Current problems in the understanding of colicin import across the Escherichia coli outer membrane (OM), involving a range of cytotoxic mechanisms, are discussed: (I) Crystal structure analysis of colicin E3 (RNAase) with bound OM vitamin B12 receptor, BtuB, and of the N-terminal translocation (T) domain of E3 and E9 (DNAase) inserted into the OM OmpF porin, provide details of the initial interaction of the colicin central receptor (R)- and N-terminal T-domain with OM receptors/translocators. (II) Features of the translocon include: (a) high-affinity (Kd ≈ 10−9 M) binding of the E3 receptor-binding R-domain E3 to BtuB; (b) insertion of disordered colicin N-terminal domain into the OmpF trimer; (c) binding of the N-terminus, documented for colicin E9, to the TolB protein on the periplasmic side of OmpF. Reinsertion of the colicin N-terminus into the second of the three pores in OmpF implies a colicin anchor site on the periplasmic side of OmpF. (III) Studies on the insertion of nuclease colicins into the cytoplasmic compartment imply that translocation proceeds via the C-terminal catalytic domain, proposed here to insert through the unoccupied third pore of the OmpF trimer, consistent with in vitro occlusion of OmpF channels by the isolated E3 C-terminal domain. (IV) Discussion of channel-forming colicins focuses mainly on colicin E1 for which BtuB is receptor and the OM TolC protein the proposed translocator. The ability of TolC, part of a multidrug efflux pump, for which there is no precedent for an import function, to provide a trans-periplasmic import pathway for colicin E1, is questioned on the basis of an unfavorable hairpin conformation of colicin N-terminal peptides inserted into TolC.
A summary of the present state of understanding of the molecular events associated with colicin translocation across the cell membrane envelope is important not only for its contribution to fundamental mechanisms of transmembrane cytotoxin import. An understanding of the mechanisms of membrane interaction and cytotoxicity of colicins is also relevant to the translocation of other macromolecules and, as well, strategies that confront antibiotic-resistant disease-causing bacteria [1,2].
The membrane envelope of Gram-negative bacteria consists of two membranes separated by a periplasmic space within which an additional barrier for protein diffusion is formed by the peptidoglycan layer. The outer membrane (OM) consists of an electrically charged inner and outer leaflet formed by phospholipid (PLp) and the anionic lipopolysaccharide.
Mechanisms of colicin action and structure–function have been reviewed in recent years [3–11]. The present discussion is limited to details of the mechanism of translocation across the OM of a subset of the ‘group A’ colicins, i.e., colicins A, E1–E9, and N. These colicins utilize the inner membrane (IM) protein complex TolQRA, with an extended trans-periplasmic domain that functions in colicin translocation across the periplasm and the inner cytoplasmic membrane, IM , discussed previously in a review  in which one of the authors of this article participated. The present review/discussion initially considers mechanistic details of colicin translocation of nuclease colicins across the OM, based on the interactions of (i) the N-terminal domain of the nuclease colicin E9 [12,13], and (ii) the C-terminal catalytic domain of the RNAase colicin E3  with the OmpF porin receptor. These considerations lead to inferences concerning the pathway of nuclease colicin import across the OM. The interaction of colicin E1 with the OM- and periplasm-spanning protein TolC  and its consequences for the OM import pathway/mechanism of E1 are discussed.
Briefly summarizing pathways for cellular import of these colicins, for which a crystal structure of the entire colicin [15–17], or of sub-colicin functional domains [18,19], has been obtained (Figure 1A–E displays ribbon diagrams based on the crystal structures of colicins: E3  [1JCH]; N , without disordered NH2-terminal 90 residues [1A87]; Ia  [PDB: 1CII]; and the carboxy-terminal channel-forming domain of colicin E1  [PDB: 2I88], and colicin A  [PDB: 1COL]). At least three to four stages in the translocation of colicins across the cell envelope can be defined: (I) binding to the OM of the Gram-negative Escherichia coli; (II) formation of an OM translocon through recruitment and utilization of OM proteins which translocate the colicin including the C-terminal domain responsible for cytotoxicity (coded green, Figure 1A); (III) transfer across the periplasmic space, and binding to, or insertion into, the cytoplasmic membrane (IM). A consequence of insertion into the IM is manifestation of the cytotoxic effect of the C-terminal domain of the channel-forming colicins A , E1 [20–23], Ia [24,25], or N [8,16,26–28], which depolarize the IM with inhibitory and lethal consequences for cellular energetics and viability. The consequences of IM depolarization can be measured through cell loss of potassium to the extracellular milieu , decrease in intracellular ATP levels , or inhibition of active transport . (IV) Colicins with degradative enzymatic action are translocated into the cytoplasm where, among the lethal activities imposed by the C-terminal cytotoxic domain, are: (a) hydrolysis of chromosomal DNA  and concomitant inhibition of DNA synthesis (colicins E2, E7, and E9) and (b) inhibition of protein synthesis through (i) RNase-dependent degradation of the ribosomal subunits [33,34] or (ii) tRNA by colicin D  or E5 . A fourth stage in which protein complexes embedded in the IM [3,37–39] can influence colicin translocation through the OM is not discussed here in detail.
Colicin domain and crystal structures.
The OM translocation quandary
It is proposed that the pathway and mechanisms by which colicins are translocated across the OM, summarized in the reviews on mechanisms of colicin cytotoxicity and import cited above (Introduction), and suggested for colicin E9 in recent studies on a ‘total thread’ model , deserve further discussion. Import of group A colicins across the OM involves receptor/translocator proteins, which include the universal porin, OmpF [40–42], BtuB [43–50], TolC [42–48,50–59], and lipopolysaccharide (LPS), the role of the latter discussed mainly for colicin N [26,28,47,60–62]. For colicins such as E3 and E9, which via their N-terminus, insert into OmpF, a translocation mechanism through the OM, which follows initial binding to the vitamin B12 translocator, BtuB, is proposed based on the observation that the C-terminal domain of colicin E3 also inserts into OmpF .
Functions of OmpF and BtuB in colicin passage across the OM
Cellular reception of the group A colicin E3 requires the OM protein BtuB, the translocator for vitamin B12 , and receptor for bacteriophage BF23 . BtuB is a 22 stranded β-barrel protein whose structure determined by X-ray diffraction is known to a resolution of 1.95 Å , and which is the primary high-affinity receptor [43,44,50,64–66] for the nuclease colicins E3 and E9. Colicin E3 binds to BtuB through the central apical region of its ‘R-domain’ (colored in red, Figure 1A,C) with a dissociation constant Kd ≈ 10−9 M [44,45], and E9 binds with a similar affinity . The binding affinity of BtuB for E1 has been reported in some [5,67], but not all  studies, to be smaller, thereby requiring a larger concentration, by 1–2 orders of magnitude, to achieve the same cytotoxicity.
Interaction between BtuB-bound colicin and OmpF: the ‘fishing pole’ model
The rate of OM diffusion of a fluorophore-labeled complex of colicin E3 and BtuB has been measured by single-cell fluorescence . The two-dimensional diffusion constant, D2d, for OM diffusion of the BtuB–colicin E3 complex  is 5–10 × 10−10 cm2 s−1, over a time scale of 25–150 ms, the diffusion spatially restricted by the dense OM population of OmpF , present at 104–105 copies/cell . A colicin translocon ‘fishing pole’ model [41,44,71] for the colicin-mediated interaction between BtuB and OmpF was proposed for the initial OM colicin–receptor interactions that precede colicin passage through the OM. Thus, the disordered N-terminal domain of the colicin ‘fishes’ for an aperture or an anchoring site in the pores of the homo-trimeric 16-stranded antiparallel β-barrel OmpF porin [42,72,73]. Successful ‘fishing’ initiates translocation across the OM.
Structure-based data on the translocation mechanism
The mechanism of colicin translocation across the OM received further focus from crystal structures obtained for the disordered N-terminal domain of colicin E3  and E9  inserted into the OmpF porin: (a) using co-crystals with OmpF, an N-terminal 83 residue disordered peptide of the RNase colicin E3 was demonstrated (resolution without/with inserted T83, 1.6 Å/3.0 Å) to occlude the ion channel function of OmpF ; this N-terminal peptide of colicin E3 was seen to penetrate and span the pore of the trimeric OmpF porin ; (b) a similar, but more detailed result was obtained with the disordered N-terminal peptide of the DNAase colicin E9, for which two binding sites/domains, OSB1 and OSB2, of the 83 residue N-terminal peptide of E9 were found to occupy or occlude two of the three pores of the tri-pore OmpF porin [12,13], characterized by a free energy of binding, ΔG ≈ −8 kcal/mol .
The ‘fishing pole’ mechanism of colicin translocation across the OM and the details of utilization of OmpF were inferred from the analysis by electrospray ionization mass spectrometry of an isolated translocon complex of colicin E9 which contained the bound TolB protein . The colicin was found to be anchored on the periplasmic side of the OM by attachment to TolB and insertion of the colicin N-terminus from the periplasmic side of the OM back into one of the two remaining open pores of the trimeric OmpF (Figure 2). The detail of the N-terminus folded back into the second pore of OmpF distinguishes this ‘total thread’ model from a similar ‘thread’ model . The question is whether the OmpF-inserted state of the N-terminal domains of colicins E3 and E9 is an initial state in the process of total threading of the entire colicin molecule, as inferred in refs [12,41,74], or the structure of the N-terminal anchor which is fixed by its insertion into the first pore of the trimeric OmpF. Its binding to TolB in the periplasm, and its reinsertion into the second pore, which locks the anchor in place, allows the C-domain which carries the cytotoxicity information to be translocated through one of the other two pores of the OmpF trimeric pore structure, as proposed  and described in Figure 3.
Import of nuclease colicins (e.g., E3, E9): schematic of established steps.
Model for translocation of nuclease colicins across cell envelope.
Energy barriers in colicin translocation
At least two defined energy barriers exist in the model proposed in Figure 3: (a) a major energetic barrier involves release of the tightly bound (Kd ≈ 10−14 M ) immunity protein (shown for colicin E3 in yellow, Figure 1C), which protects colicinogenic cells from cytotoxic action from the colicin that they produce [76,77]. While this problem will not be discussed further, it is noted that: (i) the Imm protein (immunity protein) has been inferred to be released in vivo via the transmembrane electrochemical potential gradient, , which exists across the cytoplasmic membrane (Figures 2 and 3), and which would supply the needed free energy. The energy would be transduced from the IM via a trans-periplasmic conformation of the TolA protein  and (ii) it has been possible in vitro to demonstrate release of the immunity protein (pI ≈ 4) from colicin E3 via chromatography on an anion exchange column at alkaline pH .
(b) Binding to BtuB receptor: A formidable energy barrier to the ‘total thread’ model is conferred by the required dissociation of the colicin, receptor-binding domain (coded ‘red’ in Figure 1), from its primary receptor, the vitamin B12 translocator, BtuB, to which the apex of the coiled-coil receptor-binding domain is tightly bound, with Kd ≈ 10−9 M for colicin E3  and E9 . No mechanism has been proposed for colicin release from this tight attachment of its R-domain. This bound state also precludes a simple mechanistic analogy with OM ‘auto-transporter’ pathways . Although the latter translocation pathways have been described for large polypeptides, these pathways are not known to be impeded by states or positions of tight binding of the polypeptide in the translocation pathway. The problem of necessary release or dissolution (by a protease?) is a significant aspect of the ‘quandary’ for colicin translocation through the OM.
Consequences of studies on the ‘total thread’ model
To briefly summarize the consequences of the important contribution of the binding studies of colicin E9 to OmpF and to TolB in the periplasm [12,13], they lead to the inference that the threaded structure of the N-terminal domain of the nuclease colicins in OmpF defines an anchor in OmpF and the OM, and lead to the question of the pathway for translocation of the catalytic C-domain.
Translocation of the catalytic C-domain
A pathway using the OmpF-E3 system , described in Figure 3, is proposed for passage of the C-terminal- active domain (green domain, Figure 1) across the OM. This pathway utilizes the R-domain to initially bind the colicin to the OM surface, inserts the colicin N-terminal domain into the first pore of the trimeric OmpF to initially anchor the translocation apparatus. The anchor is then secured by the binding to TolB and reinsertion into the second pore of the OmpF. It is proposed that a major auxiliary function of this ‘anchor structure’ is to bring the catalytic C-terminal domain of the colicin into proximity to the third pore of OmpF into which it can insert as documented by its ability to occlude OmpF channels in vitro . It is thus proposed that the activity-containing C-domain of colicins E2/E3/E7/E9 can lead its own translocation through the OmpF porin  (Figure 3), consistent with the results in refs [80–82]. The latter studies, which provide evidence in support of the alternative mechanism described above for OM translocation of a nuclease colicin, are: (i) colicin E2 can be bound to the BtuB receptor and the Tol import machinery, while the nuclease domain enters the cytoplasmic compartment of the target cell  where (ii) the IM protease FtsH (energy-dependent protease combining ATP-dependent unfolding & protein degradation activities) cleaves the cytotoxic domain of colicins D, E2, E3, and E7, which are then freed to enter the cytoplasm [80,82] .
Questions about the alternative ‘C-domain model’ for translocation across the OM
There are problems with the alternative model, as previously noted : a conceptual problem is that there are no known recognition sites or ‘boxes’ in the C-terminal domain comparable with the OmpF, TolB, or TolA-binding sites in the translocation (T) domain. A response to this reservation is that the existence of signaling mechanism(s) or ‘Tol boxes’ for directing the active C-domain to the cytoplasmic membrane and cytoplasm, and their mechanism is, to our knowledge, thus far uninvestigated.
Channel-forming colicins; Ia, N, and E1
The C-domain of the pore-forming colicins consists of nine α-helices which form a tightly packed globule in a hydrophilic environment (helix content ∼70%) . Considering the membrane environment in which these colicins form transmembrane ion channels, a helical conformation is thermodynamically most favorable for a polypeptide in a hydrophobic lipid environment . Differential calorimetry analysis of colicin E1 showed that the C-domain is less stable in the intact colicin and more stable when it loses its interactions with T- and R-domains .
As the one of two pore-forming colicins whose crystal structure has been solved  (Figure 1B), the description of the translocation mechanism of the 69 kDa TonB-dependent channel-forming group B colicin Ia is of major interest. Its structure also provides an additional model for intact colicin E1 , for which an atomic structure has been obtained only for the C-terminal channel domain (Figure 1F) . A unique feature of the translocation network for colicin Ia is that, instead of the OM BtuB–OmpF receptor–translocator pair, colicin Ia utilizes two copies of the Cir (Colicin I Receptor) . Following binding of colicin Ia to Cir, for which a crystal structure has been obtained , another Cir protein is recruited and inferred to be the OM translocator, thus creating a two receptor pair analogous to OmpF–TolC. The proposed mechanism of insertion and interaction with the OM receptor, Cir, ascribes the translocator function to a second copy of Cir . A previous review, which included a discussion of the pathway and mechanism of colicin Ia translocation through Cir, without using this particular terminology, assumed a ‘total thread’ model for translocation of colicin Ia through Cir .
Another study on the mechanism of formation and action of the ion channel function of colicin Ia, which is of interest in the discussion below of OM translocation of the colicin E1 C-terminal domain, is the inferred involvement of a conformational ‘flip’ in the membrane translocation of helices of the colicin Ia C-terminal channel-forming domain .
In the context of the energetic expense in the ‘total thread’ model for translocation of colicins E3 and E1 (section 3 above) arising from its tight (Kd = 10−9 M) binding to BtuB, the affinity of colicin N for OmpF, with a dissociation constant Kd = 2 × 10−6 M−1 , is somewhat smaller, although still indicative of a high affinity. The disordered amino-terminus of the colicin N T-domain is involved in translocon formation through binding of an N-terminal 27-residue epitope with the C-terminal domain of TolA [89,90].
OM binding translocation: a role of LPS
A role of LPS in the mechanism of colicin uptake in the OM, discussed below, is based significantly on studies of the mechanism for colicin N [47,60,61,91–93], the smallest colicin in group A (Figure 1D), which has a unique structure and mode of intracellular entry. In contrast with colicin E3 (Figure 1C), its structure is relatively compact: (i) a disordered N-terminal T-domain and (ii) the channel-forming C-terminal domain connected to the R-domain through a single helix. Colicin N has no known high-affinity receptor analogous to BtuB. An EM structure of colicin N together with OmpF shows colicin N to be bound at the periphery of the OmpF trimer where it has been inferred to interact with LPS [26,88].
Threading of the N-terminal unordered segment into the OmpF pore is supported by occlusion of OmpF channels in planar bilayers by an N-terminal segment of colicin N . In contrast with models for insertion of colicins E3 and E9 into a pore of the OmpF porin [41,74], the OmpF–lipid interface external to the OmpF pore has been proposed as an alternative route for translocation of colicin N [26,61].
Unfolding of the tertiary structure of the colicin E1 C-domain upon interactions with an anionic PLp membrane, or OM LPS, which results in transition to a two-dimensional helical array, proceeding without loss, and even an increase in α-helical structure , suggests a role of the negatively charged LPS in inducing conformational changes in the structure of the colicin channel domain. As for colicins E3 and E9, the OM protein BtuB is used for the receptor function that initiates the import of colicin E1. BtuB has an affinity for E1 similar to , or somewhat smaller  than that for, colicin E3. This energetic barrier, as discussed above for the nuclease colicins E3, poses a problem for models of the OM transit of colicin E1. In addition, colicin E1 uses the OM TolC protein which is well documented to function as a multidrug cellular efflux pump [95–97] (Figure 4A,B) as a receptor . The traditional mechanism for OM translocation of colicin E1 through the OM and the periplasmic space via the TolC drug export protein [42,56] has recently been reconsidered in the context of details of the binding interaction of colicin E1 N-terminal peptides to TolC :
Problems of energetics: Two well-defined energetic barriers are: (i) the initial high-affinity (nM) binding to the BtuB receptor , discussed above in the context of colicin E3 (Figure 2); (ii) there is no known energy source to unfold the globular C-domain of colicin E1 , which consists of nine closely packed helical segments (Figure 1A), to allow a conformation that can be translocated through any channel protein.
Problem of TolC structure and colicin E1 Translocation. In the set of OM receptors/translocators, the hybrid structure of TolC is unique: (i) it crosses the OM as a 12-stranded β-barrel and (ii) spans at least half the periplasmic space in a 12-stranded α-helical conformation, linked structurally to the IM–drug export AcrAB TolC  and MacAB–TolC  drug efflux complexes. TolC has been documented in vitro to have ion channel characteristics, with a small specific ion conductance, 80–90 pS in 1 M KCl [51,54]. TolC has been proposed to be a trans-periplasmic channel for passage of the entire colicin E1 molecule including the active C-terminal domain [5,55], although its small channel conductance is not consistent with a channel function for translocation of colicin E1 or any polypeptide of conventional cross-section. Crystal structure data show that the aperture of the helical channel at its terminus in the periplasm is at most 8.4 Å, ‘unlikely to be sufficient for even small molecules to pass unimpeded’ , and thus too small to allow passage of colicin E1, even in an undocumented hypothetical unfolded state; (iii) regarding the OM entry for a ‘total thread’ model, a steric problem at the entry for colicin E1 passage through TolC has been described . The structure of colicin E1 inserted into the TolC OM β-barrel domain, inferred from far-UV (ultraviolet) CD (circular dichroism) spectral and thermal stability analysis of T-domain peptides, is inferred to be a helical hairpin (Figure 4). Such a hairpin cannot be accommodated by the pore in the helical periplasmic domain of the TolC. Thus, while the OM aperture (28–37 Å) of the TolC can accommodate colicin entry, the passage through the periplasmic helical domain is unlikely. Thus, while an OM/trans-periplasm translocation function is obviously this time, the only suggestion that can be made about a translocon function of TolC is that it is similar to the of TolB/OmpF in translocon function for nuclease colicins.
Role of charged LPS: Because the cytotoxic channel-forming function of colicin E1 utilizes a folded multi-helix structure , there is no functional necessity to permanently unfold the nine-helix globular C-domain  which is, however, not geometrically amenable to translocation through an OM pore protein. The C-domain of colicin E1 is basic (pI = 9) and thus has a tendency to be positively charged, suggesting a basis for interaction with the anionic LPS in the outer half of the OM. The mechanism by which this interaction would facilitate translocation of the C-domain is not known. However, based on the precedent that an anionic membrane surface charge has been associated with unfolding of the colicin E1 channel domain into a two-dimensional membrane surface-bound array , it is suggested that the anionic surface change of the OM LPS layer can induce a structure change (e.g. ‘flattening’) in the E1 C-domain that would facilitate its passage through the OM. As suggested for OM translocation of coliicn N , and discussed below, it is suggested that the C-terminal channel-forming domain of colicin E1 can be translocated across the OM in a more ‘flattened’ conformation at a lipid–protein interface, or through the lipid phase adjacent to TolC using the amphipathic and cationic character of C-domain helices. The suggestion of an alternative ‘flattened’ structure associated with OM translocation is related to an inference derived from the fluorescence resonance energy transfer analysis of the colicin E1 C-domain binding to anionic membrane: the nine-helix globular C-terminal channel domain of colicin E1 (Figure 1F), in response to the local membrane surface charge, converts to a two-dimensional helical array anchored in the membrane by its hydrophobic hairpin .
It is noted that Figures 2–4, which depict the E. coli OM in the context of mechanisms of translocation or action of colicin, do not show the surface-bound LPS which forms a negatively charged electrostatic layer on the outer surface of the bacterium (Figure 3). The negative charge, whose magnitude depends on the extent of binding of exogenous divalent Mg2+ , originates from lipid A in the outer leaflet of the OM and from the LPS polyanionic core. Although this LPS web may serve as a defensive barrier to toxic compounds , in the context of colicin insertion, the surface anionic electrical potential might be suggested to be an energy source that contributes to the release of immunity protein from colicin E3 (Figure 2; ) and, as well, partial unfolding of the C-terminal channel domain of colicin E1 (Figure 4B). It is suggested that a role of LPS in electrostatically influenced conformational change and partial unfolding of the colicin E1 C-domain in the translocation mechanism extends the membrane structure framework in which future studies can be designed to describe the OM translocation pathway of colicin E1.
Cellular import of colicin E1.
It should also be considered that threading of the C-terminal domain of colicin E1 (loss of secondary structure) is not required for translocation across the OM, or for channel formation in the IM, because its helical structure is thermodynamically more favorable in a hydrophobic environment. The trans-OM passage of the C-domain of colicin E1 could also involve a ‘flip-flop’ mechanism with subsequent C-domain diffusion into periplasm, for which a precedent is the ‘flip-flop’ of amphipathic helices of the colicin Ia C-domain in the planar PLp membrane . Import of the colicin E1 C-domain from the OM to IM also requires a major role of the energy-transducing TolQRA complex in the IM , which is not discussed here.
Hypotheses for mechanisms of OM translocation of colicins offered in the present discussion are in a preliminary stage. Regarding nuclease colicins, recent studies on the interaction of colicin E9 with the OmpF receptor and the periplasmic TolB protein imply an anchor site for the colicin via its N-terminal translocation domain in which it utilizes two pores of the OmpF trimer; the ability of the colicin E3 C-terminal catalytic domain to also occlude OmpF channels suggests that OmpF can also serve as the translocation site for the nuclease activity domain. Regarding the IM channel-forming colicins, the present discussion mainly focused on colicin E1. Based on the known structure properties of the TolC multidrug export protein, which has a necessary but not satisfactorily described function in colicin E1 import, it is concluded that there is presently no structurally satisfactory model for OM import of colicin E1.
molar-free energy change
two-dimensional lateral diffusion constant
differential scanning calorimetry
energy-dependent protease combining ATP-dependent unfolding & protein degradation activities
molar dissociation constant
Omp, outer membrane, outer membrane protein
picoSiemen, unit of conductance
transmembrane proton electrochemical potential gradient.
The authors’ research that underlie this review were supported by NIH-GM018457 and the Henry Koffler Distinguished Professorship (W.A.C.) and recombinant DNA analysis by the Purdue Cancer Center grant (P30 CA023168).
We thank Profs W. Im, K. Jakes, and E. I. Zgurskaya, respectively, for discussions on anionic LPS as a source of electrostatic driving force for translocation of colicin E1, mechanisms of colicin import, and mechanistic details of bacterial multidrug efflux pumps. We thank R. Harding and G. Sincich for their technical support with the manuscript and illustrations.
The Authors declare that there are no competing interests associated with the manuscript.