Channel Domain of Colicin A Modifies the Dimeric Organization of Its Immunity Protein

Abstract

Proteins conferring immunity against pore-forming colicins are localized in the Escherichia coli inner membrane. Their protective effects are mediated by direct interaction with the C-terminal domain of their cognate colicins. Cai, the immunity protein protecting E. coli against colicin A, contains four cysteine residues. We report cysteine cross-linking experiments showing that Cai forms homodimers. Cai contains four transmembrane segments (TMSs), and dimerization occurs via the third TMS. Furthermore, we observe the formation of intramolecular disulfide bonds that connect TMS2 with either TMS1 or TMS3. Co-expression of Cai with its target, the colicin A pore-forming domain (pfColA), in the inner membrane prevents the formation of intermolecular and intramolecular disulfide bonds, indicating that pfColA interacts with the dimer of Cai and modifies its conformation. Finally, we show that when Cai is locked by disulfide bonds, it is no longer able to protect cells against exogenous added colicin A.

Introduction

Pore-forming colicins are plasmid-encoded bacteriocins, synthesized by Enterobacteriacae, that are lethal to other related strains. Like many toxins, colicins are organized into structural domains that perform different functions: the N-terminal and the central domains are involved in the transport through the Escherichia coli envelope, and the C-terminal domain carries the toxic activity of the protein (1).

The three-dimensional structures of the soluble form of the pore-forming domains of colicins A, E1, Ia, B, and N share the same general architecture: a bundle of eight amphipathic α-helices surrounding two hydrophobic α-helices (H8+H9), completely buried within the protein (2–6). However, crystal structures do not reveal the structure of the channel in the membrane, which remains uncharacterized. The hydrophobic helical hairpin may be the primary attachment region for channel formation and channel opening and closing require insertion and extrusion of an amphipathic segment, which has not been precisely defined (7, 8).

Each colicin plasmid also encodes a specific immunity protein that protects the producing cell against the cytotoxic activity of its colicin. Immunity proteins are integral inner membrane proteins. They are classified into two groups according to sequence similarities: the A type (immunity proteins to colicins A, B, N, and U) and the E1 type (immunity proteins to colicins E1, 5, K, 10, Ia, and Ib) (9–11). The colicin A immunity protein (Cai)² has four transmembrane segments (TMSs), and its N and C termini are in the cytoplasm (Fig. 1), whereas the immunity protein to colicin E1 (Cei) crosses the cytoplasmic membrane three times, with the N terminus located in the cytoplasm and the C terminus in the periplasm (10, 11).

FIGURE 1.

Model of Cai in the cytoplasmic membrane of Escherichia coli. H1, H2, H3, and H4 denote the transmembrane α-helices. The positions of cysteine residues are shown with black dots and labeled with numbers.

The immunity proteins to pore-forming colicins diffuse laterally in the membrane and interact with helices of the pore-forming domain of their cognate colicin just prior channel opening (12–15). Genetic studies of A-type colicins indicate that the main determinant recognized by the immunity protein is the hydrophobic helical hairpin (12, 16). Similar studies with an E1-type colicin suggested that inactivation of the toxin is due to interaction between the voltage-gated regions of colicin and the TMSs of the immunity proteins (14, 15). The first biochemical evidence of pore-forming colicin physically interacting with its immunity protein was reported by Espesset et al. (1996), who showed that a colicin A pore-forming domain, fused to a prokaryotic signal peptide (sp-pfColA), could be co-immunoprecipitated with its cognate epitope-tagged immunity protein (EpCai) (17). Using the same co-immunoprecipitation procedure, Nardi et al. demonstrated that the smallest fragment of colicin A recognized by Cai was the hydrophobic helical hairpin (18).

However, little is known about the tertiary structure of immunity proteins within the membrane. For example, there have been no investigations of whether immunity proteins are in monomeric or oligomeric states. Only limited and preliminary data are available concerning the arrangement of helices in the membrane, and the residues in close contact between TMSs have not been identified. Here, we describe the use of site-directed cysteine cross-linking to gain insight into the interaction surfaces of Cai at the structural level. Copper (II) orthophenanthroline (Cu-OP), a commonly used cross-linking reagent, was employed to catalyze the formation of disulfide bonds between SH groups. EpCai has four cysteine residues, all located in its TMSs. We show that EpCai forms a covalent dimer in the membrane upon incubation with Cu-OP, and we identified the cysteine residue involved in EpCai dimerization. In addition to dimer formation, we also detected intramolecular disulfide bonds that connect TMS1 to TMS2 and TMS2 to TMS3. Interestingly, intra- and intermolecular disulfide bonds formation was abolished when sp-pfColA was co-expressed. Moreover, if EpCai is oxidized with Cu-OP treatment, it is no longer able to protect cells against exogenously added colicin A. These various findings indicate that in vivo EpCai forms a dimer in the membrane and changes conformation upon interaction with the colicin A pore-forming domain.

EXPERIMENTAL PROCEDURES

Strain, Plasmids, and Growth Conditions

E. coli C600 was used as the recipient strain for all cloning procedures. Plasmid pImTC encodes the wild-type Cai under the control of its own promoter (19). Plasmid pEpCaiCm encodes the epitope-tagged immunity protein (EpCai) (19). pVL1 encodes the wild-type Cai with the 178 first amino acids of ColA fused to its N terminus (20). pCT1 encodes the pore-forming domain of ColA fused to the PelB sequence signal (sp-pfColA) (21). pCT3 encodes the B3 mutant form of sp-pfColA (sp-pfColA_B3) (21). pBAD/HisC contains the araBAD promoter from E. coli and was used to insert the gene encoding sp-pfColA or sp-pfColAB3. Routinely, cells were grown aerobically at 37 °C or 30 °C, in LB medium supplemented with ampicillin (100 μg/ml), kanamycin (50 μg/ml), or chloramphenicol (40 μg/ml).

Constructs

QuikChange mutagenesis PCR with complimentary pairs of oligonucleotides was used to introduce site-directed substitutions. Genes encoding sp-pfColA and sp-pfColAB3 were amplified by PCR using appropriate 5′ and 3′ primers. The 5′ primers contained the DNA sequence (30 bp) of pBAD/HisC immediately upstream from the ATG start codon and the DNA sequence encoding the first eight amino acids of the signal peptide of sp-pfColA or sp-pfColAB3. The 3′ primers correspond to the DNA sequence encoding of the N-terminal eight amino acids of sp-pfColA or sp-pfColAB3 and contained the DNA sequence (30 bp) of pBAD/HisC immediately upstream from the ATG. Once amplified, the DNA fragments encoding sp-pfColA or sp-pfColAB3 were used as megaprimers and inserted into pBAD/HisC by a second-stage PCR that amplified the whole plasmid.

Colicin Assay

Colicin activity was assessed by the formation of halos on a lawn of the target strain, as described previously (22). The data are reported as the highest dilution of colicin solution that still inhibits cell growth (Table 1). Strains sensitive or immune to ColA were spread on LB medium agar plates containing or not containing Cu-OP (200 μm).

TABLE 1.

Activity of Cai and its mutant forms

Immunity protein and mutant forms	ColA dilution resulting in clear zonea
No immunity	10−4
WT	(i)b
C31S	(i)
C74S	(i)c
C107S	(i)
C122S	(i)
L109C	(i)
V113C	(i)
Y119C	(i)c
C31S/C74S	10−3c
C31S/C107S	(i)
C31S/C122S	10−2c
C74S/C107S	10−4c
C74S/C122S	10−4c
C107S/C122S	10−1c
S106C/C107S/C122S	10−3c
C107S/L108C/C122S	10−3c
C107S/L109/ C122S	10−3c
C107S/S110/C122S	10−4c
C107S/P111C/C122S	10−4c
C107S/L112C/C122S	10−4c
C107S/V113C/C122S	10−3c
C107S/Y114C/C122S	10−4c
C107S/A115C/C122S	10−3c
C107S/A116C/C122S	10−4c
C107S/I117C/C122S	10−3c
C107S/V118/C122S	10−4c
C107S/Y119C/C122S	10−3c
C107S/L120C/C122S	10−4c
C107S/F121C/C122S	10−3c
C107S/F123C/C122S	10−2c
C31S/C74S/C107S	10−4c
C31S/C107S/C122S	10−4c
C74S/C107S/C122S	10−4c
C31S/C74S/C107S/C122S	10−4c

^a The last dilutions resulting in clear zones of growth inhibition are listed.

^b Immune phenotype for which no zone clearing with undiluted colicin was detected.

^c Results obtained at 30 °C. At 37 °C these mutated Cai were inactive. The quadruple mutant was not produced even at 30 °C.

In Vivo Disulfide Bond Formation and Immunodetection

Aliquots of 8 × 10⁸ exponentially growing cells (A₆₀₀ = 0.6) were pelleted and resuspended in 1 ml of 10 mm sodium phosphate buffer (NaPi, pH 6.8), and then treated with 1 mm EDTA (pH 8) and 5 mm N-ethylmaleimide (NEM; Sigma) to block reduced thiols groups. The cells were treated, as appropriate, with the indicated concentration of the oxidative catalyst Cu-OP (Sigma) prior to treatment with EDTA and NEM. The cells were subsequently pelleted and resuspended in Laemmli loading buffer in the presence or absence of β-mercaptoethanol (β-me). In vivo formaldehyde cross-linking was performed as described previously (23). Proteins were separated by 15% SDS-PAGE, and subjected to Western blotting with the primary anti-1C11 monoclonal antibody and anti-mouse second antibodies labeled with Alexa Fluor 488 dye (Molecular Probes); the banding pattern was analyzed using the Odyssey infrared imaging system (LI-COR Biosciences).

RESULTS

Cross-linking of EpCai by Cu-OP

We investigated the effect of Cu-OP oxidation on the cysteine thiol groups of EpCai. Cells producing EpCai were incubated with various concentrations of Cu-OP at 20 °C for 15 min. The reaction was stopped by the addition of EDTA to trap free and complexed Cu²⁺, and of NEM to block the remaining free cysteine residues. Samples were separated by SDS-PAGE, and EpCai was detected by immunoblotting with monoclonal antibody (mAb) 1C11 directed against the epitope tag of EpCai (Fig. 2). Increasing concentrations of Cu-OP decreased the intensity of the band corresponding to the monomeric form of EpCai (apparent molecular mass of 18 kDa) and increased the intensity of the band corresponding to the dimeric form (36 kDa) (Fig. 2A). After treatment with β-me, the 36-kDa band disappeared, with all EpCai migrating in the 18-kDa band (Fig. 2A). These results strongly suggest that the 36-kDa band corresponded to a dimer of EpCai, stabilized by an intermolecular disulfide bond. This dimer will hereafter be referred to as [EpCai]_2. To confirm the existence of the dimer, another cross-linking reagent, formaldehyde (FA), was used. Cells producing EpCai were treated with 1% FA for 15 min, and the proteins were separated by SDS-PAGE. Both EpCai and [EpCai]₂ were detected in FA-treated samples, but only the band corresponding to the monomeric EpCai was visible in the absence of FA treatment and in FA-treated samples that were subsequently heated to break any cross-links (Fig. 2B). Although these results suggest that EpCai exists as a dimer in the membrane, the possibility that EpCai interacts with another protein that has a similar apparent molecular mass to form an heterodimer cannot be formally excluded. To demonstrate that the observed 36-kDa band indeed corresponds to a homodimer of EpCai, we used a larger variant of Cai, called VL1 (20). In absence of Cu-OP, monomeric VL1 migrated on SDS-PAGE as a single band of 30 kDa apparent molecular mass. Following treatment with Cu-OP, a second band of 60–65-kDa apparent molecular mass was observed, undoubtedly a homodimer of VL1 because cross-linking between VL1 and a hypothetical protein of an apparent molecular mass of 18 kDa would not migrate to this position (Fig. 2C). These results clearly indicate that Cai forms homodimers in the E. coli inner membrane. Interestingly, following treatment with 20 μm Cu-OP, two other bands that migrated close to EpCai or [EpCai]₂ were detected (Fig. 2A). These bands may correspond to EpCai monomers and dimers that react with Cu-OP to form intramolecular disulfide bonds. Indeed, these bands disappeared in the presence of a reducing agent, and only the band corresponding to the monomeric form of EpCai was observed. Preincubation of cells with NEM prior to the addition of Cu-OP prevented intra- and intermolecular disulfide bond formation (Fig. 2A). We used mutagenesis experiments to identify the cysteine residues involved in these intra- and intermolecular disulfide bonds.

FIGURE 2.

EpCai forms homodimers. A, effect of Cu-OP concentration on the amount of EpCai dimer. For all the figures, samples were resolved by 15% SDS-PAGE, and anti-1C11 mAb was used for immunodetection. Aliquots of 0.2 × 10⁸ cells of the strain producing EpCai, incubated with or without one of a series of concentrations of Cu-OP, were boiled in the absence or presence of the reducing agent β-me and preincubated or not with NEM. Positions of EpCai and EpCai dimers are indicated on the right. Positions of EpCai monomers and dimers with intramolecular disulfide bond(s) are indicated as EpCai_s-s and [EpCai_s-s]₂, respectively. Molecular mass markers are indicated on the left. B, EpCai cross-linking with FA. Aliquots of 0.2 × 10⁸ cells of the strain producing EpCai were treated (FA) or not treated (−) with formaldehyde. ΔFA is similar to FA, except that samples were heated to 96 °C to break the cross-links. C, EpCai and VL1 dimer formation. Aliquots of 0.2 × 10⁸ cells of the strain producing EpCai or VL1, incubated with or without Cu-OP (400 μm), were boiled with or without β-me. Positions of EpCai, EpCai dimers, VL1, and VL1 dimers are indicated on the right. Molecular masses are indicated on the left.

Cys¹⁰⁷-Cys¹⁰⁷ Disulfide Bond Mediates EpCai Dimerization

EpCai possesses cysteine residues at positions 31, 74, 107, and 122. All of these four cysteine residues are located in TMSs. Cys³¹ is located in TMS1, Cys⁷⁴ in TMS2, and Cys¹⁰⁷ and Cys¹²² in TMS3 (Fig. 1). To identify the cysteine residue(s) involved in EpCai dimerization following incubation with Cu-OP, we constructed four mutated EpCai proteins, each with a different single cysteine replaced with serine. EpCai_C31S, EpCai_C107S, and EpCai_C122S were produced in amounts similar to the wild-type, but EpCai_C74S could not be detected by immunoblotting with the monoclonal antibody 1C11 (Fig. 3A). Furthermore, EpCai_C31S, EpCai_C107S, and EpCai_C122S each conferred the ColA-insensitive phenotype to an E. coli strain transformed with the corresponding plasmids, whereas the same strain transformed with the EpCai_C74S construct stayed ColA-sensitive (Table 1). We discovered that the Cys⁷⁴ to Ser transition led to a thermosensitive phenotype, and therefore we evaluated EpCai_C74S production at various temperatures (Fig. 3B): EpCai_C74S was detected at 25 °C, 30 °C, and 33 °C but not at 37 °C. Moreover, an E. coli strain transformed by the plasmid encoding EpCai_C74S was insensitive to ColA at 30 °C, demonstrating that EpCai_C74S is produced and functional at this temperature. To identify the residue involved in EpCai dimerization, cells producing the various mutated EpCai were incubated with each of a series of concentrations of Cu-OP. Doses of 20–50 μm Cu-OP effectively catalyzed the dimer formation for EpCai wild type, EpCai_C31S, EpCai_C74S, and EpCai_C122S whereas 100 μm Cu-OP was required for EpCai_C107S dimerization (Figs. 2A and 3C). These results suggest that Cys¹⁰⁷ is involved in EpCai dimerization at low Cu-OP concentrations and that if it is replaced by a serine residue, EpCai can dimerize at a high Cu-OP concentration through another cysteine residue. Each of the EpCai mutants was incubated with Cu-OP and subjected to Western blotting. A band corresponding to an oxidized monomeric form of EpCai was detected for EpCai_C31S, EpCai_C107S, and EpCai_C122S but not for EpCai_C74S. This suggests that Cys⁷⁴ is involved in intramolecular disulfide bonds with Cys³¹, Cys¹⁰⁷, or Cys¹²².

FIGURE 3.

Dimer formation of mutated EpCai. For all figures, samples were subjected to 15% SDS-PAGE and probed with anti-1C11 mAb. A, expression of mutated EpCai. 0.2 × 10⁸ cells of the indicated strains were loaded on SDS-PAGE and immunodetected. B, expression of EpCai C74S at various temperatures. Strains carrying the construct encoding EpCai C74S were grown at the indicated temperature and induced for 90 min with 100 μm isopropyl 1-thio-β-d-galactopyranoside. C, influence of Cu-OP concentration on the amount of mutated EpCai dimer. See legend to Fig. 2A for wild-type Cai. Cells were grown at 30 °C.

[EpCai]₂ Dimerizes through Its TMS3

To identify the cysteine residue involved in the dimerization of EpCai carrying the C107S mutation, we constructed the following double serine substitutions: Ser³¹/Ser¹⁰⁷, Ser⁷⁴/Ser¹⁰⁷, and Ser¹⁰⁷/Ser¹²² and tested them for dimerization upon Cu-OP incubation. Dimer formation was observed for all of the double mutants except Ser¹⁰⁷/Ser¹²² (Fig. 4A). Therefore, EpCai preferentially forms dimers through Cys¹⁰⁷ (Fig. 3C) but can also dimerize through Cys¹²² if Cys¹⁰⁷ is absent (Fig. 4A). EpCai_C107S/C122S is unable to form dimers following treatment with Cu-OP (Fig. 4A), so we tested whether this double EpCai mutant exists as a monomer or a dimer in the inner membrane. We found that dimers of EpCai_C107S/C122S were being observed following FA treatment (Fig. 4B), demonstrating that even in absence of Cys¹⁰⁷ and Cys¹²², EpCai exists as a dimer in the inner membrane.

FIGURE 4.

Cysteine residues involved in intra- and intermolecular disulfide bonds. A, aliquots of 0.2 × 10⁸ cells of the indicated strains incubated with or without Cu-OP at the indicated concentration, resolved by 15% SDS-PAGE, and probed with anti-1C11 mAb. B, EpCai_C107S/C122S cross-linking with FA. Aliquots of 0.2 × 10⁸ cells of strain producing EpCai treated (FA) or not treated (−) with formaldehyde were analyzed by 15% SDS-PAGE and immunoblotting with the anti-1C11 mAb after heating the samples at 96 °C (ΔFA) or not.

Systematic Cysteine Scanning Mutagenesis of TMS3

To determine which face of the TMS3 helix is involved in EpCai dimerization, we used the Ser¹⁰⁷/Ser¹²² double mutant and replaced several other residues of TMS3 with cysteine. Sixteen unique cysteine mutations covering TMS3 residues 106–123 were constructed (Fig. 5A). Following oxidation, most of the cysteine variants were observed as dimers, which reverted to monomers after treatment with β-me. At low concentrations of Cu-OP (20 and 50 μm), all constructs with substitutions at positions 106–109 were efficiently cross-linked, suggesting that this part of the helix is relatively mobile, consistent with its location at the helix end, close to the cytoplasm. Note that Cys¹⁰⁷ is in this region. For the cysteine substitutions at positions 110 to 123, only the mutants with cysteine residues on one face of the helix were able to cross-link, as depicted on the helical wheel projection in Fig. 5B showing the positions of the residues involved in EpCai dimerization; This face of the TMS3 helix is on the side opposite Cys¹⁰⁷.

FIGURE 5.

Immunoblot analysis of cross-linked EpCai proteins containing cysteine substitutions in TMS3. A, aliquots of 0.2 × 10⁸ cells of the indicated strains were incubated with or without Cu-OP at the indicated concentration (O₂₀, 20 μm; O₅₀, 50 μm) or with 25 mm β-me (R), and analyzed by 15% SDS-PAGE and immunoblotting with anti-1C11 mAb. B, helical wheel projection of the predicted transmembrane α-helix 3 of EpCai. The percentage next to each residue is the percentage of cross-linked dimer formed after oxidation with 50 μm Cu-OP at room temperature for 20 min. The strongly interacting face of TMS3 is indicated with a solid arrow.

Intramolecular Disulfide Bonds Connect TMS1 to TMS2 or TMS2 to TMS3

We show above that Cys⁷⁴ is involved in intramolecular disulfide bond formation (Fig. 3C) with Cys³¹, Cys¹²², or Cys¹⁰⁷. However single C31S, C107S, and C122S substitutions did not abolish intramolecular disulfide bonds, suggesting that Cys⁷⁴ is in close contact with at least two other cysteines. The C31S/C107S and C107S/C122S EpCai mutants, but not the C31S/C122S mutant, were able to form intramolecular bonds (Fig. 4A). Therefore, formation of intramolecular disulfide bonds required the presence of Cys⁷⁴ and Cys³¹ or Cys⁷⁴ and Cys¹²². In agreement with this result, these three cysteine residues are located on the periplasmic side of TMSs, whereas Cys¹⁰⁷ is located on the cytoplasmic side (Fig. 1).

Hydrophobic Helical Hairpin of pfColA Prevents Cross-linking of EpCai by Cu-OP

Our results indicate that EpCai dimerizes through the TMS3 in the inner membrane and that TMS2 is in close contact with TMS1 and TMS3. To determine the effect of the interaction between the pore-forming domain of ColA and EpCai, we co-produced the two proteins in the inner membrane. The pore-forming domain of ColA (pfColA) fused to a signal peptide (sp-pfColA) inserts into the inner membrane of E. coli and forms a functional channel (24). However, an E. coli strain harboring a plasmid encoding EpCai is immune to sp-pfColA, and therefore EpCai and sp-pfColA can be produced in the same cells (17). The production of EpCai was induced first through the addition of 100 μm isopropyl 1-thio-β-d-galactopyranoside. One hour later, production of sp-pfColA was induced by the addition of arabinose (0.5 mg/ml) for 30 min. In these conditions, sp-pfColA is produced in excess with respect to the EpCai and is lethal for the producing bacteria (24). Last, Cu-OP was added or not to the culture, which was incubated for 15 min and then harvested. The two proteins were detected by immunoblotting with the monoclonal antibody 1C11 and with a polyclonal antibody directed against pfColA (Fig. 6A). In the absence of sp-pfColA, EpCai was converted into [EpCai]₂ upon Cu-OP treatment, as described above. Conversely, the production of sp-pfColA led to the absence of [EpCai]₂, and only the reduced monomeric form of EpCai was detected (Fig. 6A). This indicates that the interaction between pfColA and EpCai prevents both [EpCai]₂ formation and intramolecular disulfide bond formation by Cu-OP. To confirm these findings, we constructed three new single cysteine mutants: L109C, V113C, and Y119C. These three mutated EpCai efficiently protected the producing cells against ColA at 30 °C. As for the C74S substitution, the Y119C substitution led to a thermosensitive phenotype (data not shown). These three single cysteine mutants dimerized following incubation with Cu-OP either through the native Cys¹⁰⁷ or the introduced cysteine (Fig. 6B). Furthermore, trimeric forms of EpCai were observed, suggesting that the native and the introduced cysteines can both simultaneously form disulfide bonds. In the presence of pfColA, none of the three mutated EpCai proteins dimerized upon Cu-OP incubation. This indicates that the presence of pfColA prevents dimerization of EpCai irrespective of the position of cysteines in TMS3. Analogous results were obtained with the inactive EpCai triple mutants C107S/L109C/C122S, C107S/V113C/C122S, and C107S/Y119C/C122S, which contain only one cysteine in TMS3. To corroborate these results, we performed cross-linking experiments in presence of another cross-linker, FA, which can react with primary amino groups: in the presence of sp-pfColA, wild-type EpCai was only detected as a monomer (Fig. 6C).

FIGURE 6.

sp-pfColA prevents EpCai dimerization. A, strains producing EpCai, EpCai and sp-pfColA, or EpCai and sp-pfColA_B3 grown at 37 °C and incubated for 60 min with 100 μm isopropyl 1-thio-β-d-galactopyranoside (IPTG), then 30 min with 0.5 mg/ml arabinose. Samples (0.2 × 10⁸ cells) were incubated with the indicated concentration of Cu-OP and analyzed by 15% SDS-PAGE and immunoblotting with the anti-1C11 mAb (upper panel) or the anti-pfColA polyclonal Ab (lower panel). B, as in A, with strains producing EpCai L109C, EpCai V113C, EpCai Y119C, and co-producing or not sp-pfColAB3. Samples were immunoblotted with the anti-1C11 mAb only. The arrow indicates the position of the EpCai trimer. C, sp-pfColA preventing EpCai dimerization with FA. Aliquots of 0.2 × 10⁸ cells of strains producing EpCai, EpCai, and sp-pfColA were treated (FA) or not (−) with formaldehyde. Cell extracts were analyzed by 15% SDS-PAGE and immunoblotting with the anti-1C11 mAb.

We tested whether sp-pfColA in its closed state similarly prevented EpCai cross-linking by Cu-OP. We used sp-pfColA_B3 which carries two substitutions (A492E/F493P) in the loop region connecting helices 5 and 6 of pfColA. This double substitution abolishes the channel activity of pfColA but does not prevent membrane insertion (25). We produced EpCai in the presence and absence of sp-pfColA_B3 and performed the cross-linking assay with Cu-OP (Fig. 6A). In the closed state, pfColA prevented Cu-OP-catalyzed cross-linking of wild-type EpCai (Fig. 6A) and mutated EpCai proteins (data not shown). Therefore, channel activity is not required, and the presence of sp-pfColA in the membrane is sufficient to abolish EpCai cross-linking.

EpCai Incubated with Cu-OP Is Inactive

We next tested whether covalent Cai dimers protected the cell against ColA. Cells producing Cai were totally immune to ColA, whereas ColA efficiently killed sensitive, Cai-free cells (Fig. 7A). Surprisingly, sensitive cells incubated with Cu-OP became insensitive to ColA but remained sensitive to other pore-forming colicins, including ColE1 and ColB, and to the nuclease colicins ColE2 (Fig. 7A). We suspected that at least one of the proteins involved in the ColA translocation through the E. coli envelope was affected by Cu-OP. To test this we added hybrid colicins AAB and BBA (26) to sensitive cells incubated in the presence or absence of Cu-OP. ColAAB consists of the first two domains of ColA fused to the pore-forming domain of ColB, and ColBBA consists of the first two domains of ColB fused to the pore-forming domain of ColA. As expected, the two hybrid colicins were fully active against sensitive cells, whereas only ColBBA was active against sensitive cells incubated with Cu-OP (Fig. 7B). This confirms that ColA binding or translocation through the E. coli envelope is affected by Cu-OP but that pore-forming activity is not. We then tested ColBBA activity on cells producing Cai incubated with or without Cu-OP. In the absence of Cu-OP, Cai effectively protected the cells against added ColBBA (12). Conversely, the protection conferred by Cai was completely abolished by incubation with Cu-OP (Fig. 7C). This clearly demonstrates that the oxidized Cai locked by disulfide bonds is unable to prevent ColA pore formation.

FIGURE 7.

Oxidized Cai is inactive. Agar plates containing or not 200 μm Cu-OP or 200 μm Cu-OP and 5 mm DTT (E) were overlaid with E. coli C600 cells constitutively producing or not producing Cai or mutated Cai (D), as indicated. Then, 1 μl of a serial dilution of ColA, ColE1, ColE2, or ColB (A), ColAAB or BBA (B–E) was dropped onto each plate. The absence of zones of killing indicates biological resistance to colicins. These experiments were performed at 30 °C.

To determine which disulfide bonds inactivate Cai, we tested ColBBA activity on cells producing either EpCai_C107S/C122C or EpCai_C74S. In the presence of Cu-OP, EpCai_C107S/C122C is unable to form intermolecular disulfide bonds, and EpCai_C74S does not form intramolecular disulfide bonds. These two mutated EpCai were inactive when Cu-OP was added (Fig. 7D), suggesting that the intramolecular and the intermolecular disulfide bonds are both independently able to inactivate the immunity protein.

Finally, to demonstrate that the EpCai inactivity following incubation with Cu-OP results from disulfide bond formation and not from any secondary effects of metal ions such as Cu²⁺, we tested ColBBA activity on cells producing and not producing EpCai in the presence of Cu-OP and in the presence of the reducing agent DTT (Fig. 7E). In the presence of DTT, EpCai effectively protected the cells against the lethal action of ColBBA and reversed the effect of Cu-OP.

DISCUSSION

The topological model of Cai, constructed from biochemical and genetic analyses (10, 11), indicates the presence of four transmembrane segments. However, little was known about either the three-dimensional arrangement of the TMSs within the inner membrane or the oligomeric state of Cai. The Cai sequence does not contain any canonical motif likely to mediate the noncovalent association of the TMSs: in particular, there is no leucine heptad motif and no GXXXG motif, known to mediate association of TMSs (27, 28).

To investigate the helix packing of Cai, we used Cu-OP, a zero-length cross-linker of Cys residues. In the presence of Cu-OP, EpCai migrated as a dimer in SDS-PAGE. We studied the physiological relevance of this dimerization. Indeed, the observed dimerization may have been artifactual, resulting from simple aggregation of EpCai and/or random collisions. In a previous study on the biosynthesis and insertion of Cai, Géli et al. found that noninserted Cai protein did not aggregate but was rapidly degraded (10); consequently, they indicate that the amount of Cai detected in the membrane by immunoblot analysis after SDS-PAGE reflects insertion of Cai into the membrane. Here, we verified that after induction of the genetic construct, EpCai was recovered in the membrane fraction and was only solubilized in the presence of detergent, suggesting that the protein was correctly inserted into the inner membrane and not aggregated (data not shown). Ji and Middaugh demonstrated that the formation of random collisional cross-linking of membrane proteins is a rare event occurring only under extreme conditions of high protein concentration and long times of incubation (29). Even after inducting overexpression of the gene, EpCai has been reported to be only minor protein of the inner membrane (20). Furthermore, we used short times of incubation (usually 15 min) with Cu-OP or FA to assure that cross-linking reflected specific interactions rather than random proximity. Three further observations confirm that EpCai dimers occur naturally: EpCai appears to dimerize along a specific surface (TMS3) rather than randomly; second, point mutation specifically abolished dimer formation; and finally EpCai dimer formation was completely abolished by the presence of its target, the ColA C-terminal domain.

In the presence of Cu-OP, we detected a shift in the electrophoretic mobility of EpCai (Fig. 2A). We associated this mobility shift to the formation of an intramolecular disulfide bond. Such mobility shifts have been previously used to analyze cross-linking of the Tar chemotactic receptor (30), and subunit a of E. coli ATP synthase (31). In general, cross-linking between helices results in a more compact structure in SDS leading to a faster electrophoretic migration through the polyacrylamide gel. However, the presence of a disulfide bond can also have an opposite effect and lead to lower electrophoretic mobility, as for example the cardiac Na⁺-Ca⁺ exchanger proteins (32). In our case, disulfide bond formation reduced the mobility of EpCai in SDS-PAGE (Figs. 2A, 3, and 4). This shift in the mobility can be attributed to formation of an intramolecular disulfide bond between TMSs of EpCai for the following reasons: (i) in the absence of Cu-OP, this shift was never observed (Fig. 2, A and B); (ii) the sequential treatment with Cu-OP, NEM, and then β-me, which reduces and breaks disulfide bonds, abolished the shift in mobility (Fig. 2A); and (iii) we did not observe a shift after Cu-OP treatment either with the mutants EpCai_C74S and EpCai_C31S/C122S or when wild-type EpCai was co-produced with sp-pfColA (Figs. 3C, 4A, and 6A).

To identify the cysteine residues involved in intra- and intermolecular disulfide bond formation, we used a reverse cysteine scanning approach. A mutated form of EpCai lacking all of the four natural cysteine residues was constructed. However, this EpCai quadruple mutant was unstable, even at low temperature, and could not be visualized by Western blotting. Only in those constructs with single serine mutations or the double serine mutation C31S/C107S was EpCai activity not significantly affected. All other mutations (double or triple) affected EpCai activity to various degrees (Table 1). However, we were able to visualize and discriminate all of the different forms of the mutated proteins by Western blotting (monomer versus dimer, with or without intramolecular disulfide bonds), and therefore obtain information about helix packing in the membrane. We clearly demonstrate that EpCai forms homodimers through its TMS3 and that in one monomer TMS2 is in close contact with TMS1 and TMS3. The natural cysteine 107 and the region of the TMS3 helix involved in dimer formation are on opposite side of TMS3 (Fig. 5B). When the residues at positions 106–109 are replaced with cysteines they are able to cross-link like the natural cysteine 107, suggesting that this region is flexible. Such flexibility is consistent with its location at the N terminus of TMS3 and could explain why Cys¹⁰⁷ is able to come into close proximity with the same residue of a neighboring monomer. Thus, EpCai appears to form homodimers through one face of the TMS3 helix, and the observed cross-linking through the native cysteine 107 is possible because of the high flexibility of the 106–109 region. The high mobility of the 106–109 region and Cys¹⁰⁷ in particular may explain why EpCai is able to form trimers when cysteines are introduced in positions 109, 113, and 119 (Fig. 6B). Two observations are quite surprising: the failure of the A126C construct to cross-link and the efficient cross-linking of the F121C construct (Fig. 5B). The A126C mutant may be unable to cross-link due to the tight packing of residues at the interface of the helix dimer. A126C may be disruptive as a result of steric hindrance to helix packing. Phe¹²¹ is at the end of the helix, close to the periplasm and therefore may be in a relatively mobile region. These various findings allow us to propose a preliminary model of the arrangement of the helices in the membrane (Fig. 8).

FIGURE 8.

Model of Cai helix packing and dimerization. White and gray circles represent the four TMSs (H1–H4) of two monomers of Cai that form the dimer. Double-headed arrows indicate intramolecular disulfide bonds between Cys residues, and black lines indicate intermolecular disulfide bonds. A large gray arc represents the TMS3 helix face involved in dimerization. The dotted line indicates a potential interaction between one Cai monomer represented with white circles and another Cai monomer depicted with four hatched circles.

Interestingly, the presence of pfColA modifies Cai helix packing. When pfColA is co-produced and targeted to the E. coli inner membrane, EpCai does not form any disulfide bonds, indicating that pfColA modifies both the dimer interface of EpCai and the respective orientations of the TMSs within the monomers. Disulfide bonding is also inhibited by an inactive from of pfColA, which inserts into the membrane by its hydrophobic helical hairpin, but is unable to form a channel. The hydrophobic helical hairpin of pfColA thus alters the relative distances between residues 31 and 74, and residues 74 and 122 in within the protein molecule and also between residues 107, 109, 113, 119 of adjacent protein molecules. Interaction between EpCai and its target pfColA results in major modifications of the helix packing in EpCai, and in particular an increase of the distance between the two adjacent TMSs3. This interaction also prevented Cys¹⁰⁷ from forming a disulfide bond with the same residue of a neighboring monomer, indicating decreased flexibility of the Cys¹⁰⁶–Cys¹⁰⁹ region. TMS3 has been identified as a key region for helix-helix interaction between immunity proteins and their cognate colicins (16, 33, 34). Possibly, in the absence of pfColA, the arrangement of the TMSs of EpCai are mobile as suggested by the intramolecular disulfide bonds formed in oxidative conditions being at any of various different positions. Espesset et al. (19) suggested that a certain degree of mobility between the helices of EpCai might be important for the interaction between these helices and the hydrophobic helical hairpin of the pore-forming domain and the subsequent inactivation of the colicin. In agreement with this notion, an intermolecular disulfide bond or an intramolecular disulfide bond introduced into EpCai that reduced interhelix mobility also inactivated the protein (Fig. 7). It is also clear that the conformation of EpCai in the presence of pfColA is different from that in its absence: the presence of pfColA leads to the absence of both intramolecular and intermolecular disulfide bonds (Fig. 6A).

In summary, investigating cross-linking involving introduced or native cysteine residues in Cai allowed us to show that Cai is a dimer in the membrane. The positions of cysteine involved in intra- or intermolecular disulfide bonds allow us to propose a model of the dimer showing both the arrangement of the TMSs in the monomers and that TMS3 is involved in dimerization. Most importantly, we found that pfColA interacts with the Cai dimer and that this interaction modifies both Cai dimer interaction and the arrangement of the helices inside the two monomers. Although we do not show conclusively that pfColA dissociates the EpCai dimer, preliminary results reported here are in favor of this hypothesis (Fig. 6). These findings have important implications for the understanding of ColA inactivation by Cai and direct attention the role of the Cai dimer in pfColA inactivation.