Genome-wide screens: novel mechanisms in colicin import and cytotoxicity


*E-mail; Tel. (+1) 765 494 4956; Fax (+1) 765 496 1189.


Only two new genes (fkpA and lepB) have been identified to be required for colicin cytotoxicity in the last 25 years. Genome-wide screening using the ‘Keio collection’ to test sensitivity to colicins (col) A, B, D, E1, E2, E3, E7 and N from groups A and B, allowed identification of novel genes affecting cytotoxicity and provided new information on mechanisms of action. The requirement of lipopolysaccharide for colN cytotoxicity resides specifically in the lipopolysaccharide inner-core and first glucose. ColA cytotoxicity is dependent on gmhB and rffT genes, which function in the biosynthesis of lipopolysaccharide and enterobacterial common antigen. Of the tol genes that function in the cytoplasmic membrane translocon, colE1 requires tolA and tolR but not tolQ for activity. Peptidoglycan-associated lipoprotein, which interacts with the Tol network, is not required for cytotoxicity of group A colicins. Except for TolQRA, no cytoplasmic membrane protein is essential for cytotoxicity of group A colicins, implying that TolQRA provides the sole pathway for their insertion into/through the cytoplasmic membrane. The periplasmic protease that cleaves between the receptor and catalytic domains of colE7 was not identified, implying either that the responsible gene is essential for cell viability, or that more than one gene product has the necessary proteolysis function.


Protein import and export across cellular membranes involve an assembly of integral membrane polypeptides that provide a conduit for protein transfer across the hydrophobic membrane. Colicin import into bacteria is a model for the study of protein import across bacterial membranes. Colicins are produced by Escherichia coli in response to stresses such as nutrient depletion and overcrowding, and kill closely related sibling cells that contain the set of receptors and trans-envelope import proteins, but do not contain the respective immunity protein. Based on known components of the translocation network utilized for cell entry, colicins have been divided into two groups. Group A utilizes the Tol network consisting of genes tolA, tolB, tolQ, tolR and pal (Nagel de Zwaig and Luria, 1967; Davies and Reeves, 1975a; Lazzaroni et al., 2002). Group B employs the Ton network comprised of tonB, exbB and exbD (Davies and Reeves, 1975b; Braun et al., 2002). Loss of the colicin receptors in the outer membrane, proteins in the translocation pathway, or of the immunity protein, are the only known mechanisms of colicin resistance or tolerance.

The first step in colicin import is binding to its outer membrane primary receptor. The protein component of this primary receptor is known to be BtuB for colicin (col) A, E1, E2, E3 and E7, OmpF for colicin N, and FepA for colicins B and D. Subsequent to binding to the primary binding steps, nuclease E colicins, for which the import process has been studied extensively (Di Masi et al., 1973; Mock and Pugsley, 1982; Benedetti et al., 1989; Bouveret et al., 1997; Garinot-Schneider et al., 1997; Kurisu et al., 2003; Housden et al., 2005; Duche et al., 2006; Loftus et al., 2006; Duche, 2007; Sharma et al., 2007; Yamashita et al., 2008), utilize a second receptor/translocator to accomplish translocation across the outer membrane. This secondary receptor/translocator is OmpF for colicin A, TolC for colicin E1 and OmpF/OmpC for colicins E2, E3 and E7. No secondary receptor/translocator has yet been identified for colicin N and the group B colicins. In addition to the primary receptor and secondary receptor/translocator, several other proteins have been identified that are responsible for colicin import across the double membrane of the target cell. A total of 10 genes, btuB, iutA, ompF, ompC, tolC, tolQRAB and tsx have been identified to play a role in the uptake of group A colicins, while the corresponding number for group B colicins is nine [cir, exbBD, fepA, fhuA, lepB, tonB, tsx (Cascales et al., 2007) and the recently identified fkpA (Hullmann et al., 2008)]. This number is considerably smaller than the number of proteins showed to be involved in protein transport across the mitochondrial double membrane. In the mitochondrial system, virtually all the preproteins traverse the outer membrane through the initial entry gate, the translocase of the outer membrane complex that consists of seven polypeptides. The full translocation of the preproteins across the inner membrane into the mitochondrial matrix requires 10 more polypeptides (Bohnert et al., 2007; Bolender et al., 2008). Thus, on the one hand, it might be expected that additional components, proteins or lipids as yet unidentified, could be involved in colicin uptake across the E. coli double membrane. On the other, it is possible that some gene(s) previously proposed to be required for colicin cytotoxicity are not essential. The ‘Keio collection’ ( can be used to clarify this situation.

This collection, a single-gene knockout library of the entire E. coli genome (Datsenko and Wanner, 2000; Baba et al., 2006), consists of 3985 strains, in duplicate, inoculated in 96-well plates, each well consisting of a distinct single-gene knockout mutant. In the present study, the collection was screened for the sensitivity of each knockout strain to eight colicins, A, B, D, E1, E2, E3, E7 and N, allowing for the identification of knockout strains that are able to grow in the presence of colicin. These colicin ‘tolerant’ or ‘resistant’ single-gene knockout strains were further analysed by complementation with the respective genes from the ‘ASKA’ (a complete set of E. coli K-12 ORF archive) orfeome library, which restored the colicin sensitivity to the knockout strain. As only the genes nonessential for cellular viability are represented in the Keio collection, proteins that are required by colicin but are also essential for cell growth (e.g. lepB for colD) will not be scored in the present genomic screening. It is noted that while a ‘positive hit’ implies that the particular gene is important for the colicin cytotoxicity, a ‘negative hit’, where a gene deletion does not affect colicin activity, is equally significant. A ‘negative hit’ implies that, barring a redundancy in the gene function, the gene is not critical for colicin cytotoxicity. Through this study we have been able to define the requirement of the 3985 genes in the Keio collection for colicin cytotoxicity.

Results and discussion

Colicin cytotoxicity and receptor function

The minimum colicin concentration at which cytotoxicity is expressed can be measured qualitatively by ‘spot titers’, the minimum amount of colicin that generates a clear ‘spot’ on a Petri dish covered with sensitive cells. As determined through multiple assays (n = 3), using colicin whose purity was tested and confirmed by SDS-PAGE, and whose concentration was accurately determined (see Experimental procedures), this concentration for colE1 is ∼125–250 pM, 50–100 pM for colicins E2, E3 and E7. For colB, colD and colN, ∼500 pM−1 nM generates a clear zone of inhibition. The minimum inhibitory concentration for colA is ∼2.5–5 nM. Spot titers were also used to estimate the receptor binding affinity of E colicins for BtuB (Taylor et al., 1998; Kurisu et al., 2003). Colicin E2, E3 and E7 were neutralized by BtuB at a colicin : BtuB ratio of 1:1, implying a high affinity, i.e. Kd∼10−9 M (Kurisu et al., 2003). The binding affinity of colE1 for BtuB is significantly lower (Imajoh et al., 1982; Taylor et al., 1998), so that BtuB neutralized colE1 when added at a ratio of ∼1:40 (colicin : BtuB). Because of the low binding affinity of colN for OmpF [Kd = 2 × 10−4 M (Evans et al., 1996a,b)], neutralization of colN by OmpF could not be measured. Similarly, the binding affinity of colA to BtuB could not be estimated using the cytotoxicity neutralization assay, implying a relatively low affinity of colA towards BtuB.

I. Colicin N

Screening of the Keio collection for genes that are required for colN cytotoxicity yielded a number of new genes (Table 1), all of which are involved in the biosynthesis of lipopolysaccharide (Fig. 1A and B). Knockout of genes galU, gmhA, gmhB, gmhC, gmhD, rfaH, waaC, waaF, waaG and waaP conferred cellular resistance to colN, as can be seen from analysis of their growth in the presence and absence of colN (Fig. 2A and B). Complementation with the corresponding gene from the ASKA collection restored sensitivity to colN for all of the above strains (data not shown).

Table 1.  List of genes affecting the cytotoxicity of colicins A, B, D, E1, E2, E3, E7 and N.
GeneaMutantCellular/colicin import functionsComplementationSource
Colicin N
 gmhAgmhA740::kanLPS core biosynthesisYesThis study
 gmhBgmhB725::kanLPS core biosynthesisYesThis study
 gmhCgmhC745::kanLPS core biosynthesisYesThis study
 gmhDgmhD 731::kanLPS core biosynthesisYesThis study
 waaCwaaC 733::kanLPS core biosynthesisYesThis study
 waaFwaaF 732::kanLPS core biosynthesisYesThis study
 waaGwaaG 742::kanLPS core biosynthesisYesThis study
 waaPwaaP741::kanLPS core biosynthesisYesThis study
 rfaHrfaH 783::kanRegulates the expression of genes waaB, waaG, waaR, waaP and waaQYesThis study
 galUgalU745::kanLPS core biosynthesisYesThis study
 tolAtolA788::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 tolQtolQ786::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 tolRtolR787::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 ompFompF746::kanOuter membrane porin FNTPrevious studiesd and this study
 ompRompR739::kanRegulates production of OmpF and OmpCNTPrevious studiesd and this study
Colicin A
 gmhBgmhB725::kanLPS core biosynthesisYesThis study
 yciByciB762::kanInvolved in zinc uptakeYesThis study
 rffTrffT744::kanInvolved in enterobacterial common antigen synthesisYesThis study
 tolCtolC732::kanOuter membrane component of the tripartite efflux pumpsYesPrevious studiese and this study
 hnshns746::kanRegulates OmpF and OmpC expressionNTThis study
 surAsurA765::kanPeriplasmic chaperone for OM porinsNTThis study
 cpxAcpxA771::kanInvolved in OmpF and OmpC regulationYesThis study
 yeiLyeiL790::kanInvolved in mid-term, stationary-phase viability under nitrogen starvation.NobThis study
 tolAtolA788::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 tolBtolB789::kanUptake of some group A colicinsNTPrevious studiesc and this study
 tolQtolQ786::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 tolRtolR787::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 ompFompF746::kanOuter membrane porin FNTPrevious studiesd and this study
 ompRompR739::kanRegulates production of OmpF and OmpCNTPrevious studiesd and this study
 btuBbtuB754::kanReceptor for vitamin B12 uptakeYesPrevious studiesf and this study
Colicins B and D
 tonBtonB760::kanTogether with OM receptors involved in the uptake of iron and cyanocobalaminNTPrevious studiesg and this study
 exbBexbB779::kanDeletion causes tolerance or sensitivity to group B colicinsNTPrevious studiesg and this study
 exbDexbD778::kanDeletion causes tolerance or sensitivity to group B colicinsNTPrevious studiesg and this study
 fepAfepA721::kanFerric enterobactin outer membrane receptorYesPrevious studiesg and this study
Colicin E1
 tolAtolA788::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 tolRtolR787::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 tolCtolC732::kanOuter membrane component of the tripartite efflux pumpsYesPrevious studiesh and this study
 btubbtuB754::kanReceptor for vitamin B12 uptakeYesPrevious studiesf and this study
Colicins E2, E3 and E7
 tolAtolA788::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 tolBtolB789::kanUptake of some group A colicinsNTPrevious studiesc and this study
 tolQtolQ786::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 tolRtolR787::kanTolerance to group A colicins; maintains cell envelope integrityYesPrevious studiesc and this study
 ompRompR739::kanRegulates production of OmpF and OmpCNTPrevious studiesd and this study
 btuBbtuB754::kanReceptor for vitamin B12 uptakeYesPrevious studiesf and this study
Figure 1.

A. Pathway for synthesis of ADP-L, D-heptose (▴ in B). All genes in this pathway were found to be important for cytotoxicity of colicin N.
B. Biosynthesis of LPS by addition of successive heptose (▴) and glucose moieties (Δ). The galU and waaG gene deletion has a partial phenotype, waaC, waaF and waaP genes were required, while waaO, waaQ and waaR were not required for colicin N activity (waaU is not represented in the Keio collection). WaaP is involved in phosphorylation of the first heptose (o). Based on the partial phenotype of the galU and waaG knockouts, it is inferred that colN interacts with the LPS inner core along with the first added glucose.

Figure 2.

Effect of colN on the growth of wild type and knockout strains.
A. Knockout strains ΔgalU, ΔgmhA, ΔgmhB and ΔwaaC strains and their complemented counterpart (CΔgalU, CΔgmhA, CΔgmhB and CΔwaaCCΔwaaF) were tested for colN sensitivity.
B. LPS mutant strains ΔrfaH, ΔwaaF, ΔwaaG and ΔwaaP were also resistant to colN and their complemented counterpart (CΔrfaH, CΔwaaF, CΔwaaG and CΔwaaP) were tested for colN sensitivity. The growth of these strains in the absence of colicin has not been shown for purposes of clarity. OD410 of 0.1, 0.45, 0.7, 0.1 and 1.2 correspond approximately to a cell titer per ml of 0.7 × 108, 2.6 × 108, 5.5 × 108, 25 × 108 and 28 × 108 respectively.

(i) gmhA, gmhB, gmhC and gmhD.  All of these genes encode for enzymes involved in the biosynthesis of the heptose moiety [adenosine diphosphate 5′-l-glycero-β-d-manno-heptose (ADP-L, D-Hep), ▴, Fig. 1A and B)] of the lipopolysaccharide from sedoheptulose 7-phosphate, and were found to be essential for colicin N cytotoxicity. All strains were positively complemented with the corresponding genes from the ASKA collection (Fig. 2A) (data not shown for ΔgmhC and ΔgmhD). Deletion of gmhB has been shown to cause a partial defect in lipopolysaccharide (LPS) synthesis, so that two different populations of LPS exist, heptose-minus and heptose-rich forms, suggesting that another function in the E. coli cell can at least partially compensate for the role of GmhB in LPS synthesis (Kneidinger et al., 2002). Knockout of the gmhB gene, however, creates resistance to colN (Figs 2A and 3D), suggesting that the concentration of heptose-rich complete LPS molecules in the ΔgmhB cells is not sufficient for colN activity.

Figure 3.

Comparison of the effect of colA and colN on the growth of wild type and knockout strains.
A and B. ΔwaaC cells (A) and ΔgmhA (B) are resistant to colN, but sensitive to colA. Sensitivity to colN is restored upon complementation of these strains (CΔgmhA and CΔwaaC). Thus, the resistance of these strains to colN is not attributed to the low titers (10% of wild-type levels) of OmpF in the outer membrane. Rather, it is inferred that LPS plays a direct role in colN cytotoxicity.
C. ΔtolC cells are resistant to colA, most likely due to the decreased level of OmpF in the outer membrane [2% of wild-type levels (Misra and Reeves, 1987)] in ΔtolC cells. However, due to the higher affinity of colN for OmpF, ΔtolC cells are killed by colN.
D. ΔgmhB cells are resistant to both colN and colA. Sensitivity to these colicins is restored upon complementation (CΔgmhB). Because colA is the least potent of all group A colicins and utilizes only OmpF, in contrast to nuclease E colicins that can utilize both OmpF and OmpC, the above strains were not tested individually with nuclease E colicins. The concentration of colA and colN was 5 and 0.5 nm respectively.

(ii) waaC and waaF.  The ADP-heptose moiety (ADP-L, D-Hep) is sequentially added to the 3-deoxy-D-manno-oct-2-ulosonic (Kdo) unit by the waaC and waaF gene products. Deletion of these genes renders the cells completely resistant to colN (Fig. 2A and B). Sensitivity to the colicin is restored when the knockout cells are complemented with the respective genes.

(iii) waaG, waaP and rfaH.  The rfaH gene product, which positively regulates the expression of genes waaB, waaG, waaR, waaP and waaQ (Pradel and Schnaitman, 1991), is required for colN activity so that its deletion results in resistance to colN and complementation restores sensitivity (Fig. 2B). Primary screening and further verification for sensitivity to colN showed that while WaaP is essential, ΔwaaG cells are partially resistant (Fig. 2B). WaaB, WaaO, WaaR and WaaQ are not required, implying that the effect of rfaH deletion on colN cytotoxicity is due to its positive regulation of waaP and waaG expression. WaaP, a LPS kinase involved in the addition of phosphate to the first heptose (Fig. 1B), is necessary for the sequential action of WaaQ and WaaY. However, waaQ deletion does not affect colN cytotoxicity (data not shown), implying that only the phosphorylation of the first heptose by WaaP is necessary for colN cytotoxicity.

(iv) waaG and galU.  Deletion of galU, which is involved in the production of UDP-glucose, a substrate for WaaG, renders the cells partially resistant (Fig. 2A) to colN. This partial resistance of ΔgalU cells for colN can be explained by the leaky LPS phenotype of such a strain (Schnaitman and Klena, 1993). As mentioned above, ΔwaaG cells are also partially resistant to colN (Fig. 2B). Sensitivity of both of these strains to colN is restored upon complementation (Fig. 2B). However, unlike other colN-resistant strains, ΔwaaG cells in the presence of the colicin display a long incubation lag of ∼3 h before the initiation of significant growth. The reason for this lag in growth is presently unclear. GalU and WaaG together add the first glucose to the LPS inner core and are required for colN cytotoxicity, while WaaO and WaaR, which extend the LPS by adding the next two glucose moieties, are not necessary. Therefore, the LPS inner core along with the first added glucose defines the minimum binding site for colN. In similar screening studies of the ‘Keio collection’ with T7 phage, it was concluded that the first glucose of the inner core is also essential for cytotoxicity of the phage (Qimron et al., 2006).

(v) Deep rough phenotype.  Mutation in the genes gmhA, gmhC, gmhD, waaC, waaF and waaP cause the ‘deep rough phenotype’ that has alterations in the LPS structure and results in a >90% reduction in the concentrations of the porins OmpF, OmpC, LamB and PhoE (Schnaitman and Klena, 1993). Thus, resistance of strains with the deep rough phenotype to colN could be attributed to the reduced concentration of porins in the outer membrane. However, all of these strains were efficiently killed by the seven other colicins used in this study, each of which requires an outer membrane receptor and translocator to for passage across the outer membrane.

While OmpF is the primary high affinity receptor for colicin N, it serves as the secondary low affinity receptor for colA. Therefore, colA is more susceptible to decreased levels of OmpF [see sections II (iv) and VI (iii)] and is expected to have a much lower affinity for OmpF than colN. As evidenced by the growth curves of ΔgmhA (Fig. 3A), ΔwaaC (Fig. 3B), ΔwaaF and ΔwaaP cells (data not shown), these deep rough LPS mutants were efficiently killed by colA, implying that the decreased sensitivity of these cells to colN is a consequence of an altered interaction with LPS.

II. Colicin A

Primary screening of the genes required for colicin A cytotoxicity led to the identification of several new genes (Table 1). ΔtolC cells were completely resistant to colA and sensitivity was restored upon complementation with tolC(Figs 3C and 4). The strains ΔgmhB, ΔrffT and ΔyciB showed varying degrees of resistance to colA. Complementation with the corresponding genes from the ASKA collection restored the sensitivity of these strains to wild-type levels (Fig. 4). The cpxA and hns genes, involved in regulating the expression of OmpF (Suzuki et al., 1996; Batchelor et al., 2005; Skerker et al., 2005), and SurA, required for OmpF folding and insertion into the outer membrane (Lazar and Kolter, 1996), are essential for colA activity. The primary screening also identified a requirement of yeiL gene for colA cytotoxicity. Although the ΔyeiL strain could not be complemented, it was resistant to colA, but sensitive to other group A colicins, E1, E2, E3, E7 and N (data not shown).

Figure 4.

Effect of colA on the growth of wild type and knockout strains. Knockout strains ΔyciB, ΔrffT, ΔgmhB and ΔtolC strains and their complemented counterpart (CΔyciB, CΔrffT, CΔgmhB and CΔtolC) were tested for colA sensitivity. The knockout strains were resistant to colA. Complementation with the respective genes from the ASKA collection restored their sensitivity. The growth of these strains in the absence of colicin has not been shown for purposes of clarity. It is noted that the CΔtolC strain is more sensitive than the wild-type cells to colA.

(i) rffT.  RffT, a TDP-Fuc4NAc : Lipid II transferase (Rahman et al., 2001), catalyses the synthesis of Und-PP-GlcNAc-ManNAcA-Fuc4NAc (Lipid III), the third lipid-linked intermediate involved in enterobacterial common antigen (ECA) synthesis. ECA is present in the outer membrane of all Gram-negative enteric bacteria (Makela and Mayer, 1976; Mayer and Schmidt, 1979; Kuhn et al., 1988). The absence of rffT causes a partial resistance to the colicin and complementation restores colA cytotoxicity (Fig. 4). The absence of the genes, rfe and rffM, which are also involved in the synthesis of the first and second lipid-linked intermediate in the ECA synthetic pathway, did not affect colA activity (data not shown). Although the role of this gene in colA cytotoxicity is unclear, it is possible that colA interacts weakly with the ECA. Alternatively, the effect of rffT on colA might be distinct from its role in ECA synthesis, indicating another function of the gene or a secondary effect of the gene deletion. An additional effect of mutations in the rffT gene is accumulation of lipid II, which stimulates the transcription of degP that encodes a heat shock-inducible periplasmic protease (Danese et al., 1998). Thus, stimulation of degP transcription might induce degradation of the colicin in the periplasmic space, leading to some degree of colicin resistance.

(ii) gmhB.  The gmhB gene product, a heptose 1,7-bisphosphate phosphatase, is involved in the synthesis of d-glycero-D-manno heptose 1-phosphate in the pathway of LPS synthesis (Kneidinger et al., 2002). The knockout of the gmhB gene creates resistance to colA. Complementation of the ΔgmhB cells with gmhB+ gene restores the cytotoxicity of colA (Figs 3D and 4). However, other strains deficient in LPS synthesis (including ΔgmhA) that are resistant to colN, were killed efficiently by colA (Fig. 3A and B). In addition, as mentioned above, deletion of gmhB has only a partial defect in the synthesis of LPS core (Kneidinger et al., 2002), implying that an unknown function of gmhB, distinct from its function in LPS synthesis, is required for colA cytotoxicity.

(iii) tolC.  Deletion of the tolC gene caused resistance to colA (Figs 3C and 4). This is an unexpected result, because TolC had not been recognized as a requirement for colA cytotoxicity. The likely explanation is that tolC deletion decreases OmpF expression by approximately two orders of magnitude, while at the same time increasing OmpC expression (Misra and Reeves, 1987; Benedetti et al., 1991). Complementation of ΔtolC cells with the tolC gene restored the sensitivity to colA.

(iv) Genes affecting outer membrane levels of OmpF.  The knockout of surA and yciB genes results in resistance to colA (data not shown). The products of these genes have been hypothesized to be involved in membrane integrity, and their absence causes a reduction in outer membrane levels of OmpA, OmpC, OmpF and LamB porins (Lazar and Kolter, 1996; Niba et al., 2007). The CpxA-CpxR envelope stress response system regulates the levels of OmpF and OmpC in the outer membrane, so that cpxA deletion causes a ∼20-fold reduction in the level of OmpF (Batchelor et al., 2005). Hns is also involved in the regulation of OmpF expression so that mutations in hns can decrease OmpF and increase OmpC levels (Suzuki et al., 1996). Resistance to colA, caused by deletion of these genes, is inferred to be a consequence of their effect on OmpF levels and the relative low affinity of colA for OmpF [also see section VI (iii)]. However, all of these strains are killed by colN, implying that they produce sufficient OmpF for colN to exert cytotoxicity and reinforcing the point that the affinity of colN for OmpF is greater than that of colA.

Although colA is able to kill the cells with ‘deep rough phenotype’ in which the OmpF level is reduced to approximately 10% (∼104/cell) of the wild-type levels, ΔcpxA cells, in which the amount of OmpF in the outer membranes is ∼20 times less (∼5000/cell) than the wild-type level, are resistant to the colicin. Thus, it is estimated that colA requires at least 5 × 103−104 molecules of OmpF in the outer membrane for efficient cytotoxicity.

(v) yeiL.  The gene, yeiL, was found to be required for colA cytotoxicity so that its deletion caused complete resistance to colA, but not to the other group A colicins tested. YeiL, a member of the CRP-FNR family (Gostick et al., 1999), is required for viability under nitrogen starvation (Anjum et al., 2000). However, this strain could not be complemented by the corresponding genes from the ASKA library. This was anticipated as complementation by a high-copy yeiL+ plasmid, followed by isopropyl-1-thio-β-D-galactopyranoside (IPTG)-induced overproduction, was not productive because the amplified protein was insoluble (Anjum et al., 2000). Alternatively, non-complementation raises the possibility that more than one gene is missing in the knockout strain and the colicin resistance was due to the absence of some other gene essential for the cytotoxicity of that colicin. For this reason, the yeiL knockout strain must be analysed further before any conclusion can be made about its function in colA cytotoxicity.

III. Colicin E1

Primary screening (Table 1) and further verification of the strains in which different Tol proteins were deleted revealed that TolR and TolA are essential for colE1 cytotoxicity. However, the requirement for TolQ is only partial (Fig. 5), as its absence does not result in a loss, but only diminished (∼5–10 fold), sensitivity, to colE1 (data not shown). The high concentration of colE1 used in the primary screening (10 nM) prevented observation of this small effect of tolQ deletion on colE1 cytotoxicity. The ΔtolQ strain, however, is resistant to colicins A, E2, E3, E7 and N (Fig. 5, data shown for colE3). Complementation of all the three strains ΔtolQ,ΔtolR and ΔtolA with the respective genes from the ASKA library restored the colicin sensitivity of these strains (Fig. 5).

Figure 5.

TolA and TolR are essential for colicin E1 cytotoxicity while TolQ is not. ΔtolA, LB plates containing no colicin (top left), colicin E1 (top right) or colicin E3 (bottom left) were divided into eight sections each containing different cells. These are: (1) ΔtolA; (2) CΔtolA; (3) and (4) wild-type cells containing empty vector; (5) ΔtolQ; (6) CΔtolA; (7) ΔtolR; (8) CΔtolR;. Because, IPTG-induced overexpression of TolQ, TolR and TolA interferes with cell growth, IPTG was not added to these plates and complementation occurred due to leaky expression of the ASKA clones.

Previous studies of the role of Tol proteins in colE1 cytotoxicity appear to have yielded contradictory results. TolA (TolII in older literature), TolQ, and TolR have been indicated to be important (Lazzaroni et al., 2002), although other studies have inferred an absolute requirement for TolA and TolQ, but not for TolR (Benedetti et al., 1991; Lazdunski et al., 1998). It is noted that due to a weak/defective ribosome binding site, expression of TolR depends on the successful translation of TolQ (Vianney et al., 1996), a requirement that can be bypassed by an increase in affinity of the ribosome binding site of tolR. Therefore, it is likely that at least some of the absolute requirement for TolQ, described previously, might be attributed to the polar effect of TolQ on TolR translation. Yet other studies have identified a TolQ point mutant (TolQ66/TPS66) that is insensitive to E1 (Sun and Webster, 1987; Vianney et al., 1994). However, complementation of this TolQ mutant restored envelope integrity of the cells but not sensitivity to colE1, implying that a secondary mutation is responsible for the colE1 resistance. In the present study, the ΔtolQ strain from the Keio collection was complemented by the tolQ ASKA clone (Fig. 5), implying that it is a true tolQ knockout. It is also noted that at least five Tol mutants (Tol III, IV, XIII, XIV and XV) have been identified that were sensitive to colE1 but resistant to colicins, A, E2, E3 and K (Davies and Reeves, 1975a). Tol III was later renamed TolB, while the Tol IV, XIV and XV mutations were traced to the OmpR locus (Sarma and Reeves, 1977). A similar colicin resistance pattern of each of these genes has been detected in the present study. Thus far it has not been possible to determine the relation between Tol XIII and TolQRAB or OmpR but, given the data in the present study, the simplest inference is that this mutant is a TolQ mutant. It is noted that different colE1 proteins exist that have subtle differences in the sequence when compared with the plasmid colE1 (GeneID: 2693967) sequence that was used in the present study. In the pioneering study of Davies and Reeves (1975a) four different E1 protein sources were used. Although the identity/sequence of colE1 used in the previous studies is not always clear, all of these colE1 showed a similar resistance pattern. Finally, it is noted that colicin cytotoxicity determination is dependent on the efficiency of the colicin. ColE1 preparations that have low activity would significantly exaggerate the Tol requirement of colicin. ColE1 used in the present studies is cytotoxic at subnanomolar concentration in ‘spot titer’ tests.

It has been postulated that, due to the role of RfaH and GalU in LPS synthesis, these proteins affect the export and insertion of TolC into the outer membrane (Wandersman and Letoffe, 1993). Thus, the absence of rfaH and galU results in a reduced level of TolC in the outer membrane. However, in the present studies all LPS mutants identified to be important for colN, including ΔrfaH and ΔgalU, were sensitive to colE1 (data not shown). The sensitivity of these strains to colE1 implies that the decreased level of TolC in the outer membrane is sufficient for this colicin. Because GalU affects only the export of TolC to the outer membrane and not the production of TolC, the ΔgalU strain presumably has normal/sufficient amounts of OmpF in the outer membrane to allow sensitivity to colA.

IV. Colicins E2, E3 and E7; proteolysis is necessary for import

All strains in the Keio collection except those missing the expected tolA, tolB, tolQ, tolR, btuB and ompR genes were found to be sensitive to these colicins (Table 1). One of the genes that is believed to be required for these colicins, whose identity is not yet known, is that of a protease that would enable the C-domain of these colicins to detach before its entry into the cell. Such a proteolytic event has been identified for colD (de Zamaroczy et al., 2001; de Zamaroczy and Buckingham, 2002), colE2 (Sharma and Cramer, 2007) and colE7 (Shi et al., 2005). Preliminary data suggest that a similar proteolysis occurs in colE3 (see below).

Sequence alignment of the R-/C-domains linker region of colE2 and colE3 with that of colE7 revealed a conserved basic residue Lys450 in colE3 that could functionally replace the proteolytic site Arg447 of colE7 (Shi et al., 2005) and Arg452 of colE2 (Sharma et al., 2007). However, when Lys450 of colE3 was mutated in a triple alanine mutant, E3 K450A/N451A/K452A, there was only a small effect on its cytotoxicity (Fig. 6A). To identify the putative cleavage site in colE3, basic residues in and around the linker region were changed and the relative cytotoxicity of these mutants measured through their effect on the growth of colicin sensitive cells (Fig. 6A and B). Although the triple mutants E3 K450A/N451A/K452A (E3450–452) and E3 P453A/R454A/K455A (E3453–455) were as active as the wild-type colicin, the hexa mutant E3 K450A/N451A/K452A/P453A/R454A/K455A (E3450–455) showed a significant reduction in cytotoxicity (Fig. 6A). To further analyse the hexa mutant, residues ‘Lys-Arg-Asn’, containing the proteolytic site of colE2 (Sharma et al., 2007) and colE7 (Shi et al., 2005), were introduced into E3450–455 in two formats. In the first, ‘Lys-Arg-Asn’ was introduced within the hexa mutation at site 451–453 (E3450–455/451–453E7), and in the other at site 456–458 (E3450–455/456–458E7), i.e. after the hexa mutation. Cytotoxicity was restored if ‘Lys-Arg-Asn’ was introduced within the hexa mutation, but not following it (Fig. 6B). Thus, the introduction of the colE7 proteolytic site restores cytotoxicity in a site-dependent manner, implying that the reduced activity of the E3450–455 mutant is due to the loss of a putative proteolytic site contained within the hexa mutation.

Figure 6.

Cytotoxicity of colicin E3 constructs, with mutations in the R-/C-domain linker region. The cytotoxic activities were compared by analysing the effect of the E3 constructs on growth of colicin-sensitive cells.
A. All of the mutants tested, except the hexa mutant E3450–455, had wild-type activity. Although mutating the residues 450–452 and 453–455 to alanine did not have any effect on colicin activity, the combination of the two mutations to generate the construct E3450–455 reduced colicin activity by a factor of ∼30, as assayed by spot titer (data not shown).
B. When the proteolytic site of colicin E7, residues Lys446–Arg447–Asn448, was mutated into the hexa mutant, the cytotoxicity was restored in a site-dependent manner. Cytotoxicity was restored only if the residues were placed within the hexa mutation (E3450–455/451–453E7) and not after it (E3450–455/456–458E7). Definition of colicin mutants: E3432–435, colicin E3 with mutations R432A/K433A/K434A/K435A; E3436–440, colicin E3 with mutations E436A/D437A/K438A/K439A/R440A; E3447–449, colicin E3 with mutations N447A/D448A/E449A; E3450–455, colicin E3 with mutations K450A/N451A/K452A/P453A/R454A/K455A; E3456–458, colicin E3 with mutations G456A/F457A/K458A; E3450–455/451–453E7, colicin E3 with mutations K450A/N451K/K452R/P453N/R454A/K455A; E3450–455/456–458E7, colicin E3 with mutations K450A/N451A/K452A/P453A/R454A/K455A/G456K/F457R/K458N.

However, the gene product responsible for such a proteolytic event could not be identified. OmpT, an outer membrane protease that can cleave the C-domains of colicin E2 (Duche et al., 2009) and E3 (de M. Zamaroczy, pers. comm.) was found in the present screening not to be required for the activity of any of the eight colicins tested. Only two possible scenarios can be envisaged under which we would not be able to identify the protease during the screening of the Keio collection: (i) the protein catalysing the proteolytic activity is essential for bacterial survival in which case it is not represented in the Keio collection, for which a precedent is provided by the lepB gene that cleaves colicin D (see section V below); (ii) more than one protein can catalyse proteolysis of these eight colicins.

V. Colicins B and D

Primary screening identified the genes fepA, tonB, exbB and exbD to be important (Table 1) for the activity of colB and colD (data not shown). lepB, which encodes for the inner membrane protease that has been shown to cleave colD (de Zamaroczy et al., 2001), was not represented in the Keio collection because it is an essential gene. None of the other 3981 knockout strains tested showed colicin resistance, implying that barring any redundancy of gene function, none of these genes were important for the activity of colicins B and D.

VI. Consequences of new genes for import pathways of colicins

The present screening of the E. coli knockout library has enabled the identification of several new genes involved in colicin cytotoxicity. However, TolQ, which was inferred to be required for the import of colE1, was found to be nonessential. These findings impact current models of colicin import pathways.

(i) Translocon for import of nuclease E colicins; no requirement for peptidoglycan-associated lipoprotein.  Colicins or their cytotoxic domains utilize protein networks (translocons) to cross the cell envelope and enter the cell to exert cytotoxicity. A model for the outer membrane translocon has been proposed for the group A nuclease E colicins (Kurisu et al., 2003; Housden et al., 2005; Sharma et al., 2007). BtuB and OmpF are the outer membrane components of this translocon, while the members of the Tol network, TolQRAB, form the periplasmic and the inner membrane components. The Tol network is involved in maintaining the integrity of the outer membrane so that Tol mutants leak periplasmic proteins and are hypersensitive to detergent and drugs (Lazzaroni and Portalier, 1992; Lazzaroni et al., 1999; Llamas et al., 2000; Cascales et al., 2002). One of the defined components of the Tol network is the peptidoglycan-associated lipoprotein (Pal) (Lazzaroni et al., 2002; Cascales et al., 2007). Interactions of Pal with TolA and TolB have been established by in vivo cross-linking (Bouveret et al., 1995; Cascales et al., 2000). The same region of Pal is involved in binding to TolB and peptidoglycan (Clavel et al., 1998; Ray et al., 2000) so that the interactions of Pal with TolB and peptidoglycan have been indicated to be mutually exclusive (Bouveret et al., 1999). Further, the binding sites of Pal and the colE9 T-domain on TolB were found to overlap to a significant degree, and disruption of the TolB–Pal interaction destabilized the integrity of the outer membrane (Loftus et al., 2006). However, despite these significant interactions of Pal with TolA and TolB and its function in outer membrane stability (Lazzaroni et al., 1999; Cascales et al., 2002), Pal was found in the present studies to not be essential for the cytotoxic activity of all the group A colicins tested. These results are in agreement with two previous reports that implied that Pal is not required for cytotoxicity of colicins A, E1, E2 and E3 (Fognini-Lefebvre et al., 1987; Clavel et al., 1998). However, this result contrasts with the suggestion that Pal is part of the import apparatus for these colicins (James et al., 2002; Bonsor et al., 2008) and a recent report from ‘spot titer’ data, where a partial requirement for Pal was implied for the activity of colicin S4, a Tol network-dependent colicin (Arnold et al., 2009).

(ii) BtuB–OmpF interactions; requirement for mobility of BtuB in the outer membrane.  BtuB, which has a high binding affinity (Kd < 10−9 M) for the nuclease E colicins (Kurisu et al., 2003; Sharma et al., 2007), serves as the primary receptor that captures the colicin from the extracellular medium to concentrate it on the cellular surface where it can diffuse and interact with the secondary receptor/translocator, OmpF. Consistent with their function as the primary and secondary receptor for colicin translocation, the outer membrane of each cell contains ∼200–400 copies of BtuB, a density two to three orders of magnitude smaller than that of OmpF, which is present at a level of ∼105 molecules/cell (Nikaido and Vaara, 1987). BtuB with bound colicin E3 has been shown in single-molecule studies to have a high level of lateral mobility in the outer membrane (J. Spector et al., in preparation). The diffusion constant of OmpF was also determined, and found to be much smaller. The present studies suggest that diffusion of one or more outer membrane receptors has an important role in the function of the colicin translocon. Based on the crystal structures of the R-domains of colicin E2 and E3 complexed to BtuB (Kurisu et al., 2003; Sharma et al., 2007), which show the 100 Å coiled-coil receptor binding domain fishing for OmpF, there is at least one OmpF molecule within the mobile range of a colicin bound to its BtuB receptor. In the case of the ‘deep rough phenotype’ LPS mutants, there is a decrease in the level of OmpF in the outer membrane to 10% of that in the wild type (Schnaitman and Klena, 1993), which translates to less than one OmpF molecule within the range of each colicin bound to BtuB. However, these cells are efficiently killed by colA (Fig. 3), suggesting that BtuB is able to diffuse in the outer membrane sufficiently rapidly to form a functional complex with OmpF.

(iii) Subsequent steps in import through the outer membrane.  Passage of the colicin N-terminus through the OmpF pore (Kurisu et al., 2003; Zakharov et al., 2004; Sharma et al., 2007; Yamashita et al., 2008) requires interact with the TolB protein in the periplasm (Bouveret et al., 1997; Garinot-Schneider et al., 1997; Loftus et al., 2006), followed by release of the immunity protein (Duche et al., 2006). Subsequently, the C-domain interacts with the OmpF porin (Zakharov et al., 2006), is proteolysed by the putative periplasmic protease (Shi et al., 2005; Sharma et al., 2007) and enters the cytoplasm where it exerts its cytotoxic effect.

(iv) Colicin N, OmpF and LPS.  Interaction of colN with LPS has been implied in the recent 25 Å structure obtained by electron microscopy of a complex of OmpF and colN, which found colN bound to the outer surface of OmpF (Baboolal et al., 2008). The outer surface of OmpF has been suggested to contain the binding site for LPS (Ferguson et al., 2000; Vandeputte-Rutten et al., 2001; Baboolal et al., 2008), and the binding of colN in this region displaced the LPS. It has been suggested that the import of colN might take place at the OmpF–LPS interface and an interaction between OmpF and the N-terminal helix of the colN pore forming domain has been demonstrated (Baboolal et al., 2008). However, to date there has been no direct evidence of an essential function of LPS in colN cytotoxicity. The present study provides the first conclusive evidence for LPS function in colN cytotoxicity and indicates an interaction of colN with LPS. It is possible that the low affinity of colN for OmpF, indicating differences in the colN translocation mechanism, might merely be the result of insufficient LPS for a colN footprint in the above affinity studies. It is noted that interaction with LPS has been previously found to be important for RTX toxins ApxI and ApxII (Ramjeet et al., 2005), and galU has been found to be important for the uptake of these two toxins and affects their interaction with LPS (Ramjeet et al., 2008). The above results also correlate well with the fact that the OmpF-specific bacteriophage, K20, requires both OmpF and LPS for its receptor function (Silverman and Benson, 1987). However, K20 but not colN requires WaaB, which functions to add another glucose moiety to LPS inner core after RfaG (Traurig and Misra, 1999), for its activity.

(v) Colicin A, OmpF and TolC.  Deletion of tolC resulted in complete cellular resistance to colA (Figs 3C and 4), in contrast to the small effect on colA activity found in previous studies (Benedetti et al., 1991). This effect of tolC deletion on colA cytotoxicity was attributed to tolC-mediated control of OmpF (Misra and Reeves, 1987). However, colN was able to kill ΔtolC cells (Fig. 3C), implying that it has a higher affinity for OmpF than colA, which enables colN to bind OmpF even when the outer membrane has a significantly lower OmpF population. This observation is consistent with the conclusion that colA uses OmpF as a secondary translocator while colN uses OmpF as the primary receptor. Because tolC deletion decreases OmpF, but increases OmpC, levels in the outer membrane, ΔtolC cells are sensitive to colE2, E3 and E7 that can utilize either OmpF or OmpC as the secondary receptor (Sharma et al., 2007).


The Keio collection and the ASKA library were used to provide a more complete description of the cellular network that supports colicin cytotoxicity. The major findings of the present study are: (i) ColN is unique in that it requires LPS inner core biosynthesis genes for its activity. The LPS inner core along with the first added glucose forms the minimum binding site for colN during its import into the cell. (ii) The gene gmhB is required for colA cytotoxicity. Several additional genes affecting colA cytotoxicity exert their effect by decreasing OmpF levels in the outer membrane. (iii) ColE1 absolutely requires tolA and tolR, but not tolQ, whose deletion generates only partial resistance. (iv) Nuclease E colicins E2, E3 and E7 do not require the Pal for their activity. (v) The 3900+ knockout strains that were killed by the colicins tested in the present study imply that, barring any redundancy in the gene function, these genes are not required for colicin cytotoxicity.

Experimental procedures

Bacterial strains

The ‘Keio collection’, a systematic single-gene knockout library of all nonessential genes in E. coli, and the ‘ASKA’ orfeome library, which contains each E. coli gene cloned into a plasmid vector, allow rapid screening and detection of genes involved in colicin import and cytotoxicity. The Keio collection was constructed in E. coli K-12 BW25113 [rrnB3ΔlacZ4787 hsdR514Δ(araBAD)567Δ(rhaBAD)568] (Datsenko and Wanner, 2000; Baba et al., 2006) and the ASKA collection in E. coli K-12 AG1 (Stratagene, La Jolla, CA), which is a derivative of DH1: recA1 endA1 gyrA96 thi-1 hsdR17 glnv44(supE44) relA1 (Kitagawa et al., 2005).

The E. coli XL1 Blue strain was used as the host strain for cloning of mutations and deletions. Cloning was done in the pET41b vector such that there was a His8 tag at the C terminus of the protein. E. coli BL21(DE3) was the host strain for expressing the protein. In the pET41b vector, protein expression is under the control of a strong IPTG-inducible T7 RNA polymerase promoter. All cultures were grown in Luria–Bertani (LB) media or on LB agar plates supplemented with antibiotic when required.


Plasmids for colicins B, D and E7 were obtained from K. Postle, M. de Zamaroczy and H. S. Yuan respectively. Colicind E2 and E7 were cloned between the NdeI/XhoI sites of the expression vector pET41b using standard protocols mentioned previously (Zakharov et al., 2004; Sharma and Cramer, 2007). C-terminal His8-tagged colicins E2, E3 and E7 were purified by metal affinity chromatography using an Ni-charged iminodiacetic acid-agarose column.

Colicins A (Cavard and Lazdunski, 1979), B (Hilsenbeck et al., 2004), D (de Zamaroczy et al., 2001), E1 (Lindeberg et al., 2000) and N (Cavard and Lazdunski, 1979; Izard et al., 1994) were purified using published protocols. BtuB (Kurisu et al., 2003) and OmpF (Yamashita et al., 2008) were purified as mentioned elsewhere. Colicin E1 was cloned in pT 7–7 vector and E. coli BL21(DE3) was the host strain for expressing the E1 protein. Purity of colicins was visually ascertained by SDS-PAGE. The concentration of the proteins was measured using extinction coefficients determined by the Protparam tool (Gasteiger et al., 2005).

Mutagenesis and protein purification

All colicin E3 cloning and mutagenesis were done as previously described (Sharma and Cramer, 2007). Purification of overexpressed colicin E3 constructs was carried out with a Ni-charged iminodiacetic acid-agarose column.

Screening procedure

The E. coli knockout mutants were applied as drops on LB agar plates containing colicins and incubated at 30°C overnight. Colicin-sensitive cells were killed by the colicin, thus allowing only the resistant cells to grow. As a control to check for cell growth, cells were also applied to plates from which colicin was absent. A total of 3985 non-lethal single-gene knockout strains in the library were screened for colicins A, B, D, E1, E2, E3, E7 and N. Because of the high density of cells applied to the plates, the concentration of colicins incorporated into the plates was high: 10 nM for colicins B, D and E1; 15 nM for colicins A and N; 20 nM for colicins E2, E3 and E7. The strains showing resistance or tolerance to colicin were then purified and further analysed for resistance by streaking single colonies of the colicin-resistant strains, identified in the primary screening, on colicin-containing LB plates that were incubated at 37°C. Subsequently, colicin-resistant strains were verified by PCR. In some cases, independent verification was also performed by targeted screening of a particular strain from the Keio collection. Because of the high colicin concentration used and redundancy in the gene functions, it is noted that the present search could not have identified all the gene products involved in the cytotoxicity of the colicins tested.


The ASKA library is a collection of each of the deleted genes cloned in a plasmid such that their expression is under the control of IPTG. Each ASKA clone has been named as pCA24N::geneX, where geneX stands for the gene that is cloned in the vector pCA24N, e.g. pCA24N::tolA, for the tolA ASKA plasmid that was used to complement the ΔtolA Keio strain. To verify the results obtained by screening the Keio collection, the strains identified to be colicin-resistant were tested for complementation with plasmids from the ASKA library containing the corresponding gene. As a control, the strains were also transformed with the empty vector. These transformants were then tested for sensitivity to the set of colicins. Successful complementation should restore sensitivity to a colicin. As a positive control, strains with genes known from previous studies to be necessary for colicin activity deleted: ΔtolQ, ΔtolR and ΔtolA for group A colicins A, E1, E2, E3, E7 and N; ΔbtuB for colicins A, E1, E2, E3 and E7; and ΔfepA for colicins B and D, were complemented with the respective genes and tested for restoration of colicin sensitivity. Complementation was achieved in the presence of 0.5 mM IPTG for all the mutants except the Tol knockouts. In the case of the Tol mutants complementation was achieved in the absence of any IPTG.

Cytotoxicity assays

The time-course for growth of the colicin-resistant strains carrying the corresponding ASKA genes (or the vector control) was analysed in the presence and absence of colicin. Ten microlitres from an overnight culture was added to 190 μl of LB medium in 96-well plates supplemented with chloramphenicol (25 μg ml−1), IPTG (0.5 mM) and colicin A (5 nm) or colicin N (500 pm). The cells were allowed to grow at 37°C and the growth was monitored by measuring optical density at 410 nm at 30 min intervals.

The cytotoxicity of the different colicin E3 mutant constructs was compared by analysing the effect of colicin on the growth rate of colicin-sensitive E. coli K17 indicator cells. One hundred millilitres of LB media was inoculated with an overnight culture of the indicator cells and grown to an OD600 (optical density at 600 nm) of 0.1. The culture was then aliquoted, and a different colicin construct was added to each sample at a concentration of 200 pm. Cell growth was monitored by measuring the optical density at 600 nm at 1 h intervals.

Cytotoxicity of colicins was assayed qualitatively using a ‘spot titer’, in which a lawn (∼2 × 108 cells) of colicin-sensitive E. coli K17 indicator cells in log growth phase were spread on a LB plate to which different concentrations of colicin are applied as 20 μl drops. After overnight incubation of the plates at 37°C, killing of the bacterial cells is seen as clear spots in a lawn of bacteria. The lowest inhibitory concentration was defined as the smallest concentration that would generate a clear zone of inhibition.

Colicin affinity for receptor; neutralization by BtuB and OmpF

The receptor binding affinity of colicin was assayed by the ability of BtuB (OmpF for colN) to neutralize the cytotoxicity of the colicin assayed by the microbiological spot titer. BtuB (OmpF for colN) in detergent solution was mixed with the colicin at different molar ratios before addition of a 20 μl aliquot to the Petri plate, on which a lawn of colicin-sensitive E. coli K17 indicator cells in log growth phase had been spread. The minimum molar ratio of BtuB (OmpF for colN) to colicin that prevented colicin cytotoxic activity was used as an indicator of colicin affinity for the receptor. The concentrations of colicins A, E1, E2, E3, E7 and N, respectively, used for the neutralization experiment were 5, 0.5, 0.5, 0.5, 0.5 and 1 nm. At these concentrations, colicins, diluted in detergent-containing buffers, were able to generate a clear zone of inhibition in the absence of BtuB.


These studies were supported by a grant from the NIH-GM18457 and the Henry Koffler Professorship (W.A.C.), and NIH-GM083296 (B.L.W.). We thank L.N. Csonka for suggesting the knockout library experiment and S.D. Zakharov for helpful discussions.