Analysis of Shigella flexneri Wzz (Rol) function by mutagenesis and cross-linking: Wzz is able to...

Analysis of Shigella ¯exneri Wzz (Rol) functionby mutagenesis and cross-linking: Wzz is ableto oligomerize

Craig Daniels and Renato Morona*

Department of Microbiology and Immunology, The

University of Adelaide, Adelaide, South Australia,

Australia 5005.

Summary

The modal length or degree of polymerization (dp) of

the Shigella ¯exneri O-antigen is determined in an

unknown manner by the Wzz/Rol protein. The Wzz

protein is anchored into the cytoplasmic membrane

by two transmembrane domains (TM1 amino acids

32±52; TM2 amino acids 295±315) with the central

loop of the protein located in the periplasm. Plasmids

were constructed encoding hybrid Wzz proteins con-

sisting of regions of S. ¯exneri Wzz (WzzSF ) and Sal-

monella typhimurium Wzz (WzzST ). These imparted

O-antigen modal chain lengths that implied that

the carboxy-terminal region of Wzz was involved in

chain length determination. Site-directed mutagenesis

was undertaken to investigate the functional signi®-

cance of highly conserved residues in amino-/car-

boxy-terminal domains of WzzSF. Some of the WzzSF

variants resulted in O-antigen modal chain lengths

much shorter than those of wild-type WzzSF, whereas

other mutants inactivated WzzSF function entirely and

a third class had a longer O-antigen chain length dis-

tribution. The data indicate that amino acids through-

out the length of the WzzSF protein are important

in determination of O-antigen modal chain length.

In vivo cross-linking experiments were performed to

investigate the interactions between Wzz proteins.

The experiments indicated that the WzzSF protein is

able to form dimers and oligomers of at least six

WzzSF proteins. A carboxy-terminal-truncated WzzSF

protein having the amino terminal 194 amino acids

was able to oligomerize, indicating that the amino-

terminal region is suf®cient for the Wzz±Wzz inter-

action observed. Shortened WzzSF proteins having

internal deletions in the amino-terminal region were

also able to oligomerize, suggesting that residues

59±194 are not essential for oligomerization. Cross-

linking of WzzSF proteins with mutationally altered

residues showed that loss of WzzSF function may

be correlated to a reduced/altered ability to form oli-

gomers, and that mutational alteration of glycine resi-

dues in the TM2 segment affects WzzSF±WzzSF dimer

mobility in SDS polyacrylamide gels. These results

provide the ®rst evidence of protein±protein interac-

tions for proteins involved in O-antigen polysaccharide

biosynthesis.

Introduction

Complex glycolipids such as lipopolysaccharides (LPSs)

are characteristic of the outer membrane of Gram-nega-

tive bacteria. LPSs consist of three covalently linked com-

ponents: lipid A; a core sugar region; and a polymerized

chain of sugar repeat units, representing the O-antigen.

Genes required for the biosynthesis of O-antigen in Shigella

¯exneri are located in the rfb region (Macpherson et al.,

1991; 1994). The O-antigen biosynthesis process is cur-

rently believed to involve assembly of a tetrasaccharide

repeat unit on the lipid carrier bactoprenol, transfer of the

repeat unit to the periplasmic side of the membrane by

Wzx (Liu et al., 1996; Reeves et al., 1996), then polymer-

ization of the repeat units by Wzy (O-antigen polymerase),

and ®nally ligation to the lipid A-core oligosaccharide by

WaaL (O-antigen ligase). In S. ¯exneri, the number of

O-antigen repeat units attached to the lipid A-core is non-

randomly distributed (<11±16 repeats) and this modal

length [or degree of polymerization (dp)] is regulated in

an unknown manner by the wzz/rol gene product (Morona

et al., 1995). In some S. ¯exneri strains, an additional

population of LPS molecules exists, having a modal length

of $ 90 repeats; this is determined by the Cld protein

encoded on a small plasmid (pHS-2) (Stevenson et al.,

1995). Other bacteria that produce O-antigen by a Wzy-

dependent mechanism, such as Escherichia coli (Liu

and Reeves., 1994) and Salmonella typhimurium (Batche-

lor et al., 1992), also have a homologous wzz gene which

imparts a characteristic modal length to the respective

O-antigen chains.

The Wzz proteins are characterized by two conserved

transmembrane (TM) domains located in the amino-termi-

nal (TM1) and carboxy-terminal (TM2) regions, and have

Molecular Microbiology (1999) 34(1), 181±194

Q 1999 Blackwell Science Ltd

Received 19 May, 1999; revised 19 July, 1999; accepted 21 July,1999. *For correspondence. E-mail [email protected]; Tel. (�61) 8 8303 4151; Fax (�61) 8 8303 4362.

a large hydrophilic central domain located in the periplasm

(Morona et al., 1995). The Wzz homologues in members

of the Enterobacteriaceae and other bacteria share a con-

sensus proline-rich motif `PX2PX4SPKX1X10GGMXGAG8

located just before and within (underlined) TM2 (Becker

et al., 1995; Becker and Puhler, 1998). Several residues

just before and within TM1 of Wzz proteins are also highly

conserved. Wzz homologues (paralogues), including those

from members of the Enterobacteriaceae, the Wzc and

ExoP proteins, have been grouped into the MPA1 and

MPA2 (cytoplasmic membrane periplasmic auxillary pro-

teins) families of proteins involved in assembly of bacter-

ial surface polysaccharides (Paulsen et al., 1997).

The Wzz proteins show amino acid sequence similarity

with proteins associated with a diverse range of bacterial

polysaccharide biosynthesis systems. Streptococcus pneu-

moniae 19F, which produces a capsular polysaccharide

(CPS), encodes a gene cpsC, the product of which has

amino acid similarity at its carboxy-terminal end to Wzz

proteins (Guidolin et al., 1994). Acidic exopolysaccharide

succinoglycan (EPS I) produced by Sinorhizobium meliloti

requires ExoP, which also shows amino acid similarity to

Wzz proteins and also has the proline-rich motif described

above (Becker et al., 1995). These CPS/EPS systems

have an additional protein/polypeptide domain, with an

ATP-binding motif; in the case of S. pneumoniae, this is

a separate protein encoded by cpsD, whereas S. meliloti

exoP encodes a single protein containing both the Wzz

homology region and a domain with an ATP-binding motif

(Paulsen et al., 1997).

Two models have been proposed to explain the function

of Wzz proteins. Bastin et al. (1993) suggested that Wzz

interacts with Wzy and acts as a molecular timer allowing

polymerization by Wzy to continue for a set amount of

time, thereby resulting in consistent addition of repeat

units during polymerization. The alternative hypothesis

proposed by Morona et al. (1995) suggests that Wzz

acts as a molecular chaperone, facilitating the interaction

between Wzy and WaaL (the O-antigen ligase), with modal-

ity resulting from a given ratio of Wzy and WaaL. Recently,

we published data indicating the importance of the ratio

of Wzy to Wzz in determination of O-antigen chain length

distribution (Daniels et al., 1998). Additionally, the results

of Amor and Whit®eld (1997) indicate a pivotal role for

WaaL in the process of O-antigen chain length regulation,

and provide some support for the Morona et al. (1995)

model.

It has previously been reported that Wzz proteins can

in¯uence the modal chain lengths of heterologous O-anti-

gens (Batchelor et al., 1992; Bastin et al., 1993; Burrows

et al., 1997; Klee et al., 1997). A number of studies have

attempted to de®ne which region of the Wzz protein func-

tions in determining the modal length. Klee et al., 1997

compared the modal chain lengths of O-antigen from a

S. ¯exneri strain harbouring the Wzz proteins of E. coli

K-12, S. dysenteriae 1, and S. ¯exneri Y, and found them

to be quite different. The primary amino acid sequences

of these three Wzz proteins were shown to be almost iden-

tical with differences in only nine positions, ®ve of which

were conservative substitutions. Franco et al. (1998) per-

formed site-directed mutagenesis on an E. coli O2 wzz

and assessed their effect on the E. coli O111 O-antigen

modal length distribution. They reported that amino acids

distributed throughout the length of Wzz affected chain

length and concluded that modal value determination

may be an overall property of the protein. Becker and

Puhler. (1998) reported mutagenesis studies on the

ExoP protein proline-rich motif. Two of seven mutations

in this region affected EPS I production by increasing the

production of low-molecular-weight EPS I at the expense

of high-molecular-weight EPS I, highlighting the impor-

tance of this motif.

In this study, we have investigated the structure and

function of the S. ¯exneri Wzz protein. We used hybrid

proteins and heterologous complementation to attempt to

localize functional regions. Site-directed mutagenesis of

conserved and non-conserved residues was also used

to indicate which residues were essential for function and

which were directly involved in chain length determination.

A polyclonal anti-WzzSF serum was developed and used

to follow Wzz protein production and protein interactions.

We also present the ®rst biochemical evidence obtained

by in vivo cross-linking for an interaction between Wzz pro-

teins. Wzz proteins form homo-oligomers of at least six

units and the dimeric form of Wzz is highly stable. The abil-

ity of mutant and deleted Wzz proteins to form oligomers

was also investigated and correlated with their phenotypic

impact on O-antigen modal length.

Results

Comparative analysis of S. ¯exneri and S. typhimurium

Wzz proteins

To investigate the location within the S. ¯exneri Wzz

(WzzSF ) of residues functioning in O-antigen chain length

determination, we compared the WzzSF amino acid sequ-

ence with that of Salmonella enterica Typhimurium Wzz

(WzzST ). The Wzz proteins of S. ¯exneri (WzzSF) and S.

enterica Typhimurium (WzzST ) are 72% identical, differ-

ing mainly in the periplasmic domain ¯anked by the two

transmembrane domains (TM1 and TM2). Although quite

similar, the proteins impart signi®cantly different modal

chain lengths to the O-antigen of the LPSs in their wild-

type strains WzzSF (11±16 repeats) and WzzST (19±30

repeats). Franco et al. (1998) have recently classi®ed

modal lengths imparted by Wzz proteins into three cate-

gories: short (S-type; 7±16 repeats), intermediate (I-type;

Q 1999 Blackwell Science Ltd, Molecular Microbiology, 34, 181±194

182 C. Daniels and R. Morona

10±18 repeats) and long (L-type; 16±25 repeats). Under

this classi®cation, WzzSF would be S-type and WzzST an

L-type.

The wzz genes from S. ¯exneri 2457T and S. enterica

Typhimurium LT2 were cloned into the T7 overexpression

vector pRMCD77 (see Experimental procedures ; Fig. 1A),

and the plasmids introduced into S. ¯exneri wzz strain

RMA696. LPS from trans-complemented strains were com-

pared with the LPS pattern of the S. ¯exneri parent strain

2457T on an SDS 15% polyacrylamide gel (Fig. 2). As

expected, the strain containing the wzzSF (pRMCD78)

gave an identical LPS phenotype to that of the wild-type

strain (S-type; Fig. 2) and plasmid pRMCD80 (wzzST ) con-

ferred a modal chain length of (19±31 repeats; L-type),

approximately double that determined by wzzSF (11±16

repeats). To localize the residues involved in determining

the modal chain length to either the amino or carboxy

end of the protein, we constructed hybrid Wzz proteins

by fusing the two wzz genes at the common Bgl II site

(Fig. 1A) (see Experimental procedures ). The plasmid

encoding the N-WzzSF::WzzST-C protein (pRMCD106)

when used to complement RMA696 resulted in production

of LPSs with a modal length of 17±26 repeats (L-type; Fig.

2), which is close to that observed for WzzST. The strain

containing the plasmid encoding the N-WzzST::WzzSF-C

protein (pRMCD104) produced LPSs with a modal length

of 14±19 repeats (I-type; Fig. 2), which is similar to but

longer than that produced by the action of WzzSF. This

indicated that residues involved in chain length determina-

tion may be located in the carboxy-terminal region of the

Wzz protein. Klee et al. (1997) compared the amino acid

sequences of E. coli K-12, S. dysenteriae type 1 and S.


Fig. 1. Schematic representation of the Wzzconstructs.A. Structure of the WzzSF, WzzST,N-WzzST::WzzSF-C and N-WzzSF::WzzST-Cproteins. Wild-type and hybrid genes werecloned into vector pRMCD77 (seeExperimental procedures ) so they could beexpressed from the T7 RNA polymerasepromoter. Hybrid proteins were constructedusing the common Bgl II site indicated in the®gure. TM1 and TM2 representtransmembrane segment one andtransmembrane segment two respectively.B. WzzSF truncation/deletion proteins usedin formaldehyde cross-linking experiments.Deletion points are indicated by the ¯ankingresidues (Gln-161-D-Asp-194). Constructnames are indicated on the left of the ®gure.The predicted molecular weights of theproteins are indicated on the right side ofthe ®gure (in kilodaltons).C. Location of the mutated regions withinthe WzzSF protein. Aligned proteins arefrom Shigella ¯exneri (WzzSF ) (X71790),Salmonella enterica Typhimurium LT2(WzzST ) (Z17278), E. coli E4991/76(WzzE4991/76) (AF011910), Shigelladysenteriae (WzzSD) (Y07560), E. coli K12(WzzK12) (Y07559), and Sinorhizobiummeliloti (ExoPSM) (P33698). Thetransmembrane regions are overlined andresidues of the proline-rich motif are in bold.The mutated residues are indicated (Residue#) and the amino acid change is shown. Theamino acid positions of each protein areindicated on the right of the ®gure (inbrackets). Speci®c amino acid substitutionsin WzzSF were introduced by site-directedmutagenesis (see Experimental procedures ).

Wzz is an oligomer 183

¯exneri Y Wzz proteins, and reported that, although they

differed in nine positions, WzzSF differed signi®cantly at

only two sites (amino acid 267 and amino acid 270). At

position 267, the WzzSF has a basic residue (Lys) whereas

WzzST has a polar residue (Asn). We changed this residue

in WzzSF (K267 ! N) using site-directed PCR mutagen-

esis, creating plasmid pRMCD108 (see Experimental pro-

cedures ). RMA696 containing this plasmid had LPSs with

an increased modal chain length (13±20 repeats; I-type;

Fig. 2). This result showed that residue 267 is involved in

O-antigen chain length determination, but clearly other

residues are also involved in producing the modal length

(L-type) conferred by WzzST.

Mutational analysis of conserved Gly residues of TM2

The carboxy-terminal consensus region includes three

proline residues just before TM2 and four glycine residues

located within TM2 (Fig. 1C). We investigated the role in

O-antigen modal chain length determination of this highly

conserved region by using site-directed mutagenesis to

mutate the proline and glycine residues to alanine (Fig.

1C) (see Experimental procedures ). Initially, we changed

the Gly305, Gly306, Gly309 and Gly311 to alanine resi-

dues. Mutational alteration of single residues G305 ! A

or G311 ! A had no effect on the modal chain length of

LPSs when compared with the wild type (Fig. 3). The

double mutation G305 ! A/G309 ! A and the triple muta-

tion G305 ! A/G306 ! A/G309 ! A also had no effect on

the modal chain length, however the dual change of

G305 ! A/G311 ! A resulted in a marked reduction in the

modal length [3±8 repeats; very short (VS) type] (Fig. 3).

These data indicate that relatively conservative single

residue changes in the glycine-rich domain have no effect

on Wzz O-antigen chain length regulation. However, some

multiple changes (G305 ! A/G311 ! A) result in a Wzz

which confers a signi®cantly reduced O-antigen modal

chain length.

Mutational analysis of conserved Pro residues

proximal to TM2

The function of the Pro residues in the motif was investi-

gated. While the mutational alteration of P283 ! A had no

effect on O-antigen chain length pro®le (Fig. 3), the muta-

tion of P286 ! A resulted in a Wzz protein with diminished

ability to confer a modal chain length as a reduced level of

O-antigen chains with a wild-type modal length could be

seen (Fig. 3). Mutation of P292 ! A resulted in complete

loss of function of the protein as RMA696 harbouring

this plasmid (pRMCD116) had an LPS phenotype identical

to the parent wzz mutant strain (RMA696) (Fig. 3). Hence,

at least two of the three proline residues in the highly con-

served carboxy-terminal end of WzzSF are clearly essen-

tial for wild-type function of the protein in regulating modal

length.

Mutational analysis of conserved residues proximal

to and within TM1

Wzz proteins, particularly the closely related proteins from

the Enterobacteriacea, have many residues in their amino-

terminal transmembrane domains (TM1) that are highly


Fig. 2. Effect of hybrid Wzz proteins on LPS structure. This showsa silver-stained SDS 15% polyacrylamide gel withlipopolysaccharide prepared from the indicated strains. The ®rstlane contains the 2457T wild-type strain and the second lane theS. ¯exneri wzz ::KmR strain RMA696. All other tracks containRMA696 harbouring the indicated plasmid. Rough LPS (R-LPS)uncapped by O-antigen is indicated on the right side of the ®gure.The number of O-antigen repeats are shown on the left side of thegel. Samples represent 1±2 ´ 108 cells.


conserved (Fig. 1C). We decided to test the importance of

the highly conserved KMTIII motif located just adjacent to,

and within, TM1. The mutational change I35 ! C had no

effect on the modal length conferred by WzzSF, however

WzzSF with a double change of I35 ! C and M32 ! T

was functional but conferred a reduced modal length on

the LPS O-antigen chain (3±8 repeats; VS-type; Fig. 3).

WzzSF with a single change of M32 ! T resulted in LPS

with modal length of 10±15 repeats; this was a slight

reduction in modal length compared with the wild type.

The change of K31 ! A resulted in loss of activity as

seen by the inability to complement the wzz defect in

RMA696 (Fig. 3). Residues in the TM1 region are clearly

essential for function and, like TM2, mutations can result

in either a dramatically reduced O-antigen modal chain

length (VS-type) or the entire loss of Wzz function.

The wzzSF plasmids described above were used in T7

polymerase overexpression experiments to ensure that

failure to complement the wzz mutant strain was not due

to an inability of the plasmids to express the various mutant

WzzSF proteins. All vectors and wzzSF plasmids were intro-

duced into E. coli DH5 containing pGP1-2 (E2096), which

encodes T7 RNA polymerase under lambda cI control

(Tabor and Richardson, 1985). T7 expression followed by

electrophoresis and Western immunoblotting using anti-

WzzSF antibodies (see Experimental procedures ) indicated

that WzzSF or mutated WzzSF could be produced from all

of the constructs (Fig. 6). The wzzST, wzzST::wzzSF and

wzzSF::wzzST plasmids were also checked for their ability

to produce Wzz-related proteins (Fig. 4A). The anti-WzzSF

serum was also able to detect WzzST and the hybrid

proteins. Interestingly, low-molecular-weight crossreac-

tive bands were visible in tracks containing WzzSF, N-

WzzST::WzzSF-C and N-WzzSF::WzzST-C but not WzzST.

The crossreactive species may be breakdown products,

and the differential detection suggests that either the

anti-WzzSF serum could not detect WzzST breakdown pro-

ducts or that WzzST is more stable than WzzSF.


Fig. 3. Effect of Wzz point mutations onfunction. Silver-stained SDS 15%polyacrylamide gels showinglipopolysaccharide (LPS) prepared from theindicated strains. Lanes containing RMA696and its derivatives are indicated at the bottomof the gels by arrows. The plasmidsharboured within the strains are indicated atthe top of the ®gure. Rough LPS (R-LPS)uncapped by O-antigen is indicated on theright side of the ®gure. The number of O-antigen repeats are listed on the left side ofthe gel. Samples represent 1±2 ´ 108 cells.


Detection of WzzSF oligomer formation by in vivo

cross-linking

During the assessment of Wzz production by Western

immunoblotting, we frequently detected a band with an

apparent molecular mass of 72 kDa equivalent in size to a

WzzSF dimer (Fig. 4B). The (72 kDa species was observed

in samples that had been heated to 1008C in the presence

of SDS, and no species of > 72 kDa was observed in these

samples (Fig. 4B). This suggested that WzzSF was able to

oligomerize in vivo and we undertook chemical cross-link-

ing experiments to investigate this possibility. Cross-link-

ing was performed using formaldehyde on E. coli E2096

containing pRMCD78 with the wild-type wzzSF gene (see

Experimental procedures ). Whole cell samples were

separated on SDS 10% polyacrylamide gels and analysed

by Western immunoblotting using anti-WzzSF antibodies.

As can be seen in Fig. 5, treatment with 0.5% formaldehyde

generated species migrating as <72 kDa <160 kDa and

<210 kDa in addition to the 36 kDa WzzSF protein band.

These species correspond closely to two, four or six times

the apparent molecular weight of WzzSF protein. Species

of higher molecular weight migrating more slowly than

the <210 kDa form were also detected, however these

appear to have barely entered the separating gel and

are too large to have their size extrapolated from the pro-

tein standards used. The molecular species migrating at

<72 kDa appeared to be a doublet. The doublet could

be due to the incorporation of the smaller Wzz-related pro-

tein seen in the untreated sample (Fig. 5). The reappear-

ance of the smaller product after heating to destroy the

formaldehyde cross-linking supports this suggestion.

The very large species, along with the <160 kDa and

<210 kDa species, could not be detected when the

samples were heated at 1008C for 20 min in sample buffer

containing SDS before electrophoresis (Fig. 5). The dimeric

form of <72 kDa, however, was still detected after heating

the cross-linked samples at 1008C (Fig. 5). These data

indicate that WzzSF can be cross-linked to form an oligo-

meric complex of a size corresponding to at least a hex-

amer, and the dimeric form of cross-linked WzzSF is very

stable as it is not readily dissociated. The cross-linking

data correlate well with our initial observation that the

WzzSF protein was able to form a dimer in the absence

of any cross-linking reagent, even after being heated to

1008C in the presence of SDS.

In vivo cross-linking of WzzSF with altered residues

It was possible that the altered function/lack of function

noticed for some of these proteins could be related to an

inability to form oligomers. We investigated the possible

correlation between WzzSF protein function and its oligo-

meric state by testing all of the mutant Wzz proteins for

their ability to be cross-linked. E. coli DH5 strains (E2096)

harbouring the constructs were subjected to in vivo formal-

dehyde cross-linking and whole cell samples were solubil-

ized, electrophoresed and analysed by immunoblotting

with anti-WzzSF antibodies. All of the mutant WzzSF pro-

teins generated from these constructs were able to form

oligomeric complexes similar to those of the wild-type

WzzSF (Fig. 6). The hybrid constructs producing N-

WzzSF::WzzST-C, and N-WzzST::WzzSF-C were also

able to form oligomeric complexes equivalent to that of


Fig. 4. Western immunoblot of E. coli E2096strains producing Wzz proteins.A. After electrophoresis on an SDS 15%polyacrylamide gel and transfer tonitrocellulose, proteins were detected usingaf®nity-puri®ed anti-WzzSF antibodies (seeExperimental procedures ). Lanes containwhole cell samples of bacteria (equivalentto 1 ´ 108 cells) harbouring the plasmidsindicated at the top of the ®gure. Migrationpositions of the molecular mass standards(Pharmacia) are indicated on the right side(in kilodaltons): soybean trypsin inhibitor(20.1), carbonic anhydrase (30), ovalbumin(43), bovine serum albumin (67) andphosphorylase b (94).B. Immunoblot showing WzzSF expressedin E. coli E2096 is able to dimerize in thepresence of SDS. Whole cell samples(equivalent to 1 ´ 108 cells) were heated at1008C for 5 min before electrophoresis (SDS10% polyacrylamide gel) and immunoblotting.The WzzSF (36 kDa) protein is indicated onthe right of the ®gure and the dimeric form(<72 kDa) is indicated with an asterix.Migration positions of molecular massstandards as in A are indicated in kilodaltonson the right side.


the wild type WzzSF (Table 1). However, strains contain-

ing plasmid constructs pRMCD116 (WzzSF(P292A)) and

pRMCD119 (WzzSF(K31A)), both of which are unable to

impart an O-antigen modal length on RMA696 LPS, formed

less of the dimeric form and signi®cantly less of the higher

molecular weight species (Fig. 6). Strains containing plas-

mids pRMCD112, pRMCD113 and pRMCD114, encoding

wzzSF genes with G305 ! A/G309 ! A, G305 ! A/

G311 ! A and G305 ! A/G306 ! A/G309 ! A respec-

tively, show a difference in the apparent molecular mass

of the Wzz protein oligomers compared with the wild-type

Wzz (Fig. 6). This difference is particularly noticeable in

the region where the Wzz dimer migrates; the dimeric

forms of the mutated proteins have a slightly larger appar-

ent molecular mass (<80 kDa; Fig. 6; Table 1). Taken

together, these results suggest that loss of function may

be linked to a reduced/altered ability to form oligomers,

and the glycine residues in the TM2 segment in¯uence

WzzSF±WzzSF dimer conformation as the mutations affect

WzzSF±WzzSF dimer mobility in SDS polyacrylamide gels.

Formaldehyde and dithio-bis(succinimidylpropionate)

(DSP) cross-linking of Wzz in S. ¯exneri

To con®rm the cross-linking results obtained using formal-

dehyde on overexpressed WzzSF proteins in E. coli, we

performed cross-linking on wild-type S. ¯exneri. Formal-

dehyde is a small reactive molecule capable of polymeriz-

ing to a variety of lengths (Prossnitz et al., 1988), whereas

dithio-bis(succinimidylpropionate) (DSP) is a ®xed-arm-

length cross-linking reagent 12 AÊ in length. Cross-linking

was performed on wild-type S. ¯exneri Y (PE638) and E.

coli E2096 harbouring pRMCD78, using either 0.5% for-

maldehyde or 0.2 mM DSP after which cells were fractio-

nated (see Experimental procedures ). Whole cell and

cytoplasmic membrane fractions of both treated and

untreated bacteria were electrophoresed on an SDS 10%

polyacrylamide gel and Wzz was detected by immunoblot-

ting with anti-WzzSF antibodies. The results obtained for

the formaldehyde cross-linking were identical for both

the E. coli and S. ¯exneri samples (Table 1). Electrophore-

tic species corresponding to <72 kDa, <160 kDa and

<210 kDa were observed in all formaldehyde-treated

samples as expected (Table 1). The <210 kDa species

were also present in both E. coli and S. ¯exneri samples.

The DSP cross-linked samples showed a similar Wzz pro-

®le to that observed when using formaldehyde, however in

this case the Wzz dimer had two different apparent mole-

cular masses (<72 kDa, <77 kDa; Table 1). An additional

difference was noticed between the E. coli and S. ¯exneri

DSP cross-linked samples. DSP cross-linked samples

indicated only a dimeric form of the protein in S. ¯exneri,

however the high-molecular-weight species were detectable

in E. coli E2096 (pRMCD78) using this cross-linker. This

difference is likely to be a consequence of altered cross-

linking ef®ciency due to strain differences. The cross-link-

ing with DSP largely con®rmed the results obtained with

formaldehyde, and the data obtained using S. ¯exneri

correlated with what is observed in E. coli harbouring

plasmid-encoded WzzSF. We were unable to detect

WzzSF±WzzSF dimer in any S. ¯exneri samples in the

absence of cross-linking reagent.

The N-terminal domain of WzzSF is suf®cient for

oligomer formation

Plasmid pRMCD107 was initially constructed to identify

regions within WzzSF that are required for oligomer forma-

tion (Fig. 1B). Strains harbouring this construct were able

to produce a truncated WzzSF protein of <23 kDa which,

like the wild-type WzzSF, was able to dimerize in the


Fig. 5. Formaldehyde cross-linking of WzzSF in E. coli. Aftertreatment with 0.5% formaldehyde, whole cell samples wereelectrophoresed on an SDS 10% polyacrylamide gel and analysedby immunoblotting using af®nity-puri®ed anti-WzzSF antibodies (seeExperimental procedures ). Formaldehyde (0.5% form.) treated (�)and untreated (ÿ) samples are indicated at the bottom of the ®gure.Samples were heated at 608C for 10 min or at 1008C for 5 minbefore loading. The migration positions of the prestained molecularmass standards (New England Biolabs) are indicated on the rightside (in kilodaltons): triosephosphate isomerase (32.5), aldolase(47.5), glutamic dehydrogenase (62), MBP-paramyosin (83), MBP-b-galactosidase (175). Samples represent 1±2 ´ 108 cells.


absence of cross-linking reagent (data not shown). Formal-

dehyde cross-linking of E. coli harbouring pRMCD107 con-

®rmed a dimeric species with an apparent molecular mass

of <48 kDa, and additional species were also detected

migrating as <65 kDa, <75 kDa and <210 kDa (Table 1).

These data show that the carboxy-terminal end of WzzSF

is not needed for dimerization, and that a protein±protein

interactive domain may be located in the amino-terminal

194 residues of WzzSF. To further investigate this phenom-

enon, we constructed three plasmids with sequential inter-

nal deletions within the amino-terminal residues of WzzSF.

Plasmids pRMCD138, pRMCD139 and pRMCD140 have

internal deletions of 33, 80 and 135 amino acids (Fig.

1B), resulting in Wzz proteins with apparent molecular

weights of 33, 28 and 22 kDa respectively (Table 1). Intro-

duction of pRMCD107, pRMCD138, pRMCD139 and

pRMCD140 into S. ¯exneri RMA696 did not allow restora-

tion of O-antigen modal chain length (Table 1). Initially, for-

maldehyde cross-linking of E. coli strains expressing these

proteins indicated they were able to oligomerize, however,

apart from the monomer, only very high molecular weight

species (<210 kDa) were detected. This may be a conse-

quence of reduced levels of the truncated proteins and/or

their reduced reactivity with anti-WzzSF antibodies. Sub-

sequent cross-linking experiments with strains harbouring

pRMCD138, pRMCD139 and pRMCD140, followed by

immunoblotting of over-loaded samples and increased

exposure times indicated that the internally deleted WzzSF

proteins were able to dimerize (Table 1). The apparent

molecular mass of the dimeric Wzz species generated from

cross-linking strains containing pRMCD138 (<67 kDa),

pRMCD139 (<55 kDa) and pRMCD140 (<45 kDa) were

approximately twice that of their respective monomers.

This con®rms that the <72 kDa species noted when wild-

type WzzSF is cross-linked is indeed a WzzSF dimer.

These results indicate that the TM1 region functions in

dimer formation, but do not exclude any essential roles

for the TM2 region in these interactions.


Fig. 6. Formaldehyde cross-linking of mutant WzzSF proteins in E. coli. Cells were treated as described in Experimental procedures andwhole cell samples were electrophoresed on SDS 10% polyacrylamide gels followed by immunoblotting using af®nity-puri®ed anti-WzzSF

antibodies. Formaldehyde (0.5% form.) treated (�) and untreated (ÿ) samples are indicated at the bottom of the ®gure. The migrationpositions of the prestained molecular mass standards New England Biolabs (NEB) are indicated on the right side (in kilodaltons),as describedin Fig. 5. Samples represent 1±2 ´ 108 cells.


Discussion

Recent publications have emphasized the importance of

LPS O-antigen modal chain length in bacterial pathogen-

esis and immune responses to O-antigens (Attridge et

al., 1990; Hong and Payne, 1997; Klee et al., 1997; Van

Den Bosch et al., 1997). Hong and Payne (1997) showed

that the plasmid-encoded (pHS-2) Wzz/Cld protein of S.

¯exneri 2a (2457T), which imparts a modal length of

<90 repeats, is required for serum resistance. Hong and

Payne (1997) and Van Den Bosch et al. (1997) have

also shown that the chromosomally encoded Wzz protein

of S. ¯exneri 2a is required to achieve wild-type levels of

intracellular and intercellular spread in HeLa cells. Sereny

tests have also indicated that S. ¯exneri 2a wzz chromoso-

mal mutants are avirulent (Van Den Bosch et al., 1997).

The last results appear to be due to abnormal cellular loca-

lization of the normally polar located IcsA protein in the

wzz mutant. Klee et al. (1997) found that introduction of

heterologous wzz genes into a potential vaccine carrier

strain resulted in reduced masking of a heterologous

O-antigen and a concomitant increase in the immune

response of mice to the desired O-antigen. Despite their

biological importance in determination of O-antigen chain

length distribution, very little is known about the structure

and function of Wzz proteins. In this study, we attempted

to localize functional domains within the S. ¯exneri Wzz

protein. Previous attempts to localize a functional region

within E. coli Wzz proteins indicated that a range of amino

acid positions throughout the Wzz protein affected the

modal chain length (Franco et al., 1998). We found that

the Wzz protein from S. enterica Typhimurium gave an

increased O-antigen modal length in S. ¯exneri wzz strain

RMA696 (Table 1). The hybrid constructs which encode

N-WzzST::WzzSF-C and N-WzzSF::WzzST-C showed that

residues in the carboxy-terminal end of Wzz have a major

impact on the O-antigen modal chain length. Although

neither of the hybrid proteins were able to impart a modal

chain length that is identical to the two wild-type genes

used (wzzSF and wzzST ), N-WzzSF::WzzST-C (17±26

repeats) gave a modal value very close to that of the wild-

type WzzST (19±31) (Table 1). Change of the basic residue

(Lys) at position 267 in the carboxy-terminal region of

WzzSF to a polar residue (Asn) caused an increase in

the modal length from 11±16 repeat units (S-type) to

14±19 repeat units (I-type). This modest increase in length

indicates that this residue is involved in chain length deter-

mination. However, it is now obvious from our data, and that

presented by Franco et al. (1998), that residues throughout

the Wzz protein can have subtle effects on O-antigen chain

length.

To date, no experimental data for the functional sig-

ni®cance of the highly conserved glycine residues have

been reported. However, the complete conservation of


Table 1. Summary of Wzz constructs and phenotypes.

Apparent Wzz Apparent dimerPlasmid/strain Mutation/characteristics size (kDa) size (kDa)a Apparent oligomer size (kDa)a Modal lengthb

PE638 S. flexneri Y 36 72 [72, 77] 160, 210, > 210 11±16pRMCD78c Wild-type WzzSF 36 72 [72, 77] 160, 210, > 210 [160, 210, > 210] 11±16pRMCD80c Wild-type WzzST 35 ND ND 19±31pRMCD104c N-WzzST::WzzSF-C 35 70 160, 210, > 210 14±19pRMCD106c N-WzzSF::WzzST-C 36 72 160, 210, > 210 17±26pRMCD107c WzzSFD194±325 23 48 65, 75, > 210 Non-modalpRMCD108c WzzSF(K267 ! N) 36 72 160, 210, > 210 13±20pRMCD109c WzzSF(P283 ! A) 36 72 160, 210, > 210 11±16pRMCD111c WzzSF(G311 ! A) 36 74 160, 210, > 210 11±16pRMCD112c WzzSF(G305 ! A/G309) 36 80 160, 210, > 210 11±16pRMCD113c WzzSF(G305 ! A/G311 ! A) 36 80 160, 210, > 210 3±8pRMCD114c WzzSF(G305 ! A/G306 ! A/G309 ! A) 36 80d 160, 210, > 210d 11±16pRMCD116c WzzSF(P292 ! A) 36 72d 160, 210, > 210d Non-modale

pRMCD117c WzzSF(P286 ! A) 36 72 160, 210, > 210 11±16pRMCD119c WzzSF(K31 ! A) 36 72d 160, 210, > 210d Non-modalpRMCD121c WzzSF(I35 ! C) 36 72 160, 210, > 210 11±16pRMCD122c WzzSF(I35 ! C/M32 ! T) 36 72 160, 210, > 210 3±8pRMCD125c WzzSF(G305 ! A) 36 72 160, 210, > 210 11±16pRMCD127c WzzSF(M32 ! T) 36 72 160, 210, > 210 10±15pRMCD138c WzzSFDGln-161-Asp-194 33 67 > 210 Non-modalpRMCD139c WzzSFDGlu-114-Asp-194 28 55 > 210 Non-modalpRMCD140c WzzSFDThr-59-Asp-194 22 45 > 210 Non-modal

a. After cross-linking with 0.5% formaldehyde, or 0.2 mM DSP where indicated by square brackets.b. Average length of O-antigen repeat units.c. Base vector is pRMCD77.d. Reduced amounts of dimer and oligomer were observed.e. Reduced amount of O-antigen of modal length was observed.ND, not done.


the glycine residues in the hydrophobic transmembrane

segment (Fig. 1C) led Bastin et al. (1993) to argue for

their possible involvement in protein±protein interactions.

We targeted these residues in order to determine whether

they were essential for modal chain length function in Wzz

proteins. Single-residue Wzz variants with G305 ! A or

G311 ! A did not alter the modal length from that of the

wild-type protein (Table 1). These relatively conservative

changes may not be suf®cient to signi®cantly alter this

region of the protein, and therefore no obvious effect on

function was observed. Double and triple mutant Wzz var-

iants with either G305 ! A/G309 ! A or G305 ! A/

G306 ! A/G309 ! A also did not alter the modal length.

The dual Wzz variant having G305 ! A/G311 ! A

resulted in a dramatic change in the O-antigen chain

length distribution: a modal length of 3±8 repeats was

observed (VS-type; Fig. 3; Table 1). Glycine 305 appears

to be completely conserved; it is present in Wzz-related

proteins such as the E. coli K-12 WzzE/ORF2 protein

(involved in enterobacterial common antigen biosynthesis;

Meier-Dieter et al., 1992), the Erwinia amylovora AmsA

protein (EPS biosynthesis; Bugert and Geider, 1995),

the S. pneumoniae CpsC protein (CPS biosynthesis; Gui-

dolin et al., 1994) and the S. meliloti ExoP protein (EPS

biosynthesis; Becker et al., 1995). Glycine 311 is not as

highly conserved. These residues are located at either

end of the motif in TM2 of WzzSF, and it could be that alter-

ing both of these residues in the same WzzSF protein

results in a major change in protein conformation and/or

protein±protein interaction.

Becker and Puhler (1998) have previously studied the

proline-rich motif of the S. meliloti ExoP protein. They

found that two mutations, R443 ! l and P457 ! S, resulted

in increased expression of low-molecular-weight EPS I at

the expense of high-molecular-weight EPS I. Further muta-

tions of P451 ! S and P454 ! S caused no change in EPS

I production. Amino acid sequence alignment of ExoP with

Wzz shows that the proline residues 451 and 457 from

ExoP align with the WzzSF proline residues 286 and 292

respectively (Fig. 1C). We found that altering P283 ! A

had no effect on the function of the WzzSF protein, the

P286 ! A mutation resulted in reduced Wzz activity, and

the P292 ! A mutation completely abrogated function

(Fig. 3; Table 1). These results indicate that whereas

Pro-286 is needed for ef®cient function, Pro-292, located

in the conserved SPK motif (Becker and Puhler, 1998;

Fig. 1C), is absolutely essential for Wzz function. These

observations agree with the data obtained using ExoP,

in which mutation of P457 ! S in the conserved SPK

motif had the most dramatic affect on EPS I production.

However, recent work by GonzaÂ lez et al. (1998) has indi-

cated that the ExoP protein is involved in the production of

both high- and low-molecular-weight succinoglycan by two

alternative mechanisms, and this observation along with

the presence of an ATP-binding domain within ExoP sug-

gests that ExoP potentially has a more complex role in bio-

synthesis, making it dif®cult to correlate its function to that

of Wzz.

Alteration of residues preceding and within TM1 of WzzSF

either resulted in complete loss of function or altered func-

tion respectively. The complete loss of function resulting

from the mutation of K31 ! A was not expected. This resi-

due is highly conserved between Wzz proteins, however

it is located topologically within the cytoplasm, just before

the ®rst transmembrane region. Alteration of K31 ! A

eliminates a positively charged amino acid residue on

the cytoplasmic side of the membrane which could affect

function by destabilizing the TM1 region of the protein,

allowing either the conformation of the TM segment to be

altered or interfering with protein interactions. The single

change of I35 ! C had no effect, however changing

M32 ! T resulted in a slight reduction in O-antigen

modal length (10±15 repeats). The dual change of

I35 ! C/M32 ! T resulted in a Wzz which conferred a

dramatic reduction in modal length (3±8 repeats; Table 1).

It is interesting that dual mutations in either of the

two transmembrane regions (I35 ! C/M32 ! T or

G305 ! A G31 ! A) can result in a protein that imparts

very similar modal length. These data imply strongly that

both regions are involved either directly in modal chain

length determination and/or are required for protein±pro-

tein interaction. This could be further investigated by iden-

tifying a mutation in TM1 which suppresses the mutation in

TM2 or vice versa.

In vivo formaldehyde cross-linking has been used

extensively to study protein interactions in bacteria, includ-

ing both outer membrane proteins (Mourrain et al., 1997)

and cytoplasmic membrane proteins (Prossnitz et al.,

1988; Higgs et al., 1998). It has previously been proposed

that proteins involved in O-antigen biosynthesis may inter-

act to form a complex (Bastin et al., 1993; Morona et al.,

1995). The complex could include proteins such as Wzx,

Wzy, Wzz and WaaL, however there is currently no pub-

lished biochemical evidence to prove the existence of a

complex. In this study, we concentrated on protein inter-

actions involving the WzzSF protein. Our cross-linking

experiments using overexpressed WzzSF in E. coli DH5

indicate that WzzSF forms a homo-oligomer of at least

six units. In vivo cross-linking experiments using formalde-

hyde on wild-type S. ¯exneri also allowed detection of

WzzSF oligomers. DSP cross-linked samples indicated

only a dimeric form of the protein in S. ¯exneri, however

the high-molecular-weight species were detected in E.

coli DH5 using this cross-linker. The wild-type S. ¯exneri

strain produces S-LPS whereas the E. coli K-12 strain

used for the cross-linking experiment produces R-LPS,

furthermore DSP is hydrophobic and a larger molecule

than formaldehyde and may be hindered in its access to



the WzzSF protein by the presence of the S-LPS or other

outer membrane components in S. ¯exneri. Hence, the

inability to detect complexes larger than the dimer in S.

¯exneri when cross-linked with DSP may be due to the

decreased amounts of DSP able to associate with the

WzzSF protein. However, both the formaldehyde and DSP

cross-linking data strongly support the existence of WzzSF

dimers. In the DSP cross-linked samples, the WzzSF dimer

migrated at two different apparent molecular weights: one

corresponded to the mass observed when formaldehyde

was used (<72 kDa) as seen with the WzzSF wild type

and most Wzz mutant variant proteins (Fig. 6), and the

other migrated signi®cantly more slowly (<77 kDa). The

latter is a similar size to that seen for Wzz proteins with

multiple glycine to alanine mutational alterations (Fig. 6;

Table 1). An explanation for this variation in apparent

molecular mass is that WzzSF exists in two conformations

which can only be detected in the dimeric form of the

protein.

Plasmid pRMCD107 produces a truncated protein of

<23 kDa consisting of the amino-terminal 194 residues

of WzzSF. After formaldehyde cross-linking, several new

species were generated: <48 kDa <65 kDa <75 kDa,

> 210 kDa (Table 1). The <48 kDa species observed

when this protein was cross-linked was approximately

twice that observed of the monomer and indicates that

the amino-terminal 194 residues of WzzSF are suf®cient

to allow homo-oligomer formation. The <65 kDa and

<75 kDa species, which were relatively minor in compari-

son to the other species seen, do not correlate directly with

what was observed for the wild-type protein. We would

have expected to see species of <92 kDa and <138 kDa,

equivalent to a tetramer and hexamer respectively. The

<65 kDa species could represent a trimer, however

trimeric species were not observed for wild-type WzzSF.

The lack of the carboxy-terminal 131 residues including

TM2 from the protein produced by pRMCD107 may be

contributing to the unusual cross-linking pro®le. Deletion

proteins produced by pRMCD138, pRMCD139 and

pRMCD140 had apparent molecular weights of <33 kDa,

<28 kDa and <22 kDa respectively (Table 1). Only very

large species (> 210 kDa) could be visualized when these

proteins were cross-linked and our usual amount of sam-

ple was loaded (<1±2 ´ 108 bacterial cells). Overloading

of gels with cross-linked E. coli lysates producing the pro-

teins allowed visualization of species equivalent in size to

that expected for their respective dimers (data not shown).

This indicates that residues 59±194 are not essential for

oligomerization. Taken together, these data imply that

transmembrane segment one (TM1) is involved in the

WzzSF±WzzSF interaction process, however it does not

exclude other regions of the protein from being involved.

This study has emphasized the functional signi®cance of

Wzz oligomer formation. However, the data do not prove

or disprove either of the currently suggested models for

Wzz action (Bastin et al., 1993; Morona et al., 1995).

Both models allow for possible protein±protein interactions

between enzymes involved in O-antigen assembly/trans-

location (Wzz, Wzy, WaaL), and any subsequent models

would need to take into account Wzz±Wzz interactions.

Although there appear to be no obvious interactions

between Wzz (up to the hexameric size) and any other

proteins, the ability of Wzz to associate into very large

complexes (>210 kDa) raises the possibility that Wzy

and WaaL may also be localized in these complexes. For-

mation of a large O-antigen/LPS biosynthetic complex

including Wzz would allow centralization of the LPS

assembly/transport machinery and a concomitant increase

in the ef®ciency of lipopolysaccharide expression. Investi-

gation of the presence of proteins such as Wzy and WaaL

in the large complexes will require the development and

use of speci®c antisera.

Experimental procedures

Bacterial culture conditions

All strains were grown at 378C in Luria±Bertani broth (LB;Morona et al., 1994) unless otherwise stated, except for strainscontaining pGP1-2 (308C) (Tabor and Richardson, 1985). Anti-biotics were used at the following concentrations where appro-priate: ampicillin (Ap) 150 mg mlÿ1 and kanamycin (Km)50 mg mlÿ1.

Construction of recombinant clones

All cloning/manipulations were performed using E. coli DH5a

as the recipient strain (Bethesda Research Laboratories). TheT7 polymerase expression vector pET17b (Novagen) was thebase plasmid for all wzz constructs. pET17b was ®rst digestedwith Bgl II and end-®lled with dNTPs using Klenow followed byreligation to form pRMCD77. PCR ampli®cation was performedusing standard protocols with Amplitaq DNA polymerase (Hoff-man-La Roche). wzzSF and wzzST coding regions were PCRampli®ed from S. ¯exneri 2a (2457T) (Formal et al., 1958) andS. enterica Typhimurium LT2 (EX730), respectively, using pri-mers incorporating NdeI and BamHI restriction sites (under-lined): no. 2001, 58-CAGTTAGGCATATGATGAGAGTAG-38;no. 2002, 58-TAGGATCCGAGCAGGTGTGATGTTG-38; no.2343, 58-TAGTTAGGGTACATATGACAGTG-38; no. 2344,58-CCACCATCCGGATCCGAAGC-38. The ampli®ed DNAwas then digested with NdeI and BamHI, and cloned into like-wise-digested pRMCD77, producing pRMCD78 (wzzSF ) andpRMCD80 (wzzST ) (Fig. 1A). Ampli®ed wzzSF DNA wasalso ligated into NdeI/BamHI-digested pRE1 (Reddy et al.,1989) to form pRMCD16. The sequence of the cloned PCR-ampli®ed fragments was veri®ed by DNA sequencing usingthe T7 promoter and T7 terminator primers, as recommendedby Applied Biosystems. Chimeric wzz genes were constructedby restriction enzyme digestion of pRMCD78 and pRMCD80with XbaI and Bgl II; the two fragments generated from each



plasmid were band isolated from 1% (w/v) low-temperature-gelling agarose gels using the QIAquick Gel Extraction Kit(Qiagen). The small XbaI/Bgl II fragments (encoding theamino-terminal 194 amino acids of the Wzz proteins) fromeach plasmid were ligated with the larger fragment encodingthe carboxy-terminal end of the alternate gene. This resulted inpRMCD104 (wzzST::wzzSF ) and pRMCD106 (wzzSF::wzzST )(Fig. 1A).

Site-directed mutagenesis and deleted WzzSF

construction

Site-directed mutagenesis of codons was performed by over-lap extension using PCR and pairs of complementary primers(DNA sequences are available on request). Primer pairs wereused in combination with ¯anking primers (T7 promoter andT7 terminator) to generate mutated fusion products thatwere then fused using PCR to form the intact wzzSF gene con-taining single, double or triple codon changes (Horton, 1993).The sequence of the cloned PCR-ampli®ed fragments, includ-ing the mutations, were veri®ed by DNA sequencing using theT7 promoter and T7 terminator primers, as recommended byApplied Biosystems. Plasmid pRMCD107 was constructed bydigesting pRMCD78 with Bgl II and Not I followed by end-®llingand re-ligation (Fig. 1B). wzzSF deletion constructs were pro-duced by using the Bgl II restriction site; primers were designedincorporating a Bgl II restriction site (underlined) that boundto three different locations in the wzzSF gene: no. 2918 58-CTAGATCTTGATTCACTTTATC-38; no. 2919 58-CTAGAT-CTTCCTGATTATCCAG-38; no. 2920 58-GTGATAAGATC-TGTTGACGTCC-38. PCR using these primers and the T7promoter primer ampli®ed a DNA fragment that was thendigested with XbaI/Bgl II and ligated into band-isolatedpRMCD78 (large vector fragment containing XbaI/Bgl II)encoding the carboxy-terminal end of wzzSF. The resultingplasmids were called pRMCD138, pRMCD139 andpRMCD140 (Fig. 1B).

LPS analysis

All S. ¯exneri strains were grown for 16 h at 378C in LB con-taining appropriate antibiotics. Small-scale preparationswere made by proteinase K treatment of whole cell lysates(Hitchcock and Brown, 1983). After electrophoresis on SDS15% polyacrylamide gels, LPS was detected by silver stainingas described previously (Morona et al., 1991).

Cell fractionation, SDS±PAGE and Western

immunoblotting

Whole cell samples were prepared in sample buffer (Lugten-berg et al., 1975), except DSP cross-linked samples whichwere resuspended in sample buffer without 2-mercapto-ethanol. The remainder were fractionated to identify proteinslocated in soluble and insoluble fractions by a previously des-cribed method (Achtman et al., 1979). Wzz proteins weredetected after samples were solubilized by heating at either608C for 10 min (cross-linked) or 1008C for 5 min, then sepa-rated by SDS±PAGE and transferred to nitrocellulose mem-branes (Morona et al., 1995). Rabbit anti-WzzSF serum was

used as the primary antibody and goat anti-rabbit peroxidaseconjugate (KPL) as the secondary antibody. The blot wasdeveloped using Boehringer Mannheim chemiluminescence(POD) reagents as described by the manufacturer.

In vivo protein cross-linking

Formaldehyde cross-linking was performed by the methoddescribed by Prossnitz et al. (1988). Overnight cultures(18 h) of E2096 (E. coli DH5� pGP1-2) harbouring WzzSF

constructs were subcultured and grown with aeration at308C until they reached an OD600 of 1.0. The temperaturewas raised to 428C for 20 min to induce the production ofWzz, and cultures were then grown for an additional 60 minat 378C. Overnight cultures (18 h) of S. ¯exneri were subcul-tured 1:10 and grown with aeration for 3 h before cross-linking.Cells (E. coli or S. ¯exneri ) were harvested by centrifugationand washed once in ice-cold 10 mM K2HPO4/KH2PO4, pH 6.8,and then resuspended in the same buffer to an OD600 of 1.0.Formaldehyde (Univar 37% w/w) was added to a ®nal con-centration of 0.5%, and the samples were incubated by stand-ing for 1 h at 238C. Cross-linked samples of E. coli (1 ml) or S.¯exneri (1.5 ml) were harvested by centrifugation, washedonce in 1.5 ml of ice-cold 10 mM K2HPO4/KH2PO4, pH 6.8,and resuspended in 80 ml of sample buffer (Lugtenberg et al.,1975). Samples were either stored at ÿ208C or immediatelyfractionated. Aliquots (30±80 ml; 1±2 ´ 108 E. coli cells or 7±8 ´ 108 S. ¯exneri cells) were subjected to SDS±PAGE andWestern immunoblotting. Dithio-bis(succinimidylpropionate)(DSP) (Pierce) cross-linking was carried out as describedby Thanabalu et al. (1998). Brie¯y, cultures (50 ml) to becross-linked were washed in buffer (150 mM NaCl, 20 mMNaPO4 pH 7.2) and concentrated 10-fold in the same bufferfollowed by cross-linking with 0.2 mM DSP for 30 min at378C. DSP was quenched with 20 mM Tris, pH 7.5. Cellswere then harvested and either stored atÿ208C or immediatelyfractionated.

Preparation of Wzz speci®c antibodies

Large-scale protein preparations were produced by over-expression of WzzSF from pRMCD16 in E. coli MZ1 (Reddyet al., 1989). Brie¯y, cultures (250 ml) grown at 308C toOD600 <0.8 were induced to produce WzzSF by temperatureshift (428C/20 min) followed by a further incubation (378C/3 h). The cells were fractionated essentially as above, andthe whole membranes were solubilized in 1.67% (w/v) Sarko-syl (NL-97 Geigy), 10 mM Tris (pH 8) to separate the inner andouter membranes. WzzSF used to raise rabbit anti-WzzSF

antisera was puri®ed from the soluble component (cytoplasmicmembranes) by two cycles of SDS±PAGE. The rabbit wasimmunized (intramuscular at four sites) on day 1 with gel homo-genates in Freunds complete adjuvant. Subsequent boostswere performed on days 16, 41 and 70 using gel homo-genates in Freunds incomplete adjuvant. The rabbit wasexsanguinated by cardiac puncture under anaesthesia 20days after the last immunization and the serum was stored atÿ208C. Anti-WzzSF antibodies were immunoaf®nity puri®edfrom the anti-WzzSF sera using the method described by Sal-amitou et al. (1994), and cytoplasmic membrane extracts from



E. coli (MZ1� pRMCD16) that contained WzzSF. After elutionof the antibodies, an equal volume of sterile glycerol (100%)was added, and the antibodies were stored at ÿ208C.

Acknowledgements

This study was funded by the National Health and MedicalResearch Council of Australia. C.D. is in receipt of an Austra-lian Postgraduate Research 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.

Amor, P.A., and Whit®eld, C. (1997) Molecular and functionalanalysis of genes required for expression of group IB Kantigens in Escherichia coli : characterisation of the his-region containing gene clusters for multiple cell-surfacepolysaccharides. Mol Microbiol 26: 145±161.

Attridge, S.R., Daniels, D., Morona, J.K., and Morona, R.(1990) Surface co-expression of Vibrio cholerae and Sal-monella typhi O-antigen on Ty21a clone EX210. MicrobPathog 8: 177±188.

Bastin, D.A., Romana, L.K., and Reeves, P.R. (1993) Mole-cular cloning and expression in Escherichia coli K-12 ofthe rfb gene cluster determining the O-antigen of an E.coli O111 strain. Mol Microbiol 5: 2223±2231.

Batchelor, R.A., Alifano, P., Biffali, E., Hull, S.I., and Hull,R.A. (1992) Nucleotide sequences of the genes regulatingO-polysaccharide antigen chain length (rol ) from Escheri-chia coli and Salmonella typhimurium : protein homologyand functional complementation. J Bacteriol 174: 5228±5236.

Becker, A., and Puhler, A. (1998) Speci®c amino acid substi-tutions in the proline-rich motif of the Rhizobium melilotiExoP protein result in enhanced production of low-molecu-lar-weight succinoglycan at the expense of high-molecular-weight succinoglycan. J Bacteriol 180: 395±399.

Becker, A., Niehaus, K., and Puhler, A. (1995) Low molecularweight succinoglycan is predominantly produced by Rhizo-bium meliloti strains carrying a mutated ExoP proteincharacterized by a periplasmic N-terminal and a missingC-terminal domain. Mol Microbiol 16: 191±203.

Bugert, P., and Geider, K. (1995) Molecular analysis of theams operon required for exopolysaccharide synthesis ofErwinia amylovora. Mol Microbiol 15: 917±933.

Burrows, L.L., Chow, D., and Lam, J.S. (1997) Pseudomonasaeruginosa B-band O-antigen chain length is modulated byWzz (Rol). J Bacteriol 179: 1482±1489.

Daniels, C., Vindurampulle, C., and Morona, R. (1998) Over-expression and topology of the Shigella ¯exneri O-antigenpolymerase (Rfc/Wzy). Mol Microbiol 28: 1211±1222.

Formal, S.B., Dammin, G.J., LaBrec, E.H., and Schneider, H.(1958) Experimental Shigella infections: characteristics ofa fatal infection produced in guinea pigs. J Bacteriol 75:604±610.

Franco, A.V., Liu, D., and Reeves, P.R. (1998). The Wzz (Cld)

protein in Escherichia coli : amino acid sequence variationdetermines O-antigen chain length speci®city. J Bacteriol180: 2670±2675.

GonzaÂ lez, J.E., Semino, C.E., Wang, L.-X., Castellano-Torres, L.E., and Walker, G.C. (1998) Biosynthetic controlof molecular weight in the polymerization of the octasac-charide subunits of succinoglycan, a symbiotically impor-tant exopolysaccharide of Rhizobium meliloti. Proc NatlAcad Sci USA 95: 13477±13482.

Guidolin, A., Morona, J.K., Morona, R., Hansman, D., andPaton, J.C. (1994) Nucleotide sequence analysis of genesessential for capsular polysaccharide biosynthesis in Strep-tococcus pneumoniae type 19F. Infect Immun 62: 5384±5396.

Higgs, P.I., Myers, P.S., and Postle, K. (1998) Interactions inthe TonB-dependent energy transduction complex: ExbBand ExbD form homomultimers. J Bacteriol 180: 6031±6038.

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

Hong, M., and Payne, S.M. (1997) Effect of mutations in Shi-gella ¯exneri chromosomal and plasmid-encoded lipopoly-saccharide genes on invasion and serum resistance. MolMicrobiol 24: 779±781.

Horton, R.M. (1993) In vitro recombination and mutagenesisof DNA, SOEing together tailor-made genes. In Methodsin Molecular Biology. Vol. 15. PCR Protocols CurrentMethods and Applications. White, B.A. (ed.). Totowa, NJ:Humana Press, pp. 251±261.

Klee, S.R., Tzschaschel, B.D., Timmis, K.N., and Guzman,C.A. (1997) In¯uence of different rol gene products onthe chain length of Shigella dysenteriae Type 1 lipopolysac-charide O antigen expressed by Shigella ¯exneri carrierstrains. J Bacteriol 179: 2421±2425.

Liu, D., and Reeves, P.R. (1994) Escherichia coli K-12regains its O-antigen. Microbiology 140: 49±57.

Liu, D., Cole, R.A., and Reeves, P.R. (1996) An O-antigenprocessing function for Wzx (RfbX): a promising candidatefor O-antigen ¯ippase. J Bacteriol 178: 2102±2107.

Lugtenberg, B., Meijers, J., Peters, R., van der Hoek, P., andvan Alphen, L. (1975) Electrophoretic resolution of the majorouter membrane proteins of Escherichia coli K-12 into fourbands. FEBS Lett 58: 254±258.

Macpherson, D.F., Morona, R., Beger, D.W., Cheah, K.-C.,and Manning, P.A. (1991) Genetic analysis of the rfb regionof Shigella ¯exneri encoding the Y serotype O-antigen spe-ci®city. Mol Microbiol 5: 1491±1499.

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

Meier-Dieter, U., Barr, K., Starman, R., Hatch, L., and Rick,P.D. (1992) Nucleotide sequence of the Escherichia colirfe gene involved in the synthesis of enterobacterial com-mon antigen. J Biol Chem 267: 746±753.

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 ¯ex-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 ¯exneri.J Bacteriol 177: 1059±1068.

Mourrain, P., Lasa, I., Gautreau, A., Gouin, E., Pugsley, A.,and Cossart, P. (1997) ActA is a dimer. Proc Natl AcadSci USA 94: 10034±10039.

Paulsen, I.T., Beness, A.M., and Saier, Jr, M.H. (1997) Com-puter-based analyses of the protein constituents of trans-port systems catalysing export of complex carbohydratesin bacteria. Microbiology 143: 2685±2699.

Prossnitz, E., Nikaido, K., Ulbrich, S.J., and Ames, G.F.L.(1988) Formaldehyde and photoactivatable cross-linkingof the periplasmic binding protein to a membrane compo-nent of the histidine transport system of Salmonella typhi-murium. J Biol Chem 263: 17917±17920.

Reddy, P., Peterkofsky, A., and McKenney, K. (1989) Hyper-expression and puri®cation of Escherichia coli adenylatecyclase using a vector designed for expression of lethalgene products. Nucleic Acids Res 17: 10473±10488.

Reeves, P.R., Hobbs, M., Valvano, M.A., Skurnik, M., Whit-®eld, C., Coplin, D., et al. (1996) Bacterial polysaccharide

synthesis and gene nomenclature. Trends Microbiol 4:495±503.

Salamitou, S., Lemaire, M., Fujino, T., Ohayon, H., Gounon,P., Beguin, P., and Aubert, J.-P. (1994) Subcellular local-isation of Clostridium thermocellum ORF3p, a proteincarrying a receptor for the docking sequence borne bythe catalytic components of the cellulosome. J Bacteriol176: 2828±2834.

Stevenson, G., Kessler, A., and Reeves, P.R. (1995) A plas-mid-borne O-antigen chain length determinant and its rela-tionship to other chain length determinants. FEMS MicrobiolLett 125: 25±30.

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

Thanabalu, T., Koronakis, E., Hughes, C., and Koronakis, V.(1998) Substrate-induced assembly of a contiguous chan-nel for protein export from E. coli : reversible bridging of aninner-membrane translocase to an outer membrane exitpore. EMBO J 17: 6487±6496.

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



Analysis of Shigella flexneri Wzz (Rol) function by mutagenesis and cross-linking: Wzz is able to...

Documents

Transcript of Analysis of Shigella flexneri Wzz (Rol) function by mutagenesis and cross-linking: Wzz is able to...