OmpA Protein of Escherichia coli Outer Membrane Occurs in Open

Vol. 269, No. 27, Issue of July 8, pp. 17981-17987, 1994 Printed in U.S.A.

JOURNAL OF B I O ~ C I CCHEMISTRY AL 0 1994 by The American Society for Biochemistry and Molecular Biology, Inc. THE

OmpA Protein of Escherichia coli Outer Membrane Occurs in Open and Closed Channel Forms* (Received for publication, March 14, 1994, and in revised form, May 2, 1994)

Etsuko SugawaraS and Hiroshi Nikaido From the Department of Molecular and Cell Biology, University of California, Berkeley, California 94720-3206

OmpAproteinof Escherichia coliouter membrane can produce diffusion channels when reconstituted into proteoliposomes (Sugawara, E., and Nikaido, H. (1992) J. Biol. Chem. 267, 2507-2511). The pore size is similar to that of the classical E. coli porins OmpF and OmpC, but the penetration rates of smallsolutes through theOmpA channel are about 60 times slower than that through the OmpF channel. Here we examined the possibility that only a small fraction of the OmpA molecules produces open channels. Unilamellar proteoliposomes were made so that each vesicle contained only a small number of OmpA molecules. These vesicles, containing 0.3 M urea within, were fractionated on a linear iso-osmolar density gradient madeof urea and sucrose. This resulted in the clear separation of vesicles not containing any open channel, staying on top of the gradient, from those containing at least one open channel, sedimentingclose to the bottom. Calculationusing Poisson distribution indi3% of the OmpA molecules cated that only between 2 and contained open channel. The open form is estimated to allow the diffusion of L-arabinose at a rate comparable with that through the OmpF porin channel. The open and closed states were relatively stable propertiesof the protein. Denaturation of open formOmpA and its subsequent renaturation converted it into a nonfunctionalor closed form, suggesting that the open andclosed forms represent two alternative conformers of this protein.

(Nikaido et al., 1991). OprF and OmpA show sequence homology, a n d immunologicalcross-reactivity has beenreported (Woodruff and Hancock, 1989). Two possible hypotheses exist on the slow penetration of solutes throughthe OmpA and OprF channels. 1) There may be much friction between the solute and the channel wall in these proteins, and this may slow down the penetrationof solutes. 2) Alternatively, only a small fraction of the protein population may contain the open channels,as was originally proposed for OprF (Nikaido and Hancock, 1986). We tried to distinguish these possibilities by using the technique of "transport-specific fractionation" of Goldin and Rhoden (1978). We show here the existence, within the OmpA protein population, of both open channel and closed channel forms. EXPERIMENTALPROCEDURES

Bacterial Strains and TheirCultivation-Escherichia coli HN705 (K12 Ailac-proAB) AompC ompF::Tn5 zei-298::TnIO supE rpsLl FtraB36' pro&+ lacls lacZAh415) (Sugawara and Nikaido, 19921, which .is deficient in both of the classical E. coli porins OmpF and OmpC, was used for the isolation of OmpA. Strain JF701 (Chai and Foulds, 1978)was used to purifyOmpF. These strains were kept frozen at -70 "C in 20% glycerol. Cells were grown in L broth at 37 "C, with aeration by rotary shaking. Purification of OmpA and OmpF Proteins-OmpA was purified as described previously (Sugawara and Nikaido, 1992), except that 10% glycerol was present in the elution buffer of the gel filtration column. This led to a better separation of OmpA from contaminating proteins. The purification of OmpF followedthe procedure of Nikaido and Rosenberg (1983).To remove salt, thefinal preparations were passed through OmpA, composed of 325 amino acid residues (Chen et al., a column of Sephadex G-25 (bed volume,5 ml), which was equilibrated 1980), is one of the most abundant proteinsin the outermem- with 0.1% lithium dodecyl sulfate, 10 mM Tris-C1, pH 7.5. Both OmpA and OmpF preparations were free from major contaminants, as judged brane ofEscherichiacoli. It is reported to stabilize the shape of the cell (Sonntaget al., 1978) and to functionas a phage recep- by the scanning of the silver-stained SDS-PAGE.' Trypsin Digestion-Proteoliposomes containing 50 pg of OmpA were tor (Datta et al., 1977; van Alphen et al., 1977) and as a meincubated with trypsin (20 pglml) for 45 min at 37 "C. Proteoliposomes diator in F-factor-dependent conjugation (Schweizer and disrupted by sonication in the presence of trypsin were incubated in a Henning, 1977). In addition, we showed that this protein also similar manner. Denatured OmpA prepared by boiling in 1%SDS and produces a porin-like diffusion channel (Sugawara and a control mixture without OmpAwere treated identically. Digestion was Nikaido, 1992). However, OmpAis different from the classical stopped by addition of 0.1 mM phenylmethylsulfonyl fluoride, and the porins, OmpFand OmpC, in at least two ways(Sugawara a n d samples were analyzed by SDS-PAGE. Reconstitution of OmpA and OmpF Proteins into Unilamellar Nikaido, 1992). (i) The flux of solutes through OmpAchannel is Vesicles-The reconstitution of these proteins into the unilamellar much slower, usually about 50-fold, than through the OmpF vesicles basically followed the procedure of Ambudkar and Maloney channel, in spite of the fact that the estimated sizeof the OmpA (1986). Acetone-washed egg phosphatidylcholine (50 mg), which was channel is similar to that of OmpF. (ii) The OmpA protein dissolved in 2:l (vlv) chloroform/methanol, was dried under a streamof appears to existas stable monomers when isolated, as well as nitrogen and kept in avacuum dessicator for at least 1h. The dried lipid i n the intact outer membrane. Interestingly, the outer mem- was suspended in 1 ml of 10 mM HEPES-NaOH, pH 7.0, and sonicated brane protein F (OprF) of Pseudomonas aeruginosa also i s mo- to clarity in abath-type sonicator. Typically,90 pl of sonicated lipid was mixed with 450pl of urea buffer 1344 mM urea, 10 mM KCl, 10 mM nomeric and produces a veryinefficientdiffusionchannel HEPES, pH 7.0, and 3 mM NaN,) containing known amounts of OmpF or OmpA, and octyl-6-D-glucoside wasadded to a final concentration of * This study was supported in part by Grant "09644 from the Na- 1.1%.After 30 min at 37 "C, the mixture was diluted into 14 ml of the tional Institutes of Health. The costs of publication of this article were same urea buffer at room temperature. Proteoliposomes were collected defrayed in part by the payment of page charges. This article must by centrifugation at 100,000 x g for 1 h at 4 "C and were suspended therefore be hereby marked "advertisement" in accordance with 18 carefully in 200 p1of urea buffer. U.S.C. Section 1734 solely to indicate this fact. $ To whom correspondence should be addressed: DeDt. of Molecular and Cell Biology, 229 Stanley Hall, University of California, Berkeley, The abbreviations used are: PAGE, polyacrylamide gel electrophoreCA 94720-3206. sis; S, siemens.

17981

17982

Closed Open and

Forms of E. coli OmpA Protein

Iso-osmolar Density Gradients-The gradient centrifugation was carried out according to Harris et al. (1989). In order to get maximum separation between the”open” and ”closed” vesicles, the gradient was made shallower, so that a linear gradient rangedfrom 0.3 M urea at the top to 0.15 M sucrose, 0.15 M urea at the bottom, both solutions being made up in 10mM KCI, 10 mM HEPES-NaOH, pH 7.013 mM NaN,. The gradient was made up in a 5-ml ultracentrifuge tube, the vesicle suspension (0.2 ml) was layeredon top of the gradient, and the tube was centrifuged at 300,000 x g for 16 h ina Beckman SW65 swinging bucket rotor at 10 “C. Samples were recovered from the bottom of the tube in 250-pl fractions, and portions were used for lipid determination (see below). OmpF or OmpA was detected by SDS-PAGE, followed by staining withCoomassie Brilliant Blueor silver (Heukeshoven and Dernick, 1988), or by immunoblotting with polyclonal antibodies against OmpF or OmpA(with an alkaline phosphatase-conjugated anti-rabbitIgG goat antibody (Sigma) as thesecond antibody). Vesicle Size-Proteoliposome suspension (0.5 ml) was applied to the column of Bio-Gel A-150m (0.95 x 20 cm) equilibrated with the urea buffer and eluted with the same buffer. “Nanosphere size standards” (Duke Scientific Corporation, Palo Alto, CA), 40, 96, and 156 nm in 0.08 diameter, were used for calibration of the column, which was equili156 86 41) brated and eluted with 1% Triton X-100 in bufferto prevent aggregation of these beads. E Liposome Swelling Assay-This was carried out essentially a s deC Q, scribed earlier (Nikaido et al., 1991). When proteoliposomes fractionco m ated by iso-osmolar gradient centrifugation were usedas the source of OmpA, the collected vesicles were first concentrated about 4-fold by 4m ultrafiltration t h r o u g h h i c o n (Beverly, M A ) PM-10 membrane. Most of the uredsucrose buffer was then removed by dialysis against 0.05% a 0 lithium dodecyl sulfate at 4”C, and portions of these samples were S m reconstituted into multilamellar proteoliposomes with 2.4 pmol of egg 0 phosphatidylcholine (Avanti Polar Lipids, Alabaster, AL) and 0.2 pmol L 0 of dicetylphosphate (Sigma). v, I) Recovery a n d DenaturationlRenuturation of OmpA-Gradient-fraca tionated OmpA-containing vesicles were solubilized with 1.5% octyl-& D-glucoside. The protein was separated from the lipids by chromatography ona 5-ml column of phenyl-Sepharose CL4B (Pharmacia Biotech 0 10 20 30 Inc.), which was equilibrated with 51 mM octyl-P-D-glucoside, 0.4M LiC1, 10 mM Tris-C1, pH 8.0, 5 mM EDTA, 3 mM NaN, and eluted with the Fraction same buffer. Protein was eluted earlier than the lipids. Pooled protein fraction was either used directlyfor reconstitution or divided into two FIG. 1. Characterization of egg phosphatidylcholine vesicles Vesicles were preportions for renaturation experiments. One portion was treated with 6 prepared by octyl-p-wglucose dilution method. pared as described under “Experimental Procedures.” The upper figure M guanidine hydrochloride, 0.1%2-mercaptoethanol. Forrefolding, guanidine hydrochloride wasremoved by dialysis against 2 litersof water shows an electron microscopic picture of typical vesicles, negatively after the additionof 0.1% SDS and 10-fold molar excessof lipopolysac- stained with 2% uranyl acetate on glow-discharged carbon film grids. The lower figure shows the gel filtration profiles of vesicles formed in charide from the same strain and then the dialysis was continued against 10 mM Tris-C1, pH 7.5. The other portion wasdialyzed against the presence (0)and absence(0)of OmpA protein on a Bio-Gel A-150m column (0.95 x 20 cm). Vesicles were detected by turbidity a t 589 nm. the same buffer, and this was used as control. Both samples were The column was calibrated with “nanosphere size standards,” 40, 96, reconstituted into unilamellar vesicles as described above. and 156 nm in diameter, and their positions of elution are marked at the Circular Dichroism Spectra-Circular dichroism spectra were retop of the figure. corded on a n Aviv model 62DS Circular Dichroism Spectrometer (Aviv Associates, Lakewood, NJ). All spectra were corrected for light scattering by subtracting spectra of samples not containingOmpA. eter close to 100 nm (Fig. 1).These results confirm those of Other Methods-Total lipid phosphatewasdetermined by the Chen and Wilson (1984). method ofAmes (1966). Protein was determined by using bicinchoninic The N-terminal portion of OmpAis thought to span the memacid procedure(Smithet al., 1985),using the BCA Protein Assay Reagent from Pierce Chemical Co., with bovine serum albuminas stand- brane, whereas its C-terminal portion is assumed to exist as a ard. SDS-PAGE was carried out as described by Lugtenberg et al. periplasmic domain (Vogel and Jahnig, 1986). Unfolded OmpA (1975). The samplesfor SDS-PAGEwere heated in the sample buffer for is completely digested by trypsin. When the unfolded OmpA is 5 min eitherat 50or 95 “C. Lipopolysaccharide was preparedaccording refolded and incorporated into lipid bilayers, the protein is to Galanos etal. (1969). digested only down to themembrane-protected fragment of M, RESULTS

Reconstitution of OmpA intoUnilamellar Vesicles-Estimation of the fraction of the OmpA molecules in theopen form required theproduction of proteoliposomes of uniform size and composition. We used the octyl-P-D-glucoside dilution method (see“Experimental Procedures”). Under our conditions, the centrifugal pellet contained 84 and 80%, respectively, of the OmpA protein and thephospholipids initially added. Gel filtration (see “Experimental Procedures”) showed that thevesicles were monodisperse (Fig. 11, with the mean diameter of 1132 38 and 109 * 32 nm for OmpA-containing and noncontaining vesicles, respectively. Electron microscopy also confirmed that these were unilamellar vesicles with relatively uniform diam-

24,000 (Dornmair et al., 1990; Surrey and Jahnig,19921, similar in behavior to the nativeOmpA (Schweizer et al., 1978). In order to assess if the OmpA protein was inserted correctly into the membranes of our vesicles, the vesicles were treated with trypsin. As shown in Fig. 2, some of the OmpA molecules were resistant to trypsin, and others were converted into fragments of M,24,000. (The bandjust underneath thefragment is trypsin itself.) It appears that thetrypsin-digested fraction corresponds t o the OmpAinserted withits C-terminal domain facing outside, whereas the resistant fraction represents OmpA inserted in the opposite orientation so that the C-terminal domain is protected inside the vesicles. This interpretation is supported by the observation that more of the OmpA became degraded when the vesicles were sonicated in the presence of

Closed and Open Forms of E. coli OmpA Protein

1

2

3

4

5

6

I

17983

, , , , , , , , . , ,

e L

C

aJ

z8

10

a

""-

0

10

Fraction

"

L

C

aJ 0 L

aJ

n FIG.2. Treatment of OmpA-containing vesicles with trypsin. Vesicles were made by octyl-P-D-glucoside dilution, and portions each containing 50 pg of OmpA were incubated with or without trypsin (20 pg/ml) for 45 min at 37 "C. Lane 1, M, standards (see below); lune 2, vesicles before incubation; lune 3, vesicles incubated without trypsin; lune 4, vesicles incubated with trypsin; lune 5,vesicles lysed by sonication in the presence of trypsin; lune 6, control vesicles not containing OmpA, incubated with trypsin. M,standards were, from the top, phosphorylase b (92,500),bovine serum albumin (66,200), ovalbumin (45,000), carbonic anhydrase (31,000), soybean trypsin inhibitor (21,5001, and hen egg white lysozyme (14,400).

trypsin (compare the intensityof OmpA bands inlanes 4 and 5 of Fig. 2). Iso-osmolar Density Gradient Works as Expected-Control experiments were first carried out in order to ascertain that iso-osmolar gradient of Harris et al. (1989) indeed separates vesicles without any channels from those containingopen channels. A linear iso-osmolar density gradient formed of urea and sucrose buffershad a range of specific gravity from 1.013 at the top to 1.031 at the bottom. Vesicles are made in a solution containing only urea and arelayered on top of the gradient.If there are open channels in thevesicle membrane, sucrose will enter through the channel and urea will leave, as the vesicles sediment through the gradient. In contrast, vesicles with no open channels will remain at the top. When vesicles made without any protein were layered onto the gradient,most of the vesicles indeed stayed at the top after 16 h of centrifugation (Fig.3A 1. Vesicles made with OmpFwere used as a positive control, because OmpF channel is open under physiological conditions (Nikaido and Vaara, 1985). Much of these vesicles (70% by phospholipid assay) did indeed sediment to areasclose to thebottom (Fig. 3B). However, 30% remained on top. This was surprising because each vesicle should have contained 10 OmpF trimers on average. Immunoblotting after SDS-PAGE, however, showed that the vesicles remaining on top weredevoid of any OmpF protein,the protein being present exclusively in thevesicle sedimenting close to thebottom (Fig. 3B). Because OmpF has a strong lateral interaction(Nikaido and Vaara, 1985), aggregates of trimers, rather than individual trimers, apparently became incorporated into vesicles. These results indicated that the linear iso-osmolar density gradient may be used to separateopen channel porins from proteins not containing open channels. Separation of Open Form OmpA from Closed Form OmpAOne obvious explanation of the slow solute penetration rates through OmpA channel is that only a small fraction of the protein contains an open channel. Similar fractionation through iso-osmotic gradient wastherefore performed with proteoliposomes containing OmpA. Vesicles were made from 0.5 nmol (17 pg) of OmpA protein and 6.4 pmol of phospholipids. The molar ratio between OmpA and phospholipids was therefore 1:12,800. Each vesicle has a diameter of about 110 nm

1

10

20

Fraction

FIG.3. Density gradient centrifugation of control vesicles. Vesicles were centrifuged on a linear iso-osmolar density gradient of 0.3 M urea to 0.15 M urea, 0.15 M sucrose for 16 h at 300,000 x g).Twenty fractions were collected from the bottom. Lipid distribution was monitored by assaying for phospholipid phosphorus. When OmpF-containing vesicles were used, protein distribution was determined by densitometric scanning of the OmpF band after SDS-PAGE separation of proteins followed by Coomassie Blue staining. A, vesicles formed without the addition of any protein; B, vesicles formed in the presence of 50 pg of OmpF. Filled-in burs show the amount of lipids and empty burs that of OmpF protein in each fraction, both as percent of the total found in all fractions.

(see above) and, therefore, a surface area of 38,000 nm2. Considering that each phospholipid molecule occupies an area of about 0.6 nm', each vesicle should contain 127,000 lipid molecules. Thus the ratioused corresponds to about 10 molecules of OmpA monomer per vesicle. When these vesicles were subjected to thegradient centrifugation, about 78% of total vesicles, as determined by the phospholipid assay, stayed at the top of the gradient andonly 22% moved to thebottom (Fig. 4). Protein distribution through the gradientappeared to parallel roughly that of lipid (Fig. 4), and thus the separation of the two vesicle populations was notcaused by the uneven incorporation of the OmpA protein. We considered the possibility that all OmpA channels were open and that the sedimentation or lack of sedimentation of vesicles was determined by slow, chance influx of sucrose. If this were the case, then the OmpA molecules in the top fractions should produce a similar permeability as theOmpA molecules in thebottom fraction. This was testedby collecting both the sedimented and unsedimented vesicles and by reconstituting theOmpA proteins in themwith a large excess of additional phospholipids into multilamellar proteoliposomes. These preparations were then used for the quantitative determination of average permeability through the OmpA protein, by measuring the rates of their osmotic swelling in aniso-osmotic solution of L-arabinose (see "Experimental Procedure"). Fig. 5 shows that the OmpA protein from vesicles that remained on top of the gradient did not show any pore-forming activity, whereas OmpA from the sedimented vesicles allowed the diffusion of L-arabinose at a significant rate, indicating that there are indeed two forms of OmpA, which are quite different in terms of their pore-forming activity. The osmotic swelling rate, which indicates the permeation rate of L-arabinose, was 0.054 OD,Jmin/pgof proteinwith OmpAfrom the sedimented vesicles, a rate that is much higher than the rate of 0.006 OD,Jmin/pg of protein, obtained with unfractionated OmpA protein (see Fig. 3 of Sugawara andNikaido, 1992). Clearly the

17984

Closed and Open Forms of E. coli OmpA Protein 40

A

n

Fraction

Fraction

FIG.4. Density gradient centrifugation of OmpA-containing vesicles. Vesicles formed in the presence of 17 pgOmpA were layered onto the uredsucrose gradient and centrifuged as described in Fig. 3. Protein distribution was determined by densitometric scanning of silver-stained SDS-polyacrylamide gels of the fractions. Fitted-in burs show the amountof lipids andempty bars that of OmpA protein in each fraction, both as percent of the total found in all fractions.

.

60 i

-

.-E C

0 0

m

c .--

Fraction

0

1

2

3

4

5

Protein (pg)

0)

3

rn

FIG.5. Permeation rates of L-arabinose through two OmpA fractions. OmpA-containing proteoliposomes werefractionated as shown in Fig. 4, and vesicles were collected from four peak fractionsat the top and the bottom. The samples were concentrated about 4-fold by ultrafiltration usingAmicon concentrator, and mostof the uredsucrose buffer was removed by dialysis against 0.05% lithium dodecyl sulfate. Different amountsof the sedimentedvesicles (0)and vesicles from the top of the gradient (0) were reconstituted with 2.4 pmol of egg phosphatidylcholine and 0.2 pmol of dicetylphosphate into multilamellar vesicles. Then 17 p1 of the proteoliposome suspension were diluted into 0.6 ml of iso-osmotic L-arabinose, and therate of influx of L-arabinose was measured as the initial ratesof osmotic swelling of the vesicles, as indicated by the decrease inOD,,,.

FIG.6. Fractionation of vesicles reconstituted from the open and closed form OmpA. Initial gradient centrifugation of OmpAcontaining vesicles were performedas described in thelegend to Fig.3, except that the sedimentation was carried out in eight tubes, separating a total of 136 pg of protein into nonsedimented and sedimented fractions. These two fractions were collected, and OmpA protein was isolated by solubilization with octyl-P-D-glucosideand phenyl-Sepharose chromatography as described under “Experimental Procedure.” These two OmpA preparations, onefrom the sedimentedvesicles (A), and the other from the vesicles that stayed on top (B), were then reconstituted with 6.4 pmol each of phospholipids, and the vesicles, each containing a n average of 27 (A) or 10(B)OmpA molecules, were then separated by density gradient centrifugation. In A, each of the vesicles sedimented in the first run was expected to contain at least one open form OmpA among 10 molecules incorporated. Thus each vesicle should have contained an average of 27 x 1/10 or 2.7 open channels, and Poisson distribution predicts that 95% of all vesicles should contain at least one open channel.Most of the vesicles indeed sedimented into the gradient.

tains two rather stable forms, one producing open channels in lipid bilayers and the other incapable of doing so. Assuming that OmpA protein will be randomly distributed among a large numberof vesicles, we can usePoisson statistics to estimate the fraction of open channel forms among the OmpA population. When the average numberof open form OmpA per vesicle is n, the fraction of vesicles containing no open form OmpA is predicted to be exp (-n). With the experiment of Fig. sedimented vesicles are much enriched in the open channel 4, the fraction of vesicles remaining attop was 0.78, and thus n is 0.248. The total number of OmpA per vesicle was 10 as form of OmpA. mentioned earlier.Thusthe fraction of open form protein The experiment of Fig. 5, however, could not rule out the population of OmpA was 0.248/10 or 2.5%. We possibility that OmpA occurs in a single form, whichneverthe- among the total less could take either an open or closed conformation when it have carried out similar experiments with 6, 5, and 3 OmpA and becomes inserted into a lipid bilayer. Thus if the initial inser- molecules per vesicle, and in these experiments 84,88, 94% tion of OmpA into unilamellar vesicles put it into an open of the vesicles stayed on top of the gradient,respectively. From conformation, that conformation would have been maintained these data, we can estimate that 2.9, 2.6, and 2.1%, respecin the multilamellarvesicles, as the OmpA probably remained tively, of the total OmpA population was in the open form. Are the Open and Closed Forms of OmpA ZnterconuertibZe?associated with theoriginal neighboring lipids throughout the second reconstitution step.We therefore carried out the second The open and closed forms of OmpA may differ in the primary reconstitution with OmpA dissociated from the surrounding structure, for example one of them carrying a covalent modifitwo conformers of the lipids of the unilamellar vesicles. The sedimented andunsedi- cation. Alternatively,they may represent mented vesicles were solubilized with octyl-p-o-glucoside, and identical protein. If the latter is true, thenwe may be able to by denaturation followed by lipids were removed by hydrophobic chromatography (see“Ex- convert oneconformer to the other perimental Procedure”). The essentiallylipid-free OmpAprepa- renaturation. We therefore denatured the closed form OmpA, rationswerethenreconstitutedinto new proteoliposomes, isolated from the nonsedimenting vesicles, in 6 M guanidine denaturwhich were then subjected to density gradient centrifugation. hydrochloride, 0.1% P-mercaptoethanol.Complete Vesicles reconstituted with OmpA derived from the vesicles ation was confirmed by the circulardichroism spectrum which that remained on top in the first centrifugation all stayedon showed no evidence of secondary structure (Fig. 7). When guatop, and vesicles reconstituted with OmpA derived from the nidine hydrochloride and P-mercaptoethanol were removed by sedimented vesicles allsedimented t o the bottom(Fig. 6). dialysis, refolding of the OmpA occurred, as evidenced by the These results suggest that the total OmpA preparation con- circular dichroism spectrum showing P-pleatedsheet structure.


1

- 1 Y

5

10

15

Fraction

x

m

17985

1

200

220

260

240

Wavelength (nm)

-

"

1

.,

2

3

4

5

FIG.8. Density gradient centrifugation of proteoliposomes reconstituted with closed form OmpAafter denaturation-renaturation. Closed form OmpA was isolated from the vesicle fraction that remained on top in a linear density gradient similar t o Fig. 4. A portion of it was denatured and renatured asdescribed under "Experimental Procedures." The protein was reconstituted into proteoliposomes either before ( A ) or after ( B ) denaturation-renaturation at the ratio of 25 OmpA molecules per vesicle, and these vesicles were separated in a linear density gradientof uredsucrose, a s described under "Experimental Procedure."

FIG.7. D e n a t u r a t i o n and renaturation of OmpA. OmpA was unfolded in 6 M guanidine hydrochloride, 0.1% 2-mercaptoethanol and was renatured by dialysis as described under "Experimental Procedure." The top panel (A) shows the circular dichroism spectra of OmpA at differentstages: -, nativeOmpA -- - -, OmpA unfolded in 6 M guanidinehydrochlorideand 0.1% 2-mercaptoethanol;and , refolded OmpA. Spectra wererecorded a t a protein concentration of 0.15 mg/ml. B , SDS-PAGE analysis of OmpA a t different stages. Samples weresolubilized in the samplebuffer a t 50 "C, so that the native conformers of OmpA would migrate at the position of M, 31,000 protein. Denatured OmpA migrates more slowly, closer to the position expected from its trueM , of 35,000. Lane 1,M , standards (see Fig. 2); lane 2, native OmpA lune 3, OmpA unfolded in 6 M guanidine hydrochloride and 0.1% 2-mercaptoethanol and refolded by dialysis; lune 4, native OmpA recovered from reconstituted vesicles; lune 5, protein recovered from vesicles reconstitutedwith unfolded-refolded OmpA. It is possible that the small fraction of denatured OmpA seen in lune 3 was not incorporated intovesicle membrane. " "

"he completeness of the refolding process is evident from the of magnitude of ellipticity at 215 nm, which was similar to that the native OmpA (Fig. 7). Most of the renatured OmpA also migrated at the position of native form in SDS-PAGE, although a small fraction migrated as the denaturedform (Fig. 7). When this refolded closed form OmpA was reconstituted into proteoliposomes, they failed to sediment into the density gradient, a result suggesting that the closed form could not be converted into the open form by refolding, a t least under the conditions used (Fig. 8). When a similar refolding experiment was camed out with open form OmpA, isolated from the sedimented vesicles, most of the open form OmpA was found to havebeen converted into nonfunctional (closed) form(s), as the proteoliposomes remained on top of the gradient (Fig. 9). If we assume that the refolding produces a native conformation, as indicated by the circular dichroism spectrum and the mobility in SDS-PAGE, this result suggests that open and closed forms are different conformers and that theopen conformers were converted into closed ones by the refolding process. However, at present we cannot exclude the alternativepossibility that renaturationdid

Fraction 401

1

-

I

5

10

15

Fraction FIG.9.Density gradient centrifugation of proteoliposomesreconstituted with open form-enriched OmpAafter denaturationrenaturation. OmpA enriched in open form molecules was isolated from the sedimentedvesicle fraction in a linear density gradient similar to Fig. 4. A portion of it was denatured and renatured as described under "Experimental Procedures." The protein was reconstituted into proteoliposomes either before (A) or after ( B ) denaturation-renaturation at the ratio of 22 OmpA molecules per vesicle, and these vesicles were separated in a linear density gradient of uredsucrose, asdescribed under "Experimental Procedure."

not produce a completely native conformation, but the differences were too subtle to detect by the methods used. DISCUSSION

OmpA protein of E. coli produces a nonspecific diffusion channel with the estimated diameter of about 1nm, similar to that of the classical porin channels (Sugawara and Nikaido, 1992). We attempted to find why the permeation of solutes

17986

Closed Open and

Forms of E. coli OmpA Protein

through OmpA channel occurs about 50 times more slowly than the fractionated OmpA proteins in 6 M guanidine hydrochloride, that through theclassical porin channels. For this purpose, we 0.1% 2-mercaptoethanol. The denatured protein migrated as used the technique of transport-specific density gradient frac- expected for a protein of M , 36,000 in SDS-PAGE (Fig. 7), and tionation, which has been used successfully for the isolation of the absence of secondary structure was confirmed by circular functional formsof several membrane transport proteins (Hessdichroism (Fig. 7). These preparations were then renatured in and Andrews, 1977; Goldin and Rhoden, 1978; Papazian et al., the presence of lipopolysaccharide (Schweizer et al., 1978) by 1979; Goldin et al., 1980). removing guanidinium hydrochloride by dialysis. The products We used essentially the conditions of Harris et al. (1989) for appeared t o have become fully renatured, according to several iso-osmotic gradient centrifugation and could confirm that the criteria. ( a ) They migrated mostly as a 31,000-dalton band in method separated cleanly vesicles containingOmpFporin SDS-PAGE. ( b )Their circular dichroism spectra showed extenchannels from those not containingany proteins (Fig. 3). When sive P-sheet structure, with the value of molar ellipticity close the vesicles, each containing 10 OmpA molecules on average, t o that of the nativeOmpA protein (Fig. 7).( c )Treatment with were separated on this gradient,only about 30%of the vesicles trypsin produced a protectedfragment ofM, 24,000as with the apparently contained open channels, which allowed the influx native OmpA (Schweizer et al., 1978). When the OmpA fraction of sucrose into the vesicles and resulted in the downward mi- enriched for the open form was used for denaturation-renaturgration of vesicles into the gradient (Fig. 4). This separation ation, the open form disappeared from the sample(Fig. 91, was not due to the statistically infrequent random influx of suggesting that theopen form can be converted into the closed sucrose molecules into a homogeneous population of vesicles, form without any covalent modification. However, we have not because when the sedimented vesicles and vesicles that stayed been able to achieve the converse transition, from closed to on top were collected and multilamellar proteoliposomes were open form (Fig. 8). reconstituted from these vesicles and excess phospholipids, the If the open and closed forms of OmpA are indeed two consedimented vesicles produced a far higher pore-forming activ- formers of the same protein, what differences in thefolding of ity than thevesicles remaining on top (Fig.5 ) . Thus thereexist the protein could produce these two forms? The proposed foldtwo relatively stable populations of OmpA, and Poisson statis- ing pattern of OmpA includes the N-terminal domain containCtics showed that between 2 and 3% of the OmpA molecules ing eight transmembrane antiparallel p-strands and the terminal periplasmic domain (Bremer et al., 1982; Morona et contained the open channel. A trivial explanation of these data would be that 97-98% of al., 1984; Vogel andJahnig, 1986). Clearly the majority of the OmpA protein was damaged during purification, so that OmpA molecules must fold in thisway, as much evidence exists their pore-forming activity was inactivated. Our result, how- t o support this model. These include the trypsin cleavage of native OmpA (Schweizer et al., 1978) and the observation that ever, is clearly against such an interpretation, as the outer membrane of intact cells containing OmpA but lacking in clas- all of the mutations thatdiminish the ability of OmpA to serve sical porins shows a very low permeability, corresponding t o as a receptor for bacteriophages occur in thefour external loops less than 5% of that in the wild-type strain producing the predicted by the model (Morona et al.,1984).Although an eightclassical porins (Bavoil et al., 1977). Thus OmpA confers only stranded p-barrel would have a backbone-to-backbone diamvery low permeability even in intact cells. Another trivial ex- eter of 1.1 nm, this barrel in the OmpA protein is unlikely to planation is that the "open channel form" corresponds to con- function as an efficient, large channel, as the model shows taminating classical porinsin the preparation. However, this is numerous amino acid residues withlong side chains protruding interior. Thus theopen form, a minority popuunlikely because ( a )SDS-PAGE analysis of protein in thesedi- into the channel mented vesicles did not reveal any other protein band except lation among OmpA, is likely to fold differently. One possibility OmpA (result not shown),and ( b )a mutation in the ompA gene is that the protein traverses the membrane more than eight was shown to abolish most of the pore-forming activity recov- times. Indeed, the computer program that predicts the folding ered from the outer membrane of our ompF- o m p C mutant patterns of outer membrane proteins (Schirmer and Cowan, 1993) suggests that OmpA may traverse the membrane more strain (Sugawara and Nikaido, 1992). The OmpA protein in the sedimented vesicles produced a than tentimes, as found for other classical porins. This hypothliposome swelling rate (which corresponds to the penetration esis isalso madeattractive by the suggestion that P. aeruginosa rate of L-arabinose) of 0.054 OD,,, min" pg-' (Fig. 5 ) . Although OprF, a homolog of E. coli OmpA, appears to traverse thememeach vesicle contained 10 OmpA molecules, it is likely that in brane many times, with its C-terminaldomain also embedded most cases only one OmpA among these 10 molecules was of the in the membrane (Finnen et al., 1992). Another possibility is open form, because only 2-3% of the totalOmpA contained the that it may take more than one molecule of OmpA to produce open channel. Thus the specific permeability of the pure open open channels. This hypothesis is consistent with our denaturform OmpA is estimated tobe about 10-fold higher, i.e. around ation-renaturation data, because the formation of oligomeric 0.5 OD,,, min" pg". This specific activity is close to the value, conformer would not be favored under our conditions of rena0.4 OD,,, min" pg" (Nikaido and Rosenberg, 19811,of the turation, where the protein concentration was quite low, typiOmpF porin, whose channel is completely open (Cowan et al., cally 50-100 pg/ml. Further studyis necessary to establish the 1992). Thisobservationalso rulesoutthe possibility that differences between the open and closed forms of OmpA. OprF, the major porin of P. aeruginosa OprF (Nikaido et al., OmpA confers permeability t o reconstituted vesicles simply by destabilizing the membrane,because such a mechanism is un- 1991), is a homolog of OmpA, and itspore-forming functions are reminiscent of OmpA, as the rate of diffusion of solutes is much likely to produce the permeability of this magnitude. We have attempted to obtain evidence for interconversion slower than through the channelsof classical porins (Angus et between the presumptive open and closed conformers. OmpA is al., 1982;Yoshimura and Nikaido, 1982;Yoshimura et a l . , 1983). known to be denatured in the presence of 8 M urea, or by boiling An early black lipid film study of OprF showed a singlechannel in SDS, and such denatured form migrates in SDS-PAGE with conductance of 5.6 nS in1M KC1 (Benz and Hancock, 1981). This an apparent M , of 36,000 (Schweizer et al., 1978). The dena- value is quite high in comparison with the single channel contured form could then be renatured, and the refolded OmpA ductance of E. coli OmpF porin, 0.7 nS. (Early literature remigrates inSDS-PAGE with an apparent M , of 31,000, just like ported the single channel conductance as 2 nS (Benz et al., 1978), the native OmpA (Schweizer et al., 1978). We have denatured but most likely this was created by the insertion of a trimeric


17987

Jansonius, J. N., and Rosenhusch, J. P. (1992) Nature 358, 727-733 porin containing three channels). A more recent study found Datta, D. B., Arden, B., and Henning, U. (1977) J. Bacteriol. 131, 821-829 that these largeconductance steps are infrequent, and more the Dornmair, K., Kiefer, H., and Jahnig, F. (1990) J. Biol. Chern. 265, 18907-18911 frequent steps corresponded t o 0.36 nS (Woodruff et al., 1986). Finnen, R. L., Martin, N. L., Siehnel, R. J., Woodruff, W. A., Rosok, M., and Hancock, R. E. W. (1992) J . Bacteriol. 174,49774985 Superficially these datalook similar to our data on OmpA, preC., Luderitz, O., and Westphal, 0.(1969) Eur. J . Biochem. 9, 245-249 sented inthis study, and the “0.36 n S channel maybe thought Galanos, Goldin, S.M., and Rhoden, V. (1978) J . Biol. Chern. 253,2575-2583 to correspond to theclosed form. However, it is not clear whether Goldin, S. M., Rhoden, V., and Hess, E. J. (1980) Proc. Natl. Acad. Sci.U . S. A. 77, 6884-6888 this interpretation is correct. If we consider that the hydrated A. L., Walter, A., and Zimmerberg, J. (1989) J. Mernbr. Biol. 109,243-250 K+ and C1- ions have a diameter of 0.36 nm (Robinson and Harris, Heukeshoven, J., and Dernick, R. (1988) Electrophoresis 9, 28-32 Stokes, 19551,the Renkin formula, which was shown to apply to Hess, G. P., and Andrews, J. P. (1977) Proc. Nntl. Acad. Sci. U. S. A. 74, 482486 the diffusion of nonelectrolytes through porin channels (Nikaido Lugtenberg, B., Meijers, J., Peter, R., van der Hoek, R., and van Alphen,L. (1975) FEBS Lett. 58,254-259 and Rosenberg, 19811, shows that the 0.36-nS channel should Morona, R., Klose, M., and Henning, U.(19841 J . Bacteriol. 159, 570-578 have a diameter of about 0.9 nm, incomparison with the OmpF Nikaido, H., and Hancock, R. E. W. (1986) in The Bacteria, Vol. X . The Biology of Pseudomonads (Sokatch, J . R., and Ornston,L. N., eds) pp. 145-193, Academic porin channel witha diameter of 1nm (Cowan et al., 1992). Thus Press, Orlando, FL these are hardly “closed channels, and low the permeability of Nikaido, H., and Rosenherg, E. Y. (1981) J. Gen. Physiol. 77, 121-135 Nikaido, H., and Rosenherg, E. Y. (1983) J. Bncteriol. 153, 241-252 small solutes, suchL-arabinose, as through the OprF porin channel (Yoshimura et al., 1983; Nikaido et al., 1991) cannot be ex- Nikaido, H., and Vaara, M. (1985) Microbiol. Reu. 49, 1-32 Nikaido, H., Nikaido. K., and Harayama. S. (1991) J. Biol. Chem. 266, 770-779 plained by this hypothesis. Studies of OprF porin, using a simi- Papazian, D., Rahamimoff, H., and Goldin, S. M. (1979) Proc. Natl. Acad. Sci. U. S. A . 76, 3708-3712 lar iso-osmotic gradient centrifugationprocedure, are currently Robinson, R. A,, and Stokes, R. H. (1955) Electrolyte Solutions, p. 247, Academic in progress in our laboratory. Press, New York

REFERENCES Amhudkar. S. V., and Maloney, P. C. (1986) Methods Enzymol. 125, 558-563 Ames, B. N. (1966) Methods Enzyrnol. 8, 115-117 A n g u s , B. L., Carey, A. M., Caron, D. A., Kropinski, A. M. B., and Hancock,R. E. W. (1982) Antimicrob. Agents Chemother. 21, 299-309 Bavoil, P., Nikaido, H., and von Meyenhurg, K. (1977) Mol. & Gen. Genet. 158, 23-33

Benz, R., and Hancock, R. E. W. (1981) Biochim. Biophys. Acta 646,298-308 Benz, R., Janko, K., Boos, W., and Lauger, P. (1978) Biochim. Biophys. Acta 511, 305-319

Bremer, E., Cole, S. T., Hindennach, I., Henning, U., Beck, E., Kurz, C., and Schaller, H. (1982) Eur. J. Biochern. 122,223-231 Chai, T.-J., and Foulds, J. (1978) J . Bacteriol. 135, 16P170 Chen, C., and Wilson, T. H. (1984) J . Biol. Chem. 259, 10150-10158 Chen, R., Schmidmayr, W., Kramer, C., Cohen-Schmeisser, U., and Henning, U. (1980) Proc. Nntl. Acad. Sci. U. S.A . 77, 45924596 Cowan, S. W., Schinmer, T., Rummel, G., Steiert, M., Ghosh, R., Pauptit, R. A,,

Schinmer, T., and Cowan, S. W. (1993) Protein Sci. 2, 1361-1363 Schweizer, M., and Henning, U. (1977) J . Bacteriol. 129, 1651-1652 Schweizer, M., Hindennach, I., Garten, W., and Henning, U. (1978) Eur. J . Biochem. 82, 211-217 Smith, P. K., Krohn, R.I., Hermanson, T., Mallia, A. K., Gartner, F. H., Provenzano, M. D., Fujimoto, E. K., Goeke, N. M., Olson, B. J., andKlenk, D. C . (1985)Anal. Biochem. 150, 76-85 Sonntag, I., Schwarz, H., Hirota, Y., and Henning, U. (1978) J . Bacteriol. 136, 280-285

Sugawara, E., and Nikaido, H. (1992) J . Biol. Chem. 267, 2507-2511 Surrey, T., and Jahnig, F. (1992) Proc. Natl. Acad. Sci. U. S. A . 89, 7457-7461 van Alphen, L., Havekes, L., and Lugtenherg, B. (1977) FEBS Lett. 75, 285-290 Vogel, H., and Jahnig, F. (1986) J . Mol. Biol. 190, 191-199 Woodruff, W.A., and Hancock, R. E. W. (1989) J. Bacteriol. 171, 3304-3309 Woodruff, W. A., Parr, T. R. J., Hancock, R. E. W., Hanne, L. F., Nicas, T. I., and Iglewski, B. H. (1986) J . Bacteriol. 167,473479 Yoshimura, F., and Nikaido, H. (19821J . Bacteriol. 152, 636-642 Yoshimura, F., Zalman, L. S.,and Nikaido, H.(1983) J . Biol. Chem. 258,2308-2314