Surface expression of O-specific lipopolysaccharide in Escherichia coli requires the function of the TolA protein



We investigated the involvement of Tol proteins in the surface expression of lipopolysaccharide (LPS). tolQ, -R, -A and -B mutants of Escherichia coli K-12, which do not form a complete LPS-containing O antigen, were transformed with the O7+ cosmid pJHCV32. The tolA and tolQ mutants showed reduced O7 LPS expression compared with the respective isogenic parent strains. No changes in O7 LPS expression were found in the other tol mutants. The O7-deficient phenotype in the tolQ and tolA mutants was complemented with a plasmid encoding the tolQRA operon, but not with a similar plasmid containing a frameshift mutation inactivating tolA. Therefore, the reduction in O7 LPS was attributed to the lack of a functional tolA gene, caused either by a direct mutation of this gene or by a polar effect on tolA gene expression exerted by the tolQ mutation. Reduced surface expression of O7 LPS was not caused by changes in lipid A-core structure or downregulation of the O7 LPS promoter. However, an abnormal accumulation of radiolabelled mannose was detected in the plasma membrane. As mannose is a sugar unique to the O7 subunit, this result suggested the presence of accumulated O7 LPS biosynthesis intermediates. Attempts to construct a tolA mutant in the E. coli O7 wild-type strain VW187 were unsuccessful, suggesting that this mutation is lethal. In contrast, a polar tolQ mutation affecting tolA expression in VW187 caused slow growth rate and serum sensitivity in addition to reduced O7 LPS production. VW187 tolQ cells showed an elongated morphology and became permeable to the membrane-impermeable dye propidium iodide. All these phenotypes were corrected upon complementation with cloned tol genes but were not restored by complementation with the tolQRA operon containing the frameshift mutation in tolA. Our results demonstrate that the TolA protein plays a critical role in the surface expression of O antigen subunits by an as yet uncharacterized involvement in the processing of O antigen.


Gram-negative bacteria are surrounded by a complex envelope structure consisting of cell wall peptidoglycan and an outer membrane (Nikaido and Vaara, 1985). The outer membrane prevents the rapid entry of lipophilic compounds, such as bile salts, detergents and fatty acids (Nikaido and Vaara, 1985; Baty et al., 1990; Webster, 1991), as well as larger molecules such as bacteriophages and bacteriocins. Vitamins, iron chelators and antibiotics, which also cannot readily cross the outer membrane (Nikaido and Vaara, 1985), bind to specific cell surface receptors and can be internalized via two distinctive import systems. One of these systems involves the TonB–ExbB–ExbD protein complex that interacts with certain outer membrane receptors to promote the energy-dependent translocation of receptor ligands across the outer membrane (Letain and Postle, 1997). The other is the Tol import system that comprises a multiprotein complex involved in the translocation of group A colicins and filamentous bacteriophages (Webster, 1991; Lazdunski et al., 1998). The Tol proteins are encoded by a cluster of seven genes, orf1–tolQRAB–pal–orf2, organized into two transcription units (orf1–tolQRA, and tolB–pal–orf2;Vianney et al., 1996; Lazdunski et al., 1998; Lazzaroni et al., 1999). Extensive studies have resulted in the identification and localization of Tol proteins. TolA, TolQ and TolR are associated with the plasma membrane (Guihard et al., 1994; Derouiche et al., 1995). TolA has a large periplasmic domain (Levengood et al., 1991; Guihard et al., 1994), and biochemical studies have shown specific interactions between the central part of TolA and outer membrane porins (Derouiche et al., 1996). Interactions with porin trimers in the presence of SDS have also been demonstrated for the TolB protein (Rigal et al., 1997). TolB is a periplasmic protein (Isnard et al., 1994; Vianney et al., 1996) that also interacts with Pal (peptidoglycan-associated lipoprotein), a protein anchored to the periplasmic side of the outer membrane (Vianney et al., 1996; Clavel et al., 1998). Orf2 is a periplasmic protein, whereas Orf1 appears to be a cytosolic protein (Levengood and Webster, 1989). The function of these two proteins has not been established.

Mutations in tol genes are commonly associated with pleiotropic defects in outer membrane permeability (Webster, 1991). The precise physiological role of the Tol system has not been established, but it has been proposed that the Tol proteins contribute to the organization and normal function of the outer membrane (Lazdunski et al., 1998). Recently, it has been shown that tolpal mutants of Escherichia coli release outer membrane vesicles (Bernadac et al., 1998). Furthermore, the TolAQR proteins can be isolated from a membrane fraction corresponding to putative contact sites between outer and cytoplasmic membranes (Guihard et al., 1994; Bouveret et al., 1995; Derouiche et al., 1995), which have been shown to contain newly synthesized outer membrane components (Ishidate et al., 1986). Therefore, the available evidence suggests that the Tol–Pal proteins contribute to the biogenesis of the outer membrane.

Lipopolysaccharide (LPS) is present in the outer membrane of most Gram-negative bacteria and consists of lipid A, core oligosaccharide and the O-specific polysaccharide chain (Whitfield and Valvano, 1993). The formation of LPS is a complex process involving the synthesis of activated precursors in the cytoplasm, followed by the independent assembly of the lipid A-core and O polysaccharide (Whitfield and Valvano, 1993). At least two different mechanisms have been described for the assembly of O antigens. One of them involves the synthesis of O-repeating subunits by the addition of subsequent monosaccharides at the non-reducing end of the molecule, a process that takes place on the cytosolic side of the cytoplasmic membrane (Whitfield, 1995). These subunits are translocated across the cytoplasmic membrane where they become polymerized. The undecaprenyl-linked polymer is then ligated en bloc to the lipid A-core by reactions occurring on the periplasmic face of the membrane (Mulford and Osborn, 1983; Marino et al., 1991; McGrath and Osborn, 1991a). This pathway, also referred to as the wzy (polymerase)-dependent pathway, occurs in the synthesis of the majority of O antigens, especially in those made of repeating units of different sugars (heteropolymeric O antigens; Keenleyside and Whitfield, 1999). The second mechanism involves the formation of a polymeric O antigen by reactions taking place on the cytosolic face of the cytoplasmic membrane, which are mediated by the sequential action of glycosyltransferases elongating the polysaccharide at the non-reducing end (Whitfield and Roberts, 1999). The nascent polysaccharide is transported across the cytoplasmic membrane by an ATP-binding cassette transporter (Bronner et al., 1994), and subsequently ligated to lipid A-core. This pathway has been observed especially in O antigens made of repeating units of the same sugar (homopolymeric O antigens), such as those from E. coli O8 and O9 (Whitfield, 1995), as well as in group 2 and 3 exopolysaccharide capsules (Whitfield and Roberts, 1999).

As LPS is one of the major structural components of the outer membrane, we hypothesized that Tol proteins could play a role in LPS biosynthesis and/or assembly. We explored this hypothesis using the E. coli O7 LPS gene cluster as a model system. The O7 polysaccharide results from the wzy-dependent polymerization of a pentasaccharide subunit made of N-acetylglucosamine, galactose, mannose, N-acetylviosamine and rhamnose (L'vov et al., 1984). The O7 biosynthesis gene cluster has been studied in detail in our laboratory (Valvano and Crosa, 1989; Marolda et al., 1990; 1999; Marolda and Valvano, 1988; 1993; 1995). We have shown that it encodes several enzymes involved in the synthesis of three nucleotide sugar precursors (dTDP-rhamnose, GDP-mannose and dTDP-N-acetylviosamine), four glycosyltransferases, an O antigen polymerase (Wzy) and a putative O antigen translocase (Wzx) (Marolda et al., 1999). The latter is presumed to be involved in the process of translocation of undecaprenyl-bound O antigen subunit across the cytoplasmic membrane (Liu et al., 1996; Feldman et al., 1999). In this study, we present evidence that the absence of TolA protein compromises the surface expression of polymeric O7 antigen. Furthermore, this alteration is associated with a reduced growth rate, serum sensitivity, dramatic changes in cell morphology and an abnormal accumulation of radiolabelled mannose, a sugar component unique to the O7 subunit, in cytoplasmic membrane fractions. We propose that TolA can modulate the surface expression of O antigen by an involvement in the processing of the O antigen subunits, possibly during the process of membrane translocation of O antigen or at the subsequent stages of LPS assembly on the periplasmic side of the plasma membrane.


The E. coli K-12 tolA gene has a role in the surface expression of O7-specific LPS

E. coli K-12 strains do not usually express O-specific LPS because of several mutations in the O antigen biosynthesis cluster (Liu and Reeves, 1994), but can support the expression of O-specific LPS if transformed with genes encoding the synthesis of O antigens from other sources (Whitfield and Valvano, 1993). To investigate the possible role of tol genes in the expression of O-specific LPS, we transformed tol mutants of E. coli K-12, as well as their corresponding isogenic parent strains, with the cosmid pJHCV32 encoding the O7-specific LPS biosynthesis genes (Valvano and Crosa, 1989; Marolda et al., 1990). LPS from the transformants was extracted and examined by Western blot. The tolQ and tolA strains produced less O7-specific LPS (Fig. 1A and B, lanes 2) than the respective tolQ+ and tolA+ parent isolates (Fig. 1A and B, lanes 1). By densitometric analysis, the reduction in polymeric O7 antigen in the tolQ and tolA mutants was estimated to be about 35% and 60%, respectively, relative to the amount of lipid A-core containing one O7 antigen subunit (lower bands in all lanes of Fig. 1). In contrast, the tolR and tolB mutants did not demonstrate any difference in the expression of O7 LPS relative to their isogenic counterparts (data not shown).

Figure 1.

Western blot analysis of O7 LPS in tolA and tolQ mutants and their respective isogenic parent strains. LPS was prepared from similar amounts of starting cultures (adjusted to an OD600 of 2.0) as described previously (Marolda et al., 1990). After SDS–PAGE, O7-specific polymers were detected by Western blotting using an anti-O7 polyclonal serum.

A. Lane 1, C600 (pJHCV32); lane 2, A592(pJHCV32); lane 3, A592(pJHCV32, pTPS202).

B. Lane 1, P90C(pJHCV32); lane 2, P90CtolQ(pJHCV32); lane 3. P90CtolQ(pJHCV32, pTPS202).

The O7 LPS defect in the tolA strains A592(pJHCV32) and P90CtolQ(pJHCV32) was corrected by transformation with pTPS202 (Fig. 1A and B, lanes 3), which carries orf1–tolQRA and part of tolB, as well as several other genes upstream of the tol cluster that are not related to the Tol system (Sun and Webster, 1986). As TolQ, TolR and TolA proteins form a complex in the cytoplasmic membrane (Derouiche et al., 1995), we preferred to use pTPS202 in the complementation experiment to ensure that all the components of the complex are present at similar ratios to the wild-type strains. However, this experiment did not indicate which of the tol gene(s) encoded by pTPS202 were involved in correcting O7 LPS surface expression. An involvement of tolB could be ruled out, as A593(pJHCV32) expressed O7 LPS at levels similar to the C600 isogenic parent (data not shown). The role of tolA was investigated using the plasmid pJG2, a derivative of pTPS202 with a frameshift mutation inactivating tolA(Table 1). The O7 LPS-deficient phenotype of strains A592(pJHCV32) and P90CtolQ(pJHCV32) was not corrected after transformation with pJG2 (data not shown). The tolQ allele in P90C is tolQ13, an amber mutation in codon 36 of tolQ that has been shown to reduce transcription of tolA by ≈ 50% (Vianney et al., 1996). Thus, the phenotype of the tolQ mutation regarding O7 LPS surface expression can be explained by a polar effect on tolA gene expression. Furthermore, the tolR::Cm mutation in TPS300, which did not affect O7 LPS expression, has been reported previously as non-polar on tolA, given the fact that a plasmid containing this mutation and an intact tolA gene can complement the tolA mutation in A592 (Sun and Webster, 1987). Therefore, we concluded that reduced O7 LPS surface expression in the mutant strains is primarily associated with a defect in the production of TolA.

Table 1. Properties of strains and plasmids used in this study.
Strain/plasmidRelevant propertiesSource or reference
 A592C600 tolA1 Nagel de Zwaig and Luria (1967)
 A593C600 tolB2 Nagel de Zwaig and Luria (1967)
 BW19851RP4-2::Mu-1kan::Tn7/creB510 hsdR17 endA1 zbf-5 Metcalf et al. (1994)
uidAMluI)::pir(wt) recA1 thi 
 C600 thr leu tonA lacY supE thi Laboratory stocks
 CLM11SØ874 manA::pMAV020, AmpRThis study
 GMS343 aroD6 argE3 lacY1 galK2 man-4 mtl rpsL700 tsx29? supE44?λ Novel and Novel (1973)
 JAG4CLM11, tolA::Ω, Sm/SpcR, AmpRThis study
 JAT1VW187, tolQ::Ω, Sm/SpcRThis study
 JAT2VW187, tolA::Ω, Sm/SpcR, O7This study
 P90C araΔ(lac-pro) thi Bradley and Whelan (1989)
 P90CtolQ araΔ(lac-pro) thi tolQ13 Bradley and Whelan (1989)
 SØ874 lacZ trpΔ(sbcB-rfb) upp rel rpsL Neuhard and Thomassen (1976)
 SY327-λpirλpir araDΔ(lac-pro) argE recA56 RifRnalA Miller and Mekalanos (1988)
 TPS300P90C/F′lac pro orf3::Cm(tolR) Sun and Webster (1987)
 VW187 E. coli O7:K1, prototrophic Valvano and Crosa (1984)
 pCM1115.2 kb EcoRI fragment containing the upstream region Marolda and Valvano (1995)
of the wb(rfb)EcO7 gene cluster 
 pCM1311.2 kb EcoRI–HindIII fragment containing the promoter region Marolda and Valvano (1988)
of the O7 LPS gene cluster cloned into pFZY1, ApR 
 pGP704Cloning vector, oriR6K ApRmobRP4J. Mekalanos
 pGS63 fumA and manA genes cloned in pBR322, TcR Guest and Roberts (1983)
 pJG1pTPS202 with a KmR gene cassette inserted in the ScaI site, KmR, ApSThis study
 pJG2pTPS202 with a deletion of the single NotI site within tolAThis study
 pJG11 tolA::Ω and nptI-SacBR into pGP704, AmpR, KmR, Sm/SpcRThis study
 pJHCV32 wb(rfb) EcO7 cosmid, TcR, O7+ Valvano and Crosa (1989)
 pJT10 tolQ::Ω and nptI–SacBR into pGP704, AmpR, KmR, Sm/SpcRThis study
 pMAV0200.85 kb of the manA gene cloned in the EcoRV site of pGP704This study
 pMAV11 wecA gene cloned in pACYC184 Alexander and Valvano (1994)
 pTPS2028.7 kb BglII fragment carrying orf1–tolQRAB genes cloned in pJIH12, ApR Sun and Webster (1986)
 pUM24Cm ntpI–sacBR cassette in pPR329, CmR, KmR Ried and Collmer (1987)

TolA and tolQ mutations do not affect transcription of the O7 LPS gene cluster

We investigated whether the low expression of O7 LPS in tolQ and tolA mutants could result from downregulation of the transcription of O7 biosynthetic genes. In a previous study, we have characterized the promoter region of the O7 LPS cluster using transcriptional fusions to a promoterless lac operon (Marolda and Valvano, 1988). One of these fusion plasmids containing the wbEcO7 promoter pCM131 was transformed into the tolA mutant A592 carrying pJG1, and the production of β-galactosidase was assessed as described in Experimental procedures. As both pCM131 and pTPS202 have a β-lactamase gene, we constructed pJG1 by inserting a KmR gene in pTPS202, thus facilitating antibiotic selection for the maintenance of both plasmids. Strains A592(pCM131 and pJG1) and A592(pCM131) produced 140 ± 12 and 137 ± 10 units of β-galactosidase respectively. Similar results were obtained with the P90CtolQ strain (data not shown). We also conducted reverse transcriptase–polymerase chain reaction (RT–PCR) experiments using primers that anneal to rmlC, located at ≈ 3.5 kb downstream of the O7 promoter (Marolda and Valvano, 1995). For this experiment, the tolA mutant and its parent strain were transformed with pCM111, which contains the O7 promoter and the first five genes of the O7 LPS cluster (Marolda and Valvano, 1995). As in the case of β-galactosidase production, no striking differences were found in the amount of PCR amplification products from mRNA templates (data not shown). All these results suggest that the effect of the tol mutations on O7 LPS expression cannot be explained by a reduction in transcription across the O7 LPS cluster.

The tolA mutation is associated with a reduced amount of core LPS

An intact lipid A-core is required for the surface expression of O-specific antigens (Schnaitman and Klena, 1993; Whitfield and Valvano, 1993). Therefore, a defective lipid A-core oligosaccharide could explain the O7 LPS-deficient surface expression found in our tol mutants. The migration properties of lipid A-core oligosaccharide bands in tolQ and tolA mutants were identical to those of their isogenic counterparts (Fig. 2), ruling out the presence of gross structural defects of the lipid A-core in these mutants. The three bands observed result from the heterogeneity of the E. coli K-12 core oligosaccharide (Austin et al., 1990). However, in the case of the tolA mutant (Fig. 2, lane 2), the upper core band was barely visible, and the amount of the other two bands was considerably reduced compared with those of strain C600 (Fig. 2, lane 1). This difference could not be attributed to a loading artifact, as the LPS samples were prepared from standardized amounts of bacterial cells, and the results were reproducible in repeated experiments. To quantify the LPS more precisely, we determined the concentration of keto-deoxyoctulosonic acid (KDO), a unique component of lipid A-core, in the samples (Raetz, 1996). KDO values were normalized by the protein concentration in the outer membrane to account for potential differences in the outer membrane content between tol mutants and isogenic counterparts. LPS samples prepared from the P90CtolQ and P90C showed similar specific amounts of KDO (Table 2). In contrast, the tolA mutant had a 23% reduction in the amount of KDO with respect to the control strain (Table 2). Although statistically significant, this reduction in the lipid A-core is not sufficient to explain the dramatic decrease in O7 polysaccharide expression in the tolA mutant.

Figure 2.

Analysis of LPS in tolA and tolQ mutants of E. coli K-12 and their isogenic parent strains. LPS preparations were analysed by tricine-SDS–PAGE followed by silver staining. Lanes: 1, C600; 2, A592; 3, P90C; 4, P90CtolQ. The arrow in lane 2 indicates a band barely visible in the photograph.

Table 2. KDO concentration in LPS preparations from tol mutants and their isogenic derivativesa.


KDO (nmol)
(mg ml−1)


  1. a . Data are the means of three independent determinations. Comparisons were analysed using the z-test ( Armitage and Berry, 1994).

P90C7.81 ± 0.125.49 ± 0.281.421 ± 0.15 P = 0.379
P90CtolQ8.56 ± 0.145.77 ± 0.241.485 ± 0.14 
C6009.05 ± 0.355.93 ± 0.131.526 ± 0.14 P = 0.013
A592 (tolA)6.65 ± 0.215.60 ± 0.181.187 ± 0.12 

The tolA mutant expressing O7 LPS accumulates radiolabelled mannose in the plasma membrane

As O7 polysaccharides could be detected in both tolQ and tolA mutants (Fig. 1A and B, lanes 2), the effect of these mutations is probably not in the biosynthesis of the O7 subunit, but rather in subsequent processing steps involving the assembly of O7 LPS. Therefore, we investigated the biosynthesis of the O7 subunit by radiolabelling with [2,6-3H]-mannose, as this sugar only occurs in the O7 subunit (L'vov et al., 1984). To prevent the catabolism of exogenously added mannose, we used a mutant defective in the manA gene. Preliminary experiments with the manA strain GMS343 showed a relatively high background incorporation of radiolabelled mannose even in the absence of the O7 LPS genes (data not shown). This was probably a result of the formation of radioactive GDP-fucose, a by-product of GDP-mannose metabolism that is a precursor for the synthesis of colanic acid capsular polysaccharide in E. coli K-12 (Whitfield and Valvano, 1993). Thus, to ensure that [2,6-3H]-mannose is exclusively incorporated in the O7 subunit, we constructed a manA mutation in the colanic acid-deficient strain SØ874, which has a deletion eliminating the colanic acid biosynthesis genes, including the genes involved in the synthesis of GDP-fucose. The resulting derivative, CLM11, did not grow in minimal medium with mannose as the only carbon source, and the incorporation of radioactive mannose in the cell envelope fractions was negligible (Table 3, Fig. 3A). Also, the expression of O7 LPS in CLM11(pJHCV32) was mannose dependent, confirming that this strain contains a manA mutation (data not shown). We used the mutagenic plasmid pJG11 (Table 1) to construct the isogenic strain JAG4 with a mutation in the tolA gene. CLM11 and JAG4 were transformed with pJHCV32, and labelling experiments were carried out as described in Experimental procedures. As expected, more than 99% of the total incorporated label was found in the outer membrane fraction of CLM11(pJHCV32). (Table 3, Fig. 3B), suggesting that the labelled material corresponds to O7 LPS that has been properly assembled and translocated to the outer membrane. In the case of JAG4(pJHCV32), 2.3% of the total incorporated label was found in the plasma membrane fraction, compared with 0.7% in the plasma membrane fraction of CLM11(pJHCV32) (Table 3). Although small, these differences were statistically significant (P > 0.01), suggesting a low level of accumulation of mannose-radiolabelled material in the cytoplasmic membrane. As we have observed previously that the expression of O7 polysaccharide in E. coli K-12 is poor (Marolda and Valvano, 1988; 1995), JAG4 was transformed with pMAV11 to increase the amount of WecA transferase as a way of increasing the production of O7 LPS. WecA (formerly Rfe) catalyses the transfer of N-acetylglucosamine to undecaprenol phosphate, and this is the initiating enzyme step for the synthesis of the O7 repeating subunit (Alexander and Valvano, 1994). In the presence of pMAV11 and pJHCV32, ≈ 22% of the total incorporated label accumulated in the cytoplasmic membrane fraction of JAG4 (Table 3,Fig. 3D). This accumulation was not observed in the control experiment with CLM11 (pJHCV32,pMAV11) or with CLM11(pMAV11), indicating that the overexpression of WecA per se does not cause accumu-lation of mannose-radiolabelled material in the cytoplasmic membrane (Table 3; data not shown). Thus, a defect in the TolA protein is associated with an accumulation of mannose-radiolabelled material in the cytoplasmic membrane. As the membrane-associated material appeared exclusively in the presence of pJHCV32, we conclude that it represents accumulated O7 LPS biosynthesis intermediates.

Table 3. Accumulation of O7 LPS precursors in membrane fractions.
StrainPlasmidOuter membrane
(c.p.m. mg−1 protein)a
Plasma membrane
(c.p.m. mg−1 protein)
  1. a . The values of c.p.m. mg −1 protein represent the average of three independent experiments.

  2. b . Percentage accumulation (mean ± standard error) of [2,6- 3H]-mannose in the plasma membrane fractions, as calculated by the formula:

  3. 100 × [c.p.m. (plasma membrane)–c.p.m. (background)]/{[c.p.m. (outer membrane)–c.p.m. (background)] + [c.p.m. (plasma membrane)–c.p.m. (background)]}.

CLM11pJHCV322517112 (0.7 ± 0.1)b
CLM11pJHCV32, pMAV112434118 (0.8 ± 0.05)
JAG4pJHCV32166271 (2.3 ± 0.11)
JAG4pJHCV32, pMAV111783577 (22 ± 1.2)
Figure 3.

Sucrose density gradient centrifugation of membranes from E. coli K-12 strains CLM11 (manA) and JAG4 (manA tolA::Ω) incubated in the presence of radiolabelled mannose. French press lysates were generated and resolved on 30–60% sucrose step gradients, with fractions harvested and examined as described in Experimental procedures. Closed and hatched bars indicate the fractions containing cytoplasmic (IM) and outer (OM) membranes, respectively, as determined by the distribution of specific markers throughout the gradient (see Experimental procedures).

A. CLM11.

B. CLM11(pJHCV32).

C. JAG4(pJHCV32).

D. JAG4(pJHCV32, pMAV11).

A polar tolQ mutation in the E. coli O7 strain VW187 is associated with reduced growth rate, serum sensitivity and elongated cell morphology

We have shown previously that the O7 LPS expression mediated by pJHCV32 in E. coli K-12 strains is significantly lower than in wild-type O7 isolates (Marolda et al., 1990; Marolda and Valvano, 1995). This is caused by the lack of a regulatory region necessary for efficient elongation of the mRNA that is absent in the cosmid pJHCV32 (Marolda and Valvano, 1988). Therefore, we attempted the construction of tolQ and tolA mutants in the wild-type O7 strain VW187 to determine more precisely the effect of these genes under conditions of full O7 LPS expression. For this purpose, we first generated the mutagenic plasmids pJT10 and pJG11, which contain the respective tolQ and tolA genes inactivated by the insertion of the Ω interposon (Table 1;Experimental procedures).

Plasmid pJG11 was used for the construction of a tolA mutation in VW187. The resulting mutant, strain JAT2, initially showed a phenotype similar to that of the E. coli K-12 A592(pJHCV32) displaying a very low amount of O7 LPS barely visible in polyacrylamide gels. However, JAT2 grew extremely slowly and, after a few passages, it completely lost the ability to make O7 LPS. These O7-deficient colonies also grew much faster than the original mutants. Three independent attempts at mutagenesis yielded similar results. All these strains had a tolA mutation, as judged by their sensitivity to deoxycholate, resistance to cloacin DF13 and the results of Southern blot hybridization. Therefore, we concluded that a mutation in the tolA gene is detrimental to cells expressing O7 antigen. The O7 spontaneous derivatives of JAT2 displayed lipid A-core oligosaccharide bands with the same migration in the gel as the lipid A-core of the wild-type VW187 strain (data not shown). This suggests that the lipid A-core of the mutants was structurally intact and that the mutations preventing the formation of O7 LPS were located in genes involved in the synthesis of the O7 antigen. These mutant strains were not studied any further.

Introduction of the suicide plasmid pJT10 in strain VW187 resulted in the isolation of a tolQ mutant strain designated JAT1, which was sensitive to deoxycholate and resistant to cloacin DF13. The insertion of the Ω interposon in the tolQ gene was verified by Southern blot hybridization (data not shown), and the lack of TolA protein was demonstrated by Western blot (Fig. 4, lane 2). Transformation of JAT1 with pTPS202, but not with pJG2, restored the expression of TolA (Fig. 4, lanes 3 and 4). JAT1 produced ≈ 40% less O7 LPS compared with VW187 (Fig. 5, lane 2), and transformation with pJG1, but not with pJG2, restored the wild-type phenotype of O7 LPS expression (Fig. 5, lanes 3 and 4). From these results, we concluded that the tolQ mutation in VW187 was associated with reduced O7 LPS expression, and that this phenotype results from a polar effect on the tolA gene, as was found with E. coli K-12.

Figure 4.

Western blot analysis of TolA protein expression. Cell lysates were processed as described in Experimental procedures. Lanes: 1, VW187; 2, JAT1; 3, JAT1(pTPS202); 4, JAT1 (pJG2).

Figure 5.

LPS analysis of E. coli VW187 and the tolQ mutant derivative JAT1. LPS was isolated and analysed by SDS–PAGE followed by silver staining. Lanes: 1, VW187; 2, JAT1; 3, JAT1(pTPS202); 4, JAT1(pJG2).

O-specific LPS confers a selective advantage to bacteria growing in body fluids by increasing their resistance to serum complement-mediated lysis (Joiner, 1988). Therefore, we determined whether reduced O7 LPS expression in JAT1 had any biological significance in protecting bacterial cells from the lytic activity of serum complement. Figure 6 shows that JAT1 was highly sensitive to fresh serum in contrast to the wild-type VW187. The serum-sensitive phenotype was corrected by complementation with pTPS202. Killing was complement mediated, as JAT1 survived treatment with heat-inactivated serum that was complement depleted.

Figure 6.

Percentage survival of E. coli O7 strains VW187 and JAT1 (VW187tolQ) using normal and heat-treated (H) fresh dog serum.

We also observed that JAT1 grew considerably slower than the wild-type strain VW187 in either heat-inactivated serum (Fig. 6) or regular culture medium (data not shown). In Luria broth, JAT1 displayed a doubling time of ≈ 119 min compared with 22 min for VW187. The slow growth rate of JAT1 was corrected after transformation with pTPS202, whereas pJG2 failed to complement, ruling out a secondary mutation(s) causing a growth defect and confirming that it resulted from a defect in tolA gene expression. In contrast, the E. coli K-12 counterparts P90CtolQ and P90C grew at comparable rates with or without the O7 LPS cosmid pJHCV32. As E. coli K-12 strains only express a small amount of O7 LPS in the presence of pJHCV32 (Marolda et al., 1990; Marolda and Valvano, 1995), we surmised that the growth defect in JAT1 was associated with the formation of wild-type amounts of O7 polysaccharide.

JAT1 cells were examined by phase-contrast light microscopy and found to display an elongated morphology, in some cases reaching up to sixfold or more the length of normal cells. Elongated cells were also non-motile, in contrast to the wild-type VW187. Bacterial cells were examined for viability using a dual fluorescent staining consisting of propidium iodide and the Syto9. This technique is based on the ability of the bacterial membrane to exclude the red fluorescent dye propidium iodide, whereas the green fluorescent Syto9 is membrane permeable. The majority of JAT1 cells were elongated and red fluorescent (Fig. 7B), showing round condensations heavily stained with propidium iodide corresponding to nucleoids. In contrast, the wild-type VW187 showed small green fluorescent rods, indicating that the propidium iodide was excluded. We also found non-fluorescent elongated cells that looked flattened, and they were only visible by phase-contrast microscopy (data not shown). These cells are probably empty sacculi that fail to retain the fluorescent dyes. Flattened and elongated sacculi were also observed by electron microscopy after negative staining of whole cells (Fig. 8A). The morphological alterations in JAT1 cells were corrected by transformation with pTPS202 (Fig. 7D), but not with pJG2 (Fig. 7E), and they were present in cells with or without spectinomycin in the growth medium (Fig. 7C). Therefore, we concluded that the elongated cells were a consequence of the TolA defect in these cells and not of an increased permeability to spectinomycin. As another control, we also examined E. coli K-12 P90CtolQ cells microscopically. No elongated cells were found, and > 90% of the cells appeared as green rods (data not shown), suggesting that entry of propidium iodide is not due to the tolQ mutation in these cells and is probably associated with O7 LPS expression.

Figure 7.

Fluorescence microscopy using the LIVE/DEAD kit. Photographs of VW187 (A), JAT1 grown with 80 µg ml−1 spectinomycin (B), JAT1 grown in the absence of spectinomycin (C), JAT1(pTPS202) (D) and JAT1(pJG2) (E) showing cell morphology and staining properties. Bacteria in (B), (C) and (E) show an elongated morphology and display permeability to the red fluorescent dye propidium iodide. Magnification 1000 ×.

Figure 8.

Transmission electron microscopy of E. coli O7 VW187 and JAT1 strains. Cells were negatively stained as described in Experimental procedures by transmission electron microscopy at the lowest magnification. Horizontal bar, 2 µm.

A. JAT1. Arrows indicate empty sacculi that do not stain well and show small vesicles, probably corresponding to membrane vesicles resulting from disruption of the cell envelope.

B. VW187.


Our study shows that the absence of the TolA protein in E. coli K-12 causes a detectable reduction in the surface expression of O7-specific LPS. Several lines of evidence demonstrated that this phenotype is associated with the lack of a functional TolA protein: (i) a similar phenotype was found in the tolQ13 mutant, which is polar on tolA gene expression; (ii) the phenotype was not present in a tolR::cat mutant non-polar on tolA; (iii) a normal O7 LPS surface expression occurred after complementation with pTPS202, but not with pJG2, which carries a frameshift mutation inactivating only the tolA gene. The O7 LPS-deficient phenotype could not be explained by reduced transcription of the O7 biosynthetic genes, although post-transcriptional effects were not investigated. The tolA mutant also expressed less lipid A-core oligosaccharide, but the magnitude of this reduction was not sufficient to account for the dramatic decrease in polymeric O7 antigen. The reduction in LPS content of the outer membrane could be explained in part by the formation and release of small outer membrane vesicles, as reported by Bernadac et al. (1998) in tol–pal mutants.

The synthesis of O antigens occurs independently from that of lipid A-core oligosaccharide. In the case of O7, as well as with many other wzy-dependent O antigens, the O subunit is assembled on the cytosolic face of the cytoplasmic membrane and subsequently translocated to the periplasmic side of the membrane where it is further polymerized and ligated to the lipid A-core (Feldman et al., 1999). The phenotype of the tol mutants could not be explained by defects in the biosynthesis of O7 antigen subunits, as the mutant strains formed a complete O7 antigen, as judged by the electrophoretic banding profile of LPS samples showing reactivity with the O7 antiserum. We also detected an abnormal accumulation in the cytoplasmic membrane of mannose-radiolabelled material. This material presumably corresponds to O7 LPS biosynthesis precursors, suggesting a block in either the translocation of O7 antigen units across the plasma membrane or the transfer of O7 LPS intermediates to the outer membrane, both of which occur very rapidly in normal cells (Marino et al., 1991; McGrath and Osborn, 1991b). Therefore, we propose that TolA may be involved in the process of assembly of O antigen, possibly by facilitating the transfer of O polysaccharide to lipid A-core molecules or by contributing to the translocation of the O antigen subunits across the cytoplasmic membrane. In either case, it would be conceivable to expect an accumulation of bactoprenyl-bound O7 antigen intermediates, which in turn would compromise the recycling of bactoprenyl phosphate molecules for other functions such as the synthesis of cell wall peptidoglycan. This interpretation is supported by the fact that E. coli VW187 cells with a tolA mutation were non-viable and rapidly developed variants with uncharacterized mutations in O7 antigen synthesis. Furthermore, a tolQ mutation in VW187 with a strong polar effect on tolA expression displayed dramatic morphological changes and reduced growth when concomitantly expressing O7 biosynthesis genes. The morphological changes were phenotypically similar to those caused by mutations affecting the synthesis of cell wall components and normal septation and are consistent with a reduced formation of cell wall peptidoglycan. Alternatively, abnormal accumulation of O7 polymeric LPS in the periplasmic space may cause a deleterious effect in the bacterial cell, perhaps by increasing the osmolarity within this cell compartment. Recently, Meury and Devilliers (1999) have reported that tolA mutants show modified morphology and produce DNA-free cells, especially under conditions of both high and low osmolarity. They have also described how tolA mutants were unable to locate the division site properly and septation was impaired (Meury and Devilliers, 1999). Filamentation has also been observed in tol mutants of Vibrio cholerae (Heilpern and Waldor, 2000). In order to determine the exact nature of the tolA block, biochemical studies are under way to characterize the membrane-accumulated mannose-radiolabelled material in terms of its size, composition and topology in the plasma membrane, as well as to measure the rate of peptidoglycan synthesis.

The Tol import system was initially discovered in relation to the translocation of group A bacteriocins and DNA from filamentous bacteriophages across the outer membrane (Webster, 1991). Mutations in various tol genes are not only associated with a defect in the import of these molecules but also with various pleiotropic defects. These involve a reduction in the content of outer membrane proteins, increased susceptibility to detergents and hydrophobic compounds and leakage of periplasmic proteins (Webster, 1991). All these phenotypes suggest that the Tol import system plays an important role in maintaining the integrity of the outer membrane, but its exact function remains to be elucidated. Derouiche et al. (1996) showed that the central domain of TolA forms high-molecular-weight complexes specifically with trimeric OmpF, OmpC, PhoE and LamB outer membrane proteins. These authors proposed that TolA might function in the assembly of those outer membrane proteins that require de novo LPS synthesis, which is also consistent with recent observations that tol–pal mutants release outer membrane vesicles (Bernadac et al., 1998). TolA, TolR and TolQ proteins interact with each other via their membrane-spanning segments to form a complex (Guihard et al., 1994; Derouiche et al., 1995), and they fractionate preferentially at a density corresponding to that of the proposed contact sites between outer and plasma membranes (Guihard et al., 1994). Interestingly, newly synthesized LPS molecules and capsular polysaccharides are also associated with these contact sites (Bayer, 1979; 1991; Ishidate et al., 1986; Koncke et al., 1990). Our discovery that TolA is involved with the surface expression of O polysaccharide and, to a lesser extent, with synthesis of the LPS core suggests a model in which the TolQRA complex may act as a functional site for the export of newly synthesized outer membrane protein and LPS precursors from the plasma to the outer membrane.

Despite the fact that TolQ, -R and -A proteins interact physically forming a complex (Guihard et al., 1994; Derouiche et al., 1995; Lazzaroni et al., 1995), we have shown in this study that TolQ and TolR appear not to be directly involved in the O7 LPS surface expression defect. It is possible that the genes exbB and exbD, which are functionally homologous to tolQ and tolR, respectively (Braun, 1989), are compensating the O7 LPS defect in the tolQ and tolR mutants. Alternatively, it is also possible that the effect on O7 LPS in E. coli K-12 tolQ and tolR mutants is not noticeable because of the poor expression of the recombinant O7 antigen and the fact that small amounts of TolA protein may be sufficient to ensure normal processing of O7 LPS in these cells. The latter explanation is consistent with our findings that the O7 LPS-deficient phenotype is less apparent in the tolQ13 mutation, and also with the finding that the tolA mutant in VW187 is not viable in the presence of O7 LPS synthesis, whereas a polar tolQ mutation remains viable. The lack of detection of TolA protein in the VW187 tolQ::Ω mutant by Western blotting does not preclude the synthesis of small undetectable amounts of the protein. It could also be argued that the absence of TolA alone is more deleterious for the cell than the probable absence of TolQRA in the tolQ polar mutant. As these three proteins form a complex in the plasma membrane, the absence of any one component may be more deleterious than the lack of all of them. This may especially be the case for TolA, whose function cannot be complemented by any other gene. Construction and analysis of double mutants in strain VW187 with defects in both exbB/tolQ and exbD/tolR as well as non-polar mutations in each of the tol genes will be required to examine these possibilities.

When discovered, the tol genes were not considered to be essential, as mutations in these genes could be obtained relatively easily in E. coli K-12, and the growth of the mutant strains was not compromised, at least under standard laboratory conditions. Strains of E. coli K-12 do not produce O-specific polysaccharide because of various mutations in the biosynthesis genes for the O antigen (Liu and Reeves, 1994). We have shown that a tolQ mutant derived from the wild-type E. coli O7 strain VW187 is sensitive to fresh serum despite forming a substantial amount of complete O7 polysaccharides. This observation argues that the tol system is important in vivo and may contribute to the virulence of the microorganism. More importantly, we also show in this study that our efforts to obtain a tolA mutant in VW187 were unsuccessful. Mutant cells grew very poorly and rapidly yielded derivatives with secondary mutations preventing O7 antigen synthesis that suppressed the growth defect. The essential nature of tolA and tolQ has been reported in Pseudomonas aeruginosa PAO (Dennis et al., 1996). These investigators also tried unsuccessfully to obtain stable mutations in both these genes. As in E. coli VW187, P. aeruginosa PAO also produces O-polysaccharide, we suggest that the lethality of tol mutations in P. aeruginosa may be caused by a defect in the processing of the O units. Therefore, the lethal phenotype of tol mutations in wild-type bacteria strongly supports the idea that Tol proteins may play an essential role in the biogenesis of the outer membrane. In conclusion, our data provide evidence indicating that, in addition to their proposed role in the assembly of outer membrane proteins, the Tol proteins, and especially TolA, also function in the processing of O antigen-specific LPS.

Experimental procedures

Strains and plasmids

Strains and plasmids used in this study are described in Table 1. Strain A592 carries a 4 bp deletion causing a frameshift mutation in the tolA gene (Levengood and Webster, 1989). Strain SØ874 has a large deletion eliminating the genes for colanic acid biosynthesis (Neuhard and Thomassen, 1976; Marolda and Valvano, 1995). Strains BW19851 and SY327λpir were used for the maintenance and delivery of suicide plasmids. Bacteria were cultured in Luria agar supplemented with ampicillin (100 µg ml−1), chloramphenicol (30 µg ml−1), spectinomycin (80 µg ml−1), tetracycline (20 µg ml−1), kanamycin (20 µg ml−1) and 5% (w/v) sucrose as appropriate. For some experiments, microorganisms were also cultured in M9 minimal medium with 0.5% (w/v) d-glucose or 0.5% (w/v) d-mannose. All tol mutants were sensitive to 0.2% (w/v) sodium deoxycholate (Levengood-Freyermuth et al., 1993). Sensitivity to cloacin DF13 and colicin N were used as indicators of TolA function, and the assays were conducted as described previously (Thomas and Valvano, 1992; 1993).

Recombinant DNA methods

Purification of plasmid DNA and Southern blot hybridizations were carried out as described elsewhere (Valvano and Crosa, 1984). The non-radioactive digoxygenin (DIG) DNA labelling and detection kit and enzymes for molecular biology were purchased from Roche Diagnostics. PCR amplification of the manA gene was carried out with primers 5′-ACTGAGACTAGTACGACTT-3′ (sense) and 5′-AAGCTTAGCAAGAGATG-3′ (antisense).

Construction of plasmids and mutant strains

pJG1 is a derivative of pTPS202 with a frameshift mutation in the coding region of the tolA gene. pTPS202 was digested with NotI, and the ends were filled with DNA polymerase I. Inactivation of tolA in pJG2 was confirmed by the inability of this plasmid to complement the tolA mutation in strain A592, while complementing the tolB mutation in strain A593, as assessed by colicin N and cloacin DF13 sensitivity assays.

tolA and tolQ mutants in strains VW187 and CLM11 were generated by homologous recombination using plasmids pJT10 and pJG11 (Table 1). These plasmids were based on the suicide vector pGP704 (Miller and Mekalanos, 1988). pJT10 was constructed by cloning the 2.6 kb HincII fragment of pTPS202, containing the tolQRA genes (Sun and Webster, 1986), into the EcoRV site of pGP704. Next, the Ω interposon (Prentki and Krisch, 1984) was inserted into a unique AvaI site of the tolQ gene, and the ntpI–sacBR cassette from pUM24Cm (Brooks et al., 1990) was inserted in the BglII site of the vector DNA. pJG11 was constructed in a similar manner, but using the 1.7 kb NruI fragment of pTPS202 containing the tolA gene and part of tolR, followed by insertion of the Ω fragment into the unique NotI site of tolA and the subsequent cloning of the ntpI–sacBR cassette as described for pJT10. Plasmids were constructed using E. coli SY327λpir as a background strain and then transformed into strain BW19851 as a conjugation donor. The suicide plasmids were integrated into the chromosome by homologous recombination between DNA inserts and chromosomal sequences. As sucrose is lethal to cells carrying the sacB gene (Ried and Collmer, 1987), exconjugants were selected for sucrose and spectinomycin resistance to isolate only those microorganisms in which second cross-over events resulted in the removal of the integrated plasmid and the wild-type copy of the gene. These cells were then screened for sensitivity to ampicillin and kanamycin to confirm the disappearance of vector sequences from the site of recombination. ApS KmS mutants were examined further by sensitivity to deoxycholate, slide agglutination with O7 antiserum, LPS pattern profiles, Southern blot hybridization and Western blotting with anti-TolA antibodies.

For the construction of a manA mutant in strain SØ874, a 1.3 kb fragment containing the manA gene was amplified by PCR from plasmid pGS63 (Guest and Roberts, 1983) and digested with HincII. The 0.85 kb HincII fragment internal to the manA gene was cloned directly into the EcoRV site of pGP704, resulting in pMAV020 (Table 1). This plasmid was transformed into BW19851 and conjugated to SØ874. AmpR exconjugants were examined for growth in M9 plates with mannose. One of these exconjugants that was unable to use mannose was designated CLM11 (Table 1).

LPS analysis

LPS was isolated and analysed by Tris-glycine-SDS–PAGE as described previously (Marolda et al., 1990). Briefly, this involved suspending bacterial cells in a lysis buffer containing proteinase K, followed by a hot phenol extraction. Commercially cast 16% polyacrylamide gels were purchased from Novex, and LPS bands were visualized by silver staining (Marolda et al., 1990). The analysis of lipid A-core was carried out by tricine-SDS–PAGE (Schagger and von Jagow, 1987). Western blot analysis was performed with an O7-specific rabbit polyclonal antiserum (Valvano and Marolda, 1991). Outer membranes were prepared as described by Kappos et al. (1992). The amount of KDO in membrane samples was determined (Osborn, 1963) and expressed in nmol mg−1 protein.

Detection of the TolA protein

Detection of TolA was carried out by SDS–PAGE of cell lysates, followed by transfer to nitrocellulose membranes using standard procedures. Membranes were incubated with TolA–rabbit polyclonal antiserum (kindly provided by R. Lloubès) and then with horseradish peroxidase-linked sheep anti-rabbit immunoglobulin G (Amersham Pharmacia Biotechnology). Detection by chemiluminescence assay was performed using the BM chemiluminescence blotting substrate (Roche Diagnostics) as recommended by the manufacturer.

β-Galactosidase assay

Strains A592 and P90CtolQ containing pJG1 or pJG2 (Table 1) were transformed with pCM131, which carries the O7 promoter region fused to a promoterless lac operon (Marolda and Valvano, 1988). Cells were incubated in Luria broth containing 1% (w/v) glucose to inhibit the endogenous production of β-galactosidase by the chromosomal copy of the lacZ gene. We have shown previously that glucose does not affect the expression of the O7 LPS promoter (Marolda and Valvano, 1988). β-Galactosidase activity was determined as described elsewhere (Marolda et al., 1990).

Light and electron microscopy

Colonies from overnight plate cultures were examined by phase-contrast microscopy. Cell viability and morphology was assessed using the BacLight LIVE/DEAD kit from Molecular Probes according to the supplier's recommendations. Fluorescence microscopic images were obtained with a Zeiss Axioskop II equipped with a 100 × oil immersion objective (numerical aperture = 1.3), a cooled CCD camera and fluorescein isothiocyanate (FITC) and rhodamine filters. For electron microscopy, cells were suspended in saline and washed several times before staining with 1% (w/v) ammonium molybdate and placing onto copper grids. Specimens were examined at low magnification using a Phillips 300 electron microscope.

Serum resistance

Cells were examined for resistance to 95% serum as described previously (Taylor et al., 1972) using pooled fresh dog serum. Previous experiments showed that strain VW187 is equally resistant to human, rabbit, sheep and dog normal sera. Complement-depleted serum was obtained by heating at 56°C for 30 min.

Radioactive labelling of O7 subunits

The O7 subunits were labelled according to Marino et al. (1985). Cultures were grown in M9 medium supplemented with glycerol as the sole carbon source. Cells in mid-log phase were added to 100 nM d-[2,6-3H]-mannose (46 Ci mmol−1; Amersham), incubated for an additional 2 min at 37°C and then transferred to an ice-cold water bath. Cells were harvested by centrifugation (5000 g for 10 min at 4°C), resuspended in 10 ml of ice-cold suspension buffer [10 mM HEPES, 25% (w/v) sucrose, 4 µg ml−1 RNase, 4 µl of DNase, pH 7.4] and lysed by three passages through a French press cell at 10 000 psi. Cell lysates were centrifuged at low speed (5000 g for 10 min at 4°C), and the clear supernatants were layered on a 60% (w/v) sucrose cushion (10 mM HEPES, pH 7.4) followed by a 3 h centrifugation at 100 000 g. Cell membranes were collected from the interface of the sucrose cushion and layered onto a 30–60% (w/v) sucrose step gradient. The gradient was centrifuged (200 000 g for 24 h at 4°C) and fractionated by bottom puncture. Fractions were collected and examined spectrophotometrically for protein content and for radioactivity with a scintillation counter. The distribution of the cytoplasmic membrane through the gradient was monitored by assaying lactate dehydrogenase activity (Bergmeyer et al., 1983). The distribution of the outer membrane through the gradient was monitored by SDS–PAGE analysis with Coomassie brilliant blue staining and scoring the presence of characteristic outer membrane proteins OmpF and OmpC.

Cytoplasmic and outer membrane fractions were pooled and precipitated with cold trichloroacetic acid to eliminate unbound [2,6-3H]-mannose from particulate material. The relative incorporation of radioactive material in the outer membrane fractions was compared with that of the plasma membrane fractions and used to determine the amount of O7 LPS biosynthesis intermediates that accumulated in the cytoplasmic membrane.


J. A. Thomas and J. A. Gaspar contributed equally to this study, which was used in partial fulfilment of their PhD and MSc degrees respectively. J. A. Thomas obtained a posthumously awarded PhD degree from the University of Western Ontario in 1994. An obituary can be found on the worldwide web at We thank Lydia Dafoe for technical assistance, and all the colleagues mentioned in Table 1 for kindly providing us with strains and plasmids. We also thank R. Lloubès for the gift of TolA antiserum, S. Wilton for help with the electron microscopy, and D. E. Heinrichs for critical reading of the manuscript. Light microscopy morphological analyses were conducted in a departmental facility supported by an Academic Development Fund from the University of Western Ontario. This study was supported by research grants from the Natural Science and Engineering Research Council and the Medical Research Council of Canada.