Characterization of the interaction of IpaB and IpaD, proteins required for entry of Shigella flexneri into epithelial cells, with a lipid membrane


V. Cabiaux, Université Libre de Bruxelles, Laboratoire de Chimie Physique des Macromolécules aux Interfaces, CP 206/2 Boulevard du Triomphe, 1050 Brussels, Belgium. Fax: +32 02 6505382, Tel.: +32 02 6505365, E-mail:


Entry of Shigella flexneri into epithelial cells and lysis of the phagosome involve the IpaB, IpaC, and IpaD proteins, which are secreted by type III secretion machinery. We report here the purification of IpaB and IpaD and the characterization of their lipid-binding properties as a function of pH. The interaction of IpaB with the membrane was quite independent of the pH whereas that of IpaD took place only at low pH. To support the data obtained with the purified proteins, we designed a system in which protein secretion by live bacteria was induced in the presence of liposomes, thereby allowing interaction of proteins with lipids directly after secretion and bypassing any purification step. In these conditions, both IpaB and IpaC, as well as minor amounts of IpaA and IpgD, were associated with the membrane and the ratio of IpaB to IpaC was modulated by the pH. The relevance of these results with respect to the dual roles of IpaB, IpaC and IpaD in induction of membrane ruffles and lysis of the endosomal membrane is discussed.


large unilamellar vesicle


asolectin multilamellar vesicule



phenylmethanesulfonyl fluoride





poly(vinylidene difluoride)

Members of Shigella spp., including S. flexneri, are the etiological agents of shigellosis. In this disease, destruction of the epithelium of the colonic mucosa is provoked by the strong inflammatory response that is induced by invasion of the epithelium by bacteria. Entry of bacteria into epithelial cells involves actin polymerization, generating local membrane ruffling at the site of contact between the bacterium and the host cell. Bacteria then lyse the membrane of the endocytic vacuole, multiply in the cytoplasm, and move by inducing actin polymerization at one of their poles. The movement of intracellular bacteria generates the formation of cell protrusions that contain a bacterium at their tip, which are internalized by neighbouring cells. Lysis of the two cell membranes that surround bacteria in internalized protrusions completes the process of intercellular dissemination, which allows bacteria to spread from cell to cell without being exposed to the outside medium (reviewed in [1]).

All bacterial genes necessary for entry into epithelial cells are clustered within a 30-kb region of a 220-kb virulence plasmid [2]. This region encodes the secreted IpaA-D and IpgD proteins, their cytoplasmic IpgC and IpgE chaperones, and the Mxi-Spa type III secretion machinery [3–9]. Genetic analysis indicated that IpaB, IpaC, and IpaD are essential for entry into epithelial cells [10]. IpaB and IpaC are stored in the bacterial cytoplasm in association with IpgC, which is necessary for the stability of IpaB and the partitioning of IpaB and IpaC [11]. Once secreted, IpaB and IpaC associate in a complex in the extracellular medium. This complex is necessary and sufficient to induce membrane ruffling and entry of latex beads into epithelial cells [12]. When bacteria are endocytosed by macrophages, the same proteins are responsible for the lysis of the membrane vacuole [13,14]. Both proteins have several hydrophobic regions of sufficient length to form transmembrane helices. These data strongly suggest that Ipa proteins, either independently or as a complex, might interact with and destabilize cell membranes.

To characterize the potential interaction of proteins secreted by S. flexneri with lipid membranes, we have undertaken studies adopting two approaches. One approach involves the purification of the individual proteins suspected to interact with the cell membrane and the characterization of their interaction with an artificial membrane. The other approach involves the induction of secretion of Ipa proteins in the presence of pure lipid vesicles (liposomes), the purification of the proteoliposomes and the identification of the proteins that are associated with the lipid membrane. This should allow the identification of proteins that interact with the lipid membrane in conditions close to physiological conditions and, in the future, the characterization of the mechanism(s) that are responsible for this interaction, by reconstructing the system piece by piece. We have previously demonstrated that purified IpaC was able to bind to and to induce calcein release from lipid vesicles in a pH-dependent manner [15]. Here, we report the purification of IpaB and IpaD to homogeneity and the characterization of their ability to bind to and to induce calcein release from asolectin vesicles. IpaB, but not IpaD, was highly efficient in destabilizing the membrane of liposomes. We also identified the proteins that interact with lipids following induction of protein secretion by wild-type S. flexneri in the presence of liposomes at both pH 7.4 and 5.0.

Materials and methods

Bacterial strains and plasmids

M90T is the virulent, wild-type strain of S. flexneri 5 [16]. SF2000 [15] is a derivative of M90T in which both the ipaB and sepA genes have been inactivated [10,17]. SF914 is a derivative of Escherichia coli strain DH5α (endA1 hsdR17 supE44 thi1 recA1 gyrA relA1Δ(lacZYA-argF) U169 F′[ƒ80d lacδ(lacZ)M15]) harboring pMM100, a derivative of pACYC184 that expresses the LacI repressor, and pHI2, a derivative of pQE30 (Quiagen Inc.) that expresses His6-IpgC and IpaB [18]. Plasmid pHI2 contains an isopropyl thio-β- d-galactoside inducible lac promoter, a translation start site followed by six codons specifying His residues fused in frame with the ipgC coding sequence, the entire ipaB gene, and the first 57 codons of ipaC.


Tryptic soy broth medium was from ICN (Biomedicals Inc.), ampicillin and the Premixed Protein Molecular Weight Markers, Low-range (14.4–97.4 kDa) were from Boehringer Mannheim (Germany), phenylmethanesulfonyl fluoride was from Serva (Boehringer Ingelheim Bioproducts Partnership), asolectin, isopropyl β- d-thio-galactopyranoside, l-α-phosphatidylcholine from egg yolk (egg PtdCho), l-α-phosphatidylserine from bovine brain (PtdSer), and calcein were from Sigma (St Louis), and Talon™ Metal Affinity Resin (Co2+) was from Clontech Laboratories, Inc. (Palo Alto, USA). Lysozyme was a gift of Y. Looze (Université Libre de Bruxelles, Belgium). Urea, Tris, and NaCl were of HPLC grade, and all other reagents were of analytical grade. Asolectin (mixed soybean phospholipids) was purified according to [19] and calcein was purified by gel filtration chromatography on Sephadex LH-20 (Pharmacia, Uppsala, Sweden) as previously described in [20].

IpaB purification

A preculture of SF914 grown in TCS medium containing ampicillin (50 µg·mL−1) was diluted 100 times in TCS medium and incubated at 37 °C until the culture reached an optical density of 1 at 600 nm. Expression of His6-IpgC and IpaB was induced by adding IPTG (2 m m) and bacteria were incubated for 2 h at 37 °C. Bacteria were harvested by centrifugation of the culture at 4700 gav for 15 min at 4 °C (Sorvall RC26-PLUS, Rotor: SLA-1500), the pellet was resuspended in 10 mL of 20 m m Tris, 150 m m NaCl, pH 7.9, buffer and frozen at −20 °C. After thawing, lysozyme was added (1 mg·mL−1) and another freeze-thaw cycle was performed. Phenylmethanesulfonyl fluoride was added (5 m m) and the sample was sonicated three times for 15 s on ice (50 W, Sonifier B-12, Branson Sonic Power Company). The lysate was centrifuged at 6700av g (Sorvall RC26-plus, Rotor: SA-300) for 10 min and the supernatant was then incubated with 1.5 mL of the Co2+ affinity resin at 4 °C for 1 h under gentle stirring. The mixture was poured into a 10 mL column (Biorad Econo system low pressure chromatography) and the resin was washed with 10 mL of washing buffer (20 m m Tris/150 m m NaCl, pH 7.9), 6 mL of washing buffer containing 5 m m 2-mercaptoethanol, 6 mL of washing buffer containing 4 m urea and, 6 mL of washing buffer containing 8 m urea. Proteins bound to the resin were eluted with 10 mL of washing buffer containing 0.25 m imidazole. Aliquots of elution fractions were trichloroacetic acid precipitated and analyzed by a 12% SDS/PAGE according to [21]. IpaB-containing fractions were pooled and the pH of the sample was ajusted to pH 9.5 with concentrated NaOH prior to loading onto a Mono-Q column (anion exchange, Pharmacia, Uppsala Sweden) equilibrated with 5 m m Tris/8 m urea, pH 9.5. Elution was carried out with an NaCl 0–50% gradient (with buffer A containing 8 m urea, 5 m m Tris, pH 9.5, and buffer B containing 8 m urea, 5 m m Tris, 1 m NaCl, pH 9.5) using a flow of 1 mL·min−1. The protein N-terminal amino-acid sequence was determined by automated Edman degradation as described in [15].

IpaD purification

IpaD was purified from the culture medium of SF6200. The culture conditions and the ammonium sulfate (45%, w/v) precipitation and the anion exchange chromatography have been described previously [15]. The IpaD-containing fractions were pooled and their pH was adjusted to pH 6.5. The sample was then loaded on a Mono-Q column equilibrated with 4 m urea/5 m m Tris, pH 6.5, and eluted by a 0–25% NaCl gradient (with buffer A countaining 4 m urea/5 m m Tris, pH 6.5, and buffer B containing 4 m urea/5 m m Tris/1 m NaCl, pH 6.5) using a flow rate of 1 mL·min−1.

Liposome preparation

Asolectin multilamellar vesicules (MLVs) were formed in 10 m m Hepes, 150 m m NaCl, 1 m m EDTA, pH 7.2 as previously described [15]. Large unilamellar vesicles (LUVs) with a diameter of 0.1 µm were prepared at room temperature according to the procedure of [22] by using an extruder (Lipex Biomembranes Inc., Vancouver, Canada). The liposome concentration was estimated by measuring the lipid phosphorus content as previously described [23]. LUVs containing 60 m m calcein were prepared in 10 m m Hepes, pH 7.2, according to the above procedure. The nonencapsulated dye was removed from the liposome suspension by chromatography on a Sephadex G-50 column equilibrated with 10 m m Hepes, 150 m m NaCl, and 1 m m EDTA, pH 7.2. The osmolarity of the solution (calcein and buffer) was measured with a Roebling osmometer and fixed to 310 ± 5 mosmol·kg−1 of water.

Fluorescence measurements and calcein release assay

Fluorescence spectra were recorded at 37 ± 0.5 °C on a SLM 8000 fluorimeter as described [15]. Fluorescence of calcein, which is entrapped in unilamellar vesicles at a self-quenching concentration, increases after release of calcein from liposomes. The protein-induced release of calcein was recorded by measuring the fluorescence of the liposome suspension at excitation and emission wavelengths of 490 and 520 nm, respectively. Complete release of the dye was obtained by lysing the LUVs with Triton X-100 at a final concentration of 0.1% (v/v). The percentage of fluorescence Ft at time t is defined as:


where I0 is the initial fluorescence obtained after dilution of the vesicles in the appropriate buffer, If is the total fluorescence observed after addition of Triton X-100, and It is the fluorescence at time t corrected for the dilution. An aliquot of the LUV suspension containing the dye (pH 7.2) was added to the cuvette containing the buffer at the chosen pH (total volume of 1 mL). After pH and temperature equilibration, the protein was added to the suspension and the fluorescence was immediately measured. The concentration of urea in the assay never exceeded 150 m m, a concentration that had no effect on stability of vesicles, regardless of the pH of the suspension (data not shown). Buffers used for the different pH were as follows: 10 m m Hepes, 150 m m NaCl, 1 m m EDTA, pH 7.2; 10 m m Mes, 150 m m NaCl, 1 m m EDTA, pH 6.5 to pH 6.0; and 10 m m acetate, 150 m m NaCl, 1 m m EDTA, pH 5.0 to pH 5.4.

Preparation and characterization of LUV containing proteins

Bacteria were grown under aerated conditions in 60 or 100 mL of TCS medium at 37 °C until they reached the end of the exponential growth phase. A 10 mL volume of culture was harvested by centrifugation at 3300 g for 12 min at 37 °C, and resuspended in a 1/20 volume of NaCl/Pi (1.47 m m KH2PO4, 0.87 m m Na2HPO4, 10 m m KCl, 137 m m NaCl), pH 7.4 or 5.0, at 37 °C. Asolectin LUVs prepared as described above in NaCl/Pi at both pH values were added to the bacterial suspension (7 mg LUV per mL suspension) and Ipa secretion was induced by addition of Congo red at a final concentration of 150 µm. After 15 min of incubation at 37 °C, bacteria were pelleted by centrifugation at 14 000 g for 5 min. phenylmethanesulfonyl fluoride (final concentration 5 × 10−4m) was added to the supernatant, which was mixed with an equal volume of 80% (w/v) sucrose in NaCl/Pi, pH 7.4 or 5.0. A 30 to 2% sucrose gradient was poured on top of the 40% sucrose layer and samples were centrifuged at 126 000 g for 16 h at 4 °C. Liposomes were collected, mixed with 3–4 vol. of NaCl/Pi containing 0.5 m NaCl, pH 7.4 or 5.0, and centrifuged at 126 000 g for 35 min. The pellet was resuspended in NaCl/Pi at the appropriate pH and analyzed by gel electrophoresis on a 12% SDS polyacrylamide gel according to [21]. Protein bands were excised from the gel, transferred onto a poly(vinylidene difluoride) (PVDF) membrane. The protein N-terminal amino-acid sequences were determined by automated Edman degradation using a Beckman LF 3400D protein-peptide microsequencer.


Purification of IpaB

We first tried to purify IpaB from the culture medium of the S. flexneri ipaD mutant SF622 that constitutively secretes IpaB and IpaC [10,25]. However, even in the presence of detergents, bile salts (deoxycholate), or high urea concentrations, IpaB was either copurified with IpaC or with IpaH9.8 and IpaH4.5, which are secreted proteins with molecular weights similar to that of IpaB [26]. As IpaB is stored in the bacterial cytoplasm in association with the IpgC chaperone [11], we sought first to purify the IpgC–IpaB complex from cytoplasmic extracts and then to separate IpaB from IpgC. To facilitate purification of the IpgC–IpaB complex, a His tag was added to the N-terminus of IpgC so that purified IpaB would not carry any covalent modification. Construction of the recombinant plasmid pHI2 expressing the His-tagged IpgC protein, designated His6-IpgC, together with IpaB under the control of an isopropyl thio-β- d-galtoside regulated lac promoter is described elsewhere [18]. A cytoplasmic extract of E. coli strain SF914 was loaded onto a Co2+ affinity column that was then extensively washed in presence of 5 m m 2-mercaptoethanol, and 4 m and 8 m urea to avoid nonspecific binding of E. coli proteins to the resin. Proteins were then eluted by a buffer containing 20 m m Tris, pH 7.9, 150 m m NaCl and 250 m m imidazole. The elution fractions contained, as expected, mostly IpaB (63 kDa) and His6-IpgC (19 kDa) and a contaminant of 26 kDa (not shown). IpaB-containing fractions were pooled and IpaB (pI = 8.32) was then purified by anion exchange chromatography using a Tris pH 9.5 buffer containing 8 m urea. IpaB was eluted from the column at a salt concentration of 210 m m NaCl ( Fig. 1A), as indicated by SDS/PAGE analysis of fractions of chromatography ( Fig. 1B). The N-terminal sequence of the 63 kDa protein present in fraction 19 was MHNVSTTTTGFPLAKI, which corresponded to that of IpaB. No other N-terminal sequences were detectable in the sample, indicating that IpaB had been purified with a degree of purity over 99%. In addition, this suggested that the minor band observed on the gel corresponded to a degradation product of IpaB lacking a C-terminal portion. The yield of purification of IpaB was 2 mg·L−1 of culture.

Figure 1.

Purification of IpaB and IpaD. (A) The elution profile of a representative purification of IpaB on a Mono-Q column, with the absorbance at 280 nm indicated by a continuous line and the salt gradient profile indicated by a dashed line (100% = 1 m NaCl). (B) Coomassie blue staining of a 12% SDS/PAGE of fractions of the chromatogram. Numbers of fractions are indicated on the chromatogram. (C) 10% SDS/PAGE and Coomassie blue staining of a sample of IpaD purified as described in Material and methods.

Purification of IpaD

IpaD was purified from the culture medium of the S. flexneri ipaB sepA mutant SF6200 that was previously used for purification of IpaC [15]. A first anion exchange chromatography carried out under the conditions described for the purification of IpaC (5 m m Tris, pH 9.0, 4 m urea) allowed us to obtain fractions enriched in IpaD (not shown), although these were still contaminated by small amounts of IpaA (78 kDa) and IpaC (46 kDa). A second anion exchange chromatography was performed at pH 6.5, still in the presence of 4 m urea, and IpaD was eluted at 150 m m NaCl. The purity of IpaD was estimated to be over 90% on the basis of a 10% SDS Coomassie Blue stained gel ( Fig. 1C). The yield of purification of IpaD was 1 mg·L−1 of culture.

Interaction of IpaB and IpaD with the lipid membrane as a function of pH

To investigate the potential interaction of IpaB and IpaD with a lipid membrane, we studied the ability of the purified proteins to induce the release of a fluorescent probe encapsulated into large unilamellar vesicles. IpaB was able to induce a fast and efficient release of the encapsulated calcein ( Fig. 2). In contrast, the ability of IpaD to induce calcein release was weak as a significant calcein release was only observed below pH 5.5 and at a protein concentration of 328 n m ( Fig. 2B). The possibility that the observed effect could be due to a small contamination of the IpaD sample by IpaC can be ruled out as the pH profile of calcein release by IpaC is different from that observed here [15]. Because of the low destabilizing effect of IpaD, we focused on IpaB to determine the following characteristics . As the calcein release induced by IpaC was pH-dependent and its ability to destabilize vesicles was decreased when the protein was preincubated at low urea concentration [15], we investigated the effect of the pH and the concentration of urea of the medium on the release of calcein induced by IpaB. In contrast to the pH-dependence observed with IpaC, the pH of the reaction had little effect on the calcein release induced by IpaB ( Fig. 2). IpaB was then diluted to a final concentration of 16 n m in buffers that did not contain urea and incubated for 20 min at 37 °C prior to addition of calcein loaded vesicles and monitoring of calcein release. Preincubation at pH 7.2 decreased the ability of IpaB to induce calcein release by 30%, whereas preincubation at pH 5.0 did not modify the level of calcein release (data not shown).

Figure 2.

Effect of pH on the release of calcein induced by IpaB and IpaD. (A) Kinetics of calcein release from asolectin LUVs (5 µm) induced by IpaB (3.2 n m) at 7.2, 6.5, 6.0, 5.5 and 5.0 (from a to e, respectively). The arrow indicates the addition of proteins. The fluorescence due to vesicles was subtracted from the spectra and the kinetics were rescaled between 0 (F0) and 1 (F100). (B) Percentage of calcein release from asolectin LUV (5 µm) induced by IpaB (3.2 n m) (□) and IpaD (328 n m) (▪) obtained after 1 min and calculated as described in Material and methods.

Interaction of IpaB with the lipid membrane as a function of lipid composition

Asolectin contains 20% negatively charged lipids. To investigate the role of charged lipids in the interaction of IpaB with the lipid membrane, we used vesicles of egg PtdCho, a neutral lipid. No calcein release was observed with these vesicles at either pH 7.2 (data not shown) and pH 5.0 ( Fig. 3). Addition of 20% PtdSer, a negatively charged lipid, did not restore any calcein release at pH 7.2 but increased calcein release at pH 5.0 ( Fig. 3).

Figure 3.

Effect of negatively charged lipids on the release of calcein induced by IpaB (3.2 n m) at pH 5.0. Kinetics of calcein release from LUVs (5 µm): a, asolectin; b, egg PtdCho; c, egg PtdCho-PtdSer (8 : 2, w : w). The arrow indicates the addition of the protein. Kinetics were rescaled between 0 (F0) and 1 (F100).

Determination of an apparent binding constant of IpaB to asolectin vesicles

To determine the apparent binding constant of IpaB with the lipid membrane, calcein release from asolectin LUVs was studied as a function of IpaB concentration, for four lipid concentrations (5, 12.5, 25, and 37.5 µm) at pH 7.2 ( Fig. 4A) and pH 5.0 ( Fig. 4B). As the lipid concentration increased, a higher concentration of IpaB was required to release a determined amount of calcein, which suggested that the leakage rate was determined by the amount (termed ‘r’) of membrane-bound protein per lipid molecule. This quantity r is related to such experimental parameters as Ct, the total protein concentration, Cf, the free protein concentration, and L, the lipid concentration, through the equation Ct = Cf + rL (for a more detailed description of this formalism, see [27]). Plots of Ct versus L for different extents of release (between 25% and 45%) were obtained (not shown) from a plot of the percentage of calcein release versus Ct ( Fig. 4A,B) at different L values. According to the mass equation, the corresponding r and Cf values were calculated from the slope (evaluated from the linear part of the curve) and the intercept, respectively. The relationships between r and Cf (binding isotherm) are shown in Fig. 4C. An apparent binding constant was evaluated by linearly extrapolating the curve to the zero concentration of free protein [28], which gave a Kapp of 4.9 × 106m−1 at pH 7.2 and of 0.37 × 106m−1 at pH 5.0.

Figure 4.

Effect of IpaB and lipid concentrations on the release of calcein. Percentages of calcein released from asolectin LUVs were determined at various concentrations of IpaB and lipids at pH 7.2 (A) and 5.0 (B). Lipid concentration: ◆, 5 µm; ▪, 12.5 µm; ▴, 25.0 µm; ●, 37.5 µm. (C) Evaluation of the apparent binding constant of IpaB to asolectin LUVs. The binding isotherms, are given at pH 7.2 (●) and pH 5.0 (▴).

Association of proteins secreted by S. flexneri with a lipid membrane

To support the data obtained with the purified proteins, we bypassed the purification steps and identified the proteins secreted by S. flexneri that could spontaneously associate with a membrane by using Congo red to activate the Mxi-Spa secretion machinery [29] in live bacteria incubated in the presence of liposomes. As the pH differentially affected the ability of purified IpaB (this work) and IpaC [15] to interact with and destabilize liposomes, we investigated the association of the secreted proteins with liposomes when secretion was induced at pH 7.4 or 5.0. The protein content of proteoliposomes prepared as described in Materials and methods was characterized by SDS/PAGE and silver staining. When secretion was induced at pH 7.4, four proteins were associated with the liposomes ( Fig. 5A). From their N-terminal sequences (MHNVNNTQAPTFLYK, MHNVSTTTTGFPLAKI, MEIQNTKPTQTLY, MHITNLGLHQVSFQ), these proteins were identified as IpaA, IpaB, IpaC, and IpgD, respectively, with an IpaB/IpaC molar ratio of 2. When secretion was induced at pH 5.0, five proteins were associated with the liposomes ( Fig. 5B) and identified as IpaA, IpaB, IpaC, IpaD (MNITTLTNSISTSSF), and IpgD with a IpaB/IpaC molar ratio of 1. In the absence of lipid vesicles, the same proteins were found in the bacterial supernatant whatever the pH at which secretion was induced by Congo red (data not shown). These results confirm that IpaB, IpaC and IpaD were able to interact with lipid vesicles and that the pH of the medium differentially affected their association with lipids.

Figure 5.

Association of secreted proteins to asolectin vesicles. Proteins associated with purified LUVs following incubation of bacteria in the presence of Congo red and LUV at pH 7.4 (A) or 5.0 (B) were analyzed by SDS/PAGE and silver staining. The analyzed sample corresponds to 450 µg of phospholipids (1.3 × 108 bacteria).


Entry of S. flexneri into epithelial cells involves membrane ruffles at the sites of contact between bacteria and cells. No Ca2+ fluxes have been detected during bacterial entry, indicating that the membrane integrity of the cell was maintained during this process [30]. Genetic analyses have shown that IpaB, IpaC, and IpaD are each required for both entry of bacteria into epithelial cells and lysis of the phagocytic vacuole when bacteria are internalized by macrophages [13,14]. This suggests that these proteins might adopt different behaviors depending upon their physico-chemical environment, one leading to cytoskeleton reorganization and the other to lysis of the lipid membrane. This could be analogous to what is observed with some bacterial toxins, such as diphtheria toxin and the Bacillus anthracis toxins, which can either interact with the cell surface or insert deeply into the membrane. The different behaviors of these toxins are triggered by changes in the pH that follows their internalization into endosomes (reviewed in [31]). To get insight into the roles of IpaB, IpaC, and IpaD in interactions of bacteria with cell membranes, we have purified each of these proteins, characterized their association with artificial lipid vesicles, and investigated whether this association was modulated by the pH (Ipa C [15]).

Both IpaB and IpaC were able to interact with and destabilize the membrane of liposomes. IpaD also had a membrane destabilizing activity, although this activity required a much higher protein concentration compared to that of IpaB and IpaC. Although the ability of IpaB to induce calcein release showed little dependence upon pH, its lipid-binding constant was lower at low pH than at neutral pH. This suggests that IpaB is more efficient in destabilizing the membrane at pH 5.0 than at neutral pH. In addition, the presence of negatively charged lipids in the membrane had an effect on calcein release induced by IpaB at low pH but not neutral pH, and preincubation of IpaB in conditions promoting its association with the membrane, such as low urea concentration, had no effect on calcein release at pH 5.0 but decreased the release at pH 7.2. These results indicate that IpaB does not behave similarly at neutral and low pH, which suggests that the conformation of the protein is affected by the pH. In the case of IpaC, calcein release was strictly dependent upon the pH and the binding constant was higher at low pH than at neutral pH [15]. Therefore, even though the interaction of both IpaB and IpaC with the membrane was affected by the pH, the two proteins did not respond in the same fashion to pH changes. Whether the conformational changes of IpaB and IpaC are associated with secondary, tertiary, or quaternary modifications, such as oligomerization of the proteins, remains to be determined.

We also identified the proteins that associate with lipid membrane following activation of the Mxi-Spa secretion machinery in live bacteria in the presence of liposomes. Although this system is less amenable to kinetic studies compared to that using purified proteins, it has two interesting features: all secreted proteins are present in the reaction mixture and protein purification steps that involve the presence of high concentrations of urea are bypassed. When secretion was induced at neutral pH, four proteins were associated with the lipids: IpaB and IpaC, as well as IpaA and IpgD in lesser amounts, with a molar ratio of two IpaB for one IpaC. When secretion was induced at pH 5.0, a fifth protein, IpaD, was associated with liposomes and the molar ratio was one IpaB for one IpaC. The presence of IpaD in liposomes obtained at pH 5.0 and its absence in liposomes obtained at pH 7.4 was consistent with the observation that purified IpaD could induce calcein release at low pH but not at neutral pH. The increased amount of IpaC relative to IpaB that was associated with liposomes at pH 5.0 compared to pH 7.4 was also consistent with results obtained with purified proteins that indicated that the association of IpaB with LUVs was decreased and that of purified IpaC was increased at low pH [15]. These results on the membrane association of proteins secreted by the type III secretion machinery of S. flexneri in the presence of liposomes are also consistent with the observation that IpaB, IpaC, IpaA, and IpgD were associated with the membrane of erythrocytes following a hemolytic assay [32]. Although roughly similar amounts of IpaA, IpaB, IpaC, and IpgD are secreted in response to Congo red, lower amounts of IpaA and IpgD were associated with the membranes of liposomes. This suggests that IpaA and IpgD have a lower capacity to interact with the lipid membrane compared to IpaB and IpaC. IpaA has been shown to bind vinculin [33], a cytoplasmic protein involved in actin polymerization, and IpgD is homologous to SopB, a protein translocated by the type III secretion machinery of Salmonella dublin[34]. This suggests that the interaction of IpaA and IpgD with the membrane of liposomes might be indirect, perhaps mediated by their association with IpaB and IpaC.

The consistency of results obtained using purified proteins and proteins secreted by bacteria suggests that individual proteins exhibit properties that are relevant to the membrane association of native proteins, possibly present within a complex as proposed for IpaB and IpaC [12]. Future studies will be aimed at unravelling the mode of interaction of individual proteins and of proteins associated within a complex to the membrane.


V. C. is Senior Research Associate of the Fonds National de la Recherche Scientifique (Belgium).