The hexA gene of Erwinia carotovora encodes a LysRhomologue and regulates motility and the expression ofmultiple virulence determinants

Stephen J. Harris, Yu-Ling Shih, Stephen D. Bentleyand George P. C. Salmond *

Department of Biochemistry, University of Cambridge,Cambridge CB2 1QW, UK.

Summary

We have identified a gene important for the regulationof exoenzyme virulence factor synthesis in the plantpathogen Erwinia carotovora ssp. carotovora (Ecc )and virulence and motility in Erwinia carotovora ssp.atroseptica (Eca). This gene, hexA (hyperproductionof exoenzymes), is a close relative of the Erwiniachrysanthemi (Echr ) gene pecT and encodes a mem-ber of the LysR family of transcriptional regulators.hexA mutants in both Ecc and Eca produce abnor-mally high levels of the exoenzyme virulence factorspectate lyase, cellulase and protease. In addition, EcahexA mutants show increased expression of the fliAand fliC genes and hypermotility. Consistent with arole as a global regulator, expression of hexA fromeven a low-copy plasmid can suppress exoenzymeproduction in Ecc and Eca and motility in Eca. Pro-duction of the quorum-sensing pheromone OHHL inEcc hexA is higher throughout the growth curvecompared with the wild-type strain. Overexpressionof Ecc hexA also caused widespread effects in severalstrains of the opportunistic human pathogen, Serra-tia . Low-copy hexA expression resulted in repressionof exoenzyme, pigment and antibiotic production andrepression of the spreading phenotype. Finally, muta-tions in hexA were shown to increase Ecc or Ecavirulence in planta .

Introduction

Erwinia carotovora ssp. carotovora is a member of thefamily Enterobacteriacae and causes soft-rot in a numberof plant species (Perombelon and Kelman, 1980). A vari-ety of factors have been shown to influence the ability ofthis organism to attack plant tissue. Of major importanceare a number of extracellular pectinases, including several

pectate lyases (Hinton et al., 1989) and polygalacturonase(Hinton et al., 1990). In addition, other degradative enzymes,such as cellulase and protease, are secreted. Cellulase isan important virulence determinant in Erwinia carotovorassp. carotovora (Walker et al., 1994), but the role of pro-tease in virulence is ill-defined (Dahler et al., 1990; Strom-berg et al., 1994).

The synthesis of these different exoenzymes is co-ordi-nately regulated by a complex, but ill-understood, inter-action of a variety of gene products. Of key importanceto this exoenzyme regulatory network is the productionof the cell density-dependent signalling molecule N-(3-oxohexanoyl)-L-homoserine lactone (OHHL) by the actionof the carI gene product (Bainton et al., 1992; Jones et al.,1993; Pirhonen et al., 1993). OHHL is normally requiredfor the enhanced synthesis of pectate lyase, cellulaseand protease. However, in the absence of a second globalregulator protein RsmA, OHHL is no longer necessary forexoenzyme production (Chatterjee et al., 1995). Thissuggests that the function of OHHL might be to relieveRsmA-mediated repression. The RsmA protein has strongsequence homology to E. coli protein CsrA (Cui et al.,1995), which has been proposed to regulate target geneglgC by binding to its mRNA and accelerating transcriptdegradation (Liu et al., 1995). It has, therefore, been sug-gested that RsmA might function in a similar manner andact either directly on the exoenzyme gene transcripts them-selves or via a secondary global transcription factor (Muk-herjee et al., 1996). As well as CarI and RsmA, other geneproducts have been shown to regulate exoenzyme pro-duction globally in some strains of Ecc, including aepA(Liu et al., 1993) and aepH (Murata et al., 1994).

In addition to its role in exoenzyme regulation, RsmAhas also been implicated in the control of many other viru-lence determinants including the production of antibiotic,pigment and EPS as well as the hypersensitive responseand motility (Mukherjee et al., 1996). Although the role ofmotility in many human bacterial pathogens is ambiguous(Strauss, 1995), it has been established as an importantvirulence factor in several plant pathogens including Erwi-nia carotovora (Pirhonen et al., 1991; Mulholland et al.,1993).

The importance of exoenzyme production and motilityto the virulence of Erwinia carotovora allied to the incom-plete knowledge of the regulation of these processes

Molecular Microbiology (1998) 28(4), 705–717

Q 1998 Blackwell Science Ltd

Received 19 February, 1997; revised 8 February, 1998; accepted 12February, 1998. *For correspondence. E-mail gpcs@mole.bio.cam.ac.uk; Tel. (01223) 333650; Fax (01223) 333345.

encouraged us to look for corresponding novel regulatorygenes. We initiated parallel, independent screens to lookfor exoenzyme mutants in Ecc and motility mutants inEca. By chance, we identified two Ecc mutants and oneEca mutant defective in the same gene. Mutants defectivein this gene (hexA) show increased production of pectatelyase, cellulase and protease in Ecc and both increasedproduction of exoenzymes and increased motility in Eca.In this report, we show that hexA is a close homologueof the Erwinia chrysanthemi gene pecT and encodes amember of the LysR family of transcriptional regulators.We also show that low-copy expression of hexA can sup-press the production of multiple virulence determinantsincluding exoenzymes, antibiotics and motility in both thephytopathogen Erwinia and the human pathogen Serratia.

Results

Isolation of Ecc hexA mutants

Transposon mutagenesis of Ecc strain MH1001 was under-taken using TnphoA8-2 (a lacZ derivative of TnphoA, car-rying tetracycline resistance; Wilmes-Reisenberg andWanner, 1992) to try to generate global exoenzymeregulatory mutants. Transposon-containing transductantcolonies were screened first for protease production andthen subsequently on Cel and Pel assay plates. Mutantsshowing an increased production of all exoenzymesoccurred at a frequency of about 1 in 1000. Using thismethod, we obtained several mutants with a Hex pheno-type (hyperproduction of exoenzymes). Two of these,designated M53 and C9, were selected for further study.To show that the exoenzyme phenotype was caused bythe transposon in each of these strains, the correspondinginsertions were transduced into clean genetic backgrounds(MH1000) using the generalized transducing phage fKP(Toth, 1991). In both cases, the exoenzyme phenotype ofthe Tcr transductants showed 100% co-inheritance withthe antibiotic resistance marker (data not shown). The pre-sence of a single transposon in each of these strains wasalso confirmed by Southern hybridization using a trans-poson-specific probe (data not shown). For all subsequentexperiments (with the exception of the cloning and map-ping), Ecc hexA mutant M53 was used in preference toC9, because the position of the transposon insertion wascloser to the 58 end of the hexA gene and thereforemore likely to give a null phenotype (Fig. 3A).

Isolation of an Eca hexA mutant

Transposon TnphoA8-1 (a lacZ derivative of TnphoA, carry-ing kanamycin resistance; Wilmes-Reisenberg and Wanner,1992) was used to mutagenize Eca strain Em56.2 beforesearching for motility mutants. Six thousand colonies were

screened on motility plates at 258C. One mutant (desig-nated ll6), showed both increased motility and increasedproduction of protease, cellulase and pectate lyase. Toshow genetically that the multiple exoenzyme and motilityphenotypes were caused by the transposon in this strain,the generalized transducing phage fM1 (Toth et al., 1997)was used to transduce the insertion to a clean geneticbackground. Both the motility and exoenzyme phenotypesshowed 100% co-inheritance with the Knr resistance markerpresent on the transposon. Again, the presence of a singletransposon in this strain was confirmed by Southern hybrid-ization using a transposon-specific probe (data not shown).Subsequent analysis revealed that the TnphoA8-2 mutantsof MH1001 (M53 and C9) and the TnphoA8-1 mutant ofEm56.2 (ll6) were defective in homologous genes (seebelow). The corresponding gene in each case was calledhexA.

Exoenzyme production of hexA mutants

Quantitative spectrophotometric assays were performedto assess the increase in pectate lyase, cellulase andprotease synthesis caused by the hexA mutants in Eccthroughout the growth curve (Fig. 1A–C). The hexA muta-tion was observed to have no effect on the growth rate ofEcc (Fig. 1A–C). All three enzymes are induced earlier inthe growth phase in a hexA mutant than in the wild-typestrain. Also, all three are expressed to much higher levels,with the hexA mutant showing a 2.67-fold increase in cel-lulase production, a 1.5-fold increase in pectate lyase pro-duction and a 5.6-fold increase in protease productionafter 13, 11 and 11 h respectively. In Eca, the hexA muta-tion caused a 2.85-fold, 1.75-fold and 2.75-fold increase inthe production of pectate lyase, cellulase and protease,respectively, at early stationary phase compared with awild-type control (data not shown). Plate assays indicatedthat hexA had no obvious effect on polygalacturonase pro-duction, and so liquid assays were not performed for thisenzyme.

Motility of the Eca hexA mutant

As mentioned previously, hexA mutant ll6 was isolated in ascreen for motility mutants of Eca. In Eca, hexA causes anincrease in motility on motility assay plates (Fig. 2A). Weperformed Northern hybridizations to determine whetherthis increased motility might result, in part, from an increasein flagellar synthesis. For this experiment, total RNA wasextracted from wild-type Eca (strain 1043) and from EcahexA (ll6). This was then Northern blotted and hybridizedwith probes to the Eca fliA and fliC genes. Figure 2Bshows that, in a hexA mutant, synthesis of the fliA andfliC transcripts is markedly upregulated.

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

706 S. J. Harris, Y.-L. Shih, S. D. Bentley and G. P. C. Salmond

Cloning of the hexA genes from Ecc and Eca

Chromosomal DNA from each of the mutant strains wasdigested with either BamHI (to yield DNA 58 of the trans-poson) or Pst I (to yield DNA 38 of the transposon) andthen cloned into plasmid pACYC177. Primers weredesigned to the DNA flanking the transposon insertionsin Ecc and used to amplify by polymerase chain reaction(PCR) a digoxygenin-labelled probe homologous to thecentral portion of the Ecc hexA gene (fragment a inFig. 3A). This probe was used to isolate a 6 kb EcoRIclone containing the whole of the Ecc hexA gene (plasmidpSH42 in Fig. 3A) from an Ecc mini-library constructed inplasmid pBluescript-II KS. The same probe was also usedto isolate a 6 kb EcoRI clone containing the Eca hexA gene

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

Fig. 1. Expression of pectate lyase (A), cellulase (B) and protease(C) through the growth curve in hexA and wild-type Ecc. Cultures(25 ml) of Pel minimal medium (cellulase and pectate lyase) or LB(protease) were inoculated to an OD of 0.02. Cultures wereincubated with shaking at 308C, and 0.5 ml samples were takenevery hour and the OD (600 nm) measured. Assays wereperformed in triplicate on the same 0.5 ml samples and the medianvalue is shown. Pectate lyase activity is expressed as change inOD 235 nm ml¹1 culture min¹1; cellulase activity is expressed aschange in OD 550 nm ml¹1 culture h¹1; protease activity isexpressed as change in OD 436 nm ml¹1 culture h¹1. Resultsrepresent a single experiment, although each experiment wasrepeated at least three times with similar results.

Fig. 2. A. Motility of wild-type and hexA mutant strains of Ecaassayed on 0.3% swarm agar plates at 258C.B. Northern hybridization showing the effects of hexA mutation onfliA and fliC transcription. Lanes 1 and 2 were hybridized with afliA-specific probe, and lanes 3 and 4 were hybridized with afliC-specific probe.

hexA regulates multiple virulence determinants 707

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

Fig. 3. A. Diagram of Ecc hexA locus. Position of genes in region indicated by white arrows. Shaded arrows indicate that, from sequencehomology, the gene is expected to extend outside of the region sequenced. Hatched rectangle indicates region of map that has not beensequenced. Only positions of restriction sites used to create plasmids are indicated on the diagram. Positions of transposon insertions in hexAare indicated by open triangles. The orientation of the lacZ gene on transposon insertions is indicated by black arrows. The region used toscreen a mini-library for the hexA gene is indicated by the black rectangle (a). nuo A,B,C are the first three genes of the NADHdehydrogenase operon and show 65%, 90% and 83% identity to their E. coli homologues at the protein level. aat is aspartate aminotransferase and shows 91% identity to its E. coli homologue at the protein level.B. Alignment of deduced amino acid sequences of Ecc HexA, Echr PecT and E. coli Lrha. Sequence identities are indicated by black boxes.Conservative amino acid substitutions are indicated by grey boxes. The numbers on the left refer to the position of the amino acid from theproposed start of translation.

708 S. J. Harris, Y.-L. Shih, S. D. Bentley and G. P. C. Salmond

(plasmid pYS012) from an Eca mini-library constructed inpCLI920 (Lerner and Inouye, 1990). A Pst I partial digestof pSH42, followed by digestion with SphI yielded a1.6 kb fragment containing the whole of the Ecc hexAgene and 300 bp upstream of the expected start of trans-lation. This fragment was cloned into pBluescript-II KS togive plasmid pSH50 and into low-copy vector pCLI920 togive pSH51 (Fig. 3A). Sequence analysis of this 1.6 kb frag-ment revealed only one long open reading frame (ORF) of948 bp. The direction of transcription of this ORF is inagreement with the direction indicated by the transposontranscriptional fusions. Translation of this ORF from anATG at base 290 would give a predicted protein with amolecular mass of 34 776 Da. Upstream of the ATG, thereare two potential ¹10 sequences, 267–272 (five out ofsix matches) and 261–266 (five out of six matches). Thelatter of these also conforms to the consensus for anextended ¹10 sequence (Keilty and Rosenberg, 1987)with eight out of nine matches to the conserved residues.There is also a weak match to the ¹35 consensus sequenceat 229–233 (four out of six matches). Additional sequenceinformation outside of the hexA gene was obtained fromplasmids pSH47 (HindIII/EcoRI fragment in pBluescript-IIKS) and pSH49 (SphI fragment in pBluescript-II KS;Fig. 3A).

As expected, the Eca hexA gene is extremely similar toits Ecc counterpart. Sequencing of the Eca hexA geneindicated 75 base differences from Ecc in the long ORF.These differences give rise to 11 amino acid changes, ofwhich five are conservative. These sequence data havebeen submitted to the DDBJ/EMBL /GenBank databasesunder the accession numbers AF057063 and AF057064.

The Ecc HexA sequence was compared with proteinsheld in the NCBI database using the BLAST search program.HexA shows strong homology to LysR-type transcriptionalregulators (LTTRs) especially over the conserved DNA-binding domain. The strongest homology is to PecT fromErwinia chrysanthemi (80% identity, 88% similarity) andto LrhA from Escherichia coli (64% identity, 79% simi-larity; Fig. 3B). PecT mutants of E. chrysanthemi showincreasedexpression of severalof the pel genes responsiblefor the synthesis of pectate lyase (Surgey et al., 1996).The function of LrhA is unknown (Bongaerts et al., 1995).

Self-regulation

As many LTTRs regulate their own transcription, we wereinterested to see if this was the case for hexA. To test this,strain M53 (containing a transposon insertion in hexA,resulting in a transcriptional fusion between hexA andlacZ ) was used. The position of the transposon insertionin strain M53 is so close to the N-terminus of hexA thatthe hexA–lacZ fusion formed is unlikely to have any bio-logical activity (Fig. 3A). Strain M53 was transformed

with plasmid pSH51 (containing low-copy hexA) and inde-pendently with the control vector pCLI920. As the growthcurve demonstrates, expression of HexA from the low-copy vector had no effect on the growth rate of Ecc(Fig. 4). However, throughout the growth curve, expressionof hexA–lacZ was higher in the presence of pSH51 (con-taining hexA) than in the presence of the control plasmidpCLI920. The increase in expression caused by pSH51varied from 1.5 times to 2.4 times with an average of1.82 times. A similar expression profile was also seen fora hexA–lacZ fusion in Eca (data not shown).

Low-copy hexA suppresses exoenzyme production atthe level of transcription

To test whether extrachromosomal copies of hexA couldsuppress the production of exoenzymes, we transformedMH1000 with pSH51 and Em56.1 with pYS012 and withpSH51. As a control, both strains were transformed withpCLI920. Figure 5A shows that hexA can dramaticallysuppress exoenzyme production in both strains. Ecc andEca containing pCLI920 expressed similar levels of exo-enzymes to the parent wild-type strains. In contrast, thepresence of plasmid-borne hexA in either strain resultedin a reduction in the production of all three exoenzymesto very low levels. In Eca, the level of suppression isslightly less dramatic if the Ecc hexA gene is used ratherthan the Eca gene.

LysR homologues regulate expression of their targetgenes at the level of transcription (Schell, 1993). To test

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

Fig. 4. Effect of plasmids pCLI920 and pSH51 on a hexA –lacZtranscriptional fusion. Cultures (25 ml) of Pel minimal mediumcontaining spectinomycin (0.05 mg ml¹1) were inoculated to an ODof 0.02. Cultures were incubated with shaking at 308C and 0.5 mlsamples were taken every hour and the OD measured at 600 nm.b-Galactosidase assays were performed on the same 0.5 mlsamples. b-Galactosidase activity is expressed as nmol of ONPGbroken down min¹1 ml¹1 culture. Assays were performed intriplicate. Median value is plotted in each case with higher andlower values indicated by error bars.

hexA regulates multiple virulence determinants 709

if this was the case for HexA, we used transcriptionalfusions to the pelC and celV genes of Erwinia carotovorastrain SCRI193 (L. Vincent-Sealey, unpublished). Figure5A shows that low-copy expression of HexA from plasmidpSH51 causes 5.15-fold and 4.76-fold reductions in theexpression of the pelC–lacZ and celV–lacZ fusionsrespectively.

As these results show a more pleiotropic effect for hexAthan those shown by pecT from E. chrysanthemi (Echr ),which appears only to affect pectate lyase production (Sur-gey et al., 1996), we tested the effect of low-copy hexAexpression on exoenzyme production in Echr. We found

that expression of cellulase (and protease) in Echr wasunaffected by low-copy expression of hexA, whereas pec-tate lyase expression was repressed fourfold (data notshown). These results are consistent with those obtainedfor pecT by Surgey et al. (1996).

Low-copy hexA suppresses motility in Eca

As our Eca hexA mutant is hypermotile at 258C, we werealso interested to see if low-copy hexA would suppressmotility. Eca strain Em56.1 was transformed with pSH51,pYS012 and pCLI920. Figure 5B shows that both hexAgenes suppress motility. Plasmid pCLI920 had no effecton motility in Eca, whereas the pYS012 containing EcahexA resulted in a strong reduction in motility. As withexoenzyme production, the level of suppression shown bythe Ecc gene was less than that caused by its Eca counter-part. As Ecc strain MH1000 is not motile, the motility effectsof hexA in multicopy could not be assayed in this strain.

HexA protein binds upstream of the pelC gene

To determine whether hexA regulates exoenzyme produc-tion by binding directly to the promoters of the exoenzymegenes or whether it acts via an intermediary gene orgenes, we performed DNA-binding experiments using acrude extract enriched in HexA protein as described inExperimental procedures. Gel-shift experiments were per-formed on labelled DNA fragments containing the pelCpromoter using varying concentrations of HexA proteinand specific and non-specific competitor DNA. The pro-moter from the nuoA gene, which lies immediately down-stream of hexA and which codes for NADH dehydrogenasesubunit A, was chosen as a non-specific control. Expressionof the homologue of this gene in E. coli was shown to beunaffected by mutation of the E. coli hexA homologuelrha (Bongaerts et al., 1995). A 531 bp DraI/StuI fragmentincluding the nuoA promoter was used. Figure 6 showsthat HexA protein binds to the pelC promoter. Bindingcan be outcompeted by a 100-fold excess of unlabelledprobe but not by a 100-fold excess of unlabelled non-specificcompetitor, showing the binding to be specific. Underthese conditions, we could not identify binding to the hexApromoter.

HexA delays the onset of OHHL accumulation

Given the central role of OHHL in the co-ordinate regula-tion of exoenzyme production, it was important to deter-mine whether any of the phenotypes shown by the hexAmutants might be mediated through changes in the levelof the signalling molecule. To do this, we used a lux-based detection system previously reported by Throup etal. (1995). Wild-type and hexA strains were grown in

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

Fig. 5. A. Effect of expression of HexA from a low-copy plasmid onexoenzyme production and transcription in Ecc and enzymeproduction in Eca. Ecc was grown at 308C in Pel minimal mediumfor 12 h. Eca was grown at 258C in Pel minimal medium for 20 h.Pectate lyase activity is expressed as change in OD 235 nm ml¹1

culture min¹1. Cel activity is expressed as change in OD623 nm ml¹1 culture h¹1. Prt activity is expressed as change in OD436 nm ml¹1 culture h¹1. b-Galactosidase activity is expressed innmol of ONPG broken down min¹1 ml¹1 culture. X indicates assaynot performed.B. Motility of Eca in the presence of control vector (pCLI920),pSH51 or pYS012 on 0.3% swarm agar.

710 S. J. Harris, Y.-L. Shih, S. D. Bentley and G. P. C. Salmond

PMB, and samples were taken hourly through the growthcurve. Figure 7A shows that the hexA mutant producesconsistently higher levels of OHHL throughout the growthcurve, especially during the early log phase. In addition,low-copy expression of HexA from pSH51 was shown tocause a decrease in OHHL levels in the exponential andearly stationary phases by comparison with a vector onlycontrol (Fig. 7B).

Global regulatory effects of hexA in Serratiamarcescens strains

Plasmids pSH51, pYS012 and pCLI920 were transferredby electroporation into three strains of Serratia marces-cens. Strain ATCC39006 produces pigment (prodigiosin),antibiotic (carbapenem), cellulase and pectate lyase andis motile. Strains ATCC274 and SBon1 produce pigmentand protease and are non-motile but are capable ofspreading motility (O’Rear et al., 1992).

In strain ATCC39006, both hexA genes inhibit the pro-duction of antibiotic, with the Eca hexA gene inhibitingthis process completely (Fig. 8). Similarly, both inhibit pro-digiosin production, again with the Eca hexA gene show-ing the strongest effect. In this strain, cellulase productionis only partially inhibited by the hexA genes and, althoughthe inner halo produced by pectate lyase is inhibited, theouter halo is unaffected, implying that there are multiplepectinolytic enzymes in this strain of Serratia and thatthey are differentially regulated by hexA. Interestingly,both hexA genes seem to cause a slight increase in motil-ity in this strain of Serratia (Fig. 8) in contrast to theireffects in Eca (Fig. 5B).

Again, in strains ATCC274 and SBon1 (data not shown),prodigiosin production is inhibited, although in these casesthe Eca hexA gene is slightly less effective than the EcchexA gene. This pattern of repression is repeated forprotease production. Finally, both Ecc and Ecc hexA com-pletely repressed spreading motility. Spreading motilityhas been described for a non-motile strain of Serratiawhen grown on media containing 0.3–0.4% agar (O’Rear

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

Fig. 6. DNA-binding assays performed using a digoxygenin-labelled 624 bp, SacI/Pst I, pelC fragment as probe. Amounts ofcrude protein extracts are labelled across the top. Plasmids alongthe bottom refer to those present in cells used to generate thecrude protein extract. In lanes 6 and 7, a 100-fold molar excess ofunlabelled specific competitor (pelC) or non-specific (nuoA)competitor was added.

Fig. 7. A. Production of OHHL compared with OD 600 nm in hexAand wild-type strains of Ecc.B. Production of OHHL compared with OD 600 nm in wild-type Ecccontaining vector only (pCLI920) or vector plus hexA (pSH51). Forboth A and B, strains were grown in Pel minimal medium (25 ml) at308C, and 0.5 ml samples were taken every hour. Levels of OHHLwere then measured for each sample in triplicate, as described inExperimental procedures. White and grey bars indicate the medianvalue for each of the triplicates with the error bars indicating theupper and lower values. Results represent a single experiment,although the experiment was repeated three times with similarresults.

hexA regulates multiple virulence determinants 711

et al., 1992). This movement is relatively rapid and accom-panied by the production of copious amounts of colourlessextracellular slime (ES). In strains ATCC274 (data notshown) and SBon1, both hexA plasmids completely rep-ressed spreading and the production of extracellular slime.

Effects of hexA on virulence

To assess the effects of a hexA mutation on the ability ofEcc and Eca to macerate plant tissue in vivo, we used apotato tuber virulence test (Walker et al., 1994.) Tuberswere inoculated with wild-type and hexA strains for bothEcc and Eca. The weights of wet rot produced by eachof these strains over a 4-day period is shown in Fig. 9.In Ecc, the hexA mutation results in a doubling of theamount of rot produced compared with the wild-type strainover the 4-day period. In Eca, the effects on virulence areeven more pronounced, with the hexA mutant producing

almost six times as much rot as the wild type after 4days (Fig. 9). In addition, we also assessed the effect onvirulence of expressing the hexA gene from a plasmid inwild-type Ecc. In this experiment, the Ecc strain carryingthe hexA gene on a plasmid (pSH51) produced only halfthe rot of Ecc carrying control plasmid pCLI920 after 4days (data not shown).

Discussion

The co-ordinate regulation of multiple virulence determi-nants by bacterial pathogens is a complex process. InErwinia carotovora species, several of the genes thatplay a role in this process have been identified previously.In this study, we have identified a new global negative reg-ulator of multiple virulence determinants in Erwinia caroto-vora, which we have named hexA.

HexA shares extensive homology with PecT from Echr

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

Fig. 8. Repression of multiple virulencedeterminants by hexA in Serratia strains. Asample of 5 ml of overnight culture wasspotted onto the appropriate assay plate andgrown at either 308C (cellulase, pectate lyase,protease, prodigiosin and spreading) or 258C(antibiotic and motility). Photographs weretaken after 16 h except for cellulase andpectate lyase plates, which were left for 48 hbefore being developed.

712 S. J. Harris, Y.-L. Shih, S. D. Bentley and G. P. C. Salmond

and LrhA from E. coli. The homology to these two proteinsis very strong (80% and 64% identity respectively) and,thus, HexA is likely to be the Ecc or Eca functional equiva-lent of PecT or LrhA. A role in the regulation of pectatelyase synthesis in Echr for PecT has been described pre-viously (Surgey et al., 1996). We found that, in Ecc andEca, HexA has a much more general regulatory function,repressing both production of several different exoenzymesand motility. Like PecT and LrhA, HexA shows stronghomology at the N-terminal end to many LysR-type tran-scriptional regulators (LTTRs). However, although thesequence homology suggests that HexA is undoubtedlyan LTTR, it lacks several of the features shared by many,but not all, LTTRs. For instance, most LTTRs are tran-scriptional activators (Schell, 1993), whereas HexA hasrepressive effects (although HexA could also activate theexpression of a global repressor). Also, most LTTRs aretranscribed divergently from the genes that they controland regulate their own transcription negatively as a conse-quence (Schell, 1993). In contrast, hexA is not closelylinked to any known target gene and activates its own tran-scription (Figs 3 and 4).

In agreement with the HexA phenotype of overpro-duction of exoenzymes and increased motility, we haveshown that the expression of HexA from a low-copy vectorin wild-type strains dramatically represses both exoenzymeproduction and motility (Fig. 5). Interestingly, the Ecc andEca proteins had slightly different effects when expressedin their non-cognate strains. For example, Ecc HexA rep-ressed exoenzymes and motility to a lesser degree in Ecathan did Eca HexA.

As would be expected for a LysR homologue and incommon with PecT (Surgey et al., 1996), we found thathexA affects the level of transcription of its target genes.Low-copy expression of HexA reduced the expression of

a pelC–lacZ fusion (Fig. 5A). However, in contrast toPecT, we found that HexA also repressed the transcriptionof a celV–lacZ fusion. As we also found that HexA wasunable to suppress cellulase production in Echr (Fig. 5A),this may suggest that the cellulase gene in Echr has evolvedto become independent of HexA regulation. We have alsoshown by Northern analysis that, in a hexA mutant, syn-thesis of the fliC transcript is upregulated (Fig. 2B). In E.coli, the fliC gene encodes the flagellin subunit from whichthe flagellum is made (Kuwajima et al., 1986). This increasein fliC (and presumably flagellin) expression might, there-fore, be one of the factors responsible for the increasein motility in Eca hexA. However, it is unclear whetherincreased expression of flagellin alone (in the absence ofincreased expression of other components of the flagellarapparatus, e.g. motor, hook) would cause a dramaticincrease in motility. With this in mind, it is therefore inte-resting that hexA also causes an increase in fliA expres-sion (Fig. 2B). In Salmonella, fliA encodes an alternativesigma factor specific for flagellar operons (Ohnishi et al.,1990). This suggests that, should similar regulatory hier-archies operate in Erwinia, hexA might influence multiplegenes involved in motility through interactions with oneor more regulators towards the top of the hierarchy (e.g.including fliA). A detailed analysis of hexA impact on mul-tiple Eca flagellar regulon genes will be required to deter-mine the exact point(s) of interaction.

We were also able to show by DNA-binding assays thatHexA protein binds directly to the pelC promoter (Fig. 6).As low-copy expression of HexA represses transcription ofa pelC–celV fusion (Fig. 5A), this strongly suggests thatHexA is a direct repressor of pelC transcription ratherthan acting indirectly by activating the expression of a rep-ressor of pelC. Although low-copy expression of HexAactivates transcription of the hexA gene itself (Fig. 4),

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

Fig. 9. Production of wet rot by wild-type and hexA mutant strains of Ecc and Eca after inoculation into potato tubers.A. The effect of hexA mutation on virulence in Ecc.B. The effect of hexA mutation on virulence in Eca. At each time point, six tubers were assayed. The values shown in the graphs representthe mean weight of rot calculated after the largest and smallest of the six rot weights had been discarded. 95% confidence intervals areshown only for day 4 and are indicated by error bars above and below the mean.

hexA regulates multiple virulence determinants 713

we were unable to show HexA binding to the hexA pro-moter using the same conditions as for pelC. This maybe because HexA regulates its own expression via a thirdparty gene or, alternatively, different binding conditionsmay be required for HexA to bind to its own promoterrather than to pelC. This may be related to the fact thatHexA represses pelC transcription but activates hexAtranscription.

One question that arises is whether the repression ofexoenzyme production shown by low-copy hexA is medi-ated entirely at the level of transcription. For example,although low-copy hexA substantially represses celV–lacZ, the expression of the fusion still reaches more than20% of the wild-type level (Fig. 5A). In contrast, in thepresence of low-copy hexA, the production of cellulaseenzyme is reduced to less than 0.5% of the wild-typelevel. As celV is the major cellulase in Ecc (Cooper andSalmond, 1993), this difference suggests that the repres-sion of cellulase production may be partly mediated post-transcriptionally. One possibility is that HexA activatestranscription of rsmA. If RsmA acts by degrading the exo-enzyme gene transcripts (Mukherjee et al., 1996), thenoverexpression of RsmA (induced by overexpression ofHexA) would lead to a decrease in the level of exoenzymegene transcripts and, ultimately, to a decrease in the levelof exoenzymes produced. It is reasonable to suggest that,as RsmA is unlikely to degrade all transcripts, then thosetranscripts that are degraded must contain a specificsignature that marks them for degradation. If the lacZtranscript does not contain such a signal, then RsmA-mediated transcript degradation is unlikely to affect theexpression of the celV–lacZ fusion. The construction ofa transcriptional fusion in the SCRI193 rsmA homologuewould allow us to test this hypothesis.

As high levels of OHHL are important for activatingexoenzyme synthesis, it is possible that an increase inOHHL levels, as a result of hexA inactivation, might contri-bute to the increase in exoenzyme production. We there-fore looked at OHHL levels in wild-type and hexA strainsand found that the expression of OHHL in the hexA strainwas increased throughout the growth phase (Fig. 7). Thiscould also be explained by HexA being an activator ofrsmA transcription. RsmA has been suggested as regulat-ing OHHL levels by binding to and degrading the OHHLtranscript (Mukherjee et al., 1996). Thus, if HexA was notexpressed in the early growth phase, this might lead to areduction in RsmA expression and a concomitant increasein OHHL levels. In addition, we looked for reciprocal effectsof OHHL on hexA transcription. The addition of exogenousOHHL did not appear to increase the expression of thehexA reporter gene in the presence or in the absence ofa functional hexA gene (data not shown).

To see if hexA might have closely conserved homo-logues and might affect processes other than the easily

assayed phenotypes of exoenzyme production and motil-ity, we overexpressed hexA in several Serratia marces-cens species. To our surprise, hexA was able to suppressmultiple processes in all three Serratia strains. With theexception of motility in strain ATCC39006, all the pheno-types we tested were repressed to a greater or lesserdegree by Ecc and Eca hexA (Fig. 8). These results sug-gest that hexA will have a closely conserved homologue inthe human pathogen Serratia marcescens.

Finally, we looked at the effect of the hexA mutations onvirulence in Ecc and Eca. In both cases, the hexA mutantswere substantially more virulent then the wild-type strain(Fig. 9). The Ecc hexA mutant produced twice as muchrot as the wild-type strain after 4 days, whereas the EcahexA strain produced nearly 6 times as much rot. As motil-ity has been shown to be an important virulence determi-nant in Eca (Mulholland et al., 1993), the increased effectof hexA on virulence in Eca (over Ecc) might be explainedby the increase in motility caused by the hexA mutation(given that our strain of Ecc is non-motile). However, thepotato tuber assay mainly measures maceration ability.We do not know if infectivity in whole plants is affected inhexA mutants.

One question that arises from this work is whether HexAregulates the expression of all its downstream genes bybinding directly to their promoters (as it does for pelC) orwhether it controls the expression of some of its targetgenes by regulating the expression of another global regu-lator (such as rsmA). In a preliminary attempt to try toidentify possible HexA binding sites, the promoters ofrsmA and several exoenzyme genes from Ecc weresearched for a candidate HexA-binding sequence, butnone were found (data not shown). However, such a searchis problematic, as the consensus for LysR-type transcrip-tional regulators usually consists of two 4 bp invertedrepeats separated by a distance specific to the LysR homo-logue (Schell, 1993). DNA footprinting studies on the pro-moters of rsmA and the exoenzyme genes will be requiredto aid the identification of such a HexA-binding consensus.Irrespective of its mechanism of action, it is clear from thisstudy that HexA is an important global regulator affectingmultiple and diverse physiological traits in both plant andhuman pathogens.

Experimental procedures

Bacterial strains and culture media

Strains used in this study were E. coli DH5 (Bethesda ResearchLaboratories), E. coli ESS (a b-lactam-supersensitive indica-tor strain provided by Beechams Pharmaceuticals), E. coliK38 (Tabor and Richardson, 1985), MH1000 (an EMS inducedlac¹ mutant of Ecc strain SCRI193), MH1001, (a lamB þ deri-vative of MH1000 carrying pHCP2), Em56.1 (a lac¹ mutantof Eca strain 1043; Hinton et al., 1989), Em56.2 (a lamB þ

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

714 S. J. Harris, Y.-L. Shih, S. D. Bentley and G. P. C. Salmond

derivative of Em56.1 carrying pHCP2), Erwinia chrysanthemistrain LA15 (a gift from F. Barras), Serratia marcescensstrains ATCC39006 (Bycroft et al., 1987), ATCC274 (Regueet al., 1991) and SBon1 (G. Stewart, unpublished). Exceptwhere stated, Erwinia and Serratia species were grown at308C and E. coli at 378C in Luria–Bertani (LB; Miller, 1972)or nutrient broth and on corresponding solid plates (15 g l¹1

agar). When required, antibiotics were added to a final con-centration of 50 mg ml¹1, except for tetracycline, which wasused at 10 mg ml¹1.

Transposon mutagenesis

Ecc strain MH1000 and Eca strain Em56.1 were transformedwith the plasmid pHCP2 (Clement et al., 1982), which carriesthe gene encoding the LamB receptor protein from E. coli,thus allowing phage infection. LamBþ Erwinia were thengrown up in a 5 ml overnight culture in the presence of10 mM MgSO4 and antibiotic (to select for the LamB plasmid).The culture was then inoculated with 50 ml of lysate (approxi-mately 1010 phage ml¹1) and incubated at 308C for 20 min.After this, 9 ml of LB were added, and incubation continuedfor a further 1 h at 308C with shaking. The cells were then cen-trifuged, resuspended in a small volume and spread onto pro-tease assay plates containing tetracycline (for TnphoA8-2) orkanamycin (for TnphoA8-1).

Exoenzyme assays

Pectate lyase and cellulase activities were detected on platesas described in Andro et al. (1984). Protease plates were pre-pared as follows: 400 ml of 10% (w/v) tryptone soya agar and75 ml of 10% (w/v) skimmed milk were made up in water.These solutions were autoclaved, cooled to 558C and thencombined and poured. Protease activity was evident as aclearing zone around the growing colony.

Spectrophotometric protease and pectate lyase assayswere performed as described by Reeves et al. (1993). Spec-trophotometric cellulase assays were performed as describedby Cooper and Salmond (1993) or as described by Biely etal. (1985). Cellular fractionation was performed as describedby Hinton and Salmond (1987) for separation into sonicateand supernatant fractions only. b-Galactosidase assayswere performed as described by Miller (1972).

Motility and spreading assays

Motility was assayed on tryptone swarm agar plates (10 g l¹1

bacto tryptone, 5 g l¹1 NaCl, 3 g l¹1 agar) at 258C for bothErwinia and Serratia. The spreading phenotype of non-motileSerratia strains was assayed in the same manner.

Antibiotic and prodigiosin production

Antibiotic production in Serratia strains was assayed by spot-ting 2–5 ml of an overnight culture onto the surface of a 0.7%LB agar overlay containing 200 ml of an overnight culture of E.coli ESS. Plates were then incubated at 258C overnight. Pro-digiosin production was assayed on peptone glycerol plates

(5 g l¹1 peptone, 1% v/v glycerol, 1.5 g l¹1 agar; Morrison,1966).

Recombinant DNA techniques

Electroporations were performed as described by Dower etal. (1988) using a Bio-Rad electroporator. CaCl2 transforma-tions were performed as described by Tang et al. (1994).PCR reactions were performed on a Perkin-Elmer DNAthermal cycler using standard protocols. Other standardrecombinant DNA techniques were performed as describedby Sambrook et al. (1989).

Sequencing of the Ecc and Eca hexA genes

Exonuclease III deletions were performed on pSH49 using akit supplied by Amersham. DNA was prepared from 3 ml ofculture using a Qiagen mini-spin DNA isolation kit. Sequenc-ing was performed using the Pharmacia T7 Sequenase kit.The Eca hexA gene was sequenced using primers designedto the Ecc hexA gene. Initial manipulations of the sequenceincluding conceptual translation and promoter comparisonswere performed using Genetics Computer Group software.The deduced primary protein sequence was compared withthe NCBI database using the BLAST enhanced utility programBEAUTY at http:/dot.imgen.bcm.tmc.edu:9331/seq-search/pro-tein-search.html.

Northern analysis

RNA was extracted and blotted as described in Sambrook et al.(1989). Digoxigenin-labelled DNA probes were synthesized,hybridized and detected as described in the manufacturer’sinstructions (Boehringer Mannheim). Equal loadings betweenlanes were confirmed by spectrophotometry at an OD of260 nm. The fliA-specific, digoxigenin-labelled probe wasmade from a 471 bp PCR product stretching from 13 bp to484 bp downstream from the translational start site of theEca fliA gene. The fliC-specific, digoxigenin-labelled probewas made from a 477 bp PCR product stretching from 35 bpupstream to 442 bp downstream from the translational startsite of the Eca fliC gene. Eca fliA and fliC mutants were iso-lated during the same motility screen that identified EcahexA. DNA flanking the transposon was cloned from thesemutants and used to isolated full-length wild-type clonesthat were subsequently sequenced (data not shown).

HexA protein preparation

PCR primers were designed to the 58 and 38 ends of the hexAORF to allow PCR of hexA and cloning into the expressionvector pT7-6 (Tabor and Richardson., 1985). This createdvector pYS14. The hexA insert was then cloned back out ofthis vector into pBluescript-II KSþ to allow sequencing ofthe insert to check for PCR errors. To prepare crude proteinextract, E. coli strain K38 carrying plasmid pG1-2 (which pro-vided a heat-inducible T7 RNA polymerase) was transformedwith pYS14 (or the control vector pT7-6) and then grown up toan OD at 600 nm of 1 in LB. The expression of HexA wasinduced by shifting to 428C for 2 h. Cells were then pelleted,

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

hexA regulates multiple virulence determinants 715

washed twice with water, resuspended in buffer (20 mMHEPES, pH 7.2, 50 mM KCl, 1 mM EDTA, 5% glycerol) andsonicated. Cellular debris was then pelleted, and the proteinconcentration in the supernatant was determined using theBradford method (Bradford, 1976).

Gel-shift assays

The DNA fragments to be used as probes in this assay wereend labelled with digoxygenin-labelled ddUTP as described inthe digoxygenin gel shift kit supplied by Boehringer Mann-heim. Approximately 100 ng of template DNA was labelled.After labelling, unincorporated nucleotides were removedusing a Qiagen spin gel extraction kit. The labelled DNAwas eluted in 100 ml of 10 mM Tris, pH 8.5. The buffer forthe binding reaction contained 20 mM HEPES, pH7.2, 50mMKCl, 1 mM EDTA, 5% glycerol (v/v), 25 mg ml¹1 poly-(dI-dC),200 mg ml¹1 bovine serum albumin. For the binding reactions,1 ml (approximately 1 ng) of labelled probe was incubated inbinding buffer in a final volume of 20 ml with varying amountsof crude protein extract and competitor DNA for 20 min at 308Cand then loaded onto a high-ionic-strength gel (20 mM HEPES,pH 7.2, 400 mM glycine, 2 mM EDTA, 4% acrylamide, 2.5%glycerol, pH 7.2) and run at 30 V for 5 h at 48C. The sampleswere then electroblotted to nylon membrane (400 mA for30 min) and then detected as described in the BoehringerMannheim gel shift kit.

OHHL assays

OHHL assays were performed as follows. Briefly, 100 ml ofsupernatant were mixed with 100 ml of sensor strain (E. colistrain JM109 containing plasmid pSB401; Throup et al., 1995)and placed into a 96-well microtitre plate. Emission of lightwas measured after 3 h using a Life Sciences LabsystemsLuminoskan RS.

Virulence tests

Strains were assayed for in planta virulence as described byWalker et al. (1994) using a starting inoculum of approxi-mately 104 cells.

Acknowledgements

We would like to thank Matt Holden for the use of strainMH1000 and Lois Vincent-Sealey for use of the celV –lacZand pelC–lacZ fusion strains. We would also like to thank J.Throup for the use of the lux sensor plasmid. This work wassupported by BBSRC grants P08495 and P03960.

References

Andro, T., Cambost, J., Kotoujansky, A., Cattaneo, J., Ber-theau, Y., Barras, F., et al. (1984) Mutants of Erwinia chry-santhemi defective in secretion of pectinase and cellulase.J Bacteriol 160: 1199–1203.

Bainton, N., Bycroft, B., Chhabra, S., Head, P., Gledhill, L.,Hill, P., et al. (1992) A general role for the lux autoinducer

in bacterial cell signalling: control of antibiotic biosynthesisin Erwinia. Gene 116: 87–91.

Biely, P., Mislovicova, D., and Tomas, R. (1985) Solublechromogenic substrates for the assay of endo-1,4-b-xylan-ases and endo-1,4-b-glucanases. Anal Biochem 144: 142–146.

Bongaerts, J., Zoske, S., Weider, U., and Unden, G. (1995)Transcriptional regulation of the proton translocating NADHdehydrogenase genes (nuoA-N) of Escherichia coli by elec-tron acceptors, electron donors and gene regulators. MolMicrobiol 16: 521–534.

Bradford, M. (1976) A rapid and sensitive method for thequantification of microgram quantities of proteins utilisingthe principle of protein dye binding. Anal Biochem 72:248–254.

Bycroft, B., Maslen, C., Box, S., Brown, A., and Tyler, J.(1987) The isolation and characterisation of (3R, 5R)-(3S, 5R)-carbapenem-3-carboxylic acid from Serratia andErwinia species and their putative biosynthetic role. JChem Soc Chem Commun 1987: 1623.

Chatterjee, A., Cui, Y., Liu, Y., Korsi Dumenyo, C., and Chat-terjee, A.K. (1995) Inactivation of rsmA leads to overproduc-tion of extracellular pectinases, cellulases and proteases inErwinia carotovora ssp. carotovora in the absence of thestarvation/cell density-sensing signal, N- (3-oxohexanoyl)-L-homoserine lactone. Appl Environ Microbiol 61: 1959–67.

Clement, J., Perrin, D., and Hedgepath, J. (1982) Analysis ofreceptor and b-lactamase synthesis and export using clonedgenes in a mini-cell system. Mol Gen Genet 185: 302–310.

Cooper, V., and Salmond, G. (1993) Molecular analysis of themajor cellulase (CelV) of Erwinia carotovora: evidence fora evolutionary ‘mix and match’ of enzyme domains. MolGen Genet 241: 341–350.

Cui, Y., Chatterjee, A., Liu, Y., Korsi Dumenyo, C., and Chat-terjee, A.K. (1995) Identification of a global repressor gene,rsmA of Erwinia carotovora ssp. carotovora that controlsextracellular enzymes, N-(3-oxohexanoyl)-L-homoserine lac-tone and pathogenicity in soft-rot Erwinia. J Bacteriol 177:5108–5115.

Dahler, G., Barras, F., and Keen, N. (1990) Cloning of genesencoding extracellular metalloproteases from Erwinia chry-santhemi EC16. J Bacteriol 172: 5803–5815.

Dower, W., Miller, J., and Ragsdale, C. (1988) High efficiencytransformation of E. coli by high voltage electroporation.Nucleic Acids Res 16: 6127–6145.

Hinton, J., and Salmond, G. (1987) Use of TnphoA to enrichfor extracellular enzyme mutants or Erwinia carotovorassp. carotovora. Mol Microbiol 1: 381–386.

Hinton, J., Sidebotham, J., Gill, D., and Salmond, G. (1989)Extracellular and periplasmic isoenzymes of pectate lyasefrom Erwinia carotovora belong to different families. MolMicrobiol 3: 1785–1795.

Hinton, J., Gill, G., Lalo, D., Phaston, G., and Salmond, G.(1990) Sequence of the peh gene of Erwinia carotovora:homology between Erwinia and plant enzymes. Mol Micro-biol 4: 1029–1036.

Jones, S., Yu, B., Bainton, N., Birdsall, M., Bycroft, B., Chha-bra, S., et al. (1993) The lux autoinducer regulates the pro-duction of exoenzyme virulence determinants in Erwiniacarotovora and Pseudomonas aeruginosa. EMBO J 12:2477–2482.

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

716 S. J. Harris, Y.-L. Shih, S. D. Bentley and G. P. C. Salmond

Keilty, S., and Rosenberg, M. (1987) Constitutive function ofa positively regulated promoter reveals new sequencesessential for activity. J Biol Chem 262: 6389–6395.

Kuwajima, G., Asaka, J.-I., Fujiwara, T., Fujiwara, T., Node,K., and Kondoh, E. (1986) Nucleotide sequence of thehag gene encoding flagellin of Escherichia coli. J Bacteriol168: 1479–1483.

Lerner, G., and Inouye, M. (1990) Low copy number plasmidsfor regulated low-level expression of cloned genes in E.coli with blue/white insert screening capability. NucleicAcids Res 18: 4631.

Liu, Y., Murata, H., Chatterjee, A., and Chatterjee, A.K. (1993)Characterisation of a novel regulatory gene aepA that con-trols extracellular enzyme production in the phytopatho-genic bacterium Erwinia carotovora ssp. carotovora. MolPlant–Microbe Interact 6: 299–308.

Liu, Y.M., Yang, H., and Romeo, T. (1995) The product of thepleiotropic Escherichia coli gene csrA modulates glycogenbiosynthesis via effects on mRNA stability. J Bacteriol 177:2663–2672.

Miller, J. (1972) Experiments in Molecular Genetics. ColdSpring Harbour, NY: Cold Spring Harbour Laboratory Press.

Morrison, D. (1966) Prodigiosin synthesis in mutants of Ser-ratia marcescens. J Bacteriol 91: 1599–1604.

Mukherjee, A., Cui, Y., Liu, Y., Korsi Dumenyo, C., andChatterjee, A.K. (1996) Global regulation in Erwinia speciesby Erwinia carotovora rsmA, a homologue of Esherichiacoli csrA : repression of secondary metabolites, pathogeni-city and hypersensitive reaction. Microbiology 142: 427–434.

Mulholland, V., Hinton, J., Sidebotham, J., Toth, I., Hyman,L., Perombelon, M., et al. (1993) A pleiotropic reduced viru-lence (Rvi¹) mutant of Erwinia carotovora ssp. atrosepticais defective in flagella assembly proteins that are con-served in plant and animal bacterial pathogens. Mol Micro-biol 9: 343–356.

Murata, H., Chatterjee, A., Liu, Y., and Chatterjee, A.K.(1994) Regulation of the production of extracellular pectin-ase, cellulase and protease in the soft rot bacterium Erwiniacarotovora ssp. carotovora: evidence that aepH of Erwiniacarotovora ssp. carotovora 71 activates gene expressionin E. carotovora ssp. carotovora, E. carotovora ssp. atro-septica and Escherichia coli. Appl Environ Microbiol 60:3150–3159.

Ohnishi, K., Kutsukake, K., Suzuki, H., and Iino, T. (1990)Gene fliA encodes an alternative sigma factor specific forflagellar operons in Salmonella typhimurium. Mol GenGenet 221: 139–147.

O’Rear, J., Lindianne, A., and Harshey, R. (1992) Mutationsthat impair swarming motility in Serratia marsescens 274include but are not limited to those affecting chemotaxisand flagellar function. J Bacteriol 174: 6125–6187.

Perombelon, M., and Kelman, A. (1980) Ecology of the softrot Erwinias. Annu Rev Phytopathol 18: 361–387.

Pirhonen, M., Saarilahti, H., Karlsson, M., and Palva, E.(1991) Identification of pathogenicity determinants of Erwinia

carotovora ssp. carotovora by transposon mutagenesis.Mol Plant–Microbe Interact 4: 276–283.

Pirhonen, M., Flego, D., Heikinheimo, R., and Palva, E. (1993)A small diffusible signal molecule is responsible for theglobal control of virulence and exoenyme production inErwinia carotovora. EMBO J 12: 2467–2476.

Reeves, P., Whitcombe, D., Wharam, S., Gibson, M., Allison,G., Bunce, N., et al. (1993) Molecular cloning and charac-terisation of 13 out genes from Erwinia carotovora sub-species carotovora: genes encoding members of a generalsecretion pathway (GSP) widespread in Gram-negativebacteria. Mol Microbiol 8: 443–456.

Regue, M., Fabregat, C., and Vinas, M. (1991) A generalizedtransducing bacteriophage for Serratia marcescens. ResMicrobiol 142: 23–28.

Sambrook, J., Fritsch, E., and Maniatis, T. (1989) MolecularCloning: A Laboratory Manual, 2nd edn. Cold Spring Har-bour, NY: Cold Spring Harbour Laboratory Press.

Schell, M. (1993) Molecular biology of the LysR family of tran-scriptional regulators. Annu Rev Microbiol 47: 597–626.

Strauss, E. (1995) When a turn off is a turn on. Curr Biol 5:706–709.

Stromberg, V., Lay, G., and Kyostio, S. (1994) Erwiniacarotovora ssp. carotovora: protease in soft rot (abstract).Phytopathology 84: 1109.

Surgey, N., Robert-Baudouy, J and Condemine, G. (1996)The Erwinia chrysanthemi pecT gene regulates pectinasegene expression. J Bacteriol 178: 1593–1599.

Tabor, S., and Richardson, C. (1985) A bacteriophage T7RNA polymerase/promoter system for controlled exclusiveexpression of specific genes. Proc Natl Acad Sci USA 82:1074–1078.

Tang, X., Nakata, Y., Li, H., Zhang, M., Gao, H., Fujita, A., etal. (1994) The optimization of preparation of competentcells for transformation of E. coli. Nucleic Acids Res 22:2857–2858.

Throup, J., Camara, M., Briggs, G., Winson, M., Chhabra, S.,Bycroft, B., et al. (1995) Characterisation of the yenI/yenRlocus from Yersinia enterolitica mediating the synthesis oftwo N-acylhomoserine lactone signal molecules. Mol Micro-biol 17: 345–356.

Toth, I. (1991) The isolation of novel Erwinia phages andtheir use in the study of bacterial phytopathegenicity. PhDthesis, University of Warwick, Coventry, UK.

Toth, I., Mulholland, V., Cooper, V., Bentley, S., Shih, Y., Perom-belon, M., et al. (1997) Generalized transduction in thepotato blackleg pathogen Erwinia carotovora ssp. atrosep-tica by bacteriophage fM1. Microbiology 143: 2433–2438.

Walker, D., Reeves, P., and Salmond, G. (1994) The majorsecreted cellulase, CelV, of Erwinia carotovora ssp. caroto-vora is an important soft rot virulence factor. Mol Plant–Microbe Interact 7: 425–431.

Wilmes-Reisenberg, M., and Wanner, B. (1992) TnphoA andTnphoA8 elements for making and switching fusions forstudy of transcription, translation and cell surface localis-ation. J Bacteriol 174: 4558–4575.

Q 1998 Blackwell Science Ltd, Molecular Microbiology, 28, 705–717

hexA regulates multiple virulence determinants 717