overexpression and topology of the shigella flexneri o-antigen polymerase (rfc/wzy)

Overexpression and topology of the Shigella flexneriO-antigen polymerase (Rfc/Wzy)

Craig Daniels, Christofer Vindurampulle and RenatoMorona *

Microbial Pathogenesis Unit, Department of Microbiologyand Immunology, The University of Adelaide, Adelaide,South Australia, Australia 5005.

Summary

Lipopolysaccharides (LPS), particularly the O-antigencomponent, are one of many virulence determinantsnecessary for Shigella flexneri pathogenesis. O-antigenbiosynthesis is determined mostly by genes located inthe rfb region of the chromosome. The rfc/wzy geneencodes the O-antigen polymerase, an integral mem-brane protein, which polymerizes the O-antigen repeatunits of the LPS. The wild-type rfc/wzy gene has nodetectable ribosome-binding site (RBS) and four rarecodons in the translation initiation region (TIR). Site-directed mutagenesis of the rare codons at positions4, 9 and 23 to those corresponding to more abundanttRNAs and introduction of a RBS allowed detection ofthe rfc/wzy gene product via a T7 promoter/polymer-ase expression assay. Complementation studies usingthe rfc/wzy constructs allowed visualization of a novelLPS with unregulated O-antigen chain length distribu-tion, and a modal chain length could be restored bysupplying the gene for the O-antigen chain length regu-lator (Rol/Wzz) on a low-copy-number plasmid. Thissuggests that the O-antigen chain length distributionis determined by both Rfc/Wzy and Rol/Wzz proteins.The effect on translation of mutating the rare codonswas determined using an Rfc::PhoA fusion proteinas a reporter. Alkaline phosphatase enzyme assaysshowed an approximately twofold increase in expres-sion when three of the rare codons were mutated.Analysis of the Rfc/Wzy amino acid sequence usingTM-PREDICT indicated that Rfc/Wzy had 10–13 trans-membrane segments. The computer prediction modelswere tested by genetically fusing C-terminal deletionsof Rfc/Wzy to alkaline phosphatase and b-galactosi-dase. Rfc::PhoA fusion proteins near the amino-term-inal end were detected by Coomassie blue stainingand Western blotting using anti-PhoA serum. Theenzyme activities of cells with the rfc/wzy fusions

and the location of the fusions in rfc/wzy indicatedthat Rfc/Wzy has 12 transmembrane segments withtwo large periplasmic domains, and that the amino-and carboxy-termini are located on the cytoplasmicface of the membrane.

Introduction

Shigella is the aetiological agent of bacillary dysentery orshigellosis, which is of major concern in overcrowdedareas of the developing world (Hale, 1991). Lipopolysac-charides (LPS) are a major surface component that contri-bute to virulence by providing resistance to host defences(Lindberg et al., 1991) and are also directly involved in Shi-gella flexneri pathogenesis (Van den Bosch et al., 1997).LPS molecules consist of lipid A, a core sugar region, anda chain of polymerized sugar repeat units representingO-antigen. In S. flexneri, nearly all the genes involved inO-antigen biosynthesis are located in the rfb region onthe chromosome (Macpherson et al., 1994; Morona et al.,1994). The rfb region encodes enzymes required for thesynthesis of nucleotide diphosphate sugars, nucleotidesugar transferases and proteins involved in the assemblyof the O-antigen (Makela and Stocker, 1984; Macphersonet al., 1994). In addition, a rfe/wecA gene encoding anN-acetylglucosamine transferase is required for O-antigenbiosynthesis (Sandlin et al., 1995). Assembly of a tetra-saccharide repeat unit occurs on the lipid carrier bactopre-nol, and the RfbX/Wzx protein seems likely to transfer therepeat units to the periplasmic face of the cytoplasmicmembrane (Macpherson et al., 1995; Liu et al., 1996).Polymerization of O-repeat units into long-chain O-antigenis accomplished by the O-antigen polymerase, the rfc/wzygene product (Morona et al., 1994; Reeves et al., 1996).Biochemical evidence strongly suggests that polymeriza-tion occurs on the periplasmic side of the cytoplasmicmembrane (Mulford and Osborn, 1983). The number ofO-antigen repeat units polymerized is non-randomly distri-buted (<12–17 repeat units), and this modal chain lengthis governed in an unknown manner by the rol/wzz geneproduct (Morona et al., 1995).

A number of LPS phenotypes have been reported. Thewild-type LPS phenotype is referred to as smooth (S-LPS),with mutant LPS phenotypes referred to as rough (R-LPS)or semirough (SR-LPS). R-LPS consists of only the coreoligosaccharide and lipid A, and SR-LPS consists of lipidA, core oligosaccharide and a single O-antigen unit. In

Molecular Microbiology (1998) 28(6), 1211–1222

Q 1998 Blackwell Science Ltd

Received 26 November, 1998; revised 16 March, 1998; accepted 20March, 1998. *For correspondence. E-mail [email protected]; Tel. (8) 8303 4151; Fax (8) 8303 4362.

S. flexneri, the SR-LPS phenotype arises by mutation ofthe rfc/wzy gene (Morona et al., 1994). Finally, the LPSof rol/wzz mutants lacks a modal O-antigen chain lengthdistribution and the chains have a random length.

The rfc/wzy genes of many Gram-negative bacteria havenow been cloned and characterized (Collins and Hackett.,1991; Brown et al., 1992; Klena and Schnaitman, 1993;Morona et al., 1994; de Kievit et al., 1995; Lukomski etal., 1996). All have mol% GþC lower than their respectivechromosomal average, and the open reading frames(ORFs) have comparable amino acid composition andcodon usage. Predicted rfc/wzy gene products are highlyhydrophobic. Rfc/Wzy proteins are thought to be integralmembrane proteins as they have 11 or 13 putative mem-brane-spanning domains. Attempts to visualize and to eluci-date the subcellular location of the rfc/wzy gene productshave failed (Collins and Hackett., 1991; Morona et al.,1994; Lukomski et al., 1996). The inability to detect Rfc/Wzy has generally been attributed to: (i) poor translationinitiation of the mRNA as rfc/wzy genes generally haveno detectable ribosome-binding site (RBS) and (ii) the pre-sence of minor/rare codons in the first 25 amino acids ofthe protein, allowing ribosome stalling, which may preventfurther translation initiation. In fact, rare codons, which aredecoded by minor accepting tRNA species (Grosjean andFiers, 1982), are dispersed throughout the entire length ofthe rfc/wzy gene. Rare codons have been shown to modu-late gene expression in Escherichia coli, particularly whenlocated near the 58 end of a message (Chen and Inouye,1990; Goldman et al., 1995; Zahn and Landy, 1996).

In this study, we report for the first time the expressionand visualization of the S. flexneri rfc/wzy gene productafter site-directed mutagenesis of some of the rare codonsand introduction of a strong RBS. The effects of rfc/wzyoverexpression on O-antigen chain length were investi-gated. We also show quantitatively, by use of an Rfc::PhoAreporter, the effect on translation of mutating rare codonsin the first 25 amino acids of rfc/wzy. A revised topolo-gical model is presented based on analysis of data fromcomputer prediction models, and Rfc/Wzy::PhoA andRfc/Wzy::LacZ fusion experiments.

Results

Identification of Wzy by L-[ 35S]-methionine labelling

The wzy gene of S. flexneri has no detectable RBS andhas four rare codons in the first 25 amino acids of the cod-ing sequence (Morona et al., 1994). The rare codons in thefirst 25 of wzy are those encoding isoleucine-4 (ATA), iso-leucine-9 (ATA), glycine-22 (GGA) and arginine-23 (AGA).The wzy gene has 12% rare codons overall and thereforetranslation of the wzy gene under normal growth conditionsis likely to be poor.

To test whether Wzy protein synthesis is modulated by

the above-mentioned rare codons, codon variants in threeof the four early rare codons were constructed. In all ofthe constructs, wzy was placed downstream from a T7 pro-moter and the RBS was from gene 10 of T7 phage. Threeplasmids were constructed: (i) pRMCV5 has wild-type wzycloned into pET11bYZ; (ii) pRMCD10 differed from thenative wzy by substitution of ATT for ATA at codons 4and 9, and CGT for AGA at codon 23; and (iii) pRMCD6has the mutated wzy from pRMCD10 cloned into pBC-KS, with the 58 end adjacent to the T7 promoter. Theseplasmids were introduced into E2096, E. coli DH5 con-taining pGP1-2, which encodes T7 RNA polymeraseunder lambda cI control (Tabor and Richardson, 1985).Figure 1 shows that a 43 kDa protein was expressed byconstructs containing wzy with the isoleucine 4, 9 andarginine 23 codons mutated. A protein band migrating at<43 kDa was detected in E2096 (pRMCD10) and E2096(pRMCD6) (Fig. 1), the 43 kDa product was more easilydetectable in the latter strain. This corresponds to thesize of Wzy predicted from its sequence (43.7 kDa).Labelled bands of lower than expected molecular masswere also detected in samples of E2096 (pRMCD10) andE2096 (pRMCD6); these are possibly the result of eitherpremature termination or degradation of Wzy. The produc-tion of this protein was very poor when compared with thatof the b-lactamase encoded on the positive control plasmidpT7-3 (Tabor and Richardson, 1985) (Fig. 1). No 43 kDaprotein band was observed in either of the negative con-trols E2096 (pET11bYZ) and E2096 (pBC-KS). Samplesfrom the strain harbouring the wild-type wzy construct(pRMCV5) also did not have a detectable 43 kDa protein.In addition, pET11bYZ-based constructs containing wzywith either isoleucine-4 or arginine-23 codons mutatedalso did not produce a 43 kDa protein, however a constructwith both isoleucine-4 and isoleucine-9 codons mutateddid allow detection of the 43 kDa protein (data not shown).These data indicate that mutation of rare codons near theinitiation site of wzy and introduction of a strong RBSallowed increased expression and hence detection of thewzy gene product for the first time.

Complementation studies

The wzy plasmids described above were used in comple-mentation experiments to ensure that failure to detect Wzywith some plasmids was not due to either mutations intro-duced during construction or some other inability of theplasmids to express wzy. Complementation from the con-structs in wild-type S. flexneri relied on readthrough fromthe plasmid vector, and no T7 RNA polymerase was initi-ally used.

All vectors and wzy plasmids were introduced into theS. flexneri wzy mutant strain RMM109 (Morona et al.,1994). RMM109 is a spontaneous wzy mutant and has a

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 1211–1222

1212 C. Daniels, C. Vindurampulle and R. Morona

frameshift mutation at position 9214 in the wzy sequencethat results in premature termination of Wzy synthesis(data not shown). RMM109 has a SR-LPS and is resistantto infection by the smooth LPS-specific bacteriophageSf6c. LPS samples from trans-complemented strainswere compared with the LPS pattern of the parent strainPE638 on a SDS 20% polyacrylamide gel (Fig. 2A).Strains containing pET11bYZ clones of wzy (RMM109[pRMCV5] and RMM109 [pRMCD10]) gave comparablebut only partial complementation of the wzy mutant pheno-type (Fig. 2A) as substantial amounts of the SR-LPS werestill being produced. Plasmid pRMCV5, which did not pro-duce detectable Wzy by L-[35S]-methionine labelling, couldpartially complement the RMM109 wzy mutation, indicat-ing that functional Wzy was being produced from the con-struct. LPS of the strain RMM109 [pRMCD6] showedalmost complete complementation of the wzy phenotypeas it had very low levels of SR-LPS (Fig. 2A). Introductionof control plasmids (pET11bYZ and pBC-KS) into the par-ent (PE638) or mutant strain (RMM109) had no effect onthe LPS phenotype (Fig. 2A).

Wild-type S. flexneri (PE638) is sensitive to Sf6c bac-teriophage, whereas the wzy mutant strain (RMM109) is

resistant (Morona et al., 1994). Strain RMM109 (pRMCV5)was found to be partially resistant to the bacteriophage,and RMM109 (pRMCD6) was sensitive. This correlateswith the partial and almost complete complementation ofthe LPS phenotypes observed for strains carrying plas-mids pRMCV5 and pRMCD6.

In an attempt to fully complement the SR-LPS pheno-type of RMM109, we introduced the T7 RNA polymerasegene on plasmid pGP1-2 into strains PE638, RMM109,PE638 [pBC-KS], RMM109 [pBC-KS] and RMM109[pRMCD6]. To obtain better separation of LPS with longO-antigens, an SDS 15% polyacrylamide gel was used(Fig. 2B). Strains PE638 [pBC-KS], RMM109 [pBC-KS]and RMM109 [pRMCD6] gave LPS patterns similar tothat described above (Fig. 2B). Introduction of pGP1-2into PE638 [pBC-KS] and RMM109 [pBC-KS] had no effecton the LPS pattern (Fig. 2B). However, strain RMM109[pRMCD6] with pGP1-2 had an altered LPS pattern (Fig.2B). The LPS molecules had O-antigen repeat units thatwere greater than the modal length (12–17 repeats), withlittle or no banding detectable below 11 O-antigen repeatunits. The LPS profile of strains RMM109 [pRMCV5] andRMM109 [pRMCD10] was also altered and similar to


Fig. 1. Identification of the wzy gene product.This figure shows an autoradiograph ofL-[35S]-methionine-labelled proteins. E2096harbouring the indicated plasmid was grownand labelled with L-[35S]-methionine asdescribed in Experimental procedures. Thesamples as indicated on the figure wereelectrophoresed on an SDS 15%polyacrylamide gel. Arrowheads indicate theWzy (43.7 kDa) protein, and lower molecularmass protein products are indicated with anasterix. The location of b-lactamase (Bla) isindicated on the left side of the figure. Themigration positions of the molecular massstandards (Pharmacia) are indicated on theright side (in kilodaltons): soybean trypsininhibitor (20.1), carbonic anhydrase (30),ovalbumin (43), bovine serum albumin (67)and phosphorylase b (94).

Rfc/Wzy overexpression and topology 1213

RMM109 [pRMCD6, pGP1-2] when pGP1-2 was introducedinto these strains (data not shown). From these data wehypothesized that trans-complementation of the Wzydependent O-antigen polymerization outstrips the abilityof the chromosomally encoded Wzz to control chain lengthdistribution.

To investigate this phenomenon further, we used a cloneof the S. flexneri wzz gene on a low-copy-number plasmid(pRMCD100). In this construct wzz expression is driven bythe lac promoter and therefore Wzz should be constitu-tively produced in S. flexneri. Introduction of pRMCD100into RMM109 [pRMCD6, pGP1-2] led to restoration ofa modal chain length (Fig. 2C); introduction of controlplasmid pWSK29 did not alter the LPS pattern (Fig. 2C).The LPS O-antigen modal chain length of this strain(RMM109 [pRMCD6, pGP1-2, pRMCD100]) was, however,shorter than that of the parental S. flexneri strain (PE638)(Fig. 2C). Introduction of pRMCD100, pGP1-2 and pBC-KS

into PE638 also resulted in a decreased modal chainlength (data not shown). We suggest that under the condi-tions used, an increased Wzz dosage (due to pRMCD100)was able to compensate for increased Wzy dosage inthe trans-complemented wzy mutant strain RMM109[pRMCD6, pGP1-2].

Effect of rare codons on wzy translation

A Wzy::PhoA reporter was used to quantitatively examinethe effect of mutating the rare codons to those correspond-ing to more abundant tRNAs. The constructs consisted ofthe first 78 bp of wzy (encoding 26 amino acids, wzy8)fused directly to 8PhoA. 8phoA is the E. coli wild-type alka-line phosphatase gene lacking both the promoter and sig-nal sequence. Expression of the inserts is driven by the lacpromoter and the RBS is that of gene 10 from T7 phage.As described below, a fusion of 8PhoA to this point in


Fig. 2. Silver-stained LPS of the S. flexneri SR-LPS mutant complemented with wzy. LPS was prepared from the strains indicated in thefigure by proteinase K treatment, as described in Experimental procedures; each lane represents 1 ×108 cells.A. SDS 20% polyacrylamide gel.B. SDS 15% polyacrylamide gel; strains were grown for 16 h at 308C (LPS with increased O-antigen chain length is indicated by a bracket onthe right side of the figure).C. SDS 15% polyacrylamide gel.


Wzy resulted in a positive PhoA fusion with high alkalinephosphatase activity. The four plasmids constructed were(i) pRMCD59 containing the wild-type wzy8 fused to8phoA ; (ii) pRMCD60 had wzy8 with the isoleucine-4codon altered from ATA to ATT; (iii) pRMCD62 had wzy8

with both isoleucine-4 and isoleucine-9 altered from ATAto ATT, and (iv) pRMCD64, which differed from pRMCD62by the additional substitution of CGT for AGA at position23.

Figure 3 shows results from PhoA assays performed onbacteria grown in either the presence or the absence of0.2% (w/v) glucose. The addition of glucose, results indecreased expression from the lac promoter on the vectorbecause of catabolite repression. In the presence of glu-cose, the ATA→ATT mutation at codon 4 of pRMCD60resulted in a 50% increase in PhoA activity comparedwith that of the wild-type wzy8 construct (pRMCD59), indi-cating an increase in Wzy::PhoA expression. Further sub-stitutions of the isoleucine codon at position 9 (pRMCD62)and the arginine codon at position 23 (pRMCD64) do notalter further the PhoA activity. In the absence of glucose,the single isoleucine-4 codon change (pRMCD60) resultedin a twofold increase in detectable PhoA activity when com-pared with the fusion with wild-type wzy8. The constructscontaining further substitutions (pRMCD62, pRMCD64)show a slight increase in the PhoA activity over thatfound with pRMCD60. These data show that mutation ofrare codons in the first 25 amino acids can increase thetranslation of wzy::phoA fusions. However, under the con-ditions used, the increase is limited and similar to thatobserved in other systems (Zahn and Landy, 1996).

Computer-generated models and PhoA fusions

Previous studies indicated that S. flexneri Wzy was a highlyhydrophobic cytoplasmic membrane protein (Morona et al.,1994). The model for S. flexneri Wzy had 13 transmembrane

segments with the amino-terminus in the periplasm andthe carboxy-terminus in the cytoplasm (Morona et al.,1994). We have reanalysed the primary amino acid sequ-ence of the S. flexneri Wzy using the TM-PREDICT package(Hoffman and Stoffel, 1993). The program predicted 13inside-to-outside helices, one of which was consideredinsignificant, and 12 outside-to-inside helices, with onebeing considered insignificant. Two models were gene-rated from TM-PREDICT when the transmembrane helixlength was set between 17 and 23 amino acids. Both mod-els had 10 transmembrane helices with the ‘strongly pre-ferred model’ having the amino and carboxy termini inthe periplasm and the ‘alternative model’ with both terminiin the cytoplasm.

To test these models, we constructed fusions of wzy tothe gene encoding the E. coli alkaline phosphatase (Man-oil, 1991). The wzy gene from pRMCV5 was cloned viaXbaI–BamHI into the 8phoA fusion vector pRMCD28, pro-ducing pRMCD29. Wzy::PhoA fusions were generated bysequentially deleting the carboxy terminal end of the wzygene and joining these at random to the truncated 8phoAgene (Experimental procedures). Forty-eight PhoA-posi-tive (blue) transformants were characterized using poly-merase chain reaction (PCR) and DNA sequencing.They were found to contain in-frame wzy::phoA fusionsat 28 different sites in wzy.

The alkaline phosphatase enzyme activities of DH5a

cells expressing the wzy::phoA fusions were determined.The location of the fusion junctions and their activitiesare shown on the topological model of Wzy (Fig. 4). Over-all the activities correlate well with the ‘alternative model’predicted by the TM-PREDICT program. The model presentedin Fig. 4 has 12 transmembrane helices, whereas both TM-

PREDICT models had only 10. In general, the periplasmicallylocated alkaline phosphatase fusions have five- to 10-foldgreater activity than those with junctions in putative mem-brane-spanning domains (e.g. compare fusion 3 with 4,


Fig. 3. Wzy::PhoA reporter synthesis is increased by mutating rare codons. The DNA sequences for each construct are indicated with codonchanges from wild type shown as underlined bold letters. PhoA assays were performed in strain DH5a [with or without 0.2% (w/v) glucose,to assess the effects of catabolite repression on expression from the constructs] three times, in duplicate, as described in Experimentalprocedures. Standard PhoA units are presented 6 standard deviation and ratios represent the increase in PhoA activity over that of thewild-type activity.


and 27 with 28). Each putative periplasmic region wascharacterized by at least one positive PhoA fusion (mosthave multiple fusions), and in general fusions havingincreased Wzy length show decreased levels of alkalinephosphatase activity.

Generation of wzy::lacZ fusions

Owing to the amount of screening required, the in vitrodeletion method used to generate wzy::phoA fusionsmakes it inefficient to isolate negative (white) wzy::phoAfusions that are in-frame. Because of this, we decided tocomplement the wzy::phoA data by generating wzy::lacZfusions. The Wzy::LacZ fusions would allow confirmationthat the positive PhoA fusions were genuinely exportedto the periplasmic space and complete the Wzy topologymodel by mapping the cytoplasmic face of the protein(Manoil, 1991).

Two types of wzy::lacZ fusions were generated by PCRtechniques. The first type was fusion switch constructs andwas generated by using an oligonucleotide primer thatbinds over the 58 end of 8phoA and incorporates a SmaI

restriction site. PCR using this primer and the M13rev pri-mer on pre-existing wzy::phoA constructs allowed genera-tion of products that could be cloned into the LacZ fusionvector pRMCD70. Eight wzy::phoA fusions (with at leastone fusion in five of the six periplasmic segments) wereswitched to lacZ fusions by this method (fusions 7, 8, 10,13, 14, 19, 22 and 27 were switched, Fig. 4). Attempts toswitch fusion numbers 1 and 3 were unsuccessful. b-Galac-tosidase activities of CC118 E. coli cells harbouring thesewzy::lacZ fusion constructs were measured. All fusionsgenerated by this technique gave less than 10 units ofb-galactosidase activity when assayed, regardless oftheir previous alkaline phosphatase activity. These datastrongly indicate that the positive Wzy::PhoA fusions areindeed in periplasmically located segments.

The second type of wzy::lacZ fusions were predictedto be cytoplasmically located and hence LacZ positive ifthe model in Fig. 4 was correct. Specific oligonucleotideprimers were designed to the five putative cytoplasmicsegments and to the carboxy terminus of wzy. The sixwzy::lacZ fusions generated by this method were LacZpositive, and when assayed had b-galactosidase activities


Fig. 4. Topological model of S. flexneri Wzy based on wzy::phoA and wzy::lacZ fusion analysis, and computer prediction data. Amino acidsare represented by their single letter code (GenBank accession number X71970). Residues flanking transmembrane regions are indicated withtheir primary amino acid number subscripted (e.g. I4 and G22). Wzy::PhoA fusion positions are marked at the terminal amino acid of the Wzyprotein and the alkaline phosphatase units (three duplicate assays, no added glucose) for these fusions are represented as 3591 (fusionnumbers are subscripted). Fusions switched from PhoA to LacZ are indicated with an asterix. Wzy::LacZ fusion positions and b-galactosidaseunits (three duplicate assays with cells grown in 0.2% (w/v) glucose) are indicated in their relative positions as 1802 (fusion numbers aresubscripted).


many times greater than those that were switched fromWzy::PhoA-positive to Wzy::LacZ fusions (Fig. 4).

Using the Kyte–Doolittle and TM-PREDICT data, and thewzy::phoA and wzy::lacZ data, we propose that the Wzyprotein is inserted in the cytoplasmic membrane as indi-cated in Fig. 4. Ten of the 12 transmembrane regions areas in the ‘preferred’ topology model predicted by the TM-

PREDICT program, and interestingly transmembrane regions209–226 and 353–370 were also predicted by TM-PREDICT

as being transmembrane helices (data not shown) butnot used by the program. In summary, the model has sixperiplasmic segments, two of which are relatively large(i.e. 21 and 53 amino acids), and five cytoplasmic segments,with the amino and carboxy termini in the cytoplasm.

Detection of Wzy::PhoA proteins

To assess the stability and subcellular location of theWzy::PhoA fusion proteins, Western blotting was performedusing anti-PhoA serum. Whole cells and cell fractions(cytoplasm, periplasm, and whole membrane) of sevenof the PhoA fusions were tested. Only Wzy::PhoA fusions

3 (aa 26), 6 (aa 93) and 8 (aa 167) were detectable, andWzy::PhoA fusion proteins greater in size than fusion 8,including fusions 10 (aa 207), 14 (aa 252), 17 (aa 260)and 27 (aa 350), were not detected (Fig. 5). Fractionationof the cells showed that Wzy::PhoA fusion proteins corre-sponding to fusions 3, 6 and 8 were located solely in themembrane fraction (data not shown). Some breakdownproducts (including PhoA monomer) of larger fusions 10,14, 17 and 27 were detectable by Western blotting of over-loaded whole-membrane preparations (data not shown).Surprisingly, Wzy::PhoA fusion proteins corresponding tofusions 3, 6 and 8 were easily detectable on Coomassieblue-stained SDS polyacrylamide gels (data not shown).The facile detection of Wzy::PhoA fusion proteins by Wes-tern blotting and Coomassie blue staining contrasts with ourdifficulties when attempting to detect intact Wzy. This sug-gests that rare codons throughout the wzy open readingframe may be important for translation of larger segmentsof wzy ; the decrease in production of larger Wzy::PhoAfusion proteins is consistent with this notion.

Interestingly, we observed that Wzy::PhoA fusion proteins3, 6 and 8 were able to dimerize, even under denaturing/


Fig. 5. Western immunoblot of E. coli DH5astrains producing Wzy::PhoA fusion proteins.After electrophoresis on a SDS 10%polyacrylamide gel and transfer tonitrocellulose, proteins were detected, usinganti-PhoA serum as described inExperimental procedures. Lanes containwhole-cell samples (equivalent to 1 ×108 cells)of strains loaded as indicated on the figure.Possible dimeric forms are indicated by anopen arrowhead (fusion 8 dimer is notmarked). Migration positions of molecularmass standards as in Fig. 1 are indicated inkilodaltons on the right side.


reducing conditions (Fig. 5). This dimerization appeared tobe Wzy mediated because no similar dimers were detect-able in the PhoA-positive control (E. coli K-12 strain C75A)(Fig. 5).

Discussion

Regulation of gene expression at the level of translationappears to be multifactorial. Previous studies have shownthat by manipulation of codons and translation initiation,gene expression can be altered such that it increases ordecreases (Chen and Inouye, 1990; Rosenberg et al.,1993; Zahn and Landy, 1996). In this study, we attemptedto increase the level of production of the wzy gene productof S. flexneri. Previous attempts to detect the wzy geneproduct were unsuccessful and found that the gene lackeda RBS and contained a high percentage of rare codons inthe first 25 amino acids of its sequence and throughout thecoding region (Morona et al., 1994). It was hypothesizedthat the absence of a RBS, combined with the presenceof rare codons, caused a low level of wzy gene expressionbecause of the inability of translating ribosomes to clearthe translation initiation region (TIR). We found that theinsertion of a RBS in front of wzy, in conjunction with muta-genesis of at least two of the four rare codons present inthe first 25 amino acids, resulted in a significant increasein gene expression and hence allowed detection of Wzyprotein. Only very low levels of the Wzy protein weredetectable with the pET11bYZ-based clone, with more inthe pBC-KS clone; this may be due in part to plasmidcopy number and/or stability of the plasmids as pET11-bYZ confers ampicillin resistance and pBC-KS conferschloramphenicol resistance. Truncated protein productswere visible within the pBC-KS based wzy (pRMCD6) con-struct (Fig. 1, lane 6); it is not known if these are caused bypremature termination of protein production or degradationof the full-length Wzy gene product. Trans-complementa-tion of the spontaneous wzy mutant RMM109 indicatedthat the wzy plasmids constructed were all able to producea functional Wzy protein. The pBC-KS-based constructgave more complete complementation (based on LPSphenotype and bacteriophage sensitivity) than the pET-based construct, which correlated well with the L-[35S]-methionine labelling data.

The LPS phenotype resulting from overexpression of Wzy(due to the presence of pGP1-2 and pRMCD6) has notpreviously been reported. In fact, these data are contraryto those previously published by Lukomski et al. (1996),who found that E. coli O4 wzy on high-copy-number/strong promoter constructs led to very poor complementa-tion. They predicted that this was due to the limited pool ofrare tRNAs being exhausted before complete Wzy proteinis translated. Our data indicate that this is not the casein S. flexneri, and we hypothesize that Wzy-dependent

polymerization may become uncoupled from the Wzz-mediated O-antigen modal chain length determination.We were able to restore a modal chain length by the intro-duction of a plasmid having the S. flexneri wzz gene(pRMCD100). The new modal length was slightly lower(10–15 repeats) than that of the wild-type strain (12–17repeats). A reduced modal length has previously beenreported for the KLPS of E. coli O8: K87 when its parentalwzz (cld/rol ) gene was introduced on a plasmid (Franco etal., 1996). These phenomena are difficult to explain mech-anistically, however they do indicate the importance of abalance between the O-antigen polymerase (Wzy) andthe regulator of O-antigen chain length (Wzz) in determin-ing the distribution of the repeat unit length. This supportsour previous hypothesis that the modal O-antigen chainlength is determined by the ratio of interacting components(Wzy/Wzz/WaaL) (Morona et al., 1995).

Gene fusion constructs have previously been used tomeasure the effects of rare codons on gene expression(Spanjaard and Van Duin, 1988; Zahn and Landy, 1996).Zahn and Landy used int gene–b-galactosidase fusionsto detect frameshifts caused by tandem rare Arg codons.In this study, we successfully used a short Wzy::PhoAfusion (26th amino acid) to quantitatively measure the effectof mutating rare codons to those of more abundant tRNAspecies. The results indicated that mutation of the fourth(isoleucine) codon (ATA→ATT) was sufficient to increasethe level of Wzy::PhoA fusion protein production (twofold).In the study by Zahn and Landy it was found that mutationof a single rare Arg codon at position 3 in the int genesequence could significantly enhance Int synthesis (two-to fourfold). The lack of further increases in Wzy::PhoAprotein production upon mutation of the isoleucine-9 andarginine-23 codons may be due to the proximity of thesecodons to the RBS, i.e. ribosomes stalled at positions 9or 23 may not be close enough to the RBS to prevent anew ribosome from binding.

We used the TM-PREDICT program (Hoffman and Stoffel,1993) to analyse the secondary structure of the Wzy pro-tein. The possible models generated for the S. flexneriWzy protein by the TM-PREDICT analysis agreed stronglywith data previously generated using the Kyte and Doolittleprocedure (Morona et al., 1994). The enzyme activities ofthe set of Wzy::PhoA fusions are consistent with the topo-logical model predicted for the Wzy protein. The isolationof fusion 8 (Fig. 4) showed that a region previously con-sidered to be a membrane-spanning region was actuallya periplasmic segment. Analysis of amino acids in thisdomain with the parameters suggested by Kyte and Doo-little (1982) resulted in the identification of this region asbeing relatively hydrophobic. Reanalysis of this regionwith TM-PREDICT indicated that the region was hydrophilic,agreeing with the location and activity of fusion 8. Fusion13 (9 PhoA units) is an anomalous fusion as it is flanked



by fusions of high activity [fusion 12 (100 units) and fusion14 (182 units)]. Fusions of this nature have previously beenreported (Calamia and Manoil, 1990) and they underscorethe importance of having more than one fusion in eachperiplasmic domain and/or fusions on both sides of themembrane. The wzy::phoA to wzy::lacZ switch fusionsgenerated in this study show the inverse relationshipbetween the activities of the two fusion types (i.e. highPhoA units switch to low LacZ units). Positive wzy::lacZfusions indicated that the protein traverses the cytoplas-mic membrane and the LacZ is located in the cytoplasm.The LacZ data are critical in the generation of the finalmodel because the process used to isolate wzy::phoAfusions in this study does not readily allow the isolationof negative (hence cytoplasmic) fusions.

Only relatively small Wzy::PhoA fusions could be visual-ized using Western blotting; fusions containing any part ofthe large periplasmic segment (53 amino acids) were notstable. We could detect small amounts of 8PhoA in thelanes of larger Wzy::PhoA fusions (fusions 14, 17 and 27)indicating that degradation had occurred. This may beone of the reasons why the Wzy protein was difficult todetect by L-[35S]-methionine labelling, i.e. there were smallamounts of the protein synthesized and it was veryunstable. The facile detection of the small Wzy::PhoAfusion proteins by Western blotting (Fig. 5) somewhat con-tradicts the hypothesis that rare codons restrict expressionof wzy, however, this system is a synthetic one and doesnot represent the production of full-length Wzy protein,and it is also possible that PhoA is able to stabilize the trun-cated Wzy. The dimerization of the detectable Wzy::PhoAfusion proteins observed under denaturing/reducing con-ditions (Fig. 5) appears to be mediated by the first trans-membrane region of the Wzy protein, and although thismay be non-specific, it suggests that native Wzy is ableto dimerize.

The model for the Wzy membrane topology presented inFig. 4 is based on the experimental and computer predic-tion data. Two large periplasmic segments are predicted inthe model. This was expected as the Wzy O-antigen poly-merase should have some major periplasmic regionsbecause O-antigen polymerization is considered to occuron the periplasmic face of the cytoplasmic membrane(Mulford and Osborn, 1983). The topology model for theO-antigen polymerase showing 12 transmembrane-span-ning regions is highly reminiscent of the structure of 12 seg-ment cytoplasmic membrane transporters (Saier, 1994).We have observed a low degree of sequence similaritybetween O-antigen polymerase homologues, putative O-antigen transporter/flippase homologues, transport per-meases and proteins of the electron transport chain (datanot shown). This similarity is generally between hydrophobicregions of the protein. The structure and sequence similari-ties may be fortuitous, but the possibility that the O-antigen

polymerase is also a permease is not an unreasonablehypothesis. The O-antigen polymerase may use the protonmotive force/electrochemical gradient to drive O-antigenpolymerization and couple this to proton or ion transport.Furthermore, it may act to pump bactoprenol-phosphateback to the cytoplasmic side of the membrane after theO-antigen repeat unit has been transferred to the growingO-antigen chain at the reducing end. This would recyclethe lipid carrier for use in further rounds of tetrasacchariderepeat unit synthesis, followed by transport by the O-anti-gen transporter (RfbX/Wzx) (Liu et al., 1996). Furtherstudies will be required in order to ascertain the subcellularlocation of the polymerization step, to elucidate the func-tional domain(s) of the Wzy protein, and its mode of action.

Experimental procedures

Bacterial strains and culture conditions

Table 1 lists the strains used in this study. All strains were grownat 378C in Luria broth (Morona et al., 1994), except for strainscontaining pGP1-2 (308C). Antibiotics were used at the follow-ing concentrations when appropriate: ampicillin (Ap) 100 mgml¹1; chloramphenicol (Cm) 25 mg ml¹1; kanamycin (Km)50 mg ml¹1; and rifampicin (Rif) 200 mg ml¹1. Chromogenicsubstrates 5-bromo-4-chloro-3-indolyl-phosphate (X-pho) and5-Bromo-4-chloro-3-indolyl b-D-galactopyranoside (X-gal) wereused at 40 mg ml¹1 in Luria plates.

DNA methods

Table 2 lists the plasmids used in this study. Plasmid DNAwas prepared by the alkali lysis technique, and restrictionenzyme digestion, agarose gel electrophoresis, ligation, elec-troporation and transformation were performed as describedpreviously (Morona et al. 1994; 1995). Expression vectorpET11bYZ was constructed by insertion of a 2.2 kb fragmentcontaining four T7 terminators (EcoRI–HindIII) from vectorpYZ100 (Yan Zheng-Xin, Max-Planck-Institute for Biology,Tubingen, Germany) between the EcoRI and HindIII sites ofpET11b. The base vector for pRMCD28 and pRMCD70 waspWSK29, which has a pSC101 ori and hence has a low copynumber (Wang and Kushner, 1991). A 58 truncated 8phoAPst I–XhoI fragment from pCH39 (Hoffman and Wright,1985) was treated with T4 polymerase and ligated into pBlue-script digested with HincII–XhoI (C. Clark, unpublished). Thisconstruct was digested with Pst I–XhoI and the 8phoA frag-ment cloned into similarly digested pWSK29, creatingpRMCD28. pRMCD70 has a truncated 8lacZ (EcoRI–DraI)from pRS414 (Simons et al. 1985) cloned into EcoRI–HincIIdigested pWSK29. Unidirectional deletions were generatedin plasmid DNA using the Erase-a-base kit from Promega.Briefly 2.5 mg of pRMCD29 DNA was digested with HindIII,protected with a-phosphorothioate nucleotides before beingdigested with SmaI and treated with Exonuclease III then trea-ted with Klenow and S1-nuclease, and the plasmids were thenreligated with T4 DNA ligase. After transformation into DH5a,positive PhoA fusions were detected by plating onto Luria



agar containing the chromogenic substrate X-pho. Sequenc-ing of fusion constructs was carried out using dye terminatorsequencing reactions (AB, Applied Biosystems) using theM13 reverse primer and no. 2211 58-CCATCGCCAATCAGC-AAA-38 (phoA) or no. 2254 58-TACGCCAG CTGGCGAAAG-38 (lacZ ).

Polymerase chain reaction and site-directedmutagenesis

PCR amplification was performed using standard protocolswith Amplitaq DNA polymerase (Hoffman-La Roche). Wild-type wzy clones were generated by PCR using primers no.859 58-CATCACCTTACGCATATGAATAATATA-38 and no.860 58-TTCCCTATTTTTGGATCCTTTATTTTGCTC-38 andcloning into pET11bYZ using NdeI and BamHI (underlined).Site-directed mutagenesis of the three rare codons was per-formed by overlap extension using PCR and three pairs ofcomplementary primers (DNA sequences are available onrequest). Primer pairs were used in combination with flankingprimers to generate mutated fusion products that were thenfused using PCR to form the intact wzy gene. The mutations

were confirmed by DNA sequencing using the T7 promoterprimer, as recommended by AB.

Alkaline phosphatase and b-galactosidase assays

Alkaline phosphatase assays were performed by a methodmodified from that previously described by Manoil (1991).Cultures were grown in LB with Ap at 378C for 16 h then sub-cultured 1:20 into fresh LB containing Ap and incubated untilthe O.D.600 <0.5. A 1 ml sample of culture was centrifugedfor 3 min at 14 000 r.p.m. and washed in cold 10 mM Tris-HCl, pH 8.0, 10 mM MgSO4 and the final pellet resuspendedin 1 ml of cold 1 M Tris-HCl, pH 8.0, 1 mM iodoacetamide.The OD600 was measured by placing 300 ml in the well of a96-well microtitre tray and reading on a DYNATECH MR7000 microplate reader. Washed culture (200 ml) was addedto 800 ml of 1 M Tris-HCl, pH 8.0, 0.1 mM ZnCl2, 1 mM iodo-acetamide and permeabilized by the addition of 50 ml 0.1%SDS and 50 ml of chloroform. An aliquot (275 ml) of perme-abilized cells was placed in duplicate wells of a microtitretray and the reaction was started by the addition of 25 ml of0.4% p-nitrophenyl phosphate (in 1 M Tris-HCl, pH8.0). Optical


Table 1. Bacterial strains.

Strain Characteristics Source/reference

E. coli K-12DH5 F¹ recA1 endA1 hsdr17 (rk

¹ mkþ) supE44l- thiI gyrA relA1 Bethesda Research Laboratories

DH5a endA hsdR supE44 thi-1 recA1 gyrArelAD (lacZYA-argF ) U169 [f80 dlacD (lacZ ) M15] phoA Bethesda Research Laboratories

CC118 D(ara leu) 7697 DlacX74 DphoA20galE galK thi rpsE rpoB argE recA1 Manoil and Beckwith (1985)

C75A tonA22 pho-64 ompF627 (T2R) relA1 pit10 spoT1 B. Bachman CGSC no. 5978a

E2096 DH5 þ pGP1-2 Laboratory collection

S. flexneriPE638 S. flexneri Y rpoB (RifR) Morona et al. (1994)RMM109 PE638 rfc/wzy Morona et al. (1994)

a. CGSC, E. coli Genetic Stock Centre, Yale University, New Haven, CT, USA.

Table 2. Plasmids.Plasmid Details Source

pET11b ApR, expression vector Rosenberg et al. (1987)pYZ100 T7 polymerase terminator vector Yan Zheng Xina

pBC-KS CmlR derivative of pBluescript StratagenepWSK29 Low-copy-number pBluescript derivative Wang and Kushner (1991)pRS414 lac operon fusion vector Simons et al. (1987)pGP1-2 T7 RNA polymerase KanR Tabor and Richardson (1985)pT7-3 bla T7 positive control ApR Tabor and Richardson (1985)pET11bYZ pET11b with T7 terminator of pYZ100 This studypRMCV5 Wild-type wzy in pET11bYZ This studypRMCD10 wzy ile-4, -9 and arg-23 in pET11bYZ This studypRMCD6 wzy ile-4, -9 and arg-23 in pBC-KS This studypRMCD28 E. coli ‘phoA in pWSK29 This studypRMCD29 pRMCD28 with wzy from pRMCV5 This studypRMCD70 E. coli lacZ in pWSK29 This studypRMCD100 S. flexneri wzz in pWSK29 This studypRMCD59 pRMCD28 with 78 bases of wild-type wzy This studypRMCD60 pRMCD59 with ile-4 mutated wzy This studypRMCD62 pRMCD59 with ile-4, -9 mutated wzy This studypRMCD64 pRMCD59 with ile-4, -9, arg-23 mutated wzy This study

a. Max-Planck-Institute for Biology, Tubingen, Germany.


densities were recorded at 410 nm (colour change) and 570 nm(cell debris). PhoA units were calculated using the standardequation (Manoil, 1991). b-Galactosidase assays were per-formed as previously described (Baker et al., 1997), usingstrains grown in LB with 0.2% (w/v) glucose.

T7 polymerase-driven radiolabelling assays

The method of Tabor and Richardson (1985) was used todetect radiolabelled proteins. Strain E2096 containing expres-sion constructs were grown at 308C for 16 h in LB with Ap andKm. Bacteria were subcultured and grown to an OD600 of 0.5;600 ml of the culture was pelleted, washed and resuspendedin 1 ml of methionine assay medium (Difco no. 0423-15-2)and grown at 308C for a further 2 h. After addition of 0.4 mMisopropyl-b-D-thiogalactopyranoside (IPTG) to strains con-taining pET11bYZ-based plasmids the cultures were shiftedto 428C for 20 min, Rif was added to a final concentration of200 mg ml¹1, and incubation continued for a further 20 min at428C. The culture was then transferred to 378C for a further2 h followed by pulsing with 10 mCi of L-[35S]-methioninefor 5 min at 308C. Cells were then pelleted (15 000 r.p.m./30 s Heraeus Biofuge 15) and resuspended in 120 ml ofsample buffer (Lugtenberg et al., 1975). Samples were heatedat 1008C for 5 min before electrophoresis on SDS 15% poly-acrylamide gels. Gels were then stained with Coomassie bril-liant blue G250 and dried before detection of labelled proteinsusing autoradiography (Morona et al., 1995). A scanned imageof the original autoradiograph is available on request.

Cell fractionation and Western immunoblotting

E. coli cells carrying wzy::phoA fusion constructs were grownto mid-exponential phase, whole-cell samples were preparedin sample buffer, and the remainder were fractionated to iden-tify proteins located in soluble and insoluble fractions by a pre-viously described method (Achtman et al., 1979). PhoA fusionproteins were detected after samples were solubilized by heat-ing at 1008C for 5 min, then separated using SDS–PAGE andtransferred to nitrocellulose membranes (Morona et al., 1995).Rabbit anti-E. coli alkaline phosphatase serum (5-Prime, 3-Prime) was used as the primary antibody and goat anti-rabbitperoxidase conjugate (KPL) as the secondary antibody. Theblot was developed using 4-chloro-1-naphthol as describedby Hawkes et al. (1982).

LPS and Sf6c bacteriophage sensitivity

All S. flexneri strains were grown for 16 h at 378C in LB con-taining appropriate antibiotics, except for those containingpGP1-2 (308C). Small scale preparations were made by pro-teinase K treatment of whole-cell lysates (Hitchcock andBrown, 1983). After electrophoresis on SDS 15% or 20% poly-acrylamide gels, LPS was detected using silver staining asdescribed previously (Morona et al. 1991). Sf6c bacteriophagesensitivity was performed as described by Morona et al. (1994).

Acknowledgements

Funding for this study was provided by the National Health

and Medical Research Council of Australia and the WorldHealth Organization through the Diarrhoeal Diseases VaccinesProgramme. C.D. is in receipt of an Australian PostgraduateResearch Award.

References

Achtman, M., Manning, P.A., Edelbluth, C., and Herrlich, P.(1979) Export without proteolytic processing of inner andouter membrane proteins encoded by F sex factor tra cis-trons in Escherichia coli minicells. Proc Natl Acad Sci USA76: 4837–4841.

Baker, S., Daniels, C., and Morona, R. (1997) PhoP/Q regu-lated genes in Salmonella typhi: identification of melittinsensitive mutants. Micro Pathogen 22 165–179.

Brown, P.K., Romana, L.K., and Reeves, P.R. (1992) Mole-cular analysis of the rfb gene cluster of Salmonella serovarmuenchen (strain M67), the genetic basis of the poly-morphism between groups C2 and B. Mol Microbiol 6:1385–1394.

Calamia, J., and Manoil, C. (1990) lac permease of Escheri-chia coli: topology and sequence elements promotingmembrane insertion. Proc Natl Acad Sci USA 87: 4937–4941.

Chen, G.T., and Inouye, M. (1990) Suppression of the nega-tive effect of minor arginine codons on gene expression;preferential usage of minor codons within the first 25codons of Escherichia coli genes. Nucleic Acids Res 18:1465–1473.

Collins, L.V., and Hackett, J. (1991) Molecular cloning char-acterization, and nucleotide sequence of the rfc gene,which encodes an O-antigen polymerase of Salmonellatyphimurium. J Bacteriol 173: 2521–2529.

de Kievet, T.R., Dasgupta, T., Schweizer, H., and Lam, J.S.(1995) Molecular cloning and characterization of the rfcgene of Pseudomonas aeruginosa (serotype O5). MolMicrobiol 16: 565–574.

Franco, A.V., Liu, D., and Reeves, P.R. (1996) A Wzz (cld )protein determines the chain length of K lipopolysacchar-ide in Escherichia coli O8 and O9 strains. J Bacteriol178: 1903–07.

Goldman, E., Rosenberg, A.H., Zubay, G., and Studier, F.W.(1995) Consecutive low-usage leucine codons block trans-lation only when near the 58 end of a message in Escheri-chia coli. J Mol Biol 245: 467–473.

Grosjean, H., and Fiers, W. (1982) Preferential codon usagein prokaryotic genes: the optimal codon-anticodon inter-action energy and the selective codon usage in efficientlyexpressed genes. Gene 18: 199–209.

Hale, T.L. (1991) Genetic basis of virulence in Shigellaspecies. Microbiol Rev 55: 206–224.

Hawkes, R.E., Niday, E., and Gordon, J. (1982) A dot blot-immunobinding assay for monoclonal and other anti-bodies. Anal Biochem 119: 142–147.

Hitchcock, P.J., and Brown, T.M. (1983) Morphological hetero-geneity among Salmonella lipopolysaccharide chemotypesin silver-stained polyacrylamide gels. J Bacteriol 154:269–277.

Hoffman, C.S., and Wright, A. (1985) Fusions of secretedproteins to alkaline phosphatase: An approach for studyingprotein secretion. Proc Natl Acad Sci USA 82: 5107–5111.



Hoffman, K., and Stoffel, W. (1993) TMBASE – a database ofmembrane spanning protein segments. Biol Chem Hoppe-Seyler 374: 166.

Klena, J.D., and Schnaitman, C.A (1993) Function of the rfbcluster and the rfe gene in the synthesis of O-antigen byShigella dysenteriae 1. Mol Microbiol 9: 393–402.

Kyte, J., and Doolittle, R.F. (1982) A simple method for dis-playing the hydropathic character of a protein. J Mol Biol157: 105–132.

Lindberg, A.A., Karnell, A., and Weintraub, A. (1991) Thelipopolysaccharide of Shigella bacteria as a virulence fac-tor. Rev Infect Dis 13: S279–284.

Liu, W.S., Cole, R.A., and Reeves, P.R. (1996) An O-antigenprocessing function for Wzx (RfbX): a promising candidatefor O-antigen flippase. J Bacteriol 178: 2102–2107.

Lugtenberg, B., Meijers, J., Peters, R., van der Hoek, P., andvan Alphen, L. (1975) Electrophoretic resolution of themajor outer membrane proteins of Escherichia coli K12into four bands. FEBS Lett 58: 254–258.

Lukomski, S., Hull, R.A., and Hull, S.I. (1996) Identification ofthe O-antigen polymerase (rfc) gene in Escherichia coli O4by insertional mutagenesis using a non-polar chloram-phenicol resistance cassette. J Bacteriol 178: 240–247.

Macpherson, D.F., Manning, P.A., and Morona, R. (1994)Characterization of the dTDP-rhamnose biosynthesisgenes encoded in the rfb locus of Shigella flexneri. MolMicrobiol 11: 281–292.

Macpherson, D.F., Manning, P.A., and Morona, R. (1995)Genetic analysis of the rfbx gene of Shigella flexneri.Gene 155: 9–17.

Manoil, C., and Beckwith, J. (1985) TnphoA : a transposonprobe for protein export signals. Proc Natl Acad Sci USA82: 8129–8133.

Manoil, C. (1991) Analysis of membrane protein topologyusing alkaline phosphatase and b-galactosidase genefusions. Methods Cell Biol 35: 61–75.

Makela, P. H., and Stocker, B.A.D. (1984) Genetics of lipo-polysaccharide. In Handbook of Endotoxin. Vol. 1. RietschelE.T. (ed.). Amsterdam: Elsevier Science Publishing, pp.59–137.

Morona, R., Brown, M.H., Yeadon, J., Heuzenroeder, M.W.,and Manning, P.A. (1991) Effect of lipopolysaccharidecore synthesis mutations on the production of Vibrio cho-lerae O-antigen in Escherichia coli K-12. FEMS MicrobiolLett 82: 279–286.

Morona, R., Mavris, M., Falarino, A., and Manning, P.A.(1994) Characterization of the rfc region of Shigella flex-neri. J Bacteriol 176: 733–747.

Morona, R., Van Den Bosch, L., and Manning, P.A. (1995)Molecular, genetic, and topological characterisation ofO-antigen chain length regulation in Shigella flexneri. J Bac-teriol 177: 1059–1068.

Mulford, C.A., and Osborn, M.J. (1983) An intermediate stepin translocation of lipolysaccharide to the outer membraneof Salmonella typhimurium. Proc Natl Acad Sci USA 80:1159–1163.

Reeves, P.R., Hobbs, M., Valvano, M.A., Skurnik, M., Whit-field, C., Coplin, D., et al. (1996) Bacterial polysaccharidesynthesis and gene nomenclature. Trends Microbiol 4:495–503.

Rosenberg, A.H., Lade, B.N., Chui, D-S., Lin, S-W., Dunn,J.J., and Studier, F.W. (1987) Vectors for selective expres-sion of cloned DNAs by T7 RNA polymerase. Gene 56:125–135.

Rosenberg, A.H., Goldman, E., Dunn, J.J., Studier, F.W.,and Zubay, G. (1993) Effects of consecutive AGG codonson translation in Escherichia coli, demonstrated with a ver-satile codon test system. J Bacteriol 175: 716–722.

Sandlin, R.C., Lampel, K.A., Keasler, S.P., Goldberg, M.B.,Stolzer, A.L., and Maurelli, A.T. (1995) Avirulence ofrough mutants of Shigella Flexneri : requirement of O-anti-gen for correct unipolar localization of IcsA in the bacterialouter membrane. Infect Immun 63: 229–237.

Saier, M.H., Jr, (1994) Computer-aided analyses of transportprotein sequences: gleaning evidence concerning function,structure, biogenesis, and evolution. Microbiol Rev 58: 71–93.

Simons, R.W., Houman, F., and Kleckner, N. (1987)Improved single and multicopy lac-based cloning vectorsfor protein and operon fusions. Gene 53: 85–96.

Spanjaard, R.A., and Van Duin, J. (1988) Translation of thesequence AGG-AGG yields 50% ribosomal frameshift.Proc Natl Acad Sci USA 85: 7967–7971.

Tabor, S., and Richardson, C.C. (1985) A Bacteriophage T7RNA polymerase/promoter system for controlled exclusiveexpression of specific genes. Proc Natl Acad Sci USA 82:1074–1076.

Van Den Bosch, L., Manning, P.A., and Morona, R. (1997)Regulation of O-antigen chain length is required for Shigellaflexneri virulence. Mol Microbiol 23: 765–775.

Wang, R.F., and Kushner, S.R. (1991) Construction of versa-tile low-copy-number vectors for cloning, sequencing andgene expression in Escherichia coli. Gene 100: 195–199.

Zahn, K., and Landy, A. (1996) Modulation of lambda inte-grase synthesis by rare arginine tRNA. Mol Microbiol 21:69–76.



overexpression and topology of the shigella flexneri o-antigen polymerase (rfc/wzy)

Documents