Nature Structural Biology9, 476 - 484 (2002)
Published online: 20 May 2002; | doi:10.1038/nsb797
The cytotoxic domain of colicin E9 is a channel-forming endonuclease
Khédidja Mosbahi1, 2, Christelle Lemaître1, 2, 3, Anthony H. Keeble1, Hamid Mobasheri1, 4, Bertrand Morel5, Richard James6, Geoffrey R. Moore5, Edward J.A. Lea1
& Colin Kleanthous1
1 School of Biological Sciences, University of East Anglia, Norwich, NR4 7TJ, UK.
2 These authors contributed equally to this work.
3 Present address: Laboratoire de Spectrométrie de Masse Bioorganique, Université Louis Pasteur, UMR/ULP CNRS 7509, ECPM 25 rue Becquerel, F-67087 Cedex 2, France.
4 Present address: Laboratory of Membrane Biophysics, Institute of Biochemistry and Biophysics, University of Tehran, PO Box 13145-1384, IR Iran.
5 School of Chemical Sciences, University of East Anglia, Norwich, NR4 7TJ, UK.
6 Division of Microbiology and Infectious Diseases, University Hospital, Queen's Medical Centre, University of Nottingham, Nottingham NG7 2UH, UK.
Bacterial toxins commonly translocate cytotoxic enzymes into cells using channel-forming subunits or domains as conduits. Here we demonstrate that the small cytotoxic endonuclease domain from the bacterial toxin colicin E9 (E9 DNase) shows nonvoltage-gated, channel-forming activity in planar lipid bilayers that is linked to toxin translocation into cells. A disulfide bond engineered into the DNase abolished channel activity and colicin toxicity but left endonuclease activity unaffected; NMR experiments suggest decreased conformational flexibility as the likely reason for these alterations. Concomitant with the reduction of the disulfide bond is the restoration of conformational flexibility, DNase channel activity and colicin toxicity. Our data suggest that endonuclease domains of colicins may mediate their own translocation across the bacterial inner membrane through an intrinsic channel activity that is dependent on structural plasticity in the protein.
Biological membranes present a formidable barrier to the translocation of proteins, a process that often requires large, membrane-bound protein assemblies. This is true of protein translocation whether the protein is in the unfolded state, as in sec-dependent protein translocation, or in the folded state, as occurs in the export of metalloproteins through the Tat pathway of bacteria1,
2,
3. Bacterial toxins such as anthrax are also examples of proteins that translocate across membranes but use a much scaled-down apparatus involving pore-forming subunits4. Here, we describe the first example of an endonuclease with widespread structural identity to enzymes in prokaryotes and eukaryotes that forms ion-conducting channels in membranes, which is linked to its ability to translocate into bacteria.
Colicins are protein antibiotics released by Escherichia coli to kill closely related strains during times of stress. The importance of colicins in bacterial competition and colonization is emphasized by the observation that a significant proportion of enterobacterial species are colicinogenic5. Colicins are active at nanomolar concentrations and generally induced through the SOS pathway; however, the mechanism by which they translocate into bacterial cells remains poorly understood. Their cytotoxic activities and cellular sites of action vary from those that are cytotoxic in the periplasm, including ionophores that depolarize the cytoplasmic membrane and inhibitors of peptidoglycan synthesis, to those that cross the cytoplasmic membrane to digest DNA or RNA. Although colicins are disparate in their mode of action, they nonetheless seem to use a common mechanism to traverse the outer membrane of Gram-negative bacteria. This mechanism involves a centrally located receptor recognition domain that binds to an outer membrane protein, which is normally involved in the uptake of nutrients, and an N-terminal translocation domain that contacts proteins in the periplasm, which include the Tol proteins for group A colicins or the TonB/ExbB/ExbD proteins for group B colicins5,
6. These multipartite interactions, thought to occur simultaneously, are geared toward bringing the C-terminal cytotoxic domain across the outer membrane.
Colicin E9 forms channels in membranes Pore-forming colicins, such as K, Ia, E1, N and A, have been the focus of more than two decades of research. Because of this, the mechanism by which their cytotoxic domains form voltage-gated channels in the inner membrane is reasonably well understood7,
8,
9. In contrast, there have been few studies aimed at understanding how the cytotoxic domains of enzymatic colicins cross the cytoplasmic membrane of bacteria. For example, it is not clear whether accessory proteins, such as the Tol proteins10, are involved directly in this process or whether the toxin can translocate across the cytoplasmic membrane unaided. To begin addressing these questions, we investigated the interactions of the endonuclease toxin colicin E9 with membranes in planar lipid bilayer experiments. E9, as well as its homologs E2, E7 and E8, is a group A colicin that kills bacteria through nonspecific degradation of chromosomal DNA11,
12. We tested whether DNase colicins have channel-forming domains that might be required for translocation of the DNase into the cytoplasm, by analogy with diphtheria toxin, which has a channel-forming domain responsible for translocating its ADP ribosyl transferase into the cytosol of eukaryotic cells13.
Colicin E9 is a 60-kDa toxin that is normally released from colicinogenic bacteria in the form of a heterodimeric complex with the 9.5 kDa immunity protein Im9 (ref. 14). The immunity protein protects the colicin-producing bacterium from the activity of its own toxin but is jettisoned on entry of the colicin into a susceptible cell15. Hence, the form of the toxin tested in the bilayer experiments had the immunity protein removed (see Methods). Previous work from our laboratory has shown that this form of the toxin retains complete biological activity14. Immunity-free colicin E9 (2 nM) was added to the cis chamber of a bilayer apparatus in 10 mM Tris-HCl buffer at pH 7.5, containing 0.1 M NaCl and 10 mM CaCl2, with a potential difference (p.d.) applied across the membrane. Random, fluctuating current was observed that shows evidence of opening and closing events with conductance of the order of 100 pS, although larger conductance states are also seen (Fig. 1a). To identify the region(s) of the protein responsible for this activity, we analyzed a truncated colicin in which the E9 DNase domain had been deleted16 but which retained the domains responsible for receptor binding and outer membrane translocation. Surprisingly, this construct does not produce channel activity (Fig. 1b) even at micromolar protein concentrations (data not shown). Finally, we analyzed the E9 DNase domain17 and found that the channel activity, whose characteristics are similar to those of the full-length colicin, is associated with this domain (Fig. 1c).
Figure 1. Planar lipid bilayer experiments with colicin E9.
The current traces are of 2 nM intact colicin E9 and its domains incorporated into soybean lecithin bilayers. The schematic above each set of traces indicates the protein construct used. Intact colicin E9 has three domains, the R-domain is the central BtuB-binding domain found in all E group colicins, the T-domain is the translocation domain involved in binding Tol proteins in the periplasm and the C-terminal DNase domain carries the cytotoxic activity. The dashed and dotted lines in all traces identify zero (also marked by a closed triangle) and 100 pS conductance, respectively. Open triangles identify additional conductance events, although not all are marked. a, Current traces for intact colicin. The top traces (potential difference (p.d.), +90 mV) show the absence of channel events early in the record, followed after 3 s by random opening and closing of a 100 pS channel, with larger conductances also apparent. After 1 s further recording and toward the end of these traces, only the 100 pS conductance is seen. Examples of channel opening and closing are indicated. The bottom traces (p.d., +80 mV) are from a separate experiment and highlight how conductance at 100 pS tends to dominate the traces with additional channel events superimposed on this. b, Continuous current traces (p.d., +80 mV) of a truncated form of colicin E9 in which the DNase has been deleted showing the absence of channel activity. No channel activity was evident throughout the entire record of 100 s. c, Continuous current traces (p.d., -80 mV) for the 15 kDa E9 DNase domain. Channels at 100 pS were clearly evident for the E9 DNase domain (not shown), the traces in the figure highlighting the additional conducting states that are also observed.
Characterization of colicin DNase channels We have investigated the channel activity of the colicin E9 DNase domain from seven different protein preparations involving >80 separate membrane experiments, with channels evident in every case. The channel data (Fig. 1) demonstrate that the enzyme shows discrete 'open' and 'closed' states, as well as several different conducting states with lifetimes that vary over the tens-of-milliseconds time range. To evaluate the number, size and frequency of the different conducting states displayed by the E9 DNase domain, we analyzed data over a total record time of 7,000 s from eight independent bilayer experiments (Fig. 2a). The resulting histogram of channel frequency versus channel size reveals that the most frequently observed channels are 100 pS, although a range of higher conductance states are also observed with a periodicity approximately equivalent to this unit size. This was confirmed when the current distribution for E9 DNase 'open' and 'closed' conductance states was analyzed for a single experiment using an all-points amplitude histogram in which the 100 pS channel is clearly apparent (Fig. 2b). The recordings for the E9 DNase have some similarity to those of the pore-forming colicins in that they display a range of conductance states8. However, unlike pore-forming colicins, the conductance states of the E9 DNase have much shorter lifetimes (milliseconds compared to seconds) and display larger conductance states at the single-channel level at relatively low salt concentrations. For example, colicin E1 shows a single conductance state of 10 pS at pH 6 and 1 M KCl8.
Data were collected and analyzed as described (see Methods). a, Histogram showing the range of different conductance states (in 20 pS bins) showed by the E9 DNase (x-axis) and their frequency (y-axis). All-points amplitude histograms for b, the E9 DNase and c, the E3 rRNase domains. The data were collected over 125 s at an applied p.d. of 80 mV. The data for the E3 rRNase show the absence of any channel activity (zero current) during the entire record, whereas that for the E9 DNase shows two peaks (indicated by arrows), each corresponding to a distinct current level in the recording. The area under each peak represents the proportion of the record spent at that current level; the mean single channel currents obtained from the Gaussian fits to these peaks were 8.5 pA and -11.4 pA, corresponding to conductance states of 106 pS and -142 pS, respectively. d, Current/voltage relationship for single E9 DNase channels showing that they are independent of the applied p.d. The gradient of the regression line fit to this I/V data is 105 pS.
We analyzed the DNase domains of the other E group colicins, E2, E7 and E8, under equivalent conditions. All showed channel activity, although the channels varied in size and gating behavior in each case (data not shown). In a recent paper on the channel-forming toxin colicin A, the DNase domain of colicin E2 was determined to not induce channels in bilayers18. However, the experimental conditions used in these experiments (in which channel events of the order of seconds were recorded) preclude their observation.
Because colicin endonucleases are basic proteins (pI 9.5), it could be argued that the effects we observe are due to nonspecific adsorption of these positively charged proteins to the bilayer, resulting in its disruption or destabilization. This can be discounted on the basis of three observations. First, the discrete nature of the observed gating events (Fig. 1) implies a specific protein−membrane interaction. Second, we investigated the ability of unrelated proteins, including some of similarly high pI (cytochrome c, pI 9.5 and lysozyme, pI 9.3) and bovine serum albumin (pI 5.8), in separate bilayer experiments (n = 4) under the same conditions (Fig. 1) and over a range of membrane potentials ( 50−100 mV); however, we could not detect any channel behavior (data not shown). Finally, we investigated whether an enzymatic toxin from another colicin family could induce channel activity. Colicin E3 is a ribonuclease specific for 16S ribosomal RNA. We added the 11 kDa rRNase domain of colicin E3 (pI 9.7) to bilayers but again failed to detect any channel activity (Fig. 2c). We conclude that the channel activity of the DNase domain of colicin E9 is specific to colicin endonucleases and not due to nonspecific effects of the interaction of a basic protein and the membrane bilayer.
We determined the voltage-dependence of the E9 DNase channels by collecting single-channel data for each applied voltage over a period of 125 s and analyzing the data in the form of all-points amplitude histograms. From the peaks of Gaussian curves fit to these data (Fig. 2b), mean single-channel currents were obtained and plotted against the applied membrane potential (Fig. 2d). The resulting I/V curve shows that the level of current increased linearly with the imposed voltage across positive and negative polarities, indicating that E9 DNase channels, unlike those of pore-forming colicins, are not voltage dependent. The single channel conductance derived from the regression line fit to the data is 105 pS.
To estimate the relative selectivity of DNase channels for anions and cations, measurements of the reversal potential (Erev, the potential difference of the cis relative to the trans chamber) were made with 0.1 M salt solution on the cis side and 1 M on the trans side of the membrane. Erev for three salts, sodium chloride, sodium gluconate and glucosamine chloride, at pH 7.5 were found to be -20 mV, +33 mV and -35 mV, respectively (data not shown). Thus, the order of channel permeability is Cl- > Na+, Na+ > gluconate- and Cl- > glucosamine+. Because there is significant deviation from the equilibrium potentials of the small ions used (Erev = +58 mV for Na+ and -58 mV for Cl-), these data show that the DNase channels are permeable to large ions to a significant degree, similar to findings reported for pore-forming colicins19,
20,
21.
The molecularity of the different E9 DNase conductance states is not known at the present time, although the dominance of the 100 pS channel at 2 nM of protein suggests that this is the unit channel conductance for a single DNase molecule. To address if this is the case, we further analyzed the channel activity at a 10-fold higher protein concentration (Fig. 3). The 100 pS channels are even more prevalent under these concentrations, with fewer periods where the channels remained closed, which is characteristic of traces at 2 nM. There did not seem to be any increase in the frequency of larger channel events, which might be expected if the channels were oligomeric. However, there were instances of several insertions into the membrane, as indicated by 'ladder-effects' in bilayer traces (Fig. 3a), with steps equivalent to 100 pS. We conclude from these data and from the experiments at 2 nM, that the E9 DNase channels are probably monomeric, with a unit conductance of 100 pS and that the rarer, larger conductances we observed reflect several insertions into the bilayer rather than oligomeric assemblies of protein. A unimolecular explanation for this activity would also be consistent with the antibacterial single-hit kinetics displayed by colicins. Interestingly, the concentration of one molecule per bacterial cell is 1−2 nM, which is the concentration at which we readily detect E9 DNase channel activity in planar lipid bilayers (Figs 1, 2).
Figure 3. E9 DNase channels at higher protein concentration.
To examine the molecularity of the E9 DNase channels, single-channel measurements were made as described in Fig. 1 but at 10 higher protein concentration (20 nM). The applied p.d. was +150 mV. Because the single channel conductance of the E9 DNase is independent of voltage (Fig. 2d), the data collected at this p.d. are directly comparable with those in Fig. 1. a, From a state of zero conductance (closed state) a 'ladder effect' is observed as several gating events occur to generate a series of 'open' states of differing conductance (the dotted lines correspond to intervals of 100 pS). b, Continuous channel traces later in the same experiment where a 100 pS channel (corresponding to the dotted line) is present almost throughout the recording. Superimposed on this are other channel events, some of which are identified by an open triangle, that seem largely to be multiples of 100 pS. Note that the timescale in this figure is more compressed than that in Fig. 1.
Using a bilayer composed of 70% phosphatidyl ethanolamine, 20% phosphatidyl glycerol and 10% cardiolipin, a mixture that more closely mirrors that of the inner membrane of E. coli22, we found that the E9 DNase showed channel activity essentially indistinguishable from that seen using soybean lecithin (data not shown), indicating that membrane composition is not an important factor. We also investigated the effect of transition metal ions, DNA and Im9, all of which bind to the E9 DNase23,
24,
25,
26, on channel activity. Neither metal nor DNA binding affected the ability of the DNase to induce channels. In contrast, Im9, which binds to the E9 DNase with an equilibrium dissociation constant of 10-14 M (ref. 25), completely abolished channel activity (data not shown). Hence, Im9 inhibits both the enzymatic activity of the colicin and its ability to form channels in bilayers.
Disulfide bond inhibition of DNase channels Engineered disulfide bonds are a useful tool to probe the structural reorganization/unfolding events that accompany the translocation of toxins across membranes27,
28. A single disulfide bond was introduced into the 134-amino acid E9 DNase domain between residues 20 (D20C) and 66 (E66C) to assess its influence on channel activity (see Methods; Fig. 4a). Although residues 20 and 66 are not optimally positioned for disulfide bond formation (the side chains are separated by >3.5 Å)29, each is part of a highly mobile region of the DNase30 (see below). The purified double mutant, E9 DNase D20C/E66C (E9 DNaseSH2), formed a disulfide bond (E9 DNaseS−S) when oxidized with diamide that, consistent with its location, did not affect either immunity binding or DNase activity of the domain (Fig. 4b,c). The E9 DNaseS−S failed to produce channels in bilayer experiments. However, channel activity was restored when the reducing agent dithiothreitol (DTT) was added to the cis chamber, although the resulting channel activity was noisier than that of the wild type protein (Fig. 5a,b). Prior alkylation by iodoacetamide of the reduced thiols of E9 DNaseSH2 resulted in channel data similar to that of the wild type protein but also with the formation of smaller channels of 50 pS (Fig. 5c). These data demonstrate that the channel-forming activity of the E9 DNase is an intrinsic property of the enzyme and not due to a contaminant associated with its purification, and that a single intramolecular crosslink completely abolishes this activity.
Figure 4. Engineering a disulfide bond into the E9 DNase.
a, Crystal structure of the E9 DNase29 showing the location of enzyme active site amino acids (cyan), the Im9 binding site (red) and regions of the enzyme the backbone atoms of which have been shown by NMR30 to be conformationally mobile (yellow). Two residues in these mobile regions of the DNase (Asp 20 and Glu 66) were chosen for mutation to Cys for the generation of the disulfide form of the enzyme, E9 DNaseS−S and ColE9S−S (the disulfide bond is shown in green). b, Comparing Im9 binding data of wild type (open circle), reduced (closed triangle) and oxidized (open triangle) E9 DNase D20C/E66C mutant using Trp emission fluorescence spectroscopy, as described by Wallis et al14. c, Comparing endonucleolytic digestion of 12mer dsDNA by stopped-flow absorbance at 260 nm in 50 mM triethanolamine, pH 7.5, buffer containing 1 mM MgCl2 at 25 °C, following the change in hyperchromicity. Symbols as for (b).
Figure 5. Effect of disulfide bond on E9 DNase channel activity.
a, E9 DNaseS−S was added to the cis chamber of the bilayer apparatus (p.d. 100 mV). The total record of an experiment is shown (82 s), indicating the absence of channel activity for the oxidized protein. This was confirmed for two protein preparations using several different membranes. b, The result of adding 5 mM DTT to the cis chamber containing E9 DNaseS−S. Recording ceased during the addition of DTT, and 10 s elapsed before it was resumed. Channel activity began to appear midway through the record, although it was more noisy and less well defined compared to the wild type enzyme. c, Total channel record for E9 DNaseSH2 alkylated with iodoacetamide (see Methods). Sections of the trace are shown in (i) and (ii). Clearer channel events are seen compared to the reduced protein, although these differ from those of the wild type DNase (see text).
Previous work from our laboratory has shown that the E9 DNase displays extensive conformational dynamics in solution, with this flexibility centered on the 20s and 60s regions, where the disulfide bond was engineered, and readily manifested as doubled crosspeaks in 1H-15N HSQC-NMR spectra30. E9 DNaseS−S shows a simplified HSQC spectrum, with no evidence of chemical exchange crosspeaks (Fig. 6). Nevertheless, strong similarities remained between the chemical shifts of the wild type enzyme and E9 DNaseS−S (>80% of the amide resonances had very similar chemical shifts), showing that they probably have similar structures. Doubled amide resonances equivalent to those of the wild type protein reappeared on reduction of the disulfide bond (Fig. 6). In addition, temperature denaturation experiments indicated that the stability of the E9 DNaseS−S increases by 11 °C as a result of the covalent crosslink (data not shown). Hence, it is reasonable to assume that the loss of channel activity in E9 DNaseS−S is due to its increased stability and reduced flexibility relative to the wild type DNase.
Figure 6.1H-15N HSQC spectra of E9 DNaseS−S and E9 DNaseSH2.
1H-15N HSQC spectra at 600 MHz of the oxidized (blue) and reduced (red) E9 DNase D20C/E66C double mutant. The data indicate that the oxidized protein has a single form in solution, whereas the reduced protein, like the wild type DNase, exists in two forms in an 60:40 molar ratio30. The conformational heterogeneity of the wild type protein has been shown to affect the chemical shifts of Gly 15, Asp 36, Lys 63, Lys 69 and Val 121. All these peaks are doubled in the spectrum of E9 DNaseSH2 (side panels), whereas none are doubled in the spectrum of the E9 DNaseS−S. The large chemical shift differences between the Lys 69 NH resonances of oxidized and reduced protein probably reflect the local conformational effects of Cys 66.
Colicin E9 cytotoxicity studies We tested whether the equivalent disulfide bond in colicin E9 (D468C/E514C to generate ColE9S−S) had antibacterial activity. On solid media, wild type toxin shows activity at nanomolar protein concentrations, similar to that reported31, whereas ColE9S−S is inactive even at micromolar concentrations (Fig. 7a). Indeed, the disulfide-bonded toxin behaves as a previously identified active site mutant, ColE9 H575A, which is completely devoid of cytotoxic activity32 (Fig. 7a). Unlike conventional active site mutants, however, toxin activity for ColE9S−S can be recovered on reduction of the disulfide bond with DTT, although the reduced protein (ColE9SH2) shows only wild type colicin activity when first alkylated with iodoacetamide, suggesting that it may be re-oxidizing during import into bacteria. The cytotoxic activity of the thiol-alkylated toxin was confirmed in liquid culture, where it shows wild type activity, whereas the oxidized protein is completely inactive (Fig. 7b). Our data demonstrate that a disulfide bond in the E9 DNase domain inhibits both channel activity in lipid bilayers and cytotoxic activity, with both activities restored on reduction of the disulfide bond.
Figure 7. Antibacterial activity of oxidized and reduced colicin E9.
a, Agar plate assay comparing the biological toxicities of wild type colicin E9 with oxidized, reduced and alkylated colicin E9 D468C/E514C, as well as a previously identified DNase active site mutant (H575A) (denoted by the asterisk). In each lane of the plate, a lawn of E.coli JM83 cells has grown onto which was spotted a serial dilution of each of the protein constructs indicated. Zones of clearing indicate cell death. b, Comparison of the cytotoxic activities of oxidized and alkylated colicin E9 D468C/E514C against bacterial cells grown in liquid media. The figure shows growth curves of E. coli JM83 in the absence of colicin E9 (closed circle) or with the addition of wild type colicin E9 (closed square), ColE9S−S (open triangle) or reduced and alkylated colicin E9 D468C/E514C (open square).
Discussion Enzymatic A-B toxins, such as diphtheria, cholera and pertussis, kill mammalian cells by translocating their cytotoxic enzymes (the A domain) into the cytosol with the aid of the receptor-binding B domain33,
34. In the case of diphtheria, its B domain also forms conducting channels (in planar lipid bilayers) that form a conduit for the ADP ribosyl transferase to enter the cytosol13. In contrast to this mode of action, our data suggest that the endonuclease domain of colicin E9 may act as its own membrane translocator. As well as having implications for the mechanism by which enzymatic colicins translocate into bacteria, the present study highlights the potential for similar activity in proteins homologous to E9, such as intron- and intein-encoded homing endonucleases that initiate recombination events in prokaryotes and eukaryotes35.
Membrane translocation by DNase colicins Because all enzymatic E group colicins require the same import apparatus in E. coli cells (BtuB/OmpF and the Tol proteins), the mechanism by which the cytoplasmic membrane is breached has generally been assumed to also be the same. However, only DNase (and not rRNase) toxin domains show channel activity in bilayers, raising the possibility that their passage across the inner membrane may occur by different mechanisms.
The effects of the intramolecular disulfide bond serves to link colicin E9 cytotoxicity and E9 DNase channel activity. Although this does not prove that the enzyme acts as its own translocator, this link is certainly compelling. Arriving at more direct evidence may prove difficult because unlike diphtheria toxin, where the translocating and cytotoxic activities are carried on distinct domains, the same 15 kDa domain of colicin E9 contains its channel-forming and cytotoxic DNase activities, functionalities that may not be easily separable.
Channel formation by the E9 DNase almost certainly involves substantial structural changes, which may lie at the heart of the unusual conformational dynamics shown by the enzyme in solution30. Structural rearrangements are also a prerequisite for channel-formation by the 10-helical bundle, pore-forming colicins, where a molten globule has been inferred to be the active species36,
37. In the case of colicin Ia, the pore-forming domain has been shown to translocate large segments of polypeptide, and even proteins, across the bilayer during channel formation18,
19,
38. Although the channel activity of the E9 DNase is associated with the ability of colicin E9 to kill bacterial cells, it cannot be the causative agent of cell death. This is inferred from the colicin E9 active site mutant H575A, which lacks enzymatic activity and is not cytotoxic32 (Fig. 7a) but is still able to form conducting channels in bilayers (data not shown). A key question is why the DNase channels themselves do not kill the cell during translocation. One way in which cell death could be avoided during channel-mediated translocation of the DNase across the inner membrane is the proteolytic excision of the endonuclease upon reaching the cytoplasm, thereby allowing the membrane to re-seal. Indeed, proteolytic cleavage of both rRNase and DNase colicins has been reported in which the processing site seems to be in a helical region linking the enzyme to the receptor-binding domain of the toxin39,
40,
41.
Conclusions Bacterial toxins, by their nature, are compact killing machines that engage in a variety of macromolecular associations with target cells into which they transfer folded protein domains or subunits. In general, this is accomplished in two stages: receptor-mediated translocation across one or more membranes followed by the expression of a cytotoxic activity in the desired cellular location, with each function associated with a specific domain or subunit. The present work has identified a channel-forming activity intrinsic to the cytotoxic endonuclease domain of the enzymatic colicin E9 that is indispensable for cytotoxicity. Endonuclease domains of colicins are, therefore, remarkable multifunctional proteins capable of binding metal ions, immunity proteins, DNA and lipids. This diversity of macromolecular associations likely underpins the ability of the cytotoxic endonuclease to translocate from the extracellular environment to the cytoplasm of an E. coli cell.
Methods Proteins. Colicin E9 (and the double mutant ColE9 D468C/E514C)14, E9 DNase (and the double mutant E9 DNase D20C/E66C)32 and the colicin T-R deletion construct16 were purified as described. Proteins were routinely lyophilized either from water (DNases) or 50 mM potassium phosphate buffer, pH 7.5 (colicins). Protein concentrations were determined from absorbance at 280 nm25,
31.
Planar lipid bilayers and single-channel recordings. Bilayers were formed principally from soybean lecithin (Type II-S, Sigma) using the technique of Montal-Muller42. The electrolyte used in the bilayer experiments was 10 mM Tris-HCl, pH 7.5, 0.1 M NaCl and 10 mM CaCl2. Colicin E9, its derivatives and the E9 DNase were added to the cis compartment to a final concentration of 2 nM or 20 nM. Single-channel measurements were made at room temperature as described43,
44. Briefly, using reversible Ag/AgCl electrodes and an applied p.d., the membrane current was measured using a bilayer amplifier (HAMK2TC, R.A.P. Montgomery) filtered at 1,000 Hz with a low pass filter (VBF/3, Kemo Ltd). All quoted p.d.s are referenced to the cis compartment. Recordings were made using a PC with a CED 1401 interface (Cambridge Electronic Design), and data were analyzed using patch-clamp software (PAT v7.0)45. Records typically lasted 100 s, with 250 ms sections of these records shown in the figures.
To determine the range of channel activities shown by the E9 DNase, channel data from eight separate bilayer experiments were analyzed covering 7,000 s of records. The range of current over which the signal extended was divided into a series of contiguous, equal sized 'bins'. The collective data was analyzed as the frequency with which each conductance state was observed against the range of different states detected.
All-points amplitude histograms from single channel records were obtained using the patch-clamp software and the resulting histograms fitted to Gaussian curves to derive mean single channel conductances. An I/V curve was subsequently generated by plotting the mean single-channel currents for the E9 DNase at 2 nM as a function of the applied p.d.
Cation/anion selectivity estimates of E9 DNase channels in bilayers were determined in 10 mM Tris-HCl, pH 7.5, 10 mM CaCl2 with 1 M salt solution on the trans side of the membrane and 0.1 M salt in the cis compartment. Agar bridges were used in conjunction with Ag/AgCl electrodes to avoid electrode potential problems arising from the different solutions used. The reversal potential (Erev) was estimated for sodium chloride, sodium gluconate and glucosamine chloride as the p.d. applied to the cis side of the membrane for which the DNase channel openings changed polarity.
Disulfide bond engineering. Plasmid PCR mutagenesis using Pfu-Turbo (Stratagene) was used to engineer the E9 DNase D20C/E66C double mutant in two rounds of mutagenesis. The presence of the mutations was confirmed by DNA sequencing and electrospray ionization mass spectrometry (ESI-MS) of the purified protein (theoretical Mw = 15,048 Da and experimental Mw = 15,048.58 1.70 Da). The E9 DNase D20C/E66C domain, along with the downstream His-tagged immunity gene, was also subcloned into the complementary NcoI-XhoI sites of the pCS4 plasmid32, allowing for purification of intact colicin E9 containing the double mutant in the DNase domain (ColE9 D468C/E514C).
For both E9 DNase D20C/E66C and ColE9 D468C/E514C mutants, proteins were purified in 50 mM Tris, pH 7.5, in the presence of 10 mM DTT until disulfide bond formation was induced by oxidation. The E9 DNase D20C/E66C required a 20 min incubation with 1 mM diamide at room temperature, whereas the disulfide bond in the intact colicin formed readily on removal of reductant by dialysis (the reason for this difference in disulfide stability is unclear). The presence of the disulfide bond was confirmed in both proteins by the differential alkylation of the thiols by iodoacetamide (50 mM for 30 min in the dark at room temperature) with or without previous reduction with DTT, followed by ESI-MS where only monomeric species were detected. In each case, the two thiols could be alkylated only following previous reduction with DTT and all masses were within 1−2 Da of the expected mass. In the case of E9 DNaseS−S, the presence of the disulfide bond was also confirmed by the loss of reactivity with the thiol reagent DTNB and by its faster migration on SDS gels when compared with the reduced protein. Gel filtration chromatography under native conditions, as described by Wallis et al.14,
17, confirmed the monomeric nature of the oxidized forms of E9 DNaseS−S and ColE9S−S.
Biophysical experiments. Endonuclease activity was assayed in 50 mM triethanolamine buffer, pH 7.5, and 1 mM Mg2+ at 25 °C with 5 M protein and 2 M DNA by following the hyperchromic shift resulting from the cleavage of a double-stranded oligonucleotide (after Baldwin et al.46). A 12mer dsDNA palindromic sequence 5'-GACGATATCGTC-3' (re-annealed following prior heating to 80 °C) was used as substrate and its hydrolysis monitored by a time-dependent increase of A260 after mixing the protein with DNA using a * stopped-flow CD spectrophotometer in the absorbance mode (Applied Photophysics). Im9 binding studies of the reduced/oxidized E9 DNase D20C/E66C protein were carried out on a Shimadzu RF5000 or Spex Fluoromax essentially as described by Wallis et al.14 NMR spectra were recorded on a Varian Unity Inova 600 MHz spectrometer with pulse sequences implemented in the Varian 'Protein-pack' suite of experiments using a 2 mM E9 DNase D20C/E66C sample that had been labeled with both 13C and 15N, as described by Whittaker et al.30 HSQC spectra were first recorded for the oxidized protein and then again following reduction of the sample with dithiothreitol (10 mM). Assignments were obtained from CBCA(CO)NH and HNCACB spectra as described by Boetzel et al.47 and references cited therein.
Biological activities of colicins. Determination of colicin activity against E. coli JM83 using an agar plate assay was as described by Wallis et al.31 When alkylated colicin was used, the reduced protein was first treated with iodoacetamide as described above. For liquid culture experiments, 3 g ml-1 (final concentration) colicin was added to 20 ml LB medium inoculated with an overnight culture of JM83 and growth monitored as change in optical density (OD600) at 30 min intervals.
Received 18 December 2001; Accepted 18 March 2002; Published online: 20 May 2002.
REFERENCES
Johnson, A.E. & van Waes, M.A. The translocon: a dynamic gateway at the ER membrane. Annu. Rev. Cell Dev. Biol.15, 799−842 (1999). | Article | PubMed | ISI | ChemPort |
Manting, E.H. & Driessen, A.J.M. Escherichia coli translocase: the unravelling of a molecular machine. Mol. Microbiol.37, 226−238 (2000). | Article | PubMed | ISI | ChemPort |
Berks, B.C., Sargent, F. & Palmer, T. The Tat protein export pathway. Mol. Microbiol.35, 260−274 (2000). | Article | PubMed | ISI | ChemPort |
James, R., Lazdunski, C. & Pattus, F. Bacteriocins, Microcins and Lantibiotics (NATO ASI Series H, Springer, Heidelberg; 1992)
Lazdunski, C.J. et al. Colicin import into Escherichia coli cells. J. Bacteriol.180, 4993−5002 (1998). | PubMed | ISI | ChemPort |
Schein, S.J., Kagan, B.L. & Finkelstein, A. Colicin K acts by forming voltage-dependent channels in phopholipid bilayer membranes. Nature276, 159−163 (1978). | PubMed | ISI | ChemPort |
Cramer, W.A. et al. Structure-function of the channel-forming colicins. Annu. Rev. Biophys. Biomol. Struct.24, 611−641 (1995). | Article | PubMed | ISI | ChemPort |
Gouaux, E. The long and short of colicin action: the molecular basis for the biological activity of channel-forming colicins. Structure5, 313−317 (1997). | Article | PubMed | ISI | ChemPort |
Webster, R.E. The tol gene products and the import of macromolecules into Escherichia coli cells. Mol. Microbiol.5, 1005−1001 (1991). | PubMed | ISI | ChemPort |
James, R., Kleanthous, C. & Moore, J.R. The biology of E colicins: paradigms and paradoxes. Microbiology142, 1569−1580 (1996). | PubMed | ISI | ChemPort |
Pommer, A.J., Wallis, R., Moore, R., James, R. & Kleanthous, C. Enzymological characterization of the nuclease domain from the bacterial toxin colicin E9. Biochem. J.334, 387−392 (1998). | PubMed | ISI | ChemPort |
Oh, J.K., Senzel, L., Collier, R.J. & Finkelstein, A. Translocation of the catalytic domain of diphtheria toxin across planar phospholipid bilayers by its own T domain. Proc. Natl. Acad. Sci. USA96, 8467−8470 (1999). | Article | PubMed |
Wallis, R. et al. In vivo and in vitro characterization of overproduced colicin E9 immunity protein. Eur. J. Biochem.207, 687−695 (1992). | Article | PubMed | ISI | ChemPort |
Kleanthous, C. & Walker, D. Immunity proteins: enzyme inhibitors that avoid the active site. Trends Biochem. Sci.26, 624−631 (2001). | Article | PubMed | ISI | ChemPort |
Penfold, C.N. et al. A 76 residue polypeptide of colicin E9 confers receptor specificity and inhibits the growth of vitamin B12-dependent E. coli 113/3 cells. Mol. Microbiol.38, 639−649 (2000). | Article | PubMed | ISI | ChemPort |
Wallis, R. et al. Tandem overexpression and characterization of the nuclease domain of colicin E9 and its cognate inhibitor protein Im9. Eur. J. Biochem.220, 447−454 (1994). | Article | PubMed | ISI | ChemPort |
Slatin, S.L., Nardi, A., Jakes, K., Baty, D. & Duché, D. Translocation of a functional protein by a voltage-dependent ion channel. Proc. Natl. Acad. Sci. USA99, 1286−1291 (2002). | Article | PubMed | ChemPort |
Jakes, K.S., Kienker, P.K. & Finkelstein, A. Channel-forming colicins: translocation (and other deviant behaviour) associated with colicin Ia channel gating. Q. Rev. Biophys.32, 189−205 (1999). | Article | PubMed | ISI | ChemPort |
Raymond, L., Slatin, S.L. & Finkelstein, A. Channels formed by colicin E1 in planar lipid bilayers are large and exhibit pH-dependent selectivity. J. Membr. Biol.84, 173−181 (1985). | PubMed | ISI | ChemPort |
Bullock, J.O. & Kolen, E.R. Ion selectivity of colicin E1. J. Membr. Biol.144, 131−145 (1995). | PubMed | ISI | ChemPort |
Kadner, R.J. In Escherichia coli and Salmonella; Cellular and Molecular Biology 2nd edn Vol. 1 (ed. Neidhardt, F.C.) 58−87 (ASM Press, Washington; 1996).
Kleanthous, C. et al. Structural and mechanistic basis of immunity towards endonuclease colicins. Nature Struct. Biol.6, 243−252 (1999). | Article | PubMed | ISI | ChemPort |
Pommer, A.J. et al. Homing-in on the role of transition metals in the HNH motif of colicin endonucleases. J. Biol. Chem.274, 27153−27160 (1999). | Article | PubMed | ISI | ChemPort |
Wallis, R., Moore, G.R., James, R. & Kleanthous, C. Protein-protein interactions in colicin E9 DNase-immunity protein complexes. Diffusion-controlled association and femtomolar binding for the cognate complex. Biochemistry34, 13743−13750 (1995). | Article | PubMed | ISI | ChemPort |
Pommer, A.J. et al. Mechanism and cleavage specificity of the H-N-H endonuclease colicin E9. J. Mol. Biol.314, 735−749 (2001). | Article | PubMed | ISI | ChemPort |
Falnes, P.Ø., Choe, S., Madshus, I.H., Wilson, B.A. & Olnes, S. Inhibition of membrane translocation of diphtheria toxin A fragment by internal disulfide bridges. J. Biol. Chem.269, 8402−8407 (1994). | PubMed | ISI | ChemPort |
Duché, D., Parker, M.W., González-Mañas, J-M., Pattus, F. & Baty, D. Uncoupled steps of the colicin A pore formation demonstrated by disulfide bond engineering. J. Biol. Chem.269, 6332−6339 (1994). | PubMed | ISI |
Kühlmann, U.C., Pommer, A.J., Moore, G.R., James, R., & Kleanthous, C. Specificity in protein-protein interactions: The structural basis for dual recognition in colicin endonuclease-immunity protein complexes. J. Mol. Biol.301, 1163−1178 (2000). | Article | PubMed | ISI | ChemPort |
Whittaker, S.B.M. et al. NMR detection of slow conformational dynamics in an endonuclease toxin. J. Biomol. NMR12, 145−159 (1998). | Article | PubMed | ISI | ChemPort |
Wallis, R. et al. Protein-protein interactions in colicin E9 DNase-immunity protein complexes. Cognate and noncognate interactions that span the mM-fM affinity range. Biochemistry34, 13751−13759 (1995). | Article | PubMed | ISI | ChemPort |
Garinot-Schneider, C., Pommer, A.J., Moore, G.R., Kleanthous, C. & James, R. Identification of putative active-site residues in the DNase domain of colicin E9 by random mutagenesis. J. Mol. Biol.260, 731−742 (1996). | Article | PubMed | ChemPort |
Lacy, B.D. & Stevens, R.C. Unraveling the structures and modes of action of bacterial toxins. Curr. Opin. Struct. Biol.8, 778−784 (1998). | Article | PubMed | ISI |
Falnes, P.Ø. & Sandvig, K. Penetration of protein toxins into cells. Curr. Opin. Cell Biol.12, 407−413 (2000). | Article | PubMed | ISI | ChemPort |
Chevalier, B.S. & Stoddard, B.L. Homing endonucleases: structural and functional insight into the catalysts of intron/intein mobility. Nucleic Acids Res.29, 3757−3774 (2001). | Article | PubMed | ISI | ChemPort |
van der Goot, F.G, González-Mañas, J.M., Lakey, J.H. & Pattus, F. A 'molten-globule' membrane insertion intermediate of the pore-forming domain of colicin A. Nature354, 408−410 (1991). | Article | PubMed | ChemPort |
Zakharov, S.D. et al. Membrane-bound state of the colicin E1 channel domain as an extended two-dimensional helical array. Proc. Natl. Acad. Sci. USA95, 4282−4287 (1998). | Article | PubMed | ChemPort |
Slatin, S.L., Qiu, X-Q., Jakes, K.S., & Finkelstein, A. Identification of a translocated protein segment in a voltage-dependent channel. Nature371, 158−161 (1998). | Article |
de Zamaroczy, M., Mora, L., Lecuyer, A., Géli, V. & Buckingham, R.H. Cleavage of colicin D is necessary for cell killing and requires the inner membrane peptidase LepB. Mol. Cell8, 159−168 (2001). | Article | PubMed | ISI | ChemPort |
Liao, C.C., Hsiao, K.C., Liu, Y.W., Leng, P.H. & Chak, K-F. Processing of DNase domain during translocation of colicin E7 across the membrane of Escherichia coli. Biochem. Biophys. Res. Commun.284, 556−562 (2001). | Article | PubMed | ISI | ChemPort |
Soelaiman, S., Jakes, K., Wu, N., Li, W. & Shoham, M. Crystal structure of colicin E3: implications for cell entry and ribosome inactivation. Mol. Cell8, 1053−1062 (2001). | Article | PubMed | ISI | ChemPort |
Montal, M. & Muller, P. Formation of bimolecular membranes from monolayers and study of their properties. Proc. Natl. Acad. Sci. USA69, 3561−3566 (1972). | PubMed | ChemPort |
Bishop, N.D. & Lea, E.J.A. Characterization of the porin of Rhodobacter capsulatus 37b4 in planar lipid bilayers. FEBS Lett.349, 69−74 (1994). | Article | PubMed | ISI | ChemPort |
Bishop, N.D., Lea, E.J.A., Mobasheri, H. & Spiro, S. Altered voltage sensitivity of mutant OmpC porin channels. FEBS Lett379, 295−298 (1996). | Article | PubMed | ISI | ChemPort |
Dempster, J. Single-channel analysis program PAT V7.0 PhD. thesis, Univ. of Strathclyde, Glasgow (1993).
Baldwin, G.S., Vipond, I.B. & Halford, S.E. Rapid reaction analysis of the catalytic cycle of the EcoRV restriction endonuclease. Biochemistry34, 705−714 (1995). | Article | PubMed | ISI | ChemPort |
Boetzel, R. et al. Assignment of 1H, 13C and 15N signals of the inhibitor protein Im9 bound to the DNase domain of colicin E9. J. Biomol. NMR12, 567−568 (1998). | Article | PubMed | ISI | ChemPort |
Acknowledgments We thank A. Reilly, C. Moore and N. Cull for expert technical assistance and A. Leech for help with the acquisition of all mass spectrometry data. We also thank the referees of this paper for their helpful comments. This work was supported by The Biotechnology and Biological Sciences Research Council. A.H.K. was supported by a Wellcome Trust Prize Studentship.
Competing interests statement:
The authors declare that they have no competing financial interests.
100 pS, although larger conductance states are also seen (
80 mV. The data for the E3 rRNase show the absence of any channel activity (zero current) during the entire record, whereas that for the E9 DNase shows two peaks (indicated by arrows), each corresponding to a distinct current level in the recording. The area under each peak represents the proportion of the record spent at that current level; the mean single channel currents obtained from the Gaussian fits to these peaks were 8.5 pA and -11.4 pA, corresponding to conductance states of 106 pS and -142 pS, respectively. d, Current/voltage relationship for single E9 DNase channels showing that they are independent of the applied p.d. The gradient of the regression line fit to this I/V data is 105 pS.
higher protein concentration (20 nM). The applied p.d. was +150 mV. Because the single channel conductance of the E9 DNase is independent of voltage (
M protein and 2 
* stopped-flow CD spectrophotometer in the absorbance mode (Applied Photophysics). Im9 binding studies of the reduced/oxidized E9 DNase D20C/E66C protein were carried out on a Shimadzu RF5000 or Spex Fluoromax essentially as described by Wallis et al.
