Organization of the genes encoding the biosynthesis of actagardine and engineering of a variant generation system


*E-mail; Tel. (+44) 1707 356133; Fax (+44) 1707 356131.


The biosynthetic pathway of the type B lantibiotic actagardine (formerly gardimycin), produced by Actinoplanes garbadinensis ATCC31049, has been cloned, sequenced and annotated. The gene cluster contains the gene garA that encodes the actagardine prepropeptide, a modification gene garM, involved in the dehydration and cyclization of the prepeptide, several putative transporter and regulatory genes as well as a novel luciferase-like monooxygenase gene designated garO. Expression of these genes in Streptomyces lividans resulted in the production of ala(0)-actagardine while deletion of the garA gene from A. garbadinensis generated a strain incapable of producing actagardine. Actagardine production was successfully restored however, by the delivery of the plasmid pAGvarX. This plasmid contains an engineered cassette of the actagardine encoding gene garA and offers an alternative route to generating extensive libraries of actagardine variants. Using this plasmid, an alanine scanning library has been constructed and the mutants analysed. Further modifications include the removal of the novel garO gene from A. garbadinensis. Deletion of this gene resulted in the production of deoxy variants of actagardine, demonstrating that the formation of the sulfoxide group is enzyme catalysed and not a spontaneous chemical modification as previously believed.


Lantibiotics are ribosomally synthesized antibiotic peptides that undergo enzyme-mediated post-translational modifications to generate their active form. They are characterized by the presence of the unusual amino acids lanthionine (lan) and/or methyllanthionine (MeLan). These are formed by the dehydration of selected serine and/or threonine residues followed by the intramolecular addition of cysteine. The formation of these residues generates intrachain thioether bridges that impart structural stability and give rise to a diverse range of biologically active compounds. In addition to lanthionine residues, lantibiotics may also contain many other modified residues (for reviews, see Chatterjee et al., 2005; Willey and van der Donk, 2007). Following the modification of specific residues of the prepeptide, the N-terminal leader sequence (between 23 and 59 amino acids in length) is proteolytically cleaved and the mature peptide exported. Although the exact role of this leader peptide remains unclear, its presence seems to ensure the full processing and maturation of the lantibiotic peptide (Patton et al., 2008). In contrast to multi-enzyme complexes, the direct ribosomal synthesis of lantibiotics makes them an attractive proposition for pathway engineering. The structure of the peptide can potentially be manipulated in a much more direct fashion than for other classes of metabolites.

The genes encoding lantibiotic biosynthesis, export and immunity are clustered and designated by the generic symbol lan. In addition to the gene that encodes the lantibiotic peptide lanA, each lantibiotic gene cluster sequenced to date contains either a lanM in the case of type B lantibiotics or a lanB and C for type A. The products of these genes catalyse the dehydration and cyclization of the prepeptide to the mature lantibiotic. Finally, all lantibiotic gene clusters (with the exception of epicidin 280) possess genes encoding an ABC transport system responsible for the export of the mature peptide from the cell: this gene is designated lanT. Additional genes encoding regulation (lanRK), peptidases (lanP) and transport/immunity (lanFEG) have also been reported (Chatterjee et al., 2005).

First described in 1977 (Parenti et al., 1977), actagardine (Fig. 1A) is 19 amino acids in length and is produced by the actinomycete, Actinoplanes garbadinensis. The structure of actagardine has been determined by NMR (Zimmerman and Jung, 1997), which has shown that the molecule has a compact conformation comprised of an N-terminal lanthionine ring and three intertwined C-terminal methyllanthionine rings. Binding studies suggest that actagardine, like the related lantibiotic mersacidin, inhibits cell wall biosynthesis by binding to lipid II and blocking transglycosylation (Héchard and Sahl, 2002). Actagardine demonstrates potent activity against Gram-positive pathogens and crucially its binding is not antagonized by vancomycin (Brötz et al., 1998).

Figure 1.

The structure and amino acid sequence of actagardine. The structure of the actagardine peptide is shown in (A) while the amino acid sequence of the GarA prepeptide is shown in (B). The arrow indicates the site of protease cleavage and release of the mature actagardine peptide from the leader sequence of the prepeptide. Residues involved in bridge formation are underlined.

We describe here the cloning and analysis of the biosynthetic gene cluster encoding actagardine and the development of a system for generating libraries of actagardine variants. This system was validated by substituting the amino acids not involved in lanthionine bridge formation for alanine residues. New actagardines were generated and evaluated. The results suggest that actagardine variants can be readily generated and this may constitute a valuable method for improving the antibacterial activity and pharmaceutical properties of this molecule.

Results and discussion

Analysis of the cosmid CosAG14 encoding the biosynthetic gene cluster for actagardine

A cosmid library of genomic DNA isolated from A. garbadinensis ATCC31049 was constructed and screened against a digoxigenin (DIG)-labelled oligonucleotide and a DNA probe believed to encode the actagardine structural gene. The cosmid CosAG14, hybridized to both probes and was selected for complete sequencing. This cosmid includes a 38 168 bp fragment containing 31 complete open reading frames (ORFs) and one incomplete ORF at the 3′ end of the fragment (this sequence has been deposited in the GenBank database with the accession number FJ547091). Translation of ORF 19, located close to the centre of this fragment, confirms that this gene, designated garA, encodes the actagardine prepropeptide. A putative leader peptide lies immediately upstream of the actagardine coding sequence (Fig. 1B). At positions −1 and −2 relative to actagardine there are two alanine residues, this arrangement is identical to that identified in the leader peptide of mersacidin (Altena et al., 2000). These residues are believed to represent the recognition site for protease cleavage of the leader from the mature lantibiotic. In this instance, cleavage between position −1 and −2 would result in the retention of an alanine residue at the N-terminus, generating ala(0) variants (m/z +36 for doubly charged ions) of actagardine. Such variants were described by Vértesy et al. (1999) and are routinely observed in the fermentation broths of A. garbadinensis. The abundance of ala(0) products containing an additional N-terminal alanine residue appears to decrease in prolonged fermentations of A. garbadinensis presumably due to cleavage of the N-terminal alanine. A putative stem loop begins at the eleventh base pair downstream of garA and spans a predicted 44 bases (ΔG, −157.8 kJ mol−1). Stem loop structures are a common feature of lantibiotic gene clusters (Altena et al., 2000; Widdick et al., 2003), and have been shown to stabilize mRNA transcripts (Pag et al., 1999).

The organization of the genes included in CosAG14 is shown in Fig. 2, while the proposed function of the protein encoded by each ORF is provided in Table 1. The genes necessary for the biosynthesis of lantibiotics are typically clustered around the structural gene, lanA and normally consist of not more than 17 genes. It is therefore highly unlikely that all of the 31 genes encoded by CosAG14 are involved in the biosynthesis of actagardine. The annotation of the ORFs 20–27 that lie downstream from the garA gene and are believed to be involved in the biosynthesis of actagardine is as follows.

Figure 2.

Map showing the gene organization of the cosmid CosAG14 consisting of DNA isolated from Actinoplanes garbadinensis ATCC31049. The order and direction of transcription of all of the open reading frames are shown. Genes believed to be involved in the biosynthesis of actagardine are filled, with the structural gene garA coloured black. The approximate position of the putative stem loop downstream of garA is indicated by an asterisk. pbp, penicillin binding protein.

Table 1.  Open reading frames identified in CosAG14.
ORFProposed functionLength
  • a. 

    ORF 32 is believed to be incomplete.

  • The lengths in amino acids of each protein are provided.

3ATPase involved in cell division518
4Sugar hydolase203
6Cytosine/adenine deaminase161
9Pyruvate oxidase597
10Hydrolase or acyltransferase223
11Aldose epimerase396
12ABC transport component378
13ABC sugar transport permease352
14ABC transport protein253
16ABC transport permease280
17ABC transport permease301
18Substrate binding ABC transporter445
19GarA, actagardine prepeptide64
20GarM, modification enzyme1053
21GarO, monooxygenase336
22GarR1, response regulator231
23GarH, transporter permease812
24GarT, ABC transporter259
25GarK, response regulator kinase372
26GarR, response regulator sensor220
27Penicillin binding protein800
30Response regulator222
31Fructose bisphosphate aldolase286

GarM (orf20) encodes a 1053-amino-acid protein with strong sequence similarities to lanM proteins. Comparisons using blast aligns GarM to ClvM of the recently sequenced Clavibacter michiganensis ssp. michiganensis NCPPB382 chromosome, producer of the lantibiotic closely related to actagardine, michiganin A (Gartemann et al., 2008). LanM proteins catalyse the maturation of the prepeptide produced by the structural lanA gene. The product of garM is likely to catalyse the dehydration of serine and threonine residues of the actagardine prepeptide and subsequent formation of lanthionine bridges.

GarO (orf21) encodes a 341-amino-acid protein responsible for catalysing the incorporation of oxygen, generating actagardine from deoxyactagardine (as described in the following section).

The protein product of garR1 shows sequence similarity (∼45% identity) to several two-component response regulators. The N-terminal includes a conserved aspartate residue believed to be the phosphate acceptor site. However, as reported for MrsR1 (Altena et al., 2000), the gene does not include a Lys-Pro-Phe motif necessary for conformational change in response to phosphorylation of the Asp residue. The C-terminal (from amino acid 160) aligns to helix–turn–helix transcriptional regulators.

GarH has a length of 2439 bp. The encoded protein shares sequence similarity (∼30% identity) with ABC transporter-associated permeases. Just 3 bp downstream from garH is the stop codon of garT. The protein encoded by garT is 259 amino acids in length and shares a 57% identity to SCO4666, an ABC-type antimicrobial peptide transport component of Streptomyces coelicolor (Bentley et al., 2002). Despite a lack of close similarity of both garH and garT to genes of other lantibiotic clusters, the organization of these genes and their proposed products suggests that they are involved in transport of actagardine from the cell and/or play a role in self-immunity.

GarK (orf25) overlaps the start codon of the downstream gene garR (orf26) by 4 bp. This arrangement is consistent with lanK and lanR genes identified in numerous lantibiotic biosynthetic clusters including michiganin A (Holtsmark et al., 2006), mersacidin (Altena et al., 2000) and cinnamycin (Widdick et al., 2003). The products of these genes are 367 and 220 amino acids respectively, and likely form a two-component regulatory system. garK is believed to encode a sensory histidine kinase sharing a sequence identity of 39% to ClvK. garR likely encodes a response regulator and is 51% identical to ClvR. Studies on the analogous genes mrsR2 and mrsK2 of the mersacidin cluster demonstrated that the proteins encoded by these genes regulate the expression of the immunity genes mrsE, F and G (Guder et al., 2002). It is therefore conceivable that the products of garK and garR regulate the expression of genes involved in immunity to actagardine, possibly, garH and/or garT.

Orf27 is 2403 bp in length. The encoded protein shares a 30–55% sequence identity with penicillin-binding proteins and contains conserved regions of glycosyltransferase and transpeptidase domains. An analogous gene has been identified in Listeria monocytogenes and is believed to be involved in conferring resistance to nisin (Gravesen et al., 2004). Given the proximity of orf27 to the actagardine structural gene, it is conceivable that the product of this gene performs an analogous role in A. garbadinensis.

An enzyme responsible for cleavage of the leader peptide generating mature actagardine has not been identified. This function is often performed by a protease, LanP, as in the case of nisin, epidermin and Pep5 (Willey and van der Donk, 2007) or by a protease domain incorporated into a bifunctional LanT protein as described for the lctT gene of the lacticin 481 gene cluster (Ihnkunen et al., 2008). However, as reported for the gene clusters encoding the actagardine-like michiganin A (Gartemann et al., 2008) and cinnamycin (Widdick et al., 2003), the actagardine gene cluster does not encode a LanP and despite the extended length of garH compared with cinH (cinnamycin cluster), no obvious protease domain has been identified. It is possible that cleavage of the leader peptide occurs through the action of a protease encoded outside the cluster.

With the exception of garA and garO, the roles of proteins encoded by the genes in this region have been assigned through sequence comparisons. The exact roles of the other genes encoded by CosAG14 and delineation of the biosynthetic gene cluster remain the subject of ongoing studies.

Role of the garO gene product

Actagardine shares a high degree of homology to michiganin A produced by the plant pathogen C. michiganensis subsp michiganensis (Holtsmark et al., 2006). In addition to an extra N- and C-terminal amino acid, the sequence of the michiganin A peptide differs from actagardine by just two residues (V5L and V15I; actagardine/michiganin A). Despite this homology the arrangement and genes encoding biosynthesis of the respective lantibiotic differ markedly. Most noticeable is the presence of a gene (designated garO) within the actagardine cluster, which is absent from the michiganin biosynthetic gene cluster. This gene encodes a proposed protein that is homologous to luciferase-like monooxygenases. The proximity of this gene to garA suggests that garO is involved in the biosynthesis of actagardine, although to date no monooxygenases have been described in any of the published lantibiotic biosynthetic gene clusters. Actagardine however, is unique among lantibiotics in that it is the only example that includes a sulfoxide group. Previous studies have proposed that the formation of this bond at the thioether bridge between residues 14 and 19 may result from chemical modifications arising during isolation procedures (Zimmerman et al., 1995). The identification of GarO as a putative monooxygenase however, suggests that the formation of the sulfoxide bond may well be enzyme catalysed. To examine this possibility, a strain of A. garbadinensis in which the garO gene has been removed was engineered. This strain was designated A. garbadinensisΔgarO. Analysis of supernatant from fermentations of A. garbadinensisΔgarO by high-performance liquid chromatography (HPLC) and LC-MS identified the presence of compounds with masses and retention times corresponding to deoxyactagardine and ala(O)-deoxyactagardine (Fig. 3, well 8) at yields comparable to actagardine production by the wild-type strain. No products consistent with the oxidized forms of these compounds were detected. The antibacterial activity of deoxyactagardine was marginally lower compared with oxidized actagardine (refer to Table 2), implying that the sulfoxide group may affect the activity of actagardine. However, a more comprehensive in vitro antimicrobial study would be needed to confirm the hypothesis regarding the role of this sulfoxide group to the activity of actagardine.

Figure 3.

The analysis of supernatants from the fermentation of engineered strains of A. garbadinensis and S. lividans grown in the medium GM1 for 5 and 3 days respectively. The wells in Mueller-Hinton agar seeded with Micrococcus luteus (A) were loaded with supernatants (50 μl) from fermentations of the following strains: (1 and 5) A. garbadinensis; (2) S. lividans 1326; (3) S. lividans 1326::CosAG14HEapra; (4) purified actagardine (50 μg ml−1); (6) A. garbadinensisΔgarA; (7) A. garbadinensisΔgarA::pAGvarX; and (8) A. garbadinensisΔgarO. The LC-MS traces (B) show the mass spectra of compounds detected: (a) actagardine (wells 1, 4, 5 and 7); (b) ala(0)-actagardine (wells 1, 3, 5 and 7); (c) deoxyactagardine (well 8); and (d) ala(0)-deoxyactagardine (well 8). The major ions detected were the [M + H + Na]+2 ions.

Table 2.  The antimicrobial activity of actagardine, deoxyactagardine and vancomycin against a range of Gram-positive pathogenic organisms shown as minimum inhibitory concentrations (μg ml−1)
  1. VRE, vancomycin-resistant enterococcus; VSE, vancomycin-sensitive enterococcus; EMRSA, epidemic methicillin-resistant S. aureus; MSSA, methicillin-sensitive S. aureus; MSSE, methicillin-sensitive S. epidermidis.

Enterococcus faecalis 292128, 88, 82, 2
E. faecium 7131121 (VRE)64, 6464, 64> 64, > 64
E. faecium 19579 (VSE)16, 1632, 320.5, 1
Micrococcus luteus 46981, 12, 41, 1
Staphylococcus aureus R33 (EMRSA)8, 168, 161, 1
S. aureus SH1000 (MSSA)8, 816, 322, 2
Streptococcus epidermidis 11047 (MSSE)32, 3232, 321, 2
Streptococcus pneumoniae R616, 3216, 320.5, 0.5

The generation of solely unoxidized actagardines by A. garbadinensisΔgarO indicates a likely role of the garO gene as the monooxygenase responsible for the formation of the sulfoxide bond. This represents the first monooxygenase gene, to our knowledge, to be identified among lantibiotic biosynthetic gene clusters. The lack of a homologous gene encoding a monooxygenase in the michiganin A gene cluster explains the absence of a sulfoxide bond in the corresponding lanthionine bridge (Holtsmark et al., 2006).

Heterologous expression of the biosynthetic genes encoding actagardine in Streptomyces lividans 1326

Heterologous expression systems have been used in the study of several lantibiotics, including epicidin (Heidrich et al., 1998), cinnamycin (Widdick et al., 2003) and nukacin (Nagao et al., 2007). Expression of the biosynthetic gene cluster containing garA in a heterologous host provides additional evidence for the role of these genes and potentially offers an alternative means of generating variants of actagardine. Redirect technology (Gust et al., 2004) was used to introduce a cassette (isolated as an SspI fragment from pMJCOS1) containing the Streptomyces phage ΦC31 attachment site attP and integrase (int) into the neomycin resistance gene (neo) of CosAG14, resulting in the generation of CosAG14HEapra. This cosmid was introduced into S. lividans 1326 and the exconjugants were used for fermentations in the medium GM1. These cultures were incubated for 7 days at 30°C over which time aliquots were removed and analysed by well diffusion assays against Micrococcus luteus and HPLC-MS.

Zones of inhibition (halos) indicative of the presence of a biological active compound(s) were observed around the wells loaded with supernatants of S. lividans containing the cosmid CosAG14HEapra (S. lividans::CosAG14HEapra). HPLC-MS analysis confirmed the presence of compounds with retention times and masses corresponding to ala(0)-actagardine (Fig. 3, well 3). These compounds were absent from supernatants of S. lividans 1326. The levels of production peaked at 3 days of incubation and were comparable to the combined levels of actagardine and ala(0)-actagardine synthesized by the wild-type producer following 5 days of incubation (∼50–80 mg l−1).

Although it is possible that complementing activities have been provided by the host S. lividans, the detection of ala(0)-actagardine produced by S. lividans::CosAG14HEapra demonstrates that CosAG14 includes the genes necessary for the production of ala(0)-actagardine. The accumulation of ala(0)-actagardine in the heterologous system suggests that the protease responsible for cleavage of ala(0)-actagardine to actagardine may lie outside the biosynthetic cluster and is probably distinct to the protease responsible for cleavage to generate ala(0)-actagardine. This latter protease is likely to be an unrecognized gene within the cluster or provided from outside the cluster by both A. garbadinensis and S. lividans. The detection of ala(0)-actagardine produced by S. lividans::CosAG14HEapra provides evidence that this system expresses the actagardine biosynthetic genes. It also exemplifies the potential to express variants of the garA gene, in an organism that is well characterized, grows faster than A. garbadinensis and produces ala(0)-actagardine at levels comparable to the wild-type strain.

Engineering of pAGvarX: a vector designed for generation of variants of the actagardine peptide

The biological activity of lantibiotics and lack of reported resistance make them good candidates for mutagenesis with the intention of generating compounds with improved activity. A plasmid has been constructed that enables amino acids at any position (except the bridge forming residues at positions 6 and 7) of the actagardine peptide to be replaced, deleted or inserted. The plasmid, pAGvarX, contains a modified version of the garA gene in which the unique restriction sites BglII, BsrGI and AvrII have been introduced through silent mutations. Restriction digestion of pAGvarX using BglII and AvrII allows the first five amino acids of actagardine to be replaced with annealed oligonucleotides encoding alternative amino acids at any one of these five positions. Similarly, by digesting pAGvarX with BsrGI and AvrII, the amino acids in positions 8–19 can be replaced with alternative residues. Evaluation of pAGvarX as a vehicle to generate actagardine variants initially involved assessing the vector's ability to restore production of actagardine to A. garbadinensisΔgarA, a strain of A. garbadinensis in which the garA gene has been removed (see Experimental procedures). Although A. garbadinensisΔgarA lacks the ability to produce actagardine (Fig. 3, well 6) it contains all the biosynthetic genes required to process the prepropeptide to the mature lantibiotic and hence is a suitable host strain for introducing complementing plasmids encoding variants of the actagardine peptide.

The plasmid pAGvarX was introduced by conjugation into A. garbadinensisΔgarA. Apramycin-resistant exconjugants were identified and grown in GM1 production medium for 7 days. The supernatants were then collected and analysed by HPLC and LC-MS in addition to being tested for antibacterial activity against M. luteus. The analyses of supernatants from strains of A. garbadinensisΔgarA transformed with the plasmid pAGvarX (designated A. garbadinensisΔgarA::pAGvarX) confirmed the presence of biologically active compounds with the expected masses and retention times for actagardine and ala(0)-actagardine (Fig. 3, well 7). These compounds were absent in the supernatant of A. garbadinensisΔgarA.

The introduction of pAGvarX into A. garbadinensisΔgarA successfully restored the organism's ability to produce actagardines, although at lower levels than observed for the wild-type strain. These reduced levels of production may result from integration of pAGvarX at the phage ΦC31 attachment site. Previous studies have shown that while in some host organisms insertion into the ΦC31 attB site can have a neutral effect on the levels of antibiotic production, the levels in other actinomycete hosts are reduced (Baltz, 1998; Rodriguez et al., 2003). It is plausible that the observed drop in lantibiotic production compared with the wild-type organism is a consequence of differences in the integrity and stability of the mRNAs. Despite this decrease in production, the levels were sufficient to enable comparative evaluation of mutant copies of the actagardine peptide. The construction of a host strain encoding the respective garA mutation at the garA locus may lead to the production of the actagardine variant at levels more comparable to the wild-type organism. This could prove especially worthwhile for mutants demonstrating improved antibacterial activity identified in initial complementation screens.

Alanine scanning of the actagardine molecule

Alanine scanning was performed to further validate the engineered variant generation system and provide an understanding of the tolerance of the biosynthetic machinery to modifications of the amino acid sequence, Previous studies examining alterations to residues involved in bridge formation have predominantly resulted in either a complete loss of activity or a significant reduction in the level of activity of the resulting lantibiotic variant (Ottenwälder et al., 1995; Bierbaum et al., 1996; Cotter et al., 2006; Cooper et al., 2008). Given that the pattern of lanthionine bridges of actagardine is conserved among several lantibiotics, including mersacidin (Prasch et al., 1997), lacticin 3147-A2 (Ryan et al., 1999), plantaricin Wa (Holo et al., 2001), staphylococcin C55a (Navaratna et al., 1998) and haloduracin (McClerren et al., 2006), and the apparent significance of these bridges for conformational stability and retention of activity, we chose only to mutagenize residues not involved in bridge formation of the actagardine molecule.

Mutants encoding alanine at positions 2–5, 8, 10, 11, 13, 15 and 16 were constructed in pAGvarX and used to conjugate A. garbadinensisΔgarA. Fermentation samples of four independent exconjugants per mutant were assayed for antimicrobial activity and analysed by HPLC-MS. The results of HPLC-MS analysis (Fig. 4) showed that actagardine variants in three out of the four residues of ring A and residues 13 and 15 were produced. However, antibacterial activity against M. luteus was detected only in S2A and V15A variants. The V5A variant did not display any biological activity against M. luteus. The equivalent position in the closely related lantibiotic Michiganin A, active against C. michiganensis ssp. sepedonicus, is occupied by a leucine residue (Holtsmark et al., 2006). It is not known whether michiganin A is active against M. luteus or actagardine has any activity against Clavibacter so a comparison between the activity data of these compounds is not obvious. Production of variants with alterations to any of the residues within ring B (residues 7–12) was not detected. These results are in agreement with studies of engineered mutants of the lantibiotic lacticin 3147 (Cotter et al., 2006) and suggest that this region may not be suitable for a large-scale mutagenesis approach. Antibacterial activity was observed in two of the five variants (isolated from 2 l fermentations as described in supporting information), namely S2A and V15A. Although both of these mutants demonstrate reduced biological activity compared with the parent molecule, the retention of activity reveals that these residues are amenable to change and represent sites for future mutagenesis. The lack of antibacterial activity against M. luteus exhibited by the variants G3A, V5A and G13A suggests that these residues are important for activity and may not represent targets amenable to mutagenesis.

Figure 4.

Alanine scanning of the actagardine peptide. Summary of the variants of actagardine generated through the replacement of specific residues with an alanine. HPLC-MS analysis of fermentation broths confirmed the production of five of the 10 mutants and from these, two (S2A and V15A) demonstrated biological activity against Micrococcus luteus. A plus symbol denotes detection, while the absence of detection is denoted by a minus. Abu, 2-aminobutyric acid.

The approach described here offers a direct means of introducing mutations within the peptide encoding region and does not rely on the generation and selection of a double cross-over recombinant. Individual and multiple residues of the actagardine peptide can be replaced, deleted or inserted at specific positions by the introduction of annealed oligonucleotides encoding the desired mutation. The system also offers the potential of being applied in a random mutagenic approach. Extensive libraries of mutants carrying alterations to the structural gene could be generated through the annealing of degenerate oligonucleotides to the pAGvarX backbone. In summary, the results reported here highlight the versatility of the plasmid pAGvarX and demonstrate the flexibility of the actagardine biosynthetic machinery in recognising and processing variants of the GarA peptide. At a time when the incidence of resistance to current antibacterial agents is increasing at an alarming rate, this system offers an alternative approach to the rapid production and screening of new variants of actagardine with potential application to similar biosynthetic systems. Not only could this help identify compounds with improved antibacterial activity, but may also aid rational design approaches by unlocking information surrounding structure/activity relationships.

Experimental procedures

Strains and plasmids

Actinoplanes garbadinensis ATCC31049 was maintained by growing on medium ABB13 (Fitzgerald et al., 1998) at 30°C. S. lividans 1326, Escherichia coli BW2113/pIJ790 and E. coli ET12567/pUZ8002 were kindly supplied by Professor Mervyn Bibb (John Innes Centre, Norwich, UK) and manipulated as previously described (Kieser et al., 2000; Gust et al., 2004). E. coli DH10B (Invitrogen) was used for the routine propagation of plasmids. E. coli BT340 (CGSC#7629) used as described in the Redirect procedure (Gust et al., 2004) was obtained from the E. coli Genetic Stock Center, Yale University, USA. M. luteus ATCC4698 was used as the indicator organism in well diffusion assays.

The cosmid SuperCos1 and E. coli XL1-Blue MR cells used in the construction of the cosmid library were obtained from Stratagene. The plasmids pIJ773 (Kieser et al., 2000) and pMJCOS1 were generously provided by Professor Mervyn Bibb. Plasmids pUC19 and pLITMUS28 (New England Biolabs) were used for routine cloning while the plasmid pSET152 (Bierman et al., 1992) was used in the construction of pAGvarX.

The strains used to evaluate the in vitro antimicrobial activity of actagardine and deoxyactagardine were sourced as follows: Enterococcus faecium (VRE), is a clinical isolate recovered from a patient at Leeds General Infirmary and maintained in a culture collection belonging to the Division of Microbiology, University of Leeds; Staphylococcus aureus SH1000 (Horsburgh et al., 2002) was also obtained from the Division of Microbiology, University of Leeds while E. faecium (19579), Enterococcus faecalis (29212), Streptococcus pneumoniae (R6) and M. luteus (4698) were all obtained from the ATCC; Staphylococcus epidermidis (11047) was obtained from the National Collection of Type Cultures; S. aureus (EMRSA strain R33) was obtained from the Health Protection Agency.

Identifying and cloning of the actagardine biosynthetic gene cluster from A. garbadinensis ATCC31049

The DIG-labelled oligonucleotide, O/SBDIG-1, was designed by backtranslating the amino acid sequence of actagardine (Zimmerman et al., 1995) considering the codon usage for the genus Actinoplanes. Southern hybridization analysis of genomic DNA isolated from A. garbadinensis and digested using the restriction enzyme NcoI identified a ∼3 kb fragment that hybridized to O/SBDIG-1. The NcoI digest of the genomic DNA was repeated and DNA fragments of ∼3 kb were isolated and cloned into NcoI-digested pLITMUS28. The resulting plasmids were introduced into E. coli DH10B cells and then analysed by colony hybridization using the probe O/SBDIG-1. A hybridising colony was identified and submitted for sequence analysis. Sequencing revealed that this plasmid designated pLITAG01 consists of DNA encoding the lanA structural gene for actagardine (designated garA). The primers O/ACT08F and O/ACT09R were designed based upon sequence from pLITAG01. Using these primers in a PCR reaction together with DIG-labelled dNTPs (Roche Diagnostics) and pLITAG01 as a template, a 2296 bp DIG-labelled DNA fragment was generated and designated SBDIG-2.

A cosmid library was generated by cloning Sau3AI partially digested genomic DNA fragments (approximately 35–40 kb in size) isolated from A. garbadinensis into the cosmid SuperCos1 previously digested using BamHI and XbaI. This library was screened by colony hybridization using SBDIG-2. Twenty-five cosmids believed to hybridize to SBDIG-2 were selected and analysed via Southern hybridization using the probes O/SBDIG-1 and SBDIG-2. Nine cosmids derived from the genomic DNA library from A. garbadinensis hybridized to both probes. DNA was isolated from each cosmid and analysed by restriction enzyme digestion and sequencing using the primers T3 and T7. The cosmid CosAG14 was selected and subsequently fully sequenced.

Construction of the strain A. garbadinensis ΔgarA

First, the region of DNA from the cosmid CosAG14 encoding garA was replaced with the cassette SBdel-1. SBdel-1 consists of the apramycin-resistance gene [aac(3)IV] and oriT flanked by FLP recognition target sites and was obtained by PCR amplification using the plasmid pIJ773 as the template together with the primers O/SB50F and O/SB51R. The garA gene of CosAG14 was replaced with SBdel-1 and subsequently removed by FLP-mediated excision generating the cosmid, CosAG14ΔB, using the Redirect method (Gust et al., 2004). The second stage was to engineer the cosmid so that it could be introduced into A. garbadinensis via conjugation. This began by introducing CosAG14ΔB into the E. coli strain BW25113/pIJ790 by transformation. The ampicillin gene of CosAG14ΔB was then replaced with SBdel-2 following the Redirect protocol (Gust et al., 2004) generating the cosmid CosAG14ΔC. The cassette SBdel-2, like SBdel-1, houses the apramycin-resistance gene [aac(3)IV] and oriT flanked by FLP recognition target sites but was generated using the primers O/SB52F and O/SB53R together with the template pIJ773. CosAG14ΔC was used to transform electrocompetent cells of E. coli ET12567/pUZ8002 before being conjugated with A. garbadinensis. Exconjugants were obtained and subcultured through six successive rounds of growth in tryptic soy broth (TSB) (Sigma) without apramycin at 30°C and 250 r.p.m. Cells from culture 6 were plated onto medium 65 (DSMZ) and incubated at 30°C. After 5 days colonies were transferred and patched out over an area of approximately 1 cm2 onto medium 65. After 3 days of incubation at 30°C the patched cells were transferred to medium 65 containing apramycin at a final concentration of 50 μg ml−1. Following 72 h incubation at 30°C, cells sensitive to apramycin were selected and the respective patches used to inoculate 50 ml flasks containing 10 ml TSB and grown at 30°C and 250 r.p.m. for 4 days. Genomic DNA was prepared from each culture and analysed by PCR using oligonucleotides O/AGvar01bF and O/AGvar06r. PCR products from cultures that produced bands of the size consistent with the deletion of the garA gene were selected. In parallel, analysis of fermentation broths by HPLC demonstrated that these strains did not produce actagardine.

Construction of the strain A. garbadinensis ΔgarO

The strain, A. garbadinensisΔgarO, is a deletion mutant of A. garbadinensis in which the garO gene has been removed. This strain was constructed following the protocol as described for the construction of A. garbadinensisΔgarA except, in place of using SBdel-1, the garO gene of CosAG14 was initially replaced with the cassette ‘Del_garO’. The cassette Del_garO was generated by PCR using the primers O/ΔgarO1F and O/ΔgarO2R and the plasmid pIJ773 as a template. The procedure for isolating this mutant was identical to the one described previously for garA. The genotype of this mutant was confirmed by PCR and Southern hybridization analysis (results not shown).

Construction of plasmids designed to express actagardine variants

The DNA fragment containing the first base 3′ adjacent to the ORF lying upstream of garA and up to the leucine residue within the leader peptide of the garA gene (10 amino acids upstream of the mature peptide) was amplified by PCR using the primers O/AGvar01bF and O/AGvar02bR and pLITAG01 as a template. The primers were designed to introduce a flanking XbaI site at the 5′ end and a BglII site via a silent mutation at the 3′ leucine residue with respect to garA. This fragment was introduced into pUC19 previously digested using SmaI to yield pAGvar1. The region of DNA spanning from the 3′ end of garA to the adjacent downstream ORF was amplified by PCR using the primers O/AGvar05F and O/AGvar06R and pLITAG01 as a template. The primers were designed to introduce a flanking AvrII site after the stop codon of garA and an EcoRI site at the other end. The resulting PCR product was cloned into pUC19 previously digested using SmaI to yield pAGvar2. The plasmids pAGvar1 and pAGvar2 were then digested using XbaI and the insert from pAGvar1 recovered and cloned into pAGvar2. The correct orientation of the incoming fragment was determined by restriction analysis. The resulting plasmid pAGvar3 was subsequently digested using BglII and AvrII and ligated to the annealed oligonucleotides O/AGvar03F and O/AGvar04R [20 μl of each oligonucleotide (100 μM) mixed then heated to 95°C for 7 min followed by 65°C for 2 min and allowed to cool at room temperature] generating pAGvar4. The plasmid pAGvar4 was subsequently digested using EcoRI and XbaI and the resulting ∼620 bp fragment introduced into pSET152 (Bierman et al., 1992) previously digested using EcoRI and XbaI yielding the vector pAGvarX.

Construction of vectors for heterologous expression

The cosmid pMJCOS1 is a derivative of SuperCos1 (Stratagene) in which the gene encoding for neomycin resistance has been replaced by a cassette (HEapra) that includes DNA encoding an oriT, attP, integrase (int) and an apramycin-resistance gene [aac(3)IV]. The cassette HEapra was isolated by digesting pMJCOS1 with SspI and recovering the DNA from an agarose gel. This cassette together with CosAG14 was used to generate the cosmid CosAG14HEapra following the Redirect protocol as described by Gust (Gust et al., 2004). The cosmid CosAG14HEapra was subsequently introduced into S. lividans via conjugation. Apramycin-resistant exconjugants of S. lividans::CosAG14HEapra were isolated.

Conjugation procedure

Intergeneric conjugation between E. coli and Actinoplanes sp. was adapted from the procedure described by Heinzelmann et al. (2003), using the strain E. coli ET12567/pUB8002 (Kieser et al., 2000) as the donor. E. coli ET12567/pUB8002 cells transformed with the respective plasmid and grown overnight [10 ml Luria broth (LB, Sigma) supplemented with the required antibiotics] in a 50 ml centrifuge tube at 37°C and 250 r.p.m. were used to inoculate (5% inoculum) a 50 ml tube containing 15 ml of LB supplemented as before. This culture was incubated at 37°C and 250 r.p.m. until the OD600 had reached approximately 0.6. The cells were pelleted by centrifuging for 10 min at 4000 r.p.m. (Sorvall Legend RT) then washed twice using 10 ml of TSB per wash. After the final wash the cell pellet was resuspended in 1.5 ml TSB. Meanwhile, cells of A. garbadinensis grown in 40 ml TSB (Sigma) in 250 ml Erlenmeyer flasks at 30°C and 250 r.p.m. for 6 and 7 days were pooled. The mycelia were left to settle for ∼1 h, then a 40 ml aliquot (removed from the top of the flask) was centrifuged at 4000 r.p.m. for 20 min. The cell pellet was resuspended in 20 ml TSB and sonicated for 5 × 15 second bursts at an amplitude of 13 (Sanyo Soniprep 150). 1.5 ml of sonicated A. garbadinensis recipient cells was mixed with 1.5 ml of E. coli donor cells. 400 μl aliquots were plated onto ABB13 and placed at 30°C. Following overnight incubation each plate was overlaid with soft nutrient agar (50% nutrient agar, Oxoid) containing apramycin and spectinomycin to a final concentration of 60 μg ml−1 each. The plates were incubated at 30°C for up to 14 days. Exconjugants were picked and patched out over an area of ∼1 cm2 onto an ABB13 plate supplemented with the appropriate antibiotics and incubated as before. Intergeneric conjugation between E. coli and S. lividans was performed following the procedure described by Kieser (Kieser et al., 2000).

Fermentation of exconjugants

Agar plugs from patched exconjugants were used to inoculate 50 ml conical flasks containing 8 ml of either AAS seed medium (soluble starch 10 g, glucose 10 g, Bacto-peptone 5 g, dry corn steep liquor 1 g, yeast extract 2 g, reverse osmosis water to 1 l and pH adjusted to 7) for fermentations of A. garbadinensis strains or TSB for the growth of S. lividans strains with two glass beads and supplemented with apramycin and nalidixic acid (if required), both to a final concentration of 50 μg ml−1. The cultures were incubated at 30°C and 250 r.p.m. for 72–96 h after which 250 μl from each was used to inoculate a 50 ml flask containing 8 ml GM1 [meat extract 4 g, peptone 4 g, NaCl 2.5 g, yeast extract 1 g, soyflour (Cargill) 10 g, glucose 25 g, CaCO3 5 g, reverse osmosis water to 1 l and pH adjusted to 7.6]. These cultures were incubated at 30°C and 250 r.p.m. 500 μl whole broth aliquots were removed following 3, 5 and 7 days of incubation and the supernatants (collected by centrifugation at 14 000 r.p.m. for 10 min, IEC Micromax) analysed by HPLC-MS.

Well diffusion assays

Micrococcus luteus ATCC4698 was inoculated from frozen stock into 10 ml Mueller-Hinton Broth (Oxoid, CM0405) and grown overnight at 30°C and 200 r.p.m. One millilitre of this culture was used to inoculate 300 ml of molten Mueller-Hinton Agar (Oxoid, CM0337) kept at 50°C, which was then dispensed into Petri dishes as 25 ml aliquots. Wells (6 mm diameter) were made into the seeded agar and subsequently loaded with 50 μl samples. The bioassay plate was placed into a biological safety cabinet until the loaded samples had diffused, at which point the plates were transferred to a 30°C incubator and left for 48 h.



Detection of actagardine and its variants

The HPLC analyses were performed using a Hewlett Packard 1050 series HPLC system with an Agilent Zorbax SB-C18, 4.6 mm × 150 mm, 5 μ column. The HPLC-MS analyses were performed on a HPLC system (as described) linked to a Micromass Platform LC operated with MassLynx version 3.5 software. The instrument methods are described in the supporting information.

Minimum inhibitory concentration determinations

The in vitro antimicrobial activity, recorded as minimum inhibitory concentrations, were determined by a broth microdilution assay as recommended by the National Committee for Clinical Laboratory Standards (NCCLS, 2003). Susceptibility testing of all aerobic organisms with the exception of S. pneumoniae was performed by twofold serial antibiotic dilutions in Mueller-Hinton broth (Oxoid) or in brain heart infusion broth (Oxoid. S. pneumonia only) supplemented with 50 μg ml−1 Ca2+. Actively growing broth cultures were adjusted to the turbidity of a 0.5 McFarland standard (∼108 cfu ml−1), then diluted a further 1:100 in fresh sterile broth for a final inoculum of approximately 106 cfu ml−1 (105 cfu/well). The assays were performed in duplicate in sterile 96-well microtitre plates (Fisher Scientific, Leicestershire, UK) in a total volume of 200 μl (160 μl broth, 20 μl antimicrobial agent, 20 μl inoculum) in a concentration range from 64 to 0.06 μg ml−1. Plates were incubated aerobically, shaking, for 18–20 h at 37°C with the minimum inhibitory concentration defined as the lowest concentration of drug that prevented visible growth.


We thank Professor Mervyn Bibb and Dr Sean O'Rourke of the John Innes Centre, Norwich for the provision of plasmids and strains. We also thank the sequencing facility at the Department of Biochemistry, University of Cambridge, Professor Ian Chopra of the University of Leeds for the provision of strains and gratefully acknowledge funding awarded to Novacta Biosystems by the East of England Development Agency.