Eur. J. Biochem. 210,249-256 (1992) 0FEBS 1992

Voltage dependence of liver gap-junction channels reconstituted into liposomes and incorporated into planar bilayers Jean-Luc MAZET ', ThkrZse JARRY ', Daniel GROS' and Franqoise MAZET'

' Laboratoire de Physiologie Cellulaire, Centre National de la Recherche Scientifique, URA 1121, Universitt Paris-Sud, France Laboratoire de Biologie de la Differentiation Cellulaire, Centre National de la Recherche Scientifique, URA 179, Facult6 des Sciences de Lumigny, Marseille, France (Received April 22/July 6, 1992) - EJB 92 0567

The voltage dependence of rat liver gap junctions was investigated using non-denaturing solubilization and reconstitution of gap-junction protein into proteoliposomes in controlled conditions of connexon aggregation. The presence of liver connexin 32 in reconstituted proteoliposomes was checked with specific antibodies. The proteoliposomes were inserted into planar lipid bilayers by fusion. The single-channel conductance was voltage independent, and its magnitude was 700 1900 pS in 1 M NaC1, as expected from other reports, assuming that conductance is linear with ion activity. The channels were open at zero voltage and completely closed above 40 mV in either direction. This steep voltage dependence corresponded to an open/closed-state voltage difference of 19 mV and to 3.5 gating charges moving through the field. When several channels were inserted into the bilayer, a large fraction of the membrane conductance became voltage insensitive. These results show that the isolated channel units are highly voltage dependent and are consistent with the assumption that aggregated connexons interact through links which prevent voltage-sensitive conformational changes.

Gap junctions are specialized domains of plasma membrane containing numerous intercellular channels connecting the cytoplasm of adjacent cells, but excluding access to extracellular space. The gap junctions are composed of paired multimeric membrane- spanning protein channels known as hemi-channels or connexons. Immunochemistry and molecular-biology studies have begun to characterize a family of gap-junction proteins called connexins, with considerable differences in molecular size and antigenicity. While the sequences of these proteins exhibit regions of substantial similarity in the hydrophobic transmembrane segments and in extracellular loops, they have sequence divergence in the hydrophilic cytoplasmic domains (review: Bennett et al., 1991). The question is how to correlate the activity of the channel to its structure. In the present report, we have focused on voltage-dependent gating properties. In tissues where gap-junction structure is symmetrical, voltage dependence is controversial. Symmetrical dependence of gap junctions on transjunction voltage has been described in embryonic cells of Amphibia and fish (Spray et al., 1979; White et al., 1982). However, in some adult tissues, septate axon (Johnston and Ramon, 1982; Verselis and Brink, 1984), mammalian heart (Riidisiili and Weingart, 1989; White et al., 1985),mammalian pancreatic fl cells (PCrez-Armandariz et al., 1991), no voltage dependence on transjunction voltage was observed. A more complex situation appears in insect salivary glands, where a high sensitivity to transmembrane potential Correspondence to F. Mazet, Laboratoire de Physiologie Cellulaire, Universite Paris-Sud, Bltiment 443, F-91405 Orsay Cedex, France

was reported (Obaid et al., 1983). More recently, the control of gap-junction conductance by voltage was re-evaluated in Drosophilae and was shown to be dependent on two different kinds of electrical potential (Verselis et al., 1991): the potential difference between the cells (transjunction voltage) and the potential difference between the channel interior and the extracellular space (transmembrane voltage). A quite different voltage dependence has been known for a long time in rectifying synapses (Furshpan and Potter, 1959; Auerbach and Bennett, 1969; Nicholls and Purves, 1972). In these tissues, the unidirectional transmission of the electrical signal is supported by rectifying gap junctions (Giaume and Korn, 1983). With the application of patch-clamp techniques for monitoring junction currents, it appeared that the voltage sensitivity of gap junctions might depend on the level of coupling. In rat lacrimal glands, gap junctions were found to be independent of the transjunction potential, but a high transjunction voltage dependence developed before complete uncoupling (Neyton and Trautmann, 1985). A comparable observation was made in developing tissues. In neonatal and embryonic heart, gap junctions were voltage dependent, as in blastomeres, but voltage dependence diminished as total junction conductance increased (Rook et al., 1988; Veenstra, 1990). From these results, an hypothesis was derived that junction channels would interact when embedded within large arrays (Jongsma et al., 1991). Two types of interaction have been assumed: voltage dependence could either be hindered by cytoplasmic-access resistance, or gap-junction channels embedded in large arrays would form links which result in blocking conformational changes.

250 Little is known about the voltage dependence of intercellular low-resistance pathways in liver gap junctions in vivo. The first measurements in whole liver did not exhibit any voltage dependence (Graf and Petersen, 1978; Meyer et al., 1981). However, they were subject to considerable uncertainty, due to the three-dimensional-cable electrical structure of this tissue. More recent experiments with isolated hepatocytes doublets (Graf et al., 1984; Spray et al., 1986b; Reverdin et al., 1988), and in isolated liver gap junctions (Spray et al., 1986a), lead to the conclusion that junction conductance was not voltage dependent. On the contrary, voltage dependence varied in gap junctions incorporated into a planar lipid bilayer in the presence of detergent (Young et al., 1987; Campos-deCarvalho et al., 1991) or in cells transfected with connexin 32 (Harris, 1991; Moreno et al., 1991a,b). These conflicting results lead us to reinvestigate the voltage dependence of liver gap junctions. We took advantage of several benefits of the proteinsolubilization and reconstitution method described by Mazet and Mazet (1990). First, the liverjunction plaques were totally dissociated and solubilized by the detergent mixture digitonin/ n-octyl-P-D-glucoside, whereas none of the detergents used previously (Triton X-100, Nonidet P-40, sarkosyl, Chaps) was known to split liver gap junctions (Manjunath et al., 1984). Second, the aggregation state of the gap junctions in proteoliposomes could be controlled through reconstitution conditions, and the structure of the resulting clusters are identical to the native structure (Mazet and Mazet, 1990). Third, reconstituted gap junctions were incorporated into planar bilayer in the absence of detergent and of exogenous proteins. The present results show that the conductance of isolated gapjunction unit is highly dependent on transjunction voltage, and this is consistent with the assumption that voltage dependence is decreased on junction-particle aggregation. An abstract of the present results has already been published (Mazet et al., 1991).

MATERIALS AND METHODS Preparation of anti-peptide antibodies The peptide SPGTGAGLAEKSDRCSAY was synthesized according to Merrifield (1963) and purified by HPLC (purity > 90%). The first 17 amino acids of this peptide correspond to residues 266-282 of the rat connexin-32 Cterminus (Paul, 1986). The peptide was coupled to keyholelimpet hemocyanin by means of bis(diaz0benzidine). Hens were immunized by intramuscular injection of the peptide conjugate emulsified with complete Freund’s adjuvant; they were boosted every month with the peptide conjugate emulsified with incomplete Freund’s adjuvant. Egg-yolk-IgGenriched fractions were obtained according to Polson et al. (1980). Anti-peptide IgG was purified by affinity chromatography from egg-yolk-IgG-enriched fractions, as described by Dupont et al. (1988). Egg-yolk-IgG-enriched fractions from preimmune yolks were treated and eluted from the affinity column as for immune egg-yolk IgG. The collected fractions will be referred to as preimmune-egg-yolk-IgG fractions. Biological material and reconstitution Rat liver intact gap junctions were isolated according to Zimmer et al. (1987). For proteoliposome reconstitution, they

-

1

2

3 &

4

-

94

94

94

67

67

61

43

43

43

29

29

29

21

21

21

5

6

14 14

14

Fig. 1. Characterization of liver gap-junction proteins by immunoblotting of hen 1gG specific for a peptide containing residues 266 - 282 of rat connexin 32. Lanes I and 2: SDSjPAGE analysis, Coomassieblue-stained gels. The major protein of gap junctions isolated from rat liver, as described by Manjunath et al. (1984), migrates at 2728 kDa. The 54-kDa band is a dimer of this major protein. Lanes 3 and 4: immunoblot analysis of rat liver gap junctions. An immunoblot of isolated rat liver gap junctions (lane 4)was incubated with hen IgG specific for a peptide containing residues 266 - 282 of rat connexin 32, then treated as described in Materials and Methods. The 28-kDa major protein and its dimer are both labeled. In control experiments, with either no primary antibodies or a preimmune-eggyolk-IgG fraction instead of primary anti-peptide IgG, no labeling was detected. Lanes 5 and 6: immunoblot analysis of gap-junction proteins reconstituted into liposomes. The proteoliposomes were solubilized, fractionated by electrophoresis and blotted onto nitrocellulose membranes. The control (lanes 6) was incubated and revealed in the same way as lane 4. Lanes 1, 3 and 5 indicate the position of standard proteins. Their molecular masses (kDa) are indicated. Arrows indicate the tops of the separating gels (lanes 1 and 2) and immunoblot (lanes 3,4, 5 and 6).

were solubilized in a mixture of digitonin and n-Octyl-P-Dglucoside as already described (Mazet and Blattmann, 1988). After elimination of the aggregated material by high-speed centrifugation (1 h, 110000 g), the supernatant was added to a dry mixture of lecithin and cholesterol. n-Octyl-S-D-glucoside was eliminated by dialysis, as already described in detail (Mazet and Mazet, 1990). Lecithin (Sigma, type ITS) was purified according to Bligh and Dyer (1959); the cholesterol was from Sigma (chromatography standard). lmmunoblotting Rat liver gap junctions or proteoliposome samples were fractionated by electrophoresis, then transferred according to Towbin et al. (1979) onto nitrocellulose membranes (0.22 pm; Schleicher & Schull) at constant voltage ( 5 5 V) for 2 h. Transfer of proteins was checked by staining the nitrocellulose membrane with Ponceau S. Immunoblots were first saturated with so-called Blotto solution (Johnson et al., 1984), then incubated overnight at 4°C with affinity-purified anti-peptide IgG (10 pg/ml in Blotto) as described by Dupont et al. (1988). Immunoblots were treated with a secondary antibody, biotinylated goat anti-(hen IgG) F(ab’)z (dilution 1 :2000; Jackson Immunoresearch Lab.), then with 1 : 2000-diluted peroxidase-labeled streptavidin (Jackson Immunoresearch Lab.). Peroxidase activity was detected with hydrogen peroxide and 4-chloronaphthol. Control experiments were carried out using either preimmune egg-yolk-1gG fractions as primary antibody or secondary antibody only.

251 A

C

B

38.2 mV

18.6 mV

-6.4

mV

7

-24.0 mV

-33.0 m V

5 s

Fig. 2. Characterization by immunofluorescence of hen IgG specific for a peptide containing residues 266-282 of rat connexin 32 IgG. Rat liver frozen sections were incubated with anti-peptide Ig then with fluorescein-isothiocyanate- labeled secondary antibodies, as described in Materials and Methods. The fluorescent spots around the cells highlight the connexin-32-containing gap junctions. Their distribution is similar to that observed by Traub et al. (1989) using antibodies directed against the 28-kDa junction protein. Bar, 50 pm.

Immunofluorescence

Unfixed rat liver was frozen in liquid nitrogen. Tissue samples were sectioned at 20 "C using a Riechert-Jung cryostat. Sections were incubated first with saturation solution pH 7.2, containing 1 % bovine serum albumin and 0.02 % saponin) for 30 min at room temperature, then with purified anti-peptide egg-yolk IgG (3 pg/ml in saturation solution) overnight at 4 "C. After washing, the sections were incubated with fluorescein-isothiocyanate-labeled goat anti-(hen IgG) F(ab), (Kirkegaard and Perry Laboratories), diluted 1 : 100 in saturation solution, for 1 h at room temperature. The preparations were examined with a Zeiss light microscope equipped for epifluorescence. Electrophysiology

Planar lipid-bilayer membranes were formed at 25 "C across a 0.5-mm2 hole in the partition septum of a two-compartment polytetrafluoroethylene chamber. The chamber

0.2 s -

:I

N

Fig. 3. Channel activity of proteoliposomes incorporated into a planar lipid bilayer. An aliquot of connexin 32 reconstituted into proteoliposomes was added to the cis side of a planar bilayer preformed with asolectin/cholesterol (4:1) in a double Compartment chamber tilled with electrolyte solution. It was allowed to fuse with the bilayer. In the present experiment, only a single channel was incorporated. The current was recorded at various potentials and is represented at two different time scales. (B) Underlined segment of (A) a t an enlarged time scale. (C) The probability of the channel being open is illustrated by amplitude histograms of the currents. The ordinate scale is the same as in the current recording. The amplitude of each peak (abscissa) reflects the time spent at the corresponding current level (see Materials and Methods). The zero current is indicated by the continuous line. Near zero voltage (-24.0 to 18.6mV), the major peak corresponds to the open state. At higher voltage, the channel is totally closed. The small peak of open state visible at - 33.0 mV comes from the beginning of the recording.

nearest the experimentalist was called the cis compartment. The septum was first treated with lecithin/cholesterol(4: 1, by mass) dispersed in hexane (Aldrich, gold label). After hexane evaporation under N2, each compartment was filled with 5 ml electrolyte solution (1 M NaC1, 1 mM CaC12, 5 mM Mes, pH 5.7), and a drop of squalane/decane ( 1 : 1) was injected into the hole using a Pasteur pipette. Squalane (Aldrich) was filtered twice on an alumina (Sigma, type WB5) column; decane was from Wiley Organics (99.9%). The state of the planar bilayers obtained by this technique was checked with gramicidin, the activity of which is highly sensitive to membrane composition and thickness. The single-channel characteristics (amplitude and open time) were indistinguishable from that

252 o

-60

-40

-20

+

0

m

40

60

Transmembrane voltage (mV)

Fig. 4. Probability of open channels. The probability of open channels was computed in three experiments, where only a single channel was inserted into the membrane. Each point represents the time fraction spent in the open state. It was computed from the relative area of the open-state peak in the current-amplitude histogram (Fig. 3C). The channels are closed above 40 mV. The continuous curve is the best fit computed on the assumption of two symmetrical closing processes (see Results) with the parameter values V1 = 15.8 mV, V , = -22.3 mV and z = 3.5. Given the dispersion of the results, the 3-mV shift to the left of the bell-shaped curve was not considered significant. The channel unit conductances of the three experiments were 800, 1470 and 1700 pS.

observed in membranes formed from a mixture of lipids with the solvents. It was observed that heptanol, a direct blocker of gap junctions, permeabilized the bare membranes. An aliquot ( 5 - 20 pl) of the proteoliposome preparation was added to the cis compartment after membrane formation. Sometimes channel incorporation occurred spontaneously. When no fusion was observed after 15 - 30 min, a brief 150200-mV step was applied to induce fusion. The first fusion event corresponded to a number of channels which were incorporated at the same time. After this fusion event, membrane activity was usually stable up to 2 h. Only membranes which showed no conductance at - 120 mV to 120 mV before proteoliposome addition were used. We do not report on experiments carried out in the presence of 100 mM NaCl, because we observed very few spontaneous incorporations at this concentration and the membranes were very fragile upon application of high-potential steps. Voltage pulses (1 -3 min, - 100 mV to 100 mV) were applied to the cis compartment of the chamber, while the trans compartment was maintained at virtual ground voltage. The transmembrane current was recorded on magnetic tape (Racal, DS4) and sampled (Data Translation, DT2821A) onto a 80286 PC IBM-compatible computer. In multiple-channel experiments, the mean current was computed from the total current recording during a single voltage pulse, discarding the initial transient response. With bare membranes, n-octyl-p-D-glucoside induced a steep voltage-dependent increase in conductance and disruption of the membrane. Therefore, we rejected all experiments in which membrane conductance increased with voltage, under 110 mV. It was checked that this guaranteed a n-octylfi-mglucoside concentration below 10 pM. Digitonin had no effect on bilayer conductance. When the membrane contained one or two channels, two different types of histogram were used: current amplitude and

transition-step amplitude. A frequency histogram of currentsample amplitude was built for each potential pulse. The first 2 s were discarded, because they correspond to the relaxation time of the artificial membrane. The frequency histograms of transition steps were extracted from the current sample by a shape-recognition algorithm (Andersen, 1983).

RESULTS Characterizationof anti-peptide IgG SDSjPAGE analysis of isolated rat liver gap junctions shows that they are made up of a major protein of 27 - 28 kDa (Fig. 1, lane 2), demonstrated as connexin 32 by Paul (1986). By immunoblotting, this 27 - 28-kDa protein and its dimer were specifically labeled with anti-peptide IgG raised against residues 266-282 of rat connexin 32 (Fig. 1, lane 4). In control experiments (see Materials and Methods), no labeling was detected (not shown). In frozen sections of rat liver, sites immunoreactive to antipeptide IgG were distributed between hepatocytes (Fig. 2), as expected for connexin-32-containing gap junctions. The distribution of these sites was similar to that observed, for example, by Traub et al. (1989) using antibodies raised against the 27 - 28-kDa junction protein. In control experiments (see Materials and Methods), no labeling was seen (not shown). Reconstituted proteoliposomes When the electrophoresis gel was loaded with reconstituted proteoliposomes, the predominant band migrated at the connexin-32 position. Immunoblot analysis (Fig. 1, lane 6) showed that this band was labeled by the anti-peptide IgG. The high-molecular-mass band probably corresponds to a

253

1

I

I

-40

-60

I

I

-20

I

0

1

20

I

I

I

40

60

Transmembrane voltage (mV)

Fig. 5. Current/voltage relationship of a single channel. The number of current transitions during a single voltage pulse were plotted as a frequency histogram of transition amplitudes (see Materials and Methods). Each point corresponds to a single peak of the frequency histogram. The channel conductance (1470 pS in the present experiment) was estimated from a straight-line fit. The end points correspond to few and very brief open states, and their apparent deviation from the straight line was not considered significant. The insert illustrates a transition of the open state of 1830 pS to 1470 pS in a single-channel experiment. It shows that the various conductances may reflect different states of the same channel, although the potential dependence of open-channel probability was the same before and after this conductance change. The present current/voltage relationship was derived after the occurrence of this transition.

dimer of connexin 32. This proves that the liver junction protein (connexin 32) was indeed incorporated into liposomes. Open probability of a single channel The number of channels inserted into the lipid bilayer, at the first incorporation of proteoliposomes, varied from one experiment to another. Experiments where only one channel was present in the membrane were selected by checking that only one conductance level was observed in conditions where the channel was open most of the time (low-voltage recording). Fig. 3 shows the recording of such a channel, mostly open near zero voltage and closed at 40 mV. It had only one open state. 11 experiments, in which more than one level was detectable, were not used in this study. One experiment, in which the channel was closed most of the time, was also discarded, since it was not possible to prove that the channel was single. In discarded experiments, voltage dependence was usually poor. The probability of a channel being open was measured in three experiments in which a single channel was inserted into the bilayer. Each probability was calculated as the relative area of the peak corresponding to the open state in the amplitude histograms of the currents (Fig. 4). The probability of an open channel was approximately bell shaped and centred around zero voltage. On the assumption that the closure involves two distinct symmetrical voltage-dependent process, the measured probability of an open channel was tentatively fitted with an equation based on the Boltzmann relation, P

=

+ exp[z.(V, . { 1 + exp[z. ( V ,

1/({1

~

V1).F/RT])

V,) . F / R T ] ) ),

where V1 and V z are the right and left threshold voltages, and z reflects the gradient of voltage dependence. The other

constants are F,the Faraday, R , the gas constant, and T , the absolute temperature. The fitting values of the parameters are given in the legend to Fig. 4. Not taking into account the slight shift of the curve to the left, the closing process corresponds to a potential energy of 19 mV and to 3.5 equivalent gating charges moving through the field. Single-channel conductance With seven bilayers, containing one or two channels, the single-channel conductance could be measured. In some cases, it was possible to observe changes in the conductance magnitude during experiments where only a single channel was inserted into the membrane. Sometimes the change occurred during the open state (Fig. 5, insert). This kind of change was infrequent compared to the open/closed transition, and allowed measurement of each individual conductance amplitude. The example in Fig. 5 shows that the current/voltage relationship of transition amplitude versus transmembrane voltage is linear. Channel conductance was computed either from the slope of the linear fit or from a single point when an open state did not last long enough to make a curve. They were 7001900 pS, without clear amplitude multiples. These conductances were possibly related to different conducting states of a channel, but not to the gating process, since the potential dependence was apparently not modified (not shown). Multiple channel-current recordings When more than two channels were inserted into the membrane, occasionally small conductance transitions were seen (about 300 pS), but most often a single transition could not be resolved in our conditions and only mean currents were

254 800

-Mo

I

1

1

1

1

1

l

I1

1

1

1

1

1

l

1

Transmembrane voltage (mV)

Fig. 6. Current/voltage relationship in a bilayer containing many channels. The graph represents an experiment in which many channels were incorporated into the membrane. Each point is the mean current measured during a single voltage pulse, not taking into account the initial transient current (see Materials and Methods). The maximum conductance of the membrane, near zero voltage, was 19.4 nS. Taking into account a maximal open-channel probability of 87%, this corresponds to 12-32 channels, depending on the unit conductance of the single channels. Curve A was determined from a maximal Conductance of 19 pS, assuming that all the channels behave like single channels with V1 = - V , = 19 mV and z = 3.5 (Fig. 4). Curve B was drawn assuming that 60% of the conductance behaved like single channels and that the rest was voltage insensitive.

calculated. In bilayers containing many channels, the current/ voltage relationship exhibited either no voltage dependence (two experiments) or a symmetrical voltage-dependent conductance (nine experiments), but conductance was never completely blocked up to 90 mV. In Fig. 6, the membrane conductance near zero voltage was 19 nS, corresponding to 1232 channels, depending on the unit conductance of single channels. On the basis of the reported voltage dependence of a single channel, the expected current/voltage curve was computed from the parameters V1, V , and z. The currents measured were clearly different from the result expected. In fact, they could reasonably be fitted on the assumption that 40% of the conductance was voltage insensitive. It may be inferred that part of the inserted channels were not voltage dependent. However, there was no correlation between total membrane conductance and the voltage-independent fraction, possibly because when several channels were present in the membrane, the state of aggregation varied from one experiment to the other. DISCUSSION Rat connexin 32 was reconstituted into proteoliposomes after solubilization of purified rat liver gap junctions by a detergent mixture. The soluble material might contain aggregates as well as isolated particles. Before reconstitution, the solution underwent a high-speed centrifugation to remove all aggregates. After detergent elimination, the proteoliposomes contained gap-junction-like structures and a major protein SDSjPAGE band of 28 kDa (Mazet and Mazet, 1990). We show in the present report that this band is specifically labeled with antibodies raised against residues 266 - 282 of rat connexin 32. The other weakly labeled band of higher molecular mass probably corresponds to a connexin-32 dimer.

When gap junctions, solubilized and reconstituted into proteoliposomes, were incorporated into a planar lipid bilayer, the membrane became electrically permeable. At low membrane conductance, single channels could be observed. Their unit conductances were linear and ranged over 700 - 1900 pS. Of course, this broad distribution might be due to channels of different protein composition, since it is known that liver gap junctions also contain variable amounts of connexin 26 as well as connexin 32 (Henderson et al., 1979; Nicholson et al., 1987). Our experiments do not separate these two species, therefore it cannot be excluded that reconstituted proteoliposomes containing connexin 26 may be inserted into the bilayer. However, this could not explain the conductance changes which occur during the open state in membranes containing a single channel. Therefore, at least a part of the conductance variability must be due to minor changes in the connexin state, which do not affect the gate. It is not possible to compare the conductance of reconstituted channels to that of liver gap junction, because it has never been successfully measured in situ. However, in other preparations, the reported conductance was 150 pS in 150 mM KC1 (cell pairs; Spray et al., 1986a), 140 pS in 100 mM NaCl (gap junction incorporated into planar lipid bilayers in the presence of detergent; Young et al., 1987), 135 pS in 135 mM potassium glutamate (cells transfected with rat liver connexin 32; Eghbali et al., 1990). The expected conductance in 1 M NaCl may be calculated, assuming that gapjunction conductance varies linearly with ionic activity. The activity ratio of 100 mM NaCl (150 mM KCl) to 1 M NaCl is 8.5 (5.8). If one takes into account permeability ratios of 1.2:1 :0.7 for K+/Na+/Cl- (Young et al., 1987), the expected channel conductance should be about 780 - 1200 pS in our conditions. These values are similar to those reported here. The question may be raised whether both channels and hemi-channels could be inserted into the planar bilayer mem-

255 brane. This might account for the large dispersion of the conductance values of single channels, but it could not explain why single channels with a wide range of conductance amplitude were symmetrically dependent on the voltage. Neither could it account for the observation of transitions between different conductance values occurring during the open state. In multiple channel-current recordings, the presence of hemichannels could rationalize membrane conductance remaining above 40 mV. However, on assuming that the hemi-channels are asymmetrically voltage dependent and inserted in both directions, one would require at least 80% hemi-channels to reconstruct Fig. 6, curve B. This proportion was not reflected in single-channel experiments. In single-channel experiments, the channel was completely closed above 40 mV. On the contrary, a poor voltage dependence was present in high-conductance membranes, which could never be completely blocked. Since the channels were incorporated simultaneously (see Materials and Methods), it is highly probable that they were inserted as clusters in highconductance membranes. This is consistent with the observation of preformed aggregates in reconstituted proteoliposomes. The characteristics of channels participating in the voltage-independent fraction could not be elucidated in the present experiments. In particular, it could not be determined if they are related in any way to the few 300-pS steps detected. A complete lack of voltage dependence of liver gap junctions has been observed in whole liver (Graf and Petersen, 1978), in isolated hepatocytes (Graf et al., 1984; Spray et al., 1986b; Reverdin et al., 1988) and in isolated liver gap junctions (Spray et al., 1986a). On the contrary, a voltage-dependent closing was reported with gap-junction plaques incorporated into planar lipid membranes (Young et al., 1987) and with hepatoma cell pairs transfected with connexin 32 (Moreno et al., 1991b). In all the reports cited above, the activation voltage (25 100 mV) was higher, the voltage sensitivity (1.75-3) was lower, and the voltage-independent conductance (more than loo/,) greater, than the parameters fitting the present data. Our results are more similar to blastomere gap junctions ( V , = 14 mV; z = 3.5; voltage-independent conductance, 5 % ; Harris et al., 1981). This is probably a property of the fully dissociated gap-junction units achieved with the double-detergent-solubilization method (Mazet and Blattmann, 1988). If this were true, the poor voltage dependence observed in highconductance membranes would result from interactions between junction units occurring in gap-junction plaques. Such behaviour has been observed in various tissues. Voltage sensitivity apparently diminishes as membrane conductance increases in lacrimal gland (Neyton and Trautmann, 1985), neonatal rat heart (Rook et al., 1988), chick embryo ventricle (Veenstra, 1990) and cultured astrocytes (Giaume et al., 1991). The disappearance of voltage sensitivity when cells are highly coupled has been attributed to an indirect interaction by Rook et al. (1988), who have argued that the apparent lack of voltage sensitivity was due to the existence of a cytoplasmicaccess resistance which would be more important for large gap junctions than for small ones (Jongsma et al., 1991). In liver cells, at transjunction voltages higher than 40 mV, each time a channel closes in a junction plaque, its neighbours should become more sensitive to the potential and close in turn. Thus, the whole gap junction should end in a closed state. This behaviour might account for the high voltage dependence of blastomere gap junctions (Harris et al., 1981). In rat liver, however, assuming that reconstituted gap junctions can be inserted into membranes as clusters, our results are

more consistent with the assumption that gap junctions embedded in large arrays form links which prevent voltage-sensitive changes in conformation. The peptide SPGTGAGLAEKSDRCSAY was kindly synthesized by Dr J. P. Briand (Institut de Biologie Moliculaire et Cellulaire, Centre National de la Recherche Scientifique, Strasbourg, France). We thank Christian Giaume for his fruitful comments on the manuscript. This work was supported by a grant from Institut Nationalde la Santi et de la Recherche Midicale (CRE 90-07-05).

REFERENCES Andersen, 0. S. (1983) Biophys. J . 41, 147- 165. Auerbach, A. A. & Bennett, M. V. L. (1969) J . Gen. Physiol. 53,211 237. Bennett, M. V. L., Barrio, L. C., Bargiello, T. A., Spray, D. C., Hertzberg, E. & Saez, J. C. (1991) Neuron 6 , 305 - 320. Bligh, E. G. & Dyer, W. J. (1959) Can. J . Biochem. Physiol. 37,911 91 7. Campos-de-Carvalho, A. C., Hertzberg, E. L. & Spray, D. C. (1991) Braz. J. Med. Biol. Res. 24, 527 - 537. Dupont, E., El Aoumari, A,, Roustiau-Stvtre, S., Briand, J. P. & Gros, D. (1988) J . Membr. Biol. 104, 119-128. Eghbali, B., Kessler, J. A. & Spray, D. C. (1990) Proc. Natl Acad. Sci. USA 87, 1328-1331. Furshpan, E. J. & Potter, D. D. (1959) J. Physiol. (Lond.) 145, 289325. Giaume, C. & Korn, H. (1983) Science 220, 84-87. Giaume, C., Fromaget, C., El Aoumari, A., Cordier, J., Glowinski, J. & Gros, D. (1991) Neuron 6, 133-143. Graf, J. & Petersen, 0 . H. (1978) J. Physiol. (Lond.) 284, 105-126. Grdf, J., Gautam, A. & Boyer, J. L. (1984) Proc. Natl Acad. Sci. USA 81,6516-6520. Harris, A. L. (1991) in Biophysics of gap junction channels (Peracchia, C., ed.) pp. 373 - 389, CRC Press. Harris, A. L., Spray, D. C. & Bennett, M. V. L. (1981) J . Gen. Physiol. 77,95-117. Henderson, D., Eibl, H. & Weber, K. (1979) J . Mol. Biol. 132, 193218. Johnson, D. A,, Gantsch, J. W., Sportman, J. P. & Elder, J. M. (1984) Gen. Anal. Tech. 1, 3 - 8. Johnston, M. F. &Ramon, F. (1982) Biophys. J . 39, 115-117. Jongsma, H. J., Wilders, R., van Ginneken, A. C. G. & Rook, M. B. (1991) in Biophysics ofgap junction channels (Peracchia, C., ed.) pp. 163-172, CRC Press. Manjunath, C. K., Goings, G. E. & Page, E. (1984) J . Membr. Biol. 78, 147-155. Mazet, F. & Blattmann, A. (1988) C. R. Acad. Sci. Paris 307, 619684. Mazet, F. & Mazet, J.-L. (1990) Exp. Cell Res. 118, 312-315. Mazet, F., Mazet, J.-L., Gros, D. & Jarry, T. (1991) Biol. Cell 73, 7a. Merrifield, R. (1963) J . Am. Chem. Soc. 85,2149-2154. Meyer, D. I., Yancey, S. B. & Revel, J.-P. (1981) J . CellBiol. 91,505523. Moreno, A. P., Campos-de-Carvalho, A. C., Verselis, V., Eghbali, B. & Spray, D. C. (1991a) Biophys. J. 59, 920-925. Moreno, A. P., Eghbali, B. & Spray, D. C. (1991b) Biophys. J . 60, 1267- 1217. Neyton, J. & Trautmann, A. (1985) Nature 317, 331 -335. Nicholson, B. J ., Dermietzel, R., Teplow, D., Traub, O., Willecke, K. & Revel, J.-P. (1987) Nature 329, 732-734. Nicholls, J. G. & Purves, D. (1972) J. Physiol. (Lond.) 225, 637656. Obaid, A. L., Socolar, S. J. & Rose, B. (1983) J. Membr. Biol. 73, 69 - 89. Paul, D. (1986) J. Cell Biol. 103, 123-134. Ptrez-Armandariz, M., Roy, C., Spray, D. C. & Bennett, M. V. L. (1991) Biophys. J. 59, 76-92. Polson, A., von Wechmar, B. M. & van Regnmortel, M. H. V. (1980) Immunol. Commun. 9,475-493.

256 Reverdin, E. C. & Weingart, R. (1988) Am. J . Physiol. 254, C226C234. Rook, M. B., Jongsma, H. J. & van Ginneken, A. C. G. (1988) Am. J. Physiol. 255, H770 - H782. Riidisiili, A. & Weingart, R. (1989) P’iigers Arch. 415, 12-21. Spray, D. C., Harris, A. L. & Bcnnett, M. V. L. (1979) Science 204, 432 - 434. Spray, D. C., Saez, J. C., Brosius, D., Bennett, M. V. L. & Hertzberg, E. L. (1986a) Proc. Natl Acad. Sci. USA 83,5494-5491. Spray, D. C., Ginzberg, R. D., Morales, E. A,, Gatmaitan, Z. & Arias, 1. M. (1986b) J . Cell Biol. 103, 135-144. Towbin, H., Staehelin, T. & Gordon, J. (1979) Proc. Nutl Acud. Sci. USA. 76,4350 - 4354. Traub, O., Look, J., Dermietzel, R., Brummer, F., Hulser, D. & Willecke, K. (1989) J. Cell Biol. 108, 1039- 1051.

Veenstra, R. D. (1990) Am. J . Physiol. 258, C662-Cc612. Verselis, V. & Brink, P. R. (1984) Biophys. J. 45, 147-150. Verselis, V. K., Bennett, M. V. L. & Bargiello, T. A. (1991) Biophys. J . 59, 114-126. White, R. L., Spray, D. C., Campos-de-Carvalho, A. C. & Bennett, M. V. L. (1982) SOC. Neurosci. Abstr. 8, 944. White, R. L., Spray, D. C., Campos-de-Carvalho, A. C., Wittemberg, B. A. & Bennett, M. V. L. (1985) Am. J . Physiol. 249, C447(35.5. Young, J. D., Cohn, Z. A. & Gilula, N. B. (1987) Cell 48, 733-743. Zimmer, D. B., Green, C. R., Evans, W. H. & Gilula, N. B. (1987) J . Biol. Chem. 262,7151 -7163.