The Colicin E1 TolC Box: Identification of a Domain Required for Colicin E1 Cytotoxicity and TolC Binding

ABSTRACT

Colicins are protein toxins made by Escherichia coli to kill related bacteria that compete for scarce resources. All colicins must cross the target cell outer membrane in order to reach their intracellular targets. Normally, the first step in the intoxication process is the tight binding of the colicin to an outer membrane receptor protein via its central receptor-binding domain. It is shown here that for one colicin, E1, that step, although it greatly increases the efficiency of killing, is not absolutely necessary. For colicin E1, the second step, translocation, relies on the outer membrane/transperiplasmic protein TolC. The normal role of TolC in bacteria is as an essential component of a family of tripartite drug and toxin exporters, but for colicin E1, it is essential for its import. Colicin E1 and some N-terminal translocation domain peptides had been shown previously to bind in vitro to TolC and occlude channels made by TolC in planar lipid bilayer membranes. Here, a set of increasingly shorter colicin E1 translocation domain peptides was shown to bind to Escherichia coli in vivo and protect them from subsequent challenge by colicin E1. A segment of only 21 residues, the “TolC box,” was thereby defined; that segment is essential for colicin E1 cytotoxicity and for binding of translocation domain peptides to TolC.

IMPORTANCE The Escherichia coli outer membrane/transperiplasmic protein TolC is normally an essential component of the bacterium's tripartite drug and toxin export machinery. The protein toxin colicin E1 instead uses TolC for its import into the cells that it kills, thereby subverting its normal role. Increasingly shorter constructs of the colicin's N-terminal translocation domain were used to define an essential 21-residue segment that is required for both colicin cytotoxicity and for binding of the colicin's translocation domain to bacteria, in order to protect them from subsequent challenge by active colicin E1. Thus, an essential TolC binding sequence of colicin E1 was identified and may ultimately lead to the development of drugs to block the bacterial drug export pathway.

INTRODUCTION

Escherichia coli competes for scarce resources by making plasmid-encoded protein toxins called colicins, which efficiently kill closely related bacteria. All of the few dozen or so colicins that have been identified kill their targets by one of a few basic mechanisms: (i) making an ion-permeable channel in the inner membrane of the target cell, which depolarizes and kills the cell (1); (ii) enzymatically cleaving its rRNA, tRNA, or DNA in the cytoplasm (2, 3); or (iii) degrading peptidoglycan precursors in the periplasm (4, 5). Regardless of their ultimate killing mechanism, all colicins must cross at least the outer membrane in order to reach their targets. The way one particular colicin, E1, makes that transit is the subject of this study.

Colicins have a well-defined domain structure, with the killing (catalytic or channel-forming) domain at the C-terminal end, a receptor-binding (R) domain in the central part of the molecule, and a translocation (T) domain encompassing the N-terminal part of the protein (Fig. 1A to C). For the first step in killing target bacteria, the colicins have evolved to cannibalize as their primary high-affinity cell surface receptors one of a small number of outer membrane proteins (FhuA, FepA, BtuB, and Cir) normally used by the target bacteria for the uptake of essential nutrients, such as siderophore-bound iron or cobalamin. These receptors are all 22-stranded β-barrels with an N-terminal periplasm-facing plug that fills the barrel (6, 7). Structures have been solved for a number of these receptors with bound colicin R domains: colicins E3 and E2 bound to their BtuB receptor (8, 9), and colicin Ia bound to Cir (10). Once the colicin is bound at the cell surface, however, it must still cross the membrane on which it is bound. That subsequent step of intoxication is known as translocation and is mediated by the N-terminal portion of the colicin molecule. Also involved is an outer membrane translocator protein, as well as one of two families of E. coli inner membrane and periplasmic proteins, either the Tol proteins for group A colicins, or the TonB, ExbB, and ExbD proteins for group B colicins (11, 12). It is this translocation step of colicin E1 intoxication that is the subject of this study.

FIG 1.

Ribbon diagrams of colicins E3 and Ia and of TolC trimer. (A) Domain arrangement of colicins with N-terminal translocation (T, blue) (residues 1 to 315 for E3; residues 1 to 225 for Ia), central receptor-binding (R, green) (residues 316 to 450 for E3; residues 226 to 449 for Ia), and C-terminal cytotoxicity (C, red) (residues 451 to 551 for E3; residues 450 to 626 for Ia) domains. (B) Ribbon diagram of group A Tol-dependent colicin E3 (PDB 1JGH, 3.0 Å [48]). Immunity (Imm) protein is copurified with colicin E3, shown in yellow. (C) Group B TonB-dependent colicin Ia (PDB 1CII, 3.0 Å [49]). Note the extended coiled-coil R domains of both colicins. (D) Ribbon diagram of TolC trimer (PDB 1EK9, 2.0 Å [24]), forming a 12-stranded β-barrel in the outer membrane (OM) and an α-helical ∼100-Å-long extended tunnel in the periplasm (25). Gly43 is labeled, near the bottom of the β-barrel; when G43 is mutated, colicin E1 cytotoxicity is decreased (29).

Target cell outer membrane and periplasmic proteins have been coopted for translocation, and for some colicins, the process is reasonably well understood. For the nuclease E colicins, E2, E3, and E9, once the colicin is bound via its R domain on the BtuB receptor, the elongated Y-shape of the colicin (Fig. 1B) facilitates a search in two dimensions, via lateral diffusion and a “fishing pole” mechanism, by which the highly unstructured N-terminal T domain finds a copy of its OmpF translocator (8, 13). For colicins E3 and E9, segments of their T domains were shown to be bound inside the pore of OmpF (14–16), and their T domains have also been shown to occlude OmpF channels in planar lipid bilayer membranes (16, 17). For colicin Ia, a pore-forming colicin, a second copy of its Cir receptor serves as its translocator (10, 18). The colicin's R domain binds to Cir with high affinity, allowing for a more efficient search, while anchored at the cell surface, by its T domain for another nearby copy of Cir, through which it transits into the periplasm by an as-yet poorly understood mechanism that must involve some movement of the plug domain of Cir. Cir is therefore playing two roles in colicin Ia uptake, both as its primary high-affinity receptor and as its translocator.

Thus, a current model envisions the unstructured T domain entering its translocator and passing through it into the bacterial periplasm, where it interacts with one of two homologous groups of inner and periplasmic proteins (TolA, TolB, TolQ, and TolR or TonB, ExbB, and ExbD) which somehow provide the energy to both detach the colicin from its primary receptor and move it across the outer membrane through its translocator (reviewed in reference 19). Giving weight to this model, Housden et al. (16) have demonstrated the existence of a translocon complex consisting of the E9 T domain, the OmpF trimer, and the periplasmic protein TolB, which binds the E9 N terminus after it crosses the outer membrane through OmpF.

The pore-forming colicin E1 shares the property of all the E colicins in using the vitamin B₁₂ transporter BtuB as its primary receptor in the outer membrane. However, it is unique among the E colicins in using TolC, rather than OmpF, as its translocator pathway into the target cell. TolC is normally a component of the bacterial drug and hemolysin efflux machinery, a partner in several classes of tripartite multidrug efflux machines (20–23). Thus, colicin E1 uses TolC for its import, reversing its normal role as part of an export pathway. The 3-dimensional structure of TolC has been solved (24). It is a homotrimer that assembles into what has been described as a 12-stranded “channel-tunnel,” over 140 Å long, with a β-barrel in the outer membrane connected to a 12-stranded α-helical coiled coil that crosses the peptidoglycan barrier and reaches into the periplasm (Fig. 1D) (25). The single pore of the β-barrel in the outer membrane has an inside diameter of ∼20 Å, but the relatively open pore of TolC in the β-barrel tapers to a more narrowly constrictive opening of ∼4 Å that resides in the periplasm and must somehow open in an iris-like movement to allow the passage of substrate molecules (22, 24). Recent crystallographic and mutagenic studies suggest that interactions with one or more of its periplasmic partners, such as the AcrA-AcrB complex, drive TolC into its more open configuration (26, 27) to allow the passage of protein or drug substrates through its channel.

Just as the T domains of enzymatic colicins have been shown to bind to and enter their OmpF translocator, the T domain of colicin E1 occludes TolC channels in planar lipid bilayer membranes (17). Binding of some somewhat-shorter E1 T domain peptides was also detected by similar occlusion of TolC channels in planar bilayers and by coelution of those peptides with TolC on a sizing column (28). Mutations of a glycine residue known to face the inside of the TolC channel have been shown to reduce the efficiency of killing by colicin E1, implying that passage of the colicin through the channel was obstructed by the bulkier mutant amino acids (29). However, the nature of the interaction of colicin E1 with TolC and the mechanism of its passage through the periplasm, via translocation using TolC, are presently poorly understood.

In preliminary in vivo experiments, it was shown that neither the receptor-binding central domain of colicin E1 nor its BtuB receptor protein is absolutely required for cytotoxicity by colicin E1 (30). Those results implied that the in vitro interaction of the colicin E1 translocation (T) domain with TolC, as previously reported (28), could support both binding and translocation of the colicin without a prior tight binding via the receptor-binding domain/receptor interaction. Similar binding via T domain alone and subsequent cytotoxicity by colicin Ia have also been demonstrated (18). In that system, a colicin Ia construct whose receptor-binding domain had been deleted retained cytotoxicity that was completely dependent upon the presence of the Cir protein in the target cell. Furthermore, the isolated colicin Ia translocation domain was shown to bind to the Cir translocator protein and block subsequent killing by added active colicin Ia.

In this study, in vivo protection of sensitive E. coli from killing by colicin E1 by a series of increasingly shorter T domain peptides identified a short sequence of no more than 21 residues, near the middle of the classically defined translocation domain and which is required for cytotoxicity by colicin E1. Subsequent recently published biophysical measurements using this same set of peptides showed that this sequence is also required for the in vitro binding of the T domain to TolC, as measured by either occlusion of TolC channels in planar lipid bilayer membranes or binding to a size exclusion column (31). Circular dichroism and thermal stability measurements suggested that in solution, the T domain forms a helical hairpin, with the critical 21-residue segment near its center and acting as a sort of hinge (31) that may play a role in conformational changes that occur during translocation of the colicin.

The identification of a short critical TolC binding sequence has the potential for the development of drugs that could inhibit the bacterial drug export pathway, an important player in the growing problem of antibiotic resistance.

RESULTS

Binding to its BtuB receptor is not absolutely required for killing by colicin E1.

As was first reported in a meeting summary (30), E1ΔR, an E1 colicin from which the receptor-binding domain (residues 191 to 337) had been deleted, still kills wild-type E. coli, albeit with a specific activity of about 500- to 1,000-fold less than that of full-length colicin E1, which has ∼10⁶ to 10⁷ killing units · mg⁻¹. Consistent with that result, wild-type colicin E1 can kill a btuB mutant which lacks the primary receptor, although with efficiency reduced by ∼10⁴-fold relative to the wild-type colicin indicator (Fig. 2). Thus, even without binding to its high-affinity BtuB receptor, some colicin E1 can gain access to the inner membrane of target cells. That access must be via its TolC translocator, since a tolC mutant is barely killed by 1 mg · ml⁻¹ wild-type colicin E1 and is not killed at all by E1ΔR (Fig. 2). These results suggest that E1 can bind, albeit with relatively low efficiency, to TolC on the outer membrane without first attaching via its R domain to BtuB on the cell surface, and that binding to TolC is sufficient to initiate translocation of the colicin across the outer membrane. In fact, in the absence of functional BtuB, the R domain deletion mutant E1ΔR has a specific activity ∼10-fold greater than that of full-length E1 (Fig. 2). The greater specific activity of E1ΔR relative to wild-type colicin E1 was also clear in experiments in which the growth of colicin-sensitive E. coli was monitored in liquid cultures after addition of the colicins. While 0.15 μg/ml wild-type colicin E1 did not inhibit growth of cells lacking BtuB, the same concentration of E1ΔR did (see Fig. S1A and B in the supplemental material). This may simply be a reflection of more efficient translocation of the significantly shorter protein through TolC.

FIG 2.

Dependence of killing efficiency by wild-type colicin E1, E1ΔR, and wild-type colicin E3 on the presence of BtuB, TolC, and TolA. Ten-microliter spots were made on lawns of E. coli spread in melted agar with 10-fold dilutions of 1 mg · ml⁻¹ solutions of purified colicins. Wild-type or mutant indicator strains are noted on the x axis. Killing units per milligram of colicin are the inverse of the highest dilutions of a 1 mg · ml⁻¹ colicin solution at which any clearing of the lawn was observed. The arrows indicate that no growth inhibition was observed at the highest colicin concentration tested, 1 mg · ml⁻¹. The btuB::Tn10 mutant does not express intact BtuB protein; the tolC::Km (Km, kanamycin) mutant has a total deletion of TolC; the tolA mutant is a point mutant (see Materials and Methods). Error bars represent the range from three independent experiments.

Colicin E3, an RNase colicin that also uses BtuB as its primary receptor but OmpF as its translocator, is much more dependent on its initial binding via BtuB; its specific activity on the btuB mutant is reduced by five orders of magnitude on the BtuB-lacking indicator lawn (Fig. 2). This suggests that the binding of E1 to its TolC translocator is more efficient than that of E3 to its OmpF translocator.

Both E1 and E3 are strongly dependent on TolA (5), which is thought to transduce energy from the inner membrane and somehow provide the energy to move the colicin across the outer membrane (19). However, colicin E1, but not E1ΔR, is somewhat less dependent upon a functional TolA protein than is colicin E3.

The reduced cytotoxicity of colicin E3 on the TolC mutant is likely due to the fact that the synthesis of OmpF, the colicin E3 translocator protein, has been shown to be severely depressed in tolC mutants (32).

T domain peptides protect sensitive E. coli from killing by colicin E1 or E1ΔR.

The ability of colicin E1 to kill E. coli in the absence of its primary BtuB receptor or receptor-binding domain (10³ to 10⁴ units · mg⁻¹) (Fig. 2), as well as the reported binding to TolC in vitro of both intact colicin E1 and a translocation domain construct composed of residues 41 to 190 (28), suggested that the translocation domain (T domain) could bind independently to E. coli via TolC. Therefore, the ability of an isolated T domain to bind to cells in vivo and thereby protect them from killing by either active full-length colicin E1 or E1ΔR was investigated. A full-length T domain, consisting of residues 1 to 190 of E1, did protect cells from E1ΔR. Increasingly shorter T domain peptides were then constructed and tested for their ability to block killing by the colicin. Peptides ending at residue 100 did not protect against killing by E1ΔR, while those ending in residue 120 or 140 did (Tables 1 and 2; see also Tables S2 to S9). In similar experiments, deleting at least 80 or even 99 residues from the N terminus of the T domain did not prevent such peptides from protecting cells from killing by colicin, as long as the residues between 100 and 120 or 140 were present (Tables 1 and 2). Figure 3 shows the titrations of protection from killing by E1ΔR, comparing the T domain residues 57 to 120 (T57–120) with T57–140. Near-total protection was reached with ∼2 to 4 mol either T57–120 or T57–140 per mole E1ΔR, with partial protection at an equimolar peptide (Fig. 3). Once it had been established that peptide T57–140 protected fully at a molar excess of at least 4 mol T domain/mol E1ΔR, that peptide was generally included as a positive control in subsequent experiments with other T domain peptides (Tables S3 and S5 to S7). Similarly, T1–100, which did not protect, was included as a negative control. In fact, addition of the nonprotective peptide actually enhanced killing, an effect of an unknown mechanism that has been observed before when using bovine serum albumin as a control for added protein in similar in vivo protection experiments with colicin Ia and its T domain (18). From the combined experiments collated in Tables 1, 2, and S2 to S9, it can be concluded that residues 100 to 120 comprise an essential TolC binding domain of colicin E1. It should be noted that when each T domain peptide was initially tested, the maximum amount that could be added to each reaction was used, based on the yields of the individual proteins. Therefore, the molar excess of T domain over colicin varied in the experiments shown in Tables 1 and S2 to S9. Representatives of each class of peptide were subsequently assessed for their ability to protect E. coli in vivo from cytotoxicity by colicin E1ΔR in a single experiment, where all peptides were added at 100-fold molar excess over the colicin (Table 2).

TABLE 1.

Summary of colicin E1 T domain peptide protection of E. coli from cytotoxicity by colicin E1 or E1ΔR^a

Peptide	Protection
T1–190	+
T1–190 (Δ100–120)	−
T1–140	+
T100–140-His10 tag	+
T1–120	+
T1–100	−
T41–190	+
T41–140	+
T41–120	+
T57–190	+
T57–140	+
T57–120	+
T57–120-His10 tag	+
T81–140	+
T100–190	+
T139–190	−
T100–120 synthetic peptide	−
T57–140 R103Q	+
T57–140 R108Q	+
T57–140 R103Q-R108Q	+
T57–140 H104L-H117L	+
T57–140 R103Q-R108Q H104L-H117L	+

^aProtection of exponentially growing colicin-sensitive E. coli cells from cytotoxicity by colicin E1 or E1ΔR was determined as described in Materials and Methods in separate experiments done with groups of peptides as they were made (Tables S2 to S9), and the results are summarized here. With the exception of the synthetic peptide from residues 100 to 120 and the two peptides with “His₁₀ tag” in the designation, all of the peptides had a 21-residue thrombin-cleavable C-terminal sequence that included a His₁₀ sequence, which resulted from their expression from pET52-b. Mann-Whitney tests of representative data sets for T57–140 and T1–190 showed those denoted as protecting (+) to be significantly different from the controls to which no T domain was added, with P of 0.0167, while T1–100 was not significantly different from the untreated controls.

TABLE 2.

Representative colicin E1 T domains protect sensitive E. coli from cytotoxicity by E1ΔR^a

Treatment	Addition	mol T domain/mol colicin	% surviving colonies
None	None	0	100.0
E1ΔR	None	0	0.27
E1ΔR	E1 T1–190	100	4.0
E1ΔR	E1 T1–190Δ100–120	100	0.0007
E1ΔR	E1 T1–120	100	64.0
E1ΔR	E1 T100–140	100	71.0
E1ΔR	E1 T100–140-His10 tag	100	30.0
E1ΔR	E1 T1–100	100	0.0007
E1ΔR	E1 T57–120	100	59.0
E1ΔR	E1 T57–120-His10 tag	100	94.0
E1ΔR	E1 T81–140	100	74.0
E1ΔR	E1 T57–140 R103Q-H104L-R108Q-H117L	100	45.0
E1ΔR	T100–120 synthetic peptide	100	0.13

^aLog-phase wild-type indicator K361 cells were preincubated for 5 min at 37°C with shaking, with the additions shown, in a total volume of 200 μl. Colicin E1ΔR was added to a final concentration of 0.12 μM, and incubation was continued for 20 min at 37°C with shaking. Cells were diluted immediately, appropriate dilutions were plated, and the percentage of surviving colonies was determined based on the control culture to which no colicin was added.

FIG 3.

Titration of protection from E1ΔR cytotoxicity by T57–120 and T57–140. Logarithmically growing E. coli cells were incubated with T domain peptides in the amounts shown for 5 min at 37°C. Then, colicin E1ΔR was added to a final concentration of 0.23 μM, and incubation was continued for 20 min at 37°C. Reactions were stopped by diluting an aliquot 100-fold in buffer, and then appropriate dilutions were spread on nutrient agar. The percentage of surviving colonies was calculated from control cultures to which no colicin or T domain was added. The data were plotted in Prism with nonlinear fit as plateau, followed by one phase association. The R² was 0.929 for T57–120 and 0.958 for T57–140.

In order to assess the protective efficiency of the T domain peptides as accurately as possible, E1ΔR was generally used, rather than full-length E1, for the protection experiments in liquid cultures. That way, the colicin was not initially bound at the cell surface before “searching” the cell surface in only two dimensions for its TolC translocator. Instead, both the active colicin and the T domain must search in three dimensions from the culture medium for a TolC to which to bind on the cell surface. While ∼400 times more E1ΔR than intact colicin E1 was used to achieve about the same level of killing (generally ∼0.03 to 1.0% surviving colonies), a much lower molar excess of T domain over E1ΔR than wild-type (WT) colicin gave greater protection from killing. Nonetheless, the T domain did protect sensitive E. coli from cytotoxicity by wild-type colicin E1 in a concentration-dependent manner (Fig. 4).

FIG 4.

Titration of protection from colicin E1 cytotoxicity by T57–140. Log-phase wild-type indicator cells were preincubated for 5 min at 37°C with shaking, with no added T domain or T57–140 added in the amounts shown, in a total of 200 μl. Colicin E1 was then added to a final concentration of 4.4 × 10⁻¹⁰ M, and incubation was continued for 20 min at 37°C with shaking. Cells were diluted, appropriate dilutions were plated, and the percentage of surviving colonies was determined, based on a control culture to which no colicin was added. The data were plotted in Prism with nonlinear fit as plateau, followed by one phase association. R² = 0.9412.

In absolute amounts, ∼50% protection from wild-type colicin E1 cytotoxicity required ∼15- to 30-fold more T domain peptide than with protection from cytotoxicity by E1ΔR (compare Fig. 3 and 4). However, since much less of the intact colicin was used to achieve the same degree of killing, the molar excess of T domain over colicin was much greater. This apparent lower T domain efficiency of protection can be explained by the fact that the cytotoxicity of wild-type E1 is ∼1,000-fold more efficient than that of E1ΔR, by virtue of its ability to be anchored at the cell surface via binding to BtuB.

While there was some variability from experiment to experiment in the level of killing of sensitive cells by E1 or E1ΔR, which seems to result from some loss of activity over time with storage at −20°C, the difference between peptides that prevent killing and those that do not was very clear (Tables 1, 2, and S2 to S9).

It was not possible to make a plasmid-encoded construct as small as T100–120, however, because (i) deleting the segment between residues 120 and 140 so dramatically decreased the peptide yield of the shorter peptide of each pair (for example, T1–120 versus T1–140, or T57–120 versus T57–140), although the growth of the producing cells was normal, and (ii) the yields further declined as the fragments got smaller. Thus, T81–120 yielded barely visible protein on a polyacrylamide gel, while T81–140 yielded almost 100 times more protein from the same protocol but much less total protein than T57–140 or longer peptides. However, there was no evidence of degradation products from peptides ending in residue 120 on polyacrylamide gels of the purified peptides (not shown), nor were any of the plasmids toxic to the producing cells. In addition, the lack of any residues beyond W57 that absorb at 280 nm meant that protein detection during purification was not possible. However, in each pair that was made, regardless of its N terminus, the peptides ending in residue 120 protected cells from cytotoxicity by colicin E1, as did the peptides ending in residue 140. Therefore, the smallest cloned peptide that was purified and tested was T100–140, and it protected efficiently against killing by E1ΔR. Surprisingly, a synthetic peptide consisting only of residues 100 to 120 did not protect cells in vivo from cytotoxicity by colicin E1ΔR (Tables 1, 2, and S8).

However, the critical role of the segment of T domain from 100 to 120 was confirmed by deleting only that segment, both from T1–190 and from intact colicin E1. Deletion of that 21-residue segment from T1–190 completely eliminated its ability to protect sensitive E. coli from cytotoxicity by colicin E1 (Tables 1, 2, and S8). Deleting the same 21 residues from full-length colicin E1 also completely destroyed its cytotoxicity. The deletion of residues 100 to 120 greatly reduced the yield of purified protein relative to the yield of wild-type colicin E1. As was the case for the reduced yields of T domain peptides ending in residue 120, however, there was no evidence of shorter degradation products in the purified fractions, and the resulting protein did not make spots on a lawn of sensitive E. coli, even at the highest concentration tested, 0.64 mg/ml (not shown). I therefore conclude that residues 100 to 120 contain an essential binding epitope of E1 for its TolC translocator and define it as the TolC box.

As the synthetic T100–120 peptide had no C-terminal histidine tag, unlike all of the tested T domain peptides, the tag was removed from both T100–140 and T57–120 by thrombin digestion, in order to determine whether the tag sequence was involved in TolC interaction. Removal of the His₁₀ tag did not affect the ability of those peptides to provide in vivo protection from colicin E1 cytotoxicity (Tables 2 and S9).

Effect of its positively charged nature on the binding of the TolC box to TolC, in vivo.

It had been suggested previously that the preponderance of positively charged residues in the critical TolC binding domain might be important to its binding to TolC (33). The basic residues between 100 and 120, Arg103 and Arg108, were therefore mutated, separately and together, to glutamine. Neither the R103Q or R108Q single mutant, the R103Q-R108Q double mutant, or even the R103Q-H104L-R108Q-H117L quadruple mutant affected the ability of peptide T57–140 to provide protection against the cytotoxicity of E1ΔR (Tables 1, 2, and S7). Electrostatic interaction between basic residues in the TolC box and the interior of TolC must therefore not be determinative to the binding.

DISCUSSION

By creating ever-shorter peptides from the 190-residue N-terminal T domain of colicin E1, a sequence has been identified of no more than 21 residues, from residues 100 to 120, that is required for in vivo cytotoxicity. Peptides that contain that sequence are capable of binding to sensitive E. coli cells and protecting them from subsequent killing challenge by active full-length colicin E1 or E1ΔR. Deleting that segment alone abolishes that binding and thereby abolishes in vivo protection. Surprisingly, the T100–120 peptide alone could not provide in vivo protection. It is clear that no sequence N terminal or C terminal to that segment is absolutely required for those interactions, since residues up to 100 (as in the case of T100–140) and after 120 (as for T57–120) are not required for in vivo protection (Tables 1 and 2). Given that there is a ring of critical negatively charged residues in the interior of TolC (24), it seemed possible that basic residues between 100 and 120 help position the TolC box sequence within TolC. However, mutating all of the basic residues within the putative TolC box did not eliminate in vivo protection from cytotoxicity, so those basic residues are not required for that interaction. It is also possible that the only requirement for TolC box binding is some secondary structure provided by sequence on one side of the 21-residue peptide that was tested, since both T57–120 and T100–140 could bind and protect. Using this same set of T domain peptides I provided to them to determine that the T domain assumes a hairpin structure, with residues 100 to 120 as a hinge in its middle, Zakharov et al. (31) suggest that the hairpin structure itself is necessary for the insertion in TolC, and that the randomly coiled T100–120 synthetic peptide itself cannot bind to TolC. Other models will be discussed below.

Residues 100 to 120 are thus necessary, but not sufficient, for binding to TolC. In the absence of a 3-dimensional structure of the E1 TolC box bound to TolC, neither the interaction site within TolC nor the structure or position of the bound T domain can be speculated upon. Given its central role in multidrug export, it was disappointing that the 21-residue synthetic peptide did not block TolC; it was thought that such a peptide might be used to inhibit TolC's role as a drug pump.

Multiple interactions of colicin proteins with a number of proteins in their target bacteria are required for the binding and uptake of the colicins, so that their killing domains ultimately reach their targets in the cell. Normally, the first of those interactions is the tight binding of the central domain of the colicin to its outer membrane receptor. It was shown here that that first interaction is not absolutely required for cytotoxicity by colicin E1. Deletion of either the receptor-binding domain of the colicin or of the bacteria's BtuB receptor reduces cytotoxicity by three or four orders of magnitude, but it does not totally prevent killing by the colicin. In contrast, E. coli strains devoid of the TolC translocator are virtually insensitive to colicin E1. Those results suggested that significant binding to TolC by E1 could occur directly, without prior binding to the BtuB receptor. Receptor binding must therefore serve principally to concentrate the colicin at the cell surface and make subsequent essential steps of intoxication more efficient, although such binding may also initiate some degree of unfolding to make the T domain more accessible, as suggested by crystal structures of colicins E2 and E3 bound to BtuB (8, 9).

The finding in the experiments reported here that direct binding of the colicin T domain can occur in vivo without prior interaction with BtuB, as detected by the protection of sensitive bacteria by T domain peptides from killing by colicin E1, contradicts the conclusion of Masi et al. (29). They could detect in vivo cross-linking of E1 to TolC by dithiobis(succinimidyl propionate) (DSP) in cells with a functional BtuB colicin receptor but not in btuB mutant cells. However, the in vivo killing assay employed for the present study may be more sensitive to the interaction, since the chemical cross-linking assay requires subsequent protein purification and detection by Western blotting.

Similar in vivo binding of the colicin Ia T domain to E. coli and concomitant protection from cytotoxicity by colicin Ia have been shown previously for the TonB-dependent colicin Ia (18). In the case of colicin Ia, its receptor/translocator, to which both its R and T domains bind, is the plugged β-barrel protein Cir. But unlike previous results with colicin E1 and TolC and those shown here (17, 28, 31), in vitro binding of the Ia T domain to Cir in planar lipid bilayer membranes could not be demonstrated (E. Udho, K. S. Jakes, and A. Finkelstein, unpublished data).

Earlier work by Pilsl and Braun had localized the TolA binding site of E1 to the region between residues 1 and 44 by swapping N-terminal segments between highly homologous colicin E1 and colicin 10, in order to ascertain where the specificity for Tol- or TonB-based uptake resided (34). Pilsl and Braun had also noted that the TolC binding region must lie downstream of the TolA binding sequences where colicins E1 and 10 are nearly identical, since both require TolC for uptake. In this work, I have now localized that downstream sequence for colicin E1 between residues 100 and 120 and refer to it as the TolC box.

The N-terminal ∼40 residues of colicin E1 are highly glycine rich and are therefore predicted to have little or no secondary structure. This region would therefore require little or no energy to enter the 20-Å-wide β-barrel opening of TolC in the outer membrane. In a manner analogous to the entry of the disordered N terminus of colicin E3 or E9 into an OmpF pore or entry of the disordered N terminus of colicin Ia into its Cir translocator (35), the disordered ∼40-residue N terminus of colicin E1 is assumed to enter the wide outer membrane β-barrel opening of TolC. Passage through TolC would be arrested by the binding of the TolC box at an as-yet-unidentified site within TolC. In an extended conformation, assuming an average backbone length of ∼3.6 Å per amino acid (36), those 40 residues could span ∼140 Å, which is about the length of the entire TolC molecule (37). Thus, in extended form, the N terminus of colicin E1 could theoretically reach through TolC, passing through both the ∼40-Å width of the outer membrane where the β-barrel portion of TolC resides, and across much of the periplasmic space as well. If the binding site for the TolC box is further inside the TolC channel, the N terminus could extend even further into the periplasm. The experiments of Masi et al. suggest that the binding site may be extracellular to the position of TolC residue G43, which is near the juncture of the β-barrel and α-helical domains of TolC, since G43R mutants of TolC can still be cross-linked with colicin E1 (29) in vivo. The problem remains, however, as to how even an extended polypeptide chain could exit the narrow (3.9-Å-diameter) iris-like constriction at the periplasmic end of TolC (37). In its role in drug or toxin export, the inner membrane/periplasmic partners of TolC provide the energy, via inner membrane proton antiport or ATPase activities, combined with a periplasmic adaptor, to open the iris of TolC where it reaches near the middle of the periplasmic space (25). However, none of the usual periplasmic partners of Tol C, such as AcrAB, MacAB, or EmrAB, which are involved in E. coli efflux pumps, have ever been implicated in E1 import or intoxication (11, 12, 22, 38). Conceivably, binding of the TolC box itself could trigger a small conformational change in TolC sufficient to loosen the aperture and allow further passage of the unfolded T domain of E1. In that case, the N-terminal end of colicin E1, which carries the TolA box, could be delivered directly to the vicinity of TolA in the middle of the periplasm, where the exit of TolC lies (25). In fact, its use of TolC could explain the insensitivity of colicin E1 toxicity to deletion of much of the helical portion of TolA. Colicins A, E3, and N have virtually no cytotoxic activity on cells with a deletion of ∼90% of the α-helical portion of TolA, whereas colicin E1 activity is only slightly decreased, although all of those colicins require TolA for translocation (39). I speculate this is because, in the case of colicin E1, its N-terminal TolA binding site emerges from TolC in the middle of the periplasm, not at the inside of the outer membrane where the OmpF translocator of colicins A, E3, and N presents their N termini. Therefore, in the case of those colicins, TolA requires its full length to extend across the periplasm to reach its binding partner on the colicin or on TolB (40). Since the TolC exit is much further across the periplasm, the colicin-binding domain of TolA can much more easily reach the colicin's TolA binding box where it exits TolC, and the full length of the TolA helical domain can be shortened without appreciably affecting the cytotoxicity of colicin E1. Movement of the entire colicin molecule through TolC could then be facilitated by TolA, along with its inner membrane proton motive force (PMF)-driven TolQ and TolR partners. This model for translocation through TolC assumes an unfolded conformation for the T domain entering TolC. This is in contrast to a model proposed by Zakharov et al. (31) which envisions the T domain retaining the helical hairpin structure obtained in solution with isolated T domain peptides and both the N and C termini of the colicin facing the outside of the target cell. However, the TolC box, residues 100 to 120, was shown in that work to function like a hinge. The presence of the TolC box sequence serves to weaken interhelical interactions within the T domain and could facilitate its conversion to a more extended conformation that could then enter TolC as described above. For the DNase colicin E9, it has been demonstrated that the application of very low forces (<20 pN) is sufficient to trigger the dissociation of its immunity protein, which has a K_d (dissociation constant) in the femtomolar range (41). It can be postulated that some comparably weak interaction between colicin E1 and TolC triggers its release from its BtuB receptor and an unfolding that would allow binding to and passage through TolC so that the channel-forming domain reaches its ultimate target, the bacterial inner membrane. However, without crystallographic structures of the E1 T domain or a peptide fragment of it bound to TolC, these mechanistic proposals remain purely speculative.

MATERIALS AND METHODS

Bacterial strains.

The wild-type indicator for assaying colicin killing was E. coli K361, which is the classical strain W3110 rpsL (CGSC 4474; F⁻ λ⁻) (42) from our lab collection; the insertion mutant strain K1022 btuB::Tn10, which does not express BtuB, was derived from K361, as described previously (18); the tolA point mutant is Luria strain A592, obtained as CGSC 4923 from Yale (tolA1 F⁻ thi-1 thr-1 leu B6 lacY1 tolA1 tonA21 λ⁻ supE44); and the tolC::Km^r total deletion mutant RAM1129, derived from MC4100 [F⁻ araD139 Δ(argF-lac)U139 rpsL150 flbB5301 ptsF25 deoC1 thi-1 rbsR relA] (43), was a kind gift of R. Misra.

Plasmid construction.

A plasmid encoding colicin E1ΔR and the E1 immunity protein (ImmE1) was constructed as follows. The receptor-binding (R) domain, encompassing amino acid residues 191 to 337, was deleted from the colicin E1 gene in pKSJ331, which has the genes for colicin E1 and ImmE1 cloned in pUC19 (44), using the QuikChange site-directed mutagenesis kit (Agilent Technologies). This construct, named pKSJ373, gave a poor yield of badly degraded or impure E1ΔR, so the genes for ImmE1 and E1ΔR were subsequently cloned into pET52-b (Novagen/EMD Millipore) so that the colicin deletion protein would be overexpressed with a cleavable C-terminal His₁₀ tag for subsequent purification by metal chelation chromatography. Since cells bearing a gene for colicin must also carry the ability to synthesize the cognate immunity protein for that colicin (45), the gene for ImmE1 was first cloned with its own promoter sequence by PCR from pKSJ373, with AvrII sites at either end, and inserted into the unique AvrII site in pET52-b, creating pKSJ375. The resulting plasmid confers immunity to E1 on cells bearing it. The gene encoding E1ΔR was then cloned with a 5′-terminal NcoI site and a 3′-terminal SacI site by PCR from pKSJ373 between the NcoI and SacI sites in pKSJ375. The resulting plasmid, pKSJ376, has the gene for E1ΔR with a C-terminal His₁₀ tag, transcribed under the control of the inducible pET52-b promoter, and the gene for ImmE1, transcribed in the opposite direction and under the control of its own constitutive promoter.

All of the T domain constructs listed in Table 1 and in Tables S1 to S9 in the supplemental material, with the exception of T100–120, were made by PCR amplification of the relevant colicin sequences from pKSJ331, with an NcoI site at the 5′ end and an SacI site at the 3′ end. The primers used are shown in Table S1. pET52-b purchased from Novagen/EMD Millipore was cut with the same two enzymes, treated with calf intestine alkaline phosphatase, and gel purified before ligation, using T4 DNA ligase, of the appropriate T domain nucleotide sequence with sticky ends compatible to the cut plasmid. The constructs were verified by DNA sequencing and expressed in E. coli BL21(DE3) cells, as described below.

Mutagenesis of T domain peptides and colicin E1 gene.

Single amino acid residues in T57–140 were mutated using the QuikChange site-directed mutagenesis kit (Agilent Technologies). Separate mutagenic oligonucleotides were made for the R103Q and R108Q mutations; the double mutant was made using a single pair of longer oligonucleotides that created both mutations simultaneously (Table S1). Mutations of H104L and H117L were done both in the background of wild-type T57–140 and the double mutant with R103Q and R108Q. Two pairs of mutagenic oligonucleotides were used for each pair of histidine-to-leucine mutants (Table S1).

The same mutagenesis kit was used to delete residues 100 to 120 from the colicin E1 gene in pKSJ331, using an appropriate pair of mutagenic oligonucleotides to create the deletion (mutagenic primers are shown in Table S1). The resulting plasmid DNA (called pKSJ410) was sequenced to confirm the deletion. The same mutagenic oligonucleotides were used to delete residues 100 to 120 from T1–190 in pET52-b, which had been constructed as described above.

Protein purification.

Colicin E1 with residues 100 to 120 deleted was expressed and purified from pKSJ410 in TG1 cells grown in fortified broth and induced with mitomycin C, as described previously (44, 46, 47); wild-type colicin E1 was expressed and purified in the same way.

E1ΔR and T domain constructs were expressed from BL21(DE3) bearing the appropriate plasmid grown at 37°C in Overnight Express instant TB medium (EMD/Millipore) and purified by nickel chelation chromatography, essentially as described previously (18). Most of the peptides were ultimately dialyzed into 25 mM sodium borate and 150 mM NaCl (pH 9) and stored at −80°C. T100–190 was stored in 20 mM Tris-Cl and 150 mM NaCl (pH 7.0).

None of the T domain peptides that begin after residue 80 have any amino acids that absorb in UV, so they do not have a molar extinction coefficient to allow an accurate calculation of their concentration. Concentrations of those peptides were estimated from Coomassie blue-stained SDS polyacrylamide gels, with T57–140 as the standard, since that peptide has a molar extinction coefficient that can be used to calculate its concentration.

His₁₀ tag cleavage from T domain peptides.

The C-terminal His tag was removed from T57–120 and T100–140 using restriction-grade thrombin (EMD Millipore). Digestion of the peptides was done for 5 h at room temperature, using 2 units of thrombin/mg of protein. The reactions were stopped by the addition of phenylmethylsulfonyl fluoride to 1 mM, and polyacrylamide gel electrophoresis confirmed that the digestion was complete.

Synthetic T100–120.

The peptide H-EALRHNASRTPSATELAHANN-OH (T100–120) was purchased from AnaSpec Peptide (Fremont, CA). For in vivo experiments, the peptide was dissolved in water and then brought to a concentration of 2 mg · ml⁻¹ in 25 mM sodium borate-50 mM NaCl (pH 9.0), so that the buffer composition was the same as that for the other T domain peptides used in those experiments.

Colicin spot tests, killing assays in liquid culture, and protection assays with T domain.

Sensitivity to colicins or colicin domains was measured by spotting 10 μl of 10-fold serial dilutions of 1 mg · ml⁻¹ protein solutions on lawns of indicator strains spread in melted soft agar on petri dishes. The inverse of the concentration of the highest dilution at which any inhibition of growth of the lawn is seen is defined the endpoint killing activity.

To measure killing and protection by T domain in liquid culture, colicin-sensitive E. coli K361 (18) cells were grown in tryptone broth to an optical density at 660 nm (OD₆₆₀) of 0.4 (∼1.5 × 10⁸ to 2 × 10⁸ ml⁻¹). One hundred microliters of cells was added to 100 μl of broth, broth plus buffer, or broth plus T domain. The mixture was incubated with shaking at 37°C for 5 min, and then wild-type colicin E1 or E1ΔR was added, and incubation was continued with shaking at 37°C for 20 min (The amount of colicin to use for these experiments was first determined by a titration experiment and was chosen to result in killing of at least 99% of the indicator cells, in the absence of added protective peptides. For E1ΔR, that was ∼2 × 10⁻⁷ M, and for wild-type colicin E1, the concentration was ∼4 × 10⁻¹⁰ M). At the end of the killing period, a sample of each reaction mixture was immediately diluted 100-fold into sterile 6 mM CaCl₂-410 mM NaCl to stop cell growth. Subsequent aliquots of appropriate serial dilutions were spread on tryptone agar plates and incubated overnight at 37°C, and the surviving colonies were counted. The percentage of surviving colonies was calculated from the control culture to which no colicin or T domain was added. Data from separate protection experiments were evaluated for the significance of protection using Prism.