Thomas C. Jenkins Department of Biophysics, Johns Hopkins University, ... d Department of Anatomy and Cell Biology, Harvard Medical School Boston, MA ...
Molecular Brain Research, 15 (1992) 269-280
269
Elsevier Science Publishers B.V. BRESM 70481
Ion channels in single bilayers induced by rat connexin32 A.L. Harris a, A. Walter b, D. Paul c, D.A. Goodenough d and J. Zimmerberg
e
" Thomas C. Jenkins Department of Biophysics, Johns Hopkins University, Baltimore, MD 21218 (USA), b Department of Physiology and Biophysics, Wright State University, Dayton, OH 45435 (USA), c Department of Neurobiology, Harvard Medical School, Boston, MA 02115 (USA), d Department of Anatomy and Cell Biology, Harvard Medical School Boston, MA 02115 (USA) and e Laboratory of Theoretical and Physical Biology, National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD 20892 (USA)
(Accepted 19 May 1992)
Key words: Ion channel reconstitution; Gap junction; Connexin; Lipid bilayer; Liposome; Transport-specific fractionation
The gap junction channel mediates an important form of intercellular communication, but its detailed study is hindered by inaccessibilityin situ. We show here that connexin32, the major protein composing junctional channels in rat liver, forms ion channels in single bilayer membranes. The properties of these reconstituted connexin32 channels are characterized and compared with those of gap junction channels. The demonstration that connexin32 forms channels in single membranes has implications for assembly and regulation of junctional channels, and permits detailed study of the gating, permeability and modulation of this channel-forming protein.
INTRODUCTION The gap junction channel forms a pathway for exchange of cytoplasmic factors between cells that is important for normal development and mature tissue function 4'28'36'76. Junctional channels are formed by two end-to-end hexameric hemichannels (connexons), each spanning a single plasma m e m b r a n e 44'71.. They are composed of highly homologous proteins called connexins 6. In rat liver, junctions are composed of connexin32 4°'55 along with a small amount of connexin26 53'69. The junctional aqueous pore ( ~ 14 A diameter) is permeable to large molecules 3 and maximally open at zero transjunctional voltage 64'65. Access to both ends of the pore is via cytoplasm, and the regulatory sites for channel gating are intracellular. This configuration presents difficulties for study of junctional channel physiology and ligand-activation in situ. Reconstitution of the junctional channel in planar phospholipid bilayers would permit detailed study of its permeability, gating and modulation. Reconstitution is particularly important to identify modulatory factors that act directly on the channel and not via cytoplasmic intermediates.
One approach is to reconstitute the double-membrane channel in a pair of closely-apposed bilayers, but this has serious technical impediments. Our singlem e m b r a n e approach recognizes that the junctional channel is composed of two single-membrane structures (connexons or hemichannels) that each contain an aqueous pore. Structural, biochemical and physiological data strongly indicate that each connexon contains molecular components capable of mediating gating transitions and gating sensitivities 5'44'65'68'71'73'75. Each connexon may have the capability to function as a single-membrane channel, even though it may not often do so in cells. Investigation of junctional channels may be aided by study of their single-membrane subunits, which have the potential of being studied by standard reconstitution techniques. The problem has been identification of bilayer channels as formed by connexin. Protein from gap junctions isolated by cell fractionation can induce ion channel activity in planar b i l a y e r s 43'49'67'74'77. It is difficult to identify such channels as junctional because there is no uniquely identifying physiology, and there are no specific toxins or blockers known to act on the junctional channel itself.
Correspondence." A.L. Harris, Thomas C. Jenkins Department of Biophysics, Johns Hopkins University, Baltimore, MD 21218, USA.
* We refer to the hexameric hemichannels in one membrane as connexons, and to the connexin channels through two plasma membranes as junctional channels.
270 It is not sufficient to biochemically purify a putative junctional protein because the bilayer 'assay' can reveal activity of single contaminating molecules. We use a novel approach that circumvents these difficulties (preliminary reports in refs. 30-32). We establish that connexin32 forms ion channels in single phospholipid membranes by (a) incorporating junctional protein into unilamellar vesicles, (b) isolating the vesicles that are permeable to sucrose, (c) showing that the sucrose-permeable vesicles are specifically enriched for full-length connexin32 (identified with a monoclonal antibody), and (d) fusing the vesicles into planar phospholipid bilayers to observe channel activity directly. The reconstituted single-membrane connexin32 channels have permeability, conductance and voltage-sensitivity consistent with data from junctional channels. The demonstration that connexin32 forms functional ion channels in single membranes makes possible direct and detailed study of connexin channels. MATERIALS
AND METHODS
Vesicle formation Reconstitution of protein into unilamellar vesicles followed Mimms et aL as. Egg phosphatidylcholine, bovine brain phosphatidylserine (Avanti Polar Lipids) and recrystallized cholesterol (Sigma) were dissolved in chloroform at a mole ratio of 2:1:0.75, with a trace amount of [14C]dipalmitoylphosphatidylcholine. The lipids were dried under argon, and suspended at 1 m g / m l in urea buffer containing 80 m M n-octyl-O-o-glucopyranoside (Calbiochem). Gap junctions isolated by detergent extraction from rat liver I were added, and the mixture sonicated briefly in a bath sonicator (Laboratory Supplies Co.) and incubated on ice for at least 20 min with occasional gentle swirling. Approximately 10% of the connexin32 was solubilized by this procedure. The mixture was then applied to a 1.5 x 20 cm Bio-Gel A-0.5m column pretreated with sonicated phosphatidylcholine vesicles. The material was eluted with urea buffer at 9 m l / h at 4°C. The vesicles appeared in the void volume. Vesicle size distribution was determined by gel-filtration (6000 PW) to be monodisperse with m e a n diameter 750 A. In the vesicle-forming mixture, the protein to lipid ratio was typically 1 : 50 (w/w), and the mole ratio approximately 1:13,000 (moles connexons:moles lipid). After vesicle formation, the protein to lipid ratio was approximately the same, corresponding, to an amount of connexin32 sufficient for 2 - 4 connexons per 750 A diameter vesicle.
liquid scintillation counting to determine lipid distribution. Vesicles that are not permeable to sucrose move into the gradient a short distance to a position determined by the densities of the trapped urea buffer and the lipid m e m b r a n e (which is more dense than the urea buffer). Vesicles that are permeable to sucrose move lower into the gradient as the vesicles equilibrate with the (more dense) external solution, and come to an equilibrium position near the bottom. They do not pellet due to the relative buoyancy of the lipid in the lower part of the gradient. Though the vesicle formation technique we use 48 is standard for production of unilametlar structures, in theory it is possible that multilamellar structures are formed. For such structures to migrate to the same gradient position as the unilamellar vesicles, conducting channels would have to be present in each of the lamellae. Calculated time for equilibration of external sucrose with vesicle internal volume through one open connexon is a few milliseconds. Calculations and control experiments using proteins that do not form channels show that the increase in density is not due to the density contribution of the protein in the m e m b r a n e 34. The principle of using a density-shift to fractionate vesicles was adapted from Goldin and R h o d e n 23.
Protein blots and immunoblots Samples were analyzed by S D S - P A G E with 13% acrylamide separating gels using a minigel apparatus (Bio-Rad). Protein was electrotransferred to Immobilon P V D F m e m b r a n e (Enprotuch) either in a standard wet tank system (Bio-Rad) at 54 V for 1 h or using a semi-dry blotter (American Bionetics) at 108 milliamps for 1-2 h. For immunoblots, the m e m b r a n e was probed with primary antibody (M12.13, a monoclonal against connexin32) 25 that binds to the cytoplasmic loop (Paul, D., unpublished observations) at 5 - 1 0 / x g / m l at 1 - 2 h at 37°C. Secondary antibody was an alkaline phosphatase-conjugated goat anti-mouse IgG (Boehringer Mannheim), incubated at 5 - 1 0 / x g / m l at 1 - 2 h at 37°C. Blots were developed in 0.1 m g / m l Nitro blue tetra-azolium (Sigma) and 0.05 m g / m l 5-bromo-4-chloro3-indolyl-phosphate (Boehringer M a n n h e i m ) for 1-5 min. For protein blots, the m e m b r a n e was stained with colloidal gold.
Bilayer experiments Planar phospholipid bilayers were formed from bacterial phosphatidylethanolamine and bovine phosphatidylserine 1:1, 2% in decane 5° (Fluka) and bathed in 182 m M KCI, 9 m M HEPES, and 0.1 m M E D T A at pH 7.4. After m e m b r a n e thinning, aliquots of density gradient fractions that contained vesicles were added to the front chamber, and mixed. If the bare bilayer conductance remained low and noise-free, MgCI 2 was added to the front and rear chambers to 10 mM. An osmotic gradient was imposed by addition of urea to the front chamber to 200 m M 13'81. All additions were followed by several minutes of thorough mixing. Voltages were imposed relative to the rear chamber. Bilayer currents were measured by current-to-voltage conversion (Analog Devices) with a 5 × 109 1-2 feedback resistor. Data was recorded on a Kipp and Z o n e n (freq. resp. ~ 5 Hz) a n d / o r a Gould recorder (filtered at 100 Hz).
RESULTS
Solutions Urea buffer contained 10 m M KCI, 10 m M HEPES, 0.1 m M EDTA, 3 m M sodium azide and 459 m M urea at pH 7.4. For the sucrose buffer, the urea was replaced by 400 m M sucrose so that both buffers were iso-osmolar at 490 m O s m / k g as measured with a vapor pressure osmometer (Wescor). Specific gravities (D 2°) of urea and sucrose buffers were calculated to be 1.0056 and 1.0511, respectively.
Transport-specific density shift technique The procedure used to fractionate vesicles into two populations on the basis of sucrose permeability is described fully in Harris et al. 34. Linear density gradients were formed from the urea and sucrose buffers in 5 ml ultracentrifuge tubes. Aliquots of vesicles (100-200/zl) were layered on each gradient. Gradients were spun at 300,000× g for 3 - 8 h at 4°C, fractionated, and aliquots taken for
Sucrose permeability of connexin-containing vesicles Protein from isolated rat liver gap junctions, predominantly connexin32, was solubilized in octylglucoside and exogenous purified lipid, and incorporated into single-walled lipid vesicles. The amount of protein in the reconstitution corresponded to ~ 2-4 connexons per vesicle. Vesicles permeable to a large solute (sucrose; Stokes radius ~ 5 A ) w e r e identified and isolated by the transport-specific vesicle selection protocol described in Materials and Methods and characterized in Harris et al. 34. Vesicles are formed in a urea
271 buffer and separated in a urea-to-sucrose iso-osmotic density gradient during an equilibrium spin. Vesicles permitting sucrose entry become more dense and move to a characteristic lower position in the gradient (Fig. 1A-C). Vesicles that are impermeable to sucrose do not increase in density and remain at a higher position in the gradient. In this assay, the permeability induced
by a single sucrose-permeable channel in a vesicle will result in a full shift in density. A significant fraction of the vesicles formed in the presence of gap junction protein consistently showed such an increase in density (Fig. 1D, E). Vesicles formed without gap junction protein remained in a single, low density band. We conclude that a
sucrose
A
B
C
DENSITY GRADIENT 40
3O E "6
lO
, ~
rt_tll 1o
15
fraction n u m b e r
D
E
Fig. 1. Transport-specific selection of vesicles containing protein from isolated rat liver gap junctions. Junctional protein was incorporated into unilamellar vesicles formed in a urea buffer, and the vesicles segregated on the basis of sucrose permeability on iso-osmolar density gradients (A-C) as described in Harris et al. 34. Vesicles permeable to sucrose increase in density, while those that are not remain at a lighter density (C). An example of the density shift is seen in the gradient tube (D; spun 3 h). In this example, a rhodamine°phosphatidylethanolamine label was incorporated into the vesicle membrane to reveal the vesicles. The upper arrow indicates the band formed by the unshifted (sucrose-impermeable) vesicles and the lower arrow that formed by the shifted (sucrose-permeable) vesicles. In a different experiment (spun 8 h), vesicle position was followed by radioactive phospholipid label (E; top of the gradient is at the left). Vesicles formed without junctional protein were found in a (sucrose-impermeable) band at fractions 6 and 7 (open bars). Vesicles formed with junctional protein separated into two bands, with approximately three-quarters of the vesicles at the higher density (sucrose-permeable) position. With longer spins, the unshifted band tended to slowly drift down (described in ref. 34); this accounts for the lower relative position of the unshifted vesicles in E.
272 detergent-soluble protein from the isolated junctions induced a sucrose permeability in the vesicles. To identify the specific protein responsible for the sucrose permeability, the proteins in the density gradient bands were run on SDS gels, transferred to blotting membrane, and stained for protein (using colloidal gold) and for connexin32 (using a monoclonal antibody against connexin3225 that binds to the cytoplasmic loop). The vesicles in the upper gradient band will contain only non-functional protein or protein that forms sucrose-impermeable channels. Vesicles in the lower band will contain protein that forms sucrose-permeable channels as well as some non-functional protein coincidentally present in the same vesicles. Therefore functional protein should be specifically enriched in the lower band, relative to non-functional protein, and vice versa.
The blots of vesicles showed bands corresponding to monomeric and multimeric forms of connexin32, and a proteolytic fragment of the connexin32 monomer (Fig. 2A). These bands are commonly seen in gets of isolated gap junctions 37"4°'s4'5s. Little or no non-connexin protein was detected in the vesicles. When the proteins in the two transport-selected populations of vesicles were compared, it was cleat that the relative amount of tull-length connexin32 to its fragment differed in the two populations of vesicles (Fig. 2B). The proportion of full-length connexin32 was greater in the sucrose-pc> meable vesicles (Fig. 2B, lanes 3 and 5), relative to its fragment, than it was in the sucrose-impermeable vesicles (Fig. 2B, lanes 4 and 6). Full-length monomer was almost totally absent from the sucrose-impermeable vesicles. Thus, sucrose-permeability selects for fulllength connexin32, and sucrose-impermeability selects
A
B
,,
TRIMER
D,MER
t
~
47kDa
