which differed little in density betweenlanes. Surface Cx43 from Cx43 HeLa cultures under control conditions (lane 5)and after 10-min treatment with 10 mM DTT (lane 6) showed similar bands inthe two lanes (p2-p3, phosphorylated; np, nonphosphorylated). Molecularmass markers on the right (BSA, 80 kDa; ovalbumin, 49.1 kDa; carbonicanhydrase, 34.8 kDa).

Retamal et al. PNAS � May 15, 2007 � vol. 104 � no. 20 � 8325

CELL

BIO

LOG

Y

applied DTT at various times after beginning metabolic inhibi-tion [induced by iodoacetate (270 �M) and antimycin A (5ng/ml)]. After 20 min of metabolic inhibition, DTT still increasedthe rate of dye uptake (to �260% of control; Fig. 6 A and D, ***,P � 0.001; n � 4). After 30 min, DTT caused a much smallerincrease (to �140% of control; Fig. 6 B and D, *, P � 0.05 vs.control; n � 4). After 40 min, DTT decreased the rate of dyeuptake (to 40% of control; Fig. 6 C and D, **, P � 0.01; n � 4).The increase in dye uptake during metabolic inhibition can beaccounted for by insertion of additional hemichannels into thesurface membrane (10). The action of DTT to decrease dyeuptake during metabolic inhibition is rapid enough to suggest aneffect on open probability, rather than a reduction in surfaceexpression. In contrast to the effect of DTT during metabolicinhibition, an increase in dye uptake by DTT in normoxicconditions is accompanied by little change in surface expression(Fig. 5C) and therefore is likely to be due to increased openprobability.

DiscussionHere we demonstrated that activity of Cx43-EGFP hemichannels inHeLa transfectants is sensitive to reduction in intracellular redoxpotential. We found that: (i) DTT increased dye uptake in Cx43 orCx43-EGFP cells, but not in parental cells or cells expressingEGFP-Cx43, which does not form functional hemichannels; (ii) anincrease in dye uptake was inhibited by hemichannel blockers; (iii)DTT increased hemichannel opening at positive voltages in Cx43-EGFP cells, but not in parental cells; (iv) GSH, a poorly membranepermeant, physiological-reducing molecule, enhanced hemichan-nel activity only when it was applied intracellularly; and (v) the DTTeffect was not due to changes in distribution, abundance, orphosphorylation state (as assessed by electrophoretic mobility) ofhemichannels or changes in intracellular pH (6). DTT presumablyincreased open probability of active channels and/or fraction ofsurface hemichannels that were active because it had at most a smalleffect on single-hemichannel conductance and surface expression.

Cx43 has nine cysteine residues, three in each of the twoextracellular loops and three in the cytoplasmic C-terminaldomain. Oxidation/reduction of intracellular cysteines may beinvolved in the regulation of the opening of Cx43-EGFPhemichannels. Hemichannels of other connexins with cysteineresidues in the C terminus may also prove to be sensitive to redox

potential; Cx30 and Cx56 have three and one C-terminal cys-teines, respectively, whereas Cx26 and Cx45 have none. Not allcysteines may be sites for redox changes because of the influenceof neighboring amino acids (22, 23). Cx32 has four cysteineresidues in its C terminus, but Cx32 hemichannels expressed inHeLa cells are rather insensitive to DTT under normoxicconditions (H. Sanchez and J.C.S., unpublished data).

During metabolic inhibition, dye uptake ascribed to Cx43hemichannels is increased, and most or all of the increase can beaccounted for by increased surface expression. However, thisincrease is reversed by scavengers of reactive oxygen species,such as Trolox, melatonin (9, 11), DTT, and GSH ethyl ester(10). Extracellular GSH had no effect on hemichannel openinginduced by metabolic inhibition (11), similar to its lack of actionin normoxic conditions (Fig. 5), suggesting an intracellular siteof action. Taken together, these data suggest that oxidation ofCx43 or a regulatory molecule during metabolic inhibition leadsto increased surface expression and increased permeationthrough open hemichannels. Metabolic inhibition causes S-nitrosylation of Cx43, which may increase channel activity, andnitrosylation is reversed by DTT (10). The decrease in perme-ation by reducing agents during metabolic inhibition is in theopposite direction of the increase in permeation and openinginduced by DTT under normoxic conditions, a difference not yetunderstood. The transition between increasing and decreasingpermeation requires many minutes of metabolic inhibition (Fig.6). Phosphorylation state may be relevant, and metabolic inhi-bition induces dephosphorylation of Cx43 hemichannels, asindicated by increased electrophoretic mobility.¶ However, de-phosphorylated Cx43 hemichannels remain open in a reconsti-tuted system (25) and when expressed in Xenopus laevis oocytes(26). Trolox and DTT do not prevent dephosphorylation of Cx43hemichannels during metabolic inhibition (10, 24). Moreover,cyclosporin A, a blocker of the phosphatase, calpain, reducesdephosphorylation during metabolic inhibition, but does notprevent dye uptake (11). These findings argue against a role ofdephosphorylation in hemichannel opening during metabolicinhibition, although dephosphorylation of a critical subset ofhemichannels cannot be excluded. Further study with mutationof specific cysteine residues and phosphorylation sites shouldresolve these issues.

How might reducing agents reverse the increase in perme-ability of hemichannels caused by oxidative stress during meta-bolic inhibition, yet increase opening under normoxic condi-tions? One possibility is that the same cysteine residues aresubstrates of different redox reactions, including formation andreduction of disulfide bonds, cysteine S-nitrosylation, and/orglutathionation (27). Alternatively, in different phosphorylationstates, the same modifications could cause different conforma-tional changes or the cysteines modified could be different.

Bath-applied DTT increased dye uptake proportionally to theexpression of Cx43-EGFP (Fig. 1C). However, DTT did not leadto an obvious opening of hemichannels determined by whole-cellrecording at potentials less than approximately �40 mV, al-though it increased EtdBr uptake in cells at their restingpotential. The increase in Cx43-EGFP hemichannel activity atVm � 40 mV and the increase in dye uptake at the restingpotential appear too rapid to be due to insertion of newhemichannels into the surface membrane. Moreover, the open-ings at positive potentials are rapidly reversed on returning thepotential to inside negative values (data not shown). Finally, theincrease in surface expression is minimal when evaluated byfluorescence and surface biotinylation (Fig. 5). At the resting

¶Retamal, M. A., Cortes, C. J., Bukauskas, F. F., Bennett, M. V. L., Saez, L., 44th AnnualMeeting for the American Society for Cell Biology, December 4–8, 2004, Washington, DC,presentation 1718.

Fig. 6. The effect of DTT changes during metabolic inhibition. (A) Twentyminutes after addition of iodoacetate (270 �M) and antimycin A (5 ng/ml), DTTincreased EtdBr uptake. (B) After �30 min of metabolic inhibition, DTT hadless effect. (C) After �40 min, DTT decreased EtdBr uptake. (D) Quantitationof four experiments. *, P � 0.05; **, P � 0.01; ***, P � 0.001.

8326 � www.pnas.org�cgi�doi�10.1073�pnas.0702456104 Retamal et al.

potential, the hemichannel openings are very infrequent, but theupper limit of opening can still account for the dye uptake (4).However, change in cytoplasmic constituents associated withwhole-cell recording might also result in reduced hemichannelopening at the resting potential. Recently, pannexin 1 (Px1), amember of the family of proteins forming gap junctions ininvertebrates, has been implicated in the formation of activehemichannels in vertebrate cells (28, 29). Our hemichannels areunlikely to be pannexin-based because of the dependence onCx43 expression, single-channel properties corresponding tothose of Cx43 gap junctions, and sensitivity to octanol, La3�, andextracellular Ca2� to which pannexin hemichannels are relativelyinsensitive (29). In addition and in agreement with a recentreport (30), we did not detect pannexin 1 by Western blotanalysis or immunofluorescence (data not shown).

In other cell types expressing Cx43, the normal intracellularredox potential may allow more hemichannel activity than thatobserved in HeLa cells and permit significant release of smallmolecules such as ATP, NAD�, glutamate, and PGE2 (6).Redox potential-sensing mechanisms are known to be relevantin physiological conditions, and cells can modify their intracel-lular redox potential in response to autocrine or paracrinesignals. For example, EGF (31), PDGF (32), angiotensin II (33),and IL-1� (24) can induce production of reactive oxygen species.Because Cx43 is expressed by many cell types, modulation of itshemichannels by the intracellular redox potential may affectnumerous cellular processes in which paracrine signaling occurs.

Materials and MethodsFor additional methods, see SI Methods.

Cell Cultures. Experiments were performed on HeLa cells(CCL-2; ATCC, Rockville, MD) transfected with cDNAs en-coding mouse Cx43-EGFP or EGFP-Cx43 as described (4).Parental HeLa cells served as controls.

Time-Lapse Fluorescence Imaging. Fluorescence images of cells in150 mM NaCl/4 mM KCl/1.2 mM CaCl2/5 mM Hepes at pH 7.4 plus5 �M EtdBr were recorded every 30–90 sec by using an Olympus(Tokyo, Japan) BX 51W1I upright microscope with water-immersion lenses (528-nm excitation, 598-nm emission).

Electrophysiological Measurements. Cells grown on coverslips (#0)were transferred to a chamber on an inverted Olympus IX-70microscope. The pipette solution contained 130 mM KCl, 10 mMpotassium aspartate, 0.26 mM CaCl2, 2 mM EGTA, 5 mM tetra-ethylammonium-Cl, 1 mM MgCl2, 3 mM MgATP, and 5 mMHepes at pH 7.2. Single-channel conductance was calculated fromsingle-channel currents, and the reversal potential, approximately�6 mV, was determined by extrapolation.

Immunoblots. Proteins were analyzed by immunoblotting as de-scribed (10). Aliquots of cell lysates (50 �g of protein) or totalbiotinylated surface proteins were resuspended in a final con-centration of 1� Laemmli sample buffer, separated on 8%SDS/PAGE, and electrotransferred to nitrocellulose sheets asdescribed. Densitometric analyses of immunoblot signals wereperformed with National Institutes of Health IMAGE software.

Cell Surface Biotinylation. Confluent cell cultures were treatedwith Sulfo-NHS-SS-biotin for 30 min at 4°C, followed by quench-ing with glycine. Cells were lysed, and NeutrAvidin beads wereadded followed by three rounds of centrifugation and washing.The resulting pellet was suspended in Laemmli buffer, whichcontained DTT that released labeled proteins from the beads,and centrifuged at 14,000 � g for 2 min. Cx43 in the supernatantwas determined by immunoblotting.

This work was partially funded by National Institutes of Health GrantNS045287 (to M.V.L.B.) and Fondecyt Grant 1030945 (to J.C.S.).M.A.R. was a postdoctoral fellow of the MIDEPLAN project P04/030-F.

1. Harris AL (2001) Q Rev Biophys 34:325–472.2. Saez JC, Berthoud VM, Branes MC, Martınez AD, Beyer EC (2003) Physiol Rev

83:1359–1400.3. Saez JC, Retamal MA, Basilio D, Bukauskas FF, Bennett MVL (2005) Biochim

Biophys Acta 1711:215–224.4. Contreras JE, Saez JC, Bukauskas FF, Bennett MVL (2003) Proc Natl Acad Sci

USA 100:11388–11393.5. Valiunas V, Weingart R (2000) Pflugers Arch 440:366–379.6. Evans WH, De Vuyst E, Leybaert L (2006) Biochem J 397:1–14.7. Goodenough DA, Paul DL (2003) Nat Rev Mol Cell Biol 4:285–294.8. John SA, Kondo R, Wang SY, Goldhaber JI, Weiss JN (1999) J Biol Chem

274:236–240.9. Contreras JE, Sanchez HA, Eugenın EA, Speidel D, Theis M, Willecke K,

Bukauskas FF, Bennett MVL, Saez JC (2002) Proc Natl Acad Sci USA99:495–500.

10. Retamal MA, Cortes CJ, Reuss L, Bennett MVL, Saez JC (2006) Proc NatlAcad Sci USA 103:4475–4480.

11. Contreras JE, Sanchez HA, Veliz LP, Bukauskas FF, Bennett MVL, Saez JC(2004) Brain Res Rev 47:290–303.

12. Eu JP, Sun J, Xu L, Stamler JS, Meissner G (2000) Cell 102:499–509.13. Mason HS, Bourke S, Kemp PJ (2004) Mol Pharmacol 66:1525–1535.14. Aarts M, Iihara K, Wei WL, Xiong ZG, Arundine M, Cerwinski W, MacDonald

JF, Tymianski M (2003) Cell 115:863–877.15. Soto MA, Gonzalez C, Lissi E, Vergara C, Latorre R (2002) Am J Physiol

282:C461–C471.16. Barros LF, Stutzin A, Calixto A, Catalan M, Castro J, Hetz C, Hermosilla T

(2001) Hepatology 33:114–122.17. Kondo RP, Wang SY, John SA, Weiss JN, Goldhaber JI (2000) J Mol Cell

Cardiol 32:1859–1872.

18. Walker KM, Urban L, Medhurst SJ, Patel S, Panesar M, Fox AJ, McIntyre P(2003) J Pharmacol Exp Ther 304:56–62.

19. Bukauskas FF, Bukauskiene A, Bennett MVL, Verselis VK (2001) Biophys J81:137–152.

20. Cooper CD, Lampe PD (2002) J Biol Chem 277:44962–44968.21. De Vuyst E, Decrock E, Cabooter L, Dubyak GR, Naus CC, Evans WH,

Leybaert L (2006) EMBO J 25:34–44.22. Stamler JS, Toone EJ, Lipton SA, Sucher NJ (1997) Neuron 18:691–696.23. Greco TM, Hodara R, Parastatidis I, Heijnen HF, Dennehy MK, Liebler DC,

Ischiropoulos H (2006) Proc Natl Acad Sci USA 103:7420–7425.24. Hwang YS, Jeong M, Park JS, Kim MH, Lee DB, Shin BA, Mukaida N, Ellis

LM, Kim HR, Ahn BW, Jung YD (2004) Oncogene 23:6603–6611.25. Kim DY, Kam Y, Koo SK, Joe CO (1999) J Biol Chem 274:5581–

5587.26. Bao X, Chen Y, Reuss L, Altenberg GA (2004) J Biol Chem 279:9689–

9692.27. Aracena P, Sanchez G, Donoso P, Hamilton SL, Hidalgo C (2003) J Biol Chem

278:42927–42935.28. Bruzzone R, Hormuzdi SG, Barbe MT, Herb A, Monyer H (2003) Proc Natl

Acad Sci USA 100:13644–13649.29. Pelegrin P, Surprenant A (2006) EMBO J 25:5071–5082.30. Huang Y, Grinspan JB, Abrams CK, Scherer SS (2007) Glia 55:46–56.31. Bae YS, Kang SW, Seo MS, Baines IC, Tekle E, Chock PB, Rhee SG (1997)

J Biol Chem 272:217–221.32. Sundaresan M, Yu ZX, Ferrans VJ, Irani K, Finkel T (1995) Science 270:296–

299.33. Seshiah PN, Weber DS, Rocic P, Valppu L, Taniyama Y, Griendling KK (2002)

Circ Res 91:406–413.

Retamal et al. PNAS � May 15, 2007 � vol. 104 � no. 20 � 8327

CELL

BIO

LOG

Y