An asparagine oxygenase (AsnO) and a 3-hydroxyasparaginyl phosphotransferase (HasP) are involved in the biosynthesis of calcium-dependent lipopeptide antibiotics

The calcium-dependent antibiotics (CDAs) from Streptomyces coelicolor (Hojati et al., 2002; Kempter et al., 1997) belong to the group of structurally related acidic lipopetide antibiotics, which include A54145 (Fukuda et al., 1990; Miao et al., 2006), daptomycin (Baltz et al., 2005; Miao et al., 2005; Raja et al., 2003), friulimicins and amphomycins (Vértesy et al., 2000). All of these nonribosomally biosynthesized lipopeptides contain N-terminal fatty acid side chains, which is a trans-2,3-epoxyhexanoyl moiety in the case of CDA (Fig. 1), along with decapeptide lactone or lactam cores. Within the decapeptide cores are a number of conserved amino acids, including acidic residues responsible for co-ordination of calcium ions, which are essential for antimicrobial activity. Notably, daptomycin was recently approved for clinical use, becoming the first new structural class of natural antibiotic to reach the clinic in over 30 years (Baltz et al., 2005; Raja et al., 2003). As a result of this, there has been considerable interest in establishing the biosynthetic origins of the acidic lipopeptide group of antibiotics, with the specific aim of engineering new lipopeptide variants with improved therapeutic properties. To this end we have focused on engineering the biosynthesis of CDAs, using adenylation domain active site modifications to replace Asp at position 7 in CDAs with Asn (Uguru et al., 2004) and a mutasynthesis approach to replace the D-hydroxyphenylglycine (D-HPG) at position 6 with alternative arylglycines (Hojati et al., 2002).

(32K): Fig. 1. Structures of CDA variants. The b series CDAs contain L-Trp at position 11, whilst the a series possess Z-dehydrotryptophan residues at position 11 (Hojati et al., 2002; Kempter etal., 1997). CDA5a and 6a are novel biosynthetically engineered CDAs possessing Asn at position 9.

In addition to D-HPG, CDA contains a number of other non-proteinogenic amino acids, including L-3-methylglutamic acid (3-MeGlu; at position 10), which is also present at the same relative position in the decapeptide cores of daptomycin and A54145 (Milne et al., 2006), and a C-terminal Z-dehydrotryptophan residue (Hojati et al., 2002). CDA variants which contain either (2R,3S)-3-hydroxyasparagine (3-OHAsn) or (2R,3S)-3-phosphohydroxyasparagine (3-OPAsn) at position 9 are also known. Whilst 3-OHAsn is common to a number of peptide antibiotics, including A54145 (Fukuda et al., 1990; Miao et al., 2006), katanosin B (Kato et al., 1988) and ramoplanin (Walker et al., 2005; McCafferty et al., 2002), 3-phosphohydroxyasparagine has not been found elsewhere in nature, to date.

Previously, we analysed the cda biosynthetic gene cluster and identified a gene, SCO3236, which encodes a protein that shows high similarity to clavaminate synthase (CAS), an Fe(II)/2-oxoglutarate-dependent oxygenase from Streptomyces clavuligerus. CAS catalyses the β-hydroxylation of the Arg side chain of deoxyguanidinoproclavaminic acid, and subsequent cyclization and dehydrogenation steps during the biosynthesis of clavulanic acid (Baldwin et al., 1993; Salowe et al., 1990). On this basis, we suggested that SCO3236 encodes an asparagine oxygenase (AsnO) that was predicted to catalyse the β-hydroxylation of an Asn residue in a CDA precursor to generate the 3-OHAsn found in CDA (3a, 3b, 4a and 4b variants) (Fig. 1).

β-Hydroxylation of amino acids is a common phenomenon in the biosynthesis of nonribosomal peptides (Chen et al., 2001, 2002). Indeed the subsequent processing of the amino acid β-hydroxyl groups by further oxidation (Chen et al., 2002), glycosylation (Lu et al., 2004), macrolactonization (Kato et al., 1988; Walker et al., 2005; McCafferty et al., 2002), methylation (Miao et al., 2006) or in the case of CDA, phosphorylation, serves to increase the structural diversity of nonribosomal peptides and related products, resulting in a wide range of biological activities. Of the many β-hydroxylated amino acids and derivatives found in nonribosomal peptides, some arise as a result of the oxidation of free amino acids prior to peptide assembly (Yin & Zabriskie, 2004; Ju et al., 2004; Haltli et al., 2005). For example, in the biosynthesis of viomycin and mannopeptimycin, L-arg and the non-proteinogenic amino acid L-enduracididine are hydroxylated as free amino acids by non-haem Fe(II)/2-oxoglutarate-dependent oxygenases VioC (Yin & Zabriskie, 2004; Ju et al., 2004) and MppO (Haltli et al., 2005), respectively. In addition, there are a number of other non-haem Fe(II)/2-oxoglutarate-dependent oxygenases which are known to hydroxylate specific amino acid residues within ribosomal polypeptides. For example, mammalian asparaginyl hydroxylases EGFH and FIH are known, which catalyse the posttranslational β-hydroxylation of Asn residues within epidermal growth factor (EGF) domains (Jia et al., 1994) and the hypoxia-inducible factor (HIF) C-terminal transactivation domains (Hewitson et al., 2002). Finally, several monooxygenase enzymes have been characterized that function on amino acid thioesters that are tethered to carrier proteins via flexible phosphopantetheine prosthetic groups (Chen et al., 2001, 2002). These include haem protein hydroxylases (ORF20, NovI and NikQ), which are responsible for the β-hydroxylation of peptidyl carrier protein (PCP)-tethered tyrosine and histidine residues during the biosynthesis of chloroeremomycin, novobiocin (Chen et al., 2001) and nikkomycin (Chen et al., 2002).

Downstream of asnO, in the CDA gene cluster, is a gene, SCO3234, which encodes a protein that shows sequence similarity to a number of ATP-dependent aminoglycoside phosphotransferases, including SpcN, which has been shown to phosphorylate spectinomycin as part of the host self-resistance mechanism in the producer strain Streptomyces flavopersicus (Lyutzkanova et al., 1997). Based on this, we suggested that the SCO3234 gene product is a putative 3-hydroxyasparagine phosphotransferase (HasP), which phosphorylates the 3-hydroxyasparaginyl residue of CDA3a/b and CDA4a/b to produce the 3-phosphohydroxyasparaginyl residues found in CDA1b and CDA2a/b (Fig. 1). In this paper, we aim to determine the in vivo function of the asnO and hasP gene products, and test the hypothesis that these products are involved in the hydroxylation and subsequent phosphorylation of the asparaginyl residues in CDA. It was envisaged that this insight would provide a means to control the functionality of the Asn-derived residues in CDA which could be further utilized to engineer other therapeutically relevant acidic lipopeptides based around the A54145 and daptomycin biosynthetic templates.

Bacterial strains, plasmids and culture conditions.
The bacterial strains and plasmids used in this study are listed in Supplementary Table S1. Escherichia coli strains were propagated routinely in LuriaBertani (LB) broth or on agar plates at 37 °C. Antibiotics were used in selective media at the standard concentrations (Sambrook et al., 2000). For DNA manipulation, Streptomyces strains were routinely propagated in YEME liquid media or on R2YE agar plates at 30 °C (Kieser et al., 2000).

DNA preparation, manipulation and sequencing.
S. coelicolor cosmid ScE29 encompassing SCO3232SCO3249 of the CDA biosynthetic gene cluster was obtained from the John Innes Centre, Norwich, UK, and was isolated from E. coli using the Qiagen Midiprep Kit. Small-scale plasmid preparation was carried out using the alkaline lysis method (Birnboim & Doly, 1979), unless a higher quality of DNA was required for sequencing, in which case DNA was prepared using a Midiprep Kit. Streptomyces genomic DNA was prepared as described elsewhere (Kieser et al., 2000). Restriction enzymes were purchased from Roche, New England Biolabs and MBI Fermentas. Alkaline phosphatase, T4 DNA ligase and Pwo polymerase were all purchased from Roche. All enzymes were used according to the manufacturers' recommendations, unless otherwise specified.

PCR amplification was carried out on a GeneAmp PCR system 2400 (Perkin-Elmer) using 100 pmol of each primer, ∼100 ng cosmid or 200 ng genomic DNA template per 100 µl reaction, and 1 unit of the high fidelity Pwo polymerase. PCR reactions also contained 5 % (v/v) DMSO to overcome potential secondary structure formation in the GC-rich Streptomyces DNA. Following amplification, PCR products were purified using the Qiaquick extraction kit (Qiagen). PCR products and primers used in this work are detailed in Supplementary Table S2. DNA sequencing of PCR products was carried out by the chain-termination method using the ABI/PRISM dye terminator cycle sequencing core kit (Perkin-Elmer). E. coli plasmid construction and transformations where carried out using standard procedures (Sambrook et al., 2000).

Construction of the asnO deletion cassette.
An upstream region of SCO3236 was PCR-amplified using primers AsnOF2 and AsnOR2, containing XhoI and PstI restriction sites, respectively (Supplementary Table S2), with ScE29 as a template to generate a 1558 bp amplicon (fragment A) containing the first 226 bp of SCO3236. Fragment A and pMT3000 (Paget et al., 1994) were digested with XhoI and PstI, and ligated to form plasmid pJN9a. Primers AsnOF3 and AsnOR3, containing EcoRI and XhoI restriction sites, were used in a PCR reaction with ScE29 as a template to generate a 1662 bp amplicon (fragment B) containing the last 227 bp of SCO3236. Fragment B and pMT3000 were digested with EcoRI and XhoI, and ligated to form pJN9b. Fragment A was excised from pJN9a via PstI/XhoI and ligated into similarly digested pJN9b to give the plasmid pJN10 containing the asnO (AB) deletion construct (Fig. 2). This was excised from pJN10 via the flanking BglII sites and ligated into the BamHI site of pKC1132 (Bierman et al., 1992), resulting in pJN12, which was used for delivery of the truncated asnO into the chromosome of S. coelicolor 2377 (Hopwood & Wright, 1983) and MT1110 (Hindle & Smith, 1994) strains.

(15K): Fig. 2. The generation of S. coelicolor ΔasnO and ΔhasP mutant strains. Plasmids pLG4 and pJN12, which contain truncated hasP (A'B') and asnO (AB) deletion constructs, respectively, were used to perform standard double-crossover gene replacements, which resulted in mutants possessing in-frame truncated hasP and asnO genes, respectively.

Construction of the hasP deletion cassette.
An upstream region of SCO3234 was PCR-amplified using primers HasPF1 and HasPR1, containing PstI and XbaI restriction sites, respectively, with ScE29 as a template to generate a 1830 bp amplicon (fragment A') containing the first 70 bp of SCO3234. Fragment A' and pMT3000 were digested with PstI and XbaI, and ligated to form pMUT1. Primers HasPF2 and HasPR2, containing XbaI and XhoI restriction sites, respectively, and the template ScE29 were used in the PCR amplification of a 1939 bp amplicon (fragment B') containing the last 215 bp of SCO3234. Fragment B' and pMT3000 were digested with XbaI and XhoI, and ligated to form pMUT2, excised from pMUT2 via XbaI/XhoI, and ligated into similarly digested pMUT1 to give the plasmid pLG3 containing the hasP deletion construct (A'B') (Fig. 2). The deletion construct was excised from pLG3 via the flanking BglII sites and ligated into the BamHI site of pMAH (Paget et al., 1994), resulting in pLG4, which was used for delivery of the truncated hasP into the chromosome of S. coelicolor 2377 and MT1110 strains.

Deletion of asnO and hasP from S. coelicolor MT1110 and 2377.
ΔasnO and ΔhasP delivery plasmids pJN12 and pLG4 were first passed through E. coli strain ET12567 to remove Dam/Dcm methylation. The plasmids were then alkaline-denatured as described elsewhere (Oh & Chater, 1997) and introduced into Streptomyces protoplasts using the PEG-mediated transformation method (Kieser et al., 2000). Primary integrants of pJN12 and pLG4 were selected for with apramycin and hygromycin, respectively, and grown for two generations on non-selective agar to allow double-crossover events to occur. Colonies were screened for the loss of antibiotic resistance by replica plating. Total genomic DNA was isolated from antibiotic-sensitive recombinants, and used as a template for PCR amplification using primers flanking the deleted region. Primers 29.05seqF2 and 29.05R2 amplified the region of asnO, giving products of size 1044 bp from strains that had not undergone gene replacement and 495 bp from strains that had a truncated form of asnO (Supplementary Fig. 1A). Similarly, primers HasPKO1 and HasPKO2 were used to amplify the region of hasP, resulting in products of size 1818 bp in the strains that had not undergone gene replacement and 1200 bp in the strains that contained a truncated form of hasP (Supplementary Fig. 1B). DNA sequencing of PCR products corresponding to truncated forms of asnO and hasP was also carried out, confirming that an in-frame deletion of 549 bp of the asnO and 618 bp of the hasP genes had occurred (data not shown). Two S. coelicolor 2377-ΔasnO strains and four S. coelicolor MT1110-ΔasnO strains were generated. In addition three S. coelicolor 2377-ΔhasP and S. coelicolor MT1110-ΔhasP strains were also generated.

CDA production, extraction and analysis.
The MT1110 parent, ΔasnO and ΔhasP mutant strains were grown for 6 days at 30 °C in SV2 liquid medium (Hojati et al., 2002). S. coelicolor 2377 parent, ΔasnO and ΔhasP mutant strains were grown for 6 days at 30 °C in SM22 liquid medium (Hojati et al., 2002). The culture supernatants were acidified to pH 2 with 0.5 M HCl and passed through Varian MegaBond Elute SPE cartridges (Varian) as described previously (Hojati et al., 2002). The CDA extracts were then eluted with 40 % acetonitrile in H₂O and dried under reduced pressure. Liquid chromatography (LC)-MS analysis of the crude extracts was carried out on a Micromass LCT orthogonal acceleration time of flight mass spectrometer equipped with an electrospray ionization source run in positive mode, combined with a Waters 2790 separation module. Gradient elution was carried out on a Luna C18 150x4.6 mm 3 µm analytical column (Phenomenex). Solvent A was 0.1 % formic acid and solvent B was acetonitrile/0.1 % formic acid. A gradient of 20 % B rising to 70 % over 10 min, followed by 100 % B held for 5 min at a flow rate of 1 ml min¹ was applied.

Complementation of S. coelicolor MT1110-ΔasnO with plasmid-borne asnO.
To construct a Streptomyces expression plasmid bearing asnO, the gene sequence was PCR-amplified from cosmid ScE29 with primers 29.05F2 and 29.05R2, containing restriction sites for XhoI and NdeI, respectively. The purified product was restricted and ligated into E. coli expression vector pET-15b, resulting in the 6708 bp plasmid pJN14. The asnO gene was excised from pJN14 by digestion with BamHI and XhoI, and ligated into restriction-digested pIJ6021 (Takano et al., 1995), resulting in the 8904 bp plasmid pJN15. The ligation products were used to transform Streptomyces lividans 1326, and positive transformants were selected by growth on medium containing kanamycin. pJN15 was isolated from transformants, checked by restriction analysis, and introduced into the S. coelicolor MT1110-ΔasnO mutant by protoplast transformation. The plasmid-complemented mutant was cultivated in SV2 medium supplemented with kanamycin, and thiostrepton was added at a concentration of 10 µg ml¹ to induce expression of AsnO. CDA was extracted from the culture supernatant and analysed by LC-MS, as described above.

Cell-based bioassay of Asn-containing CDAs.
Spores of S. coelicolor parent and ΔasnO mutant strains were used to inoculate nutrient agar plates in the presence and absence of Ca²⁺ (12 mM). After 2 days' incubation at 30 °C, an overnight culture of the CDA-sensitive Bacillus mycoides indicator strain was used to overlay the plates in soft nutrient agar (per litre: 13 g nutrient broth, 7 g Bactoagar). Calcium-dependent bioactivity was observed as a zone of inhibition after one night of incubation at 30 °C (Fig. 5A).

(60K): Fig. 5. Bioactivity of Asn-containing CDAs. (A) Cell bioassay of 2377-ΔasnO strains overlaid with B. mycoides as the indicator strain in the presence of Ca2+ (12 mmol). Strains 13 and 29 are ΔasnO mutant strains, whilst strains 23 and 34 had been transformed with the ΔasnO deletion construct, but proved not to possess the asnO deletion and therefore retained the same phenotype as the parent strain 2377. The parent 2377 strain is shown in the centre of the plate. No zones of inhibition were present in the control plates which contained no Ca2+. (B) Well-plate bioassay with purified CDA6a (100 µg in 100 µl) and M. luteus as the indicator strain, in the presence of Ca2+ (16 mM). Again, zones of inhibition were completely absent in the control plates which possessed no Ca2+.

Production and isolation of CDA6a from 2377-ΔasnO.
Sterile SM22 liquid medium (2 l) was inoculated with S. coelicolor 2377-ΔasnO and then incubated at 28 °C and 180 r.p.m. for 5 days. The cells were removed by centrifugation and the CDA6a in the resulting supernatants was extracted onto Varian C-18 Bond Elute SPE cartridges (2 g), as described previously (Hojati et al., 2002). Purification of the crude extracts was achieved by preparative and semi-preparative reversed-phase HPLC: Phenomenex C-18 10 µm, 250x21.2 mm and 5 µm, 250x10 mm columns; solvent A H₂O with 0.1 % HCO₂H; solvent B acetonitrile with 0.1 % HCO₂H. The flow rates were 20 and 5 ml min¹, respectively, with a starting gradient of 0 % B and 100 % A, increasing to 100 % B over 30 min and then held for a further 5 min. Fractions for the peak containing mainly CDA6a (retention time ∼17.3 min) were pooled and evaporated under reduced pressure to give CDA6a (1.0 mg, ∼90 % pure).

Well-based bioassay of CDA6a.
CDA-sensitive strain Micrococcus luteus was grown overnight at 30 °C in Oxoid nutrient broth (5 ml). This culture was used to seed Oxoid nutrient agar plates in the presence or absence of added Ca²⁺ (16 mM). Agar plugs were cut from the dried plates, creating wells into which a sterile aqueous solution of CDA6a (∼100 µg in 100 µl) was added. The plates were then incubated overnight at 30 °C (Fig. 5B).

AsnO and HasP sequence analysis
BLAST sequence similarity searches on the SCO3236 (asnO) and SCO3234 (hasP) gene products were performed. In the case of HasP, no recently deposited protein sequences, in addition to the aminoglycoside phosphotransferases noted earlier (Hojati et al., 2002), were identified which might provide insight into the function of the putative phosphotransferase. On the other hand, several new sequences, in addition to CAS, were identified which exhibit high similarity to the putative asparagine oxygenase (AsnO). For example, LptL possesses 45 % identity across 310 out of 330 amino acid residues, and is predicted to be involved in the hydroxylation of the Asn residue at position 3 of A54145 in Streptomyces fradiae (Miao et al., 2006). In addition L-enduracididine hydroxylase, VioC (Yin & Zabriskie, 2004; Ju et al., 2004), exhibits 33 % identity across a 292 amino acid alignment with AsnO, and the L-arginine hydroxylase MppO (Haltli et al., 2005) possesses 31 % identity across 300 amino acid residues. All of these proteins possess the His-1 (HXE) and the C-terminal His-3 (DNXXXXH) motifs which make up the Fe(II) coordination sphere, which supports their known or putative function as Fe(II)/2-oxoglutarate-dependent oxygenase enzymes (Haltli et al., 2005; Khaleeli et al., 2000).

Generation of S. coelicolor ΔasnO and ΔhasP mutants
In order to establish the role of the asnO and hasP gene products, both genes were separately deleted from the S. coelicolor chromosome using a standard double-crossover recombination approach (Fig. 2) (Kieser et al., 2000). The successful generation of the required in-frame mutants was confirmed by PCR amplification across the truncated region (Supplementary Fig. S1), followed by DNA sequencing. Knockout experiments were performed in both 2377 and MT1110 strains, since these strains have different growth-media requirements and produce different profiles of CDAs. The combined information can thus provide valuable complementary information which can assist in validating the results of the separate experiments.

Analysis of S. coelicolor ΔhasP and ΔasnO
The S. coelicolor 2377 parent strains and 2377-ΔhasP mutant strains were grown in liquid culture, using media and conditions which we have previously shown (Hojati et al., 2002) result in the production of phosphorylated CDAs. The culture supernatants were analysed by LC-MS, which confirmed the previous observation (Hojati et al., 2002) that the parent strain 2377 produces D-3-phosphohydroxyasparagine-containing peptide CDA2b as the major product, along with a minor amount of D-3-hydroxyasparagine variants CDA4b and CDA3b (Fig. 3A). In contrast, the 2377-ΔhasP mutants produced only the non-phosphorylated variants CDA3b and CDA4b under identical growth conditions (Fig. 3B). Indeed, repeated fermentation and analysis of several 2377-ΔhasP mutant strains clearly revealed a complete absence of CDA2b, or any other phosphorylated CDAs. This supports the hypothesis that the gene product HasP is the CDA 3-hydroxyasparaginyl phosphotransferase. Analysis of the parent strain MT1110, grown on a variety of different media, showed only D-3-hydroxyasparagine-containing CDAs and failed to reveal any phosphorylated CDAs. Comparison of the phenotype change between MT1110 and MT1110-ΔhasP strains was therefore not possible in this case.

(22K): Fig. 3. LC-MS analysis of S. coelicolor 2377 parent and 2377-ΔhasP mutant strains. (A) S. coelicolor 2377 parent strain shows CDA2b ([M+H]+ found 1577.5, requires 1577.5; [M+Na]+ found 1599.5, requires 1599.5), CDA4b ([M+H]+ found 1497.7, requires 1497.5; [M+Na]+ found 1519.7, requires 1519.5), CDA3b ([M+H]+ found 1483.7, requires 1483.5; [M+Na]+ found 1505.7, requires 1505.5). The LC-MS of CDA2b standards also gave a very broad peak, with a retention time (Rt) of 10.512.0 min. This was due to the multiple ionization states of the phosphorylated peptide and the fact that it was not possible to use a buffered eluant, because this interferes severely with mass detection on LC-ESI-MS. (B) S. coelicolor 2377-ΔhasP. Production of phosphorylated CDA2b is completely abolished in this mutant, which produces only CDA4b ([M+H]+ found 1497.5; [M+Na]+ found 1519.5) and CDA3b ([M+H]+ found 1483.4; [M+Na]+ found 1505.4).

S. coelicolor MT1110 and MT1110-ΔasnO strains were grown in liquid media under identical conditions. LC-electrospray ionization (ESI)-MS analysis of MT1110 revealed the 3-MeGlu-containing peptide CDA4a as the major product along with a minor amount of the Glu-containing CDA3a (Fig. 4A). The mutant MT1110-ΔasnO, however, does not produce any CDA that is common to the MT1110 or 2377 strains (Hojati et al., 2002; Kempter et al., 1997), but instead exhibits a major new product with a retention time of 6.90 min, which exhibited protonated, sodiated and potassiated ([M+H]⁺, [M+Na]⁺ and [M+K]⁺) molecular ions in the ESI-MS consistent molecular weight (mw) 1478.5 Da (Fig. 4B). This CDA variant was therefore assigned as a non-hydroxylated Asn-containing variant, CDA6a (Fig. 1), which is related to CDA4a by the loss of the oxygen (1494.516.0 Da). Similarly, the S. coelicolor 2377 parent and 2377-ΔasnO mutant strains were grown in liquid culture. As above, the parent strain 2377 produced CDA2b as the major product with a minor amount of CDA4b and 3b. In contrast, the mutant 2377-ΔasnO produced none of the previously characterized CDAs. Instead, two new CDA products were identified. The faster-eluting compound, with an LC retention time of 6.90 min, exhibited protonated, sodiated and potassiated molecular ions in the ESI-MS consistent with CDA6a. In addition, a slower-eluting variant (R_t=6.79) was observed with molecular ions that indicate mw 1464.5 (Fig. 4C). This product corresponds to a non-hydroxylated variant of CDA3a (1480.516.0 Da), which we have assigned as CDA5a (Fig. 1).

(24K): Fig. 4. LC-MS analysis of MT1110 and ΔasnO mutants. (A) S. coelicolor MT1110 produces CDA3a ([M+H]+ found 1481.9, requires 1481.5; [M+Na]+ found 1503.8, requires 1503.5; [M+K]+ found 1519.8, requires 1519.5), CDA4a ([M+H]+ found 1495.9, requires 1495.5; [M+Na]+ found 1517.8, requires 1517.5; [M+K]+ found 1533.8, requires 1533.5). (B) S. coelicolor MT1110-ΔasnO produces CDA6a ([M+H]+ found 1479.0, requires 1479.5; [M+Na]+ found 1501.0, requires 1501.5; [M+K]+ found 1517.1, requires 1517.5). (C) S. coelicolor 2377-ΔasnO produces CDA5a ([M+H]+ found 1465.3, requires 1465.5; [M+Na]+ found 1487.2, requires 1487.5; [M+K]+ found 1503.2, requires 1503.5) and CDA6a ([M+H]+ found 1479.2; [M+Na]+ found 1501.2; [M+K]+ found 1517.2). Rt=retention time (min).

It is thus clear from these complementary results that deletion of asnO abolishes the normal CDA production profile of 3-OHAsn- and 3-OPAsn-containing peptides, and instead results in the production of two previously uncharacterized Asn-containing CDA variants (designated CDA5a and 6a).

Expression of plasmid-borne asnO in the MT1110-ΔasnO mutant
In order to gain further confirmation that AsnO is responsible for the hydroxylation of the Asn residue in CDA, the MT1110-ΔasnO strain was complemented by transformation with a thiostrepton-inducible asnO expression vector derived from the Streptomyces plasmid pIJ6021. The plasmid containing MT1110-ΔasnO mutants was grown on media containing the inducer thiostrepton. LC-MS analysis of the extracts revealed the same CDA profile as that of the MT1110 parent strain, with the production of 3-OHAsn-containing CDA4a and CDA3a being re-established. The results of the asnO knockout and complementation experiments are thus entirely consistent with the AsnO, a single polypeptide, functioning as an asparaginyl oxygenase.

Bioactivity of Asn CDAs
Parent and ΔasnO mutant strains of S. coelicolor 2377 and MT1110 were each separately incubated with and without Ca²⁺, and overlaid with the CDA-sensitive strain B. mycoides. A calcium-dependent zone of inhibition was seen around the two S. coelicolor 2377-ΔasnO strains, indicating that Asn-containing CDAs retained calcium-dependent antimicrobial activity (Fig. 5A). Similarly, all four of the S. coelicolor MT1110-ΔasnO strains exhibited calcium-dependent antibiosis with B. mycoides (data not shown). To further confirm that the bioactivity resulted from Asn-containing CDAs, 2377-ΔasnO was cultivated on a large scale (2 l), and the resulting CDA6a was purified from the culture supernatant using preparative and semi-preparative HPLC. High-resolution ESI-MS (m/z 1479.5333, C₆₇H₇₉N₁₄O₂₅ requires 1479.5340) and UV spectroscopy (λ_max 347 nm in H₂O) characteristic of Z-dehydrotryptophan-containing a-series CDA (Hojati et al., 2002) further confirmed the structure of CDA6a. Well-plate bioassays of CDA6a with M. luteus as the indicator stain showed significant calcium-dependent zones of inhibition (Fig. 5B). Given that only ∼1 mg of CDA6a could be purified from a 2 l culture of S. coelicolor 2377-ΔasnO, it was not practical to obtain sufficient quantities of CDA6a to determine accurate MIC values, for comparison with the other known CDAs. Nevertheless, from the size of the zones of inhibition and the concentrations of CDAs used, it is possible to conclude that CDA6a is less active than the corresponding 3-OHAsn-containing CDA4a and CDA3a (Milne et al., 2006).

This work clearly shows that AsnO and HasP are involved in the biosynthesis of 3-OHAsn and 3-OPAsn in CDA. However, the results do not indicate directly whether the hydroxylation and phosphorylation reactions take place before, during or after peptide assembly. Despite this, clues about the true nature of the substrates for AsnO and HasP can be derived by analysing the amino acid specificity conferring code of the module 9 adenylation domain (A domain) of the nonribosomal peptide synthetase (cdaPS2) responsible for activating Asn or its derivatives. Indeed, the 10 essential active site residues of the module 9 A domain (DLTKIGEVGK) within CdaPS2 have similarity to many other Asn activating A domains from different sources (Stachelhaus et al., 1999; Challis et al., 2000). This indicates that the module 9 A domain will be able to activate Asn, and indeed this must necessarily be the case given the production of Asn-containing lipopeptides CDA5a and 6a in the ΔasnO mutants. Furthermore, the module 9 A domain might possess broad substrate specificity, allowing the activation of 3-OHAsn, which is only slightly larger than Asn. On the other hand, it is unlikely that the A domain could activate both 3-OPAsn and Asn, given the very large difference in molecular volume and polarity of these two substrates. Furthermore, we have shown that it is possible to change the active site of the Asp-7 A domain (at the positions underlined: DLTKIGAVNK) to match exactly the sequence of the module 9 (Asn) A domain (DLTKIGEVGK). The resulting mutant produces a new lipopeptide product, CDA-7N, with Asn at position 7 instead of Asp (Uguru et al., 2004). Conceivably, if free Asn was hydroxylated to 3-OHAsn prior to A-domain activation, then this experiment could have resulted in CDAs with 3-OHAsn or 3-OPAsn at position 7, which was not the case. However, the upstream condensation domain of module 7 was unaltered, and thus retained specificity for the Asp-S-PCP precursor. Thus, the condensation of any 3-OHAsn-S-PCP intermediate (which may have accumulated) with the upstream hexapeptidyl intermediate would have been severely impaired. Indeed, this could account for the fact that the Asp-7 A domain double point mutant produces considerably more of the hydrolysed CDA hexapeptide intermediate than the Asn-containing CDA-7N (Uguru et al., 2004).

Whilst the results presented here and previously (Hojati et al., 2002; Uguru et al., 2004) do not indicate the timing of the Asn hydroxylation, the evidence suggests that phosphorylation is most likely to occur after A-domain activation. Therefore, 3-OHAsn-containing CDAs (e.g. CDA3a, 3b, 4a and 4b) are the most likely substrates for HasP. Alternatively, the possibility that phosphorylation occurs during peptide assembly cannot be ruled out, as other nonribosomal peptide-tailoring enzymes are suggested to modify peptidyl-thioester intermediates tethered to PCP domains. Indeed, there is evidence that oxidative coupling reactions between the phenolic side chains most likely occur on the peptidyl-thioester intermediates during the assembly of vancomycin and related glycopeptides (Zerbe et al., 2004; Bischoff et al., 2005).

It is also interesting to note that Asn-containing CDAs retain calcium-dependent antimicrobial activity (Fig. 5) and that 3-OHAsn or 3-OPAsn motifs are therefore not essential for calcium binding. Interestingly, the 3-OHAsn/Asp residues found in EGF are suggested to be involved in calcium binding (Valcarce et al., 1999; Lancaster et al., 2004). Thus, whilst our results indicate that 3-OHAsn residues are not essential for coordination of Ca²⁺ ions to CDA, the possibility that β-hydroxylation of Asn in nonribosomal as well as ribosomal peptides is evolutionarily linked to the propensity of the β-hydroxylated peptide products to bind Ca²⁺ ions cannot be discounted (Lancaster et al., 2004). In addition to this, it is also worth noting that the S. coelicolor MT1110-ΔhasP and 2377-ΔhasP strains were able to produce non-phosphorylated CDAs and maintain growth on media supplemented with Ca²⁺. This suggests that HasP is not part of the host self-resistance mechanism, as is the case for other phosphotransferases which share sequence similarity (Lyutzkanova et al., 1997).

This work was supported by the Biotechnology and Biological Sciences Research Council (BBSRC) through research grants (36/B12126 and BB/C503662) and through PhD studentships to L. G., C. M. and A. P., and by the Engineering and Physical Sciences Research Council (EPSRC) through the award of a PhD studentship to J. M. N. L. G. and C. M. thank GlaxoSmithKline, and A. P. thanks Biotica Technology Ltd, for CASE awards.

Edited by: R. P. Mellado

Footnotes

^,† and Jason Micklefield¹ †Present address: School of Biomedical and Molecular Sciences, University of Surrey, Guildford GU2 7XH, UK. ‡Present address: Manchester Interdisciplinary Biocentre, The University of Manchester, 131 Princess Street, Manchester M1 7DN, UK.