FtsH-dependent Processing of RNase Colicins D and E3 Means That Only the Cytotoxic Domains Are Imported into the Cytoplasm

Abstract

It has long been suggested that the import of nuclease colicins requires protein processing; however it had never been formally demonstrated. Here we show that two RNase colicins, E3 and D, which appropriate two different translocation machineries to cross the outer membrane (BtuB/Tol and FepA/TonB, respectively), undergo a processing step inside the cell that is essential to their killing action. We have detected the presence of the C-terminal catalytic domains of these colicins in the cytoplasm of target bacteria. The same processed forms were identified in both colicin-sensitive cells and in cells immune to colicin because of the expression of the cognate immunity protein. We demonstrate that the inner membrane protease FtsH is necessary for the processing of colicins D and E3 during their import. We also show that the signal peptidase LepB interacts directly with the central domain of colicin D in vitro and that it is a specific but not a catalytic requirement for in vivo processing of colicin D. The interaction of colicin D with LepB may ensure a stable association with the inner membrane that in turn allows the colicin recognition by FtsH. We have also shown that the outer membrane protease OmpT is responsible for alternative and distinct endoproteolytic cleavages of colicins D and E3 in vitro, presumably reflecting its known role in the bacterial defense against antimicrobial peptides. Even though the OmpT-catalyzed in vitro cleavage also liberates the catalytic domain from colicins D and E3, it is not involved in the processing of nuclease colicins during their import into the cytoplasm.

Introduction

Colicins are antibacterial toxins of Escherichia coli that are released into the extracellular medium in response to environmental stress conditions. Colicin D is an RNase that cleaves the anticodon loop of all four isoaccepting tRNA^Arg (1). Colicin E3 cleaves 16 S ribosomal RNA (2). Both colicins provoke cell death by inactivating the protein biosynthetic machinery. Colicin producer cells are protected against both endogenous and exogenous toxin molecules by the constitutive expression of a cognate immunity (Imm)³ protein, which forms a tight heterodimer complex with the nuclease domain of the colicin (3, 4).

Colicin E3, like most colicins, has structurally identifiable N-terminal, central, and C-terminal domains. The first two domains are required for translocation and receptor binding, and they “hijack” certain functions of the target cell (i.e. the BtuB receptor and the Tol system) during colicin import. The C-terminal domain carries the cell-killing RNase function (5, 6). The colicin D protein has an unusual tripartite organization. The N-terminal domain is required for both the binding of the colicin to the high affinity, iron siderophore receptor FepA and for its subsequent translocation across the outer membrane. The 280-residue central domain is essential for uptake (and thus for cell killing), and it is also involved in the formation of the colicin D-ImmD protein complex (7). The passage of colicin D through the outer membrane is dependent on the proton motive force of the cytoplasmic membrane, transduced by the TonB/ExbB-D system (8, 9).

During the import process, when E-type nuclease colicin-Imm complexes bind to the cell surface receptor BtuB, their Imm protein dissociates from the bound colicin and is released into the external medium (10, 11). The binding of the complex per se is however not enough to liberate the Imm protein. The dissociation of colicins E9- and E2-Imm complexes requires the unfolding of the colicin, as it contacts the energy-transducing Tol system in the periplasm (12, 13).

To transfer the cytotoxic domain of the colicin molecule across the inner membrane into the cytoplasm, nuclease colicins need to parasitize more of the cell functions than the pore-forming colicins. The DNase domains of colicins E9 and E2 exhibit nonvoltage-gated channel forming activity in planar lipid bilayers, which involves changes in their conformation (14). Such channels, unlike those formed by pore-forming colicins, are not directly responsible for cell killing but may allow “self-propulsion” of the toxic domains into cytoplasm, driven by an electrostatic association of the DNase domain of colicin E9 with the inner membrane (15). Although such an association between the RNase colicin E3 and anionic phospholipid surfaces has been reported, no RNase colicin has been found to exhibit any channel forming activity (16).

Using cell extracts the in vitro cleavage of full-size colicin D and more recently of the DNase colicin E7 was observed at the boundary of the C-terminal killing and central domains (17–19). The inner membrane signal peptidase LepB was shown to be essential for cell killing by colicin D and also for the cleavage of colicin D in vitro in the presence of a cell extract. We and others have proposed that proteolytic processing may be a common step in cell killing by nuclease type colicins and that it should occur prior to or concomitant with the translocation of their C-terminal catalytic domain through the inner membrane (17, 19–21).

OmpT, a member of the family of outer membrane endopeptidases (omptins), plays a role in the virulence of a variety of pathogenic Gram-negative bacteria (22). Moreover, in agreement with earlier observations OmpT has been shown to cleave several receptor-bound colicins and thus to improve the survival of target bacteria exposed to colicins or other antimicrobial peptides (23–25).

In this work, we identified the in vivo processed forms of both RNase colicins D and E3 as the final colicin forms present in the cytoplasm of colicin-treated bacteria. The endoproteolytic cleavage was shown to require the inner membrane protease FtsH, which is required for the cytotoxicity of all nuclease colicins (26). Additionally the signal peptidase LepB is specifically required for the processing and/or translocation of colicin D across the inner membrane. In vitro cleavage of colicins D and E3 requires the outer membrane protease OmpT. However, OmpT is not required for the import and toxicity of colicins.

EXPERIMENTAL PROCEDURES

Bacterial Strains and Plasmids

E. coli K12 strains C600 and JM101 were used as wild-type strains. DH5α was used as the host strain for cloning and mutagenesis. All other strains and plasmids used or constructed are listed in Table 1.

TABLE 1.

E. coli strains and plasmids

Strain or plasmid	Genetic description	Source
Strain
A38a	C600 lepB (N274K), colBS/colDR	Ref. 17
AD202	ompT1000::kan	Ref. 59
C600ompT	C600 ompT1000::kan (by P1 transduction from AD202)	This work
BL21(DE3)	ΔompT	Ref. 60
UT5600	Δ(ompT-fepA-C)266	Ref. 61
C600tonB	C600 tonB::TN5	Ref. 62
D1a	D10 tonB (R158S), colBS/colDR	Ref. 40
JCL8789	188 nadA::Tn10, ΔtolB-pal, TetR	Ref. 63
JCL11650	169 btuB::Tn10, TetR	Ref. 63
AR3291	sfhC zad::Tn10, ftsH3::kan	Ref. 64
Plasmid
pLEPB(wt)	pACYC184, lepAB (at SphI/BamHI), CmR	This work
pRD8	pING-1 (araBC), lepAB, AmpR	Ref. 65
pRD8(K145A)	pING-1 (araBC), lepAB(K145A), AmpR	This work
pLEPB	pET23, (tag-His6)lepB, AmpR	Ref. 35
pLEPB(K145A)	pET23, (tag-His6)lepB(K145A), AmpR	Ref. 35
pLEPB(N274K)	pET23, (tag-His6)lepB(N274K), AmpR	This work
pML19	pUC19, ompT, AmpR	Ref. 66
pSTD113	ftsH, AmpR	Ref. 67
pFTSH	pACYC184, ftsH (at SalI/HinDIII), CmR	This work
pJF129b	pColD-CA23, colicin D operon cda-cdi-cdl	Ref. 68
pColDI	pET11a, cda-cdi(tag-His6) (at NdeI/BamHI), AmpR	This work
pCDM590	pET11a, the 3′-domain of cda (starting at Met590)-cdi(tag-His6) (at NdeI/BamHI), AmpR	This work
pIMMD	pTRAC, cdi(tag-His6) (at NdeI/BamHI), AmpR	This work
pIMME3	pTRAC, cei(tag-His6) (at NdeI/BamHI), AmpR	This work
pKSJ28b	pColE3-CA38, colicin E3 operon cea-cei-cel	Ref. 69
pColE3I	pET11a, cea-cei(tag-His6) (at NdeI/BamHI), AmpR	This work
pGDL81	sipS (lepB orthologue from B. amyloliquefaciens)	Ref. 70

^a Strain sensitive to colicin B but resistant to colicin D.

^b cda, cdi, and cdl are genes for colicin D its immunity protein and bacteriocin release protein; cea, cei, and cel are the corresponding genes for the colicin E3 operon.

Purification of Colicins

Colicin D, unlabeled or labeled in vivo with [³⁵S]Met (1000 Ci/mmol, Amersham Biosciences), in complex with its immunity protein (ImmD) was purified from an E. coli strain carrying pJF129 as described previously (17, 27). Colicin D was separated from its ImmD by gel filtration in the presence of 9 m urea (Superose 12 HR column, GE Healthcare) and refolded by dialysis against 20 mm Tris/HCl, pH 8.0.

Colicin E3-ImmE3 complex was purified from an E. coli strain carrying pKSJ28 and then dissociated by two chromatographic steps on a Mono S column (GE Healthcare) without and with 6 m urea, respectively, according to de Zamaroczy et al. (17).

Alternatively, colicin D in complex with the ImmD protein carrying a C-terminal His₆ tag (colicin D-ImmD(Ct-His₆)) was produced from BL21(DE3) cells carrying the plasmid pColDI in LB medium with ampicillin. At A₆₀₀ = 0.5 the expression of the colicin operon was induced by isopropyl 1-thio-β-d-galactopyranoside (1 mm), and incubation was continued for 3 h at 37 °C. After centrifugation the cells were resuspended in 20 ml of loading buffer (20 mm Tris, pH 8.0, 0.5 m NaCl, and 200 μg of DNase) and then broken by one passage through a French pressure cell. The supernatant was loaded onto a 1-ml Ni-NTA resin column (His-select nickel affinity gel, Sigma) equilibrated previously with the loading buffer and washed with the same buffer, and the His-tagged colicin D-ImmD complex was then eluted with 150 mm imidazole. After the complex was denaturated in 9 m urea, His-tagged ImmD was retained by two passages over a Ni-NTA column. Colicin D freed of ImmD present in the “flow-through” was concentrated (via ultrafiltration in an Amicon Ultra-4, nominal molecular weight limit 10,000; Millipore) after dialysis. The same protocol was used to produce colicin E3-ImmE3(Ct-His₆) from plasmid pColE3I.

Purification of the 12.4-kDa Colicin D Catalytic Domain in Complex with ImmD

The cloned catalytic domain (CAT) corresponds to amino acids Met⁵⁹⁰–Leu⁶⁹⁷ from the C-terminal part of colicin D. Met⁵⁹⁰ was chosen as initiating amino acid because it corresponds to a spontaneous break in the protein, which occurs during the crystallization of colicin D in complex with ImmD (27). The catalytic domain in interaction with the 10.9-kDa ImmD(Ct-His₆) was expressed from BL21(DE3) carrying the plasmid pCDM590 and purified as described above. The catalytic domain, used for antiserum production, although separated from its ImmD partner by 9 m urea still contained some ImmD protein. Thus, the polyclonal antiserum against the catalytic domain also cross-reacts with the ImmD protein.

Purification of His-tagged LepB Signal Peptidase

BL21(DE3) cells containing plasmid pLEPB, pLEPB(K145A), or pLEPB(N274K) were grown, and LepB protein expression was induced as described above. His-tagged LepB was purified on a Ni-NTA column in the presence of 1% Triton X-100 and eluted by a gradient of 25 to 200 mm imidazole (28).

N-terminal Residue Determination of the in Vitro Obtained Small Cleaved (SC Form) Colicin Peptides

The 10.5-kDa in vitro cleaved SC colicin D form was transferred from 15% SDS-PAGE to a ProBlott membrane (Applied Biosystems) for N-terminal peptide sequencing with an ABI 473A automatic sequencer at the Pasteur Institute, Paris. The in vitro cleavage site at the start of the 11-kDa SC fragment of colicin E3 was localized by mass spectrometric analysis performed on a MALDI-TOF-MS Voyager System 4106 at the École Supérieure de Physique et de Chimie Industrielles (ESPCI), Paris.

Detection of LepB Interaction with Colicin D by Far Western Blotting

Purified colicins or colicin D domains (2–5 μg) were separated on 15% SDS-PAGE, transferred to nitrocellulose membrane, and then stained by Ponceau red. A denaturation-renaturation step according to Wu et al. (29) was performed to refold the proteins. After incubation for 1 h at 37 °C with LepB (1 μg/ml) and successive washing steps, the binding of LepB to colicins was analyzed by an anti-LepB antiserum and ECL.

In Vitro Analysis of Colicin Cleavage

Extracts of E. coli periplasmic proteins were prepared by a classical osmotic shock procedure adapted from the QIAexpressionist protocol (71). Alternatively, cell fractionation was performed using a spheroplast-producing protocol based on lysozyme treatment of the cells (30). Aliquots of periplasmic extract were stored at −20 °C in 20 mm sodium phosphate, pH 6.5, 5 mm MgSO₄. Periplasmic protein extracts, with or without the addition of purified OmpT (with lipopolysaccharides (LPS), 0.4 μg) (31) and/or LepB (1.5 μg) or SipS, were incubated for 1 h at 37 °C in 20 mm sodium phosphate, 5 mm MgSO₄, pH 7.0, with unlabeled or [³⁵S]Met-labeled colicin (4 and 1.5 μg, respectively). Full-size (FS) and cleaved colicin forms (large cleaved (LC) and small cleaved (SC)) were precipitated with acetone and then separated by SDS-PAGE, and the proteins were detected by Coomassie Blue staining or phosphorimaging (Typhoon, GE Healthcare), respectively. The periplasmic extracts obtained by the two methods were used in parallel for the colicin cleavage tests and gave the same results in three repeated experiments. Polymixin B, known to form a complex with LPS, which are necessary for the activity of OmpT (32), was added (3 μg/ml) to the reaction mixture to test the OmpT dependence of in vitro colicin cleavage.

Identification by “Fishing” for the in Vivo Processed Colicin Peptides from Colicin D- or E3-treated Target Cells

50-ml cultures of wild-type or mutant E. coli strains carrying the plasmid pImmD were grown in LB medium to an A₆₀₀ = 0.4 at 37 °C and then induced for 1 h with 1 mm isopropyl 1-thio-β-d-galactopyranoside to express the plasmid-encoded His-tagged colicin D immunity protein and with 0.05 mm 2,2′-dipyridyl. 1 mg of colicin D-ImmD complex was added to the cultures for 1–3 h. The cells were harvested, washed twice, resuspended in 10 ml of 20 mm sodium phosphate, 150 mm NaCl, pH 7.0, and treated with proteinase K at 100 μg/ml for 1 h at 37 °C. After four additional washings, the colicin d- and proteinase K-treated cells were resuspended in 2 ml of buffer (Tris 20 mm, and 0.5 m NaCl, pH 7.5) with a protease inhibitor complex (Complete EDTA-free; Roche Applied Science) and then disrupted by sonication. The soluble fraction recovered after centrifugation was then loaded onto a Ni-NTA affinity column. The bound His-tagged ImmD, carrying any attached colicin molecules that had been “fished” from the cytoplasm, was eluted with 300 mm imidazole, dialyzed, and concentrated (via ultrafiltration in an Amicon Ultracel 3K, Millipore). The processed colicin D peptide (ColD PF; recovered from about 10 ml of the initial cell culture) was separated from the His-tagged ImmD on 15% SDS-PAGE, identified by Western blot analysis using an antiserum to the colicin D catalytic domain, to the immunity protein or to the full-size colicin D, and detected by ECL chemiluminescence (Immun-Star, Bio-Rad) associated with a charge-coupled device camera (ChemiDoc XRS+ System, Bio-Rad). A similar protocol to that used for colicin D, but without 2,2′-dipyridyl, was used to look for processing of colicin E3. Wild-type or mutant strains carrying plasmid pImmE3 encoding the His-tagged colicin E3 immunity protein were grown in LB or minimal medium (isopropyl 1-thio-β-d-galactopyranoside) and treated with colicin E3-ImmE3 for 1–2 h. Proteins bound to the ImmE3-His₆ were purified by Ni-NTA chromatography, and Western blot analysis was performed with an anti-colicin E3 antiserum. The His-tagged ImmE3 was detected from the bottom part of the same SDS-PAGE blot by an anti-His₆ antiserum.

Direct Immunodetection of Processed Colicin D and Colicin E3 Peptides in the S100 Cytoplasmic Fraction of Colicin-sensitive Target Cells

The direct immunodetection of the in vivo processed product required an increased colicin D import into target cells and further purification to remove other cytoplasmic proteins. A 50-ml culture of strain AD202 at A₆₀₀ = 0.3 and 37 °C was induced by 0.05 mm 2,2′-dipyridyl. At A₆₀₀ = 0.8, 0.7–2 mg of colicin D-ImmD was added to the cells for 1–2 h. After harvesting, the cells were treated with proteinase K and washed, and the cytoplasmic fraction was prepared as described above. Ribosomal proteins were eliminated from cytoplasmic extracts by micro-ultracentrifugation at 100 S (Beckman rotor TLA 100.2) to give an S100 fraction. To increase the threshold of detection of the processed colicin D form, S100 fractions were enriched for lower molecular weight proteins. Thus, the S100 fraction was precipitated with acetone. Proteins were resuspended in 350 μl of 20 mm sodium phosphate buffer, pH 7.0, and separated by FPLC-monitored gel filtration (Superdex 75 GL 10/300). Fractions containing proteins with molecular weights lower than 35,000 were pooled and precipitated by acetone. The concentrated and selectively enriched cytoplasmic proteins were separated on 15% SDS-PAGE and analyzed by Western blotting as described above.

In the case of colicin E3, S100 fractions, prepared from cells grown in LB or in minimal medium (to enhance the expression of BtuB receptor) and treated with 2–4 mg of colicin E3-ImmE3 for 0.25–2 h, were separated on 15% SDS-PAGE and analyzed directly by Western blotting with anti-colicin E3 antiserum. When necessary, ECL-detected proteins were quantified by ImageLab software (Bio-Rad).

RESULTS

In Vitro Cleavage of Colicin D in Periplasmic Extracts

We showed previously that only when a crude extract of the colicin D-resistant strain A38 carrying the lepB (N274K) mutation was supplemented with purified LepB was it capable of the endoproteolytic cleavage of full-size colicin D molecules. However, LepB alone was not able to cleave colicin D (17) (Fig. 1C, lane 4). We sought therefore to identify an additional component that might be required with LepB for colicin D cleavage. We chose to focus our investigation on periplasmic protein extracts, because the active site of the inner membrane protein LepB is oriented toward the periplasmic space (33). When LepB was added to the periplasmic extracts, prepared from the lepB mutant, about 50 ± 7% of FS 75-kDa colicin D (free of its immunity protein) was found to be cleaved. A 65-kDa LC form (Fig. 1A, lane 2) and an about 11-kDa SC form (data not shown) were visualized by SDS-PAGE. The N-terminal amino acid sequence of the SC peptide was determined to be ⁶⁰⁸KYKHAGDFGISD⁶¹⁹ (Fig. 2A). The in vitro cleavage site is located between two lysines residues at positions 607 and 608, so that cleavage liberates a C-terminal peptide of 10.5 kDa, which corresponds in size to the previously defined minimal tRNase domain of colicin D (17).

FIGURE 1.

LepB-dependent in vitro cleavage of colicin D in the presence of periplasmic extract or OmpT. A, ColD, free of its immunity protein, was treated with periplasmic extracts (PE) prepared from lepB (N274K) mutant strain A38 alone (lane 1) or in the presence of purified LepB (lane 2) or LepB and polymyxin B (Poly B, lane 3). The products were separated by 8% SDS-PAGE and detected by Coomassie Blue staining: FS ColD, 75 kDa, and LC form, 65 kDa. B, comparison of the sizes of the SC form of colicin D produced in vitro by purified OmpT and LepB with the cloned 12.4-kDa catalytic domain (CAT 12.4) and the 10.9-kDa ImmD(6His) protein. Samples were analyzed by 15% SDS-PAGE and stained with Coomassie Blue. C, cleavage of colicin D in vitro by OmpT in the presence of purified LepB (lane 1) or LepB-K145A (lane 2). The positions of migration of OmpT and LepB are indicated. (Refer to supplemental Fig. S1 for an analysis of colicin D cleavage by SipS of B. amyloliquefaciens, a LepB orthologue.) D, [³⁵S]ColD was treated with periplasmic extracts of an ompT-inactivated strain overproducing OmpT (PE pompT, lane 3) or not (PE ΔompT, lanes 1 and 2) and supplemented with purified LepB (lanes 2 and 3). FS ColD and the LC form were separated by 8% SDS-PAGE and detected by phosphorimaging.

FIGURE 2.

Location of the OmpT-catalyzed in vitro cleavage sites and the in vivo processing sites in RNase colicins D and E3. A, peptide sequence of the junction between the central domain and the tRNase domain of colicin D. The location of the OmpT cleavage (at position Lys⁶⁰⁷–Lys⁶⁰⁸), producing in vitro the 10.5-kDa SC form, and the approximate location (near Met⁵⁹⁰) of the in vivo processing site producing the 12.4-kDa ColD PF are indicated. Met⁵⁹⁰ is the first amino acid of the cloned catalytic domain (Met⁵⁹⁰–Leu⁶⁹⁷) of colicin D. B, peptide sequence of the junction between the central R-domain and RNase domain of colicin E3. The OmpT cleavage sites, located between residues Arg⁴⁵⁴ and Lys⁴⁵⁵ (within the linker) and within the peptide motif Arg⁴³²–Arg⁴⁴⁰, and the corresponding LC and SC forms are indicated. (Refer to supplemental Fig. S2 for a complementary analysis of the OmpT-dependent cleavage of colicin E3.) The approximate location of the in vivo processing site around position Asp⁴²⁰ producing the 15-kDa ColE3 PF is indicated.

Outer Membrane OmpT Is the Protease Necessary for Colicin D Cleavage in Vitro

The cleavage site, located between two basic residues, satisfies the amino acid requirement for cleavage by the outer membrane OmpT protease (34). In the presence of LepB, about 45 ±7% colicin D was cleaved in vitro with purified OmpT-(LPS) (Fig. 1C, lane 1). However, only a marginal cleavage of colicin D (<5%) was detected with OmpT alone (data not shown). We checked that colicin D cleavage was completely inhibited when polymixin B was added to the reaction mixture (Fig. 1A, lane 3). Polymixin B is known to form a complex with the LPS, which are necessary for the activity of OmpT (32). The N-terminal sequence of the SC form, liberated by the purified LepB and OmpT together (Fig. 1B), showed that it corresponds to exactly the same cleavage site observed with LepB and the periplasmic extract (Fig. 2A). We confirmed that OmpT protease was responsible for the in vitro cleavage by preparing periplasmic extracts from ΔompT mutant strains (AD202 and C600ompT). These periplasmic extracts supplemented with LepB were not able to cleave colicin D. But the endoproteolytic cleavage of colicin D was efficiently restored using extracts from the same strains overproducing wild-type OmpT from plasmid pML19 if supplemented with LepB (Fig. 1D, lanes 2 and 3).

LepB Catalytic Activity Is Not Involved in Colicin D Cleavage by OmpT

To further analyze the function of the LepB peptidase, we studied in vitro the cleavage of colicin D by OmpT in the presence of the purified mutant LepB protein (K145A). This mutation affects an essential residue in the active site of LepB without causing significant conformational change. Thus, it prevents, both in vitro and in vivo, the processing of the natural substrates of LepB (35). Despite the loss of its normal catalytic function, LepB (K145A) was able to promote the in vitro cleavage of colicin D by OmpT (Fig. 1C, lane 2). This indicates that the catalytic activity of LepB is not required for colicin D cleavage in vitro. We also showed that the LepB orthologue SipS of Bacillus amyloliquefaciens could replace LepB for the cleavage of colicin D (supplemental Fig. S1), and we observed that the SipS protein from plasmid pGDL81 efficiently restored the sensitivity of the strain A38 lepB mutant to colicin D.

LepB Interacts with Colicin D in Vitro

Further evidence that the role of LepB in colicin D processing is structural came from the detection of a direct interaction between purified wild-type LepB and colicin D. We detected by Far Western blotting a strong signal indicating the formation of a complex between LepB and colicin D (Fig. 3B, lane 7). No interaction was detected with the DNase colicin E2, RNase colicin E3, or the pore-forming colicin B (Fig. 3B, lanes 1, 2, and 8), which is consistent with the in vivo observation that LepB is not required for the toxicity of any other colicins than colicin D. Analysis of the separated domains of colicin D showed that only the central domain (Glu³¹⁴–Met⁵⁹⁰) was targeted by LepB (Fig. 3B, lanes 5 and 6) but not the N-terminal domain (Met¹–Glu³¹³, 96% identical to that of colicin B) (lane 3), the catalytic domain (Met⁵⁹⁰–Leu⁶⁹⁷) (lane 4), or the ImmD protein (lanes 4, 6, and 7). Thus, LepB recognizes the central domain, which is unique to colicin D. Moreover, we also detected a strong decrease in the interaction of colicin D, and especially its central domain, with the purified LepB-mutant protein (N274K) (<10%) (Fig. 3E, lanes 7, 6, and 5) as compared with that obtained with the wild-type LepB (100%) (Fig. 3D, lanes 7, 6, and 5), whereas the interaction with the catalytic mutant LepB protein (K145A) remained strong (85 ± 10%) (Fig. 3F, lanes 7, 6, and 5). In each blot (Fig. 3, C–F) the same amount of wild-type LepB was included (lanes C) and used as a standard to validate the quantitative comparison of LepB-specific bands from these blots. We checked that the variations in the interaction were not due to a difference in the affinity of the anti-LepB antiserum for the LepB-mutant proteins. As shown in Fig. 3G, this antiserum has equal affinity for the recognition of wild-type and each mutant LepB protein. Because the wild-type and non-catalytic LepB protein exhibited a similar capacity to form a complex in vitro with the central domain of colicin D, we verified that the plasmid pRD8(K145A), expressing the LepB mutant protein K145A, complemented the lepB(N274K) mutant strain A38 for colicin D sensitivity, as well as the wild-type LepB expressed by plasmid pRD8 (data not shown). These experiments indicate that colicin D import requires LepB but not it's catalytic activity and presumably necessitates a direct interaction of the toxin with LepB in the periplasm.

FIGURE 3.

In vitro interaction of the LepB signal peptidase with colicin D. A, DNase colicin E2 (lane 1), pore-forming colicin B (lane 2), the N-terminal domain of colicin D (lane 3), colicin D CAT with ImmD (lane 4), colicin D central domain (CD, lane 5), colicin D central and catalytic domains (CD+CAT) with ImmD (lane 6), tRNase colicin D (FS ColD) with ImmD (lane 7), and RNase colicin E3 with ImmE3 (lane 8) were separated by 15% SDS-PAGE and detected by Ponceau red staining. Molecular masses are given in kDa. B, samples are as described in A. Polypeptides interacting with purified wild-type LepB were detected by Far Western blotting with anti-LepB antiserum. C, a subset of the proteins analyzed in A was tested in the absence of any LepB (control). The numbers above each lane identify the proteins as related to blot A. Contaminant signals (lane 6) are also visible in D–F. D, same blot as in C; proteins interacting with purified wild-type LepB were tested as indicated in B. E, same blot as in C; proteins interacting with purified mutant LepB (N274K; resistance to colicin D) were tested as indicated in B. The efficiency of interaction between mutant LepB and colicin D was quantified and expressed as a percentage of values measured with wild-type LepB (D, 100%). To validate the quantitative comparison of LepB specific bands, the same amount of wild-type LepB was included (lanes C) in each blot (C–F). F, same blot as in C; proteins interacting with purified mutant LepB (K145A; inactive enzyme) were tested as indicated in B. G, affinity test of the anti-LepB antiserum for the wild type (wt) and each of the mutant LepB proteins (LepB K145A and LepB N274K). The same amount of protein (about 50 ng) was loaded for each sample on 15% SDS-PAGE and then detected by Western blotting.

In Vitro Demonstrated Cleavage of RNase Colicins by OmpT Has No Role in Their in Vivo Processing

We investigated whether other colicins were subject to OmpT cleavage in vitro by examining the RNase colicin E3, which parasitizes the Tol system for its import. Using [³⁵S]Met-labeled colicin E3, free of its immunity protein, two LC fragments, of 47 and 45 kDa (Fig. 2B), were obtained from the 58-kDa FS colicin in the presence of periplasmic extracts prepared from either the wild-type or lepB A38 mutant strains (Fig. 4A, lanes 2 and 4). As expected, LepB is not a requirement for the in vitro cleavage of colicin E3 (20). Purified OmpT restored the cleavage of colicin E3 when the assay was performed in an extract from an OmpT-deficient strain (Fig. 4A, lanes 5 and 7). However, only the shorter, 45-kDa LC form was detected with purified OmpT, even in the absence of periplasmic extract (Fig. 4A, lanes 3 and 7). This suggests that the amount of OmpT is limiting in the periplasmic extracts so that partial cleavage at one or the other of two sites in colicin E3 is observed (Fig. 2B). When purified OmpT is added in excess, both sites are fully cleaved and only the shorter LC form is detected (Fig. 4A). The OmpT protease is located in the outer membrane and oriented so that its active site is extracellular (36). The limited OmpT activity we observed in periplasmic extracts may be either the consequence of its release from the inner membrane into the periplasmic space during its passage to the outer membrane (37) or due to some contamination deriving from the methods used for the preparation of the periplasmic extracts. Mass spectrometry analysis of the 47-kDa LC form localized an OmpT-dependent cleavage site between residues Arg⁴⁵⁴ and Lys⁴⁵⁵ located inside the short linker peptide (38) (Fig. 2B). The peptide motif ⁴³²RKKKEDKKR⁴⁴⁰ carries several other consensus OmpT cleavage sites, which are presumably responsible for producing the 45-kDa LC form. Two short fragments (13- and 11-kDa SC forms) detected by Western blotting with anti-colicin E3 antiserum carry the C-terminal nuclease domain of colicin E3 and correspond respectively to the 45- and 47-kDa LC products, respectively (Fig. 2B and supplemental Fig. S2).

FIGURE 4.

Comparison of the in vitro and in vivo cleavage products of colicin E3. A, [³⁵S]ColE3 was treated with periplasmic extracts prepared from wild-type (lane 4, PE wt), lepB (N274K) mutant (lane 2, PE A38), ompT-inactivated (lanes 5 and 7, ΔompT) strains, and/or purified OmpT (lanes 3 and 7). FS ColE3, 58 kDa, and the LC forms, 47 and 45 kDa, were separated on 8% SDS-PAGE and detected by phosphorimaging. B, comparison of the migration of the in vivo ColE3 PF (lane 3) with various C-terminal peptides of colicin E3 synthesized in vitro by a coupled transcription-translation Zubay-30S system (7). Peptides Lys⁴³⁹–Leu⁵⁵¹ (lane 1, 12.8 kDa), Lys⁴¹⁶–Leu⁵⁵¹ (lane 2, 15.4 kDa), and Met⁴²⁹–Leu⁵⁵¹ (lane 4, 14.1 kDa) were separated on 15% SDS-PAGE and analyzed by Western blotting with anti-colicin E3 antiserum. C, comparison of the migration of the 15-kDa in vivo processed form (ColE3 PF15, lane 1) and the 11- and 13-kDa SC forms of colicin E3 produced in vitro (lane 2), analyzed as described in B. D, comparison of cytotoxicity of colicins D ± ImmD and E3 ± ImmE3 in the presence or absence of OmpT. The in vivo halo assay was quantified by spotting aliquots of undiluted colicin (numbered 0) and serial 10-fold dilutions (numbered 1 to 5) directly onto a lawn of wild-type (wt), OmpT-deficient (ΔompT), or OmpT overexpressing (pOmpT) cells. Dark halos indicate a clear zone of growth inhibition (toxicity), and no clearing indicates resistance to colicin. Turbid zones indicate marginal toxicity.

We examined the sensitivity of OmpT-deficient strains toward nuclease colicins. The ΔompT strains (AD202, C600ompT, and the E. coli B BL21) were all shown to be fully sensitive to colicins D and E3, as observed by the cytotoxicity test (Fig. 4D), and to DNase colicin E2 (25). Therefore, OmpT is not necessary for colicin import and toxicity. We also examined the sensitivity of the wild-type and OmpT-overexpressing strains to colicins D and E3. The toxicity of these RNase colicins in complex with their cognate immunity proteins, as judged by the growth inhibition test, diminished by 1 order of magnitude when OmpT was overexpressed from a high copy number plasmid, pML19. More importantly, colicin E3, free of its Imm protein, almost completely lost its toxicity against wild-type cells. In the case of colicin D freed of its Imm protein, the wild-type cells were sensitive, whereas OmpT-overexpressing cells were resistant (Fig. 4D). Thus, the Imm proteins appear to efficiently prevent the bacterial defense system mediated by OmpT from functioning, as shown previously in the case of colicin E2 (25). Moreover, the Imm proteins may become essential for the killing action of these colicins when the OmpT activity of target cells increases.

Identification of a Processed Form of Colicin D in the Cytoplasm of Target Cells by “Ligand Fishing” with ImmD Protein

To verify that processing of nuclease colicins did occur and was required for import and toxicity, we set up an in vivo fishing assay to detect a processed form of colicin D in the cytoplasm based on its high affinity interaction with its immunity protein (27). Wild-type or OmpT-deficient bacteria expressing plasmid-borne His-tagged ImmD protein were treated with the purified colicin D-ImmD complex. To enhance the penetration of colicin D into the periplasm, target cells were treated previously with 2,2′-dipyridyl, which increases the expression of FepA and TonB (39) necessary for colicin D translocation across the outer membrane. Any colicin D fragments complexed with the ImmD-His₆ protein and fished from the cytoplasm were analyzed by Western blotting with anti-colicin D catalytic domain antiserum. The same two bands were detected in both the wild-type and OmpT-deficient strains (Fig. 5A, lanes 1 and 3). Because this antiserum cross-reacts with ImmD (see “Experimental Procedures”), the band with a lower mass corresponds to ImmD-His₆ (10.9 kDa) as verified by a specific reaction with anti-ImmD antiserum (data not shown). The slightly heavier band corresponds to the in vivo processed colicin D form (called “PF” to distinguish it from the in vitro SC form; Fig. 2A). The in vivo processed form is absent from the cytoplasm of control cells, which are not treated with colicin D (Fig. 5A, lanes 5–8 and 10), or from cells treated with colicin D but not induced with dipyridyl (data not shown). As expected, no processed colicin D form was found in tonB- or fepA-inactivated mutants, in which translocation of colicin D across the outer membrane is blocked (Fig. 5A, lanes 2 and 4). In particular, there is no colicin D processing in the tonB mutant strain D1 (R158S), which completely prevents the import of colicin D but remains competent for FepA-dependent iron uptake (Fig. 5A, lane 11) (40).

FIGURE 5.

Analysis of the in vivo processed form of the tRNase colicin D detected by fishing. A, wild-type (wt) strain or strains carrying deletions (Δ) in tonB (lanes 2 and 6), ompT (lanes 3 and 7), fepA (lanes 4 and 8), or the R158S point mutation in tonB (lanes 10 and 11), all expressing the 10.9-kDa His₆-tagged immunity protein (ImmD(6H)), were treated (lanes 1–4, 9, and 11) or not (lanes 5–8 and 10) with ColD-ImmD. His₆-tagged ImmD and bound colicin molecules were purified from the cytoplasm as described under “Experimental Procedures,” separated by 15% SDS-PAGE, analyzed by Western blotting with anti-colicin D catalytic domain antiserum, and detected by ECL. The positions of migration of the 12.4-kDa ColD PF and the 10.9-kDa His₆-tagged ImmD protein are indicated. B, N274K lepB mutant strain A38 alone (lane 1) or carrying the pLEPB(wt) plasmid (lane 2) and the FtsH-deficient strain alone (ΔftsH) (lane 3) or carrying the pFTSH plasmid (lane 4) were treated with ColD-ImmD and analyzed as described in A. C, comparison of the in vivo ColD PF obtained by “direct” immunodetection from the S100 cytoplasmic fraction of a ΔompT strain as described under “Experimental Procedures” (lanes 1 and 2) and that obtained by fishing with His₆-tagged ImmD from a ΔompT strain (lane 3). Treatment with ColD-ImmD (mg) and Western analysis are as described in A. The cloned 12.4-kDa CAT is used as the molecular mass reference.

Significantly, colicin D processing was abolished in the lepB(N274K) mutant strain A38, which specifically prevents colicin D toxicity. This processing defect was complemented by the plasmid pLEPB(wt) (Fig. 5B, lanes 1 and 2). This demonstrates that LepB does indeed have a specific functional role in the processing of colicin D.

The essential inner membrane FtsH protease was previously observed to be required for sensitivity to nuclease colicins (26). As in the case of the lepB mutant, no processed colicin D was detected in an ftsH-inactivated mutant strain (in a lethality-suppressed background of strain AR3291 (Table 1)), but the processing was fully restored when the plasmid pFTSH was introduced (Fig. 5B, lanes 3 and 4). This result raised the possibility that FtsH may be the translocator of colicin D across the inner membrane. By hijacking FtsH function, colicin D could be processed in the inner membrane, thus releasing the catalytic domain, which is able to reach its cytoplasmic target. In this scenario, the LepB interaction with colicin D could be essential and thus may facilitate the access of colicin D to the FtsH active site.

Direct Immunodetection of the Processed Form of colicin D in the S100 Cytoplasmic Fraction of Sensitive Cells Exposed to Colicin D

In the ligand-fishing experiment, the ImmD-expressing cells are protected against cell killing by colicin D treatment, but we could not exclude the possibility that the PF may be a consequence of proteolytic degradation, destined to rid the cytoplasm of the inactive colicin D-ImmD complexes formed during the import. Thus, we attempted to directly detect the processed form in the cytoplasm after a shorter (0.5–1.5 h) treatment with colicin D. The C-terminal PF of colicin D, either detected directly in the S100 fraction of colicin D-treated cells (Fig. 5C, lanes 1 and 2) or captured by using the fishing technique (Fig. 5C, lane 3), migrated to an identical position on 15% SDS-PAGE and appeared to have exactly the same migration as the 12.4-kDa cloned catalytic domain (CAT; Met⁵⁹⁰–Leu⁶⁹⁷). Comigration with the CAT implies that the in vivo processing site is located near position Met⁵⁹⁰ (Fig. 2A), which is the first residue of the CAT. It should be noted that the product (SC form) of the OmpT/LepB-dependent cleavage of colicin D in vitro is significantly smaller than 12.4 kDa (Fig. 1B). The difference corresponds to about 18 amino acids separating the start of the in vivo PF form from that of the minimal catalytic domain (Lys⁶⁰⁷–Leu⁶⁹⁷), defined in vitro as sufficient for tRNA^Arg hydrolysis (17).

Identification of the Processed Form of Colicin E3 in the Cytoplasm of Target Cells

We wondered whether proteolytic processing was a characteristic of all RNase colicins and thus looked for the processed form of colicin E3 by direct immunodetection in the cytoplasm. The same unique peptide, with a molecular mass of about 15 kDa, was detected by an anti-colicin E3 antiserum from the S100 fractions of ompT-inactivated (Fig. 6A, lanes 1 and 2) and wild-type strains (Fig. 6B, lane 8), confirming that OmpT is not involved in the in vivo processing of colicin E3. The former experiments were performed in minimal medium, in which the low level of vitamin B₁₂ was expected to increase BtuB receptor expression. The same peptide, with a 3-fold lower intensity, was present in colicin E3-treated cells grown in LB (Fig. 6A, lane 6). In both media, colicin E3 PF was detected efficiently after only 15 min of treatment of cells with colicin E3-ImmE3 (Fig. 6C, lane 4). Detection of a similar level of PF with colicin D within 15 min necessitated about 8-fold more S100 extract (data not shown). The rapidity of the appearance of the PF is consistent with our observation (Fig. 5C) that it was generated during the translocation step (allowing the entry of the catalytic domain into the cytoplasm through the inner membrane) rather than by some proteolytic degradation of longer colicin molecules that had accumulated in the cytoplasm after their translocation. Quantification of the time course of the colicin E3 PF appearance (Fig. 6C) showed that it increased 4-fold during the first hour of treatment with colicin.

FIGURE 6.

Detection and analysis of the in vivo processed form of the RNase colicin E3. A, cells grown in minimal (Min) or LB medium were treated with the quantities indicated (mg) of ColE3-ImmE3 complex. The proteins of the S100 cytoplasmic fractions (S100 ΔompT (lanes 1–3, 6, and 7) or S100 ΔtolB (lanes 4 and 5)) were separated directly on 15% SDS-PAGE and analyzed by Western blotting with anti-colicin E3 antiserum. ColE3 PF was revealed by ECL, quantified, and expressed as a percentage of the value measured in LB (lane 6, 100%). The molecular mass of reference proteins is given in kDa. FS ColE3 is shown (58 kDa). B, cells of FtsH-deficient (lanes 1 and 2, S100 ΔftsH), btuB-inactivated (lanes 3 and 4, S100 ΔbtuB), lepB mutant alone (lanes 5 and 7, S100 lepB-N274K) and carrying the pLEPB(wt) plasmid (lane 6, S100 lepB+pLEPB), and wild-type (lanes 8 and 9, S100 wt) strains grown in LB medium were treated with ColE3-ImmE3 and analyzed as described in A. ColE3 PF was quantified as in A and expressed as a percentage of the value measured in S100 wt (lane 8, 100%). C, time course of the processing of colicin E3 performed in ompT-inactivated cells grown in minimal medium and treated with ColE3-ImmE3 for the times given (min). The amount of ColE3 PF in S100 ΔompT extracts was quantified and expressed as a percentage of the amount detected after 15 min (lane 4, 100%). D, comparison of the size of ColE3 PFs detected after treatment with ColE3-ImmE3 either by fishing (lane 2) in the ΔompT-ImmE3(6H) strain grown in LB or by direct immunodetection (lane 3) from an S100 ΔompT cytoplasmic fraction. The ImmE3 protein (lanes 1 and 2) was revealed by immunodetection with an anti-His₆ tag antiserum, performed separately with the bottom (boxed) part of the same blot.

No processed forms were detected in the cytoplasm of tolB- or btuB-inactivated mutant strains (Fig. 6, A, lane 5, and B, lane 3), which is in agreement with the involvement of the BtuB receptor and the Tol system in the translocation of colicin E3 across the outer membrane. In addition, we showed that FtsH deficiency, as in the case of colicin D, prevented detection of the processed form of colicin E3 (Fig. 6B, lane 1). However, unlike colicin D, colicin E3 toxicity is not eliminated by the lepB (N274K) mutation (20). We thus unexpected to observe lower amounts of the colicin E3 processed form in extracts from the A38 strain than from the wild type (Fig. 6B, lanes 7 and 8). The amount of processed colicin E3 was restored in extracts from the A38 strain carrying the pLEPB(wt) plasmid (Fig. 6B, lane 6). This result suggests that even though LepB is not required for colicin E3 toxicity, the rate of processing might be modulated by LepB, although no interaction between colicin E3 and LepB was detected by Far Western blotting (Fig. 3B, lane 8).

The same processed fragment was detected in vivo either by using the fishing technique, with an ImmE3-His₆ protein expressed endogenously, or directly from the cytoplasm (Fig. 6D, lane 2 compared with lane 3). The comparison confirms that the PF of colicin E3, which is trapped by ImmE3, corresponds mainly to the catalytic ribonuclease domain. This in vivo detected PF has a size close to 15 kDa, as deduced from a comparison with the migration of a set of in vitro synthesized C-terminal peptides of colicin E3 overlapping the catalytic domain (Fig. 4B). This finding allowed us to estimate that the cleavage site generating PF is close to residue Asp⁴²⁰ of colicin E3 (Fig. 2B). The ColE3 PF is about 2–4 kDa heavier than the SC forms detected after cleavage in vitro in the presence of OmpT (Fig. 4C). In fact, the in vivo processing site of colicin E3 is located inside the C-terminal part of the central receptor-binding domain.

DISCUSSION

In the present work we have demonstrated, in the case of two different RNase colicins, that an endoproteolytic processing step is essential for their import into the cell and for subsequent cell killing. Using two distinct experimental protocols, we have shown for both colicins D and E3 that only their cytotoxic C-terminal catalytic domains (12.4 kDa for colicin D and 15 kDa for colicin E3) can be detected in the cytoplasm of cells exposed to the colicins. The same processed forms were detected in both sensitive or immunity-protein protected cells. The PFs found in the cytoplasm are significantly longer than the minimal size of the C-terminal domains required for catalytic activity. In the case of colicin D, the 18 additional amino acids (positions 590–607) present in the PF (Fig. 2A) were implicated in the interaction with the immunity protein (ImmD) and could also be required for tRNA target recognition (27). In the case of colicin E3, the 31 additional residues in the PF (positions Asp⁴²⁰–Lys⁴⁵⁰) (Fig. 2B) are derived from the 100 Å long hairpin structure of the receptor-binding (R) domain (38). This suggests that the R-domain should be partly unfolded to allow the RNase domain to reach the periplasmic side of the inner membrane. Such a conformational change of the R-domain may be compatible with its stable interaction with the receptor BtuB because this latter interaction involves only residues of the R-domain that are near the tip of the coiled coil structure (41). The partial unwinding of the termini of the coiled coil R-domain (41, 42) facilitates both the insertion of the N-terminal part of the translocation-domain through OmpF and then the subsequent unfolding of the C-terminal part of colicin E3, thus allowing the catalytic domain to reach the cytoplasmic membrane so that it can enter the cytoplasm (Fig. 7) (42). In agreement with a specific cascade of events during their import (depicted for both colicins in Fig. 7), we show that the processing of colicins D and E3 does not occur if their translocation across the outer membrane is prevented by mutations affecting their receptors (FepA or BtuB) or the energy transducer Ton or Tol system (Figs. 5 and 6).

FIGURE 7.

Model of the import mechanisms for RNase colicins E3 and D. Path 1a, the BtuB receptor with bound colicin E3 and the recruited OmpF porin constitute the outer membrane translocon, which requires the Tol system to drive colicin import toward the periplasm. The assembly of the translocon may lead to the unfolding of the CAT prior to its translocation across the outer membrane (Kurisu et al. (41) and reviewed by Kleanthous (55)). Path 1b, the binding of colicin D to the FepA receptor followed by the interaction of TonB with the FepA cork domain and the N-terminal part of colicin D triggers the translocation of the CAT across the outer membrane (reviewed by Braun et al. (6) and de Zamaroczy and Chauleau (9)). The translocation mechanism of the CATs is unknown, although for both colicins the Imm protein should be released first from the receptor bound colicin. Recent evidence suggests that the Imm protein release is energy-dependent (56). The crystal structure of the conserved N-terminal domain of colicin D shown bound to FepA is adapted from that of colicin B (57). Colicin E3 and possibly colicin D remain attached to their receptors, whereas the CATs traverse the periplasm and contact the inner membrane (indicated by dotted lines and a question mark). Path 2a, electrostatic association of the RNase domain of colicin E3 (CAT) with the anionic phospholipid surface of the inner membrane (16) (indicated by a pink X) may monitor the contact with FtsH. Path 2b, interaction of LepB with the central domain (CD) of colicin D is presumably an essential requirement for colicin D to enter the cytoplasm. Residues Lys¹⁴⁵ and Asn²⁷⁴ are localized in the LepB crystal structure by red and blue spheres, respectively (adapted from Paetzel et al. (58)). Path 3, the processing and translocation mechanisms of both RNase colicins are unknown but require an active FtsH (ATPase/protease) to allow the PFs to reach the cytoplasm. Path 4, the nucleases imported by PFs cleave either 16 S rRNA (colicin E3) or tRNA^Arg (colicin D), preventing protein synthesis (1, 2) and thus leading to target cell death.

Our work has shown that the inner membrane, ATP-dependent, and membrane-anchored protease FtsH (43) is essential for the processing of both colicins D and E3 and/or the translocation of the PF into the cytoplasm (Fig. 7). In particular, we show that there is a strict correlation between the loss of sensitivity of ftsH mutants to nuclease colicins, as reported previously (26), and the absence of PF in the cytoplasm of colicin-treated, FtsH-deficient cells. It was shown previously that sensitivity to nuclease colicins requires both the protease and ATPase activities of FtsH (26). Our tentative conclusion is that the FtsH endopeptidase is the catalytic enzyme required for colicin processing.

The usual biological function of FtsH consists of the dislocation and degradation of misfolded or damaged membrane proteins (43) by “pulling” them into cytoplasm (44, 45). FtsH rather than unfolding the substrates itself, preferentially acts on already unfolded proteins (46, 47). The presence of an unstructured terminal tail of membrane substrates that penetrate into the cytoplasm appears to be important in initiating FtsH-dependent proteolysis (48). It is conceivable that DNase colicins have a similar access to FtsH, because their nuclease domain exhibits an endogenous channel-forming activity, possibly allowing its self-propulsion into the cytoplasm (14, 15).

An electrostatically mediated interaction of the nuclease domain of E type colicins with the inner membrane has been reported (16). This could lead to (or maintain) a partial unfolding of the catalytic domain so that the unstructured C-terminal end may be recognized by FtsH as a substrate (Fig. 7). FtsH-dependent ATP hydrolysis could contribute to the translocation of the RNase domain toward the proteolytic active site residing in the central pore of FtsH (49, 50).

The normal mode of action of FtsH is a processive proteolytic degradation, which liberates 10–26-residue oligopeptides into the cytoplasm (43). However the in vivo cleavage of RNase colicins D and E3 generates exceptionally long C-terminal peptides of 108 and 132 residues, respectively. This major difference compared with its usual substrates may be correlated with a particular degradation pathway shown in vitro for FtsH. After translocating an internal loop of a model substrate to the protease chamber, the proteolysis is initiated from an internal site. This liberates some discrete fragments of about 30 kDa, and the proteolysis proceeds processively in the C- to N-terminal direction (50, 51).

Previously we isolated a mutation in the gene for the signal peptidase lepB (N274K) (strain A38), which renders the cell resistant to colicin D killing. We have now demonstrated that LepB plays a specific role in colicin D processing and translocation (Fig. 5). The analysis of two mutations in lepB proves that the function of LepB in colicin import is not enzymatic. K145A LepB has no catalytic activity (35), but it still restores sensitivity to colicin D of the N274K lepB mutant strain and is able to interact in vitro with colicin D almost as well as wild-type LepB (Fig. 3). In contrast, strain A38 with the N274K lepB mutation shows no apparent deficiency with respect to the processing of normal LepB substrates (17). In addition to the absence of any PF in the N274K lepB mutant strain, we also showed in vitro a large decrease in the interaction between the N274K LepB protein and the central domain of colicin D (Fig. 3). Together these results are consistent with a structural role of LepB, which could interact with the central domain of colicin D and modify the structure of the tRNase domain so as to allow proteolysis by FtsH. Specific “adaptator proteins” have been shown to be required for recognition of other soluble protein substrates by FtsH (43).

An interesting question is: why does colicin D processing require a specific interaction with LepB, but colicin E3 does not? Cytotoxicity of DNase colicins has been shown to depend upon the net positive charge of their DNase domain, essential for their interaction with the inner membrane (26). Similarly, a productive interaction of the RNase domain of colicin E3 with the inner membrane could be ascribed to its high net positive charge (+11). In contrast, the weak positive charge of the tRNase domain of colicin D (+2) may be insufficient to allow an electrostatic interaction with the inner membrane. Thus, the interaction of colicin D with the membrane may be ensured or stabilized by a direct contact between its central domain and the LepB protein at the periplasmic side of the inner membrane (Fig. 7).

It is striking that there are four parallel requirements for bacterial toxin import into mammalian cells and nuclease colicin import into E. coli (52, 53). Import into both the eukaryotic and prokaryotic systems necessitates: (i) hijacking of a quality control system of protein folding; (ii) endoproteolytic cleavage of the toxin during translocation; (iii) a chaperone protein to access the membrane translocator; and (iv) ATP hydrolysis.

In addition to colicin D, other nuclease colicins, e.g. rRNase colicin E3 (Fig. 4) and DNase colicins E7 (18, 19) and E2⁴ (25), have been shown to be cleaved in vitro by periplasmic extracts, and in all cases the in vitro cleavage was strictly dependent upon OmpT. In this work we have shown that the cleavage of colicins D and E3 by purified OmpT (or by periplasmic extracts) occurs at a different site from the authentic in vivo processing site we have mapped (Fig. 2). The OmpT-dependent colicin cleavages observed in vitro (Figs. 1 and 4) appear to reflect the role of OmpT at the cell surface as a part of the bacterial defense system to rid itself of extraneous and nefarious proteins (54). The degradation of colicin D by OmpT at the cell surface presumably requires the attachment of colicin-Imm complex to the FepA receptor (Fig. 7) and then is accelerated by the release of the immunity protein (Fig. 4). In contrast, we have shown that OmpT cleavages clearly are not involved in the import of colicins into target cells. Contrary to a previous hypothesis (19), there is no evidence for periplasm-dependent processing during the import of nuclease colicins into the cytoplasm.

It is interesting to note, however, that the productive in vivo processing of colicin D by FtsH and the in vitro observed cleavage by OmpT both require LepB. It is conceivable that these endoproteolytic cleavages of colicin D freed of its Imm protein require a similar folding of the C-terminal part of the colicin D molecule to allow access of FtsH or OmpT to their distinct but rather closely located cleavage sites preceding the catalytic domain. The appropriate conformation of the C-terminal part of colicin D for these cleavages seems to be mediated by the specific structural interaction of the colicin D central domain with LepB.