Lantibiotics are post-translationally modified antimicrobial peptides which are active at nanomolar concentrations. Some lantibiotics have been shown to function by targeting lipid II, the essential precursor of cell wall biosynthesis. Given that lantibiotics are ribosomally synthesized and amenable to site-directed mutagenesis, they have the potential to serve as biological templates for the production of novel peptides with improved functionalities. However, if a rational approach to novel lantibiotic design is to be adopted, an appreciation of the roles of each individual amino acid (and each domain) is required. To date no lantibiotic has been subjected to such rigorous analysis. To address this issue we have carried out complete scanning mutagenesis of each of the 59 amino acids in lacticin 3147, a two-component lantibiotic which acts through the synergistic activity of the peptides LtnA1 (30 amino acids) and LtnA2 (29 amino acids). All mutations were performed in situ in the native 60kb plasmid, pMRC01. A number of mutations resulted in the elimination of detectable bioactivity and seem to represent an invariable core within these and related peptides. Significantly however, of the 59 amino acids, at least 36 can be changed without resulting in a complete loss of activity. Many of these are clustered to form variable domains within the peptides. The information generated in this study represents a blue-print that will be critical for the rational design of lantibiotic-based antimicrobial compounds.
Lantibiotics are ribosomally synthesized, post-translationally modified, antimicrobial peptides (McAuliffe et al., 2001; Pag and Sahl, 2002; Chatterjee et al., 2005; Cotter et al., 2005a). There are at least 11 lantibiotic subclasses (nisin, epidermin, streptin, Pep5, lacticin 481, mersacidin, LtnA2, cytolysin, lacticin S, cinnamycin and sublancin; Cotter et al., 2005b). Interest in these peptides has been stimulated by the observation that a number of these molecules inhibit drug-resistant Gram-positive bacteria often referred to as ‘superbugs’, including methicillin-resistant Staphylococcus aureus (MRSA) (Galvin et al., 1999; Brumfitt et al., 2002; Kruszewska et al., 2004). Moreover, it has been demonstrated that at least some lantibiotics are active at single nanomolar concentrations through a dual mechanism of action, which is facilitated by binding of lipid II, the essential precursor of bacterial cell wall peptidoglycan synthesis (Breukink et al., 1999; Wiedemann et al., 2001; 2006a;Breukink and de Kruijff, 2006). The fact that lantibiotics are gene-encoded makes them amenable to classical site-directed mutagenesis approaches to drug design. In theory, rational (or random) strategies to bioengineer lantibiotic structural genes could yield huge numbers of derivatives. Despite this advantage the large scale creation of rationally designed lantibiotic-derived analogues with potential medical applications has not yet been described. In particular, rational lantibiotic design is hindered by the lack of detailed peptide ‘maps’ indicating which residues are immutable and which can be altered. The identification of essential and variable residues is complicated by the fact that lantibiotics undergo extensive post-translational modifications that result in the formation of lanthionine and β-methyl lanthionine bridges – the structures synonymous with lantibiotics – dehydroalanine (dha) and dehydrobutyrine (dhb) as well as a variety of other residues such as d-alanines (rare even among lantibiotics). Lanthionine bridges are formed through a two-step post-translational modification that is initiated by the dehydration of serine and threonine residues to dha and dhb respectively. The subsequent reaction of these modified amino acids with intrapeptide cysteines results in the formation of lanthionine (Ala-S-Ala; in the case of dha) or β-methyl-lanthionine (Abu-S-Ala; in the case of dhb) bridges. Thus, the alteration of one residue (ser, thr or cys) will have consequences for the fate of the other residue involved in bridge formation. A number of other, less typical, modifications occur in a variety of lantibiotics (for review see Jack and Jung, 2000; Chatterjee et al., 2005), including the stereospecific hydrogenation of dha to d-alanine (Skaugen et al., 1994; Ryan et al., 1999; Cotter et al., 2005c).
Although site-directed mutagenesis of lantibiotics was first described some years ago (Dodd et al., 1992; Kuipers et al., 1992; Liu and Hansen, 1992), nisin is the only peptide for which more than eight bioengineered derivatives have been reported. A close examination of the literature indicates that less than half of the 34 amino acids in the nisin structural propeptide have been subjected to mutagenesis (for review see Cotter et al., 2005b). In the absence of such fundamental information, it will be difficult to rationally design lantibiotic-derived antibiotics. We have chosen the two-peptide lantibiotic lacticin 3147 as a suitable target for complete analysis for several reasons. First, it is active at low nanomolar concentrations via interaction with lipid II (most likely at the MurNAc-PP moiety) (Morgan et al., 2005; Wiedemann et al., 2006a); second, it inhibits all Gram-positive pathogens tested [including MRSA and vancomycin-resistant enterococci, VRE (Galvin et al., 1999)]; third, it represents a relatively rare class of lantibiotic which requires two peptides for optimal activity and lastly, it has been successfully bioengineered previously (Cotter et al., 2005c; O'Connor et al., submitted). Like other two-peptide lantibiotics, lacticin 3147 is active through the synergistic activity of two peptides, LtnA1 (30 amino acids) and LtnA2 (29 amino acids). There are already encouraging signs that the two peptides are tolerant of change as each of the individual lacticin 3147 components can function in combination with the corresponding companion peptide from another two component lantibiotic, staphylococcin C55 (i.e. LtnA1/Sacβ and LtnA2/Sacα; O'Connor et al., submitted), despite having only 55% identity with respect to the A2/β components. This flexibility, coupled with the fact that the involvement of two peptides facilitates the examination of distinct functional domains in isolation, makes the two component lantibiotics excellent candidates for bioengineering.
A systematic mutagenesis of each of the 59 amino acids of the LtnA1 and LtnA2 propeptides was carried out. In each case the relevant amino acid was converted to an alanine (existing alanine residues were converted to glycine) and the impact of these alterations on overall bioactivity (i.e. the combined impact on production and specific activity) was ascertained. High-performance liquid chromatography (HPLC) and mass analysis was carried out in all cases and in nine instances the specific activity of the bioengineered peptide, alone and with its unmutated companion peptide, were determined. Using these technologies it was possible to identify apparently essential and variable residues/regions, representing a major step towards the rational design of novel lantibiotic-derived antimicrobials.
Identification of regions conserved among two component lantibiotics
Lantibiotics can be subgrouped on the basis of the amino acid sequence of their unmodified propeptide (Cotter et al., 2005b). All of the A1/α components of two-component lantibiotics identified to date, with the exception of cytolysin, cluster within the mersacidin subgroup of lantibiotics. Alignment of peptides within the mersacidin and lacticin 481 subgroups, which together constitute greater than 40% of all lantibiotics described to date, reveals a number of conserved residues, representing a GxxxxTxS/TxD/EC[3–10x]C motif, that may represent essential core amino acids in the structure–function relationship of these peptides (Fig. 1). The locations of the (β-methyl)lanthionine bridges within a number of these peptides has been ascertained (Kogler et al., 1991; van den Hooven et al., 1996; Krull et al., 2000; Martin et al., 2004) revealing that of the hyper-conserved residues, only the glycine and glutamate/aspartate residues are not involved in bridge formation. Of the mersacidin-like peptides, six represent the A1 (or α) component of a two-component lantibiotic (lacticin 3147, Dougherty et al., 1998; staphylococcin C55, Navaratna et al., 1998; plantaricin W, Holo et al., 2001; Smb,Yonezawa and Kuramitsu, 2005; BHT-A, Hyink et al., 2005 and haloduricin, Lawton et al. submitted). When the corresponding A2 (or β) peptides are aligned it is apparent that six residues are conserved within the C-terminus of all of these peptides (i.e. T/S[5–6x]S/TTxCS/TxxC) and that a further two are located in the majority (Fig. 1). It would thus appear that the β peptides consist of a variable N-terminus and a conserved C-terminus. If conservation of amino acids is indicative of an intolerance to change then the corollary, taken to its extreme, would indicate that within the lacticin 3147 peptides as many as 22 (LtnA1) and 21 (LtnA2) residues could be amenable to change. These figures, if validated, would make these peptides excellent candidates for bioengineering as part of rational drug design. To determine to what extent LtnA1 and LtnA2 are in fact amenable to change we initiated a complete scanning mutagenesis of the two propeptides in which each of the residues was converted to alanine (or glycine in situations where an alanine was naturally present).
Alanine scanning mutagenesis of LtnA1 and LtnA2
The presence of complex post-translational modifications in lantibiotics complicates the creation of synthetic peptide analogues and lantibiotics have yet to be efficiently produced by a Gram-negative host. For these reasons mutation of the lacticin 3147 peptides was performed on the original plasmid, pMRC01, in Lactococcus lactis MGMR. This involved a two-step process involving polymerase chain reaction (PCR)-based mutagenesis of these genes on a shuttle vector, followed by double crossover recombination to introduce the engineered gene in place of the corresponding wild-type version in the pMRCO1 plasmid. Identification of each mutant was facilitated by PCR through the use of ‘check’ primers which discriminated on the basis of their extreme 3′ sequence. Though relatively time-consuming (the creation of each individual mutant took at least 6 weeks if no problems were encountered), the use of this procedure, rather than one whereby the mutated gene was provided in trans, ensured that the sole alteration within the bioengineered strains was the replacement of a single amino acid. At the concentrations of lacticin 3147 peptides present in the supernatant of a producing culture, neither LtnA1 nor LtnA2 alone possess sufficient activity to bring about detectable growth inhibition (McAuliffe et al., 2000), and thus the presence of an unmutated companion gene is beneficial as it facilitates rapid assessment of the antimicrobial capacity of newly created derivatives. Bioactivity was assessed by well diffusion assays against a sensitive indicator L. lactis HP (Fig. 2). Steps were taken to ensure that the companion unmutated peptide was produced in all instances (Experimental procedures). Mass analysis of the mutated peptides was also carried out. In each case the peptides were stripped from the surface of the producing cell (a procedure which results in greater peptide yield and minimizes the presence of contaminating peptides from the growth medium) and purified/analysed by HPLC. Fractions were subjected to mass analysis to determine the mass change resulting from mutagenesis and to identify the peak corresponding to the mutated peptide. In the absence of detectable peaks, colonies of the relevant mutant strain were subjected to colony mass spectrometry (CMS) which facilitated mass analysis in a number of instances where mass analysis of HPLC fractions was unsuccessful. In all cases the masses determined were as predicted unless clearly stated otherwise (Table 1).
Table 1. Mass analysis of bioengineered peptides.
In all cases signal to noise ratios were greater than the 5:1 cut-off.
ND, not detected.
The goal of this process was to discriminate between residues that were variable and inviolate with respect to the production of active peptide (bioactivity). Thus, initially no attempt was made to distinguish between mutations that impact on production and those that impact on specific activity. As 59 mutants were involved in the initial scanning strategy, we will discuss the effect of changing each residue on bioactivity in discrete groups according to their predicted function, e.g. residues involved in lanthionine formation, those involved in other modifications, etc.
Mutagenesis of residues involved in (β-methyl)lanthionine formation
Such is the importance of (β-methyl)lanthionine bridges that in all previous studies on other lantibiotics their mutagenesis resulted in the dramatic reduction or elimination of antimicrobial activity (Bierbaum et al., 1994; 1996; Ottenwalder et al., 1995; Kuipers et al., 1996; Chen et al., 1998). It was thus unsurprising that 12 (of 14) mutations that resulted in the replacement of residues involved in 6 (of 7) lanthionine bridges in LtnA1/LtnA2 resulted in the elimination of bioactivity. Curiously, although undetectable following HPLC purification, CMS revealed the presence of small quantities of peptide of 3273 Da produced by both LtnA1:C19A and LtnA1:C25A. These masses are consistent with the presence of an additional dehydrated residue in both cases (Dha9 and Dhb22 respectively). In contrast to the other 6 (β-methyl)lanthionine bridges, bridge 1 of LtnA1 exhibited an unusual tolerance to change in that both LtnA1:C1A and LtnA1:S2A mutants retained a significant amount of bioactivity (45 and 30 AU ml−1, respectively, relative to 152 AU ml−1 for the parental control). LtnA1, mersacidin and possibly Sacα, are unusual in possessing the smallest possible (β-methyl)lanthionine bridge (i.e. a bridge between adjoining residues) which, in each case, is found at the N-terminus. This represents the first instance in which such a bridge has been altered and is also the first occasion that a bridge has been disrupted in which the contributing serine- or threonine-derived residue lies closer to the C-terminus than the partner cysteine. The mass of the LtnA1C1A peptide is reduced by 31.7 Da, indicating that in the absence of a partner cysteine, the lanthionine bridge formed by Cys1 and Ala2 is replaced by Ala1Dha2. LtnA1S2A has a mass of 3307.51 Da. Unfortunately it is not possible to predict precisely the modified N-terminal structure of this peptide on the basis of mass alone, as a 2 Da increase could reflect the presence of a free Cys1 or its formation of a MeLan with Dhb3 or Dhb5.
Mutagenesis of other modified residues
The other modified residues include three serine-derived d-alanines in LtnA1 and LtnA2 (LtnA1S7, LtnA2S9 and LtnA2S12), in addition to four Dhb's resulting from dehydration of threonines (LtnA1T3, LtnA1T5; LtnA2T2 and LtnA2T5) and one 2-oxybutyrate (2-06) resulting from the conversion of LtnA2:T1 to Dhb and its subsequent deamination to 2-ob. The conversion of the three d-alanines to other amino acids, including l-alanine, glycine, l-threonine and l-valine, has been described previously (Cotter et al., 2005c). The consequences of replacing Dha or Dhb residues in lantibiotics would seem to vary on a case by case basis (see Cotter et al., 2005b for a review). Although mutagenesis of the 5 relevant residues in LtnA1/LtnA2 does not eliminate activity, the drop in bioactivity as a consequence of replacement with alanine is variable [in order of importance for bioactivity LtnA1:T5A (7 AU ml−1), LtnA2:T2A (27 AU ml−1), LtnA2:T1A (40 AU ml−1), LtnA2:T5A (60 AU ml−1) and LtnA1:T3A (67.5 AU ml−1)]. Although, in general, mutagenesis of dehydrated amino acids resulted in reduced HPLC peak-size, LtnA1T5A and LtnA2T1A were particularly affected. Although these peptides could not be detected by HPLC there are trace levels of LtnA2:T1A peptide production, based on CMS analysis. The predicted consequence of a Dhb to Ala change is a 12 Da reduction in mass. This is in close agreement with that observed for the LtnA1T3A (3293.58), LtnA2T2A (2834.77) and LtnA2T5A (2834.98) peptides. The mass of LtnA2T1A was 2834.50, indicating a 2-oxybutyrate to Ala (−13 Da to 2834) conversion.
Mutagenesis of residues in close proximity to d -alanines
As mentioned earlier, possibly the most unusual feature of the lacticin 3147 peptide is the presence of d-alanine residues that arise as a consequence of the dehydration of serine followed by hydrogenation of the resultant Dha. While the three d-alanine residues have already been subjected to extensive mutagenesis (Cotter et al., 2005c), the replacement of the flanking residues is interesting as flanking residues have been shown to impact on the post-translational modification of lantibiotics (Kupke et al., 1995; Rink et al., 2005) and on the introduction of d-amino acids into eukaryotic peptides (Heck et al., 1996). Thus, it was interesting to observe that mass spectrometric analysis of LtnA2 peptides in which the relevant stretch of residues was altered (LtnA2A7G to LtnA2A13G) revealed that two, LtnA2I8A and LtnA2A13G, had masses that were 2 Da lighter than expected. Such a difference indicates the presence of a Dha rather than a d-ala at residues 9 and 12 respectively. This suggests that post-translational modification by LtnJ is influenced by the flanking regions. In the wild-type LtnA1 peptide the single d-alanine is flanked by F6 and L8. The LtnA1:F6A mutant exhibited no bioactivity and because the mutant peptide could not be detected by either u.v. or mass spectrometry it was not possible to determine whether the lack of activity was due to the removal of this bulky residue alone or due to an indirect impact on the adjacent d-alanine. LtnA1:L8A exhibited some bioactivity (35 AU ml−1) and the mass detected by CMS was consistent with that of a Leu to Ala change at position 8, with no consequence for D-Ala formation at position 9.
Mutagenesis of charged residues
LtnA1 has a net neutral charge due to the presence of two positively (H23, K30) and two negatively (D10, E24) charged amino acids. In contrast LtnA2 is cationic, possessing two positively charged residues (K24, R27). Mutagenesis of the four positively charged residues was found to have a relatively minor impact as the corresponding bioactivities were greater than 95 AU ml−1 in each case. The conversion of LtnA2K24 to alanine corresponds to a natural lysine to alanine alteration at this location in related peptides and may explain its high bioactivity (96 AU ml−1). The fact that LtnA2:R27A also retains a high level of bioactivity (118 AU ml−1) indicates that, despite showing features of a cationic antimicrobial peptide, a reduction in the overall charge of the peptide to +1 does not have major consequences. The creation of additional charge mutants in the future, including some derivatives of LtnA2 with a net neutral charge, will reveal the importance of the overall charge to this peptide. The high bioactivity of LtnA1:H23A (101.25 AU ml−1) is noteworthy in that it is the only one of eight mutants in the 18–25 region of LtnA1 that exhibits high levels of bioactivity. This is despite the fact that a histidine residue is also present at the corresponding locations in Sacα, Plwα and the majority of lacticin 481-like peptides. Tolerance to mutation at this location was also surprising as conversion of the Dha residue at the corresponding position in mersacidin to isoleucine, so that it would more closely resemble actagardine, has been shown to result in a greatly reduced peptide yield and a reduction in the specific activity of the peptide by nearly 1000-fold (Szekat et al., 2003).
Mutagenesis of the negatively charged amino acids in LtnA1 had a greater impact. While the bioactivity of LtnA1:D10A was reduced to 47.5 AU ml−1, it was the LtnA1:E24A mutant that was most affected, with no detectable bioactivity. This phenomenon coincided with an inability to detect the peptide either by u.v. or mass spectrometric analysis. This residue corresponds to E17 of mersacidin, which is thought to play a critical role in lipid II binding (Szekat et al., 2003). The detrimental consequences of the LtnA1E24 A mutation thus further highlights the importance of this conserved residue.
Mutagenesis of other conserved residues
As stated previously there are a number of residues that are highly conserved in two-peptide lantibiotics. The consequences of the mutation of a number of these have been described above (i.e. residues involved in bridge formation –LtnA1S9, C19, T20, T22, C25 and C29, LtnA2S16, C20, T22, C25, T26 and C29; or charged residues – LtnA1E24) and all result in the absence of bioactivity. Leucine residues corresponding to L21 of LtnA1 are present in five of the other nine mersacidin-like peptides. In agreement with previous observations mutation of this conserved residue also results in the elimination of bioactivity (O'Connor et al., 2006). Despite the conserved nature of this residue, it was surprising that there was such a dramatic impact on bioactivity as an alanine is naturally present at the corresponding location in the closely related Sacα. Mass spectrometric analysis failed to detect an LtnA1L21A peptide. Residues LtnA1G13 and LtnA1N14 are interesting in that both are highly conserved among the A1/α peptides of two-peptide lantibiotics but not among other mersacidin-like peptides. Although the LtnA1G13A peptide remains active, the LtnA1:N14A mutant lacks any discernible activity. It may yet be that the conservation of a glycine at this position is indeed reflective of its importance with respect to structure : function relationships and that alanine alone, because of its similarity to glycine, is tolerated. While the side-chain conversion is more dramatic for an asparagine to alanine change this alone does not adequately explain the inactivity of LtnA1:N14A, given the contrasting activity of LtnA1:N15A (see below). This result indicates that this specific region of ring 2 of LtnA1 has an important functional role.
There are four other mutants in which conserved amino acids are mutated (LtnA1:G16A, LtnA1:M26A, LtnA2:P21A and LtnA2:T23A) and yet a significant retention of bioactivity is apparent. This indicates that within lantibiotics a high degree of conservation does not necessarily indicate that an amino acid is inviolate. The retention of bioactivity of LtnA1:G16A was especially remarkable given its conservation across all mersacidin-like and lacticin 481-like peptides. It will be interesting to determine the consequences of altering this glycine to residues other than alanine. The bioactivities of LtnA1:M26A (conserved in 5 of 6 A1/α peptides), LtnA2:P21A (conserved in 4 of 6 LtnA2-like peptides) and LtnA2:T23A (conserved in all LtnA2-like peptides) were 20, 45 and 53 AU ml−1 respectively. LtnA2:P21A was notable for the fact that the mass of the bioengineered peptide was consistently 2817 Da, rather than the expected 2821 Da. It is not apparent how a proline to alanine conversion could impact on neighbouring residues in a manner that would explain a 4 Da difference. Thus this matter will require further inspection in the future.
Mutagenesis of aromatic residues
Aromatic amino acids are said to have membrane-seeking properties. More specifically it has been reported that in membrane acting peptides these residues are most likely to be located at the lipid–water interface (Pebay-Peyroula and Rosenbusch, 2001; de Planque et al., 2002; Jing et al., 2003; Sanderson and Whelan, 2004). Other than LtnA1F6, the mutation of which has been described above, there are five other aromatic residues (i.e. LtnA1 Y11, W12, W18 and W28; and LtnA2 Y14). The consequences of replacing these residues with alanine are remarkably consistent in that, in all three cases, a W to A replacement results in the elimination of bioactivity and an inability to detect the mutated peptide while the two Y to A mutations cause large decreases in bioactivity. It is important to note however, that the conversion of an aromatic amino acid to an alanine is a dramatic change, and replacement with another aromatic amino acid could theoretically be less disruptive.
Mutagenesis of remaining amino acids in LtnA1
Four residues in LtnA1 do not fit into any of the above categories. These are N4, N15, A17 and A27. All of the corresponding mutants display at least some activity and the activity of three of these (LtnA1:N15A, LtnA1:A17G, LtnA1:A27G) is greater than 55 AU ml−1. The tolerance to change at these three locations is unsurprising given that these represent three of the four locations where LtnA1 and Sacα differ (L21 was discussed above). It has previously been established that mutation of these residues in LtnA1, so that it will more closely resemble Sacα (i.e. N15K and A17N), does not severely impact on the associated bioactivity (O'Connor et al., submitted). Finally, the mutagenesis of LtnA1:N4A results in a drop in bioactivity to 13 AU ml−1, again highlighting the variable consequences of Asn to Ala conversions.
Mutagenesis of remaining amino acids in LtnA2
Of the remaining eight LtnA2 mutants (LtnA2:P3A, A4G, P6A, I15A, T17A, N18A, T19A and A28G) only LtnA2:N18A results in the elimination of bioactivity. Of the LtnA2 mutants described, it is the only alteration, other than those in which bridge-forming residues are changed, that confers this phenotype. However, in contrast to LtnA1N14, this asparagine is not conserved in that although there is a corresponding amide side-chain in Sacβ, all of the other A2/β peptides have an alanine at this position. Like N14, residues T17, T19 and A28 are located within rings formed by (β-methyl)lanthionine bridges. However, the corresponding alanine mutants in these cases are active, with LtnA2:T17A and LtnA2:A28G being the two most active of the LtnA2 mutants. As with other mutations that impact on the N-terminus of LtnA2, the LtnA2:P3A, A4G, P6A and I15A changes resulted in the retention of bioactivity. It is thus apparent that all 15 of the N-terminally located amino acids of LtnA2 can be mutated without the elimination of bioactivity, indicating it to be a variable domain.
High-performance liquid chromatography analysis of LtnA1/LtnA2 derivatives
It is apparent from the HPLC profiles of peptides purified from the surface of producing strains that peak sizes varied dramatically across the different mutants (Figs. S1 and S2). It was also evident from the MG1363 LtnA1 mutant profiles that there was a correlation between the size of the wild-type LtnA2 peaks and that of the mutant LtnA1 peak (Fig. S1). Given that complementation studies indicated a plentiful supply of LtnA2 (data not shown), it would seem that the attachment of LtnA2 to the surface of the producer cells is mediated through LtnA1. This attachment is not facilitated by the immunity protein LtnI as significant levels of LtnA1 and LtnA2 can be found attached to the surface of an ΔltnI mutant (data not shown). The alternative conclusion is that the binding of the peptides to the surface of the producing cells is, as is the case for sensitive target cells, due to lipid II binding. However, it is not apparent how the producing cell could continue to synthesize peptidoglycan were this to be the case. When the profiles of the LtnA2 mutants were examined a peak corresponding to LtnA1 was identifiable in all cases (Fig. S2). However, the size of the LtnA1 peak was reduced in almost all cases. Taken together, these data suggest that binding of LtnA1 and LtnA2 to the surface of the producer cells is a cooperative process in which LtnA1 plays the primary role and is stabilized by LtnA2.
Specific activity of a number of LtnA1/LtnA2 derivatives
Nine of the mutant peptides (LtnA1C1A, N14A, N15A, K30A; and LtnA2P3A, P6A, I8A, I15A and P21A) were purified and minimum inhibitory concentration (MIC) determination assays were carried out (Table 2). When present with equimolar concentrations of companion peptide, the specific activity of the bioengineered LtnA1 peptides followed the pattern N15A > K30A > C1A > N14A, while the corresponding LtnA2 derivatives followed the pattern P6A = I8A > P3A > I15A > P21A. By combining specific activity and peptide production values it was possible to make predictions of the overall bioactivity that closely match the experimentally determined bioactivity. Differences between agar- and broth-based assays are most likely to be responsible for discrepancies such as the relatively low predicted bioactivity of LtnA1:N15A with LtnA2 compared with the actual bioactivity of the supernatant from the corresponding mutant. The purification of pure peptide also facilitates determination of the specific activity of the individual peptides. All four LtnA1 bioengineered peptides possessed greatly reduced specific activity with LtnA1:N15A and LtnA1:K30A, but not LtnA1:C1A and LtnA1:N14A being active at 10 μM. The specific activity of the individual LtnA2 peptides (Table 2) is particularly revealing. Of the five peptides the specific activity of LtnA2:P3A (10 μM) and LtnA2:P6A (> 10 μM), i.e. the two peptides in which alterations are closest to the N-terminus, is lowest thus indicating that the N-terminal region plays a greater role in solo LtnA2 activity. In contrast LtnA2:I8A, LtnA2:I15A and LtnA2:P21A retain higher levels of specific activities (5 μM, 2.5 μM and 5 μM respectively) which are comparable to that of wild-type LtnA2 (2.5 μM). When one contrasts these values with the low activities of LtnA2:I15A and LtnA2:P21A combined with LtnA1, it would appear that a reduced ability to function synergistically with LtnA1 is the primary deficiency of these two peptides. This information supports the theory that the more highly conserved, C-terminal region of the peptide is responsible for synergism with LtnA1. It may also be, given the central location of LtnA2I15, that it functions as a hinge between the N-terminal and C-terminal domains of the peptide.
Table 2. Specific activity of selected mutant peptides.
Order reflects the order of f, with i being the most active supernatant.
Rational drug design has long been utilized by the pharmaceutical industry to improve antibiotics. The phenotypic consequences of rational derivation may lead to new rounds of hypothesis, synthesis and testing until a comprehensive understanding of the antimicrobial is attained. However the degree to which most antibiotics can be altered is extremely limited as a consequence of the multistep, multienzymatic biosynthetic process. In contrast the lantibiotics are gene-encoded and are thus more amenable to change through the use of precise genetic engineering. To take advantage of this fact we set ourselves the task of producing the first detailed ‘map’ of a lantibiotic in an effort to distinguish between residues that are inviolate and those that are variable. Our endeavours to achieve this goal have resulted in the creation of a total of 59 lacticin 3147 alanine (or glycine) scanning mutants and a total of 71 such mutants in total (Cotter et al., 2005c; O'Connor et al., submitted). These figures mean that, not only are there more bioengineered derivatives of lacticin 3147 than of any other lantibiotic, the number of lacticin 3147 mutants is greater than the combined number of all other lantibiotic mutants published to date. The benefits of producing such a map are significant. It is immediately apparent that, of the 59 mutants, at least 36 retain some bioactivity. It is also highly likely that some of the remaining 23 residues will tolerate some, albeit minor, changes. Although none of the mutants created display bioactivity levels greater than that of the parent strain, the retention of activity indicates that a particular residue is amenable to change. Additional bioactivity assays, with more than 50 Gram-positive indicators, established that the data presented here (using L. lactis HP as an indicator) is a true reflection of the consequences of mutagenesis in each case (data not shown). Thus, the alanine scanning mutagenesis strategy has achieved its goal and will immeasurably benefit future rational design of derivatives of the lacticin 3147 peptides.
This study has also revealed information that has enhanced our understanding of lacticin 3147 and of related peptides. While the consequences of mutating individual amino acids have been described in great depth already, there would seem to be merit in analysing the consequences of mutagenesis with respect to stretches of amino acids that represent distinct structures within the peptides (Fig. 3). When the structure of LtnA1 was elucidated (Martin et al., 2004), one of the interesting features observed was the presence of a lanthionine bridge at the N-terminus. Although its role has yet to be ascertained, it is apparent from this study that its presence is not a prerequisite for antimicrobial activity. While this ring represents only one of a number of structural features that mersacidin and LtnA1 share, mersacidin differs from LtnA1 and other A1/α peptides by lacking an extended stretch of six amino acids between ring one and two of LtnA1. The non-production of a number of peptides mutated within this region resulted in our being unable to determine whether these residues contributed to the key functional difference between LtnA1 and mersacidin, i.e. the ability to function synergistically with LtnA2 (Wiedemann et al., 2006a). Bioengineering of the residues within ring 2 of LtnA1 reveals it may represent a variable region in that six of the nine corresponding mutants retain bioactivity when altered. While none of these mutants are more active than the parental strain, the results suggest that random mutagenesis of this specific region would be expected to yield a large number of active derivatives. These results contrast with the essential nature of many residues at the C-terminal end of LtnA1 (T20-K30) that includes two intertwining lanthionine bridges. Seven of the 11 mutants in which this region is altered lack bioactivity and if one considers the stretch of residues from W18 to C25, seven of the eight corresponding mutants lack bioactivity. Given the apparent essential nature of these mutations, and the observation that the Cys19-Cys25 stretch in LtnA1 contains a core region (TxS/TxD/EC) of the conserved motif common to all mersacidin- and lacticin 481-peptides, it would appear that all of these peptides, constituting approximately 40% of all lantibiotics described to date, share a common essential domain. Significantly this region in mersacidin represents the first point of contact between mersacidin and the bacterial cell (Hsu et al., 2003). It may be that these residues form a lipid II binding domain analogous to that previously described in nisin- and epidermin-like peptides (Hsu et al., 2004) and, by extension, that all lacticin 481-like peptides possess the ability to bind this essential peptidoglycan precursor. It has already been established that plantaricin C possesses this ability (Wiedemann et al., 2006b). With respect to biotechnological advances, the conserved nature of this region (amino acids 19–25) and its inflexibility to alanine-conversion indicates that it is not suited to large scale mutagenesis and that even minor alterations, with the exception of His23, may not be permitted. Alanine-scanning mutagenesis of LtnA1 indicates that it is highly likely that the activity of this peptide can be improved upon, at least against some strains. It is already apparent that there is room for some improvement in that although LtnA1 and mersacidin are equally active against L. lactis HP, mersacidin is 30 times as active against Micrococcus flavus (Wiedemann et al., 2006a). Now that we have identified variable and inviolate residues and domains within LtnA1, it will be possible to design a strategy that is more likely to result in the production of derivatives with enhanced activity against M. flavus and, more importantly, pathogenic bacteria. Given its location in the middle of the putative lipid II binding stretch saturation mutagenesis of His23 could prove to be particularly interesting.
It is exciting to note that LtnA2 would appear to be an even better target than LtnA1 with respect to the use of bioengineering to enhance activity. Although the three (β-methyl)lanthionine bridges, which are highly conserved among LtnA2-like peptides, would appear to be essential, 21 of the remaining 23 are amenable to change (Fig. 2). Given that conservation is particularly high at the C-terminus (LtnA2Y14-C29 shares 81% identity with Sacβ) it is likely that this region represents a common functional domain. As the ability to act synergistically with an A1/α peptide is the primary common function of all LtnA2-like peptides, it is most likely that the C-terminus of LtnA2 performs such a function. The fact that LtnA2I15A and LtnA2P21A retain solo activity comparable with that of LtnA2, while their activities with LtnA1 are greatly reduced relative to LtnA1-LtnA2, and that Sacβ can function synergistically with LtnA1 (O'Connor et al., submitted) both support this theory. In contrast to the C-terminal region, the composition of the elongated N-terminal region among the LtnA2-like peptides indicates a high degree of natural variation. This stretch is predicted to have an elongated conformation and adopts an α-helical conformation in methanol (Martin et al., 2004), two characteristics typical of cationic pore-forming antimicrobial peptides. In addition, following modification 11 of the 13 N-terminal residues of LtnA2 are aliphatic. While this may contribute to the retention of bioactivity following alanine/glycine mutagenesis, the presence of such a stretch of residues may in itself be significant as the presence of a high percentage of aliphatic amino acids at the N-terminus has been shown to be another common feature of membrane disruptive peptides (Han and Kang, 2004). If the N-terminus of LtnA2 does play a major role in pore-formation its flexibility to change makes it an excellent target for large scale bioengineering.
In addition to providing information pertaining to individual residues and regions, scanning mutagenesis has also revealed that the activity of LtnJ, the enzyme responsible for Dha to d-ala conversion in lacticin 3147 (Cotter et al., 2005c), is likely to be dependent on motif recognition. Prior to this study the possibility existed, given the absence of Dha residues from LtnA1 and LtnA2, that LtnJ converted all Dhas to d-ala regardless of their location position. However the apparent presence of free Dhas in LtnA1:C1A and LtnA1:C19A and the impact of the I8A and A13G mutations in LtnA2 on the formation of d-ala residues at positions 9 and 12, respectively, indicate that the activity of LtnJ is site-specific.
In one of the first reports of site directed mutagenesis of a lantibiotic (Liu and Hansen, 1992) it was suggested that these inhibitors could be regarded as primitive antibodies produced by bacteria, possibly even possessing constant and variable regions. This report lends further credibility to that proposal. While bacteria have no mechanism to generate a diverse population of lantibiotic peptides this can now be achieved on a large scale in the laboratory. It is apparent that the C-terminus of LtnA2 and, to a lesser extent, ring 2 of LtnA1, constitute variable domains while amino acids involved in bridge formation and a stretch of seven out of eight core residues in LtnA1 constitute the main essential regions. The information described here represents the first time that a lantibiotic blueprint has been generated upon which rational residue-specific and semirandom domain-specific design can be superimposed to bioengineer enhanced lantibiotic derivatives and ultimately novel antimicrobials.
Strains and growth conditions
Lactococcus lactis MG1363 pMRC01 (MGMR), MGMR pVE6007, LtnA1:S7A, LtnA1:L21A, LtnA2:S9A, LtnA2:S12A, LtnA1:N15K, LtnA1:A17N, MG1363 pOM31 and MG1363 pOM39 have been described previously (McAuliffe et al., 2000; Cotter et al., 2005c; O'Connor et al., submitted). L. lactis strains were routinely grown at 30°C without aeration in M17 broth (Oxoid) supplemented with 0.5% glucose (GM17) unless stated otherwise. Escherichia coli EC101 was grown at 37°C in Luria–Bertani broth with vigorous agitation unless stated otherwise. Antibiotics were used, where indicated, at the following concentrations: erythromycin (E), 150 μg ml−1 for E. coli and 5 μg ml−1 for L. lactis; chloramphenicol (Cm), 5 μg ml−1 for L. lactis. Xgal was used at a concentration of 50 μg ml−1.
Mutagenesis of pMRC01, the natural 60.2 kb plasmid on which the genetic determinants associated with lacticin 3147 production and immunity are located, was carried out using a combination of the Quikchange site-directed mutagenesis strategy (Stratagene) and double crossover mutagenesis with pORI280 (RepA–, LacZ+) as described previously (Cotter et al., 2003; 2005c) using the Quikchange protocol according to the manufacturers instruction (Table S1) with the exception that EC101 (RepA+) was used as a host. To screen for transformants containing an altered pORI280-ltnA1A2 in each case, a PCR reaction using a specific ‘check’ primer, designed to amplify mutated plasmid template only, and ltnA1A or ltnA1D (depending on the orientation of the check primer) was carried out. DNA sequencing was used to confirm that mutagenesis was successful and that no other alterations had been incorporated. Each plasmid was in turn introduced into MGpMRpVe (Cotter et al., 2003) by electroporation (Holo and Nes, 1995) and transformants were selected by plating on GM17-E-X at 30°C. Loss of the temperature sensitive plasmid pVe6007, and identification of cells in which the pORI280-ltnA1A2 derivative had integrated by single crossover recombination, was brought about by growth at 37°C in GM17-E broth followed by streaking onto GM17-E-X agar at the same temperature. Resultant colonies were then checked to ensure their inability to grow on GM17-Cm agar at 30°C. These colonies were subcultured in GM17 at 37°C and cultures were plated on GM17-X at regular intervals with a view to identifying LacZ- colonies, i.e. colonies in which a second crossover event had occurred resulting in the excision and loss of pORI280. A PCR reaction again using the relevant check primer and ltnA1A or ltnA1D was carried out to distinguish between mutants and wild-type revertants. The relevant region in candidate mutants was again sequenced to confirm the ‘check’ PCR reaction. To ensure that any changes in bioactivity from mutated strains were not attributable to an indirect impact on the unmutated companion peptide, supernatant from the LtnA1 and LtnA2 mutant strains was combined with that of MGpOM31 (LtnA1+, LtnA2– subclone) and MGpOM39 (LtnA2+, LtnA1– subclone) respectively. It was established that in every case there was significant production of the unmutated peptide (data not shown).
Antimicrobial activity assays
The antimicrobial activity of cell-free culture supernatant was determined by critical dilution assay (Ryan et al., 1996). Basically, molten agar was cooled to 48°C and seeded with the indicator strain L. lactis ssp. cremoris HP (∼2 × 107 fresh overnight-grown cells). The inoculated medium was dispensed into sterile Petri plates, allowed to solidify and dried. Wells (4.6 mm diameter) were made in the seeded agar plates. Aliquots of a twofold serial dilution of the culture supernatant were dispensed into wells, and the plates were incubated overnight at 30°C. The arbitrary units (AU ml−1) were determined as described previously (Ryan et al., 1996).
An overnight culture of the strain of interest was inoculated into 1 l of modified TY broth (1% inoculum). The inoculated medium was incubated overnight at 30°C and the cells were harvested by centrifugation (7000 g for 20 min) and resuspended in 250 ml of 70% propan-2-ol, pH 2.0 (adjusted to pH 2.0 by addition of conc. HCl). Following stirring for 4 h at 4°C, the cell debris was removed by centrifugation and the bacteriocin-containing supernatant reduced to approximately 60 ml by removing propan-2-ol via rotary evaporation. The resultant preparation was applied to a 10 g (60 ml volume) Varian C18 Bond Elut Column (Varian, Harbor City, CA) pre-equilibrated with methanol and water. The column was subsequently washed with 120 ml of 30% ethanol and inhibitory activity was eluted with 100 ml of 70% propan-2-ol, pH 2. From the 100 ml of bacteriocin-containing eluate 10 ml or 20 ml (depending on peak size) was concentrated to a volume of 1 ml through the removal of propan-2-ol by rotary evaporation. Aliquots of 700 μl were then applied to a Phenomenex (Phenomenex, Cheshire, UK) C12 reverse phase (RP)-HPLC column (Jupiter 4u proteo 90 Å, 250 × 10.0 mm, 4 μm) previously equilibrated with 25% propan-2-ol, 0.1% trifluoroacetric acid TFA. The column was subsequently developed in a gradient of 30% propan-2-ol containing 0.1% TFA to 60% propan-2-ol containing 0.1% TFA from 4 to 40 min at a flow rate of 1.2 ml min−1. Relative production was quantified on the basis of peak size as calculated by the method used previously (Cotter et al., 2005c), i.e. with Shimadzu class VP software (Shimadzu Biotech, Manchester, UK).
For CMS bacteria were collected by sweeping sterile plastic loops across colonies and mixed with 50 μl of 70% iso-propanol adjusted to pH 2 with HCl. The suspension was vortexed, the cells spun down in a benchtop centrifuge at 14 000 r.p.m. for 2 min, and the supernatant was removed for analysis. Mass spectrometry in all cases was performed with an Axima CFR plus MALDI TOF mass spectrometer (Shimadzu Biotech, Manchester, UK). A 0.5 μl aliquot of matrix solution (alpha-cyano-4-hydroxy cinnamic acid (CHCA), 10 mg ml−1 in 50% acetonitrile-0.1% (v/v) trifluoroacetic acid) was deposited onto the target and left for 1–2 min before being removed. The residual solution remaining was allowed to air-dry and the sample solution (HPLC fraction or CMS supernatant) was deposited onto the precoated sample spot. Matrix solution (0.5 μl) was added to the deposited sample and allowed to air-dry. The sample was subsequently analysed in positive-ion reflectron mode.
Minimum inhibitory concentration determination
Minimum inhibitory concentration determinations were carried out in microtitre plates as described previously (Wiedemann et al., 2006a). L. lactis HP was grown in M17 broth plus 0.5% glucose (Oxoid). Serial twofold dilutions of the peptides were made in the growth medium of the respective indicator strain. Bacteria were added to give a final inoculum of 105 colony-forming units per millilitre in a volume of 0.2 ml. After incubation for 16 h at 30°C the MIC was read as the lowest peptide concentration causing inhibition of visible growth. Results given are mean values of three independent determinations.
This work was supported by the Irish Government under the National Development Plan (2000–2006), and Science Foundation Ireland.