Antimicrobial peptides (AMPs) are a group of antibiotics that mainly target the cell wall of Gram-positive bacteria. Resistance is achieved by a variety of mechanisms including target alterations, changes in the cell's surface charge, expression of immunity peptides or by dedicated ABC transporters. The latter often provide the greatest level of protection. Apart from resistance, ABC transporters are also required for the export of peptides during biosynthesis. In this review the different AMP transporters identified to date in Firmicutes bacteria were classified into five distinct groups based on their domain architecture, two groups with a role in biosynthesis, and three involved in resistance. Comparison of the available information for each group regarding function, transport mechanism and gene regulation revealed distinguishing characteristics as well as common traits. For example, a strong correlation between transporter group and mode of gene regulation was observed, with three different types of two-component systems as well as XRE family transcriptional regulators commonly associated with individual transporter groups. Furthermore, the presented summary of the state-of-the-art on AMP transport in Firmicutes bacteria, discussed in the context of transporter phylogeny, provides insights into the mechanisms of substrate translocation and how this may result in resistance against compounds that bind extracellular targets.
As bacterial infections are becoming increasingly difficult to treat due to rising numbers of multidrug-resistant strains, efforts are being directed towards the discovery of new drugs, ideally targeting different bacterial structures to circumvent the development of cross-resistance. One group of antibiotics that has received considerable attention in recent years are the antimicrobial peptides (AMPs). They include structurally very diverse compounds, such as the heavily modified lantibiotics (e.g. nisin or mersacidin), unmodified bacteriocins (e.g. pediocin), non-ribosomally synthesized cyclic AMPs (e.g. bacitracin), glycopeptides (e.g. vancomycin) or lipodepsipeptides (e.g. ramoplanin) (reviewed for example in Ennahar et al., 2000; Guder et al., 2000; Cotter et al., 2005; Breukink and de Kruijff, 2006).
Their target spectra can also differ considerably, but most of the compounds mentioned are active against Gram-positive bacteria with a low G+C content (Firmicutes). All of these AMPs share a mode of action that inhibits the lipid II cycle of cell wall biosynthesis, albeit at different steps. Bacteriocins of class I (lantibiotics) and class II (non-lantibiotic bacteriocins) bind to lipid II, preventing incorporation of cell wall precursors into the growing peptidoglycan layer. Some bacteriocins are able to form pores in the cytoplasmic membrane, often using lipid II as a docking molecule (Guder et al., 2000; Cotter et al., 2005; Breukink and de Kruijff, 2006). Ramoplanin and enduracidin were shown to interfere with the transglycosylation reaction of cell wall biosynthesis, again by binding to lipid II (Fang et al., 2006), while vancomycin binds the terminal d-Ala–d-Ala of the pentapeptide chain, thus inhibiting transpeptidation (Perkins, 1969). Finally, bacitracin binds to undecaprenyl-pyrophosphate (UPP) that remains after incorporation of the cell wall precursors and thereby prevents recycling of the lipid carrier molecule undecaprenyl-phosphate (UP) (Storm and Strominger, 1973).
To defend themselves against AMPs, Firmicutes bacteria have developed a number of different mechanisms. For example vancomycin resistance is achieved by changing the terminal d-Ala residue of the pentapeptide chain to d-Lac, which drastically reduces binding by vancomycin (Arthur and Courvalin, 1993). Many AMPs are positively charged, and a general resistance mechanism can therefore be a reduction in the net negative charge of the bacterial cell wall. One way this is achieved is by incorporation of d-alanine into teichoic acids, thus adding a positive charge. d-alanylation of teichoic acids is catalysed by the products of the dltABCDE operon (Neuhaus and Baddiley, 2003; Cao and Helmann, 2004; McBride and Sonenshein, 2011). Another mechanism is lysinylation of the membrane lipid phosphatidylglycerol, catalysed by MprF, which also adds a positive charge to the cell envelope (Ernst and Peschel, 2011). Producer strains of lantibiotics often possess specific immunity peptides, collectively referred to as LanI, which function in self-protection (Draper et al., 2008). However, the most efficient resistance mechanisms often involve ATP-binding cassette (ABC) transporters. Several different types of such transporters have been described as self-resistance mechanisms in AMP producing strains as well as for protection against foreign AMPs. In addition to providing resistance, ABC transporters are also required for peptide export during biosynthesis of AMPs.
A number of excellent reviews have summarized the biosynthesis, structure and mode of action of AMPs (Ennahar et al., 2000; Guder et al., 2000; Cotter et al., 2005; Breukink and de Kruijff, 2006; Draper et al., 2008; Kjos et al., 2011; Alkhatib et al., 2012). However, no dedicated article has been published on the various ABC transporters involved in AMP export, self-immunity and resistance. The aims of this review were therefore to collect the wealth of information published to date on such transporters from Firmicutes bacteria, to derive a classification that can be easily applied to newly identified transporters, and to compare the identified groups regarding their physiological role, transport mechanism and gene regulation.
Classification of AMP transporters based on domain architecture
As mentioned above, several different types of ABC transporters for AMPs have been identified in Firmicutes bacteria. Annotation of newly identified transporter genes is hereby usually based on sequence similarity to previously described systems. One aim of this review was to pool the currently available functional information on these transporters to facilitate more informed predictions. While the databases contain many ABC transporters that either are encoded in AMP biosynthetic loci or have been implicated in resistance from physiological experiments, there is as yet no convincing classification. A search of the available literature showed that AMP transporters from Firmicutes could be divided into five groups by a simple comparison of their predicted domain architecture using the SMART database (Letunic et al., 2012). These proposed groups, each named after one characterized member, can be described as follows.
This group contains large transporters of approximately 700 amino acids with the ATPase domain fused to the C-terminus of the permease domain (Fig. 1A, dark blue frame, and Fig. 4A). Additionally, they contain an N-terminal peptidase C39 domain, which is responsible for pre-peptide processing. Their permease domains consist of five or six predicted transmembrane helices, depending on the employed algorithm. Available functional data suggest an intracellular localization of the peptidase domain, thus six transmembrane helices appear more likely as discussed below (Håvarstein et al., 1995; Franke et al., 1999). The SMART database (Letunic et al., 2012) contains over 800 proteins with this composition (January 2012), mostly from Firmicutes and Proteobacteria, but also from Actinobacteria and Cyanobacteria. After removal of duplicates, 181 distinct SunT-type transporters could be identified in Firmicutes bacteria (Table S1).
These proteins are similar to SunT-type transporters except that they do not contain a peptidase domain and are thus smaller with approximately 550–600 amino acids (Fig. 1A, purple frame, and Fig. 4B). It is not feasible to perform database searches for these transporters based on either domain architecture or sequence similarity alone, because they resemble prototypical fused ABC transporters and therefore produce large numbers of hits with unrelated function. A search of the transport classification database (TCDB) (Saier et al., 2009), supplemented with a manual search of known AMP biosynthetic loci for transporters with the correct domain architecture resulted in a list of nine NisT-type proteins (Table S1).
This group contains transporters with two separate permeases of 200–250 amino acids and six transmembrane helices each and one ATPase (Fig. 1A, green frame, and Fig. 4C). A homology-based analysis of all available bacterial genomes in the MicrobesOnline database (Dehal et al., 2010) using the ‘tree’ search function and the NukFEG transporter of Staphylococcus warneri (Okuda et al., 2008) as query, supplemented with known LanFEG transporters, produced a list of 29 systems (Table S1).
These transporters are comprised of two proteins, one ATPase and one permease of approximately 650 amino acids and 10 transmembrane helices with a large (c. 200 amino acid) extracellular domain located between helices VII and VIII (Fig. 1A, yellow frame, and Fig. 4D). Helices II to IV form an FtsX domain, which can be used as query in domain-based searches of the SMART database (Letunic et al., 2012). An earlier study had shown the existence of over 260 such transporters in genomes of Firmicutes bacteria (Dintner et al., 2011) (see Table S1 for 24 example sequences).
Systems from this group consist of one permease of approximately 230 amino acids with six predicted transmembrane helices and one ATPase (Fig. 1A, light blue/red frame, and Fig. 4E). As with the NisT-type transporters, a domain-based database search was not feasible. However, a search of all available bacterial genomes based on sequence similarity to the permeases of two well-characterized members, BcrB of Bacillus licheniformis (Podlesek et al., 1995) and YydI of Bacillus subtilis (Butcher et al., 2007), using blastp and the MicrobesOnline database (Altschul et al., 1990; Dehal et al., 2010) resulted in a list of 18 such transporters (Table S1).
Classification and phylogeny of AMP transporters
Phylogenetic trees calculated for the ATPase components of all transporters (see supplemental text S1 for details on the phylogenetic analyses) showed that the proteins largely clustered according to the domain-based classification described above (Fig. 1A), even though the latter was mainly based on differences in the permeases. An analysis of the permease regions resulted in a similar phylogenetic tree (Fig. 1B). It should be noted that the BceAB-type transporters were not included in the second analysis, because their 10-transmembrane-helix permeases could not be aligned well with the six-helix permeases of the remaining transporter groups (see supplemental text S1). Comparison of the phylogenetic trees to the domain-based classification described above showed that the BceAB and LanFEG groups of transporters formed distinct phylogenetic clusters. The SunT and NisT transporters formed one intermixed group, suggesting that presence or absence of the peptidase domain does not reflect the evolutionary relationship between both groups. Further, the BcrAB-type transporters fell into two branches according to sequence similarity either to BcrB of B. licheniformis (red lines) or to YydI of B. subtilis (light blue lines). It therefore appears reasonable to split this group into two subfamilies, based on sequence similarity. A striking observation was that the BcrAB-like transporters were found to be closely related and ancestral to the LanFEG-type transporters. The latter appear to have originated from a duplication of the permease gene, followed by divergence of the two new permeases (dark green lines for LanF, light green lines for LanG). The observed mirror-image of the LanF and LanG clusters suggested a co-evolution of the two subunits within each transporter (Pazos and Valencia, 2008).
Overall, the domain-based classification of AMP transporters described here presents a fast and convenient tool for the analysis of newly identified transporters and accurately reflects the phylogeny of these protein families. Therefore the nomenclature suggested above will be used throughout this review.
Genomic context of AMP transporter genes
Antimicrobial peptide (AMP) transporters have often received only cursory attention during the characterization of novel biosynthesis loci, mainly regarding their role in producer self-immunity or in export of the synthesized peptides. Therefore limited experimental data are available on their precise function, transport mechanism or even regulation. Often, valuable functional information about a gene can be derived from the conservation of its genomic context. A detailed analysis of the neighbouring genes of a subset of transporters from each group showed that, while rarely two loci had an identical arrangement, a correlation between transporter type and associated genes nevertheless existed (Fig. 2 and Table 1). Functional categories included for example structural genes for AMPs and their modification, transcriptional regulators or immunity proteins. It is important to keep in mind that likely many functionally important genes have been missed because of low sequence conservation, particularly of AMP structural or immunity genes. This may be reflected in the large number of small hypothetical genes, often with one or more predicted transmembrane helices, found in the neighbourhood of the transporter genes (Table 1). With the exception of BcrAB- and BceAB-type transporters, the transporter loci also often contained transposases or other mobility-associated elements, indicating a high degree of horizontal gene transfer (Table 1). It should be noted that genomic context was analysed only for a few randomly selected example transporters from each group and any numbers given refer to this limited subset. In the following sections each type of transporter is reviewed regarding their physiological role, transport mechanism and regulation, connected where possible to a discussion of genomic context conservation. For reasons of conciseness the main focus was placed on systems that have been at least partially characterized. A summary of all transporters discussed, including their substrate peptides, is given in Table 2.
Table 1. Occurrence of genes by functional categories in genomic context of AMP transporters
aGenes found in each locus sorted by functional category; distributions are given as per cent of loci containing at least one gene in each category, a dash indicates no such gene found; the total number of loci analysed for each transporter type is given in parentheses in the title row. See body text for detail on classification of TCSs; 1–4 TM HP, hypothetical protein with one to four predicted transmembrane helices; for more details see body text.
bAll these loci also contain a NisFEG-like transporter.
cAll these loci also contain a BceAB-like transporter.
dNumber in parentheses shows loci with an XRE regulator containing four transmembrane helices at the N-terminus.
eTwo or more transporters of the same type found in one locus.
aWhere applicable, the classification according to Cotter et al. (2005) was applied; for reasons of simplicity, lantibiotics (class I bacteriocins) were only separated into single (‘Lantibiotic’) and dipeptide lantibiotics.
The transporters of the SunT group are involved in export of AMPs or peptide pheromones with concomitant cleavage of the leader peptides (Håvarstein et al., 1995). However, analysis of the genomic context of 40 example transporters showed that only about half were actually associated with AMP biosynthetic genes (Table 1). Whether this means that the remaining transporters truly exist in the absence of such genes or is simply due to poor gene annotation is difficult to answer. AMP transport by a SunT-type transporter has for example been demonstrated for MutT (mutacin II), LctT (lacticin 481), LcnC (lactococcin A), PedD (pediocin PA-1), AvcT (avicin A) and LcaC (leucocin A) (Marugg et al., 1992; Stoddard et al., 1992; van Belkum and Stiles, 1995; Birri et al., 2010). Experimental evidence for export is usually presented as absence of the target AMP in culture supernatants of transporter mutants. Additionally, it was shown that cells harbouring transport-deficient mutants of NukT accumulated the nukacin precursor peptide intracellulary, demonstrating a lack of peptide export (Nishie et al., 2011). Interestingly, the SunT group does not only contain exporters for AMPs, but also for small peptides used for quorum sensing by Gram-positive bacteria. The best-characterized example is the ComA transporter, which is involved in competence regulation in Streptococcus pneumoniae and is required for export of the competence stimulating peptide (CSP) ComC (Hui et al., 1995). A second example is SilE of Streptococcus pyogenes, which is responsible for excretion of the autoinducer peptide SilCR required for invasive infections (Hidalgo-Grass et al., 2002).
Many SunT-type transporters were found to be associated with transport accessory proteins (Table 1). Experimental evidence that these proteins are required for export of substrate peptides by the ABC transporter is available for several examples, including LcnD (for LcnC) (Stoddard et al., 1992; Franke et al., 1999) and ComB (for ComA) (Hui et al., 1995). The transport accessory proteins possess one predicted transmembrane helix with the N-terminus located intracellulary and the bulk of the protein extracellularly, as was confirmed experimentally for LcnD (Franke et al., 1996). The precise mechanism of transport involving such accessory proteins is unknown. Some accessory proteins discovered during the genomic context analysis, including LcnD, were considerably larger than others (about 450 compared with about 170 amino acids) and contained an additional biotinyl-lipoyl domain of unknown function (Fig. 2). Conversely, more than half of the SunT-type transporters analysed here lack an obvious accessory protein (Table 1), including MutT, MrsT, NukT, ScnT and SunT. Export of the bacteriocins sakacin T and sakacin X by StxT, which also lacks an accessory protein, was experimentally shown (Vaughan et al., 2003), demonstrating that not all SunT-type transporters require such a protein.
A distinguishing feature of SunT-type transporters is their possession of an N-terminal peptidase domain (Figs 1A and 4A), which is responsible for cleavage of the leader peptide. The N-terminal 150 amino acids of the lactococcin G transporter LagD, containing the predicted peptidase domain, were shown to be sufficient for in vitro cleavage of the precursor peptide (Håvarstein et al., 1995). Similarly, the N-terminal domain of LcnC was shown to be required for cleavage of pre-lactococcin A (Franke et al., 1999). An important question in this is whether the peptidase domain is located extra- or intracellularly. Predictions of the transmembrane topology of SunT-type transporters are unclear: the in silico analyses carried out to derive the classification presented above showed mainly five or six transmembrane helices, depending on the algorithm used. However, topology assays of LcnC using fusions to PhoA and LacZ suggested the existence of only four helices (Franke et al., 1999). Experimental data for LcnC and NukT show cytoplasmic localization of the peptidase domain (Franke et al., 1999; Nishie et al., 2011), thus an even number of transmembrane helices, whether four or six, has to be assumed. Such an arrangement would also be consistent with the proposal that substrate recognition by the transporter occurs via peptide binding to the proteolytic domain (Håvarstein et al., 1995). Interestingly, ATP hydrolysis by the ATPase domain and leader peptide cleavage by the peptidase domain are closely coupled cooperative activities during substrate processing and transport, that is mutants of NukT in either domain were incapable both of transport and of leader cleavage (Nishie et al., 2011).
An early observation suggested a correlation between the sequence of leader peptide and type of exporter: substrates with a leader sequence of the double-glycine type are exported by large transporters with a fused peptidase domain, while those with unrelated leader sequences are exported by smaller transporters (Håvarstein et al., 1995; Paik et al., 1998). Applying this observation to the classification presented above, SunT- and NisT-type transporters should recognize substrates with differing leader sequences. An alignment of the first 30 amino acids of eight substrate peptides each for SunT- and NisT-type transporters showed an excellent agreement with this proposition (Fig. 3): all SunT-type substrates possessed the GG/GS motif for cleavage, preceded by a conserved EL/ES motif, while none of the NisT-type substrates contained these elements. Importantly, this applied to all SunT-type substrates tested, including lantibiotics (e.g. sublancin, mutacin II), class II AMPs (e.g. pediocin, lactococcin A) or quorum sensing peptides (e.g. ComC, SilCR), as noted previously (Paik et al., 1998).
The expression of SunT-type transporters appears mainly to be regulated in context with the surrounding biosynthetic gene clusters. Analysis of regulatory genes in the genomic neighbourhood of these transporters showed that many were associated either with a two-component system (TCS) consisting of a six-transmembrane-helix histidine kinase belonging to the group of peptide quorum sensors (Mascher et al., 2006) and a LytR family response regulator, or with an XRE-type transcriptional regulator (Table 1, Fig. 4A). In a few cases a different type of TCS was found, which can be explained by the presence of additional AMP transporters in the same locus, as discussed below.
Peptide quorum sensor TCSs were named based on the role of some characterized members in quorum sensing by Gram-positive microorganisms (Mascher et al., 2006). For example the histidine kinase ComD of S. pneumoniae induces competence in response to the CSP peptide, and a connection between quorum sensing and competence regulation has been drawn (Morrison and Lee, 2000). Similarly, SilB of S. pyogenes responds to the autoinducer peptide SilCR to induce expression of the streptococcal invasion locus (Hidalgo-Grass et al., 2002; Eran et al., 2007). Both CSP and SilCR are first exported by a SunT-type transporter, whose expression is also induced by the TCS in the presence of the signalling peptides (Morrison and Lee, 2000; Eran et al., 2007). Export of the peptide is essential for gene regulation, because mutations in the corresponding SunT-type transporter abolish signalling, which can be restored by externally added peptide (Hui et al., 1995; Eran et al., 2007). Stimulus detection by these quorum sensing kinases appears to involve non-specific interactions of the inducing peptide with a hydrophobic pocket formed by the transmembrane helices, followed by sequence-specific interactions with one of the kinase's extracellular loops (Mascher et al., 2006).
Regulation of AMP biosynthetic loci containing a SunT-type transporter and associated peptide quorum sensor kinase appears to follow the same principle. For example, expression of the avicin A production locus is induced by binding of the pheromone peptide AvcP to the histidine kinase AvcT (Birri et al., 2010), and expression of sakaxinX/T production is regulated via binding of the peptide StxP to the histidine kinase StxK (Vaughan et al., 2003). Many biosynthetic loci containing a SunT-type transporter and a peptide quorum sensor TCS lack a designated inducing peptide. It is however possible that regulation of these loci is autoinduced by the presence of the encoded AMP itself. Analyses of the promoters regulated by the peptide quorum sensor TCSs revealed the presence of conserved binding sites consisting of 9 or 10 bp direct repeats with a spacing of 11 or 12 bp (Belotserkovsky et al., 2009; Birri et al., 2010).
The high similarity in regulation of quorum sensing, competence and AMP production, together with the conservation of transporter type used for peptide export indicates a common evolutionary origin, or even development of one feature out of the other. This close connection is nicely exemplified by the Smb biosynthetic locus of Streptococcus mutans: it does not contain any regulatory genes of its own, but is regulated by quorum sensing in response to the CSP peptide (Yonezawa and Kuramitsu, 2005). A sequence with high similarity to the com-box was identified upstream of the smb operon, thus regulation may occur via the alternative sigma factor for quorum sensing, ComX (Morrison and Lee, 2000; Yonezawa and Kuramitsu, 2005).
Some SunT-type transporters were not found associated with a TCS but instead with an XRE-type transcriptional regulator (Table 1, Fig. 4A). For example, expression of the mutacin II biosynthetic operon mutAMTFEG is regulated by the XRE regulator MutR (Qi et al., 1999), and a mutR-disrupted strain was found to produce no mutacin II (Chen et al., 1999). It is thought that MutR responds to some component of complex growth media to induce mutacin II production, but the precise signal has not been identified (Qi et al., 1999). Because these regulators do not possess any transmembrane helices, they most likely detect an intracellular stimulus, in contrast to the extracellular sensing by TCSs. The brevicin 925A biosynthetic locus also contains an XRE regulator, BreG, but its role in gene regulation has not been investigated (Wada et al., 2009). The presence of such regulators is shared by loci containing NisT- and BcrAB-type transporters (Table 1 and see below).
Like the SunT-type transporters, the NisT-type systems are responsible for the export of newly synthesized AMPs. However, because they lack a peptidase domain, the substrate peptides are translocated without cleavage and are subsequently processed by an extracellular serine protease (Fig. 4B). Consistent with this, a gene encoding a protease was found in the genomic neighbourhood of 67% of the analysed NisT-type transporters (Table 1). The proteases are collectively referred to as LanP proteins, and cleavage of the leader peptide has been experimentally shown for, e.g. NisP (nisin) (Kuipers, 2004) or PepP (Pep5) (Meyer et al., 1995). Some substrates for NisT-type transporters, for example the four-peptide bacteriocin aureocin A70, are not cleaved (Netz et al., 2001). Export activity has been shown experimentally for some NisT-type proteins. For example Pep5 production was reduced to 10% in the absence of PepT, and the AMP was found to accumulate intracellularly in this strain (Meyer et al., 1995). Deletion of NisT abolished secretion of nisin, while modified and thus active nisin was found to be present intracellularly, showing that modification was independent of transport (Qiao and Saris, 1996). Similarly, SpaT was shown to be required for subtilin export (Klein et al., 1992) and AurT for aureocin A70 production (Netz et al., 2001). Low amounts of AMPs detected in the culture supernatants of such transporter mutants have been attributed to export by alternative, as yet unidentified transport systems (Netz et al., 2001), but could also be due to cell lysis caused by intracellular accumulation of the AMP, as was shown for subtilin (Klein et al., 1992).
Prior to export, lantibiotic AMPs are modified by dehydration and thioether ring formation catalysed by LanB and LanC proteins respectively. For subtilin and nisin, the existence of a multimeric complex between the modification and transport proteins, i.e. SpaBCT and NisBCT, respectively, was shown (Siegers et al., 1996; Kiesau et al., 1997). The arrangement of proteins in the former complex was suggested to be SpaC–SpaB–SpaT, because no contacts between SpaC and SpaT could be detected (Kiesau et al., 1997). However, complex formation was not found to be necessary for transport, because in the absence of NisC, a NisBT complex could still dehydrate and export pre-nisin that was lacking thioether rings. Furthermore, NisT alone could also export unmodified pre-nisin or even non-lantibiotic peptides, as long as these carried the nisin leader sequence, showing that it is a rather broad-spectrum transporter and specificity is determined by the leader peptide (Kuipers, 2004). The protease NisP can only cleave the leader if the peptide contains lanthionine residues, but cleavage is not coupled to transport, because externally added pre-nisin is also cleaved by NisP (Kuipers, 2004). Thus, in contrast to the SunT-type transporters, transport and leader processing are separable processes in the biosynthesis of substrate peptides for NisT-type transporters.
Analysis of the genomic context of NisT-type transporters did not reveal any genes for the accessory proteins that were commonly found with SunT-type transporters (Table 1). It should be noted, however, that the data set for NisT-type transporters is small and likely not fully representative, because these systems could not be identified from sequence- or domain-based database searches and were instead collected from known AMP biosynthetic loci. Often, these loci were cloned and sequenced without knowledge of the genetic neighbourhood, and thus relevant genes outside the sequenced regions could be missed.
Interestingly, the two highly similar transporters GdmT and EpiT for the lantibiotics gallidermin and epidermin, respectively, were found to possess accessory proteins, named GdmH and EpiH (Peschel et al., 1997). Because EpiT contains a frameshift mutation, all experiments were conducted with GdmT, and it was shown that this transporter requires the presence of either GdmH or EpiH for substrate export. These accessory proteins differ considerably from those described above for SunT-type transporters and contain six predicted transmembrane helices with a hydrophilic domain located between helices V and VI (Peschel et al., 1997). Their precise role in transport is unknown, and no corresponding genes were found in the loci of the remaining NisT-type transporters analysed here.
Expression of NisT-type transporters is regulated together with the other genes of the biosynthetic loci. About half of the loci analysed here contain a TCS consisting of a two-transmembrane-helix histidine kinase with an extracellular sensory domain and an OmpR-like response regulator (Table 1, Fig. 4B). This type of histidine kinases has been termed ‘prototypical periplasmic sensing’ (Mascher et al., 2006). However, regulation of NisT-type transporters by periplasmic sensing TCSs is only found in loci that also contain a LanFEG-type transporter, for example the nisin biosynthesis locus that is regulated by NisRK (Kuipers et al., 1995), and is therefore discussed in the following section. None of the NisT-type transporters was found associated with a peptide quorum sensor TCS, although again the small size of the data set should be kept in mind.
As described for the SunT-type transporters, the NisT-type systems were also commonly found associated with an XRE family transcriptional regulator (Table 1, Fig. 4B). One example for this is LasX, which is required for the expression of the lactocin S biosynthetic operon lasA–W of Lactobacillus sakei (Skaugen et al., 2002). The signal leading to activation of las operon expression by LasX is not known, but it does not appear to be autoinduction by lactocin S, because expression is unaltered in a lasT mutant, which should be defective in AMP export (Skaugen et al., 2002). The biosynthetic locus for mutacin I in S. mutans is also positively regulated by an XRE-type regulator, MutR, which is encoded upstream of the mutA operon, and again the signal for MutR activation is not known (Kreth et al., 2004). Mutacin I production is furthermore subject to regulation via quorum sensing, resulting in AMP production during growth at high cell densities. In contrast to the AMP Smb described above, this regulation does not occur via the ComCDE system, but instead via autoinducer-2 (AI-2) signalling: the AI-2 synthase LuxS somehow represses transcription of irvA, which itself encodes a transcriptional repressor of mutR and the mutA operon, which inhibits mutacin I biosynthesis (Merritt et al., 2005).
Expression of the genes for gallidermin and epidermin export, gdmT, gdmH, epiT and epiH, is positively regulated by the DNA-binding protein EpiQ that also regulates the remaining biosynthetic genes (Peschel et al., 1993; 1997). EpiQ does not belong to the XRE family of transcriptional regulators, but instead contains a DNA-binding domain with similarity to OmpR-like response regulators (Peschel et al., 1993). How EpiQ is activated, for example if it requires phosphorylation by a histidine kinase (although a typical receiver domain is missing), or if it is acting as a stand-alone regulator is not known.
Nearly all transporters of this type characterized to date are involved in self-protection of lantibiotic producing strains, and accordingly all recognize a lantibiotic substrate (Table 2). Interestingly, however, only about a third of LanFEG-type transporters included in the phylogenetic analysis above were found to be associated with lantibiotic biosynthesis genes (Table 1), suggesting that their function is not restricted to self-protection. Furthermore, 24% were neighboured by transposases or insertion sequences, indicating a high degree of genetic mobility (Table 1). As an example, the best-characterized biosynthetic locus, that of nisin, is located on a conjugative transposon (Dodd et al., 1990). Consistently with their role in self-protection, most LanFEG transporters studied to date have a very narrow substrate range, providing resistance only against the produced lantibiotic or structurally very similar peptides [e.g. gallidermin and epidermin but not nisin are transported by EpiFEG (Otto et al., 1998)]. In contrast, the CprABC transporter (for cationic AMPresistance) of Clostridium difficile, which is not associated with any obvious biosynthesis genes, has a broader substrate range providing resistance against both nisin and gallidermin (McBride and Sonenshein, 2010). Although more such stand-alone transporters will have to be investigated to draw any general conclusions, it is conceivable that in systems not associated with biosynthetic loci a more relaxed specificity is advantageous for protection against a range of lantibiotics.
The direction of substrate transport has been determined for several LanFEG-type transporters as a movement of the lantibiotic from the cytoplasmic membrane to the culture supernatant, resulting in a shift of the distribution equilibrium between target-associated and free peptide (Otto et al., 1998; Stein et al., 2003; 2005; Okuda et al., 2008; 2010) (Fig. 4C). However, the question arises as to how transport can be a resistance mechanism against cell wall-active AMPs at all: the cellular targets (UPP, lipid II etc.) of these compounds are located on the surface of the cytoplasmic membrane. Thus, at first sight, the AMP will simply re-associate with its target in a time limited only by the rate of diffusion. For LanFEG-type transporters, the answer may be found in the high co-occurrence (78%) with so-called immunity proteins (Table 1, Fig. 4C). Two types of such proteins have been described. LanI-type proteins are tethered to the membrane surface via an N-terminal lipoprotein anchor, while LanH-type proteins contain three transmembrane helices with the N-terminus located intracellularly. The high number of membrane-associated hypothetical proteins found in the neighbourhood of LanFEG-type transporters (Table 1) suggests that the co-occurrence of these transporters with immunity proteins has even been underestimated. Both types of immunity proteins appear to function in concert with their LanFEG-type transporters as described for well-understood example systems below.
Self-protection against nisin in producer strains of Lactococcus lactis is mediated by the transporter NisFEG and the LanI-type immunity protein NisI. Each system by itself can confer some level of resistance, yet full resistance is only achieved in the presence of both (Stein et al., 2003). Similar results have been obtained for self-protection of B. subtilis against subtilin by SpaFEG and SpaI (Stein et al., 2005). There is as yet no experimental evidence to unequivocally show whether full resistance is due to a cooperative action of transporter and immunity protein or whether both act independently. Stein and colleagues reported that the resistance levels provided by NisFEG and NisI are additive, arguing for the latter mechanism (Stein et al., 2003). Another study showed that NisI is present as a membrane-anchored protein but additionally exists in a lipid-free form that is also able to bind nisin (Takala et al., 2004). The lipid-free form mediated only a low level of resistance in the absence of NisFEG, but this was significantly increased in the presence of the transporter, suggesting a cooperative mode of action (Takala et al., 2004). The authors propose a mechanism of cooperativity where the NisFEG transporter removes nisin from the membrane producing a high local concentration of free nisin, which is then bound by lipid-free NisI and moved away from the cell by diffusion (Takala et al., 2004).
Resistance of L. lactis against the two-peptide lantibiotic lacticin 3147 is mediated by the transporter LtnFE (missing the second permease gene) and the immunity protein LtnI (Draper et al., 2009). LtnFE or LtnI alone each provided some degree of protection, but together mediated higher resistance than the additive contribution of each system, again arguing for cooperativity (Draper et al., 2009). This effect was only observed for the two-peptide Ltnα:Ltnβ complex of lacticin 3147, whereas LntI and LtnFE acted independently against the individual Ltnα or Ltnβ peptides (Draper et al., 2009). Interestingly, orthologues of both systems, e.g. an LtnI homologue from B. licheniformis and an LtnFE homologue from Enterococcus faecium could also impart lacticin 3147 immunity to L. lactis and also showed synergistic effects upon coexpression, indicating that cooperativity is a general trait in such systems (Draper et al., 2009).
Protection of S. warneri against nukacin ISK-1 is provided by the NukFEG transporter and the LanH-type immunity protein NukH. Heterologous studies in L. lactis showed that, as with the LanFEG–LanI systems, NukFEG and NukH together mediated higher levels of resistance than the additive effect from expressing either determinant by itself (Aso et al., 2005). Individual expression of NukFEG imparted a stronger resistance than expression of NukH alone, showing that the transporter was the major resistance determinant (Aso et al., 2005). A more detailed analysis of the resistance mechanism showed that L. lactis expressing NukH bound more nukacin than a control strain and that coexpression of NukFEGH resulted in an energy-dependent decrease in cell-associated nukacin (Okuda et al., 2008). This led the authors to the hypothesis that NukH captures nukacin and this NukH-bound substrate is then transported to the extracellular space by NukFEG, implying that NukH functions as a substrate-binding protein for the transporter (Okuda et al., 2008). Of note, for gene annotation and functional studies care should be taken that these LanH proteins are not confused with homologues of GdmH. The former have three transmembrane helices and are functionally associated with LanFEG-type transporters, while the latter have six transmembrane helices and act as accessory proteins for NisT-type transporters as described above.
While most ABC transporters contain a so-called Q-loop motif in their ATPase domain, Okuda and colleagues reported that LanFEG-type transporters appear to possess a conserved glutamate (E) instead of the glutamine (Q) residue (Okuda et al., 2010). Mutation of this E-loop motif of the nukacin transporter NukFEG to the conventional Q-loop however only had minor effects on immunity (Okuda et al., 2010). An analysis of the multiple sequence alignment of ATPases generated during the phylogenetic analysis presented in Fig. 1A showed that indeed all LanFEG-type transporters contained a glutamate at the position corresponding to position 85 in NukF. Furthermore, all BcrAB-type transporters also contained this residue (position 84 in B. licheniformis BcrA), while the NisT-, SunT- and BceAB-type transporters possess the conventional Q-loop motif (not shown), again supporting the common evolutionary origin of LanFEG and BcrAB transporters.
Nearly three quarters of the LanFEG-type transporters analysed here were found associated with a TCS containing a prototypical periplasmic sensing histidine kinase (Table 1, Figs 2 and 4C). The best-characterized examples for such TCSs, NisRK and SpaRK regulating biosynthesis and immunity for nisin and subtilin, respectively, have been previously reviewed in detail (Kleerebezem, 2004). Therefore I will here just give a brief summary on their function.
As with many lantibiotics, production of nisin A is autoinduced in response to extracellular nisin A, nisin Z or nisin U (but not other lantibiotics) (Kuipers et al., 1995; de Ruyter et al., 1996; Ra et al., 1996; Wirawan et al., 2006). This induction is highly sensitive, requiring only 30 pM nisin corresponding to about five molecules per cell and is triggered by nisin binding to the sensor kinase NisK (Kuipers et al., 1995). Interestingly, the observed cross-induction between different variants of nisin can occur even between L. lactis (nisin A producer) and Streptococcus uberis (nisin U producer), showing that AMPs can function in quorum sensing even between different bacterial genera (Wirawan et al., 2006). Another example for inter-species communication is the cross-induction of salivaricin A and A1 production by Streptococcus salivarius and S. pyogenes respectively (Upton et al., 2001). NisRK regulates nisin production from two promoters, PnisA (nisin production and nisI) and PnisF (nisFEG), whereas expression of the nisRK structural genes is constitutive (de Ruyter et al., 1996; Ra et al., 1996). Regulation of subtilin biosynthesis by SpaRK is organized in a similar fashion, but expression of all immunity determinants, i.e. SpaI and SpaFEG, is regulated from a common promoter, PspaI (Klein and Entian, 1994; Stein et al., 2002b).
The mersacidin biosynthesis locus of Bacillus sp. HIL Y-85 contains a dedicated TCS, MrsR2–MrsK2, that regulates expression of the immunity transporter, MrsFGE, but not of mersacidin production (Guder et al., 2002). Instead, production (but not immunity) is positively controlled by the orphan response regulator MrsR1. It is not known whether MrsR1 requires phosphorylation for activation and, if so, which is its cognate sensor kinase, but a potential role for MrsK2 in MrsR1 regulation was excluded (Guder et al., 2002).
An exception to this mode of regulation is the lacticin 3147 self-resistance operon ltnRIFE. It is not controlled by a TCS but instead by the XRE family transcriptional regulator LtnR. Mutants in ltnR were found to be hyperresistant, while overexpression of the regulator caused increased sensitivity to lacticin 3147, showing that LtnR acts as a repressor of ltnRIFE (McAuliffe et al., 2001). The lacticin 3147 biosynthesis operon is transcribed divergently from the resistance genes and appears to be constitutively expressed (McAuliffe et al., 2001).
BceAB-like transporters mediate resistance against AMPs (Table 2) but are not found associated with biosynthetic loci (Table 1). The only exception to this is the SalXY transporter that is part of the salivaricin A biosynthesis locus (Upton et al., 2001). For most BceAB-type transporters several substrates have been identified, and often one transporter will recognize structurally very different AMPs (summarized in Gebhard and Mascher, 2011). Substrates for BceAB-type transporters not only include lantibiotics, class II bacteriocins or bacitracin, but also glyocpeptides, β-lactam antibiotics and defensins and cathelicidins of mammalian origin (Table 2). In that regard BceAB transporters differ markedly from the other resistance transporters, which are more restricted in their substrate range and transport either bacitracin (BcrAB group) or lantibiotics (LanFEG group) (Table 2). The molecular mechanisms underlying this surprisingly broad substrate range of BceAB-type transporters are not understood, as discussed below.
Genomic context analyses show no significant co-occurrence of genes with one exception: the vast majority of BceAB-type transporters are encoded next to a TCSs where the histidine kinases contain two transmembrane helices without an extracellular domain, so-called ‘intramembrane-sensing’ histidine kinases (Table 1, Fig. 4D) (Mascher, 2006; Dintner et al., 2011). As was first described for the eponymous bacitracin-resistance transporter BceAB of B. subtilis, these transporters are not only responsible for resistance but are also required for signalling: the associated sensor kinase BceS was found to be unable to detect bacitracin in the absence of BceAB, which led to the proposition that the transporter contains the actual sensory domain of the system (Bernard et al., 2007; Rietkötter et al., 2008). Since then, the same functional relationship between transporter and TCS has been reported for a number of related modules (Ouyang et al., 2010; Hiron et al., 2011; Staroń et al., 2011; Falord et al., 2012). Furthermore, BceAB-like transporters and BceRS-like TCSs were found to have co-evolved, supporting the tight functional link between these systems (Dintner et al., 2011).
In contrast to the previously described transporter groups, the direction of substrate translocation by BceAB-type transporters is not yet known. It has been proposed that these systems may act as importers, removing the AMP from its site of action (i.e. the cell surface) followed by intracellular inactivation through degradation (Rietkötter et al., 2008; Hiron et al., 2011). Yet it is equally possible that the transport mechanism is more similar to that of LanFEG-type transporters, removing the AMP from the cell membrane to the extracellular space. However, no common association of BceAB-type transporters with any potential auxiliary proteins such as the LanI/LanH proteins was found (Table 1) (Dintner et al., 2011). The distinguishing feature of BceAB-type transporters is their large extracellular domain (ECD), and it could be speculated that this domain might play a role akin to that of LanI or LanH proteins, but to date no experimental data are available to support such a mode of action.
Substrate binding by BceAB-type transporters appears to occur via their ECD, which was shown by a domain-swap experiment between two such transporters from Staphylococcus aureus. A derivative of the colistin transporter VraFG that contained the ECD of the bacitracin transporter VraDE provided resistance against bacitracin, but no longer against its original substrate, colistin (Hiron et al., 2011). Despite this, it is still unclear how the different substrates are recognized and distinguished. As mentioned above, most BceAB-type systems transport a range of AMPs, and the substrates for a single transporter can differ considerably in their structure while at the same time other structurally related compounds are not recognized. For example PsdAB of B. subtilis transports the lantibiotic actagardine, but not the similar mersacidin. The same transporter also recognizes one lipodepsipeptide, enduracidin, but not another, ramoplanin (Staroń et al., 2011). Further, while VraDE of S. aureus can detoxify both nisin and bacitracin, B. subtilis employs two separate transporters, PsdAB and BceAB, respectively, for resistance against both compounds (Ohki et al., 2003b; Hiron et al., 2011; Staroń et al., 2011). An in-depth phylogenetic analysis of BceAB-like transporters further showed that no correlation existed between phylogenetic group and substrate range (Dintner et al., 2011).
Another interesting aspect of BceAB-type transporters is their role in signal transduction. Again, the mechanism of communication between the transporter and TCS is not known, but parallel phylogenetic analyses (Dintner et al., 2011) and bacterial two-hybrid assays (Falord et al., 2012) suggest a direct interaction between the transport permease and histidine kinase. Importantly, the transport process itself is required for the signalling process, because mutations that prevent ATP hydrolysis and thus substrate translocation abolish signal transduction (Rietkötter et al., 2008). This requirement for ATP hydrolysis even holds true for the dedicated sensing transporters described below, implying that they must be capable of transport, even though they do not mediate resistance (Hiron et al., 2011). It therefore appears that the ability to transport an AMP does not necessarily equate to providing resistance. An explanation for this may be that the transport rate of the sensing transporters is too low to counteract the inhibitory effects of the AMP while it is still sufficient for signalling, but experimental evidence for this is missing.
All BceAB-type transporters studied to date are regulated by a TCS that is most commonly encoded in an adjacent operon (Fig. 2) (Joseph et al., 2002; Dintner et al., 2011). The histidine kinases of these systems belong to the group of intramembrane-sensing kinases (Mascher, 2006; Mascher et al., 2006), and the response regulators to the OmpR group (Fig. 4D). The consensus binding sequence for the response regulators associated with BceAB-type transporters has been identified as TNACA-N4-TGTAA with an AT-rich central 4 nt spacer (de Been et al., 2008; Dintner et al., 2011). Expression of the transporter is induced in the presence of sublethal concentrations of its substrate AMP, and this requires the presence of both the TCS and the transporter itself, as described above. In some cases, a dedicated sensing transporter is present to aid in induction of a second transporter that in turn mediates the actual resistance. Such a scenario is for example found in S. aureus, where the TCS BraRS (also called NsaRS or BceRS) together with its sensing transporter BraDE (also called NsaAB or BceAB) induces expression of the transporter VraDE that provides resistance against bacitracin or nisin (Pietiäinen et al., 2009; Hiron et al., 2011; Kolar et al., 2011; Yoshida et al., 2011). Additionally, S. aureus contains a third BceAB-like transporter, VraFG, that acts as a sensor for the TCS GraRS (also termed ApsRS) (Li et al., 2007a; Meehl et al., 2007; Falord et al., 2011; 2012). The TCS GraRS is itself unusual in that it requires an accessory protein, GraX, in order to induce gene expression and has therefore been termed a three-component system (Li et al., 2007b; Falord et al., 2012). The exact role of GraX is not understood and no other intramembrane-sensing TCS studied to date possesses such an accessory protein.
Many Firmicutes bacteria contain several, in some cases up to six paralogues of BceAB-like transporters (Dintner et al., 2011). Their respective roles in resistance and regulatory interplay have been studied in most detail for B. subtilis and S. aureus, and one striking difference between the two organisms is the degree of cross-regulation among the paralogous modules. The three systems of B. subtilis (BceRS-AB, PsdRS-AB and YxdJK-LM) are well insulated from each other, with only a minor degree of cross-phosphorylation observed from the histidine kinase BceS to the response regulator PsdR (Rietkötter et al., 2008). In contrast, in S. aureus the transporter VraDE, which does not possess its own TCS, is regulated by the TCS BraRS and the sensing transporter BraDE, and according to some studies also by the TCS GraRS and its sensing transporter VraFG (Li et al., 2007a; Gebhard and Mascher, 2011; Hiron et al., 2011). This difference in regulatory complexity between B. subtilis and S. aureus is also reflected in the respective regulons of the involved TCSs. In B. subtilis the associated BceAB-type transporter is the sole target of regulation (Ohki et al., 2003b; Rietkötter et al., 2008; Staroń et al., 2011), while GraRS and BraRS of S. aureus additionally regulate genes of the cell wall stress response, including mprF, the dtlABCD operon (GraRS) and genes for cell wall biosynthesis (BraRS) (Kolar et al., 2011; Falord et al., 2012). Induction of these additional genes will doubtlessly contribute to resistance against cell wall-active antibiotics, and it therefore remains to be determined which of the broad range of inducing substrates, particularly of the three staphylococcal systems (Gebhard and Mascher, 2011), are actually detoxified by the transporter itself or simply counteracted by the general changes in the cell envelope.
A similar extended regulon might also exist in S. mutans, because in addition to the transporter MbrAB (also termed BceAB) three hypothetical proteins were identified as potential targets for the TCS MbrCD (also termed BceRS) from an in silico search for the response regulator binding consensus (Ouyang et al., 2010). In Listeria monocytogenes, the TCS VirRS regulates expression of a BceAB-like transporter, AnrAB, as well as of the dlt operon and mprF (Mandin et al., 2005; Collins et al., 2010). The sensing transporter for VirRS has not yet been identified, but a likely candidate is the uncharacterized BceAB-like transporter Lmo1747-1746, which is encoded in the same locus as VirRS (Mandin et al., 2005; Dintner et al., 2011). An explanation for the much simpler regulation in B. subtilis may be the presence of several extracytoplasmic function (ECF) sigma factors, e.g. SigW or SigM, which regulate the more general cell wall stress genes like dlt and mprF (Mascher et al., 2007). S. aureus, S. mutans and L. monocytogenes all possess no or only a single ECF sigma factor (Jordan et al., 2008), and accordingly the TCSs associated with BceAB-like transporters may have evolved to control a broader regulon.
BcrAB- and YydIJ-type transporters
BcrAB-type transporters are also involved in resistance against AMPs, but they act specifically against the cyclic peptide bacitracin. The eponymous system is found in the bacitracin producer B. licheniformis ATCC10716, where it is encoded in the biosynthesis locus bacABC–bacRS–bcrABC and confers self-resistance (Podlesek et al., 1995; Neumüller et al., 2001). A second characterized example is BcrAB from Enterococcus faecalis AR01/DGVS, a clinical isolate that displayed high-level bacitracin resistance (Manson et al., 2004).
Both transporters are co-transcribed with genes encoding putative undecaprenyl-pyrophosphatases (UppP), bcrC in B. licheniformis and bcrD in E. faecalis. UppPs can mediate bacitracin resistance by dephosphorylation of the cellular target of the AMP, undecaprenyl-pyrophosphate (UPP). While a BcrC orthologue was indeed shown to confer resistance to B. subtilis (Cao and Helmann, 2002; Mascher et al., 2003; Ohki et al., 2003a), no function of BcrD in E. faecalis has been identified to date (Manson et al., 2004; Gauntlett et al., 2008). Additionally, both transporter operons are preceded by genes for transcriptional regulators, bacRS in B. licheniformis and bcrR in E. faecalis, which are discussed in detail below. Overall, analysis of the genomic context of these transporters showed that their loci are small and restricted to the transporter itself, often with genes for a regulatory system and a UppP (Table 1, Fig. 2), supporting their involvement in resistance only. Among environmental isolates of enterococci, bcrAB genes appear widely distributed (Matos et al., 2009), which suggests genetic mobility of the encoding regions. Indeed, efficient transfer of the bcr genes from the clinical E. faecalis isolate AR01/DGVS to the bacitracin-sensitive laboratory strain JH2-2 has been demonstrated (Manson et al., 2004).
The YydIJ-like transporters possess the same domain architecture as the BcrAB transporters, but form a distinct branch in the phylogenetic tree (Fig. 1). Only one example has been characterized to date, YydIJ of B. subtilis (Butcher et al., 2007). This transporter is encoded as part of the yydFGHIJ locus, which additionally contains genes for a small peptide (YydF) and its modification (YydG) and processing (YydH). YydF most likely has a cell envelope-disturbing effect, because it induces the LiaRS-mediated cell wall stress response in B. subtilis (Butcher et al., 2007). YydIJ may act as an exporter of YydF or alternatively in YydF self-resistance, as discussed below. The distribution of YydIJ-like transporters among Firmicutes appears to be rather restricted. The complete yydF–J locus was only found in some strains of B. subtilis, while Bacillus cereus possessed the transporter neighboured by a truncated yydH gene and several genes for transposases and an integrase. Other loci contained only the transporter (Table 1, Fig. 2). As with the BcrAB-type transporters, acquisition of yyd loci by horizontal gene transfer has been discussed (Butcher et al., 2007).
To date, no information is available on the mechanism of AMP transport by BcrAB-type transporters. As mentioned previously, for BceAB-type transporters the possibility of AMP import followed by enzymatic inactivation inside the cell has been proposed (Rietkötter et al., 2008; Hiron et al., 2011). Experimental evidence, however, is available only for the LanFEG-type transporters, which remove their lantibiotic substrates from the cytoplasmic membrane and release it to the supernatant as described above (Otto et al., 1998; Okuda et al., 2008; 2010). The phylogenetic tree of AMP transporters (Fig. 1) shows that the BcrAB-type transporters are ancestral to the LanFEG transporters, thus a similar transport mechanism in both groups should be assumed.
Some indirect evidence for such a ‘hydrophobic vacuum cleaner’ mechanism (Chang, 2003) of BcrAB-type transporters may be found in the transcriptional response of the bcrR–bcrABD locus of E. faecalis to bacitracin. In the presence of the BcrAB transporter, the sensitivity of bacitracin detection by the regulator, BcrR, is greatly decreased (Gauntlett et al., 2008). These data were interpreted as removal of bacitracin by BcrAB from the site of detection by BcrR (Gauntlett et al., 2008), which is most likely at or within the cytoplasmic membrane, because BcrR has no extracellular domains (Gebhard et al., 2009). Similarly, cells of B. subtilis lacking YydIJ were found to display a stronger YydF-induced Lia response compared with YydIJ-positive strains (Butcher et al., 2007). The authors discussed this as caused by accumulation of the YydF peptide inside the cell and therefore propose YydIJ to act in peptide export (Butcher et al., 2007). In light of the data available now for LanFEG- and BcrAB-type transporters, an alternative explanation could be that YydIJ confers self-resistance by removing YydF from the membrane, and thus deletion of the transporter might increase the cell wall stress detected by the Lia system. Such a mechanism would also be consistent with the finding that cells lacking YydIJ can still induce the Lia response in a neighbouring yyd-deleted strain (Butcher et al., 2007).
Still the question remains how the removal of an AMP from the cell membrane to the culture supernatant can confer resistance. For the bacitracin-resistance transporters, the actual resistance mechanism may be tightly linked to the activity of UppPs. As shown in Table 1, 77% of the analysed BcrAB-type transporters are associated with UppP-encoding genes located downstream of the transporter (Fig. 2), and it is therefore likely that all three genes form an operon. Upon induction of this operon, the shift in the distribution equilibrium between cell-associated and free bacitracin mediated by the transporter may be sufficient to enable rapid dephosphorylation of UPP by the UppP (Fig. 4E). Because bacitracin has poor affinity for monophosphates (Storm and Strominger, 1973), resistance is thus ensured. This model is supported by an earlier report that all three genes (bcrABC) of B. licheniformis were necessary to confer bacitracin resistance to the heterologous host B. subtilis, which led to the original assumption that BcrC was a third component of the transporter (Podlesek et al., 1995). In E. faecalis the UppP BcrD was shown to be dispensable for bacitracin resistance, but it could not be ruled out that a second paralogue was encoded elsewhere on the genome (Gauntlett et al., 2008).
The majority of BcrAB transporter encoding loci analysed here contained the same type of TCS as found with BceAB-like transporters, i.e. possessing an intramembrane-sensing histidine kinase and an OmpR family response regulator (Figs 2 and 4E, Table 1). For the bacitracin biosynthesis locus from B. licheniformis, the TCS BacRS was shown to regulate expression of bcrABC, but not of bacitracin biosynthesis itself (Neumüller et al., 2001). Experimental data suggest that BacR acts as a negative regulator, repressing transcription of bcrABC in the absence of bacitracin (Neumüller et al., 2001), which is in contrast to other Bce-like TCSs studied to date that function by positive regulation (Ohki et al., 2003b; Li et al., 2007b; Collins et al., 2010; Hiron et al., 2011). It is not known how bacitracin is sensed by BacS. The intramembrane-sensing histidine kinases associated with BceAB-type transporters are known to strictly depend on their corresponding transporter for stimulus detection (Rietkötter et al., 2008; Gebhard and Mascher, 2011; Hiron et al., 2011), and it is therefore questionable whether the structurally very similar BacS is able to perceive a stimulus directly. While it is possible that BcrAB-type transporters may also be involved in the signalling process, it has been shown for E. faecalis BcrAB that it is not required for sensing (Gauntlett et al., 2008). It should however be noted that this particular transporter is not regulated by a TCS.
One-third of the BcrAB-type transporters were found to be associated with transcriptional regulators of the XRE family, often containing four predicted transmembrane helices in their C-terminal domain (Fig. 4E). One example for this, BcrR of E. faecalis, has been studied in detail, and it was shown that it is indeed a membrane-bound DNA-binding protein (Gauntlett et al., 2008). BcrR binds bacitracin directly, most likely at or within the cytoplasmic membrane as mentioned above and subsequently induces transcription from a single promoter upstream of the bcrABD operon (Gauntlett et al., 2008; Gebhard et al., 2009). The target promoter contains two inverted repeat sequences that are both necessary for induction, and it is therefore thought that it is bound by two dimers of BcrR. DNA binding was shown to be independent of bacitracin, suggesting that the promoter is constitutively occupied by BcrR, which may be important for a membrane protein in order to find its target sequence (Gebhard et al., 2009). It has been proposed that binding of bacitracin by BcrR leads to a change in the interaction between individual BcrR monomers or between BcrR and RNA polymerase (Gebhard et al., 2009), but the exact mechanism for transcription activation remains to be elucidated.
The yydIJ operon of B. subtilis is neighboured by a convergently transcribed gene, yydK, which encodes a putative GntR-type regulator (Fig. 2); however, its function is unknown.
This review to my knowledge presents the first study with a focus on the different types of AMP transporters found in Firmicutes bacteria without taking its primary motivation from AMP biosynthesis or resistance. Importantly, the genomic context analysis showed that even SunT-type and LanFEG-type transporters, which are normally considered tightly linked to lantibiotic production, are not always associated with biosynthetic genes (Table 1). To facilitate an independent comparison of the AMP transporters themselves, a new classification system based on domain architecture was developed. ABC transporters are often categorized according to the TCDB established by the Saier group and available online (http://www.tcdb.org) (Saier et al., 2009). While this database also includes the AMP transporters reviewed here, their classification appears in parts somewhat inconsistent and only very few examples are included (two to seven per family). The TCDB classification matches the one proposed here for the LanFEG-type transporters (TCDB Peptide-5 Exporters) and the BceAB-type transporters (TCDB Peptide-7 Exporters). However, SunT-type transporters are found in the TCDB families Peptide-1 Exporters and Peptide-2 Exporters, while the NisT group is found in the Peptide-1 Exporters and Peptide-4 Exporters families. The domain-based classification presented here appears more meaningful, because it accurately reflects the phylogeny of the transporter groups, for example showing that the BcrAB-type transporters are ancestral to the LanFEG group (Fig. 1). Furthermore, genomic context analyses and a review of the available literature based on this new classification revealed distinguishing as well as common characteristics regarding transport mechanism and regulation, which can be attributed to each group of transporters.
One striking observation is that the majority of AMP transporters from Firmicutes appear to function together with accessory or auxiliary proteins. This is most obvious for the SunT group, where nearly half of the transporters analysed possessed a dedicated accessory protein, and functional evidence was available on their requirement for peptide export. While the genomic neighbourhood of the analysed NisT-type transporters did not appear to contain homologous proteins, the example of the gallidermin transporter GdmT, which depends upon the unrelated accessory protein GdmH for transport, shows that at least some NisT-type transporters cannot function by themselves. This is perhaps not too surprising, given that SunT- and NisT-type systems form a common evolutionary group in the phylogenetic tree of AMP transporters (Fig. 1). It will be interesting to see whether more GdmH-like proteins are found as more NisT-like transporters are identified. Among the resistance transporters, no additional proteins are directly necessary for the transport process, but the high degree of co-occurrence of LanFEG-type transporters with LanI or LanH immunity proteins, and of BcrAB-type transporters with UppP-encoding genes nevertheless suggests a tight functional link (Fig. 4C and E). As discussed above, these proteins may be an important facet of the immunity mechanism: if the transporter's role is to remove cell membrane-associated AMPs and release them to the culture supernatant, a second protein may be needed to ensure the AMP cannot re-associate with the cell, either by binding it (LanI or LanH proteins) or by removing its cellular target (UppP proteins). The lack of association of BceAB-type transporters with any additional genes may hint at a differing transport mechanism. Alternatively such a function might be implied for their large extracellular domain, but as no experimental evidence is available to date, this remains speculation.
Association with distinct regulatory systems
Genomic context analysis showed that each group of transporters was found to be associated with a distinct type of regulatory system (Table 1, Fig. 4). SunT-type transporters are commonly regulated by TCSs with a peptide quorum sensor kinase and LytR family response regulator; LanFEG-type transporters are regulated by TCSs with a prototypical periplasmic sensing kinase and an OmpR family response regulator; BcrAB and BceAB-type transporters are regulated by TCSs with an intramembrane-sensing histidine kinase and again an OmpR family response regulator. NisT-type transporters were often found associated with the same type of TCS as the LanFEG transporters, but this is most likely due to the fact that these genetic loci in all cases also contained a LanFEG-type transporter (Table 1). Where no TCS was found in the neighbourhood of a transporter, the most common regulatory genes encoded XRE family transcriptional regulators (Table 1, Fig. 4). Such regulators were found with transporters of the SunT, NisT, BcrAB and LanFEG groups, and in several cases it has been shown experimentally that they indeed regulate expression of their respective transporter.
Parallels to quorum sensing
All three types of TCSs as well as the membrane-bound XRE regulators are activated in the presence of the transporter's substrate AMP, with the exception of some SunT-associated systems that are induced in response to a dedicated pheromone. Particularly in the context of biosynthesis loci for AMPs this mode of regulation is highly similar to quorum sensing mechanisms, and such connections between AMP production and cell–cell communication have been drawn repeatedly (Upton et al., 2001; Kleerebezem, 2004; Wirawan et al., 2006). The question that is usually asked next is whether AMPs have primarily evolved as a means for inter-species warfare or rather as a means for communication within bacterial populations. The strong correlation between the type of transporter and its associated regulatory system described here may open up new and promising avenues for investigations aimed at finding the answer.
Despite the considerable knowledge gained on AMP transporters to date, several questions remain to be answered. One central point is the direction of substrate transport by the BcrAB- and BceAB-type resistance transporters. Do they function the same way as LanFEG transporters, and if so, do the associated UppPs or other as yet unidentified proteins contribute to providing maximal resistance? A second important question concerns the site of AMP binding, not only by the transporters but also by the sensory proteins. For BceAB-type transporters, the ECD is the most likely candidate region, but where do transporters without any obvious ligand-binding domains, i.e. LanFEG- and BcrAB-type systems, bind their substrates? The same can be asked of regulatory proteins such as the membrane-bound XRE regulators or the intramembrane-sensing histidine kinases associated with BcrAB-like transporters. And finally, what forms the basis for substrate specificity? This question was posed above for the ECD of BceAB-type transporters, but also no information is available for example on how the periplasmic sensing histidine kinases found with LanFEG-type transporters distinguish between the different AMPs to provide specificity in signalling. No doubt future research will be aimed at shedding light on these aspects and continue to deepen our understanding on AMP transport in Firmicutes bacteria.
The author would like to thank Thorsten Mascher for stimulating discussions before and during preparation of this review and for critical reading of the manuscript. I also extend my apologies to the authors of the numerous publications on AMP transporters that were not cited due to space restrictions. Work in the author's lab is supported by grants from the Deutsche Forschungsgesellschaft (GE2164/3-1) and the Fonds der Chemischen Industrie.