Correspondence: Mercedes Maqueda, Departamento de Microbiología, Facultad de Ciencias, Universidad de Granada, Fuentenueva s/n, E-18071 Granada, Spain. Tel.:/fax: +34 958 242857; e-mail: email@example.com
This review highlights the main genetic features of circular bacteriocins, which require the co-ordinated expression of several genetic determinants. In general terms, it has been demonstrated that the expression of such structural genes must be combined with the activity of proteins involved in maturation (cleavage/circularization) and secretion outside the cell via different transporter systems, as well as multifaceted immunity mechanisms essential to ensuring the bacteria's self-protection against such strong inhibitors. Several circular antibacterial peptides produced by Gram-positive bacteria have been described to date, including enterocin AS-48, from Enterococcus faecalis S-48 (the first one characterized), gassericin A, from Lactobacillus gasseri LA39, and a similar one, reutericin 6, from Lactobacillus reuteri LA6, butyrivibriocin AR10, from the ruminal anaerobe Butyrivibrio fibrisolvens AR10, uberolysin, from Streptococcus uberis, circularin A, from Clostridium beijerinckii ATCC 25752, and subtilosin A, from Bacillus subtilis. We summarize here the progress made in the understanding of their principal genetic features over the last few years, during which the functional roles of circular proteins with wide biological activity have become clearer.
The phenomenon of microbial antagonism mediated by peptides has gained considerable attention in recent years because of their rapid bactericidal activity over a broad spectrum and the low propensity of bacteria to develop resistance against them. Antimicrobial peptides differ considerably in their primary structure although they are all normally fairly short molecules (12–100 amino acids) and are very often amphiphilic (Hancock & Diamond, 2000). They tend to kill their target cells by permeabilizing the cell membrane (Nissen-Meyer & Nes, 1997), although novel modes of action such as the inhibition of nucleic acid synthesis, protein synthesis, enzyme activity and cell-wall synthesis have recently been described (Brogden, 2005). Although their primary function was believed to be antimicrobial, it is now becoming clear that these peptides have a wide repertoire of activities that do not function solely against microorganisms and have interesting repercussions on the immune system (Jenssen et al., 2006). Interestingly, this antimicrobial arsenal has been isolated from almost all species of life – prokaryotic and eukaryotic microorganisms, plants, insects and other invertebrates, fish, amphibians, birds and mammals, including humans – and it has been proposed that they represent crucial components of their immediate nonspecific defence against infections, protecting the host from virtually all kinds of invading bacteria, enveloped viruses and fungi (reviewed by Jenssen et al., 2006), which all have a high electric potential in their cytoplasmic membrane.
Bacteriocins stand out among the wide variety of antimicrobial ribosomal peptides synthesized by bacteria. They have been found in all major lineages of Bacteria (revised by Riley & Wertz, 2002), and some members of the Archaea have also been seen to produce similar antimicrobial proteins (O'Connor & Schand, 2002). According to Klaenhammer (1988), 99% of all bacteria may produce at least one bacteriocin and the only reason why researchers have not isolated more is that one has not been looking for them. Bacteriocidal proteins produced by Gram-negative bacteria, e.g. the colicins produced by Escherichia coli, are often larger than 20 kDa and inhibit the growth of closely related strains (Konisky, 1982). Colicin V and the Microcins, a heterogeneous group of small peptides (>10 kDa) produced by enterobacterial strains, are exceptions. Microcins can be distinguished from colicins by their smaller size and the fact that their synthesis, nonlethal for the producer strains, is not SOS inducible (Moreno et al., 1995). In contrast to Gram-negative bacteria, Gram-positive bacteria usually produce bacteriocins smaller than 8 kDa, which, because of their amphipathic properties, are often membrane-permeabilizing peptides. In this way, they resemble many of the antimicrobial peptides produced by eukaryotes, although they are less self-toxic. These bacteriocins have so far been grouped into two major classes, which display great diversity with regard to their mode of action, structure, genetics, mode of secretion and choice of target organisms (Nes et al., 2007): class I: lantibiotics; which are small, posttranslationally modified peptides that contain unusual amino acids such as lanthionine, and class II: the heat-stable nonlantibiotics. A third class of bacteriocins has been suggested that includes secreted heat-labile, cell-wall degrading enzymes, but the classification of such enzymes as being bacteriocins has recently been thrown into doubt (Cotter et al., 2005a). In contrast to this, Heng et al. (2007) propose to retain this class for large bacteriocins, which is subdivided into IIIa (bacteriolysins) and IIIb (nonlytic proteins). The lantibiotics have been divided into two subgroups: type A, elongated molecules with a flexible structure in solution, and type B, which tend to have a more rigid, globular structure (although there are some exceptions). Lack of consensus also exists in the distinction between subgroups of the nonlantibiotics belonging to class II, which include a large number of small peptides that only have in common their marked thermostability and a lack of modified residues (Cleveland et al., 2001; Eijsink et al. 2002; Ross et al., 2002; Cotter et al., 2005a; Drider et al., 2006). In this new classification scheme, Nes et al. (2007) subdivided class II into four subclasses: IIa (antilisteria pediocin-like bacteriocins), IIb (two-peptide bacteriocins), IIc (leaderless peptide-bacteriocins) and class IId (circular bacteriocins). Nevertheless, according to the recommendations of several research groups (Kemperman et al., 2003b; Kawai et al., 2004b; Maqueda et al., 2004; Heng & Tagg, 2006; Heng et al., 2007), the circular, posttranslationally modified bacteriocins deserve to be upgraded to a new class IV.
The circular bacteriocins: description of the gene clusters
Of particular interest among the most typical circular bacteriocins described (Table 1) are the peptide enterocin AS-48 (reviewed by Maqueda et al., 2004), which is homologous with no other bacteriocin described so far, except circularin A, produced by Clostridium beijerinckii ATCC 25752 (Kemperman et al., 2003b; Kawai et al., 2004b) gassericin A and reutericin 6, two hydrophobic bacteriocins produced by species from Lactobacillus, which show 98% similarity to acidocin B, produced by Lactobacillus acidophilus (Kawai et al., 2004a). Others include butyrivibriocin AR10, produced by the anaerobic ruminal bacterium Butyrivibrium fibrisolvens AR10 (Kalmokoff et al., 2003), which displays significant amino acid similarity to gassericin A (Leer et al., 1995; Kalmokoff et al., 2003), uberolysin, the first circular bacteriocin to be characterized (biochemically and genetically) from Streptococcus uberis (Wirawan et al., 2007), and the only circular lantibiotic described to date, and subtilosin A, produced by Bacillus subtilis (Marx et al., 2001; Kawulka et al., 2004). These bacteriocins range in size from 58 to 78 amino acids and are produced by Gram-positive bacteria via gene translation. Their only difference from conventional linear proteins is that the gene-encoded precursor protein must be posttranslationally modified to join their termini to produce a seamless circle of peptide bonds. The structural details of subtilosin A (Fig. 1a) (Marx et al., 2001; Kawulka et al., 2003, 2004) and AS-48 (Fig. 1b), obtained by nuclear magnetic resonance (NMR) (González et al., 2000) and X-ray diffraction (Sánchez-Barrena et al., 2003), reveal a well-folded three-dimensional structure with α-helix motifs repeated regularly in their secondary structure, but no notable differences from those of their linear counterparts. Cross-linking the termini of a chain decreases the conformational entropy of the disordered and highly flexible linear polypeptides, thus stabilizing their bioactive conformation and rendering them more resistant to proteolysis due to a lack of exopeptidase cleavage sites.
Table 1. Characteristics of circular bacteriocins from Gram-positive bacteria
Previous reviews in this field have focused on the biochemical and structural aspects of this new generation of proteins. For the sake of introduction and continuity this review refers to some of these earlier works but deals primarily with the more recent aspects of genetic research. A comparative analysis of their genetic determinants leads to the recognition of complex but common features involved in the expression and secretion of these inhibitors and also immunity against them. Their complexity, together with the biosynthetic resources necessary to produce such circular proteins, must have evolved because of the advantages conferred on them by circularization, i.e. increased resistance to protease digestion and enhanced thermodynamic stability and integrity in the protein structure, both of which seem to improve their biological activity in vivo. In fact, due to their wide spectrum of action and stability across considerable pH and temperature ranges, they constitute one of the most attractive groups of inhibitors described to date. We update the composition of the group and define the common properties of these circular bacteriocins before highlighting the most recent observations regarding the transcriptional regulation of their genes, together with the biosynthetic machinery and characterization of immunity proteins.
The numerous studies carried out on this bacteriocin have helped to clarify its molecular composition and structure (reviewed by Maqueda et al., 2004). Analysis of the composition of purified AS-48 reveals a high proportion of basic amino acids, which give it a strong basic character with a pI of nearly 10.5. It does not contain lanthionine, β-methyllanthionine or dehydrated residues, thus making it clearly different from lantibiotics and defensins in that it lacks disulphide bridges, and so its additional stability must derive from its circular backbone (Cobos et al., 2001). Determination of the DNA sequence of the AS-48 structural gene (Martínez-Bueno et al., 1994), the sequences of its protease digestion fragments and complementary mass-spectrometry results have led to the unambiguous identification of its circular structure. This was the first example of a posttranslational modification described in which a circular structure arises from a ‘head-to-tail’ linkage (Samyn et al., 1994). Reports of the complete 1H NMR assignment of AS-48 proton resonances, together with the resulting secondary structural pattern, were essential for the determination of the high-resolution, three-dimensional solution structure resolved by NMR (González et al., 2000) (Fig. 1b). AS-48 folds in an all-α-helix manner with five helical regions spanning residues 9–21 (α1), 25–34 (α2), 37–45 (α3), 51–62 (α4) and 64–5 (α5) (Langdon et al., 1998; González et al., 2000) (Fig. 1b). The backbone circularization of the AS-48 precursor occurs in the middle of helix α5 and appears to confer considerable stability upon the three-dimensional structure, the thermal denaturation temperature of which is 102 °C (Cobos et al., 2001).
Enterocin AS-48 interacts with the cytoplasmic membrane of susceptible bacteria, into which it inserts itself in a voltage-independent manner, inducing ion permeation accompanied by a collapse of membrane potential. As has been suggested, pore formation may require a certain distribution of the electrostatic and/or hydrophobic surfaces, and an analysis of structure–function relationships is essential to a final understanding of the interaction between AS-48 and cell membranes. The binding of negatively charged membranes may be initiated by the high positive electrostatic potential of the polypeptides (cf. review Maqueda et al., 2004). Thus, the positive charges in the side chains of the residues in helix α4 and in the turn linking helix α4 to helix α5 form a cluster that might determine its antibacterial activity by promoting the formation of pores in cell membranes. Analysis of the oligomeric structure found in AS-48 crystals, however, probes deeper into the mechanism of its antibacterial activity, which involves a transition from the water-soluble DF-I to the membrane-bound DF-II at the membrane surface (Fig. 1b), implying an effective insertion into the membrane that involves a structural reorganization of a hydrosoluble dimeric form of AS-48 (Sánchez-Barrena et al., 2003). In fact, the results obtained with a 21-residue peptide fragment of bacteriocin AS-48 containing only its putative membrane-interacting region confirm that the mechanism of membrane disruption by this bacteriocin is more complicated than one merely driven by a deposition of positively charged molecules on the bacterial membrane. Some other regions of the protein must be present, such as, for instance, the hydrophobic patches surrounding the dimer DF-II, to enhance the accumulation of the peptide and favour membrane permeation (Jiménez et al., 2005).
Sequence analysis of the as-48 region
The as-48 gene cluster for enterocin AS-48 and the identical bacteriocin 21 are located on conjugative, pheromone-response plasmids pMB2 (68 kb) (Martínez-Bueno et al., 1990) and pPD1 (59 kb) (Tomita et al., 1997), respectively, both from Enterococcus faecalis strains. The only exception is the AS-48RJ variant, from Enterococcus faecium, in which the as-48 gene cluster is located on the bacterial chromosome (Abriouel et al., 2005). Interestingly, analogous genetic information (pBt136–140 determinants) involved in the production and export of a peptide similar in length and sequence to the processed propeptide of AS-48 has been found in pBtoxis, a toxin-coding plasmid of Bacillus thuringiensis ssp. israelensis (Berry et al., 2002), although no study has been undertaken into the possible production of circular peptide antibiotics at the protein level. The as-48A structural gene encodes a 105-amino-acid prepeptide consisting of a 35-amino-acid signal peptide and a 70-amino-acid propeptide, coinciding with the previously determined AS-48 amino-acid sequence described by Martínez-Bueno et al. (1994) (Fig. 2). The signal peptide contains a typical positively charged N terminus (four residues) followed by a stretch of hydrophobic residues and a polar region next to the cleavage site. The data presented in Fig. 2 show that AS-48 maturation requires a series of modifications during secretion via the following events: the AS-48-translated primary product is a proprepeptide of 105-amino-acid residues processed between H−1 and M+1 by a specific peptidase to produce a 70-amino-acid prepeptide, which must undergo a new posttranslational change that results in the linking of the N-terminal methionine (M+1) to the C-terminal tryptophan (W+70), with the elimination of a water molecule, to result in the active and mature AS-48 molecule (Martínez-Bueno et al., 1994) (Fig. 2).
According to the results published by Martínez-Bueno et al. (1998) and Díaz et al. (2003), the full expression of AS-48 and immunity to AS-48 depend on the co-ordinated expression of ten genes: as-48A, B, C, C1, D, D1, E, F, G and H (Fig. 3) (GenBank accession number X79542). The characteristics of the DNA sequence analysis and genes identified are described in Table 2. Interestingly, the two as-48BC overlapping genes are located 73 nt downstream of the TAA termination codon of the as-48A structural gene, separated by a characteristic inverted repeat (IR) with an estimated ΔG energy of −22.4 kcal mol−1 and the capacity to form a complex secondary structure (Fig. 4), which was initially thought to serve as a transcription terminator signal (Martínez-Bueno et al., 1998). Nevertheless, recent results from site-directed mutagenesis experiments, in which three bases within the IR were replaced, have demonstrated that this structure represents a characteristic substrate for endoribonucleases involved in mRNA processing and is essential for the controlled expression of the as-48BC genes (Fernández et al., 2008). Moreover, the following modules of genes described in the as-48 cluster also overlap: as-48C1DD1, followed by an intergenic region of 204 nt, after which come the contiguous as-48EFGH genes, followed by a strong transcriptional terminator sequence (Martínez-Bueno et al., 1998; Díaz et al., 2003).
Hydrophilic except for the N-terminal segment. Coiled-coil domain.
Auxiliary protein involved in As-48G recycling
Hydrophilic ATP-dependent protein
ABC transporter immunity
Integral membrane protein (4 TM)
ABC transporter immunity
Functional and transcriptional DNA analysis of the as-48 cluster
Most of the gene products involved in AS-48 production and immunity are predominantly basic, hydrophobic proteins, several of which have a high number of potential transmembrane domains (TM) (twelve in As-48B, six in As-48C1 and four in As-48C, As-48E and As-48H). Accurate predictive analytical methods have been used to investigate the possible function of As-48B, a large, hydrophobic protein (563 residues, pI of 10.48) previously described by Martínez-Bueno et al. (1998). The existence of similar internal sequences detected in this protein seems to indicate a putative duplication in the molecule, which would imply a structure of 6+6 helical spanning segments. The homologies found have been only with membrane proteins involved in transport or pore formation (unpublished results), in which the identification of domains and functional assignment of the sequences were not available. Nevertheless, phenotypic studies carried out upon the as-48ABC operon using mutagenesis with Tn5 revealed that inactivation of as-48B unambiguously prevented AS-48 production, thus confirming that it plays an essential part in AS-48 biogenesis. Disruption of the neighbouring as-48C gene, which encodes a hydrophobic 178-amino-acid protein (Table 2), decreases AS-48 production and resistance (Fernández et al., 2008). Using homology comparisons, Martínez-Bueno et al. (1998) determined that the As-48D protein encodes the functional domain of an ATP-binding cassette (ABC) transporter, and As-48C1 (with six TMs) constitutes the membrane-spanning domain for this ABC transporter. On the basis of the phenotypes of different insertion mutants, Díaz et al. (2003) concluded that the as-48C1D transporter was devoted to exporting newly synthesized bacteriocin and providing low levels of immunity. The function of this transporter cannot be replaced by the second multicomponent ABC transporter, As-48EFGH, which is clearly involved in self-protection against exogenously administered AS-48 and works as a complementary immunity mechanism (Díaz et al., 2003).
The expression of the as-48 gene cluster has been studied by Northern blotting using total RNA extracted from several JH2-2 transformants (Fig. 3). Two mRNA populations (T1 and T2 transcripts) were detected in the transcriptional analysis, thus ruling out the possibility that the as-48 genes are transcribed as a single operon. Nevertheless, transcript T1, driven by the promoter PA, transcribes the as-48ABC genes (TABC) and undergoes endonucleolytic cleavage, involving the accumulation of a very stable mRNA degradation product (TA) and a TBC with a very short lifespan, thus making its detection almost impossible (Fig. 3). The reasons why the as-48ABC genes are cotranscribed must be investigated through the functional analysis of their gene products in the cell: As-48B alone or in association with the As-48C1D translocation complex might form a pore to facilitate the exit of AS-48 molecules (Martínez-Bueno et al. 1998), while the integral membrane protein As-48C is involved in protection functions (Fernández et al., 2008). Apart from this, the long T2 transcript (c. 6.4 kb) transcribes the last seven genes (as-48C1DD1EFGH), although a second and shorter mRNA (T3, 5.4 kb) also translates the last four genes (as-48EFGH), a fact that confirms the existence of an internal promoter (Díaz et al., 2003). All in all, the results of Díaz et al. (2003) corroborate the importance of this second set of gene products in the cell, the constitutive expression of which is safeguarded by the two mRNAs T2 and T3. Finally, it is worth highlighting that a quorum-sensing mechanism similar to those described in most of the bacteriocin systems characterized to date, lantibiotics (reviewed by Kleerebezem, 2004) and nonlantibiotics (reviewed by Eijsink et al., 2002), does not exist in the as-48 region analysed.
Gassericin A, reutericin 6 and acidocin B
Biochemical and biological characteristics
Gassericin A and reutericin 6 are two highly hydrophobic, circular bacteriocins produced by two species of Lactobacillus, Lactobacillus gasseri LA39 (Kawai et al., 1994) and Lactobacillus reuteri LA6 (Toba et al., 1991a, b), which were isolated within a 2-month interval from faeces of the same human infant. Both bacteriocins were thought to be the same substance because they have the same structural genes and cannot be distinguished by their molecular weights (MWs) or primary amino-acid sequences (Kawai et al., 2001). Mature bacteriocins (58-amino-acid residues, in which I1 and A58 must be covalently linked) were purified to homogeneity from culture supernatants by reverse-phase chromatography. They present more than 74% hydrophobic residues on their surfaces and are resistant to several peptidases and proteases, as are the other circular bacteriocins (Kawai et al., 1998, 2001). The MWs of both reutericin 6 and gassericin A, determined by time-of-flight mass spectroscopy, are 5652 and the only difference found between these bacteriocins is the presence of one (gassericin A) or two (reutericin 6) d-Ala residues in the molecule, which could be the cause of the differences in their modes of action and secondary structures (Kawai et al., 2004a). Bacteriocins with the same primary structures produced by different bacterial species have been reported for other lactic-acid bacteria (LAB) species such as Pediococcus acidilactici and Lactobacillus plantarum (pediocin AcH) (Motlagh et al., 1992; Ennahar et al., 1996) and Lactobacillus curvatus and Lactobacillus sake (curvacin A-sakacinA) (Axelsson & Holck, 1995), but the study by Kawai et al. (2004a) was the first to report the different characteristics of identical native primary bacteriocins such as gassericin A and reutericin 6, conferred by a difference in d-Ala contents in the molecules, apparently arising from an l-Ala to d-Ala conversion. The importance of the chirality of d-alanines in the activity of posttranslationally modified lantibiotics was studied by Skaugen et al. (1994). As has been demonstrated, the l-serines from lactocin S are enzymatically converted to d-Ala, giving rise to an apparent mistranslation of serine codons to alanine residue. It has been suggested that in such a case, this conversion results from a two-step reaction initiated by a lantibiotic synthetase converting the gene-encoded l-serine to dehydroalanine (dha). Recently, Cotter et al. (2005b), using lacticin 3147 as a model system, identified the enzyme responsible for the conversion of dha to d-Ala, confirming the vital biological role for d-amino acids in the activity of this ribosomally synthesized peptide.
As with the other bacteriocins analysed here, their secondary structures are organized mainly in α-helices (closer to the α/β type than the α+β), which yield a compact, stable structure and maintain their activity after having been heated to 100 °C for 60 min. Both bacteriocins have been described as acting in the cytoplasmic membrane of target cells and causing the death of the cell via an efflux of potassium ions, although they induce different quantities and patterns of this efflux. The activity of reutericin 6 is influenced by substances on the surface of the indicator cells, which may cause the time lag in the mode of action and the low efflux pattern (Kawai et al., 2004). Nevertheless, their inhibition spectra against a number of Gram-positive, food-borne, pathogenic bacteria, such as Listeria monocytogenes, Bacillus cereus and Staphylococcus aureus, but not against Gram-negative organisms, do not completely coincide, suggesting that the secondary structures of the two bacteriocins are slightly different (Itoh et al., 1995; Kawai et al., 2004b). Interestingly, gassericin A inhibits the growth of Lactobacillus reuteri LA6, but reutericin 6, which has a narrower spectrum, does not inhibit the growth of Lactobacillus gasseri LA39, and in addition loses <50% of its activity at low pH values.
Acidocin B (58 amino acids) is another hydrophobic class IV bacteriocin produced by Lactobacillus acidophilus M46, which is active against Listeria monocytogenes, Clostridium sporogenes, Brochothrix thermosphacta, Lactobacillus fermentum and Lactobacillus delbrueckii ssp. bulgaricus, but inactive against most other Lactobacillus species (Leer et al., 1995). According to in silico analysis, although this peptide was not identified as a circular protein (Leer et al., 1995) it might be considered to be a natural variant of gassericin A (Kawai et al., 2004a) because of its high sequence similarity (98%), only differing in one residue (M24 V). According to alignment with gassericin B, acidocin A should have a MW of 5620 [although the weight deduced by Kawai et al. (2004a) is 5750], all this in spite of its apparent value of 2400 on sodiumdodecyl sulphate polyacyrlamide gel electrophoresis due to its hydrophobic nature (Kawai et al., 2004b). Acidocin B is susceptible to trypsin, remains stable after heating to 80 °C for 20 min and can be extracted from the fluid culture supernatant with butanol (ten Brink et al., 1994). It has been established that growth is not necessary for its production because washed producer cells can synthesize the bacteriocin in a defined production medium.
Identification of the gaaA and acdB structural genes
The structural gene of gassericin A (gaaA), located on the chromosomal DNA of Lactobacillus gasseri LA39, was found as an ORF that encoded a protein of 91 amino acids with a calculated MW of 9268, which was sited downstream of another ORF encoding a 60-amino-acid protein that remains to be identified (Kawai et al., 1998) (GenBank accession number AB007043). Both ORFs were deduced to have a putative Shine–Dalgarno (SD) sequence, although no promoter-like sequence was identified upstream. More recently, however, a gene encoding a protein with a putative ATP-binding domain probably involved in the secretion of gassericin has been located downstream of gaaA (Kawai et al., 2004a). From the DNA sequence and results of mass analysis, however, the amino-acid sequence of mature gassericin A was predicted to be 58 residues long, beginning with Ile as the N-terminal residue. The putative leader peptide sequence of this bacteriocin contains 33 residues with no Gly–Gly motifs but with a cleavage site (NH2… TTAN*IYWIADQF …) (Fig. 2) (Kawai et al., 1998). In a similar manner, the bacteriocin structural gene of reutericin was located on the chromosome of Lactobacillus reuteri LA6 (Kawai et al., 2004b).
The genes encoding the production of acidocin B have been located on the pCV461 plasmid of Lactobacillus acidophilus M46 (GenBank accession number Z34920). DNA sequence analysis revealed the presence of three consecutive ORFs, which potentially encode hydrophobic peptides of 60, 91 and 114 amino acids, and a fourth one of opposite polarity, which may well encode a peptide of 59 amino acids (Leer et al., 1995) (Table 3 and Fig. 3). By comparing the amino-acid composition of highly purified acidocin B with the deduced amino-acid sequence, ORF-2 was identified as the structural gene (acdB) while the next gene encoded a protein with a putative ATP-binding domain involved in the secretion of gassericin A (Kawai et al., 2004b). Moreover, Leer et al. (1995) suggest the existence of a gene that encodes an immunity protein for one of the three nonassigned ORFs. Finally, analysis of the intergenic region carried out here with the mfold program revealed the existence of a long IR with an estimated ΔG free energy of −20.70 kcal mol−1 and the capacity to form a complex secondary structure (Fig. 4).
Table 3. Characteristics of predicted proteins specified by the gene cluster involved in the production of and immunity to gassericin A, acidocin B and butyrivibriocin AR10 (adapted from and Kawai et al., 2004a, b and Kalmokoff et al., 2003, respectively)
Gassericin A precursor
Acidocin B precursor
ATP-binding protein (peripheral)
Integral membrane protein (2 TM)?
Butyrivibriocin AR10 precursor /
Gas A (38%) AcdB (47%)
Membrane immunity protein/ ORF3 Acidocin B (29%)
Biochemical and biological characteristics
Butyrivibriocin AR10, produced by the ruminal anaerobe Butyrivibrium fibrisolvens AR10, is a new member of the circular molecule group and shares a degree of homology with gassericin A (Kawai et al., 1998) and acidocin B (Leer et al., 1995). This bacteriocin has a wide spectrum of activity among Butyrivibrio isolates but its activity against other genera is fairly limited (Teather et al., 1999). Amino-acid analysis of butyrivibriocin AR10 purified to homogeneity confirmed a very high content of nonpolar residues, which is consistent with its extreme hydrophobicity, evident from both its elution characteristics with the reverse-phase column and the enhancement of activity in the presence of nonionic surfactants. As established by negative-ion, time-of-flight, mass-spectroscopic analysis, the MW of the isolated bacteriocin is 5981.5 Da, which corresponds to the molecular mass generated by the removal of the leader peptide and subsequent circularization, thus explaining the resistance to N-terminal Edman degradation (Kalmokoff & Teather, 1997). Its N-terminal sequence is homologous with that of gassericin A and both share features of sec-dependent leader peptidase cleavage sites (Kalmokoff et al., 2003). Moreover, the similarities between the N- and C-terminal regions of mature gassericin A before circularization (propeptide) and those predicted for butyrivibriocin AR10 (Fig. 2) would imply that they may be involved in the processing and cleavage events that go towards formation of the circular molecules, or else may be important for bacteriocin activity (reviewed by Kawai et al., 2004a).
Functional DNA analysis of the bvi cluster genes
A single sequence encoding a peptide of 80 amino acids was identified in the genome of Butyrivibrium fibrisolvens AR10 using a probe based on the amino-acid sequence of a cyanogen bromide (CNBr) fragment of the producer strain. Analysis in silico of the product of bviB, later renamed bviA (Kalmokoff et al., 2003), showed some sequence identity with gassericin A (38%) (Kawai et al., 1998) and acidocin B (47%) (Leer et al., 1995) (Table 3). The predicted amino-acid sequence of BviA is identical to the internal amino-acid sequence determined, except that found at position 18 of the internal peptide sequence (A18S), which has been interpreted as an amino-acid sequencing error (Kalmokoff et al., 2003). The predicted prepropeptide BviA (22 amino-acid residues) lacks the conserved double Gly leader peptidase processing motif common to the majority of previously characterized class II bacteriocins (Nes et al., 1996) but has a putative N-terminal cleavage site (NH2…LIPN*YIFIADKM…) similar to that of the leader peptidase cleavage site in gassericin A (Kawai et al., 1998) (Fig. 2).
Five nonoverlapping ORFs (bviBCDE and ORF 6) have been identified in the cloned region flanking the bacteriocin structural gene (Fig. 3) encoding peptides of 216, 53, 80, 169 and 64 amino acids (Table 3) (Kalmokoff et al., 2003) (GenBank accession number AF076529). Both Northern blotting and primer extension analysis indicate that the bviBCDAE region forms an independent transcriptional unit driven by a P1 promoter found 500-bp upstream of the start of bviB. Interestingly, the main transcript of 1 kb encompassed the bviA and bviE genes, although longer transcripts were also detected. This suggests the existence of a processing or degradation of a longer message encompassing the bviBCDAE genes. In the region preceding bviA, the presence of a small IR might explain the differences observed in the transcription levels of the bviBCD genes, present at relatively low levels compared with bviDE (Fig. 3) (Kalmokoff et al., 2003).
Data concerning the enhanced levels of transcription of bviA in the presence of glucose, which increases culture density, suggest that butirivibriocin production may be regulated via a quorum-sensing mechanism, which would be in accordance with the existence of two ORFs (1 and 2) located immediately before the bvi operon and encoding a two-component regulator system, similar to those described in other bacteriocins (Fig. 3).
Owing to the homology of BviB with the ATP-binding proteins and the predicted membrane-spanning domains of the hydrophobic proteins BviC and BviD (six and two TMs, respectively), it had been suggested that BviBCD might act as an ABC transport system involved in enhancing the production of and resistance to butyrivibriocin (Teather et al., 1999; Kalmokoff et al., 2003), while BviE has been considered to be the immunity protein (Kalmokoff et al., 2003).
Biochemical and biological characteristics
Streptococcus uberis 42 produces two bacteriocins, a nisin U variant (Wirawan et al., 2006) and a second inhibitor, which has recently been identified as a novel member of the circular bacteriocins and named uberolysin (Table 1). This is an environmental microorganism belonging to the LAB that can be isolated from the lips and skin of cows, raw milk and udder tissue and is a major cause of bovine mastitis (Cullen, 1966; Khan et al., 2003). In contrast to other circular bacteriocins, uberolysin is an unusual 7048 Da protein because of its thermolability and capacity to lyse actively growing Gram-positive bacteria. Susceptible bacteria include most of the streptococci tested (except Streptococcus rattus and Streptococcus mutans), Listeria spp., enterococci, Micrococcus and Staphylococcus aureus (Wirawan et al., 2007).
As usual, the determination of the primary structure of uberolysin encountered an initial difficulty in that it was not susceptible to Edman degradation, suggesting that the N-terminal residue was blocked. For this reason, the authors analysed a linear fragment of 19 residues, spanning the Ala65 to Lys13 residues, which they obtained by tryptic digestion. MS/MS collision-induced fragmentation (CID) analyses confirmed that this fragment included the head-to-tail bond between the end residues from mature uberolysin and lacked further posttranslational modifications.
Functional DNA analysis of the ubl cluster
UblA is a 76 amino-acid polypeptide encoded by the ublA structural gene, having a short leader peptide consisting of only six amino acids, after the removal of which circularization occurs between L+1 and W70 (Fig. 2). Sequencing of chromosomal DNA flanking the ublA gene revealed the existence of several ORFs, called ORF1, ublB, ublC, ublD and ublE (Table 4), which are probably involved in posttranslational modification, transport and immunity, and also possibly in the regulation of the biosynthetic gene cluster (GenBank accession number DQ650653). ORF1, located 303-bp upstream of ublA but in the opposite transcriptional direction (Fig. 3), encodes a ComE-like protein with conserved-domain homology to the LytR family of response regulators of the two-component signal transduction system. This type of regulator is involved in quorum sensing, biosynthesis of extracellular polysaccharides, fimbriation and expression of exoproteins, including toxins and bacteriocins (Nikolskaya & Galperin, 2002). It is also thought to be essential to uberolysin production, in spite of the absence of the corresponding ORF with homology to sensor kinases in its vicinity. The ubl cluster is in fact flanked (52 bp from each end) by two large 159 bp direct repeats (DR), and the authors wonder whether it may be involved in the acquisition of this gene cluster (Fig. 3).
Table 4. Characteristic of predicted proteins specified by the ubl cluster from Streptococcusuberis 42, involved in the production of and immunity to uberolysin (adapted from Wirawan et al., 2007)
Regulation of bacteriocin production
Integral membrane protein
Cir D ATP-binding protein
As usual, in the circular bacteriocins, the following ublBCDE genes, located 90 nt downstream of ublA, overlap in this intergenic region but an IR similar to that found in the majority of the cluster genes analysed here does not exist. Based on their amino-acid sequence similarities, the products of ublCD probably encode a transporter complex whereas UblB appears to be the equivalent of the CirB proteins (23% identity). The presence of an additional ABC-transporter has not, however, been detected in the uberolysin cluster but, as in the other bacteriocin clusters, a gene that may function as the putative immunity gene (ublE) has also been proposed.
Biochemical and biological characteristics
A new member of the circular bacteriocins, circularin A, a 69-residue peptide produced by Clostridium beijerinckii ATCC 25752, has been described by Kemperman et al. (2003a). Its inhibitory spectrum is quite narrow, although it inhibited all the Clostridium tyrobutyricum strains tested and also lactococci, enterococci and some Lactobacillus strains. Initial attempts to determine the N-terminal amino-acid sequence of the purified bacteriocin failed, indicating that the N terminus may be blocked. Its circular structure was resolved after purification to homogeneity from culture supernatants by reverse-phase chromatography and CNBr cleavage of the native protein into two fragments, which were then purified by HPLC to determine their amino-acid sequences (Kemperman et al., 2003b). Using blast searches against the NCBI nonredundant protein database or by direct comparison, Kemperman et al. (2003b) established a 60% homology with the amino-acid sequence of mature circularin A (69-amino-acid residue, pI 10.60) (30% identity) with that of AS-48, although it lacks the extended leader of this enterocin (Martínez-Bueno et al., 1994; Kemperman et al., 2003b). In fact, only the first three N-terminal amino acids of pre-CirA (72 residues) are cleaved by hydrolysis between L−1 and V+1, allowing the formation of a peptide bond between V+1 and the C-terminal Y+72 (Fig. 2b).
Functional and transcriptional DNA analysis of the cir cluster
The structural cirA gene was identified in the chromosome of Clostridium beijerinckii ATCC 25752 (GenBank accession number AJ566621). The cirA gene – which is preceded by a Shine–Dalgarno sequence (AAGGAGGT) and a putative promoter region – encodes a peptide of 72 amino-acid residues, followed by a large palindromic structure (with an estimated ΔG free energy of −22.90 kcal mol−1) similar to that described for the majority of the structural genes reviewed (Fig. 4). In addition, an 11-kb region embracing the structural gene cirA encompasses two clusters of genes: the cir genes (cirABCDEGHI) involved in bacteriocin biosynthesis, maturation, secretion and immunity, and the cfg genes (cfgR, cfgK, cfg01, cfg02), two of which encode a two-component regulatory system. It is noteworthy that most of the gene products of the cir cluster show some degree of homology to the proteins involved in the production of AS-48, although no homologue of the putative As-48C1 protein is found in the cir operon. Two modules of genes (cirABCDE and cirGHI) are putatively transcribed as polycistronic messengers (Fig. 3). The cirABCDE genes represent the minimum region required for bacteriocin biosynthesis, processing and extracellular production (Kemperman et al., 2003a): cirABCD gene products are required for the production of active circularin, while cirE seems clearly involved in bacteriocin immunity (Table 5), an assumption that will be discussed below. According to homology analysis, CirD and CirH both contain ATP-binding domains (Fig. 3) and CirG belongs to the HlyD family of accessory proteins of ABC transporters, while CirI, with four putative TMs and a Duf214 domain of predicted permeases, may be the membrane-spanning domain of a second ABC transporter (CirGIH). From the results obtained from single deletions in these genes, it has been concluded that CirB and CirD are required for partial resistance in the absence of CirE, but it seems clear that CirBD is principally involved in circularin secretion. It is worth noting, however, that the functional analysis proposed by Kemperman et al. (2003a) for the gene products CirB to CirI was performed mainly on the basis of homologies to proteins involved in the production of enterocin AS-48, because the roles of the Cir proteins were not investigated. There is no experimental evidence concerning the protein(s) involved in the processing and/or circularization of the CirA prepeptide, and the suggestion for an essential CirC protein involved in performing this function (Table 5), either alone or together with CirB and/or CirD, was only made on the basis of its homology with As-48C, which has, however, recently been identified as an accessory immunity determinant (Fernández et al., 2008).
Table 5. Characteristics of predicted proteins specified by the cir gene cluster (adapted from Kemperman et al., 2003b)
Protein putative function
TM, transmembrane domains.
Hypothetical AgrB regulatory protein
Enterocin AS-48 (30%)
Cir A precursor
As-48B (19%) (11 TM)
As-48C (21%) (4 TM)
ATP-binding protein As-48D (31.6%)
ABC transporter secretion/immunity
As-48D1 (30%) (2 TM)
HlyD family (1TM)/As-48F (17%)
Accesory protein ABC transporter
ATP-binding protein As-48G (31.6%)
Integral membrane (4 TM) As-48H (32%)
ABC transporter permease
As in the other cluster genes in circular bacteriocins, the genetic organization of the cir region is somewhat compressed in the form of several overlapping genes. Overlapping between the end of one gene (cfgR, cirC, cirD, cirH and cirI) and the predicted start of the downstream gene suggests regulation of expression by translational coupling, a common feature throughout the cir region (Fig. 3). For cirH, translational coupling to cirG would be the only means of expression, as it lacks any obvious ribosome-binding site. As far as the cfg cluster is concerned, CfgR and CfgK are homologous to the response regulators and histidine kinases of two-component regulatory systems, which are often involved in the regulation of bacteriocin expression, and their genes are normally located near the bacteriocin operon. Unfortunately, there have been no further reports about the possible involvement of CfgR and CfgK in CirA expression.
Biochemical and biological characteristics
Subtilosin A (35 amino acids) is the only anionic, antimicrobial, circular peptide produced by the soil bacterium Bacillus subtilis. In the same way as several of the bacteriocins reviewed here, subtilosin A is highly resistant to endoproteinases, is very stable at high temperatures and defies complete sequence analysis by Edman degradation and mass spectral examination (Babasaki et al., 1985; Marx et al., 2001). It shows bactericidal activity against a diverse range of Gram-positive and Gram-negative bacteria, aerobes and anaerobes, although it is less effective against capsulated forms of some Gram-negative bacteria (Shelburne et al., 2007). Heat shock increases its effectiveness against certain Gram-negative strains but protease has no effect on its activity (Shelburne et al., 2007). The widespread occurrence of subtilosin A reflects important physiological roles for this bacteriocin, which is produced at the end of exponential growth, particularly under stress conditions, and with specific functions during both the anaerobic and biofilm growth of Bacillus subtilis (Stein et al., 2004).
Structural studies have shown that subtilosin A is posttranslationally modified in a singular way, placing it in a new type of lantibiotic, classified as class I (Table 1). According to the structure proposed by Babasaki et al. (1985), the removal of a short leader peptide (only eight amino-acid residues) leads to a peptide of 35 amino acids in length with a circular peptide backbone (Fig. 2). Subtilosin A contains several small amino-acid residues (seven Gly, five Ala), which are distributed throughout the whole molecule, thus allowing considerable flexibility, although as a consequence of its circular structure the conformational space of the backbone is highly restricted. NMR studies of labelled subtilosin A (Kawulka et al., 2003, 2004) (Fig. 1a) are consistent with its determined molecular mass (3399.7 Da), chemical reactivity and amino-acid composition. These studies also show that in addition to its circular structure, presubtilin undergoes several processing steps including proteolytic cleavage at the N+1 and a new posttranslational modification, the oxidative linkage of cysteine sulphur to the α-carbon of another amino-acid residue, which has never been observed before in any ribosomally synthesized peptide (Kawulka et al., 2004). In fact, three cross-links are formed between the sulphurs of Cys13, Cys7 and Cys4 and the α-positions of Phe22, Thr28 and Phe31, respectively, with the loss of a hydrogen from each of these six residues. The thioether linkages between Cys and Ser/Thr are characteristic of lantibiotics (Kleerebezem, 2004), but the Cys–Phe linkage constitutes a new thioether type in this posttranslationally modified peptide. Finally, the circularization of the presubtilosin peptide involves covalent linking of the C-terminal G35 with the N-terminal N+1 (Zheng et al., 1999) (Fig. 2). All these features confer high rigidity upon this molecule, the plausible antimicrobial activity mechanism of which seems to follow a receptor-binding and -mediated lipid-perturbation model (Thennarasu et al., 2005). Solid-state NMR and fluorescence data have led to a model depicting the topology of subtilosin A embedded in the hydrophilic/hydrophobic interface of the lipid bilayer (Fig. 1a). It has been proposed that subtilosin A may bind to lipid bilayers and orientate itself in such a way that the tryptophan-bearing edge of the molecule is buried within the hydrophobic core so as to disturb the positioning of both the phosphate head group and the methylene units near the glycerol backbone. This binding results in a leakage of the aqueous contents of small unilamellar vesicles, albeit at higher peptide concentrations. Thennarasu et al. (2005) have recently investigated the type of interactions of subtilosin A with various model membranes and have come to the conclusion that it is modulated by the lipid composition of the target, the concentration used in the assay being an important factor.
Functional and transcriptional analysis of genes from the sbo-alb cluster
Current findings indicate that the production of mature subtilosin A requires the expression of eight (sboA-albABCDEFG) of the nine genes identified in the chromosome of the Bacillus subtilis producer strains (GenBank accession number AJ430547). These clustered genes are transcribed from a promoter residing upstream of the sboA structural gene (Fig. 2) and their products are involved in the posttranslational modification and processing of presubtilosin, secretion and also immunity (Zheng et al., 1999, 2000).
The first four genes from the spo-alb operon are highly conserved in various subspecies of Bacillus subtilis (96–100% amino-acid identity) while the remaining ones are less conserved (83–88% identity). The putative coding sequences of the alb operon (Table 6) encode proteins that potentially function in the processing and export of peptides: the albA and albF products are critical for production while the albB, albC and albD products have a role in export and immunity. AlbA, involved in the posttranslational modification of presubtilosin (Zheng et al., 2000), was identified as a member of the MoaA/NifB/PqqE family (Menendez et al., 1995) and possesses two Cys clusters, located at an Fe–S centre, which serve as the active sites in reactions involving hydration or dehydration of substrate compounds. The polypeptides encoded by albC, -E, and -F show primary structural similarity to known proteins: AlbC probably functions in the export as an ABC transport complex, AlbE is a processing protease and AlbF is a member of a family of zinc endoproteinases that may perform critical modifications of the presubtilosin peptide. It has been demonstrated that AlbB, -C and -D are required for full immunity. There is another gene, sboX, which presumably encodes a bacteriocin-like precursor peptide bearing a Gly–Gly motif resembling type II prebacteriocin cleavage sites (Zheng et al., 2000).
Table 6. Characteristics of the sbo and alb genes involved in the production of and immunity to subtilosin A (adapted from Zheng et al., 2000)
Homology/protein putative function
Subtilosin A precursor (pro-peptide)
ABC transporter. Export and immunity
ATP-binding cassette (ABC transporter). Export and immunity
ABC transporter. Export and immunity
Bacteriocin-like precursor peptide
A 55-bp sequence that potentially encodes a region of RNA secondary structure was identified 93 bp upstream of the albA start codon, overlapping the C-terminal coding end of sboA (Zheng et al., 1999) (Fig. 4). It is similar to a factor-independent transcriptional termination sequence that may reduce transcriptional read-through into the alb genes, thereby ensuring that the alb products are generated in smaller quantities compared with the SboA substrate peptide, as occurs in other similarly organized bacteriocin biosynthesis operons (Jack et al., 1995; Martínez-Bueno et al., 1998; Qi et al., 1999; Kemperman et al., 2003a). The regulation of the expression of the sbo-alb operon is quite complex in that it is induced in late growth cultures apparently in response to starvation and also to oxygen-limited and anaerobic conditions (Nakano et al., 2000). In fact, it has been demonstrated that the transcriptional regulation of the sbo-alb operon is under dual control and involves the negative control of the global transition-state, transcriptional regulatory protein AbrB (a regulatory protein that controls the nutritional stress response and processes of cell differentiation in Bacillus subtilis) and the two-component regulatory proteins ResD and ResE, which could positively regulate sbo-alb transcription by repressing the abrB gene or indirectly affecting the concentration of AbrB and/or its activity (Nakano et al., 2000). Besides this, it is exciting to notice that the variance of this operon may be involved in an evolutionary relationship among different producer strains (Stein et al., 2004).
Immunity determinants for circular bacteriocins
Bacteriocin producers are immune to their own bacteriocin, which results from the presence of a specific immunity protein. Genes encoding immunity determinants are usually situated within or close to related operons containing the genes required for bacteriocin biosynthesis. To date, at least 20 putative immunity proteins from the expressing strains have been identified from DNA sequences and subgrouped according to sequence similarities (Drider et al., 2006) in spite of there appearing to be little sequence homology among them. Products with functions related to immunity have been identified as small, highly charged, membrane-associated peptides. Their selectivity seems to be located in the C-terminal parts, the bacteriocin and immunity proteins being located on opposite sides of the cell membrane without any direct contact between the two molecules. Therefore, either the membrane itself or a specific membrane component seems to play a crucial role as a recognition intermediary between the bacteriocin and the immunity protein (Drider et al., 2006). In contrast to this, products concerned with immunity in bacteriocins other than those in class IIa have also been identified as membrane-interacting immunity proteins, some having a lipid modification that is thought to anchor the peptide to the cytoplasmic membrane (Ra et al., 1996; Sarik et al., 1996; Martínez-Bueno et al., 1998; Drider et al., 2006). When expressed in susceptible cells, immunity proteins provide strong protection against attack by exogenous cognate bacteriocins (Drider et al., 2006). It has been proposed that the immunity protein could interact with the bacteriocin-induced pore and thereby blocks leakage, or alternatively, with a bacteriocin receptor that inhibits its action either by altering the receptor conformation or by hiding its bacteriocin-binding site (Drider et al., 2006). In spite of their importance, however, these proteins' mode of action is still poorly understood. Furthermore, the success of their protective function often depends on the coexistence of complementary ABC systems having specific secretory/immunity functions. This type of transporter has been described in bacteriocins that open pores in the cytoplasmic membrane, and their way of providing immunity could be by preventing them from reaching the high local densities required for pore formation, i.e. by keeping the bacteriocin concentration in the cytoplasmic membrane below the critical level necessary for pore formation, either by active transport of the mature molecules to the inside of the cell or by the active extrusion of bacteriocin molecules. These ABC transporters seem to represent the conservation during evolution of a wide resistance mechanism with secretion/protection functions, probably interacting co-operatively with the immunity protein. It is obvious that co-operatively acting, multiple immunity systems provide a higher degree of immunity than does each system alone.
The function of as-48D1 plus as-48EFGH and a-48C products in AS-48 immunity
The results published by Martínez-Bueno et al. (1998) confirm that As-48D1 (56 amino acids) affords some degree of bacteriocin resistance by itself and thus it has been identified as the immunity determinant. Judging by its amino-acid sequence, the as-48D1 immunity gene encodes a small (6.3 kDa) cationic hydrophobic protein with the structure of an integral cell-membrane protein rather than being weakly associated with the membrane. In fact, the predicted hydrophobic profile and helical wheel representation of As-48D1 suggest the existence of two transmembrane segments with a typical α-helix conformation and a C-terminal hydrophilic region (unpublished results).
More recently, Díaz et al. (2003) have proposed that the internal, and possibly regulated, operon as-48EFGH is related to higher resistance against exogenously added AS-48 and indeed must be involved in self-protection against AS-48. The function of As-48EGH seems to be to enhance the secretion of AS-48 and resistance to exogenous enterocin, expelling it from the membrane into the surrounding medium, helped by the As-48F protein, which seems to be involved in accelerating the release of ADP from the ATP-binding protein As-48G (Díaz et al., 2003).
Finally, the as-48C gene, which is co-ordinately expressed with the as-48ABC operon, has recently been identified as an auxiliary immunity determinant. It has been postulated that the bacteriocin and the immunity-protein encoding genes must be cotranscribed to ensure that the producer strain is not killed by its own bacteriocin. The posttranscriptional regulation of as-48ABC, recently identified by Fernández et al. (2008), should provide producer cells with maximized production of functional AS-48 without deleterious effect, before the entire immunity machinery, as-48D1 and as-48EFGH determinants, which are located in an independent operon, begins to work.
The bviE gene confers butyrivibriocin AR10 immunity
The direct linkage and relative abundance of the transcript encompassing bviDE has led Kalmokoff et al. (2003) to propose that BviE is the immunity protein. In addition, the hydrophobic character of BviE and the prediction of membrane-spanning domains, indicating an envelope location, are both consistent with the protective function surmised for this protein.
The ublE plus ublCD secretory genes confer uberolysin immunity
According to the results presented by Wirawan et al. (2007), the ubl cluster contains only one gene that may function as the putative immunity gene (ublE), because no additional ABC-transporter has been detected in the uberolysin locus. It has been suggested, however, that a supplementary protective mechanism could be provided by the UblCD transporter involved in secretory function, which may be capable of conferring low levels of immunity.
The cirE and cirBD(?) products confer circularin A immunity
Reduced circularin A sensitivity could be acquired via two independent mechanisms, one of which is based on the expression of cirE and the other on the combined expression of an ABC transporter. CirE is a small protein with a high isoelectric point and is organized into two predicted transmembrane helices, indicating that the protein is possibly located within the membrane, in a manner similar to As-48D1, the immunity protein of AS-48, and also other bacteriocin immunity proteins. According to the results presented by Kemperman et al. (2003a), mutation in either cirB or cirD together with deletion of cirE led to the loss of the circularin A-resistant phenotype, in a similar way to the results obtained in the as-48 cluster after mutation of their respective as-48C1 or as-48D homologous genes (Martínez-Bueno et al., 1998). Nevertheless, it is plausible that the second system conferring CirA immunity would depend on the combined expression of CirE with CirGHI. This suggestion is in accordance with the proposed homology with the BacGHI proteins, identical to the as-48FGH multicomponent ABC transport system (Tomita et al., 1997; Díaz et al., 2003; Kemperman et al., 2003b).
albB plus albCD, with secretory functions, confer subtilosin immunity
Zheng et al. (2000) proposed that the small 59-amino-acid hydrophobic peptide encoded by the albB gene plays a critical role in immunity because its alteration severely diminishes subtilosin self-protection. The small, hydrophobic AlbB is also required for maximum bacteriocin production, just as the other immunity peptides encoded by lantibiotic biosynthesis operons are. How these products function in conferring bacteriocin production and also self-protection is still not known, but mutations in albC and, to a lesser extent, in albD reduce the cell's immunity to subtilosin and the amount of active peptide produced. This is explicable because the product of albC is a member of the ABC family of transport proteins, which participate in the export of subtilosin and have been shown in several bacteriocins required for complete immunity to the specific peptide produced.
Unanswered questions regarding the machinery involved in processing and circularization
Despite present progress in defining the biology of circular bacteriocins, the understanding of the enzymatic mechanisms involved in the circularization processes is still incomplete and lacking experimental evidence. Nonetheless, considerable attention is being directed towards identifying the posttranscriptional machinery and processing steps involved in the reaction(s). The function of the leader peptides, the method of leader cleavage and the steps involved in the circularization of the peptide remain unknown. It is still not known, for example, whether the reactions to join their ends from linear cleaved AS-48 precursors are prompted by two different enzymes in almost simultaneous steps or by two reactions catalysed by the same enzyme. It has been confirmed, however, that all the circular structural genes are translated into precursor peptides and it is proposed that the production of mature linear peptides requires the removal of the N-terminal leader and the joining of the pro-peptide ends by a covalent peptide bond (Fig. 2). Intriguingly, the leaders examined so far are either extended (33 or 35 residues for gassericin A and AS-48, respectively) or have only a few amino-acid residues (three, six and eight for circularin A, uberolysin and subtilosin A, respectively) and lack any conserved sequences.
Hegde & Bernstein (2006) have recently reported the complexity of the leader sequences of secreted proteins with cleavable N-terminal signal sequences that mediate their targeting to and translocation across bacterial cytoplasm membranes. They propose that signal sequences contain information that specifies the choice of targeting pathway, the efficiency of translocation, the timing of cleavage and even postcleavage functions, and play an important role in modulating protein biogenesis. Nevertheless, the absence of homology between the amino acids involved in joining the termini and the differences found in the leaders of these circular proteins remains a puzzle. Some of the many questions still to be answered are: Is there some relationship between the sites at which this process takes place? Are the cleavage and secretion coupled? What is the nature of the putative translocation intermediate? Is ATP required for folding at the outside of the membrane? Are there different mechanisms governing the circularization process, or might they use the same secretory pathways?
Nevertheless, in the scientific literature reviewed there have been attempts to assign such functions to some of the coding products identified. For example, when Kawulka et al. (2003) arrived at their conclusions about the genetic machinery required for the production of mature subtilosin A, they proposed that the two processing peptidases AlbE and AlbF might be candidates for cutting the leader from the prepeptide and circularizing the linear product. In addition, the albA product, belonging to the radical SAM protein family, which is involved in the catalysis of several unusual reactions, would form the required Cys–Phe linkage described above (Sofia et al., 2001). As far as circularin A is concerned, there is no experimental evidence to identify the protein(s) involved in the processing and/or circularization of the CirA prepeptide, although Kemperman et al. (2003b) suggest that CirC protein could be the candidate to perform this function, either alone or together with the CirB and/or CirD proteins. A similar situation has been described for ublB, which has been proposed as the protein involved in the circularization of uberolysin (Wirawan et al., 2007). Nevertheless, these suggestions are based exclusively on homologies shared with the genes encoding circularin A or enterocin AS-48. Interestingly, Martínez-Bueno et al. (1998) have proposed that As-48B, alone or in association with the As-48C1D translocation complex, might open a pore specifically to facilitate the exit of AS-48 molecules. This hypothesis could imply that the ultimate activity of the putative multicomponent As-48BC1D proteins would be to remove the leader peptide and promote head-to-tail circularization concomitantly with the export of the molecule. Recent results show, however, that cleavage or circularization (or both) must be fulfilled by specific machinery from bacteria belonging to the Enterococcus genus (Fernández et al., 2007), assuming that processing may require chromosomally derived auxiliary proteins similar to those described for the circularization of pilin in Agrobacterium tumefaciens (Eisenbrandt et al., 2000; Lai et al., 2002). An understanding of the precise mechanism underlying the processing of AS-48 and its transport to the cell surface is still a main challenge because no product with the necessary specific peptidase activity and the ability to produce the circular AS-48 backbone has yet been identified in the as-48 cluster. Speculations still exist about the requirement of either one or two enzymes involved in this process: e.g. an atypical transpeptidase that would catalyse the necessary double reaction or alternatively a signal peptidase involved in the proteolytic cleavage to remove the 35-amino-acid leader peptide, followed by a second enzyme linking the N-terminal M+1 to the C-terminal W+70. Even more, if a normal signal peptidase is used and cleaves a peptide bond upstream of the H−1 residue, a second enzyme would be needed to remove the propeptide N-terminal amino acid(s) in excess and link the N-terminal M+1 with the C-terminal W70 (Martínez-Bueno et al., 1994). Finally, it may be very interesting to confirm whether AS-48 can only be expressed by its natural host (Enterococcus species) or also by species from Clostridium, as has been observed in circularin A, the closest circular bacteriocins described so far. Nevertheless, in order to further our understanding of the mechanisms by which they function, the current level of knowledge about prebacteriocin processing needs to be extended to include the enzymes involved.
This review is devoted to describing the genetic characteristics of a unique group of circular bacteriocins produced exclusively by Gram-positive bacteria. Particular attention is paid to the genetic description of those bacteriocins that have been characterized at the molecular level, highlighting the most recent developments concerning the identification of the genes involved in the production and maturation of these extremely special proteins and the immunity of their producer bacteria. Whenever possible, the elaborate transcriptional regulation of the gene clusters involved in their expression is also explored.
The genes of this new family of circular bacteriocins are chromosomally encoded, except those encoding acidocin B and enterocin AS-48, for which genes have been located on plasmids, with the exception of an AS-48 variant (AS-48RJ), produced by Enterococcus faecium, the as-48 gene cluster of which has been located on the bacterial chromosome (Abriouel et al., 2005). As far as the similarities and specificities of the gene composition of these clusters are concerned, they are organized into polygenic operons, a well-known strategy of bacterial genomes to facilitate their co-ordinated regulation. This is true of the nine genes of the sbo-alb operon (7 kb) from Bacillus subtilis (Zheng et al., 2000) and the five genes ublABCDE and bviBCDAE) from Streptococcus uberis and Butyrivibrium fibrisolvens– although a two-component regulatory system has been identified in an operon upstream of the bvi cluster – (Kalmokoff et al., 2003), and possibly also true of the tentative triple-gene gaa operon (Kawai et al., 1998, 2004b). No information is available about the transcription of the ubl cluster. In other cases, the genes are spread between two gene modules transcribing as polycistronic messengers, where one operon carries the essential genes and a second one carries genes for bacteriocin secretion and immunity, as is the case of the cirBCDE and cirGHI genes in the cir region (Kemperman et al., 2003a, b) and the as-48ABC and as-48C1DD1EFGH operons from the as-48 cluster (Martínez-Bueno et al., 1998; Díaz et al., 2003). Interestingly, these modules are quite compressed, with overlapping genes requiring translational coupling, an additional strategy that ensures the appropriate ratio of all their products (McCarthy & Gualerzi, 1990; van de Guchte et al., 1992). An analysis of available DNA sequences (as-48, acd, bvi, ubl, cir, sbo/alb) with the mfold program revealed the existence of a long IR with the potential to form a complex secondary structure after the respective structural genes (Fig. 4), the ubl and bvi clusters being the exceptions. The positioning of the IR has usually been seen as evidence that it is a transcriptional terminator, acting as an attenuator to ensure that the proteins encoded by the last genes are present in small, catalytic amounts while the peptide bacteriocin is produced in larger quantities (Martínez-Bueno et al., 1998; Zheng et al., 1999). Fernández et al. (2008) have recently proposed, however, that in the as-48 cluster this structure really represents specific sites for mRNA processing, allowing posttranscriptional processing and differential stability of the primary transcript to control the levels of the encoded proteins. In a similar way, it has been suggested that the differences observed in the transcript levels in the bvi cluster may be regulated at the level of mRNA stability (Teather et al., 1999). This processing uncouples transcription and translation during the expression of polygenic operons, favouring differential mRNA decays, which provide a high level of mRNA for an efficient pre-protein product translation in comparison with a less abundant RNA substrate for the synchronized synthesis of accompanying polypeptides, thus ensuring the maintenance of an appropriate stoichiometry between them. Finally, there is an alternative way of controlling the differential production of proteins from the same operon, i.e. via the use of internal promoters such as those that putatively precede the as-48D1 and as-48EFGH immunity genes in the multigenic as-48C1DD1EFGH operon, thus ensuring the appropriate levels of immunity/resistance determinants against the high concentrations of AS-48 produced (Díaz et al., 2003).
It is known that most secreted and many membrane proteins contain extended, cleavable, N-terminal leader sequences that mediate their targeting to and translocation across the bacterial cytoplasmic membrane. Nevertheless, in the case of subtilosin, uberolysin and circularin, with three-, six- and eight-residue leader peptides, respectively, it is not likely that these sequences contain information to control their maturation/secretion. On checking the different secretion pathways of the bacteriocins analysed for this review, it was observed that there was no conserved Gly–Gly leader peptidase processing motif in any of the prebacteriocins investigated. Kawai et al. (1998) and Kalmokoff et al. (2003) suggest, in accordance with common features with sec-dependent peptidase (a positively charged residue at the N-terminus, a hydrophobic core region and cleavage following a neutral amino acid in position −1) and the absence of an ABC transporter, that acidocin B, gassericin A and butirivibriocyn AR10 are secreted via the prepeptide translocase (sec) pathway. Enterocin AS-48 is, however, secreted by the products of an ABC transporter identified as As-48C1D, despite a histidine residue found at position-1, which is very unusual for a signal peptide cleavage site (Martínez-Bueno et al., 1994, 1998). On the basis of this hypothesis, Kemperman et al. (2003a, b), Wirawan et al. (2007) and Zheng et al. (2000) have proposed that irBD, UblCD and AlbCD constitute putative ABC transporters for circularin, uberolysin and subtilosin, respectively, in spite of the fact that their three- six- and eight-residue leader peptides do not seem to contain any information to control their maturation/secretion.
Permeation of the membrane, leading to cell death, has been the accepted mechanism for the action of a large number of membrane-lytic polypeptides (Jenssen et al., 2006). All these bacteriocins share common antimicrobial activity at the cytoplasmic membrane of Gram-positive bacteria (except for AS-48, which is also active against Gram-negative) through a transient pore formation that results in their depolarization, thus causing the death of the susceptible cell. Nevertheless, and contrary to the principal mechanisms for the activity of hydrophobic and low-MW bacteriocins (Moll et al., 1999), gassericin A and reutericin A do not cause efflux of ATP through the cytoplasmic membrane of susceptible strains, indicating that neither of these bacteriocins causes cell lysis or the leakage of substances with MWs of more than 507 (ATP). The binding of subtilosin A to the lipid bilayer results, however, in the leakage of aqueous contents from small unilamellar vesicles, albeit at higher peptide concentrations. Interestingly, the mechanism proposed for the molecular function of AS-48 involves transition from a water-soluble dimer form (DF-I) to a membrane-bound dimer form (DF-II) at the membrane surface by a nonspecific interaction, which leads to the formation of nonselective pores and the free diffusion of low-MW solutes (Fig. 1b). Nevertheless, prolonged incubation with increased AS-48 concentrations has also been shown to cause fusion of the membranes and even more chaotic disorganization as secondary effects (Gálvez et al., 1991).
The high cost in energy involved in the production of circular bacteriocins, associated with the complex immunity mechanism for the cells that include the immunity gene, encoding an immunity protein to protect the bacteriocin producer from its own bacteriocin, and the other accessory proteins described above must also be emphasized. It is known that the expression of several genes must be involved in a high level of precursor peptide production and ATP hydrolysis for the bacteriocin-mediated secretion of neo-synthesized bacteriocins and also for the second ABC-transporter-dependent immunity when it exists. Moreover, it is still not known whether the posttranslational circularization also has some ATP requirements. Clearly, exciting challenges lie ahead during the years to come.
The authors' investigation into enterocin AS-48 was supported by the Spanish Dirección General de Investigación Científica y Técnica (Projects BİO 98-0908-C02-01, BIO2001-3237 and BIO2005-01544) and the Junta de Andalucía (PAI CVI 160). The authors are grateful to Dr Soler-González and Dr Ramírez-Rodrigo of the Departments of Biochemistry and Molecular Biology at Granada University for their help in the prediction and functional assignment of the As-48B protein. The authors thank A.L. Tate for revising our English text.