Topology of PhoE porin: the 'eyelet' region

Molecular Microbiology (1993) 7(1), 131-140 Topology of PhoE porin: the 'eyelet' region Marlies Struyve,^ Jan Visser,^ Henri6tte Adriaanse,^ Roland Benz^ and Jan Tommassen^'^* ^Department of Molecular Ceil Biology and ^Institute of Biomembranes, University of Utrecht, Padualaan 8, 3584 CH Utrecht, The Netherlands. ^Lehrstuhl fur Bioteohnoiogie, Biozentrum der Universtat Wurzburg, Am Hubland, D-8700 Wurzburg, Germany. Summary A model for the topology of the PhoE porin has been proposed according to which the polypeptide traverses the outer membrane sixteen times mostly as amphipathic [3-sheets, thereby exposing eight loops at the cell surface. Until now, no evidence has been obtained for the surface exposure of the third loop. Recently, the structure of porin of Rhodobacter capsulatus has been determined. The proposed model of PhoE is very similar to the structure of the R capsulatus porin, which has an 'eyelet' region, extending into the interior of the pore. The proposed third external loop of PhoE might form a similar 'eyelet' region. To determine the location of the predicted third external loop of PhoE, multiple copies of an oligonucleotide linker encoding an antigenic determinant of VP1 protein of foot-and-mouth disease virus (FMDV) were inserted. All hybrid proteins were properly inserted in the outer membrane. The monoclonal antibody MA11, directed against the linear FMDV epitope, was able to bind only to intact cells expressing a hybrid PhoE protein with at least three copies of the FMDV epitope present. Antibiotic sensitivity tests and singlechannel conductance measurements revealed that the insertions influenced the channel size. These results are consistent with a location of the third loop of PhoE within the pore channel. Introduction Proteins with single or multiple transmembrane a-helices are found in the plasma membrane of both prokaryotes and eukaryotes, whereas the p-type structures seem to be associated with the outer membranes of bacteria and Received 4 August, 1992; revised and accepted 8 September 1992. *For correspondence. Tel. (030) 532999; Fax (030) 513655. mitochondria. Porins belong to the p-type structure (Kleffel et ai, 1985; Vogel and Jahnig, 1986) and constitute a major class of integral membrane proteins in the outer membrane of Gram-negative bacteria. Despite the fact that they are integral membrane proteins, they lack segments of hydrophobic amino acids long enough to span the membrane. The porins are able to function in a lipid environment by making barrels with an apolar external surface. To obtain more information about the molecular organization of these proteins, we use the phosphatelimitation-inducibte PhoE porin of Escheriohia coti K-12 as a model. The native structure of PhoE is a trimer. A topology model for PhoE protein has been proposed in which the potypeptide chain of each monomer traverses the outer membrane 16 times, mostly as amphipathic t3-strands, thereby exposing eight regions at the cell surface (van der Ley et ai, 1986; Tommassen, 1988). These exposed regions correspond with the hydrophilic peaks in the hydrophilicity profile of PhoE and are most variable when the PhoE sequence is compared with those of the related E coli porins OmpF and OmpC or to those of the PhoE proteins of other Enterobacteriaceae. Until now, evidence has been obtained for the surface exposure of seven of these regions. By using genetic approaches, including the isolation of point mutants (Korteland ef ai. 1985; van der Ley et ai, 1986), deletion (Agterberg ef a/., 1989) or insertion (Agterberg etai, 1987; Bosch and Tommassen, 1987; Bosch et ai, 1989) mutagenesis and the construction of hybrid porins in which parts of PhoE were replaced by the corresponding parts of OmpF or OmpC fTommassen etai, 1984; 1985; van der Ley etai, 1987), amino acid residues or PhoE segments were identified that are part of the receptor-site for the PhoE-specific phage TC45 or of the epitopes recognized by PhoE-specific monoclonal antibodies (mAbs). Since the phage-receptor site and the epitopes are suriace-exposed, the identified amino acids and PhoE segments must be surface-exposed. However, until now, no support was found for surface exposure of the third loop. Of all the postulated exposed loops, this one is the least variable and also its hydrophilicity is less pronounced (Tommassen, 1988). Weiss ef al. (1991a.b) have recently determined the three-dimensional structure of the porin from Rhodobacter capsulatus by X-ray diffraction analysis of porin crystals. Like in the PhoE model, each monomer of this 132 M.Struyve eta\. Third cell surface-eHposed region: 112 lie gly gly asp ser sec ala gin GGT GGC GAT TCC TCG GCG CAG CCA CCG CTA AGG AGC CGC GTC Site-directed mutagenesis gly gly gly ser ser ala gin GGT GGC GGA TCC TCG GCG CAG CCA CCG CCT aaG A G C C G C G T C Linker insertion pMS38 144 4150 gly qly aer gly M p l*u ^ly sar l«u ala qly aec aer GGC ^ A TCT tO.» OAT TTA COA TCT TTA GCT GJ3A TCC TCG CCG CCT A i ^ CCA. CTA AAT CCT AOA AAT COA CCT A3S AGC BamHI Fig. 1. Sequence of the relevant pan of the postulated third external loop of PhoE prolein and the FMDV epitope. Piasmid pMS37 was created after PCR-direcled site-specitic mutagenesis of pJP12 according to the Experimental procedures wilh the mismatch oligonucleotide dGCGCCGAGGATCCGCCACCAAA- A 205bp Mlu\~Cla\ fragment containing the mutation was subcloned in a W/ul-C/al-digested PhoE expression vector pJP29 yielding a piasmid, pMS37, of 5.2kb. The mutant piasmid contained only the desired substitutions. The numbers correspond to the amino acid positions in the wild-lype PhoE protein. In the unique BamHI restriction site of pMS37 two complementary 27-mer oligonucleotides, coding tor ammo acids 144 to 150 of VP1 protein ot FMDV, were inserted. A piasmid witfi a single copy ot the linker inserted was designated pMS38. A new SamHI site appears al one end of the insertion. The BamHI sites are underlined and in pMS38 the VP1 epitope is shown in bold and the nucleotides from Ihe vector are boxed. protein contains 16 antiparallel p-strands which form a 3-barrel. The upper barrel rim contains irregularly folded chain segments; the largest of these regions, which is located between the fifth and the sixth p-strand, is 44 residues long and runs into the inside of the barrel where it defines the channel size and shape. This structure was designated 'eyelet'. Recently, Jap etal. (1991) determined the structure of PhoE porin by electron crystallography and found a density region situated furthest away from the symmetry axis, in contact with part of the wall and extending across ttie entire length of the channel. These results suggest that the PhoE porin may have a similar 'eyelet' as the R. capsulatus porin. Thus, the postulated third external loop in the PhoE model could correspond to the eyelet and could be important in determining the channel size. The aim of this study was to localize the postulated third external loop of PhoE with respect to the membrane, using a strategy originally developed by Charbit et al. (1986) in their studies on the topology of outer membrane protein LamB. After insertion of a viral epitope into sites of LamB that were predicted to face the cell surface, it was demonstrated that the epitope was recognized on intact cells by virus-specific antibodies, which proved that the corresponding LamB regions are indeed cell-surfaceexposed. Similarly, antigenic determinants of the VP1 protein of foot-and-mouth disease virus (FMDV) have been inserted in several postulated external loops of PtioE protein(Agterbergefa/., 1987,1990a). Consistent with the proposed topology model of PhoE, the epitopes were accessible for virus-specific mAbs at the cell surface in intact cells. In the present study, an antigenic determinant of VP1 protein of FMDV was inserted into the region between the postulated fifth and sixth p-strand of PhoE protein. Cell-surface-exposure of the corresponding region would be expected to lead to the binding of a virus-specific mAb to the surface of intact cells, expressing the hybrid protein. Alternatively, if this region is embedded in the pore channel, and thus important in determining the channel size, the insertion would be expected to reduce the channel opening. Results Piasmid constructions Site-directed mutagenesis was applied to create pMS37 with a unique BamH\ restriction site in the region coding for the postulated third cell-surface-exposed loop of PhoE protein (Fig. 1). The mutant PhoE protein encoded by this piasmid contains a Gly residue at position 114 instead of Asp. To create FMDV epitope insertions in PhoE, oligodeoxyribonucleotides coding for amino acids 144 to 150 of VP1 protein (Fig. 1) were ligated into BamHI-linearized pMS37. The linker was inserted up to four times yielding pMS38, pMS39, pMS50 and pMS51, respectively. For control experiments, epitope insertions were also created in a region of PhoE postulated to be at the periplasmic side of the membrane. This was done by ligating the linker into the unique SamHI site of pJP321 (Bosch etal., 1989), located between the codons for amino acids 220 and 221 of PhoE. The linker was inserted up to three times, yielding pMS66, pMS67 and pMS68 (Table 1). Also pS2 encoding a PhoE-FMDV hybrid protein with a tandem insertion of the epitope in the eighth exposed loop of PhoE (Agterberg et al., 1990a) was used in control expenments. Expression of PhoE-FMDV hybrid proteins To examine the expression of the mutant and hybrid proteins, phoRphoE mutant strain CE1248 was transformed with the different plasmids. Cell envelopes were isolated and their protein patterns were analysed by SDS-PAGE (see for examples Fig. 2). As expected, the electrophoretic mobility of the PhoE-FWDV hybrid Topology of PhoE porin "fPhoE FMDV PhoE-' OmpApJP pMS 29 37 38 39 50 51 Fig. 2, SDS-PAGE protein patterns ot cell envelopes from plasmid containing derivatives of strain CE1248 after overnight growth at SZX. Plasmids present were pJP29 (wt PhoE), pMS37 (Asp-IM - - Gly-114), pMS38(1x FMDV), pMS39(2x FMDV), pMS50 (3x FMDV) and pMS51 (4x FMDV). Only the relevant part of the gel is shown. The positions ot OmpA, PhoE, PhoE-FMDV hybrids and OmpT are indicated. proteins was reduced compared with the wild-type PhoE protein and the pMS37-encoded mutant protein carrying the Asp-114 to Gly substitution. Aftergrowth at 37*'C, the expression of the hybrid proteins with more than two copies of the FMDV epitope inserted at the postulated third external loop was somewhat reduced and an additional protein band with an apparent molecuiar weight (Mr) of 40000, i.e. identical to wild-type PhoE, was detected (Fig. 2). However after growth at 30°C, the expression of all hybrid proteins was comparable with wild-type PhoE protein and the additional 40 K band was very faint (results not shown). This 40 K protein did not react with the PhoE-specific mAb mEI in Western immunoblots (results not shown) and probably corresponds to the cuter-membrane protease OmpT whose expression is temperature regulated (Rupprecht et ai, 1983). Table 1, Binding of mAbs to cells expressing the PhoE-FMDV hybrid proteins.^ Disintegrated cells Plasmid MA11 PP1-1 pJP29 wt PhoE - ++ Eighth cell-surface-exposed region 2x FMDV + + pS2 MA11 PP1-1 The native conformation of PhoE protein in the outer membrane is a trimer that is highly resistant to SDS denaturation (Lugtenberg and van Alphen, 1983). When cell envelopes of strain CE1248 expressing wild-type PhoE protein were extracted with salt and incubated at room temperature instead of 10O^C prior to SDS-PAGE, a smear representing PhoE trimers could be detected (Fig. 3). In similar analyses, trimers of all hybrid proteins were observed (see Fig, 3 for examples). Even when the samples were heated at eCC prior to SDS-PAGE, no denaturation of the trimers was observed, showing that trimer-stability was not (drastically) affected by the insertions. PhoE trimer PhoE monomer OmpA _ m tryp pJP29 tri tryp pMS38 tri m tryp tri pM539 Fig, 3. Trypsin accessibility and trimer formation of the hybrid proteins. Cell envelopes of strain CE1248 expressing PhoE wild-type or PhoEFMDV hybrid proteins were treated in the presence (lanes tryp) or absence (lanes m and tri) of trypsin or were salt-extracted (lanes tri) as described in the Experimental procedures. Ceil envelopes were resuspended in sample buffer and incubated at I W C (lanes m and tryp) or at room temperature (lanes tri) prior to SDS-PAGE. Results are shown for the wild-type PhoE (pJP29) and for the hybrid proteins containing one (pMS38) or two (pMS39) copies of the FMDV epitope. Only the relevant part of the gel is shown. The positions of PhoE monomer, trimers and OmpA are indicated. heated RT Whole cells Ml itant protein 133 MAIl PP1-1 - ++ ++ + Third cell-surtace-exposed region Gly-114 ++ pMS37 pMS38 1X FMDV ++ 2x FMDV -f--tpMS39 3x FMDV + + +-f pMS50 pMS51 4X FMDV ++ ++ + + ++ ++ ++ ++ ++ ++ ++ Periplasmically exposed region 1 X FMDV ++ pMS66 pMS67 2x FMDV ++ pMS68 3x FMDV ++ ++ -f + ++ -f + ++ ++ ND ND ND PhoE protein functions as (part of) the receptor for phage TC45. To investigate whether the hybrid PhoE proteins expressed in strain CE1248 were correctly assembled in the outer membrane, phage sensitivity tests were performed. All mutant and hybrid proteins appeared to function as phage receptor. Thus, the proteins are correctly assembled into the outer membrane. -1- Exposure of the Ff^DV epitopes ND ND ND a. CE1248 cells containing the plasmids indicated in first column were grown in L-broth at 3 0 ^ and either directly coated to the surface of the wells of a microtitre plate or first disintegrated. Disintegrated cells were kept at room temperature (RT) or heated tor 10 min at lOCC prior to coating. The binding of the PhoE-specific, conformation-dependent mAb PPl-1 and the virus-speciftc mAb MA11 were measured tn an ELISA: + + , binding; +, reduced binding; - , no binding; ND = not done. To determine whether the FMDV epitopes inserted in PhoE are accessible at the cell surface, whole-cell enzyme-linked immunosorbent assays (ELISAs) were performed. The mAb PP1 - 1 , which recognizes a cell-surfaceexposed, conformational epitope of PhoE protein, was able to bind to cells expressing the wild-type, mutant and hybrid PhoE proteins (Table 1, Fig. 4) confirming the correct incorporation of all these proteins into the outer 134 M. Struyve et a\. 37 38 39 50 51 Fig, 4, Binding of monoclonal antibodies in whole-cell ELISAs to piasmid containing derivatives of strain CE1248. Monoclonal antibody MA11 is directed against the FMDV epitope. The monoclonal antibody PP1 -1 is directed against a cell surface-exposed part of PhoE protein. Results are shown for derivatives of strain CE1248 containing pJP29 (wt PhoE), pMS37(Asp-114 —Gly), pMS38(lx FMDV), pMS39 (2>; FMDV), pMS50 (3x FMDV), pMS51 (4x FMDV) or pS2 which encodes a PhoEFMDV hybrid protein with two copies of the FMDV epitope inserted in tandem in the eighth cell surface-exposed loop of PhoE protein {Agterberg e( a/., 1990a). membrane. As described previously, the mAb MA11, directed against the linear FMDV epitope, did bind to cells containing pS2 which encodes a hybrid protein with the epitope inserted twice in the eighth exposed region of PhoE (Agterberg et ai, 1990a). However, the mAb did not bind to intact cells expressing the PhoE proteins with one or two copies of the epitope inserted in the postulated third external loop {Fig. 4, Table 1). Only with three or four copies of the FMDV epitope inserted, was MA11 able to bind in a whole-cell ELISA. Disintegration of the pMS38and pMS39-containing cells by sonication was not sufficient to allow MA11-binding in ELISAs. Only when the disintegrated cells were heated for 10 min at 100°C prior to coating to the microtitre plate, could binding of MA11 be observed fTable 1). Apparently, the region of the PhoE polypeptide around amino acid 114 is not sufficiently exposed to allow antibody MA11 to bind to a single or tandemly inserted epitope. Only after at least triple insertions are some of the epitopes sufficiently exposed to allow antibody binding. Similar ELISA experiments were performed with pMS66-, pMS67- and pMS68-containing cells, expressing PhoE hybrid proteins with the FMDV epitope inserted at the postulated periplasmic side of the protein (Table 1). MA11 was not able to bind to these cells in whole-cell ELISAs. In ELISAs with disintegrated cells, the epitope appeared to be accessible to MA11, confirming the periplasmic exposure of the corresponding region of PhoE. Protease accessibility experiments When wild-type PhoE protein is correctly assembled in the outer membrane, it is completely resistant to treatment of cell envelopes with proteases (Tommassen and Lugten- berg, 1984). Mutant proteins that are not correctly assembled are completely degraded (Agterberg et al., 1989). Insertion mutations may create exposed protease-sensitive sites. When a mutant protein containing such an insertion is correctly assembled into the outer membrane, one may expect to find protected fragments of the protein after treatment of cell envelopes with proteases (Agterberg et al., 1990a). Treatment of cell envelopes containing the PhoE-FMDV hybrid proteins with trypsin did not result in degradation of the hybrid proteins (see Fig. 3 for examples) which confirms the correct localization of the proteins in the outer membrane. The proteolytic enzyme trypsin is relatively specific in that it splits peptide bonds on the carboxyl side of lysine and arginine residues only. These residues are not present in the FMDV epitope used (Fig. 1). Proteinase K cleaves after aliphatic, aromatic and hydrophobic amino acids and the FMDV epitope contains several of these substrate residues. Thus, proteinase K was used to study the accessibility of the inserts in the third external loop of PhoE in further detail. Wild-type and the pMS37-encoded Gly-114 mutant proteins were completely resistant to treatment of the cell envelopes with proteinase K, as was the hybrid protein with one copy of the FMDV epitope inserted (Fig. 5). However, proteinase K-treatment of cell envelopes containing the hybrid protein with two copies of the FMDV epitope inserted resulted in partial digestion of the hybrid protein. Two degradation products, one with an M, of 28 000 (Fig. 5) and one with an M, of 14000 (results not shown), were detected when the PhoE-specific mAbs mE2-1 and mEI, respectively, were -68 -43 PhoE-29 -14 ms pJP pMS 29 38 39 50 Fig. 5. Proteinase K accessibility of PhoE-FMDV hybrid proteins in cell envelopes. Cell envelope preparations of cells containing the different plasmids were treated with proteinase K and analysed by SDS-PAGE followed by Westem immunoblotting. In the biot shown, the mAb mE2-1, which recognizes an epitope located between amino acids 250-298 of PhoE. was used. Results are shown for the wild-type PhoE protein (pJP29) and the hybrid proteins with one (pMS38), two (pMS39) or three (pMS50) copies ot the FMDV epitope. The positions of PhoE and molecular mass standard proteins (ms), in kilodaltons, are indicated. The positions of the PhoE degradation products are indicated by asterisks. Only the relevant part of the blot is shown. Topology of PhoE porin used. The former antibody recognizes an epitope between amino acid 250 and 298 of PhoE, whereas the latter recognizes an epitope between amino aoids 47 and 55. Since the linkers are inserted between the amino acids 114 and 115 of PhoE, the protected degradation products correspond to the C- and N-terminal fragments of PhoE respectively, after cleavage at the site of insertion. When the hybrid proteins with three (Fig. 5) or four (results not shown) oopies of the epitope were analysed in a similar way, only the two degradation products were detected. Apparently, the region around amino acid 114 of PhoE with less than two oopies of the epitope inserted is not sufficiently exposed to allow proteinase K cleavage. However, insertion of two copies of the FMDV epitope in this region results in sufficient exposure to allow partial proteinase K cleavage. With at least three copies of the epitope, the insert is sufficiently exposed to allow full cleavage of the hybrid proteins. Pore function of the hybrid proteins in vivo The results described in the previous paragraphs suggest that the postulated third external loop of PhoE is not completely exposed at the cell surface, but rather is embedded in the pore interior, like the eyelet in the R. capsulatus porin. If this hypothesis is correct, the channel size is expected to diminish when FMDV-epitopes are inserted in this region. Consequently, the sensitivity of cells, producing the hybrid PhoE proteins as the only porins, to hydrophilic antibiotics like cephaloridine should be reduced. Strain CEI 249, which lacks the major outer membrane pore proteins, was transformed with the different piasmids. Filter discs containing cephaloridine were placed on agar top layers containing cells expressing the different hybrid proteins and, after overnight incubation at 30°C, the growth-inhibition zone around the filters was measured. In the case of pJP29- and pMS37containing cells, expressing the wild-type and the Gly-114 mutant protein, respectively, a growth inhibition zone of 2.9mm around the filters was measured. When the FMDV epitope was inserted once in PhoE protein, the cells were very resistant to cephaloridine (growth inhibition zone of 0.5 mm}; with multiple copies of the insert, the sensitivity increased again (growth inhibition zone of 1.1 mm), but did not reach the level of cells expressing the wild-type PhoE. This indicates that the channel size is diminished by insertions in the third loop of the PhoE protein. Pore function of the hybrid proteins in vitro The pore characteristics of the hybrid proteins were further studied in vitro in black lipid-bilayer experiments. Porins were isolated from strain CE1249 carrying the different piasmids. Ali porins were able to increase, at 135 Table 2. Average single-channel conductance of the hybrid protein pores in different salt solutions.^ Porin encoded by 1 MKCI 3MKCI 1 M LiCI 1 M KCH3COO pJP29 pMS37 pMS38 pMS39 pMS50 pMS51 1.50 1,60 0,17 0,60 0.60 0,80 4.80 4.70 0,40 2,10 2,00 2.40 1,20 1,10 0.12 0,38 0.35 0,45 0,60 0,70 0,10 0,45 0.52 0,65 a. Average single-channel conductance was measured with diphytanoyi glycerophosphocholine/n-decane membranes in the presence of different hybrid pores, A was calculated from recordings similar to those given in Fig. 6 and by averaging at least 100 conductance steps. The pH of the unbuffered aqueous salt solutions was around 6; Vm = lOmV, T = 25°C, nanomolar concentrations, the specific conductance of the lipid bilayer by many orders of magnitude. The kinetics of the increase were similar to those described previously for wild-type PhoE (Bauer et al., 1988a) and other porins (Benz etal., 1980). After a rapid increase during 15 to 20 min, the membrane conductance increased at a much slower rate. Addition of the detergents SDS or Genapol X-80 alone in control experiments did not lead to any appreciable increase of the membrane conductance. In the case of the wild-type and the Gly-114 mutant porins, the time-course of the increase was similar to that described previously (Bauer et ai, 1988a). However, in the case of the hybrid porins the conductance was considerably lower, which indicated either a different conductance ot the single conductive units or partial inactivation of the channels. The addition of the porins at much lower concentrations to the aqueous solution on one or both sides of a lipid bilayer membrane allowed the resolution of step increases in the membrane conductance; each step corresponds to the incorporation of one channel-forming unit into the membrane (Fig. 6). These steps showed that all the different porins form defined pores in the lipid bilayer membranes. Single-channel measurements were performed with different salt solutions. The single-channel conductance for the wild-type PhoE and the Gly-114 mutant was 1.5 nS in 1 M KCI, and only a small number of steps had twice this value, indicating simultaneous insertion of two channels (Bauer et aL. 1988b). The singlechannel conductance was influenced by the insertions; with porins containing a single copy of the epitope, the single-channel conductance was reduced by a factor of 10 compared with the wild-type pores (Fig, 6, Table 2). With multiple copies of the FMDV epitope inserted, the single-channel conductance increased by a factor of four compared with the single insertion, but it did not reach the wild-type level (Table 2). The results obtained with 3 M KCI suggested a linear relationship between bulk aqueous 136 M. Struyve et a\. Measurements with other salts, i.e, 1 f^ LiCI and 1 M KCH3COO (which represents combinations of cations and anions of different aqueous mobilities), supported these results since the single-channel conductance of the PhoE porins with epitope insertions was also reduced compared with the conductance of wild-type PhoE and Gly-114 mutant porins (Table 2), Furthermore, they suggested that the selectivity of the pores had been altered by the insertions. Therefore, the selectivity of the pores was studied in further detail. JinS Selectivity of the hybrid proteins in vitro 1nS Zero-current membrane potentials were measured to study the ion selectivity of the wild-type PhoE porin and the hybrid porins in detail. Porins were isolated from strain CE1249 carrying the different plasmids and added to the aqueous phase in a Teflon cell containing an optical black membrane. The ratio of the permeabilities, Pcatior/Panion (Pc/Pa), for the wild-type and the Gly-114 mutant were approximately the same as published previously (Benz et ai, 1985). The Pc/Pa for the hybrid with one copy of the epitope indicated that this porin was almost neutral, whereas for the hybrids with at least two copies of the epitope, the Pc/Pa value indicated that these hybrids had a preference for cations over anions (Table 3), These results are in agreement with the single-channel conductance data obtained with LiCI and KCH3COO. Discussion Fig. 6. Chart recording of stepwise increase of the membrane current after the addition of porins to a diphytanoyi glycerophosphocholine/ n-decane bilayer membrane. Both recordings start on the left and trace 2 continues in the upper trace. Trace 1 represents wild-type PhoE porins and trace 2 hybrid porins with one copy of the FMDV epitope inserted. Current amplification was 10® and 5.10^ in the case of the wild-type and the hybrid porins, respectively. The aqueous phase contained 3M KCI, pH 6,0, the temperature was 25"C and the applied voltage was lOmV, In this paper, the epitope insertion method, originally proposed by Charbit ef al. (1986), was used to study the topology of PhoE protein, with special emphasis on the postulated third external loop of the protein. Previously, Table 3. Zero-current membrane potentials, V^,, of the hybrrd protein pores for a 10-fold gradient of KCI." Porin conductivity and single-channel conductance. No closing events were observed with any of the porin preparations, which indicates that the lifetime of the pores was long, usually exceeding 10 min. Even when the transmembrane potential was increased up to 150mV, no closing events were observed. This shows that the pores are not voltage sensitive at physiological potentials. Thus, insertion of nine amino acids in the postulated third loop of PhoE is sufficient to reduce the channel opening by 90%. With at least 18 amino acids inserted, the loop becomes too large and probably crawls up against the pore wall towards the exterior of the pore. encoded by V^ Po/Pa pJP29 PMS37 pMS38 PMS39 pMS50 pMS51 -26 -22 -2.7 19 30 15 0.27 0.33 0.88 2.50 4.30 2.10 a. Zero-current membrane potentials were measured with diphytanoyi glycerophosphocholine/fi-decane membranes in the presence of different hybrid pores, Un, is defined as the difference between the potential on the diluted side (10mM KCI) and the potential at the concentrated side (lOOmM KCI). The pH of the unbuffered aqueous salt solutions was around 6; T = 25''C, Pg/Pg was calculated from the Goldman-HodgKin-Katz equation from at least three individual experiments. Topology of PhoE porin FMDV epitopes have been inserted in the fourth (Agterberg etai, 1987), fifth (Agterberg etai, 1990b) and eighth (Agterberg et al., 1990a) external loops of PhoE. Single epitope insertions were sufficient to allow the binding of the virus-specific mAb to intact cells expressing these hybrid proteins. Also insertions in the second postulated loop of PhoE protein were created (Agterberg et ai, 1990a). Once the epitope was present, intact cells expressing the hybrid protein were not able to bind the mAb MA11 in ELISAs. In our proposed topology model for PhoE protein (Tommassen, 1988), this insertion site is at the border of a membrane-spanning segment and a cell surface-exposed region, which might be the reason that the epitope was not sufficiently exposed to allow MA11 binding. However, the insertion enlarged the second loop sufficiently to make it accessible in cell envelopes to proteolytic attack (Agterberg et ai, 1990a). When two copies of the epitope were present at this site, the mAb was able to bind to intact cells. In the present work, we also inserted the epitope in a region exposed at the periplasmic side of the membrane. When the FMDV linker was inserted up to three times at this site of the protein, the MA11 antibody was not able to bind in a whole-cell ELISA. In an ELISA with disintegrated cells as immunobilized antigen, the periplasmic-exposed region became accessible to the antibody and binding could be observed. These results indicate that epitope insertion is a very suitable method by which to study details of the topology of outer-membrane proteins and that the proposed PhoE model is very accurate. In this paper, the epitope insertion method was applied to locate the postulated third loop of PhoE. When one or two copies of the FMDV epitope were inserted, no positive reaction with mAb MA11 was observed in a whole-cell ELISA or in an ELISA with disintegrated cells. This means that the epitope was accessible neither from the outside nor from the periplasmic side of the membrane. This indicates that the third loop is hidden in the interior of the pore. When at least three copies of the epitope were present, the mAb oould bind in a whole-cell ELISA. Thus In this case, the epitopes became accessible from the outside. This indicates that the third loop is close to the cell surface, like the eyelet region of the R. capsulatus porin. Consistent with the location of the third loop of PhoE in the channel interior, the insertions were found to reduce the channel size. While insertions reduced the channel size, one would expect that deletions in this region should enlarge the pore size. Benson et al. (1988) have selected mutants that were able to grow on a medium containing malto dextrins as the sole carbon source in the absence of a functional IamB gene, which encodes the normal maltodextrin channel. These mutant strains contained an altered OmpF protein with short deletions that altered the porin to allow the passage of larger solutes. The deletions 137 found were six to 15 amino acid residues long in the region between Ala-108 and Val-133. Similarly, a mutant OmpC protein with a larger channel size has been described that contained a deletion of amino acids 103-110 (Misra and Benson, 1988; Rocque and MoGroarty, 1990). Point mutations in this region, affecting the aspartate residue at position 105 of OmpC, also resulted in an increased pore size (Misra and Benson, 1988; Lakey ef ai, 1991). The amino acids 108-133 of the OmpF protein and the amino acids 103-110 of the OmpC protein are situated in the region corresponding to the third loop of PhoE protein. These results provide additional evidence that the third loop of the porins determines the pore size and this is consistent with a location within the pore channel. The OmpC mutant proteins mentioned above decreased trimer stability and increased voltage sensitivity when reconstituted in lipid bilayers (Rocque and McGroarty, 1990; Lakey efa/., 1991). The insertions in the third loop of PhoE protein did not drastically affect the stability of the trimers. Furthermore, no voltage-dependent closing of the channels was observed. The latter may be related to the black lipid-bilayer system used (Benz etai, 1978), in which high-voltage-induced closing of bacterial porins is not generally observed. Previously, Bauer etai (1989) showed that the positively charged lysine residue at position 125 of PhoE is of special importance in determining the anion-selectivity of the PhoE pores. This Lys-125 is present in the postulated third loop of PhoE. Thus, this region is not only important in determining the channel size, but it also has a function in determining the selectivity of the pores. The insertions in the PhoE hybrids presented in this paper changed the selectivity of the pores from anion to cation selective when at least two copies of the FMDV epitope were present. This could be caused by shielding or misplacement of the critical Lys-125 residue by the long stretch of inserted residues and/or by the presence of the negatively charged aspartate residue in the FMDV epitope which oould attract positively charged ions. Very recently, the crystal structures of the E. coli porins OmpF and PhoE have been elucidated (Cowan ef ai, 1992). These structures show that our original topology model of PhoE protein (van der Ley et ai, 1986; Tommassen, 1988) was very accurate, not only with respect to the number of membrane-spanning segments and exposed domains, but also with respect to the approximate positions of these segments in the primary structure of the protein. Furthermore, as we predicted from the experiments described in this paper, these structures show that the third loop folds into the p-barrel and constricts the size of the pore at approximately half the height of the barrel, which is similar to that whioh occurs with R. capsulatus porin. It was possible to insert at least up to 36 amino acids in 138 M. Struyvee\a\. the third loop of PhoE protein without affecting the biogenesis of the protein. These results are consistent with those of Agterberg ef al. (1990a) who found that the maximal number of amino acids that can be inserted without disturbing outer-membrane assembly is between 30 and 50 amino acids, depending on the site of insertion. However, the expression level of the larger hybrids was drastically reduced, in contrast to the hybrid proteins described in this paper. This indicates that the third loop of PhoE may be very suitable, when PhoE is used as an exposure vector for the expression of foreign antigenic determinants in order to develop new vaccines (Agterberg efa/., 1987; 1988; 1990a,b). Experimental procedures linker in the region of the DNA coding for amino acids Asn-220 and lle-221 of PhoE protein, which are postulated to be exposed at the periplasmic side. Plasmid pS2 (Agterberg et at., 1990a} contains a double in-tandem insertion of a synthetic linker, encoding an antigenic determinant of VP1 protein of FMDV, in the DNA coding for the eighth ceil-surface-exposed loop of PhoE protein. Plasmid pMS37 (Fig. 1), with a unique SamHI restriction site in the DNA coding for the postulated third exposed region of the PhoE protein, was created by site-directed mutagenesis. For epitope insertion, the plasmids pJP321 and pMS37 were digested with SamHI and dephosphorylated. Two complementary oligodeoxynucleotides, coding for amino acids 144 to 150 of VP1 protein of FMDV (Metoen ef al.. 1987) (Fig. 1), were phosphorylated and ligated into the linearized plasmids. The linker was designed such that a new SamHI restriction site is generated at the 3' end of the linker but not at the 5' end (Fig. 1). Thus, by repeating the procedure, piasmids could be constructed containing multiple copies of the linker (Table 1). Bacterial strains, phages and growth conditions E. CO//K-12 strain CE1248 {Korteland etat., 1985) is deleted for the p h o f gene and does not express the related OmpF and OmpC proteins as a result of an ompR mutation. The strain also carries a phoR mutation, resulting (n constitutive expression of the pho regulon. Strain CE1249 was isolated as a spontaneous derivative of CE124a that is resistant to phage \ vir (Jacob and Wollman, 1954). The mutation in this strain results in the failure to express LamB protein during growth in the presence of 1 % maltose and could therefore either be in tamB or in the regulatory gene rr)atT. Bacteria were grown overnight under aeration at SO'C or 37X in L-broth (Tommassen et at., 1983). Where necessary, the medium was supplemented with chloramphenicol (25 |xg ml '). Sensitivity to the PhoE-specific phage TC45 (Chal and Foulds, 1978) was determined by cross-streaking. Generat DNA manipulations Plasmid DNA purifications were performed as described by Birnboim and Doly (1979) followed by anion-exchange chromatography on Qiagen-columns (Diagen). Recombinant DNA techniques were performed essentially as described by Maniatis er al. (1982). Restriction endonuciease reactions and bacteriophage T4 DNA ligase treatments were performed as described by the manufacturers of the enzymes. DNA fragments were analysed on 1-2% (w/v) agarose gels. Oligonucleotides were synthesized on a Biosearch 8600 DNA synthesizer. Site-directed mutagenesis was carried out by a two-step polymerase chain reaction (PCR) as described by Mikaelian and Sergeant (1992), with the modification that the primer, marked primer 2 in their protocol, had no mismatched 3' end. Consequently, the efficiency of the method was expected and found to be 50%. Correct mutant plasmids were elected on the basis of the presence of a SamHI enzyme restriction site (Fig. 1). DNA sequencing was performed using the T7 DNA polymerase sequencing kit (Pharmacia LKB Biotechnology Inc.). Isoiation and characterization of cett fractions Cell envelopes were isolated by centrifugation after ultrasonic disintegration of the cells (Lugtenberg etat., 1975). Porin trimers were extracted from cell envelopes by incubation for 30 min at 37°C in a buffer containing lOmM Tris-HCI (pH 8.0), 0.5M NaCI, lOmM EDTA. To test the accessibility of membrane proteins to proteases, cell envelopes were resuspended in 20|i.l lOmM Tris-HCI (pH 8.0), lOmM MgCb containing 0.5 mg ml ' trypsin or 0.05mg ml"^ proteinase K. The samples were kept on ice for 30 min, after which the cell envelopes were reisolated by centrifugation. The protein patterns of cell fractions were analysed by SDS-PAGE (Lugtenberg etat., 1975), followed, where necessary, by Western immunoblotting (Agterberg et at., 1988). In Western immunoblotting mAbs mEI, which binds to an epitope located between amino acids 47 and 55 of the PhoE protein (Agterberg ef at., 1990a), ormE2-1, which binds to an epitope located between amino acids 250-298 of the PhoE protein (M. Kleerebezem, personal communication), were used. Enzyme-tinked immunosorbent assay Binding of mAbs to intact or disintegrated cells was measured in ELISAs (van der Ley ef ai, 1985). Disintegrated cells were obtained by resuspending cells of an ovemight culture in a buffer containing 15mM Tris-HCI (pH 9.6), I m M EDTA and 65mM Na2C03 followed by sonication for 15s. The samples were incubated at room temperature or heated for 10 min at 10O''C prior to coating on the surface of the wells. Intact or disintegrated cells were coated on the surface of the wells of a polystyrene microtitre plate. The PhoE-specific monoclonal antibody PP1 -1 recognizes a conformational epitope at the ceil-surface-exposed part of PhoE protein (van der Ley ef al., 1985; 1966). Monoclonal antibody MA11 recognizes a continuous epitope, constituted by the amino acids 144 to 150 of VP1 protein of FMDV (Meloen efa/., 1987). Ptasmids Plasmids pJP12 (Tommassen etal., 1982) and pJP29 (Bosch ef al., 1986) are derived from the cloning vector pACYC184 and carry the phoE gene. Plasmid pJP321 (Bosch et al., 1989) is derived from plasmid pJP29 by insertion of a 12-mer SamHI Antibiotic sensitivity assay Sensitivity to antibiotics was determined as described by Benson and Decloux (1985) with several modifications: strains were Topology of PhoE porin grown to the mid-log phase in L-broth. L-broth agar plates were overlayed with 3 ml 0,6% L-broth agar, containing 0.1 ml of the mid-log cultures. Whatman no. 1 paper fillers (5 mm diameter) with 5|xl of an antibiotic suspension (0.4jig ^.1 ') were placed on the lawn of cells. After overnight incubation at SOX, the growth inhibition zone around the filters was measured. Lipid bilayer experiments Porin trimers were isolated as described by Verhoef ef a/. (1987). The methods used for black lipid-bilayer experiments have been described previously (Benz ef ai. 1978). Briefly, membranes were formed across a circular hole, with an area of 0.2 mm^ tor single-channel conductance and an area of 2mm^ for zero-current membrane potentials, in a thin wall, separating two aqueous compartments in a Teflon cell, from a 1 % (w/v) solution of diphytanoyi glycerophosphocholine (Avanti Bioohemicals, Birmingham, AL) in n-decane, Bilayer formation was indicated when the membrane turned optically black in reflected light. For single-channel experiments the current through the membranes after application of a transmembrane potential of lOmV was measured with two calomel electrodes switched in series with a voltage source and a current amplifier (Keithley 427). The amplified signal was monitored with a storage oscilloscope and recorded on a strip-chart recorder. Zero-current membrane potentials were performed as described previously (Benz etai. 1970). Briefly, membranes were formed in a Teflon cell containing a lOmM KCI salt solution and the porins were added to the aqueous phase when the membranes were in the black state. After incorporation of 100 to 1000 porin channels into the membrane, salt gradients were established by addition of small amounts of concentrated KCI solution to one side of the membrane. The membrane current was measured with a pair of calomel electrodes switched in series with a voltage source and a electrometer (Keithley 602), The membrane potentials were analysed using the Goidman-HodgkinKatz equation (Benz etai, 1978). Acknowledgements The authors would like to thank Angela Schmid for helpful discussions and Gabriele Grutzner for help with the membrane experiments. This work was supported by Netherlands Foundations for Chemical Research and for Medical and Health Research (grant number 900-515-003), with financial aid from the Netherlands Organization for the Advancement of Research. The membrane experiments were supported by grants of the Deutsche Forschungsgemeinschaft (project B9 of the Sonderforschungsbereich 176) and of the Fonds der Chemischen Industrie, References Agterberg. M.. Adriaanse, H., and Tommassen. J. (1987) Use of outer membrane protein PhoE as a carrier for the transport of a foreign antigenic determinant to the cell surface of Escherichia coliK-^2. Gene59: 145-150, Agterberg, M., Fransen, R., and Tommassen, J. (1988) Expression of Escherichia coli PhoE protein in avirulent Salmonella typhimurium aroA and galE strains. FEMS Microbiot Lett 50: 295-299. 139 Agterberg, M., Adriaanse, H., Tijhaar, E,. Resink, A., and Tommassen, J, (1989) Role of the cell surface-exposed regions of outer membrane protein PhoE of Eschericfiia coli K12 in the biogenesis of the protein. Eur J Biochem 185: 365-370. Agterberg, M., Adriaanse, H., van Bruggen, A., Karperien, M.. and Tommassen, J. (1990a) Outer-membrane PhoE protein of Escherichia coli K-12 as an exposure vector: possibilities and limitations. Gene 88: 37-45. Agterberg. M., Adriaanse, H., Lankhof, H,, Meloen. R.. and Tommassen. J. (1990b) Outer membrane PhoE protein of Escherichia coli as a carrier for foreign antigenic determinants: immunogenicity of epitopes of foot-and-mouth disease virus. VaccirieS: 85-91, Bauer, K,. van der Ley, P., Benz, R., and Tommassen, J, (1988a) The pAio-controlled outer membrane porin PhoE does not contain specific binding sites for phosphate or polyphosphates. J Biol Chem 263: 13046-13053. Bauer, K,. Schmid, A., Boos, W., Benz, R., and Tommassen, J. (1988b) Pore formation by pho-controlled outer-membrane proteins of various Enterobacteriaceae in lipid bilayers. Eur J Biochem ^74: 199-205. Bauer. K., Struyve, M., Bosch, D., Benz, R., and Tommassen, J. (1989) One single lysine residue is responsible for the special interaction between polyphosphate and the outer membrane porin PhoE of Escherichia coli J Biol Chem264:16393-16398. Benson. S.A., and Decloux, A. (1985) Isolation and characterization of outer membrane permeability mutants in Esctierichia co/fK-12.JSac(er/o/161: 361-367. Benson, S.A., Occi, J.L.L.. and Sampson, B.A, (1988) Mutations that alter the pore function of the OmpF porin of Escherichia coli K12. J Mol Biol 203: 961-970. Benz. R,. Janko, K,. Boos, W., and Lauger, P. (1978) Formation of large, ion-permeable membrane channels by the matrix protein (porin) of Escherichia coli. Biochim Biophys Acta 511:305-319. Benz, R., Janko, K., and Lauger. P. (1979) Ionic selectivity of pores formed by the matrix protein (porin) of Escherichia coli. Biochim Biophys Acta 551: 238-247. Benz, R., Ishii, J,, and Nakae, T. (1980) Determination of ion permeability through the channels made of porins from the outer membrane of Salmonella typtiimurium in lipid bilayer membranes. J Membrane Biol 56: 19-29. Benz. R., Smid, A., and Hancock, R.E.W, (1985) Ion selectivity of Gram-negative bacterial porins, J Bacterioi 162: 722-727. Birnboim, H.C. and Doly, J. (1979) A rapid alkaline extraction procedure for screening recombinant plasmid DNA. Nucl Acids Res 7: 1513-1523, Bosch, D., and Tommassen, J. (1987) Effects of linker insertions on the biogenesis and functioning of the Escherichia coli outer membrane pore protein PhoE. Mol Gen Genet 208: 485-489. Bosch, D,, Leunissen, J., Verbakel, J,, de Jong, M., van Erp, H., and Tommassen, J. (1986) Periplasmic accumulation of truncated forms of outer-membrane PhoE protein of Escherichia coli K-12. J Mol Biol 189: 449-455. Bosch, D.,Scholten,M.,Verhagen,C,, and Tommassen, J. (1989) The role of the carboxy-termina! membrane-spanning fragment in the biogenesis of Escherichia coti K12 outer membrane protein PhoE. Mol Gen Genef 216:144-148. Chai, T., and Foulds, J. (1978) Two bacteriophages which utilize a new Escherichia coli major outer membrane protein as part of their receptor. J Bacteriol 135: 164-170. Charbit, A., Boulain, J.C, Ryter, A., and Hofnung, M. (1986) Probing the topology of a bacterial membrane protein by genetic insertion of a foreign epitope; expression at the cell surface. EMBO J 5: 3029-3037, 140 M. Struyve etal Cowan, S,W., Schirmer, T., Rummel, G., Steiert, M.. Ghosh, R,. Pauptit, R.A., Jansonius, J.N., and Rosenbusch, J.P. (1992) Crystal structures explain functional properties of two E coli porins. Nature. 358: 727-733, Jap, B,K,, Walian, P.J., and Gering. K. (1991) Structural architecture of an outer membrane channel as determined by electron crystallography. Nature 350: 167-170. Jacob, F., and Wollman, E.L, (1954) etude genetique d'un bacteriophage tempere d'Escherichia coli. I, Le systeme genetique du bacteriophage A. Ann Inst PasteurQ.7: 653, Kleffel. B., Garavito, R.M,, Baumeister, W., and Rosenbusch, J.P. (1985) Secondary structure of a channel-forming protein: porin from E coti outer membranes. EMBO J 4: 1589-1592. Korteland, J., Overbeeke, N., de Graaff. P., Overduin, P., and Lugtenberg, B. (1985) Role of the Arg'^^ residue of the outer membrane PhoE pore protein of Escherichia coli K12 in bacteriophage TC45 recognition and in channel characteristics. EurJBiocfiem 152: 691-697. Lakey, J.H., Lea, E.J.A., and Pattus, F. (1991) ompC mutants which allow growth on maltodextrins show increased channel size and greater voltage sensitivity. FEBS Lett 278: 31-34. Lugtenberg, B., and van Alphen, L. (1983) Molecular architecture and functioning of the outer membrane of Escherichia coti and other Gram-negative bacteria. Biochim Biophys Aota 737: 51-115. Lugtenberg. B,, Meijers, J., Peters, R., van der Hoek, P., and van Alphen, L. (1975) Electrophoretic resolution of the 'major outer membrane protein' of Escherichia coli K12 into four bands. FESS LeH 58: 254-258. Maniatis, T., Fritsch, E.F., and Sambrook, J, (1982) Motecutar Ctoning. A Laboratory Manuai Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press. Meloen, R.H., Puijk, W.C, Meijer. D.J.A., Lankhof, H., Postumus, W.P.A., and Schaaper, W.M.M. (1987) Antigenlcity and immunogenicity of synthetic peptides of foot-and-mouth disease virus. J Gen Virol 68: 305-314. Mikaelian, I., and Sergeant, A, (1992) A general and fast method to generate multiple site directed mutations. Nucl Acids Res 20: 376. Misra, R., and Benson, S.A. (1988) Genetic identification of the pore domain of the OmpC porin of Escherichia coli K-12. J Bacteriol 170: 3611-3617. Rocque, W.J,, and McGroarty. E.J. (1990) Structure and function of an OmpC deletion mutant porin from Escherichia coli K-12. Biochemistry 29: 5344-5351. Rupprecht, K.R.. Gordon, G., Lundrigan. M., Gayda. R.C, Markovitz. A,, and Earhart, C. (1983) ompT: Escherichia coli K-12 structural gene for protein a (3b). JSacferio/153: 1104-1106. Tommassen, J. (1988) Biogenesis and membrane topology of outer membrane proteins in Esctierichia coli. In Membrane Biogenesis. Op den Kamp. J.A.F. (ed.) NATO ASI series, pp. 351-373. Berlin: Springer-Verlag. Tommassen, J., and Lugtenberg, B. (1984) Amino terminus of outer membrane PhoE protein: localization by use of a btaphoE hybrid gene. J Bacteriol 157: 327-329. Tommassen, J., Overduin, P., Lugtenberg, B., and Bergmans, H. (1982) Cloning of phoE. the structural gene for the Escherichia coli phosphate limitation-inducible outer membrane pore protein. J Bacteriol 149: 668-672. Tommassen, J., van Tol, H., and Lugtenberg, B. (1983) The ultimate localization of an outer membrane protein of Escherichia coli K-12 is not determined by the signal sequence. EMBO J2: 1275-1279. Tommassen, J., Pugsiey, A.P., Korteland, J., Verbakel, J., and Lugtenberg, B. (1984) Gene encoding a hybrid OmpF-PhoE pore protein in the outer membrane of Esctierichia co//K12. Mot Gen Genet 197: 503-508. Tommassen. J., van der Ley, P., van ZeijI, M.. and Agterberg, M. (1985) Localization of functional domains in E coli K-12 outer membrane porins. EMBO J 4: 1583-1587. van der Ley, P., Amesz, H., Tommassen, J,, and Lugtenberg, B. (1985) Monoclonal antibodies directed against the cell-surface-exposed part of PhoE pore protein of the Escherichia coli K-12 outer membrane. Eur J Biochem 147: 401-407. van der Ley, P., Struyve, M., and Tommassen, J. (1986) Topology of outer membrane pore protein PhoE of Escherichia cott. Identification of cell surface-exposed amino acids with the aid of monoclonal antibodies, J Biot Chem 261:12222-12225. van der Ley, P., Burm, P., Agterberg, M., van Meersbergen, J., and Tommassen, J. (1987) Analysis of structure-function relationships in Escherichia coli K12 outer membrane porins with the aid of ompC-phoE and phoE-ompC hybrid genes. Moi Gen Genet 209: 585-591. Verhoef, C , Benz, R., Poon, A.P.W., and Tommassen, J. (1987) New pore protein produced in cells lysogenic for Escherichia coli phage UK253hrk. f u r J SiDc/iem 164:141-145. Vogel, H., and Jahnig, F, (1986) Models for the structure of outer-membrane proteins of Escherichia coti derived from raman spectroscopy and prediction methods. J Mot Biol 190: 191-199. Weiss, M.S., Kreush, A., Schiltz, E., Nestel, U,, Welte, W., Weckesser, J., and Schuiz, G.E. (1991 a) The structure of porin from Rhodobacter capsulatus at 1.8 A resolution. FEBS Lett 280; 379-382. Weiss, M.S., Abele, U., Weckesser, J., Welte, W., Schiltz, E., and Schuiz. G.E. (1991b) Molecular architecture and electrostatic properties of a bacterial porin. Science 254:1627-1630.