The Tol/Pal system of Escherichia coli is composed of the YbgC, TolQ, TolA, TolR, TolB, Pal and YbgF proteins. It is involved in maintaining the integrity of the outer membrane, and is required for the uptake of group A colicins and DNA of filamentous bacteriophages. To identify new interactions between the components of the Tol/Pal system and gain insight into the mechanism of colicin import, we performed a yeast two-hybrid screen using the different components of the Tol/Pal system and colicin A. Using this system, we confirmed the already known interactions and identified several new interactions. TolB dimerizes and the periplasmic domain of TolA interacts with YbgF and TolB. Our results indicate that the central domain of TolA (TolAII) is sufficient to interact with YbgF, that the C-terminal domain of TolA (TolAIII) is sufficient to interact with TolB, and that the amino terminal domain of TolB (D1) is sufficient to bind TolAIII. The TolA/TolB interaction was confirmed by cross-linking experiments on purified proteins. Moreover, we show that the interaction between TolA and TolB is required for the uptake of colicin A and for the membrane integrity. These results demonstrate that the TolA/TolB interaction allows the formation of a trans-envelope complex that brings the inner and outer membranes in close proximity.
The Tol/Pal system is a multiprotein complex embedded into the envelope of Escherichia coli. The proteins of this system are encoded by a cluster of genes organized into two operons. The first one encodes YbgC, TolQ, TolR, TolA, TolB, Pal and YbgF and the second one from an inner promoter encodes TolB, Pal and YbgF (Vianney et al., 1996; Muller and Webster, 1997). YbgC is a cytoplasmic protein, TolQ spans the inner membrane three times whereas TolR and TolA have one transmembrane domain only that anchors the protein in the inner membrane, the remaining of the protein being periplasmic. YbgF and TolB are periplasmic proteins and Pal is an outer membrane lipoprotein associated to the peptidoglycan (Lazzaroni and Portalier, 1992; Kampfenkel and Braun, 1993; Isnard et al., 1994; Vianney et al., 1994).
The Tol/Pal system has been parasited by filamentous bacteriophages and group A colicins (Davies and Reeves, 1975; Bénédetti and Géli, 1996). Colicins are protein toxins produced by E. coli and related species. They are organized into three distinct structural domains, each one being implicated in a stage of their mechanism of action. First, the central domain binds to a receptor at the cell surface. This binding results into substantial unfolding of the colicin polypeptide chain (Duchéet al., 1994). Then, the N-terminal domain translocates across the outer membrane, interacting with the translocation machinery. Finally, the C-terminal domain that carries the lethal activity, such as pore-forming or enzymatic activity, reaches its cellular target and kills the cell (Lazdunski et al., 1998; 2000).
The organization of the Tol/Pal system in the envelope has been extensively studied. Several experiments, such as cross-linking with formaldehyde, coimmunoprecipitations and genetic suppression experiments, have shown that these proteins interact with each other and are organized into two complexes. The inner membrane complex contains TolA, TolQ and TolR. The TolA transmembrane domain interacts with the first transmembrane domain of TolQ and with TolR transmembrane domain (Dérouiche et al., 1995; Germon et al., 1998). The first and the third transmembrane domains of TolQ seem to be in close contact (Lazzaroni et al., 1995). The third transmembrane domain of TolQ interacts with TolR transmembrane domain and with TolR dimer. TolRII and TolRIII, the two periplasmic domains of TolR, have the intrinsic capacity to dimerize (Journet et al., 1999). The second complex, associated to the outer membrane, is composed of TolB, Pal, Lpp and OmpA. TolB interacts with Pal, but also with Lpp and OmpA (Bouveret et al., 1995; Clavel et al., 1998; Ray et al., 2000). The only direct evidence of an interaction between the inner and the outer membrane complexes is the recently reported interaction between TolA and Pal. This interaction depends of the proton motive force (p.m.f) (Cascales et al., 2000).
To further investigate these interactions, we used the yeast two-hybrid system to identify and characterize new interactions between the Tol/Pal proteins or the periplasmic domains of these proteins, and between these proteins and colicin A. Using this technique, most of the interactions described previously were confirmed. Three new interactions never detected before have been detected. YbgF interacts with TolA, TolB dimerizes and TolA interacts with TolB. We show that the central domain of TolA (TolAII) is sufficient to interact with YbgF and that the C-terminal domain of TolA (TolAIII) and the N-terminal domain of TolB (D1) are sufficient for the TolA/TolB interaction. Futhermore, we show that the interaction between TolA and TolB is necessary for colicin A uptake and membrane integrity.
Complete network of interactions between Tol/Pal proteins reveals three new interactions
We used the yeast two-hybrid system to investigate new interactions between the Tol/Pal proteins, and between these proteins and colicin A. We chose the mature sequences of TolB, Pal, YbgC and YbgF, the periplasmic domains of TolA (TolAII-III: starting from the 66th codon of the tolA gene) and TolR (TolRII-III: starting from the 45th codon of the tolR gene), the cytoplasmic domain of TolQ (TolQc from the 36th codon to the 132th one), the translocation domain (AT1: first ATG codon to the 178th codon) and the cytotoxic domain (C-ter colA: starting from the 339th codon) of colicin A. We fused these sequences either to the DNA binding domain of LexA (in plasmid pEG-202) or to the B42 acidic transcriptional activation domain (in plasmid pMW-104). When a LexA protein fusion interacts with a B42 protein fusion in an appro-priate yeast reporter strain Saccharomyces cerevisiae EGY48lacZ, lacZ and LEU2 reporter genes are activated. As a consequence of this interaction, the yeast colonies are blue on Xgal plates and grow on minimal plates lacking leucine. All pairwise combinations of pEG-202 and pMW-104 derivatives were introduced into EGY48lacZ and screened for their ability to activate the two reporter genes. An interaction was considered positive only if both reporter genes were activated and if the LexA and B42 fusion proteins alone did not activate the reporter genes. The results are shown in Table 1. LexA-YbgF, LexA-AT1 and LexA-C-ter colA fusion proteins activate on their own transcription of LEU2 and/or lacZ reporter genes and were not further used. Most of the known interactions among the Tol/Pal proteins and between Tol/Pal proteins and colicin A were confirmed by our yeast two-hybrid system. Indeed, we found that AT1 interacts with both TolB and TolAII-III as described (Bénédetti et al., 1991a; Bouveret et al., 1997; 1998; Gokce et al., 2000), we found that TolB interacts with Pal as described (Bouveret et al., 1995; Clavel et al., 1998; Ray et al., 2000) and that TolRII-III dimerizes as described (Journet et al., 1999). The above data indicate that a good correlation exists between interactions previously detected and the ones detected with our two-hybrid analysis. This demonstrates the reliability of this system for studying the interactions between protein components of the Tol/Pal system.
Three new interactions, that had not been previously detected, have been found using the yeast two-hybrid system. TolAII-III interacts with YbgF, TolB dimerizes and strikingly, TolAII-III interacts with TolB. TolAII–III/TolB interaction was proved in the two orientations.
From these results, we conclude that a direct interaction probably exists in vivo, in the bacteria, between TolAII-III and TolB, between TolAII-III and YbgF and that TolB may dimerize.
TolAII domain is sufficient to mediate a two-hybrid interaction with YbgF
To localize the region(s) of TolA involved in its interaction with YbgF, a series of TolA mutants containing deletions were created and fused to LexA and B42 in pEG-202 and pMW-104 respectively (Fig. 1A).
TolA is 421 amino acids long. It has been subdivided into three structural domains: TolAI (amino acids 1–42) is the transmembrane domain; TolAII (amino acids 48–310) is the central domain; and TolAIII (amino acids 314–421) is the C-terminal domain. Each domain is delimited from the other by a stretch of glycine residues (Levengood et al., 1991).
Initially, we checked using Western blot analysis that all of the five LexA-TolAII-III mutants and B42-YbgF were expressed in yeast (not shown). As the β-galactosidase activity in the yeast two-hybrid system assay is a relative measure of the interaction between proteins (Estojak et al., 1995), it provided a mean to determine the strength of binding between the TolA derivatives and YbgF. Each of the TolAII-III mutants fused to LexA was coexpressed with B42-YbgF in the yeast reporter strain, and β-galactosidase activity of the different transformants was measured.
As shown in Fig. 1B, a LexA–TolAII–III fusion, when 100 residues at the C-terminal were deleted (LexA–TolAII), exhibited a significant strength of interaction with B42-YbgF, whereas LexA–TolAII–III was completely defective in this interaction when N-terminal 235 residues were deleted (LexA–TolAIII). Furthermore, a 40-amino-acid deletion in the C-terminal region of LexA–TolAII–III (LexA–TolAII–IIIΔ) does not differ in the strength of its interaction with B42–YbgF. These findings show that the YbgF interaction domain is located in the central region of TolA (TolAII).
Our results clearly show that the central domain of TolA (TolAII) is the region of highest affinity for YbgF and that, in contrast to TolAII, TolAIII is dispensable for TolA/YbgF interaction.
TolAIII domain is required to mediate an interaction with TolB
To localize the region(s) of TolA involved in its interaction with TolB, we used the above TolA mutants. Each of them fused to LexA (Fig. 1A) was coexpressed with B42-TolB, and β-galactosidase activity of the different transformants was measured (Fig. 1C). Interestingly, LexA–TolAIII does not significantly differ in the strength of interaction with B42-TolB to full-length LexA–TolAII–III, whereas LexA–TolAII displayed only background levels of β-galactosidase activity. These findings show that the TolB interaction domain is located in the C-terminal region of TolA (TolAIII), between residues 301 and 421. Furthermore, the background levels of β-galactosidase activity displayed by LexA–TolAII–IIIΔ and LexA–TolAIIIΔ with B42–TolB suggest that the TolB interaction interface might involve residues 336–377. However, we cannot exclude that the 40-amino-acid deletion may have destructured the TolAIII domain. Similar results were found when β-galactosidase activity of transformants coexpressing LexA–TolB and B42–TolAII–III mutants was measured (data not shown). We checked that all the pairs of fusion protein that activate the lacZ gene were able to activate the transcription of LEU2 gene (data not shown).
Our results clearly show that TolAII is dispensable for the TolB interaction, and that the region that interacts with TolB is located in the C-terminal part of the TolA protein: TolAIII.
The TolB-D1 domain is sufficient for the interaction with TolAIII
As the interaction between TolA and TolB may reflect, in E. coli, a link between the cytoplasmic membrane complex and the outer membrane complex, we focused our attention on it.
The TolB protein is 409 amino acids long. It consists of two domains: the N-terminal domain (D1) and the β-propeller C-terminal domain (D2) (Abergel et al., 1999; Ponting and Pallen, 1999; Carr et al., 2000). In an attempt to define the domain of TolB which is responsible for interaction with TolA, TolB-D1 (Δ145-409) and TolB-D2 (Δ1-119) were fused to LexA and B42 in pEG-202 and pMW-104 respectively. After checking that each of the B42-TolB mutants were correctly expressed in yeast (data not shown), their ability to interact with LexA–TolAIII was analysed in the two-hybrid assay by measuring β-galactosidase activity of the different transformants (Fig. 2). We observed that the B42–TolB–D1 deletion mutant presented a significant reduction in the β-galactosidase activity but still retained the capacity to interact with LexA–TolAIII. The deletion of the N-terminal 119 residues (B42–TolB–D2) abolished the interaction with LexA–TolAIII as no β-galactosidase activity could be detected in this assay. Similar results were found when β-galactosidase activity of transformants coexpressing B42–TolAIII and LexA–TolB derivatives was measured (data not shown). We checked that all the pairs of fusion protein that activate the lacZ gene were able to activate the transcription of LEU2 gene (data not shown).
In contrast to the N-terminal part of TolB, TolB-D1, the C-terminal part of TolB, TolB-D2, is not sufficient and probably not necessary for interaction, although it seems that TolB-D2 might strengthen the interaction.
To confirm that TolB was interacting with TolAIII by the D1 domain, we screened for mutants of TolB-D1 that do not interact with TolAIII any more. A library of randomly mutagenized TolB-D1 domains was generated by polymerase chain reaction (PCR) amplification. The mutagenized TolB-D1 pool was digested with restriction sites flanking the TolB-D1 coding sequence and the DNA fragments were subcloned into the yeast expression vector pMW-104 as a fusion with wild-type B42 transcriptional activation domain (see Experimental procedures). The TolB-D1 mutants that lost the ability to interact with TolAIII were sequenced. We found that a triple substitution Q22P G25R Y91H in TolB-D1 strongly decreases the LexA–TolAIII/B42–TolB–D1 interaction (data not shown). The expression of the mutant fused to B42 was checked in yeast. We analysed the presence of B42-TolB-D1 and B42-TolB-D1 Q22P G25R Y91H, using HA antibody, by immunoblotting identical amounts of total protein extracts. Quantification analysis indicates that the amount of B42–TolB–D1 Q22P G25R Y91H remains unchanged compared with the wild-type B42–TolB–D1 fusion (not shown). Moreover, each of the three substitutions is required to inhibit TolB–D1/TolAIII interaction. Similar results were found when β-galactosidase activity of transformants coexpressing B42–TolAIII and LexA–TolB–D1 Q22P G25R Y91H mutant was measured (data not shown).
Taken together, these data are consistent with the idea that TolB-D1 domain may contain the region of highest affinity to mediate an interaction with TolAIII.
TolAIII and TolB-D1 can be cross-linked in vitro
To determine whether the two-hybrid interaction between TolAIII and TolB is due to TolAIII/TolB binding or is mediated by yeast proteins, in vitro cross-linking experiments were performed. Purified TolAIII (Deprez et al., 2000) and purified TolB (Abergel et al., 1999) were treated with formaldehyde or EGS (Ethylene Glycol-bis Succinimidylsuccinate) as cross-linkers. A specific 57 kDa band was recognized by anti-TolB and anti-TolAIII antibodies only in the presence of formaldehyde (Fig. 3A, lane 4). This band disappears in the absence of TolAIII or TolB (lanes 1 and 2), in the absence of cross-linker (lane 3) and when the sample is heated for 15 min at 96°C (lane 5), a treatment that dissociates formaldehyde cross-linked proteins. Similar results were obtained using EGS instead of formaldehyde (data not shown).
These results reveal a direct interaction between TolB and TolAIII.
In the course of this experiment, we observed the formation of a higher molecular weight specie that was recognized only by anti-TolB antibody, when cross-linking experiments were performed using TolB alone or TolB with TolAIII (Fig. 3A, lanes 2 and 4). These results indicate that TolB dimerizes in vitro.
The two-hybrid analysis has suggested that TolAIII interacts with TolB-D1. We thus tested purified TolB-D1 for its ability to cross-link in vitro with TolAIII (Fig. 3B). A specific band corresponding to a TolAIII/TolB-D1 cross-linked product was detected with formaldehyde (lane 4). In addition, TolB-D1 is cross-linked efficiently to a dimer size (lanes 2 and 4) and possibly trimer and tetramer sizes (lane 2). Other bands are detected. Owing to the purity of the preparation, the presence of another protein in the complex can be ruled out. These bands might therefore reflect intramolecular cross-links, which alter the mobility of TolB-D1 dimer. In absence of TolB-D2, TolB-D1 could adopt several conformations that lead to different cross-linking sites. It must be noted that in the presence of TolAIII (lane 4), high oligomers of TolB-D1 disappear; TolB-D1 preferentially dimerizes and heterodimerizes with TolAIII. All these bands were absent in the control without formaldehyde (lane 3) and disappeared when the sample was heated (lane 5).
All these results are consistent with our two-hybrid analysis that detected an interaction between the N-terminal region of TolB and TolAIII and a dimerization of TolB. The D1 domain of TolB is sufficient for TolA/TolB interaction and TolB dimerization.
The interaction between TolA and TolB is involved in colicin A translocation and envelope integrity
To demonstrate a physiological significance for the interaction between TolAIII and TolB, we screened for mutants of TolAIII that do not interact with TolB anymore. We then studied the effects of these mutations in vivo in E. coli. A library of randomly mutagenized TolAIII domains was generated by PCR amplification. The mutagenized TolAIII pool was digested with restriction sites flanking the TolAIII coding sequence, and the DNA fragments were subcloned into the yeast expression vector pEG-202 as a fusion with wild-type LexA binding domain (see Experimental procedures). The expression of the mutants fused to LexA was checked in yeast. The LexA–TolAIII mutants were screened for the lost of their ability to interact with TolB but not with AT1 in the two-hybrid system. In this screening, three mutants were identified and sequenced. The first mutant carries the I368T substitution and a point mutation creating a stop codon, which cause premature termination of the translated TolAIII after an alanine residue at position 394. The second mutant presents the A315V S334A substitutions and a point mutation creating a stop codon after a lysine residue at position 403. The third mutant presents the double mutation Y354H V409E. As, in contrast to the other mutants, the Y354H V409E mutant presents a full length TolAIII, we focused our attention on it. We found that the double mutation Y354H V409E in TolAIII abolishes the TolAIII/TolB interaction without affecting the TolAIII/AT1 interaction (Fig. 4A). In a second step, these mutations were transferred in the plasmid pTolA, allowing expression of TolA in E. coli (Cascales et al., 2000). Plasmids expressing wild-type TolA and the mutated TolA were used to transform a tolA strain, JC7782 (J. C. Lazzaroni). pTolA complements perfectly the tolA phenotype as it restores the growth of JC7782 in presence of sodium dodecyl sulphate (SDS) (Fig. 4B) and as it restores the complete sensitivity of JC7782 to colicin A (Fig. 4C). We found that cells producing TolA Y354H V409E mutant exhibited a tol phenotype. They were sensitive to SDS and presented a higher tolerance to colicin A. JC7782 tolA pTolA Y354H V409E cannot grow in Luria–Bertani (LB) agar medium supplemented with 1% SDS and, in liquid medium, this strain is two times more sensitive to SDS when SDS concentration is higher than 1% compared with JC7782 tolA pTolA. Futhermore, JC7782 tolA pTolA Y354H V409E is more tolerant to colicin A lethal activity than JC7782 tolA expressing wild-type TolA. In the plate assay, a dilution of 10–3 is unable to kill JC7782 tolA pTolA Y354H V409E, whereas a lysis halo is formed with JC7782 tolA pTolA. As well, in liquid assay, eight times more colicin A is required to kill 50% of JC7782 tolA pTolA Y354H V409E compared with the quantity of colicin A required to kill 50% of JC7782 tolA pTolA. Moreover, we tested the sensitivity of JC7782 tolA pTolA Y354H V409E to other group A colicins. We found that the double mutation that abolished the TolA/TolB interaction also affects the biological activity of colicin E3, but had no effect on the biological activity of colicin E1 (data not shown).
Finally, we tested the interaction of TolA Y354H V409E with Pal in E. coli, using a technique of coimmunoprecipitation after cross-linking the cells with formaldehyde (Cascales et al., 2000). The results revealed that the in vivo interaction of TolA Y354H V409E with Pal was reduced but still existed (not shown).
Our results indicate that the interaction between TolA and TolB is important for the maintenance of cell envelope integrity. Moreover, our results indicate an involvement of the TolA/TolB interaction in the uptake of colicin A.
In this study, we have used the yeast two-hybrid tech-nology to investigate the interactions between Tol/Pal proteins and between Tol/Pal proteins and colicin A. This technique confirms the TolR dimerization and the TolA/AT1, TolB/AT1 and TolB/Pal interactions (Bénédetti et al., 1991a; Bouveret et al., 1995; 1998; Journet et al., 1999; Ray et al., 2000). Moreover, we confirm that the C-terminal domain of TolB (TolB-D2) is responsible for the interaction of the protein with Pal (data not shown) as previously shown using a genetic suppressor approach (Ray et al., 2000). It can be noted that no interaction was detected between TolAII-III and TolRII-III, confirming that the TolA/TolR interaction must occur between the transmembrane domains of these proteins (Dérouiche et al., 1995; Germon et al., 1998; Journet et al., 1999). Surprisingly, no interaction was detected between TolRII-III and AT1 and between Pal and TolAIII, with which other techniques succeeded (Cascales et al., 2000; Journet et al., 2001). Concerning the AT1/TolRII–III interaction, it has been shown by Journet and colleagues (Journet et al., 2001) that the region of interaction of AT1 with TolR is localized in the very end of the N-terminal AT1 domain. This region might be hidden by the domain fused in the N-terminal of AT1 preventing the TolRII–III/AT1 interaction. Nothing is known concerning the sites of interaction between TolAIII and Pal but it could be suggested that the fusion affects the conformation of the Pal interaction domain. This has not been further investigated but the insertion of a linker at the N-terminus of AT1 and Pal could provide results for this possibility.
We observed that YbgF interacts with TolAII-III. We know that ybgF is part of the tol/pal operon (Duan et al., 2001) but ybgF inactivation induces no obvious tol phenotype (Vianney et al., 1996). Thus, its participation in the Tol/Pal system has long been a question. Through its interaction with TolAII-III, we can say now that YbgF is fully related to the Tol/Pal system. Futhermore, we have clearly demonstrated that the C-terminal domain of TolA is not essential for this interaction in contrast with the TolAII domain, which is necessary and sufficient to mediate the interaction with YbgF. It is of interest to note that the existence of a coiled-coil region in the TolAII domain has been proposed (Dérouiche et al., 1999) and coiled-coil research in YbgF indicated a high probability of formation of such a structure (between residues 50 and 100, data not shown). Thus a coiled-coil interaction may be involved in the interaction between TolA and YbgF. In this case, TolAII might have a specific function and not only act as a tether allowing TolAIII to interact with components present in the periplasm. Our quantitative β-galactosidase assays indicate a lower strength of interaction for TolAII compared with TolAII-III. If the C-terminal part is not necessary for interaction, it seems yet that this part might strengthen the interaction, prob-ably by stabilizing the TolAII structure.
The most striking finding is the existence of an interaction between TolA and TolB. Indeed, although it was clear that two complexes existed (TolAQR associated to the inner membrane and TolB/Pal associated to the outer membrane), it is only very recently that cross-talk between these two complexes has been detected (Cascales et al., 2000). We present here the second direct evidence for an interaction between the inner membrane and outer membrane complexes. The interaction between TolA and TolB is a strong interaction as measured by transcriptional activation of the reporter lacZ gene, as strong as the well known interaction between TolB and Pal (data not shown). It occurs through the C-terminal domain of TolA. Futhermore, we have clearly demonstrated that the D2 domain of TolB is not essential for protein– protein interaction in contrast with the TolB-D1 domain that is necessary and sufficient to mediate the interaction with TolAIII. It is noteworthy that the N-terminal D1 domain is more conserved than the C-terminal D2 domain (Abergel et al., 1999), probably because it is involved in a specific function possibly via its interaction with TolA. Our quantitative β-galactosidase assays indicate a lower strength of interaction for TolB-D1 compared with that of full-length TolB. It is known that hydrophobic interactions and hydrogen bonds exist between the D1 and the D2 domains and that some conserved positions in the D1 domain are at the centre of a number of interactions between the β-propeller, the linker and the D1 domains (Abergel et al., 1999). Thus, the proper folding of TolB-D1 domain is probably conditioned by its interaction with the D2 domain, which seems to have a stabilizing effect in the TolB structure. If the C-terminal part is not necessary for interaction, it seems yet that this part might strengthen the interaction by maintaining a TolB-D1 conformation needed for a strong interaction. Moreover, we cannot rule out the possibility that some residues in TolB–D2 interact with TolAIII. We confirmed the TolAIII/ TolB interaction observed in vivo using the yeast two-hybrid system by in vitro chemical cross-linking experiments. The demonstration of a physical complex between TolAIII and TolB, and between TolAIII and TolB-D1, using chemical cross-linking provides rigorous evidence for these interactions.
These results confirm the existence of an important network of interactions between the Tol/Pal proteins. Among the different partners of this system, TolA and TolB are involved in multiple interactions with seven different partners. Consequently, we can assume that the interaction of TolA with TolB might represent a crucial interaction for the organisation and the function of the Tol system. Previous membrane fractionation experiments have shown that TolB protein was mainly recovered with the other Tol proteins in contact sites between the inner and outer membranes and in the inner membrane (Guihard et al., 1994; Isnard et al., 1994; Lazdunski et al., 1998; 2000). Our findings suggest that the association of TolB with the inner membrane fraction is due to its interaction with TolA. We can imagine that TolB links the two membranes via its simultaneous interactions with TolAIII and Pal. It may be physically possible for the D2 domain of TolB to interact with Pal whereas the D1 domain of TolB interacts with the C-terminal domain of TolA and the N-terminal domain of TolA interacts with TolQ and TolR. Thus, TolQRAB and Pal are probably part of an oligomeric structure that participates in the association between the inner and the outer membranes through the interaction between TolA and TolB and the p.m.f.-dependent interaction between TolA and Pal (Cascales et al., 2000).
Colicin A is a pore-forming bacteriocin that depends upon the Tol proteins to be transported from its receptor at the outer membrane surface to its target, the inner membrane. Similarly to the Tol proteins, colicin A is mainly recovered in the inner membrane and contact sites between the inner and outer membranes (Guihard et al., 1994). Cell fractionation experiments indicated that when sensitive cells were treated with colicin A, the number of Tol proteins localized at contact sites increased, suggesting that the colicin stabilizes the Tol complex at the contact sites upon translocation (Guihard et al., 1994). In this work, we find that no interaction occurs between the pore-forming domain of colicin A and Tol proteins. This is in good agreement with previous data showing that the cytotoxic domain of colicin A in the cytoplasm of E. coli was able to assemble in the inner membrane and to kill the bacterial strains by a mechanism independent of the tol genes (Espesset et al., 1994). We observed that the N-terminal domain of colicin A (AT1) interacts with TolB and TolAIII, confirming that colicin A interacts with the Tol proteins upon translocation in vivo. In our two-hybrid screen, no interaction of colicin A was detected neither with TolB-D1 nor with TolB-D2. These results contrast with the results of Carr and co-workers who have shown that the translocation domain of colicin E9 interacts with TolB-D2 (Carr et al., 2000), and probably reflect the difference between the translocation of the pore-forming colicin A which acts on the surface of the cytoplasmic membrane and the DNase E9 colicin, which has to reach the cytoplasm. The crucial role played by the interaction between TolA and TolB in colicin A uptake was shown using a TolA mutant Y354H V409E. This mutant, similarly to tolA and tolB mutants, presented a higher tolerance to colicin A, an envelope integrity defect, and the mutated protein was affected in its ability to interact with TolB but still retained its capacity to interact with AT1 and, to a lower extent, with Pal. Because, unlike TolA and TolB, Pal is not involved in the entry of colicin A (Lazzaroni and Portalier, 1992; Ray et al., 2000), we can assume that the interaction between TolA and TolB is directly involved in colicin A uptake. Moreover, the biological role for the TolAIII/TolB interaction was confirmed by the interesting finding that the TolA mutant Y354H V409E presented a higher tolerance to colicin E3, which requires TolB to be translocated (Bouveret et al., 1997), but remained fully sensitive to colicin E1, which requires the outer membrane protein TolC but not TolB (Nagel and Luria, 1967). Taken together, these observations are pointing out the physiological relevance of TolA/TolB interaction and strongly suggest that the Tol system function requires this interaction. We do not know if the Tol system function needs the formation of a stable protein complex of a precise stoichiometry between the different partners or results from a subtle balance between different interactions. It is tempting to hypothesize that all these interactions occur simultaneously, maintaining colicin A in an unfolded conformation but in a transient way as no trimeric TolA/TolB/Pal and TolA/TolQ/TolR complexes have ever been detected. This is consistent with the observation that TolB could exist, at a given time, free in the periplasm or associated with the membranes (Isnard et al., 1994).
Neither the recently determined 3D-structure of TolB (Abergel et al., 1999; Carr et al., 2000) nor previous studies of TolB in vivo and in vitro detected a dimerization (Bouveret et al., 1995; 1998; Clavel et al., 1998). However, we reproducibly detected a TolB/TolB interaction using the yeast two-hybrid system and in vitro cross-linking experiments. Similar results were obtained using the D1 domain of TolB, suggesting that TolB has the intrinsic capacity to dimerize and that the N-terminal domain promotes TolB dimerization. Moreover, the presence of multiple bands suggests that TolB-D1 multimerizes. However, we have no direct proof that these cross-linked products are not due to an anomalous folding of the TolB-D1 protein. Much more information is needed before concluding about the real physiological significance of this interaction.
In conclusion, the yeast two-hybrid technology proved to be a powerful tool to further investigate the interactions between components of the Tol/Pal system. It has revealed three new interactions between various components of this system. We demonstrate for the first time that TolB interacts with the C-terminal part of TolA and itself, that in both cases the interaction involved the N-terminal part of the protein and that TolAII interacts with YbgF. Furthermore, we provide evidence supporting the fact that the interaction between TolA and TolB linking the inner and outer membrane complexes is of a crucial importance for the functioning of the whole system. To further characterize the interaction between TolA and TolB and determine the exact relationships between these two components, we are now turning to mutagenesis experiments and nuclear magnetic resonance (NMR) studies. Preliminary NMR results confirm the interaction between TolAIII and TolB (C. Deprez, personal communication). However, much more work is needed to determine the exact relationships between these two proteins. We believe that the usefulness of the yeast two-hybrid system for testing various Tol/Pal and colicin mutants will facilitate further in vivo studies allowing to reach a full characterization and understanding of the structure–function relationships in this bacterial machinery.
Restriction enzymes and T4 DNA ligase were obtained from New England Biolabs. PCRs used the Expand High fidelity PCR System (Boehringer Mannheim). Sequencing reactions were performed with the thermo sequenase radiolabelled terminator cycle sequencing kit from Amersham Life Science. Western blots were performed on Immobilon (Millipore). LexA, TolB and TolAIII antibodies were a generous gift from A. Rigal and R. Lloubès (CNRS, Marseille). HA antibody (3F10) was from Boehringer Mannheim. Alkaline phosphatase-conjugated secondary antibodies were purchased from Jackson Immuno Research. Xgal (5-bromo-4-chloro-3-indolylβ-D-galactopyranoside) was obtained from Euromedex. Formaldehyde and Ethylen Glycol-bis Succinimidylsuccinate (EGS) were purchased from Pierce. ONPG (o-nitrophenyl-β-D-galactoside) was purchased from Sigma-Aldrich.
All bacterial cells were grown in Luria–Bertani (LB) medium or on LB plates. The antibiotics and concentration were used as follows: ampicillin, 100 μg ml–1 and chloramphenicol, 25 μg ml–1. Yeast media were purchased from Difco and Bio 101. Yeast strains were grown in appropriate minimal drop-out media (CM) in which 2% glucose or 2% galactose were added.
Plasmids and constructs
The parent vectors pEG-202 and pMW-104 for the yeast two-hybrid analysis were kindly provided by the laboratory of Roger Brent (Massachussetts General Hospital, Boston, MA) (Zervos et al., 1993). All the plasmids, derived from pEG-202 and pMW104, resulted from the cloning of PCR generated EcoRI–XhoI fragments encoding the different domains tested: TolAII-III (amino acids 66–421), TolAII-IIIΔ (corresponding to a deletion in TolAII-III between amino acids 337–376; Bénédetti et al., 1991b), TolAIII (amino acids 301–421), TolAIIIΔ (amino acids 301–421 including the deletion between 337 and 376), TolB, TolB-D1 (amino acids 1–144), TolB-D2 (amino acids 120–409), AT1 (the translocation domain of colicin A, amino acids 1–178), C-ter colA (the cytotoxic domain of colicin A, starting from the amino acid 339), YbgC, YbgF, TolQc (amino acids 36–132), TolRII-III (amino acids 45–143) and Pal. The plasmids are listed in Table 2. The identity of DNA constructs was confirmed by sequencing.
The yeast two-hybrid system developed in the laboratory of Roger Brent was employed (Brent, 1993). The yeast strain consists of EGY48 (Matα, ura3, trp1, his3, 6LexAop-LEU2). This strain contains six LexA binding sites replacing the upstream activating sequence of the LEU2 gene; therefore, growth on leucine-deficient media may serve as an indicator of interaction between the two partners. The various pEG- and pMW- constructs were transformed into EGY48 along with a reporter plasmid, pSH18-34. pSH18-34 contains eight LexA binding sites upstream of the β-galactosidase gene (lacZ).
Yeast strains were transformed by the chemical PEG/lithium acetate method according to the Clontech manual. Plates were incubated for 72 h at 30°C until colonies appeared. Positive transformants were grown for 2 days on SD and SGal media.
Purification of TolB-D1
A 1-L culture of MC4100 pTTolB-D1 cells was induced at OD600 = 0.6 with 0.01% of arabinose and incubated at 22°C overnight. The cells were harvested and resuspended in 6 ml of 10 mM Tris buffer, pH 8, containing 20 mM NaCl and 5 mM imidazol. Protease inhibitors (1 mM PMSF, 1 μg ml–1 of pepstatin, 1 μg ml–1 of leupeptine and 1 μg ml–1 of aprotinin) were added and the cells were lysed by French Press. After a 30 min centrifugation at 105 000 g, the cleared lysate obtained was applied to a 8 ml column of TALON metal affinity resin (Clontech). The column was then washed with 10 mM Tris buffer, pH 8, containing 20 mM NaCl and 5 mM imidazol. The polyhistidine-tagged protein was eluted with an imidazol gradient from 5 mM to 1 M. The eluates were collected and the presence of TolB-D1 and its purity in the different fractions was assessed by SDS–PAGE and Western blotting with AbTolB antiserum (Rigal et al., 1997). The fractions corresponding to the elutions with 300 to 380 mM imidazol contained at least 90% pure TolB-D1 and were therefore pooled. The resulting preparation was concentrated and dialysed against 10 mM phosphate buffer, pH 6.8, containing 5 mM NaCl. The concentration of the final TolB-D1 preparation was ≈ 1.75 mg ml–1.
A formaldehyde and an EGS cross-linking assay were performed. TolAIII (317–420 amino acids) (10 μM) was mixed with 10 μM of TolB in a 10 μl volume, for 15 min at room temperature. Then samples were incubated for 20 min at room temperature with 1% formaldehyde, or for 30 min at room temperature with 1 mM EGS. Cross-linking with formaldehyde was stopped by the addition of 10 mM Tris, pH 6.8 and cross-linking with EGS by the addition of 50 mM Tris, pH 7.2, and a 15 min incubation at room temperature. Sample buffer was added, and the samples were heated at 37°C for 15 min, or heated at 96°C for 15 min to destroy the cross-links with formaldehyde. Finally, the samples were analysed using SDS–polyacrylamide gel electrophoresis (SDS–PAGE) and Western blotting.
pMW-B-D1 and pEG-AII were constructed using the QuikChange Site-Directed Mutagenesis Kit (Stratagene). To construct pMW-B-D1, a stop codon was introduced at position 145 of the TolB sequence, using pMW-B as a DNA matrix. To construct pEG-AII, a stop codon was introduced at the end of domain II of TolA at position 322, using pEG-AII-III as a DNA matrix. Each construct was then digested with EcoRI–XhoI and was ligated between the same sites in pEG-202 or pMW-104 to create pEG-B-D1 and pMW-TolAII. Also, we introduced the Y354H and the V409E mutations in TolA in pTolA (Cascales et al., 2000).
Random mutagenesis by PCR
Polymerase chain reaction (PCR) fragments encoding TolB-D1 (amino acids 1–144) were amplified from pMW-B vector. To generate a mutational library, the PCR was performed following the guidelines of Fromant and colleagues (Fromant et al., 1995). The incubation mixture (25 μl) contained 10 mM Tris-HCl (pH 8.7 at 25°C), 50 mM KCl, 5 μg ml–1 of bovine serum albumin (BSA), 0.5 μM of each of the two primers, 600 pM of template DNA and 1.4 units of Expand DNA polymerase. The concentration of the non-forcing and forcing nucleotide dNTPs were 0.2 mM and 3.4 mM respectively. The MgCl2 and MnCl2 concentrations in the assay were 9.5 and 0.5 mM respectively. The reaction mixtures were submitted to 16 cycles of amplification under the following conditions: 91°C for 1 min (94°C for the first cycle), 51°C for 1 min, and 68°C for 1 min (5 min for the last cycle). Four reaction mixtures varying in the nature of the forcing nucleotide were used. The mutagenic PCR products were combined, digested by EcoRI and XhoI endonucleases, and ligated into the yeast two-hybrid vectors pMW-104 and pEG-202. Ligation products were transformed into E. coli strain C600. The expression of the various pEG-B-D1 and pMW-B-D1 mutants was checked in yeast, and the mutants were screened for interaction with TolAIII in the two-hybrid system. The TolB-D1 mutants that lost the ability to interact with TolAIII were sequenced.
In a similar way, mutagenic PCR fragment encoding TolAIII (amino acids 301–420) was amplified from pBPAII (Journet et al., 1999) vector. The mutagenic PCR products were cloned in pMW-104 and pEG-202. Ligation products were transformed into E. coli strain C600. The expression of the various pEG-AIII and pMW-AIII mutants were checked in yeast and the mutants was screened for interaction with TolB in the two-hybrid system. The TolAIII mutants that lost the ability to interact with TolB were sequenced.
Western blotting detection
Plasmids expressing fusion proteins were transformed singly into the EGY48 strain of yeast and positive cells were grown in SGal media -Trp for B42 fusions, and -His for LexA fusions, for 2 days. The equivalent of 2 × 107 cells was subjected to Western blot analysis. Proteins resolved on 12% acrylamide gels by SDS–PAGE were electrotransferred to nitrocellulose membranes. To detect the haemagglutinin epitope-tagged or LexA epitope–tagged proteins, the monoclonal antibody 3F10 or the polyclonal LexA, were used respectively.
For relative quantification of protein–protein interactions, β-galactosidase assays were performed. Yeast strains co-transformed by pEG- and pMW- constructs were grown in supplemented minimal galactose medium lacking uracil, tryptophan and histidine, in a shaking incubator, at 30°C for 72 h. They were then spun down for 2 min at 5000 r.p.m., washed with water and resuspended in Z buffer (100 mM NaPO4 at pH 7.0, 10 mM KCl, 1 mM Mg(SO4)2, 38 mM β-mercaptoethanol). Cell density was determined by measuring the OD600 for the washed cells. Then, 40 μl of 0.1% SDS was added to 500 μl of cell suspension and mixed vigorously for 30 s, followed by the addition of 60 μl of chloroform with repeated vortexing. The enzymatic reaction was started by the addition of 100 μl of ONPG solution (4 mg ml–1 in Z buffer) and the reaction was incubated at 30°C until a yellow colour began to appear, or for 15 min, after which 0.5 ml of 1.0 M Na2CO3 was added to terminate the reaction. The samples were centrifuged at top speed for 2 min and the absorbance at 420 nm was determined. The activity of β-galactosidase in Miller units was calculated with the formula units = A420 × 1000/A600 × volume × reaction time (Miller et al., 1992). Each condition was performed in triplicate.
Sensitivity tests to SDS
Plate assay. JC7782 cells harbouring (or not) pTolA and pTolA Y354H V309E were grown in LB medium at 37°C to OD600 = 0.8. Then, they were streaked independently on a LB agar medium supplemented with 1% SDS for one night at 37°C. Cells are considered sensitive when they form a translucide streak whereas resistant cells are opaque.
Liquid culture assay. JC7782 cells harbouring (or not) pTolA and pTolA Y354H V309E were grown in LB medium at 37°C to OD600 = 0.5. Then, they were 100-fold diluted in LB supplemented with various concentrations of SDS (0.1% to 3% SDS) and grown for 180 min more. As a control, water was used instead of SDS dilution. The percentage of surviving cells was estimated from the turbidity ratio of the SDS-treated and the control samples. The experiment has been performed four times.
Sensitivity tests to colicin A
Plate assay. Dilutions from 10–1 to 10–3 of a 1 mg ml–1 of stock solution of colicin A were assayed by spotting 5 μl of each dilution on freshly seeded lawns of the JC7782 tolA (J.C. Lazzaroni) strain harbouring (or not) the pTolA parental plasmid (Cascales et al., 2000) or the pTolA Y354H V309H mutated plasmid.
Liquid culture assay. JC7782 cells harbouring (or not) pTolA and pTolA Y354H V309E were grown in LB medium at 37°C to OD600 = 0.2. Then, 100 μl of colicin A diluted in Tris buffer (10 mM Tris, pH 7.2) was added to 1.25 ml of this culture and incubated for 30 min at 37°C with shaking. As a control, Tris buffer was used instead of colicin dilution. Subsequently, 100 μl of SDS (7.5 mg ml–1) was added and turbidity at 600 nm was measured after 10 min of shaking at 37°C. The percentage of surviving cells was estimated from the turbidity ratio of the colicin-treated and the control samples. The experiment has been performed four times.
We are grateful to A. Rigal for his gift of purified TolB, M. Gavioli for her gift of purified TolAIII and B. Py for her gift of the yeast strain EGY48. Our thanks also go to D. Duché, E. Bouveret, R. Lloubès and E. Cascales for reading the manuscript. We would like to gratefully thank V. Géli, who initially proposed this project, for helpful advice, discussions and critical reading of the manuscript.