Molecular Microbiology (1993) 10(2), 283-292

Linker-insertion mutagenesis of Pseudomonasaeruginosa outer membrane protein OprF

Rebecca S. Y. Wong, Helen Jost andRobert E. W. Hancock*Department ol Microbiology, University of BritishColumbia, 300-6174 University Boulevard, Vancouver,British Columbia V6T1Z3, Canada.

Summary

The oprFgene, expressing Pseudomonas aeruginosamajor outer membrane protein OprF, was subjectedto semi-random iinker mutagenesis by insertion of a1.3 kb H/7icll kanamycin-resistance fragment frompiasmid pUC4KAPA into muitipie blunt-ended restric-tion sites in the oprF gene. The kanamycin-resistancegene was then removed by Pst\ digestion, which lefta 12 nucieotjde pair linker residue. Nine uniqueciones were identified that contained such iinkers atdifferent locations within the oprF gene and werepermissive for the production of f uii-length OprF vari-ants. In addition, one permissive site-directed inser-tion, one non-permissive insertion and one carboxy-terminal insertion leading to proteolytic truncationwere aiso identified. These mutants were character-ized by DNA sequencing and reactivity of the OprFvariants with a bank of 10 OprF-specific monoclonalantibodies. Permissive clones produced OprF vari-ants that were shown to be reactive with the majorityof these monoclonal antibodies, except where theinsertion was suspected of interrupting the epitopefor the specific monoclonal antibody. In addition,these variants were shown to be 2-mercaptoethanolmodifiable, to be resistant to trypsin cleavage inintact cells and partly cleaved to a high-molecular-weight core fragment in outer membranes and, wherestudied, to be accessible to indirect immunofluores-cenee labelling in intact cells by monoclonal antibod-ies specific for surface epitopes. Based on thesedata, a revised structural model for OprF is proposed.

Introduction

The ouler membranes of Gram-negative bacteria areasymmetric biiayers consisting of iipids or phosphoiipids

Received 29 January, 1993; revised 11 June, 1993; accepted 1 July,1993. "For correspondence. Tel. (604) 822 2682; Fax (604) 822 6041.

in the inner monolayer and lipopolysaccharide (LPS) inthe outer monolayer with a restricted number of proteinsinserted into either or both monolayers. The outer mem-brane represents a primary barrier between the cell andthe environment. Therefore, it plays an important role indetermining the entry of substances into the celi. Certainproteins in the outer membrane, known as porins, formaqueous channels which allow the passage of hydrophilicmolecules. The major outer membrane protein of Pseu-domonas aeruginosa, OprF, is a porin (Benz and Han-cock, 1981; Bellido et al., 1992). In addition, OprF con-tributes to the structural integrity of the outer membraneand the cell in maintaining cell morphology and enablingcell growth in low osmolarity media (Woodruff and Han-cock, 1989).

Based on data obtained from circular dichroism studiesand the protein structure-prediction method of Paul andRosenbusch (1985), Siehnel et al. (1990) described aworking model for the topological organization of OprF inthe outer membrane. This model has been tested in partby TnphoA mutagenesis and deletion analysis of the oprFgene, which permitted the partial delineation of the epi-topes recognized by 10 different OprF-specific mono-clonal antibodies (Finnen et al., 1992), nine of which rec-ognize surface epitopes (Martin, 1992). The epitopelocations were further defined by utilizing chemical andproteolytic digestion products of purified OprF (Martin,1992). Both of these studies have generated a significantamount of data that can be used in defining the mem-brane topology of OprF. These studies provided evidencefor the presence, on the surtace of the bacterium, of sev-eral portions of the carboxy-terminus of OprF, despite thefact that the homologous region of the related (Ducheneetal., 1988; Woodruff and Hancock, 1989) protein OmpAfrom Escherichia coli had been assigned to the peripiasm(FreudI etal., 1986). Nevertheless, these approaches hadlimitations in that the epitopes could only be narroweddown to regions of between 26 and 65 amino acids.

Linker-insertion mutagenesis. either random or restric-tion site-directed, has been employed to study the topol-ogy of several E. coli outer membrane proteins includingthe maltoporin LamB (Boulain et al.. 1986), the phos-phate-starvation-inducible porin PhoE (Bosch and Tom-massen, 1987) and the major peptidoglycan-associatedprotein OmpA (FreudI et al., 1986). Linker insertion intro-duces a stretch of several amino acid residues in the

284 R. S. Y. Wong, H. JostandR. E. W. Hancock

amino acid sequence of these membrane proteins, whileleaving the rest ot the protein intact. As a result, it repre-sents a more subtle modification of the protein in compar-ison to the approaches previously utilized for OprF. Arecent crystal structure of PhoE (Cowan etal., 1992) hasdemonstrated that this protein contains a rigid p-barrelstructure of 16 transmembrane p-sheet strands of 9-16amino acids, separated by loop regions. Generally speak-ing, these loop regions of outer membrane proteins aremore likely to accommodate extra amino acids withoutgross perturbation of the protein structure. Indeed all ofthe known PhoE insertion sites (Cowan etal,, 1992) occurwithin these loops. If such insertions affect the bindingof monoclonal antibodies that recognize cell surface-exposed epitopes, it is thus possible to determine the sur-face exposure of the insertion sites.

In this paper, we present our work on semi-randomlinker-insertion mutagenesis of the cloned oprFgene. Theresultant mutants have permitted the generation of arefined membrane topology model for OprF and definitionof specific sites involved in reactivity of certain mono-clonal antibodies.

Results

Construction ofpRW3

It was not possible to subclone the native oprF gene intohigh-copy-number plasmids, probably because of the effi-cient expression of OprF from its own promoter in E. colileading to overexpression lethality (Woodruff etai. 1986).Therefore the putative -10 site of the oprF gene wasmutated by adding a G:C nucleotide pair betweennucleotides -9 and -10 to create a H/ndlll site. This weak-ened the oprF promoter, permitting the oprF gene to besubeloned into high-copy-number plasmids. The finalconstructed plasmid, pRW3 {Fig.1), contained no Pst\sites, permitting subsequent introduction of a Pst] siteduring linker-insertion mutagenesis, as well as a uniqueSal\ site within the oprF gene. Furthermore, the removalof most of the oprF gene-flanking DNA sequences maxi-mized the efficiency of linker-insertion mutagenesis. E.co//strains containing pRW3 expressed OprF, which wasdetectable by all 10 OprF-specific monoclonal antibodieson Western immunoblots. Induction of the lac promoterwith IPTG (1 mM) for 4 h before harvesting only increasedthe expression by approximately twofold, despite the factthat the lac promoter was oriented in the same directionas the oprF gene.

Linker mutagenesis

The plasmid pRW3 was linearized separately by partialdigestion with each of the four restnction enzymes as

Sail

Fig. 1. Restriction map of plasmid pRW3. The position and direction oftranscription ol oprF and the ampicJIIin-resistance marker (Amp) are indi-cated. Abbreviations: U-ori. f i origin of replicalton; pBR322-or', pBR322origin of replication; P-lac. /acpromoter; bp, base pairs.

described in the Experimental procedures. There were atotal of 74 cleavage sites in pRW3 that were recognizedby the four enzymes utilized; 37 of these were within theoprF gene. Low enzyme concentrations and/or ethidiumbromide were used to favour the production of singly cutplasmids, which were further purified by elution frompreparative agarose gels. After ligation of the restrictionenzyme-linearized plasmid pRW3 with the 1.3 kb kana-mycin-resistance cassette, 240 of the 450 screenedtransformants showed an OprF-deficient and kanamycin-resistant phenotype, indicating that the insertion hadoccurred within the oprF gene. Plasmid DNAs were pre-pared from 100 clones and digested with Pst\ to removethe kanamycin-resistance cassette but leaving a residueof 12 nucleotide pairs, which was between the H/r7cll sitesand Ps^ sites flanking both sides of the cassette. Afterreligation, 44 of the 100 kanamycin-sensitive clones hadregained the ability to express OprF, as determined bycolony immunoblot with monoclonal antibody MA5-8.Thus these clones represented insertion sites that werepermissive for the insertion of four extra amino acids. Therest of the kanamycin-sensitive clones were unable toexpress OprF or reacted only weakly with the OprF-spe-ciflc monoclonal antibody MA5-8 on colony immunoblots.Restriction enzyme analysis was used to map the inser-tion sites in each of the 100 plasmids, which categorizedthe sites in the 44 OprF-expressing plasmids into 10unique groups. The remaining 56 plasmids included anumber which demonstrated gene rearrangements ordeletions, probably owing to multiple cleavages of pRW3by the restriction enzymes prior to ligation with the muta-genesis cassette. Sequencing of the plasmid DNA from

Linker-insertion mutagenesis of OrpF 285

eight of the kanamycin-sensitive, OprF-non-reactivemutants revealed that two of the mutants had incorpo-rated the 12 nucleotide pair insert in a reading frame thatleads to the translation of a stop codon, while five othermutants represented deletions of part of the oprFsequence. Only one (pRW303} of the eight clones anal-ysed showed the incorporation of a 12 nucleotide pairinsert without any genetic alteration or change of readingframe. The inability of this clone to demonstrate an OprFpositive phenotype on colony or Western immunoblotsuggested the 'non-permissiveness' of this insertion site.

Characterization of linker insertion mutants

The exact linker-insertion sites in at least one representa-tive from each of the 10 groups of the OprF-expressingmutant pfasmids were determined by DNA sequencing.The nucleotide positions of the linker insertions and theidentities of the inserted four amino acids for 11 linkerinsertion mutants (including one non-permissive mutant)and one site-directed insertion mutant (pRW307) aresummarized in Table 1. In addition, two other plasmidscontained inserts at nucleotides 491 and 853 that weretranslated to premature stop codons (pRW304 and 315respectively) and thus were not further considered here.The expression of OprF from the 12 mutated plasmidswas examined by SDS-PAGE of outer membrane sam-ples containing the mutated proteins (Fig. 2). All of the

mutated proteins had mobilities that were modified bypretreatment with 2-mercaptoethanol, indicating that theinserted amino acid residues did not perturb the formationof the OprF cysteine disulphides. The apparent molecularmasses of two of the mutated proteins produced by plas-mids pRW301 and pRW305 were noticeably greater thanthat of the wild type (Table 1). However, the proteinexpressed by pRW305 ran with a mobility similar to theunfolded, heat-modified form of OprF as visualized on aWestern blot with an OprF-specific monoclonal antibody(Fig. 3, lane 3), suggesting that the four-amino-acid linkermay have influenced the susceptibility of the protein todenaturation by heating in SDS. Plasmids pRW309(Fig. 2. lane 9) and pRW311 {data not shown) directed theexpression of an intense band of apparent molecularweight 70000. probably corresponding to a protein OprFtrimer (Mutharia and Hancock, 1985). After 2-mercap-toethanol treatment, a much more intense monomer bandof apparent molecular weight 35 000 was observed in thesame samples (e.g. Fig, 2, lane 5). This suggested thatinsertion of the linker at amino acid positions 211 and 231may enhance the association of SDS-stable oligomers.Two of the mutated proteins, encoded by pRW302 and306, were produced in lesser amounts than the others, asdetermined by their abundance relative to the other E. coliproteins, indicating that insertions al these sites may leadto reduced protein production or unstable products. Thiswas not, apparently, caused by a growth defect since the

Table 1. Insertion siles of the 11 linker insertionmutants and one siie-directed insertion mutantand idenlities ol the inserted ammo acids.

Plasmids

pRW301pRW302pRW303pRW30SpRW306PRW3O7*pRW308pRW309pRW310pRW311pRW312pnW314

Insertionsites(nucleotide)^

1352062565215346947167687758229971059

Insertionsites(amino acid)

Gly-2Ala-26''Glu-42Ala-iat"Gly-135Val-188Ala-196"Afg-211Asp-215Ser-231'Arg-290Gly-310

Aminoacidsinserted

TCRSPAGPDLQVPAGPTCRSDLQVPAGPTCRSDLQVTCRSTCRSTCRS

Apparentmolecularmass''(kDa)

4136-

40"3535353535363528'

Apparent Molecular MassatterTrypsinTreatment" - (kDa)

outermembranes

ND36.28_

24.20.18"35,2824,3535,24,2828282828ND

wholecells

ND35-

40"353535353 6 ^NO2828

a. Position 1 IS ihe iranscrtption start siteb. The alanine residue at ihe insertion site was replaced by a glycine.c. The serine residue al the insertion site was replaced by an arginme.d. As estimated on Western immunobtot o( outer membrane samples with MA7-t. Molecular mass otunmodified OprF was 35 kDa: -. no OprF detected.e. pRW307 was generated by inserting a Sail adaptor that contained a Pst\ site into the Sal\ site cor-responding to amino acid 188.(. 35 kDa was obsen/ed as a minor band reactive with MA5-8g. Where more than one band appeared, they are listed m order of abundance, — no product: ND, notdetermined. pRW3-derived OprF gave a result identical to that of pRW306.h. Tested with monoclonal antibody MA4-4 since this mutant OprF derivative was non-reactive withMA7-t.

286 R. S. Y. Wong, K Jost and R. E. W. Hancock

1 2 3 4 5 6 7 8 9 10 reading frame occurred, it did not produce any forms ofOprF that could be detected by the OprF-specific mono-clonal antibodies or visualized in Coomassie briliiant blue-stained gels of outer membrane samples or whole-celllysates. Thus we assume that this represents a non-per-missive insertion site (Charbit etal., 1991).

Fig. 2. SDS-PAGE of ouler membrane samples of E. coli DH5aF' strainsexpressing tfie pRW3-derivative plasmids. Samples were incubated at37'C for 10 min in reduction mix (2% SDS. 10% glycerol, 62.5 mM TrispH6.8) wilh {lanes 2-8) or Without (lanes 9 and 10) 4% 2-mercap-toethanol before loading. Each lane contained -16 pg protein from eachsample. Plasmids present were lane 2, pRW302; lane 3, pRW305; fane4, pRW306; lane 5, pRW309; lane 6. pRW310; lane 7. pRW3; lane 8,pTZ19R; lane 9. pRW309 and lane 10. pRW310. Lane 1 contained themolecular weight markers; pfiosphoryjase B (106 000), bovine serumalbumin (80 000). ovaibumin (49500). carbonic anhydrase (32 500). soy-bean trypsin inhibitor (27 500), lysozyme (18500). OprF monomer bandsare indicated by triangles; the OprF SDS-stable tnmer is indicated by anarrow head.

growth rate of E. co//strains carrying parental (pRW3) orany of the mutated plasmids was not different. Mutantswith insertion sites in the C-terminal end of the proteins(e.g. those encoded by pRW312 and pRW314) producedOprF variants that were substantially but not completelydegraded to smaller fragments, which included a predom-inant 28 kDa fragment (Table 1), confirming the results ofFinnen et al. (1992), which indicated that the C-terminalregions are required for the resistance of the protein tocellular proteases.

Configuration of OprF in the outer membrane of E. coli

To permit conclusions to be drawn based on linker inser-tion mutagenesis in E. coli, regarding the structure ofOprF, it was necessary to examine whether the structureof OprF and its mutated variants in E. co//reflected that ofOprF in P. aeruginosa. It was noted that all OprF variantscofractionated with the E. co//outer membrane in fraction-ation experiments. In addition, both OprF and its variantsproduced in E. celi were 2-mercaptoefhanol modifiable,suggesting that they contained the usual cysteine disul-phide bonds observed with OprF in P. aeruginosa (Han-cock and Carey, 1979). In order to probe the configura-tions of the OprF derivatives in the outer membrane of E.coli, we also conducted trypsin-accessibility assays andindirect immunofiuorescence labelling.

Outer membrane porins tend to be protease resistant(Paul and Rosenbusch, 1985) by virtue of their posses-sion of extensive (i-structures with the linking surfaceloops tightly packed or folded in towards the porin chan-nel (Cowan et al.. 1992). It was previously demonstratedthat pure OprF or OprF In outer membranes was partlycleaved by trypsin to a core 28 kDa fragment, but thatincreasing concentrations of trypsin or increasing lengthof treatment time failed to cause further proteolysis

Monoclonal antibody reactivities of mutated OprFproteins

The mutated OprF proteins of outer membrane sampleswere further characterized by Western immunoblot andcolony immunoblot analyses, using a series of OprF-spe-cific monoclonal antibodies (Table 2). The results demon-strated that the OprF derivatives expressed by five of theplasmids (pRW301, 302, 306, 309 and 310) were reactivewith all 10 monoclonal antibodies, indicating the retentionof native OprF structure. In six other mutants, specificOprF epitopes were disrupted by the insertion of the four-amino-acid linker. However, the positive reactivities ofthese mutated proteins with the majority of the mono-clonal antibodies suggested the retention of substantialnative OprF structure in these mutants. OprF expressedby pRW303 was an exception. Despite the fact that DNAsequencing demonstrated that only 12 nucleotides wereinserted and no premature stop codons or changes in

Fig. 3. Western immunoblot analysis of outer membrane samples of E.coll DH5rtF' strains expressing the pRW3 derivative plasmids. The OprF-specific monoclonal antibody MA7-5. whose epitope is not interrupted bythe insertions, was used. Samples were heated at 90 C lor 10 mm inreduclion mix before loading. Each lane contained about 8 jig protein.Plasmids present were pRW302 (lane 2), pRW305 (lane 3), pRW306(lane 4), pRW309(lane5), pRW310 (Iane6). pRW3 (lane 7), pTZ19R(lane 8). Lane 1 contained the molecular weight markers as in Fig. 2.OpfF monomer bands are indicated by arrow heads.

Table 2. Summary of monoclonal antibody^ reac-tivities of mutated OprF variants contained inouter membrane samples.

Linker-insertion mutagenesis of OrpF 287

Plasmids

pRW301pRW302pRW303pRW305pRW306pRW307pRW308pRW309pRW310pRW311pRW312pRW314

insertion siles(amino acidposition) 7-1

Gly-2 WAla-28 +Glu-42Ala-131Gly-135 WVal-188 +Ala-196 +Arg-211 +Gln-215 +Ser231 +Arg-290 +Gly-310 +

7-2 7

+ H

+ -1

-

-t-

+-1-

-1-

+++

w

Monoclonal Antibody Reactivities"

7-4 7-5 7-6 7-7 7-8 4-4 5-8

W

a. The epitopes recognized by the monoclonal antibodies are: ammo acid 24 - amino acid 188 (MA7-1), amino acid 198 - ammo acid 275 (MA7-2), amino acid 188-ammo acid 245 (MA7-3). ammo acid188 - amino acid 275 (MA7-4. MA7-5. MA7-7), amino acid 198 - amino acid 240 (MA7-6}, amino acid178 - amino acid 187 (MA7-8), ammo acid 178 - amino acid 187 (MA4-4), amino acid 300 - aminoacid 320 (MA5-8), (Martin, 1992).b. Measured both by colony immunoblot and Western immunoblot of outer membrane samples. Sym-bols: +, reactivity equivalent to wild type (pRW3 plasmid-containing £ coli}. -. no reactivity; W, weakreactivity.

(Mutharia and Hancocl<. 1985). Trypsin treatment of outermembranes from E. coli DHSaF' containing the parentalpiasmid pRW3 resulted in substantial retention of full-sized OprF and partial proteolysis to a 28 kDa fragment(Fig. 4, cf. lanes 8, 9. and 10). Similar results wereobtained after trypsin treatment of outer membranes fromcells containing plasmids pRW302 and pRW306 with N-terminal insertions in OprF (Fig. 4, lane 1; Table 1). Plas-mids pRW309, pRW310, pRW311 and pRW312 with C-terminal insertions in OprF produced oniy the 28kDafragment after trypsin treatment (Fig. 4, lanes 4-7;Table 1). Plasmids pRW307 and pRW308, with insertionsin the central cysteine disulphide region resulted in pro-duction of an OprF derivative that was cleaved by trypsinto a 24 kDa fragment, instead of (pRW307; Fig. 4, lane 2)

or in addition to (pRW308; Fig. 4, lane 3) the 28 kDa frag-ment. Based on previous studies (Finnen et al.. 1992),such a 24 kDa fragment might be expected if cleavageoccurred near amino acid 200 within the cysteine disul-phide region, suggesting localized modification of OprFby this insertion which rendered the region susceptible totrypsin. Plasmid pRW305 directed expression of an OprFvariant, which showed unpredictable trypsin-cleavagepatterns, demonstrating a predominant 24 kDa, and minor20 kDa and 18 kDa products (Table 1). Trypsin treatmentof Intact cells of E. coli DH5aF' containing the above plas-mids did not result in proteolysis of OprF or its insertion-mutant derivatives (Table 1), with the exception of thoseC-terminal insertion mutants produced by plasmidspRW310andpRW312.

MWstd(kDa) 10

Fig. 4, Westem blot analysis of trypsinized outermembrane samples containing OprF vanants.Samples were treated with trypsin (0.1 mgml"') at37' C for 80 mm, and then heated at 88 C tor10 mm in reduction mix. The plasmids corre-sponded to the OprF variants contained in eachlane are: 1. pRW302: 2. pRW307: 3. pRW308; 4,pRW309; 5. pRW310; 6. pRW311; 7. pRW312; 8,pRW3: 9. pRW3; 10, pRW3 (untreated). OprFmonomer band is indicated by an arrow head.The 28 kDa and 24 kDa trypsin-resistant corefragments are indicated by open and solid triar-gles. respectively. Monoclonal antibody MA7-1was used for immunodetection.

288 R. S. Y. Wong, H. Jost and R. E. W. Hancock

To examine if the OprF variants are surface exposed,immunoftuoreseenee labelling was carried out usingselected plasmJd-containing E. coli strains and mono-clonal antibodies against surface epitopes from the /V-ter-mlnus (MA7-1), central region (MA7-8), and C-terminus(MA5-8) of OprF (Tables 2 and 3). In each case, regard-less of the trypsin susceptibility of the respective OprFvariants in intact celis, immunoreactivity followed pre-cisely the pattern observed in both colony immunoblotsand Western blots (Table 3). Taken together, the OprFvariants are probably inserted in the outer membrane inthe proper manner, as reflected by protection from trypsinand surface exposure.

Discussion

The results presented here demonstrate the constructionof 10 different permissive OprF variants (including onesite-directed mutant pRW307) with four extra amino acidslocated at sites spread throughout the OprF sequence. Inaddition, one partially permissive variant that was trun-cated by cellular proteases to a 28 kDa core in cells con-taining plasmid pRW314, and one non-permissive clonepRW303 were identified. The result for the plasmidpRW314-containing strain was reminiscent of the dataobtained for C-terminal deletion variants of OprF pro-duced by JnphoA mutagenesis and subcloning (Finnen etal., 1992).

To permit utilization of these data for building a revisedOprF model it was first necessary to demonstrate thatOprF adopted a configuration in E. co//identical or similarto that adopted In P. aeruginosa. It was previously shownthat OprF in E. coli was both heat- and 2-mercap-toethanol-modifiable (Woodruff etal., 1986; Finnen etal.,1992). implying the existence of extensive p-sheet struc-ture (Siehnel etal.. 1990) and the presence of intra-chaindisulphide bonds (Hancock and Carey. 1979; Siehnel efal., 1990; Finnen etal., 1992). respectively. Furthermore,

Table 3. Results from indirect immunofiuorescence labelling of intact E.co/'C388 celts containing various plasmids.

Plasmids

No plasmidpRW3PRW302pRW307pRW308pRW312

Immunofiuorescence wrth Monoclonal Antibodies"

MA7-1'' MA7-8''

a. ++. positive labelling; +. weak labelling; - . negative labelling.b. The epitopes recognized by these antibodies are listed in Ihe footnoteot Table2

OprF purified from E. coli strains demonstrated similarporin function to OprF from P. aeruginosa (Woodruff etal., 1986) and OprF had a similar shape-determining rolein an appropriate E. coli genetic background (Woodruffand Hancock, 1989). In addition, OprF was exported tothe E. coli outer membrane (Woodruff et al., 1986) andreacted in intact cells with monoclonal antibodies specificfor three surface epitopes (Finnen ef al., 1992) as con-firmed here (Table 3). Most of the above properties werealso shared by the OprF insertion mutants studied here.The demonstration of a trypsin-resistant core structure inmost variants in isolated outer membranes and the gen-eral resistance of OprF to trypsin cleavage in intact cells(Table 1, Fig. 4) argued against an aberrant configurationfor OprF, as did the apparently correct formation of cys-teine disulphides as judged by the normal 2-mercap-toethanol modifiability of the OprF variants. In addition,most clones reacted with monoclonal antibodies MA7-3,4, 5, 7 and 8 and MA4-4 which apparently recognize con-formational epitopes (Finnen et al., 1992; Martin, 1992).Furthermore, cofractionation with outer membranes andaccessibility to immunofiuorescence labelling in intactcells were consistent with correct surface localization ofthe OprF insertion variants. Taken together, the abovedata provide a compelling argument that the structures ofOprF and its variants in E. coli ciones are highly similar tothe structure of OprF in P. aeruginosa, with some local-ized perturbations.

The 10 permissive insertion-mutant derivatives of OprFdid vary somewhat when compared to unaltered OprFdespite their many similarities to wild type. For example,the insertion at amino acid two of the mature protein(pRW301) directed the expression of a 41 kDa band (datanot shown). Since one of the inserted amino acids wasarginine, it is possible that this 41 kDa band representedOprF with the signal sequence still attached, perhapsbecause of defective signal peptide processing (Li et al.,1988). Despite the observation that this OprF variant co-fractionated with the outer membrane, results obtained byindirect immunofiuorescence labelling with monoclonalantibody suggested that a derivative of this OprF variantwas not surface localized. Therefore this apparent local-ization may have been caused by contamination of theouter membrane fraction with inclusion bodies. PlasmidspRW302 and pRW306 directed the expression of loweramounts of OprF than the other ptasmids. Since the resid-ual OprF demonstrated wild-type properties and since no28 kDa protease cleavage fragment was observed inintact cells, we assume that these insertions influencedexpression or secretion rather than stability. PlasmidpRW305 directed the expression of an OprF variant ofmolecular mass identical to the heat-modified, unfoldedform of OprF and with a distinct trypsin-cleavage pattern.Thus we assume that this insertion influenced the stability

Linker-insertion mutagenesis of OrpF 289

of the profein to SDS, probably because of a slightchange in membrane configuration. Clones containingplasmids with C-terminal insertions had enhanced sus-ceptibility to trypsin cleavage in outer membranes, and toa lesser extent in intact cells. We assume that the inser-tions caused some localized disordering of the C-terminalregion of OprF, perhaps preferentially exposing a trypsin-cleavage site on the periplasmic side of OprF. PlasmidspRW307 and pRW308 led to OprF variants with a differ-ent trypsin-cleavage pattern in outer membranes, possi-bly owing to exposure in these clones of a trypsin-acces-sible cleavage site in or adjacent to the cysteinedisulphide loop. The insertions in pRW309 and pRW311appeared to promote OprF trimer stability in the absenceof 2-mercaptoethanol (Fig.1, lane 9). The insertion inplasmid pRW312 led to an OprF variant that had reducedstability to cellular proteases; however 50% or more ofthis OprF variant remained intact and this clone was reac-tive with the C-terminal specific monoclonal antibodyMA5-8 (Tables 1 and 2).

According to restriction enzyme site analysis, therewere 37 sites within the oprF gene that were potential tar-gets for the linker insertion mutagenesis, although we iso-lated insertions in only 13 of them in this study (includingtwo of them where the inserted 12 nucleotide pairs weretranslated to a stop codon). This could be because insuffi-cient clones were analysed to exhaust all the possibilities.Alternatively, it could be due to the incompatibility of theinserted amino acids with the local OprF sequences (e.g.insertion into [i-strands), which might then lead to lethalityof the corresponding clones or proteolytic degradation ofOprF (since relatively few non-permissive insertions werefurther analysed in this study). It is reasonable to specu-late that the nature of the amino acids in the linker mightaffect the result. For instance, we do not exclude the pos-sibility that cysteine residues in the linkers translated asTCRS could disrupt OprF structure if inserted near theproposed disulphide bonds of OprF. However, the obser-vation that the OprF variants expressed by pRW306 andpRW309 (insertion sites at amino acid 135 and 211respectively) migrated with similar mobility on SDS-poly-acrylamide gels, showed similar 2-mercaptoethanol mod-ifiability and reacted with MA7-8 and MA4-4, which recog-nize epitopes sensitive to reduction of the OprFdisulphide bonds, suggested that the cysteine residuepresent in the linkers of these variants did not participatein disulphide bonding with the endogenous cysteineresidues in the protein. Nevertheless, the inserted linkerscould cause minor perturbations in local conformation,even though our findings strongly suggested that theoverall wild-type arrangement of OprF was still main-tained. For example, the variant protein expressed bypRW309 showed a slightly different 2-mercaptoethanolmodifiability compared to the wild-type protein (Figs 2 and

3, lane 5), which may be due to a change of conformationin the disulphide bond region.

Overall, the above differences in properties betweenwild-type and insertion-mutant derivatives are similar tothose observed for LamB (Boulain ef al., 1986) and PhoE(Bosch and Tommassen, 1987). Interestingly, despitethese differences in properties, a recent three-dimen-sional structure of PhoE (Cowan etal.. 1992) was consis-tent with the placement of all known PhoE insertion siteswithin loop regions between p-sheet strands. Therefore,we propose that the insertion sites defined here wouldhave similar value in defining OprF topology.

Siehnel etal. (1990) previously presented a model forOprF based on circular dichroism data suggesting that62% of the secondary structure of OprF was in the form ofP-sheet (a value typical for outer membrane proteins(Cowan etal., 1992)), the apparent existence of two disul-phide bridges between the four cysteines of OprF (Han-cock and Carey, 1979) and the p-turn prediction rules ofPaul and Rosenbusch (1985). However, the assignmentof amino acids to transmembrane p-sheets or p-turns inthis earlier model was inconsistent with the data pre-sented here since the structural model of PhoE (Cowan etal., 1992) had proved the assumption that the sites ofinsertion of extra amino acids were the loops connectingadjacent transmembrane p-strands. Furthermore, thethree solved outer membrane porin structures to date(Cowan ef al., 1992; Weiss ef ai, 1991) reveal that thelongest loop regions are those exposed to the surfacerather than the periptasm. leading to the conclusion thatinsertion might occur preferentially at the surface. There-fore the primary assumption utilized in constructing ournew topological model of OprF (Fig. 5) was the placementof the 13 insertion sites. It was considered that the site ofamino acid number two in pRW301 was probablyperiplasmic since all outer membrane proteins studied todate have /V-termini that face the periplasmic side. Inaddition, the insertion in pRW303 was considered non-permissive since it resulted in no detectable product.Thus the site in pRW303 was placed within the mem-brane. Although pRW314 resulted in production of trun-cated OprF, its insertion site was within a flexible proteinsegment adjacent to a variable sequence (see below) andthus was placed at the surtace. The remaining nine siteswere placed in surtace loops. Six of these interrupted thebinding of specific monoclonal antibodies (Table 2). Ineach case, these sites were within the regions of OprF towhich these epitopes had been previously mapped(Finnen etal., 1992; Martin, 1992). Two of the insertionsinterrupted binding of four monoclonal antibodies, a resultconsistent with data indicating that these monoclonal anti-bodies MA7-3, 4, 5 and 7 recognize overlapping confor-mational epitopes on the surface of OprF (Finnen et al.,1992; Martin, 1992). In addition, MA7-8 and MA4-4 were

290 R. S. Y. Wong, H. Jost and R. E. W. Hancoct<

Y GIf. q L E

A NF ^ V T

T V I )D P A

C:C C; ND N a P ^

D t; H ' -̂C:C D \ f R | "VD

DT A

G Y

N "

P TS

AV

GY

GE

S

ATA

ND

AV

R P

Fig. 5. Proposed membrane topology of OprF,The insertion sites in the linker mutants aremarked by shaded boxes.

previously demonstrated to recognize overlapping butdiscrete epitopes within the cysteine disulphide region ofOprF, a result consistent with the data obtained for theOprF variants produced by pRW307 and pRW308, Theother four insertion sites did nol interrupt specific mono-clonal antibody-binding sites but were placed on the sur-face in keeping with the PhoE precedent discussed aboveand on the basis of preliminary information derived fromthe insertion of a malarial epitope into these sites (R, S. Y.Wong and R, E, W. Hancock, unpublished). The secondcriterion utilized for the model in Fig. 5 was based on thealignment of the oprF genes from P. aeruginosa, Pseudo-monas syringae and Pseudomonas ftuorescens (de Motet at., 1991). Since it is the surface regions of outer mem-brane proteins that undergo maximal variation (Sikkemaand Murphy, 1992), it was considered that the non-homol-ogous regions and/or small deletions should be preferen-tially located at surface loops. The third criterion was pre-diction of flexible segments by the Nowotny method usingthe computer program pc GENE. These flexible segmentswere considered to be candidates for loop regions. Thefourth criterion was the prediction from circular dichroismof 62% p-sheet in OprF (Siehnel etat., 1990), In fact theresultant model in Fig, 5 contains 56% p-sheet, and 61% ifthe central transmembrane domain is considered to be p-sheet. Finally, we included two disulphide bonds, a resultconsistent with 2-mercaptoethanol titration studies (Han-cock and Carey, 1979) and, based on previous moleculargenetic studies (Finnen et at., 1992). placed the first disul-phide between the first pair of cysteines. While havingsome similarities to the OmpF-like porin structural mod-els, we have drawn it as a two-domain structure since, inthe interconnecting regions of E. coti OmpA protein and

P, ftuorescens OprF. there are large proline, alaninerepeats that are reminiscent of hinge regions. In P. aerug-tnosa and P, syrtngae OprF the cysteine loop region hasapparently been inserted at this region, but we felt that the'hinge' concept should be retained.

The topological model in Fig. 5 will provide the basis forfuture testing. Having an understanding of how insertionsat different sites affect the structure and stability of theprotein, it will be interesting to study the influence of theinsertions on the porin function of OprF.

Experimental procedures

Bactertat strains and ptasmids

E. cott strain DH5«F'o8Od/acZAM15A(/acZy>l-a/-gF)U169endAI recA1 hsdFI17(rt, m^,')deof{ thi-1 supE44 X gyrA96retAI) was used as the host for transformations and as thebackground strain for the expression of plasmids encoding themodified OprF proteins unless otherwise stated. The plasmidpUC4KAPA in a background strain E. coli HB101 (F~supE44hsdS20{rB~me~)recA-13ara-14 proA2 tacYI galK2 rpsL20 xyl5mtt-1) was obtained from Dr J, Smit (University of BritishColumbia, Canada), It contained a 1,3 kb fragment, derivedfrom Jr\901, which encoded the enzyme aminoglycoside 3'-phosphotransferase (conferring kanamycin resistance), andwas flanked by symmetrical restriction enzyme-recognitionsites. For indirect Immunofluorescence labelling and trypsiniza-tion studies, E. coti strain C386 (tpp, ompA) (Woodruff andHancock, 1989) was used as the background strain in some ofthe experiments for the expression of modified OprF from plas-mids. When plasmids were present, media containing 75)ag ml"' of ampicillin or 50|.ig mP' each of kanamycin and ampi-cillin were used.

The OprF-encoding plasmid for linker mutagenesis was con-structed as follows. A Pst\~Sal\ fragment from plasmid

Linker-insertion mutagenesis of OrpF 291

pWW2200 (Woodruff and Hancock. 1989) containing theupstream sequence and W-terminus of the oprF gene wascloned into the vector pRK404 and was mutated by site-directed mutagenesis using the Mutagene kit (BioRad) in con-junction with the mismatch primer 5-'AAGTTCTGATAAGCT-TGCCACCCAA-3', This procedure introduced a H/ndlll siteinto the putative -10 region of the oprF promoter by adding aG:C nucleotide pair between nueleotides - 9 and -10 (using thenumbering of Duchene et al. (1988)), This had the effect ofsubstantially weakening the oprF promoter. A 0,37 kb H/ndlll/Smal fragment corresponding to the /V-terminus one-third ofOprF was cloned into pTZ18R to construct pHJ12. A 4.5 kbSmal fragment from pWW13 (Woodruff etat., 1986) containingthe rest of the oprFgene was then cloned into the Smal site otpHJ12 to construct pHJ14. The 1,45kb H/ndlll/Kpn! partialfragment containing the entire oprFgene from pHJ14 was thensubcloned into the vector pTZI 9R to obtain pRW3 (Fig, 1),

Linker-insertion mutagenesis

The plasmid pRW3 was linearized separately by partial diges-tion with restriction enzymes: Rsa\, Haelll, Tha\ or Atu\(Boehringer Mannheim), all of which leave blunt ends afterdigestion. In the cases of Alu\ and Tha\, partial digestions wereperformed in the presence of ethidium bromide (20 and50^gmr^ respectively) to improve chances of the recovery ofDNA molecules cut at a single site. After partial digestions, thereaction mixtures were loaded onto preparative agarose gelsand the full-sized linear form of the plasmid was isolated by elu-tion onto DEAE paper (Schleicher and Schuell). The four poolsof linearized pRW3, each corresponding to a separate restric-tion enzyme used, were ligated separately with a 1,3 kb H/ncllfragment containing the kanamycin-resistance cassette frompUC4KAPA. Following ligation and transformation, cells wereplated on Luria agar plates containing 50) jgmr ' each ofkanamycin and ampicillin. The doubly resistant colonies werefurther screened on colony immunoblots for loss of expressionof OprF using the monoclonal antibody MA5-8, which is spe-cific for the C-terminus of OprF (Finnen et ai, 1992), PiasmidDNA from the clones that did not express OprF was isolated bythe alkaline lysis method (Sambrook et al.. 1989), Theextracted plasmid DNA from each clone was then digested withPst\, which only recognized sites in the flanking sequences ofthe kanamycin-resistance cassette, and hence cleaved thecassette from the plasmid. Following re-ligation of the Pst\-digestion mixtures and transformation, cells were plated onampicillin-containing medium. Colonies whicfi appeared werescreened for the recovery of both immuno reactive OprF, usingmonoclonal antibodies and colony immunoblots (Mutharia

, and Hancock, 1985), and kanamycin sensitivity. The OprF-expressing, kanamycin-sensitive clones presumably containedmutated forms of pRW3 with a 12 bp insertion at sites originallyinterrupted by the kanamycin-resistance cassette, PlasmidDNA was prepared trom these clones and the insertion siteswere mapped by restriction pattern analysis by double diges-tion with Psft/H/ndlll and PstUSatl, where H/ndlll and Sa/Iwere unique sites outside and within the oprF gene codingsequence, respectively. Clones with the same restriction pat-tern were grouped and at least one representative from eachgroup was chosen for further analyses.

Construction ofpRW307

The plasmid pRW307 was constructed by inserting, into theSa/I site corresponding to ammo acid 188 in the OprFsequence, a self-hybridizing Sa/I adaptor otigonucleotide (5'-TCGACCTGCAGG-3') which carried a Pst\ site. As a result,the four amino acids DLQV were added after the valine residueat position 188,

DNA sequencing

DNA sequencing was performed to determine the exact posi-tion of the insertion sites of the representative linker-insertionmutants. Plasmid DNA from the corresponding kanamycin-resistant clones were used as templates. Template DNA wasprepared using Oiagen columns (Qiagen Inc) according to themanufacturer's protocols. The sequencing primers used were25-mer oligonucieotides (5'-ATGTAACATCAGAGAnTTGA-3'and 5'-TATGAGTCAGCAACACCTTCT-3') that hybridized toopposite strands of the kanamycin-resistance cassette,approximately 50 bp from the ends of the cassette (Oka et at.,1981), Automated DNA sequencing was carried out with theApplied Biosystems Incorporated (ABI) model 373A DNAsequencing system using polymerase chain reaction and dye-terminator chemistry as described in ABI's protocols.Sequence analyses were performed using the ABI 675 DNAsequence editor program,

Ctiaracterizatlon of outer membranes

E. coli strains containing the pRW3-derived plasmids weregrown overnight at 37"C in Luria broth supplemented with75(igml ^ ampicillin and 1 mM IPTG, Outer membrane sam-ples were isolated by selective Triton X-100 solubilization ofcell envelopes (Schnaitman, 1971), The final outer membranepellet from 50 ml of cells was resuspended in 500 \i\ of distilledwater. SDS-PAGE (Hancock and Carey, 1979) and Westernimmunoblotting (Mutharia and Hancock, 1985) utilized the pro-cedures described previously, loading respectively 16 and 8 fjgof outer membrane proteins into each lane. The OprF-specificmonoclonal antibodies used in these analyses were describedpreviously (Finnen et at., 1992),

Trypsinization studies

E. CO//strain C386 containing various pRW3-derived plasmidswere grown in Luria broth supplemented with 75Mgmr' ofampicitlin to an Agoo of 0-8- Samples of 1,5 ml of the cultureswere harvested and washed twice with 20 mM Tris-HCI pH7,4containing 5mM MgClg. The cell pellets were resuspended in500 |jl of the same buffer, Trypsin (TPCK treated, purchasedfrom Sigma Chemicals) was added at a final concentration of0,1mgmr^ of cell resuspension, followed by incubation at37'C for 60 min. Untreated samples were incubated in thesame conditions, except trypsin was omitted. Proteolysis wasstopped by heating at 88"'C for 10 min in solubilization-reduc-tion mix (Hancock and Carey, 1979). OprF in outer membranesamples was digested at a trypsin concentration of 0.1 mgmP'.The reactions were carried out as described above. In bothcases, the trypsinized samples were run on SDS-polyacry-lamide gels and analysed by Western immunoblotting with the

292 R. S. Y. Wong, H. Jost and R. E. W. Hancock

specified monoclonal antibodies. As controls, the cleavage ofOprF in R aeruginosa intact cells to a 28 kDa core fragmentand the complete cleavage of bovine serum albumin by trypsinto small-molecular-weight peptides were demonstrated.

Indirect immunofiuorescence labelling of intact cells

Immunofiuorescence labelling was performed as follows.Briefly, overnighf cultures of strains containing the specifiedplasmids were harvested and washed twice in PBS. Slideswere coated with poly-L-lysine (Sigma Chemical Co.; averagemolecular weighf -25000) by flooding wifh poly-L-lysine solu-tion ( I m g m r ' ) in a moist chamber for 15-20min and thenrinsed thoroughly with distilled water. Samples of washed cellswere smeared onto the poly-L-lysine coated slides and allowedfo air dry briefly. Slides were then incubated wifh OprF-specificmonoclonal antibodies in a 1/100 dilution in phosphafe-buffered saline (PBS) containing 1% fetal calf serum (FCS) for30 min at room temperafure. After washing with excess PBS,slides were incubated with 1/20 dilutions of fluorescein isothio-cyanate-conjugated goat anti-mouse IgG (BRL) in PBS/1%FCS for 30 min at room temperature. Following PBS washings,one drop of Sigma mounting medium was added to fhe slidesand the cells were examined under a Zeiss microscope fittedwith a halogen lamp, a condenser and filters for fluorescencemicroscopy at 525 nm for emission of fluorescein isothio-cyanate.

Acknowledgements

Funding for this research was provided in the early stages bythe Medical Research Council of Canada and subsequenfly bythe Canadian Bacterial Diseases Network. R. Wong was arecipient of the Natural Sciences and Engineering ResearchCouncil Studentship. We thank Mr M. D. Island for providingadvice regarding the method of linker insertion mutagenesisand Dr R. J. Siehnel for helpful technical advice and discus-sion. We also thank Dr J. Smit for assistance with the immuno-fiuorescence labelling of intact cell experiments.

References

Bellido, F., Martin, N.L., and Siehnel, R.J. (1992) Reevaluationin intact celis of the exclusion limit and role of porin OprF inPseudomor)as aeruginosa outer membrane permeability. JSacfer/o/174: 5196-5203.

Benz, R., and Hancock, R.E.W. (1981) Properties of the largeion-permeable pores from protein F of Pseudomonas aerug-inosa in lipid bilayer membranes. Biochim Biophys Acta 646:298-308.

Bosch, D., and Tommassen, J. (1987) Effects of linker inser-tions on the biogenesis and functioning of the Escherichiacoli outer membrane pore protein PhoE. Mol Gen Genet208: 485^89.

Boulain, J.C., Charbit, A., and Hofnung, M. (1986) Mutagene-sis by random linker insertion into the lamB gene ofEscherichia coliK^2. Mol Gen Genet 205: 339-348.

Charbit, A., Ronco, J., Michel, V., Werts, C , and Hofnung, M.(1991) Permissive sites and topology of an outer membraneprotein with a reporter epitope. J Bacteriol 173: 262-275.

Cowan, S.W., Schirmer, T., Rummel. G., Steiert, M., Ghosh,R., Pauptit, R.A., Jansonino, J.N., and Rosenbusch, J.P.

(1992) Crystal Structures explain functional properties of twoE. CO//porins. Nature 35B: 727-733.

De Mot, R., Proost, P., van Damme, J., and van Der Leyden, J.(1992) Homology of the root adhesion of Pseudomonas fluo-rescens OE 28.3 wifh porin F of Pseudomonas aeruginosaand Pseudomonas syringae. Mol Gen Genef 231: 489-493.

Duchene, M., Schweizer, A., Lottspeich, F., Krauss, G., Mar-get, M., Vogel, K., von Specht, B.-U., and Domdey, H.(1988) Sequence and transcriptional start site of the Pseu-domonas aeruginosa outer membrane porin protein F gene.J Bacterion70: 155-162.

Finnen, R.L., Martin, N.L., Siehnel, R.J., Woodruff, W.A.,Rosok, M.. and Hancock, R.E.W. (1992) Analysis of themajor outer membrane protein OprF from Pseudomonasaeruginosa using truncated OprF derivatives and mono-clonal antibodies. J Bacteriol 174: 4977-4985.

FreudI, R., Maclntyre, S., Degen, M., and Henning, U. (1986)Cell surface exposure of the outer membrane protein OmpAof Escherichia co//K12. J/Wo/e/o/188: 491-494.

Hancock, R.E.W., and Carey, A. (1979) Outer membrane ofPseudomonas aeruginosa: heat- and 2-mercaptoethanol-modifiable proteins. J6acteno/140: 902-910.

Li, P., Beckwith, J., and Inouye, H. (1988) Alteration of theamino terminus of the mature sequence of a periplasmicprotein can severely affect export in Escherichia coli. ProcNatl Acad Sci USA 85: 7685-7689.

Martin, N.L. (1992) Outer membrane proteins OprP and OprFfrom Pseudomonas aeruginosa: purification and analysis ofsecondary and tertiary structure. PhD Thesis, University ofBritish Columbia, Vancouver, British Columbia, Canada.

Mutharia, L.M., and Hancock, R.E.W. (1985) Characterizationof two surface-localized antigenic sites on porin protein F ofPseudomonas aeruginosa. Canad J Microbiol31: 381-386.

Oka, A., Sugisaki, H., and Takanami, M. (1981) Nucleotidesequence of the kanamycin resistance transposon Tn903. J/Wo/6/0/147:217-226.

Paul, C , and Rosenbusch, J.P. (1985) Folding patterns ofporin and bacteriorhodopsin. EMBO J A: 1593-1597.

Sambrook, J., Fritsch, E.F., and Maniatis, T. (1989) MolecularCloning: A Laboratory Manual. 2nd edn. Cold Spring HarborNew York: Cold Spring Harbor Laboratory Press.

Schnaitman, C.A. (1971) Solubilization of the cytoplasmicmembrane of Escherichia coli by Triton X-100. J Bacteriol108:545-552.

Siehnel, RJ., Martin, N.L., and Hancock, R.E.W. (1990) Func-tion and structure of the porin proteins OprF and OprP ofPseudomonas aeruginosa. In Pseudomonas Biotransforma-tions. Pathogenesis. and Evolving Biotechnology. Silver, S.(ed.). Washington, D.C: American Society for Microbiology,pp.328-342.

Sikkema, D.J., and Murphy, T.F. (1992) Molecular analysis ofthe P2 porin protein of non-typeable Haemophilus influen-zae. Infect Immun 60: 5204-5211.

Weiss, M.S., Abele, U., Weckesser, J., Welte, W., Schiltz, E.,and Schuiz, G.E. (1991) Molecular architecture and electro-static properties of a bacterial porin. Science 254: 1627-1630.

Woodruff. W.A., and Hancock, R.E.W. (1989) Pseudomonasaeruginosa outer membrane protein F; structural role andrelationship to the Escherichia co//OmpA protein. J Bacteriol171:3304-3309.

Woodruff, W.A., Parr, T.R., Hancock, R.E.W,, Hanne, L.,Nicas, T.I., and Iglewski, B. (1986) Expression in Escher-ichia coli and function of porin protein F of Pseudomonasaeruginosa. J Bacteriol IGl: 473-479.