Bacteriocins are narrow-spectrum protein antibiotics released to kill related bacteria of the same niche. Uptake of bacteriocins depends critically on the presence of an uptake receptor in the outer membrane, a translocation pore and an energy-dependent activating system of the inner membrane. Most bacteriocins act on the inner membrane as pore-forming toxins or they target cytoplasmic DNA/RNA and ribosomal synthesis respectively. Only two bacteriocins are known to become activated in the periplasmic space and to inhibit the renewal process of the peptidoglycan structure. In Escherichia coli, the Cma (colicin M) phosphatase is activated in the periplasmic space by the FkpA chaperone and subsequently degrades the C55-PP precursor unit of the peptidoglycan. Pst (pesticin) from Yersinia pestis carries a lysozyme homology domain to degrade peptidoglycan. Import of Pst is only achieved if the N-terminal translocation domain can span the outer membrane and if extensive unfolding of the protein during membrane passage is permitted. There is considerable plasticity in the import pathway since a chimaera comprising the activity domain replaced by T4 lysozyme is also translocated and active in killing those bacteria carrying the FyuA receptor.
Bacteria living in natural habitats compete against each other under conditions of stress and nutrient limitation by the secretion of bacteriocins. Production of these toxins (or antibiotics) in combination with the co-expression of protective immunity proteins can provide advantage for producing bacteria under ecological conditions . These weapons are both selective towards related species carrying the uptake receptor and highly efficient in their killing activity. The uptake mechanism underlies three subsequent steps: (i) recognition of the toxin by the receptor, (ii) translocation over the outer membrane, and (iii) activation in the destined compartment  (Supplementary Figure S1 at http://www.biochemsoctrans.org/bst/040/bst0401560add.htm).
Bacteriocins typically share the general architecture of three structurally independent domains: the N-terminal translocation domain (T), an intermediate receptor-binding domain (R), and the C-terminal activity domain (A) with antibiotic power . The modular genetic assembly is an architectural principle of these proteins and highlighted by the exchange and independent evolution of activity domains. This exchange has especially occurred within the class of colicins and yielded a repertoire of 11 pore-forming colicins. Although the activity domain is strongly conserved between bacterial species, translocation and receptor-binding domains are mostly unrelated and variable due to variability of the targeted receptor structure.
In Escherichia coli, more than 20 different colicins are known to target various structures in the cell. Cma (colicin M) is unique among these proteins as it targets the periplasmic peptidoglycan . Cma hijacks the FhuA iron receptor for initial contact and subsequent translocation over the outer membrane [3–7]. Energy for this process is provided by the Ton complex after direct interaction with the N-terminal TonB box of Cma and TonB has been established . In contrast with other colicins, folding of Cma depends on the presence of the periplasmic FkpA chaperone, which is essential for activity [7–10]. The toxin activity required to suppress cell growth is in the nanogram range (equivalent to approximately 50 molecules per cell) .
In Yersinia pestis and several pathogenic E. coli strains, only the pesticin (Pst) toxin has been identified as the only toxin and it was demonstrated to be active against cell division and stability through degradation of the peptidoglycan layer [11–13]. This activity was refined further in the characterization of this protein as muramidase, which was demonstrated to resemble T4 lysozyme activity whereby spheroplast formation in E. coli and Y. pestis cells is observed [14–19]. Cells are self-protected from this activity by co-expression of a small periplasmically located immunity protein Pim [1,20–22]. Uptake of Pst requires the TonB-dependent FyuA receptor and the products of exbB, exbD and tonB genes (Ton complex) in the target cell [11,23].
Evolution of the modular Pst and Cma architecture
Two toxins of the group of bacteriocins are currently known to target peptidoglycan: Cma acts on a precursor of peptidoglycan as phosphatase, whereas Pst shows clear activity as muramidase targeting the peptidoglycan network  (Supplementary Figure S1). The classical three-domain architecture has been preserved in both enzymes with their elongated and unstructured N-terminal translocation domain (CmaT and PstT), an intermediate receptor-binding do-main (CmaR and PstR) and the C-terminal activity domain (CmaA and PstA) (see the descriptive scheme of these domains in Figures 1B and 2A). Sequence analysis of the modular structures provides insights into the individual evolution of these domains as follows.
Structure and active-site representation of Cma
Structural and functional features of Pst
The N-terminal domain of both proteins covers approximately 40 residues with no sequence relationship to protein sequences in databases. The PstT sequence only occurs in proteins derived from Yersinia and typically pathogenic E. coli strains and shows a high degree of sequence identity (>95%). CmaT is also unique to E. coli genomes and one conserved relative in Salmonella enterica (UniProt entry H7P3E8). The intermediate domain PstR displays high identity with Pst derivatives from Yersinia and E. coli strains (>95%), but only weak homology with proteins from Vibrio, Yersinia and E. coli groups. By contrast, the CmaR domain sequence is unique and shows no relationship to proteins other than Cma homologues from E. coli and S. enterica. The PstA domain is highly conserved among proteins from Yersinia and E. coli according to translocation and receptor-binding domains. However, there are also representatives of this domain with an identity of 40% in a two-domain protein from the plant pathogen Pectobacterium carotovorum. In this bacterium, the PstA homology domain is fused to ferredoxin (UNIPARC UPI0001A445A9). In addition, there are several sequences from Pseudomonas, Desulvovibrio and Vibrio species with ~20% identity and ~45% similarity. Moreover, a large number of lysozyme-like proteins are related to PstA on the sequence level, most of which have a weak homology . The activity domain of Cma is observed in strains of E. coli, Pseudomonas, Burkholderia and Pectobacterium either as Cma-like three-domain proteins or as two-domain construct fused to ferredoxin such as observed in P. carotovorum and similar to the arrangement seen in Pst [3,24].
Together these data suggest a recent evolution of Pst with the PstA domain captured from related bacterial species and the PstT and PstR domains with no or negligible relationship to currently available protein sequences. Consequently, this domain resembles a novel domain fold within the group of bacteriocin domains. Owing to the stronger homology between PstA and bacterial sequences such as the VgrG protein cluster (Supplementary Figure S2 at http://www.biochemsoctrans.org/bst/040/bst0401560add.htm), it is likely that the protein has evolved from a bacterial precursor, which may have acquired a phage lysozyme homologue at an earlier point of evolution. The evolutionary ancestor of the Cma protein is even more difficult to trace. There is obvious sequence similarity only in the CmaA domain to proteins from several pathogens including plant pathogens, but no sequence identity was determined for the T and R domains. In summary, with the discovery of this peptidoglycan-targeting domain, the ‘pool’ of bacteriocin killing domains is expanded further.
Structure and function of Cma
Cma is an enzyme with unusual phosphatase activity cleaving the C55-PP precursor of peptidoglycan and LPS (lipopolysaccharide) into a hydrophobic membrane moiety and the hydrophilic PP-MG (pyrophosphate-MurNAc-peptapeptide/GlcNAc) portion [1,6] (Supplementary Figure S1). The toxin consists of three domains: an unstructured N-terminal translocation domain firmly attached to an intermediate receptor-binding domain and a terminal activity domain (Figures 1A and 1B). Interestingly, both CmaT/CmaR and CmaR/CmaA interfaces show extensions of approximately 1200 Å2 (1 Å=0.1 nm) to form strong biological domain contacts (Figure 1A and Supplementary Figure S3 at http://www.biochemsoctrans.org/bst/040/bst0401560add.htm). Consequently, the interaction between FhuA and Cma must induce a significant conformational change to the long N-terminal tail (possibly via a competitive binding mode) to release the CmaT from the protein to become accessible to the TonB protein.
The CmaR domain is fully α-helical with a hydrophobic and exposed patch of ten residues in the first α1 helix indicating possible membrane association (Figure 1A). The terminal domain CmaA resembles a mixed α/β-fold with α-helices forming the R/A interface and a strongly bent four-stranded β-sheet forming the terminal part of the structure. The phosphatase active site of CmaA has been mapped to six residues in the Ser223–Arg236 residual range [6–9,25] (Figures 1C and 1D). These residues are hydrophilic and responsible for the recognition and the catalysed processing of the C55-PP substrate in a magnesium-dependent manner. Five of these residues are strongly conserved within the small class of CmaA domains (Figure 1D).
Structure and mechanism of Pst, a functional phage T4 homologue
Y. pestis produces and secretes only one toxin to kill related bacteria of the same niche. Uptake of Pst is similar to Cma and requires the FyuA receptor instead of FhuA for translocation followed by hydrolysis of the peptidoglycan [13,17,26,27]. In analogy to Cma, the Pst comprises three domains: the N-terminal domain is similar in length to CmaT and partially unstructured. The PstR domain comprises a new mixed α/β-fold, whereas the C-terminal activity domain is almost entirely α-helical with only three short β-strands contributing to the architecture of the active site of PstA. This domain shows a cleft with an approximate diameter of 1 nm, which is shown to harbour all residues of the active site (Figure 3A). The active-site residues (Glu178, Thr201 and Asp207) are observed inside the cleft and can be superimposed structurally to related phage lysozyme residues (in T4 lysozyme, Glu11, Asp20 and Thr26). The cleft is flanked by positively charged residue patches, which probably attract and align the negatively charged peptidoglycan (Figures 2B and 2C).
Mechanistic insights into Pst translocation
To understand the uptake mechanism and to investigate the function of Pst, we combined data of crystal structures of the wild-type enzyme, active-site mutants of Pst and a chimaeric protein with in vivo and in vitro activity assays. The Pst protein is toxic to bacteria carrying the Pst receptor FyuA. Uptake studies of truncation mutants in the translocation domain demonstrate their critical size for import. Whereas a five-residue truncation in the PstT domain is tolerated, it decreases the activity to only 10%; truncation of ten residues inactivates the protein completely (Figures 3A and 3B). To test further the plasticity of Pst during uptake into bacterial cells, the activity domain was replaced by T4 lysozyme. Surprisingly, this replacement resulted in an active chimaera protein, which is not inhibited by the immunity protein Pim. The structural basis of both fused domains is maintained. Activity of Pst and the chimaeric protein (Pst–T4L in Figure 3D) was blocked through introduction of disulfide bonds, which suggests unfolding as the prerequisite to gain access to the periplasm  (Figures 3C and 3D). Our study is the first report of a protein chimaera targeting E. coli cells. This study has implications for the general use of phage lysins in targeting pathogenic bacteria. Furthermore, by using translocation and receptor-binding domains of bacteriocins engineered to proteins of interest, the selective uptake of non-E. coli proteins may be enhanced. Chimaeric proteins may be labelled before uptake for studies of the protein in vivo using EPR and fluorescence spectroscopy.
How Bugs Kill Bugs: Progress and Challenges in Bacteriocin Research: A Biochemical Society Focused Meeting held at University of Nottingham, U.K., 16–18 July 2012. Organized and Edited by Colin Kleanthous (Oxford, U.K.), Chris Penfold (Nottingham, U.K.) and Dan Walker (Glasgow, U.K.).
My research is supported by the German Science Foundation [grant number ZE522/4-1], the Max Planck Society and the Human Frontier Science Programme [grant number RGP61/2007].