Antibacterial toxin colicin N and phage protein G3p compete with TolB for a binding site on TolA

Most colicins kill Escherichia coli cells by membrane pore formation or nuclease activity and, superficially, the mechanisms are similar: receptor binding, translocon recruitment, periplasmic receptor binding and membrane insertion. However, in detail, they employ a wide variety of molecular interactions that reveal a high degree of evolutionary diversification. Group A colicins bind to members of the TolQRAB complex in the periplasm and heterotrimeric complexes of colicin–TolA–TolB have been observed for both ColA and ColE9. ColN, the smallest and simplest pore-forming colicin, binds only to TolA and we show here that it uses the binding site normally used by TolB, effectively preventing formation of the larger complex used by other colicins. ColN binding to TolA was by β-strand addition with a KD of 1 µM compared with 40 µM for the TolA–TolB interaction. The β-strand addition and ColN activity could be abolished by single proline point mutations in TolA, which each removed one backbone hydrogen bond. By also blocking TolA–TolB binding these point mutations conferred a complete tol phenotype which destabilized the outer membrane, prevented both ColA and ColE9 activity, and abolished phage protein binding to TolA. These are the only point mutations known to have such pleiotropic effects and showed that the TolA–TolB β-strand addition is essential for Tol function. The formation of this simple binary ColN–TolA complex provided yet more evidence of a distinct translocation route for ColN and may help to explain the unique toxicity of its N-terminal domain.


INTRODUCTION
Colicins are plasmid-encoded bacteriocins secreted by Escherichia coli which kill closely related competing bacteria by penetrating their outer membranes and delivering a toxic domain into or beyond the inner membrane (Cascales et al., 2007). Initially, they bind a cell surface receptor, after which they recruit a translocator protein to cross the outer membrane. Colicins are classified into Group A or B according to the periplasmic proteins they use for translocation. Group B uses the Ton system, whilst Group A utilizes the Tol proteins Q, R, A and B, and includes ColA, ColN, ColE1-9 and ColK (Kim et al., 2014). TolA is a 421-residue protein (Levengood & Webster, 1989) consisting of three domains: TolA I , TolA II and TolA III . TolA I is a transmembrane helix, anchored to the inner membrane, which interacts with TolQ and TolR , whilst TolA II is a 250-residue helical region with an opaque role in colicin function (Schendel et al., 1997). TolA III is involved in a range of protein-protein interactions (Kim et al., 2014). TolA has an important role in outer membrane integrity (Fognini-Lefebvre et al., 1987;Webster, 1991) and is also involved in the import of filamentous bacteriophage DNA (Click & Webster, 1998;Webster, 1991). TolB is a 47 kDa twodomain soluble protein which interacts both with TolA III and via a high-affinity interaction with the outermembrane-bound, peptidoglycan-associated lipoprotein Pal (Abergel et al., 1999;Bonsor et al., 2009;Bouveret et al., 1995). DTol mutants are able to replicate, but members of the Tol complex move to the central division zone during cell division, implying a non-essential role in this process (Gerding et al., 2007).
One emerging characteristic of colicin biology is that although the general mechanisms are similar, there is an astonishing variation in the specific molecular interactions employed by different colicins (Grinter et al., 2014;Jakes, 2014;Kleanthous, 2010). ColA and ColN are Group A pore-forming colicins and cell killing is via formation of an ion channel in the inner membrane (Baty et al., 1990). ColA and ColN comprise three distinct domains having roles in each of the three insertion steps. ColN, uniquely, uses LPS as its primary receptor on the cell surface and binds to it with its central receptor-binding domain  ) (Johnson et al., 2014). Like ColE9 (Housden et al., 2010(Housden et al., , 2013 it also binds to outer membrane protein F (OmpF) via an OmpF-binding site at the extreme N terminus of its disordered translocation domain (T domain), ColN-T (ColN 1-90 ) (Johnson et al., 2013). It then binds to TolA III (Anderluh et al., 2003) via a TolA-binding site (TABS) in the central region of  ). Like most other colicins, ColA-R (ColA 173-388 ) binds to a protein receptor, the vitamin B 12 uptake protein, BtuB, as its primary receptor, and then recruits OmpF before ColA-T (ColA 1-172 ) binds to both TolA and TolB (Bouveret et al., 1998;Kim et al., 2014). The pore-forming domains are sited at the C-terminal end. In this study of TolA binding we were thus mainly interested in the colicins' N-terminal T domains which contain intrinsically disordered regions.
A high-resolution X-ray structure of ColA-T 53-107 in complex with TolA 329-421 is available  and the TolAbinding 'box' of  ) has been identified by mutagenesis, NMR and bioinformatics (Anderluh et al., 2003(Anderluh et al., , 2004Gokce et al., 2000;Hecht et al., 2008Hecht et al., , 2009Raggett et al., 1998). The TolQRAB complex interacts with other proteins of the outer membrane and inner leaflet such as Pal, Lpp and OmpA, establishing an energy-dependent link between the inner and outer membranes (Cascales et al., 2000). As well as interactions with colicin T domains, TolA III is also known to bind the N-terminal (N1) domain of G3p (gene-3-protein) -a minor coat protein located at the end of the phage capsid of filamentous bacteriophage Riechmann & Holliger, 1997). A high-resolution structure of TolA 295-421 in complex with G3p-N1 1-86 has been described by X-ray crystallography [Protein Data Bank (PDB) ID: 1Tol] (Lubkowski et al., 1999) and NMR (PDB ID: 1S62) (Deprez et al., 2005). In the native structure of G3p-N1-N2 (PDB ID: 2g3p) (Holliger et al., 1999), the N1 domain binds to the N2 domain via~30-residue contacts and TolA III was found to interact with the same region (Lubkowski et al., 1999). This situation is mirrored in ColN where the ColN-T TABS epitope binds to both ColN-R and TolA III (Hecht et al., 2008). Both ColN-T and G3p-N1 bind TolA III with similar affinity, K D~1 mM (Anderluh et al., 2004;Gokce et al., 2000;Karlsson et al., 2003;Raggett et al., 1998).
The crystal structure of ColN has been resolved by X-ray crystallography to 3.1 Å resolution (Vetter et al., 1998); however, the T domain, a highly dynamic region, was unresolved. It was later shown by NMR that the disordered ColN-T is not entirely unstructured but, in an analogous way to G3p, ColN-T TABS is associated with the rest of Col-N (Hecht et al., 2008). Furthermore, it was shown that ColN-T TABS binds to similar regions of TolA III as G3p-N1 (Hecht et al., 2009). Here, we used mutagenesis to map the ColN-T-binding site of TolA III . Mutants were designed using the current X-ray structures of TolA III in complex with G3p-N1 (Lubkowski et al., 1999) and ColA-T , and previous NMR studies. These mutants were analysed both in vitro using surface plasmon resonance (SPR) and in vivo using whole-cell killing assays to demonstrate a direct link between protein-protein interactions and toxicity for each mutant.

METHODS
Bacterial strains, plasmids and protein purification. See supplementary methods available in the online Supplementary Material.
Spot test killing assay. The activities of ColN and ColA were assayed using the established spot test dilution assay (Pugsley & Schnaitman, 1978). Dilutions of colicin were spotted onto a lawn of JC207 (DTolA) E. coli cells harbouring the TolA-encoding WT pSKL10 or mutant plasmids and incubated at 37 uC for 16 h. The degree of complementation of the TolA deletion was taken as the lowest concentration of colicin that produced a zone of clearance.
Liquid culture killing assay. A single colony of E. coli JC207 cells carrying mutant or WT pSKL10 plasmid was inoculated into 5 ml lysogeny broth (LB) containing 100 mg ampicillin ml 21 [LB(amp)] and grown overnight at 37 uC. Cells were diluted to OD 600 1.7 with LB(amp) and 5 ml added to each well of a 96-well flat-bottomed microtitre plate containing 135 ml LB(amp) (pre-warmed to 37 uC). These were grown at 30 uC, with 4 mm double-orbital shaking at 150 r.p.m. in a FluoStar Optima plate reader, 60 cycles of 600 s with 20 flashes per cycle, until OD 600~0 .4 was reached (cycle 26, 273 min). Then, 15 ml 100 nM ColN (in LB medium) was added. This gave a final ColN concentration of 10 nM, which was above the MIC of 0.5-1 nM (Sharma et al., 2009). Blank wells containing LB(amp) only were subtracted from the data. Growth curves were plotted using SigmaPlot software and represent the means of two separate sets of three wells. The final OD 600 was taken at cycle 42 (153 min after colicin addition) and was used to calculate the percentage killing of ColN (see Table 1) with data normalized to 0 % killing for cells only and 100 % killing for cells complemented with WT pSKL10. SDS sensitivity assay. Using a method similar to that described previously (Walburger et al., 2002), JC207 DTolA cells harbouring pSKL10 (WT or mutant) plasmid or a pUC19 vector control were grown to OD 600 0.5 at 37 uC and then diluted 100-fold into LB(amp) (pre-warmed to 37 uC) supplemented with 0 (LB added), 0.1, 0.5 or 1 % SDS (w/v in LB) and grown at 30 uC. The percentage of surviving cells was calculated from the OD 600 after 180 min growth from the OD 600 of the control pUC19 vector at 0 (100 % survival) and 180 min (0 % survival).
Alkaline phosphatase activity. We used the method described by Torriani (1966). Briefly, cells were cultured in minimal Tris medium (pH 7.4). Single colonies of JC207 periplasmic-leaky cells and cells complemented with WT TolA (pSKL10) or TolA mutant plasmids were inoculated into 5 ml medium and grown for 8 h at 37 uC. An aliquot of 1 ml was inoculated into 50 ml medium and grown overnight at 37 uC, 180 r.p.m. to a final OD 600~2 .5. Cells (1 ml) were centrifuged at 14 000 r.p.m. for 10 min to recover the extracellular extract. Then, 600 ml was mixed with 400 ml p-nitrophenylphosphate substrate (0.2 mg ml 21 in 1 M Tris, pH 8 buffer). After a 10 min incubation, alkaline phosphatase activity was measured by A 410 .

Circular dichroism (CDS) spectroscopy
Far-UV CD (250-185 nm). Samples were exchanged into 20 mM sodium phosphate (pH 7.0) buffer using PD10 columns (GE Healthcare) as per the manufacturer's protocol and concentrations were typically 0.5-0.7 mg ml 21 . Spectra were recorded in a 0.2 mm path-length demountable cuvette using a Jasco-810 spectropolarimeter at 25 uC. Scans (buffer baseline performed under identical conditions subtracted) are given as the differential mean residue extinction coefficient De. The signal at 222 nm is an approximate measure of the a-helical content (Chen & Yang, 1971).
Near-UV CD (320-250 nm). Samples, in the same buffer, were typically 0.5 mg ml 21 . Spectra were recorded in a 1 cm path-length cuvette at 25 uC. Results are given as the molar ellipticity.
Thermal denaturation. Samples in 20 mM sodium phosphate (pH 7) buffer at a concentration of~0.15 mg ml 21 were monitored (in a 1 mm path-length cuvette) at 222 nm from 20 to 90 uC for TolA III constructs and at 288 nm from 25 to 95 uC for G3p-N1 constructs. The fraction of TolA III unfolded (f u ) was calculated at x uC using: where V u represents the far-UV CD value at 222 nm for unfolded TolA III (at 90 uC), V n represents the far-UV CD value at 222 nm for native TolA III (at 20 uC) and V x represents the far-UV CD value at 222 nm for TolA III at x uC. Data were plotted using SigmaPlot software and T m values derived from sigmoidal fits.
Binding studies: SPR. Experiments were performed using CM5 sensor chips using a Biacore X-100 (GE Healthcare) at 25 uC. The running buffer was HEPES-EP (10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA and 0.05 %, v/v, P20 detergent). All immobilizations were performed at 5 ml min 21 using standard aminecoupling carbodiimide chemistry procedures with the ligand proteins at 10 mg ml 21 in 10 mM sodium acetate (pH 4.5) buffer. TolA III , TolA III K420P, TolA III D418P and G3p-N1 V44P were immobilized at 596, 564, 550 and 423 resonance units (RU), respectively, where 1 RU~1 pg protein mm 22 on the surface. For ColN-T binding to TolA III , an improved model fit and more accurate K D were obtained by using a lower immobilization level of 155 RU TolA III to eliminate mass transport effects. In all cases, an activated/blocked flow cell 1 surface was used as the control and subtracted from the data. Analyte proteins were exchanged into HEPES-EP buffer at concentrations described in the text. Binding kinetics were performed at 30 ml min 21 using the Kinetics Wizard (Biacore X-100 control software) and .10 000 10 000 ,0 9 3 Binding of ColN, G3p and TolB to TolA binding analyses were performed at 10 or 30 ml min 21 using manual mode. Surfaces were regenerated by washing with 10 mM glycine (pH 1.8) buffer. Binding models were fitted using BIAevaluation software (v2.0.1) to simultaneously obtain association k a and dissociation k d rate constants, and the equilibrium constant K D. K D was also determined by applying a steady-state affinity model.

ColN-T 1-90 secondary structure predictions
High-resolution structures of TolA III in complex with ColA and filamentous bacteriophage M13 G3p proteins Lubkowski et al., 1999) show that binding occurs through b-strand addition such that a b-strand from one binding partner associates with a b-strand from the other, forming an extended b-sheet structure (Remaut & Waksman, 2006). NMR measurements showed that ColN-T TABS 40-67 folds upon binding TolA III (Anderluh et al., 2003) and analysis of the amino sequence by the program PONDR (Romero et al., 1997) predicted that this region has a high probability of being ordered (Hecht et al., 2008). Thus, it seemed likely that ColN-T TABS forms at least one new b-strand in order to bind TolA; to investigate this possibility we used the structural prediction algorithms I-TASSER (Roy et al., 2010;Zhang, 2008), TALOS (Cornilescu et al., 1999) and Jpred (Cole et al., 2008). An I-TASSER model of T 1-90 predicted two b-strands within the TABS (Fig. 1a, b). Although the C score of the model is low (23.88), it is supported by TALOS and Jpred algorithm results in the S61-H67 region (Fig. 1a). Previous alaninescanning mutant studies using SPR, isothermal titration calorimetry (ITC) and fluorescence spectroscopy showed that two regions, W44-W46 and Y62-F66, are essential for binding TolA III (Anderluh et al., 2003;Gokce et al., 2000) ( Fig. 1a (Anderluh et al., 2004;Gokce et al., 2000;Raggett et al., 1998) is shown as WT binding (,10 mM) (red squares), intermediate binding (green diamonds) and no binding (blue circles). The self-recognition region determined by NMR by Hecht et al. (2008) is highlighted in grey. Schematic representation of the secondary structure predictions by the I-TASSER three-dimensional model (Zhang, 2008) (top line), Jpred algorithm (Cole et al., 2008) (middle line) and TALOS prediction (Cornilescu et al., 1999) of the b-structure using the backbone chemical shifts of the T domain bound to TolA III (bottom line). (b) I-TASSER ribbon model of ColN-T 1-90 ; coloured regions as in (a). (c) Binding of ColA-T (red) and G3p-N1 (green) to TolA III . Drawn in PyMOL (http://www.pymol.org/) using PDB ID: 3QDR and PDB ID: 1Tol, respectively. The ColA and G3p b-strand binding regions are shown in brown and magenta, respectively. Site of G3p-N1 V44P mutation shown in yellow. (d) Primary structure of TolA III S333-P421. Secondary structure elements: ahelices (red blocks) and b-strands (blue arrows). Contacts with G3p-N1, inter-atomic distances ,4.5 Å (yellow dots) (PDB ID: 1Tol; Lubkowski et al., 1999). Residues which showed the strongest NMR chemical shift variations (.0.55 p.p.m.) upon binding ColN-T 1-90 (Hecht et al., 2009) are shown as red dots. The ColA-binding box  is shown. The residues mutated in this study are shaded pink. The ColA (b2) and G3p (b3) b-strand binding regions are shown in brown and magenta, respectively.
The structures of the ColA-T and G3p complexes are shown superimposed in Fig. 1(c), and the secondary structure of TolA III is represented in Fig. 1(c) (taken from PDB ID: 1Tol; Lubkowski et al., 1999). Although we used TolA III296-421 in this study, only residues 333-421 are shown in Fig. 1(d) as aa 296-332 are not resolved in any published structure or thought to be involved in interactions, and therefore were not mutated in this study. Structural contacts with G3p-N1 are indicated by yellow circles and the ColA-binding site  is shown (Fig. 1d). Residues which previously showed the strongest NMR chemical shift variations (.0.55 p.p.m.) upon binding ColN-T 1-90 (Hecht et al., 2009) are indicated by red circles and the residues mutated in this study are shaded pink (Fig. 1d).
Colicin resistance of TolA mutants pSKL10, a plasmid expressing TolA, successfully complemented E. coli JC207 DTolA cells in both liquid culture and spot test assays, restoring both resistance to SDS and sensitivity to ColA and ColN. The selected mutations were made within the TolA III region of pSKL10 and used to complement DTolA E. coli cells. Fig. 2 shows a selection of spot test assays and growth curves representing mutants displaying a WT TolA phenotype (e.g. A415R, cyan), partial phenotype (Y340A, blue; F419A, green) and complete resistance to colicin (I344D, orange) (see Fig. S1 for the full set of mutants). For spot test assays, the degree of complementation conferred by pSKL10 mutant plasmids to DTolA E. coli cells was measured as the lowest concentration of colicin that resulted in a zone of clearance on agar plates (Figs 2 and S1) and is given in Table 1. The resistance to both ColN and ColA was measured, and compared with WT pSKL10 complemented (TolA positive control) and noncomplemented JC207 cells (TolA negative control). For the WT pSKL10 plasmid these values were 100 and 500 nM for ColN and ColA, respectively. Most mutations within the a2 region and those highlighted as targets by NMR showed WT-like activity. Only more structurally disrupting mutations, such as I344D and I348D (Figs 2 and S1), showed complete resistance (.10 000 nM) to both colicins. Growth curves were measured in liquid culture to investigate the effect of ColN in real-time. Again, most mutations displayed a WT colicin-sensitive TolA phenotype, with I344D and I348D showing complete resistance, and Y340A showing a partially resistant phenotype. The 'Y. enterocolitica' TolA chimera displayed a slightly greater resistance to ColN (500 nM) than WT (100 nM), but confirmed that a2 was not a critical site for ColN (Fig. S2). The ColA TABS mutant was resistant to ColA, but sensitive to ColN, indicating that the two colicins bound to TolA at different sites (Fig. S2).

Structural characterization of TolA III mutants
We made the same mutations in a plasmid that expressed only TolA III296-421 , a well-characterized soluble domain truncate lacking the long helices of TolA I-II , and used far-UV CD spectroscopy to measure the secondary structure content. A selection of spectra is presented in Fig. 2(c). The WT TolA III gave a typical ab spectrum characterized by minima at 222 and 208 nm, and was comparable with previously published data (Anderluh et al., 2004;Derouiche et al., 1999). I344D and I348D (data not shown) showed spectra typical of an unfolded structure with negative maxima ,200 nm, which explained their inability to complement the DTolA mutation. Some mutations did not affect the structure, whilst others showed limited loss of a-helical content compared with the WT protein ( Fig.  2c; Table 1). We believe this was due to a partial or full collapse of the a2 or a4 helix.
Testing the hypothesis that ColN-T binds TolA III via b-strand complementation at b3 The previous NMR data (Hecht et al., 2009) and secondary structure predictions (Fig. 1) indicated that the binding region ColN-T residues S61-F66 (SYHITF) form a bstrand upon binding to TolA. Mutating the (b-strand addition) ColA-binding site on the TolA mutant displayed ColA but not ColN resistance in both spot test (Fig. S2) and liquid culture (data not shown) assays, indicating that this was not the binding site of ColN. As G3p-N1 binds to b3 on the opposite side of TolA III , via b-strand addition, and NMR data indicated a common binding site to G3p (Hecht et al., 2009), we then selected this strand for mutation; however, side-chains play a limited role in b-strand addition (Remaut & Waksman, 2006). In the PDB ID: 1Tol structure (Lubkowski et al., 1999), two TolA III b3 residues, D418 and K420, form critical backbone amide-carbonyl hydrogen bonds to G3p-N1 to stabilize the b-sheet interaction between the two proteins. We removed these amide protons by mutation to proline residues which lack the amide protons required for hydrogen bond donation and can only accept hydrogen bonds via the remaining carbonyls. Thus, each proline mutation removed one backbone hydrogen bond which might otherwise stabilize the b-strand addition. These mutations were modelled in the known TolA structure and, following a simple energy minimization, showed, as expected, that prolines had minimal influence on the b-strand backbone conformation (Figs 3a, b). D418V was used as a positive control mutant as valine is a strong b-strand former, can form hydrogen bonds and is one of the 'allowed' residues at this position based on the screening work of Karlsson et al. (2006). They found that only residues A, V, G and E were both resistant to SDS and sensitive to phage infection.
All three mutants had near WT far-UV CD spectra (Fig. 3c), although the proline mutations did cause a shift of the spectra minima towards 200 nm, which may indicate a small degree of unfolding compared with WT. The structural integrity was also determined by thermal denaturation up to 90 u C using far-UV CD at 222 nm (Fig. 3c, inset). A cooperative unfolding curve was observed in all cases and used to calculate the transition midpoint temperatures (T m ) (Fig. 3d). The WT TolA III had a T m of 62.4 u C, as expected (Anderluh et al., 2004). D418P and K420P showed cooperative unfolding curves with reduced T m values of 52.9 and 50.3 u C, respectively. Interestingly, the D418V mutant showed both an increased amount of secondary structure, indicated by a more negative minima at 222-208 nm in the far-UV CD, and a higher T m than WT (64.1 u C).
Spot test assays (Fig. 3e) and liquid culture killing assays (Fig. S1) showed D418P and K420P were resistant to both ColN and ColA, whereas D418V was as sensitive as WT TolA. Furthermore, D418P and K420P were sensitive to  Table 1). (c) A representative selection of far-UV CD spectra of TolA III variants, measured in a 0.2 mm path-length cuvette at 25 6C. The protein concentration was typically 40 mM in 20 mM sodium phosphate, pH 7. Data are shown as WT TolA III (black), Y340A (blue), I344D (orange), A415R (cyan) and F419A (green). 0.1 % SDS, whilst WT and D418V were not (Figs 3d and  S3). This agreed with the known phenotype of mutations within the Tol-Pal system (TolA, TolB, TolQ, TolR and Pal) which caused cells to be hypersensitive to detergents due to disruption of the cell envelope structural integrity (Davies & Reeves, 1975). Finally, both TolA proline mutants were shown to be resistant to ColE9 (Fig. S4), which binds directly to TolB only (Bonsor et al., 2009), and then is bound to TolA via the subsequent interaction of TolB with TolA . It is worth noting that this region in Y. enterocolitica has the sequence VKFPQ instead of the E. coli DFKP, which may allow TolA binding, but not ColN (Weitzel & Larsen, 2008).

Neither ColN-T nor G3p-N1 binds to TolA III proline mutants
For SPR experiments, TolA III , D418P and K420P were immobilized on CM5 sensor chips (~550 RU) before ColN-T and G3p-N1 (0.2-50 mM) were injected over the surface (Table 2). Duplicate injections were performed and the signal from a reference (activated/blocked) blank surface was subtracted from each dataset. Our previous work, carried out using glutathione S-transferase (GST)-ColN-T fusions binding to immobilized TolA III , fitted to a two-site model with K D 50.85 and 0.19 mM for GSTN40 (T domain residues 40-90). In the same study, the results with immobilized GST fusions binding soluble TolA II-III fitted to a one-site model with K D 50.94 mM (Gokce et al., 2000). Thus, the second site was possibly an artefact of TolA immobilization. In the current study, a heterogeneous ligand model fit gave two sites with K D 1 53.0 mM and K D 2 54 nM. The K D 1 is similar to the K D obtained when applying steady-state affinity fitting (1.8 mM) (Fig. S5), a global measure of affinity taking in contributions from both sites, and this is comparable with the published ITC (1 mM) (Raggett et al., 1998), SPR (1.25 mM) (Anderluh et al., 2004) and stopped flow fluorescence (2.3 mM) (Gokce et al., 2000) data. Crucially, no binding to either of the proline mutants occurred. For G3p-N1, we initially followed the previous work of Karlsson et al. (2003) immobilizing~950 RU TolA III on the chip surface, which   gave a similar K D to the published value (~1 mM); however, the fits were poor due to mass transport effects as indicated by high deviation of the residuals plot and a high x 2 (198 RU 2 ). Using a lower immobilization level of 550 RU, better fits (x 2 50.574) were obtained and a 1 : 1 Langmuir binding fit gave a K D of 14.9 mM (Fig. S5). G3p-N1 displayed very fast k on (2.056610 4 s 21 ) and k off (0.307 s 21 ) rates. Hence, steady-state affinity is the most accurate measure of K D , which gave 11.9 mM. No binding was seen by G3p-N1 to either of the proline mutants (Fig. 4).
G3p-N1 V44 is in the TolA III -binding site V44 of G3p-N1 forms two hydrogen bonds with b3 during the b-strand addition to TolA III . We mutated this residue to proline to remove one hydrogen bond by removal of the V44 amide proton. Far-UV and near-UV CD spectroscopy showed that the V44P mutant had a similar fold to the WT protein (Fig. S6), and at 288 nm a cooperative thermal unfolding with T m 49.6 u C was observed (Fig. S6, inset). The WT G3p-N1 protein had a T m of 60.1 u C, similar to the published T m (59.8 u C) calculated from the 230 nm Table 2. SPR data for the interaction of ColN-T, G3p-N1 and G3p-N1 V44P with TolA III and b3 mutants.
No binding was noted for G3p-N1 or ColN-T 1-90 on TolA III D418P or K420P surfaces. No binding was observed for TolA III versus G3p-N1 V44P.

Ligand Analyte
Binding fits Steady-state affinity K D (mM) k a 1 (M "1 s "1 ) k d 1 (s "1 ) k a 2 (M "1 s "1 ) k d 2 (s "1 ) K D 1 (mM) K D 2 (mM)  transition in far-UV CD by Martin & Schmid (2003). For SPR, TolA III was immobilized on a CM5 sensor chip (~550 RU) and varying concentrations of V44P, up to 250 mM, injected over the surface. A small amount of binding was noted .200 mM (~10 RU), but this was confirmed as non-specific by reversing the ligand/analyte and immobilizing G3p-N1 V44P on the chip surface (488 RU) and passing over TolA III (to 250 mM), where no binding occurred.
ColA-T binds to TolA III D418P and K420P. The interactions of isolated ColN-T, ColA-T and G3p-N1 domains with TolA III , D418P and K420P were compared using SPR. Comparable levels of each TolA version were immobilized to the CM5 chip surface. ColN-T, ColA-T and G3p-N1 all bound to the WT TolA III surface (Fig. 4a), but only ColA-T bound to the mutant surface, with ColN-T and G3p-N1 showing no binding (Fig. 4b, c), even at concentrations up to 250 mM (data not shown).
TolA D418 and K420 are in the TolB-binding site.
Although ColA-T can bind to TolA III D418P and TolA III K420P (Figs 4b,c), no killing by ColA was seen in spot test assays (Fig. 3e). We believed that this was due to the disruption of TolA-TolB interactions by these mutants. To test this, 10 mM TolB was injected over TolA III WT and mutant surfaces in SPR experiments. Binding was seen on the WT TolA III surface, but not to either of the mutants (Fig. 4d). As the TolA-TolB interaction was thought to be important for Tol function, we tested the outer membrane integrity of E. coli cells expressing the TolA point mutants. Extracellular levels of alkaline phosphatase, which is normally trapped in the periplasm, were measured for JC207 DTolA cells only and cells complemented with WT TolA III , D418P and K420P plasmids. Baseline alkaline phosphatase activity in culture supernatants, measured as OD 410 , was 0.23, whereas the value for JC207 DTolA cells was 0.72. For cells complemented with the WT plasmid it was 0.24, but TolA III D418P and K420P gave values of 0.67 and 0.69, respectively. Furthermore, WT complemented cells released approximately one-fifth of the amount of extracellular proteins visible from mutant cultures on 10 % SDS-PAGE (data not shown). Thus, the proline point mutants destabilized the outer membrane and created a clear tol phenotype.

TABS on TolB
A TABS has been proposed in the disordered 12 Nterminal residues of TolB (Bonsor et al., 2009;Carr et al., 2000;Walburger et al., 2002), which thus could also engage in b-strand addition, whilst Dubuisson et al. (2002) identified D120 as also being potentially involved in binding. Here, we used alignments of the amino acid sequences of TolB, ColN and G3p-N1 to identify possible TolA III -binding motifs (Fig. S4). First, TolB shares the sequence DGSY (109-112) with the core of the ColN-T TABS (59-62) and VVV (99-101) with residues 61-63 of G3p which are central to its TABS b-strand (Fig. S4). To test if these revealed shared binding sequences, we made the following mutations in TolB: DGSY-AGAA, VVV-AAA and V101P (to inhibit backbone hydrogen bonds). DGSY-AGAA and V101P showed a WT far-UV CD spectrum with VVV-AAA showing a more unfolded spectrum (Fig. S7A).
Cooperative unfolding curves at 217 nm gave T m 550.9, 52.8 and 50.8 u C for DGSY-AGAA, V101P and VVV-AAA, respectively (data not shown). The T m for the WT TolB was 55.4 u C. DGSY-AGAA and V101P mutants displayed WT near-UV CD (Fig. S7B).
TolA III (645 RU) was immobilized on a CM5 chip and 5 mM WT TolB, DGSY-AGAA, V101P or VVV-AAA passed over the surface (Fig. 4e). The 1 : 1 Langmuir binding fits were obtained as described previously (Table 3), giving a K D of 37.3 mM for WT TolB binding to TolA III , in agreement with the published data of 40 mM obtained using ITC (Bonsor et al., 2009). The DGSY substitution was comparable with WT, but TolB VVV and V101P mutants did not bind TolA.

ColN-T-TolA III binding
Previous studies (Anderluh et al., 2003(Anderluh et al., , 2004Gokce et al., 2000;Hecht et al., 2009;Raggett et al., 1998) have defined the TABS of ColN and the structural changes involved upon binding. However, no high-resolution structure is yet available for this complex and the ColN-binding site of TolA III is unknown. TolA has been identified as a translocator for several colicins and phage, with little sequence or structural identity between the proteins involved (Hecht  . It has previously been shown by NMR that the proteins G3p-N1, of filamentous bacteriophage M13, and ColN-T interact with the same region of TolA III (Hecht et al., 2009). Thus, we selected mutant targets using the structure of TolA III complexed with G3p-N1 (PDB ID: 1Tol; Lubkowski et al., 1999) and our previous NMR data (Hecht et al., 2009). Single point mutations along the entire length of TolA a2, and which preserved the native structure, did not affect the toxicity of ColN and ColA. Residues identified by NMR chemical shift changes of .0.55 p.p.m. upon ColN binding (Hecht et al., 2009) were mutated to alanines, but showed little resistance to colicin. Hence, these must be involved in subtle structural rearrangement of TolA III upon binding ColN-T, but the amino acid side-chains are not part of the binding site. The importance of this distinction will become clear below.

TolA III b-strand addition promiscuity
G3p-TolA (Deprez et al., 2005;Lubkowski et al., 1999), ColA-TolA ) and a complex of TolA III of Vibrio cholerae with the N-terminal CTXØ minor coat protein pIII of F bacteriophage (CTXØpIII-TolA) (Ford et al., 2012) are all examples of b-strand addition-type mechanisms (Remaut & Waksman, 2006). As only weak sequence constraints determine the b-strand propensity of peptides, promiscuity of binding partners is possible and binders may be either folded or disordered domains. Secondary structure predictions (Cole et al., 2008;Cornilescu et al., 1999;Roy et al., 2010;Zhang, 2008) all suggest that the ColN sequence Y62-H67 adopts a b conformation and, in our previous NMR study, the binding of G3p caused large 15 N-1 H-heteronuclear single quantum coherence chemical shift changes in b2 and b3 of TolA. These were largely imitated by ColN-T binding with even larger effects on b3 (Hecht et al., 2009). Deprez et al. (2005) described a 'stretching' of the TolA III b-sheet (b2+b1+b3) upon G3p-N1 binding; in particular, b3 is extended and a 'bulge' present in b2 disappears. The PDB ID: 1Tol structure shows that residue V44 of G3p contributes to backbone hydrogen bonding in b-addition and we made a G3p V44P mutant to remove the critical amide proton. This mutant did not bind TolA III .
Two TolA III residues, D418 and K420 (Figs 1 and 3), are similarly involved in the hydrogen-bonding network with G3p-N1. D418 does show a chemical shift upon ColN-T binding, but was below our 0.55 p.p.m. cutoff for the initial round of mutagenesis, and K420 was not assigned in the NMR spectrum due to its flexibility at the terminus of the protein. Proline mutants should not strongly disrupt the b structure (Chou & Fasman, 1974) and are the clearest way to test for b-strand addition (Fig. 3) (Joseph et al., 2013). Both mutations showed small structural changes by CD and still bound the ColA-T domain in a SPR assay, but showed clear ColN-resistant phenotypes. These results confirm that ColN-T 1-90 binds via a b-strand addition to TolA III b3 with ColN-T contributing at least one b-strand from within the TolA box, probably residues Y62-H66. Interestingly, Karlsson et al. (2006) performed a random mutagenesis screen on this region of TolA III , A415-K420, and proline mutants at these positions were screened out as being phage resistant. Their work shows that sequence conservation in this area is second to structural preference with at least one of the b-promoting WT residues (L417, D418 or K420) being retained or replaced with another strong b-former (such as V, L, M and I). It would be interesting to challenge these libraries with ColN and ColA.

Importance of D418P and K420P mutations in ColA and ColE9 killing
Although the ColN-T-and ColA-T-binding sites are on opposite sides of the TolA III structure, and ColA-T also binds the mutant proteins, the proline mutations within TolA III b3 still cause resistance to ColA. ColA clearly has a different mechanism of translocation to that of ColN and it has been proposed that ColA-T interacts first with TolB in the periplasm via its low-affinity TolB box, followed by high-affinity interaction with TolA . TolA has been shown to interact with TolB in vitro by ITC (K D~4 0 mM) (Bonsor et al., 2009), cross-linking and yeast two-hybrid experiments (Walburger et al., 2002), and this interaction is enhanced (K D~1 3 mM) when ColE9 binds to TolB (Bonsor et al., 2009).
However, the interpretation of ColA and ColE9 activity is complicated by the discovery that these point mutations give cells a DTol phenotype including sensitivity to SDS. This implies a disruption of the cell envelope due to the inhibition of the TolA-TolB complex. We used an alkaline phosphatase assay (Lazzaroni & Portalier, 1981;Torriani, 1966) to show that periplasmic alkaline phosphatase, which is larger (79 kDa) than TolB (47 kDa), leaches into the extracellular medium of DTolA periplasmic-leaky cells (JC207). Crucially, similar amounts of alkaline phosphatase leached out from TolA III D418 and K420 mutants. As we cannot determine any effects on the TolB-Pal interaction, we cannot thus rule out that the inability of ColA and ColE9 to kill these mutants may be due to very low levels of periplasmic TolB rather than a need for a TolA-TolB interaction to maintain the ColA-TolB-TolA III or ColE9-TolB-TolA III complex.
The two different binding sites for TolB and ColA-T on TolA III support the formation in vivo of a heterotrimeric complex which has been observed by gel filtration in vitro (Hecht et al., 2010), but it is not clear if a direct TolB-TolA interaction is required for the toxic activity of ColA, as it is for ColE9, or whether the bifunctional ColA-T cross-links the complex. The disordered extreme N terminus of TolB has been shown to bind TolA III and the NMR spectrum of a TolA III -TolB 1-13 complex resembles that of a TolA III -ColN-T complex . Thus it is likely that TolB and ColN-T bind via a similar b-strand addition mechanism.
The combined data for three different Group A (Toldependent) colicins give an insight into their use of the same Tol proteins in fundamentally different ways, but for apparently the same purpose. ColE9 binds only TolB, but by doing so stabilizes the TolA-TolB complex (Bonsor et al., 2009). ColA requires direct interaction with both TolA and TolB, but it is not clear if a direct TolA-TolB complex is required . Finally, ColN is unique in not requiring TolB and binds TolA at the precise site where TolB would bind. ColN consistently displays minimalistic behaviour as demonstrated by its small size, LPS receptor (Johnson et al., 2014), single protein receptor (Clifton et al., 2012;Johnson et al., 2013) and lack of TolB dependence. This simplicity makes ColN unique in colicin biology, and the disruption of TolA-TolB may be relevant to our recent discovery that ColN-T alone can target and kill E. coli (Johnson et al., 2013).