Biosynthesis of Aurodox, a Type III Secretion System Inhibitor from Streptomyces goldiniensis

ABSTRACT The global increase in antimicrobial-resistant infections means that there is a need to develop new antimicrobial molecules and strategies to combat the issue. Aurodox is a linear polyketide natural product that is produced by Streptomyces goldiniensis, yet little is known about aurodox biosynthesis or the nature of the biosynthetic gene cluster (BGC) that encodes its production. To gain a deeper understanding of aurodox biosynthesis by S. goldiniensis, the whole genome of the organism was sequenced, revealing the presence of an 87 kb hybrid polyketide synthase/non-ribosomal peptide synthetase (PKS/NRPS) BGC. The aurodox BGC shares significant homology with the kirromycin BGC from S. collinus Tϋ 365. However, the genetic organization of the BGC differs significantly. The candidate aurodox gene cluster was cloned and expressed in a heterologous host to demonstrate that it was responsible for aurodox biosynthesis and disruption of the primary PKS gene (aurAI) abolished aurodox production. These data supported a model whereby the initial core biosynthetic reactions involved in aurodox biosynthesis followed that of kirromycin. Cloning aurM* from S. goldiniensis and expressing this in the kirromycin producer S. collinus Tϋ 365 enabled methylation of the pyridone group, suggesting this is the last step in biosynthesis. This methylation step is also sufficient to confer the unique type III secretion system inhibitory properties to aurodox. IMPORTANCE Enterohemorrhagic Escherichia coli (EHEC) is a significant global pathogen for which traditional antibiotic treatment is not recommended. Aurodox inhibits the ability of EHEC to establish infection in the host gut through the specific targeting of the type III secretion system while circumventing the induction of toxin production associated with traditional antibiotics. These properties suggest aurodox could be a promising anti-virulence compound for EHEC, which merits further investigation. Here, we characterized the aurodox biosynthetic gene cluster from Streptomyces goldiniensis and established the key enzymatic steps of aurodox biosynthesis that give rise to the unique anti-virulence activity. These data provide the basis for future chemical and genetic approaches to produce aurodox derivatives with increased efficacy and the potential to engineer novel elfamycins.

where they target protein translation through inhibition of elongation factor thermounstable (EF-Tu; (4). Direct EF-Tu binding by kirromycin/aurodox-type elfamycins prevents EF-Tu:GDP from dissociating from the ribosome, preventing elongation and inhibiting protein synthesis (4). Aurodox also has an additional mode of action, originally discovered during a screen for type III secretion system (T3SS) inhibitors in Enteropathogenic Escherichia coli (EPEC; (5)). More recently, it was demonstrated that aurodox inhibits T3SS and virulence in enterohemorrhagic E. coli (EHEC) and EPEC through an EF-Tu-independent mechanism, involving the downregulation of transcription of the master virulence regulator, Ler (6).
Aurodox was discovered in 1973 (2). It is a linear polyketide compound that is highly similar to kirromycin (7) differing only in methylation of the pyridone moiety. Kirromycin biosynthesis has been characterized and the BGC encodes five large polyketide synthase (PKS) units, which act to form the polyketide backbone, before tailoring enzymes decorate the molecule (7)(8)(9)(10). Given the similarity of the molecules, we hypothesized that the aurodox biosynthetic gene cluster (BGC) may be homologous to the hybrid PKS/nonribosomal peptide synthetase (NRPS) BGC of kirromycin, with the addition of an ORF responsible for pyridone-associated methylation (Fig. 1).
In EHEC infections, antibiotic treatment is not recommended due to the prevalence of systemic side effects (11,12), and the upregulation of the bacterial SOS response in EHEC, resulting in Shiga toxin expression (13)(14)(15)(16). The consequences of this are severe for patients with increased Shiga toxin levels being associated with systemic infections and nephrotoxicity (12). Therefore, inhibiting virulence in a specific and targeted manner, which does not induce Shiga toxin production, represents a promising strategy for repurposing aurodox as an anti-virulence treatment (6).
Here, we presented the identification and cloning of the aurodox BGC followed by confirmation of the genes responsible for aurodox biosynthesis through heterologous expression and gene disruption. The aurodox biosynthetic cluster was found to differ in organization from that of kirromycin. However, based on homology, we proposed a biosynthetic scheme for aurodox. Additionally, we demonstrated through heterologous expression of the putative methyltransferase (aurM*) that methylation of the pyridone moiety is the last step in aurodox biosynthesis. This improved understanding of aurodox biosynthesis will enable greater exploitation and engineering of aurodox as an anti-virulence therapy and extends our knowledge of an important group of antimicrobial compounds.

RESULTS
Whole-genome sequencing of S. goldiniensis revealed a putative aurodox BGC with homology to the kirromycin BGC. To identify the aurodox BGC, the whole genome of S. goldiniensis ATCC 21386 was sequenced using a hybrid approach where Illumina, PacBio, and Oxford Nanopore technologies were used to generate a highquality draft genome (PRJNA602141; (17)). Analysis of the sequence using antiSMASH (18) identified 36 putative BGCs within the S. goldiniensis genome (see Table S1 at https://doi.org/10.6084/m9.figshare.19140005.v1; (17)). A large region of the S. goldiniensis genome was identified (position 4,213,370 to 4,484,508; 271 kb) that was rich in BGCs, including an 87 kb region with homology to the kirromycin BGC (7). This 87 kb region consisted of 25 open reading frames (ORFs) with 23 ORFs exhibiting .60% similarity to homologs in kirromycin BGC from S. collinus (Table 1; see Table S2 at https:// doi.org/10.6084/m9.figshare.19140005.v1). Despite the homologous ORFs within the BGCs, clear differences were apparent between the kirromycin and aurodox gene clusters, such as the inversion of NRPS/PKS genes and rearrangements of genes that encode the decorating enzymes of the polyketide backbone (Fig. 2). Two additional genes were identified in the aurodox BGC that lacked homologs in kirromycin cluster (Table 1). A gene encoding a SAM-dependent O-methyltransferase (aurM*), which we proposed catalyzes the addition of the methyl group to the pyridone moiety and a hypothetical protein with no predicted homology to genes of known function (aurQ). Given the homology to the kirromycin BGC, it was hypothesized that this putative BGC was responsible for aurodox production in S. goldiniensis (Table 1).
Heterologous expression of the putative aurodox BGC in Streptomyces coelicolor M1152 resulted in aurodox biosynthesis. To determine if the putative aurodox BGC was responsible for aurodox biosynthesis, a phage artificial chromosome (PAC) library was constructed from S. goldiniensis genomic DNA, and the resulting pESAC-13A vectors (19) Table S5). Two PCR-positive PAC clones were identified that contained the entire region of interest. PFGE was used to confirm the size of the inset and one clone (pAur1) was taken forward for further study (Fig. S1 to S3). The complete PAC was sequenced to identify the boundaries of the pAur1, which contained a 129Kbp insert (see Fig. S2 at https://doi.org/10.6084/m9.figshare.19140005.v1 for pAur1) and the complete region proposed to encode the aurodox BGC.
The introduction of pAur1, which integrates at the UC31 integration site, to the S. coelicolor M1152 'superhost' was achieved via conjugation from the nonmethylating ET12567/pR9604 strain to avoid the methyl-specific restriction system (20). Importantly, S. coelicolor M1152 encodes three copies of EF-Tu, including one copy of the elfamycinresistant type EF-Tu, tuf2 suggesting this strain would be a suitable host for expression of aurodox. Exconjugants containing the putative aurodox BGC (pAur1) and empty vector controls (pESAC-13A) were screened via PCR (see Fig. S3 at https://doi.org/10.6084/ m9.figshare.19140005.v1) and the resulting strains were cultured in liquid media. Culture supernatant extracts were then subjected to Liquid Chromatography-Mass Spectrometry (LC-MS) analysis and compared to an authentic aurodox standard (Fig. 3A). An equivalent aurodox peak was also observed in the trace from S. goldiniensis extract (Fig. 3B). Extracts from the S. coelicolor M1152/pESAC-13A, empty vector control lacked the distinct peak of aurodox (Fig. 3C), whereas S. coelicolor M1152/pAur1, the strain containing the putative aurodox cluster, exhibited the characteristic peak (Fig. 3D). Mass spectrometric analysis revealed peaks at m/z 793, corresponding to the molecular ion of aurodox from cultures of S. coelicolor M1152/pAur1 and S. goldiniensis. This peak was absent from the empty vector control strain (Fig. 3C), indicating that the putative aurodox BGC encodes aurodox biosynthesis in S. goldiniensis.
To unequivocally confirm the role of the putative aurodox BGC genes encoded on the pAur1 clone, deletion of the type IPKS (aurAI) was performed using ReDirect (21), resulting in pAur1DaurAI. We hypothesized that the deletion of the primary PKS would prevent aurodox biosynthesis and confirmed the putative BGC was required for aurodox biosynthesis. Analysis of extracts from S. coelicolor M1152/pAur1DaurAI lacked a peak at the aurodox-associated retention time (Fig. 3E) confirming the role of these genes in aurodox biosynthesis.
Bioassays of these culture supernatants using Staphylococcus aureus (ATCC 43300) as an indicator organism, further support the LCMS data, with S. coelicolor M1152/pAur1 inhibiting S. aureus growth, whereas the empty vector control (S. coelicolor M1152/pESAC-13A) and the deletion strain (S. coelicolor M1152/pAur1DaurAI) display reduced bioactivity (Fig. 3F). The presence of a peak at 6 min in the LC trace from S. coelicolor M1152/ pAur1DaurAI could be either a truncated derivative of aurodox produced from the BGC following the mutation of aurAI or more likely the activation of a silent BGC elsewhere in the genome, which is responsible for the bioactivity observed against S. aureus. Proposed biosynthesis of aurodox followed that of kirromycin. The ClusterTools algorithm from antiSMASH was used to annotate the core PKS/NRPS genes of the aurodox BGC, including specific module assignments (18,22). This facilitated the prediction that the catalytic domains of AurI to AurVI largely follow those of KirI to KirVI of the kirromycin gene cluster despite the rearrangements in the overall cluster architecture ( Fig. 2 and 4; (7,9)). In AurAI and AurAII, acetyl Co-A extension is via Claisen condensation reactions (8). However, the PKSs are atypical in arrangement, with two additional dehydratase domains present. While these were not previously identified in the kirromycin pathway (7), reanalysis of the kirromycin pathway using ClusterTools (22) does predict the presence of these domains in KirI to KirVI. Remarkably, no homolog of kirP, the kirromycin phosphopanthetheinyl transferase (PPTase) was identified in the aurodox BGC.
It is predicted that aurAIII encodes a hybrid NRPS/PKS enzyme consisting of consecutive condensation and adenylation domains that catalyze the condensation of glycine and the incorporation of the amide bond, a process conserved with the kirromycin pathway. The enzymes AurAIV-AVII are predicted to extend the aurodox backbone before AurB (which possesses the conserved DTLQLGVIWK motif (23), catalyzes the incorporation b-alanine, presumably synthesized by the putative aspartate-1-decarboxylase, AurD ( Fig. 4 and Table 1).
A SAM-dependent methyltransferase, aurM*, catalyzed the conversion of kirromycin to aurodox. An additional SAM-dependent methyltransferase (AurM*) was identified in the aurodox BGC, which was absent in the kirromycin BGC. It was hypothesized that AurM* catalyzes the conversion of kirromycin to aurodox and may be the last step in aurodox biosynthesis. To test this, aurM* from S. goldiniensis was cloned into an integrating vector (pIJ6902; (24)) and introduced into Streptomyces collinus Tü 365, a natural kirromycin producer. Empty vector controls of S. collinus Tü 365 containing pIJ6902 showed no species corresponding to aurodox but did show the presence of kirromycin compared to an authentic standard ( Fig. 5A and 5B). LCMS analysis of solvent extracts from S. collinus Tü 365 expressing aurM* revealed characteristic peaks corresponding to aurodox and kirromycin (Fig. 5C), with negative scan MS showing an m/z ratio of  (Continued on next page) 793, corresponding to aurodox in addition to an m/z ratio 795 which corresponds to kirromycin (Fig. 5A). This indicates that AurM* is responsible for the methylation of kirromycin as a precursor to aurodox formation.

DISCUSSION
Understanding the mode of action, resistance mechanisms and the biosynthesis of useful natural products is key to their development for clinical application. The linear polyketide, aurodox, while being known for almost 50 years, was recently found to exhibit novel anti-virulence activity via a previously unknown target (6), yet nothing was known about its biosynthesis. While aurodox is structurally similar to kirromycin, it is now well known that structurally identical or highly similar natural products can be biosynthesized via diverse chemical routes (1,(25)(26)(27)(28), suggesting that there is still much to be learned from studying the biosynthesis of structurally similar natural products in terms of novel activity and evolution of natural products.
Despite similarities in structure, anti-bacterial mode of action, and core BGC machinery between aurodox and kirromycin, key differences in biosynthesis were identified. The aurodox cluster is ;80% similar to the kirromycin BGC, sharing 23 of the 25 genes, with seven genes within both clusters encoding hypothetical proteins with no assigned function. There is no apparent PPTase (kirP homolog) encoded in the aurodox cluster, which would normally posttranslationally modify the acyl carrier protein (ACP) to facilitate the extension of the PK/NRP chains during assembly, suggesting that this function may be fulfilled by a promiscuous PPTase encoded elsewhere in the genome (29). Remarkably, no thioesterase domain was identified in the PKS-NRPS megasynthases. There is a putative Dieckmann-like cyclase encoded within the cluster (aurB), which may be responsible for the cleavage and cyclization of the aurodox molecule, a mechanism that has recently been proposed in a few other pyridone natural products (30). The anti-virulence activity of aurodox requires methylation of the pyridone ring, which is catalyzed by aurM*. The 'magic methyl' effect is well known in drug discovery for enhancing the activity and pharmacological properties of drugs (31,32), and the kirromycin to aurodox transformation indicates that this has been exploited in nature to alter the activity of natural products. Moreover, this indicates that the derivatization of the pyridone ring of kirromycin may be a useful strategy for diversifying elfamycin activities. Regarding the origin of aurM*, two additional O-methyl transferases are encoded within the S. goldiniensis genome, each with ;35% amino acid similarly to AurM*, which may suggest that aurM* was acquired by horizontal gene transfer (HGT) rather than duplication of an existing O-methyltransferase from the genome of S. goldiniensis, and offers insight into the evolution of novel activities in natural products and the potential of this to be driven through HGT of single ORFs.
Overall, this greater understanding of the biosynthesis of aurodox and the steps that contribute to unique modes of action will enable us to explore the potential of aurodox as a therapeutic agent for EHEC treatment in the food chain and the clinic

MATERIALS AND METHODS
Growth and maintenance of bacterial strains. The bacterial strains used in this study are detailed in Table S3 at https://doi.org/10.6084/m9.figshare.19140005.v1. Genetic constructs used and their antibiotic selection is described in Table S4. Routine growth and maintenance procedures were carried out according to Kieser et al. (33). A list of oligonucleotides used in this study can be found in Table S5.
Whole-genome sequencing of Streptomyces goldiniensis. Genomic DNA was extracted from S. goldiniensis using the Streptomyces DNA isolation protocol described by Kieser et al. (33). Nanopore reads were obtained using a genomic DNA library prepared in accordance with the Nanopore 1D ligation protocol, using MinION SPOT ON MK1 R9 flow cells and the raw data were converted to sequence data via MinKnow  base-calling software. Illumina data were obtained from Microbes NG (Birmingham, UK) using the HiSeq 2500 sequencing platform. Reads were adapter trimmed using Trimmomatic 0.30 (34) with a sliding window quality cutoff Q15. PacBio sequencing was provided by Nu-omics at (the University of Northumbria, UK) using the PacBio Sequel instrument, and contigs were assembled in HGAP4.
Streptomyces goldiniensis genome was assembled using SPAdes (34) using data provided by all three platforms. AutoMLST (35) was used to determine the closest neighbor S. bottropensis ATCC 25435, Taxonomy ID: 1054862) which was used for scaffold-based assembly via MeDuSa (36) with the quality analysis carried out using QUAST (37). Annotation of the Streptomyces goldiniensis genome was created using Prokka (38) and can be accessed at the GenBank Bioproject PRJNA602141. Biosynthetic gene cluster identification was carried out using antiSMASH bacterial version 5.0.0 with modular enzymatic domains analysis carried out using the PKS/NRPS domain annotation function in antiSMASH (18).
Aurodox production, purification, and detection. Streptomyces spore stocks (1 Â 10 8 spores) were pregerminated in 10 mL of GYM medium (32) overnight at 30°C with shaking at 250 rpm. Cells were then washed three times to remove antibiotics (if used) and resuspended in 1 mL of GYM, which was used to inoculate a 50 mL seed culture of GYM which was incubated at 30°C with shaking at 250 rpm. After 7 days, biomass was removed by centrifugation (4000 Â g, 10 min) and culture supernatant was filter sterilized through a 0.2 mM Millipore filter. Supernatants were mixed with an equal volume of chloroform and a separation funnel was used to remove the lower, solvent phase. Solvent extracts were dried under nitrogen and extracts were solubilized in ethanol for LC-MS analysis. Authentic aurodox standard (1 mg/mL) was purchased from Enzo (New York, USA). LC-MS was carried out on an Agilent 1100 HPLC instrument in conjunction with a Waters Micromass ZQ 2000/4000 mass detector. Electrospray ionization (ESI) was used in all cases. The RP-HPLC analysis was conducted on a Zorbax 45 mm x 150 mm C18 column at 40°C. Ammonium acetate buffers were used as follows: Buffer A (5 mM Ammonium acetate in Water) and Buffer B (5 mM Ammonium acetate in acetonitrile). Positive and negative electrospray methods were applied with 100 to 1000 AMU positive, and 120 to 1000 AMU negative with a scanning time of 0.5 s. The UV detection was carried out at 254 nm.
Construction of aurodox expression strains and deletion mutant. Aurodox encoding vector pAur1 (Bio S & T, Canada) and empty vector pESAC-13A (39) were transferred to the nonmethylating Escherichia coli strain ET12567 via tri-parental mating. Briefly, fresh overnight cultures of E. coli DH10b containing pAur1 or the parental vector pESAC-13A (apramycin resistant), E. coli Top10 containing the driver plasmid pR9604 (Beta-lactam resistant), and ET12567 (Chloramphenicol resistant) were to an optical density at 600 nm (OD 600 ) of 0.6. Cells were washed three times in fresh LB by centrifugation at 4000 Â g and resuspended in 1 mL of LB. A 20 mL aliquot of each strain was plated in the center of an LB plate and incubated at 37°C overnight. The resulting growth was restreaked on to the required antibiotic selection and the presence of the conjugating vector in the E. coli ET12567/pR9604 strain was confirmed via colony PCR. Mutation in pAur1 was carried out according to the ReDirect PCR targeting system in Streptomyces (21), using the hygromycin resistance cassette, pIJ10700 as the disruption cassette (http://streptomyces.org.uk/redirect/RecombineeringFEMSMP-2006 -5.pdf). Disrupted PACs were introduced into E. coli ET12567/pR9604 via tri-parental mating and subsequently moved into Streptomyces coelicolor M1152 via conjugation as described above.
Bioassays. Bioassays were conducted using disc diffusion assays with Streptomyces fermentation extracts on soft nutrient overlays containing Staphylococcus aureus ATCC 43300 as the indicator organism.