Revealing the first uridyl peptide antibiotic biosynthetic gene cluster and probing pacidamycin biosynthesis

Abstract

There is an urgent need for new antibiotics with resistance continuing to emerge toward existing classes. The pacidamycin antibiotics possess a novel scaffold and exhibit unexploited bioactivity rendering them attractive research targets. We recently reported the first identification of a biosynthetic cluster encoding uridyl peptide antibiotic assembly and the engineering of pacidamycin biosynthesis into a heterologous host. We report here our methods toward identifying the biosynthetic cluster. Our initial experiments employed conventional methods of probing a cosmid library using PCR and Southern blotting, however, it became necessary to adopt a state-of-the-art genome scanning and in silico hybridization approach to pinpoint the cluster. Here we describe our “real” and “virtual” probing methods and contrast the benefits and pitfalls of each approach.

Pacidamycins (Scheme 1) are uridyl peptide antibiotics (UPAs) produced by Streptomyces coeruleorubidus which exhibit a narrow spectrum of activity, specifically inhibiting Pseudomonas aeruginosa, a Gram-negative pathogen inherently resistant to many antibiotics.^1,2 The UPAs are inhibitors of translocase I, an essential yet clinically untargeted bacterial enzyme involved in bacterial cell wall biosynthesis; the novel UPA scaffold is thus attractive for the development of new antibiotics. The UPAs contain many structurally unusual features including a pseudopeptide backbone linked to an atypical 3′-deoxyuridine aminonucleoside. The sense of the peptide chain undergoes a double inversion caused by the incorporation of a diamino acid residue and a rare internal ureido moiety. Access to the unique biosynthetic enzymes mediating the assembly of these biosynthetically unusual features is appealing, as it would not only enable combinatorial biosynthetic approaches to designer UPA analogues, but potentially also provide useful tools for the biotransformation of fine chemicals.

Scheme 1.

Chemical structure and retro-biosynthesis of pacidamycins produced by S. coeruleorubidus.

The biosynthetic distinctiveness of the UPAs, that makes them so attractive for research, provides the greatest challenge to finding the genes encoding their assembly. Bacterial natural products are usually encoded by gene clusters where all or most of the genes may be found within the same region of genomic DNA. This feature of microbial biosynthetic pathways greatly facilitates the localization of biosynthetic genes. Gene clusters are often revealed by screening the genomic DNA of the organism of interest for genes that encode enzymes that are likely to operate in the assembly of the natural product. This task is simplified if close homologues of candidate genes are known; for novel pathways this prior knowledge is absent.

Over the past decades, our understanding of the genetics and enzymology of natural product biosynthesis has greatly increased not least due to cheaper, faster, more efficient DNA sequencing technologies. Whilst a vast number of microbial genome sequences can now be accessed and mined for biosynthetic gene clusters, a complete genome sequence can still be time consuming and costly to obtain. The recent development of massively parallel sequencing,³ on the other hand, enables the acquisition of a genome scan at relatively low cost and high speed. In this paper we utilize our recent identification of the pacidamycin gene cluster⁴ as an example to highlight the advantages of in silico probing of genome scan data as compared to working with the more traditional PCR screens and heterologous probes.

Rationale for Probe Design

In order to identify the pacidamycin biosynthetic genes we used a hypothesis-driven approach to design probes. As mentioned above, this presented an exciting challenge due to the metabolite's highly unusual structure; predictions of enzyme/gene involvement are more accurate for well understood biosynthetic routes and become more speculative the more atypical the pathway is likely to be. Given the peptidic nature of the backbone, it was envisaged that pacidamycin assembly could potentially be mediated by a nonribosomal peptide synthetase (NRPS). NRPSs are modular multi-domain complexes responsible for the biosynthesis of a plethora of clinically important metabolites such as vancomycin and daptomycin.^5,6 Notably, NRPSs are known to incorporate nonproteinogenic amino acids such as the diamino acid (2S, 3S)-DABA and m-tyrosine residues found in pacidamycin. NRPS adenylation domains have proven to be particularly useful as probes to locate NRPS biosynthetic clusters and sets of validated degenerate primers can be found in the literature.⁷

The most significant handle within the pacidamycin structure is the nonproteinogenic, N-methylated DABA residue. Genes for the generation of DABA in friulimicin and for diaminopropionate in viomycin are known.^8,9 In both cases, the diamino acid is postulated to be produced by a PLP-dependent enzyme through a β-replacement mechanism (Scheme 2). More recently, Bugg and co-workers have indicated that the biosynthesis of DABA in mureidomycin, a close structural homologue of pacidamycin, might be generated in a similar manner from threonine.¹⁰ The isotopic enrichment of the pacidamycin molecular ion, when S. coeruleorubidus is cultured in medium supplemented with [¹⁵N]-ammonium chloride, provides further evidence to support the occurrence of this β-replacement reaction in pacidamycin biosynthesis (data not shown).

Scheme 2.

Proposed mechanism of PLP-dependent diamino acid formation.⁹,¹⁰ The reaction is analogous to that catalysed by cysteine synthase or the tryptophan synthase β subunit and proceeds through a β-elimination-addition mechanism. The incoming ammonium nucleophile is shown in bold.

At the time of the study, there was no precedent for the generation of the aminonucleoside and it was not possible to design probes for the assembly of this moiety. Neither was it possible to rationally design probes for the ligation of (amino-)nucleosides to the peptide, as the few relevant ligations reported are mediated by a range of different enzyme classes.^11–14

DNA-Based Screens

A PCR screen for putative NRPS adenylation domains using published degenerate primers⁷ revealed the wealth of secondary metabolism encoded within the S. coeruleorubidus genome. This was not unexpected as Streptomyces are prolific producers of natural products including nonribosomal peptides. PCR products of the anticipated size were observed for 53 cosmids from a cosmid library with 1,536 members. Restriction digest analysis of the cosmids with positive hits allowed them to be grouped into 24 families with distinct restriction patterns. Approximately half of these families (13) contained only one member. A single adenylation domain PCR product from each family was subsequently sequenced. This revealed that seven clones were identical, thus reducing the number of cosmid families to 17. Two hits resulted from unspecific primer binding and three families showed sequence homologies to AMP-binding domains but were otherwise unrelated to adenylation domains. This left 12 distinct cosmid families (with a total of 44 different cosmids) that showed good sequence homologies to NRPS adenylation domains. Six adenylation domains that, from their sequence,^15,16 looked as though they might adenylate pacidamycin-related amino acids were selected for PCR-mediated gene disruption.¹⁷ Intriguingly, all amino acid adenylation domain mutants remained able to generate pacidamycins and this approach failed to identify a gene essential for their biosynthesis. Subsequently, a hybridization screen for a putative DABA synthase was employed. Southern blots using the putative DAP synthase gene vioB from Streptomyces vinaceus (viomycin biosynthesis) as heterologous probe resulted in almost 200 hits even under stringent conditions indicating unspecific binding and were thus deemed inconclusive.

In silico Data Mining

A clear disadvantage of the DNA-based screens described above is that multiple experimental steps are involved before a hit can be validated. This is in contrast to in silico analysis of genome sequence data which enables levels of homology to be quantified immediately. The genome scan of S. coeruleorubidus provided 10,369 contiguous DNA sequences (contigs) with a predicted 3.6-fold coverage of a 9.4 Mb genome.⁴ The average read length was unfortunately only 258 bp, probably due to the high GC content of the DNA. This limited information about the surrounding genes.

BLAST searches for NRPS and DABA synthase genes were carried out to mine the S. coeruleorubidus genome for potential pacidamycin biosynthetic genes. In this way, 69 contigs with fragments of NRPS adenylation domains were identified reflecting the large number of hits found with the earlier PCR screens. Screens for putative DABA synthase genes identified 17 contigs encoding candidate gene fragments; four of these contigs contained sequence hits with greater than 95% identity to genes involved in primary metabolism in other Streptomyces species and so could be dismissed immediately. Due to the manageable number of hits and the strong likelihood of a DABA synthase operating in pacidamycin biosynthesis, these sequences were selected for further investigation. After dereplication of identical genes by PCR with contig-specific primers six distinct DABA synthase candidates (dab1–dab6) could be deduced (Fig. 1). Interestingly, DNA sequences coding for ATP-grasp domains were clustered with dab1 and dab2 as deduced from inspection of the genome scan data, and dab1 is further associated with a PduX domain coding region. The same family of genes is also found in association with the DABA synthase gene from the friulimicin pathway;⁸ this analogy made dab1 and dab2 the most promising DABA synthase candidates. Gene disruption of dab1–dab6 demonstrated that only the dab1 mutant lost the ability to produce pacidamycins. Further corroboration of the involvement of dab1 was derived from chemical complementation of the mutant strain with synthetic DABA to restore pacidamycin production. Finally, the dab1-containing cosmid was identified and its introduction into Streptomyces lividans conferred the ability to produce pacidamycins on the heterologous host.⁴

Figure 1.

Organization of the putative DABA synthase genes dab1–dab6. The arrows are not drawn to scale. Annealing sites for specific PCR primers are indicated by half-arrows and contigs (ctg) and gene function assigments are specified. CysSyn: cysteine synthase-like, putative DABA synthase; Transp: transporter; Hyp: hypothetical protein; Decarb: decarboxylase; Des: cysteine desulfurase; MTase: methyl transferase.

Proposed Pacidamycin Biosynthetic Pathway

Sequencing of the dab1-containing cosmid revealed the pacidamycin gene cluster consisting of 22 genes and allowed a tentative biosynthesis to be postulated; this is described in more detail by Rackham et al. NRPS modules, discrete NRPS domains and associated proteins are encoded by nine genes; these are thought to be responsible for the construction of the peptide backbone including the ureido motif.¹⁸ A subcluster of four genes is likely to be involved in DABA biosynthesis, and a methyltransferase gene for N-methylation of the diamino acid forms the 3′-end of the cluster. The remaining eight genes, some of them encoding hypothetical proteins, are thought to be involved in the biosynthesis of the aminonucleoside, and in the pathway regulation and export.

Conclusions

This paper is mainly targeted at non-geneticists and novices in the field of cluster identification and uses the pacidamycin biosynthetic cluster to provide a comparison between the benefits and pitfalls of real and virtual probing for candidate genes. The in silico screening provides more, quantifiable and immediate information regarding the candidate gene and is experimentally less challenging than traditional hybridization experiments or the preparation and screening of a mutant library.

Abbreviations

ATP

adenosine triphosphate

BLAST

basic local alignment search tool

DABA

2,3-diaminobutyric acid

DAP

2,3-diaminopropionate

NRPS

nonribosomal peptide synthetase

PLP

pyridoxal phosphate

SAM

S-adenosylmethionine

UPA

uridyl peptide antibiotic