Metagenomic small molecule discovery methods

Abstract

Metagenomic approaches to natural product discovery provide the means of harvesting bioactive small molecules synthesized by environmental bacteria without the requirement of first culturing these organisms. Advances in sequencing technologies and general metagenomic methods are beginning to provide the tools necessary to unlock the unexplored biosynthetic potential encoded by the genomes of uncultured environmental bacteria. Here, we highlight recent advances in sequence- and functional- based metagenomic approaches that promise to facilitate antibiotic discovery from diverse environmental microbiomes.

Introduction

Many important antibiotic compounds have been isolated from cultured bacteria; however, the vast majority of bacteria remain recalcitrant to culturing.(1) It is estimated that soil contains as many as 10⁵ unique species per gram and that uncultured microorganisms outnumber cultured ones by two to three orders of magnitude.(2-4) Metagenomics is a culture-independent approach that seeks to access the biosynthetic capacity of the “uncultured majority” of bacterial species. By directly capturing DNA from the environment (environmental DNA, eDNA) and subsequently identifying, isolating, and expressing biosynthetic gene clusters in heterologous hosts, metagenomics has the potential to provide a complete toolkit for bringing biosynthetic diversity from the environment into drug discovery pipelines. Two general approaches are employed for interrogating and exploiting metagenomic eDNA for the production of small molecules. Sequence-based approaches profile the biosynthetic content of metagenomic samples, identify high-value targets, and aid in the targeted recovery of complete biosynthetic pathways from eDNA cosmid libraries. These recovered clusters often require genetic manipulation to activate small molecule production in a heterologous host. In contrast, function-based approaches aim to identify clones that are already biosynthetically active in a heterologous host by detecting a clone-induced phenotype in a host organism. This review covers recent technological and experimental advances that are accelerating metagenomic small molecule discovery efforts with a focus on a) sequence homology-based techniques that facilitate metagenome profiling and gene cluster recovery and b) advances in function-based methods that expedite the identification of bioactive clones.

Sequence-based metagenomic studies

The precipitous reduction of DNA sequencing cost is transforming the process of natural product drug discovery. Whereas classic, culture-based studies required isolation of compounds in the search for novel bioactivity, the availability of sequence data has driven the development of bioinformatic tools that can streamline the identification of target gene clusters without requiring chemical isolation. The methods used to identify gene clusters of interest in metagenomes generally fall into one of two categories: shotgun sequencing or PCR-based sequence tag approaches.

Shotgun studies

Genome-based approaches to natural product discovery stand to benefit from the proliferation of sequencing technologies and the accompanying bioinformatic analyses they enable. The torrent of genome sequences of cultured bacteria (>500/month at NCBI (5)) is sparking renewed interest in natural product discovery. This is due in large part to identification of previously unknown gene cluster in many organisms, including those that have been thoroughly studied.(6) Computational tools that scan assembled genomes and identify biosynthetic gene clusters, such as AntiSMASH and np.searcher, are now able to predict the expected natural products encoded by these clusters.(7, 8) Application of such tools to all newly sequenced genomes is becoming a routine part of new genome analysis, providing a way to identify and rank new clusters for genome mining. These tools can also be applied to assembled contigs generated from metagenomic sources and used to identify clusters from uncultured organisms, a strategy that has been particularly useful in the elucidation of the small molecule producing clusters of uncultured endosymbionts of marine (9-11) and terrestrial (12) metazoans. The assembly of symbiont genomes from metagenomic samples has been used to identify the gene clusters encoding a potent cytotoxin, patellazole, a novel polyketide, nosperin, as well as to guide the discovery of a genus of bacterial symbionts, Entotheonella, with promising biosynthetic potential.(9, 12, 13) By coupling deep-sequencing with other tools like whole genome amplification (14) and single-cell isolation, extensive biosynthetic information can be gleaned from otherwise difficult-to-access organisms, as is the case with the recent elucidation of the apratoxin cluster from a marine cyanobacterium.(15) The application of whole-genome sequencing is a useful tool for the characterization of endosymbionts and other relatively small metagenomes; however, other techniques are necessary for the complex metagenomes found in many natural environments like soil.

Sequence Tag Tools

A typical soil metagenome may contain 10⁴-10⁵ unique species.(2, 3) Shotgun assembly of such metagenomes is still very challenging. Fortunately, substantial information about biosynthetic genes can be obtained through the use of simpler, PCR-based sequence tag approaches.(16) Sequence tags are PCR amplicons generated using primers targeting conserved biosynthetic genes that can be used for phylogenetic analysis. They take advantage of the modularity of biosynthetic systems, which have evolved for horizontal transfer of useful phenotypes (e.g. small molecule production), while facilitating the creation of chemical novelty through well-established genetic mechanisms.(17, 18) Sequence tags can be mapped to related sequences within known biosynthetic clusters, which is the basis of the eSNAPD and NaPDoS programs.(19, 20) At high degrees of sequence similarity (eSNAPD: E-value <10e⁻⁴⁰, NaPDoS: 90% sequence identity), short sequence tags of only several hundred base pairs can effectively match a read to a reference gene cluster. Remarkably, the general structure of an entire gene cluster can generally be inferred from the tag, as validated by the recovery of eDNA clones predicted to encode novel glycopeptide, lipopeptide, and bisintercalator like natural products.(19) The true utility of these approaches is not in the identification of known gene clusters but instead in rapidly identifying gene clusters encoding congeners of valuable compounds, or in finding potentially novel gene clusters that have remained undetected in the environment. (19)

Earlier applications of sequence tags to natural product characterization were for genotyping of strains of the prolific marine Actinomcyete genus Salinaspora.(21, 22) It allowed a quick, inexpensive way to profile biosynthetic potential of cultured organisms without requiring chemical isolation. The use of 454-pyrosequencing enabled scaling of this approach to profile entire metagenomes. Pyrosequencing of ketosynthase and condensation domains from marine sponge metagenomes uncovered hundreds of previously unknown sequences, including several clades that had not been previously observed by extensive Sanger sequencing of these sponge metagenomes.(23) Furthermore, a head-to-head comparison of shotgun sequencing of the same samples demonstrated that PCR-based approaches were often 10 to 100 times more sensitive in identifying unique sequences from a metagenome of interest.(23) Even greater biosynthetic biodiversity was observed in desert soil microbiomes, where 1,000s of unique adenylation domain sequences were detected with only a fraction of them shared among distinct microbiomes. A similar pattern was observed for amplicons derived from Type I (ketosynthase) and Type II (ketosynthase alpha) polyketide biosynthesis, suggesting that soil metagenomes are likely to be a rich source of novel bioactive compounds.(24, 25) While purified metagenomic DNA is suitable for profiling the biosynthetic diversity, arrayed, large-insert libraries are needed to facilitate isolation of the identified pathways.(26) Prior to choosing a sample for use in library construction, sequence tag-based methods can be employed to screen eDNA to identify the most biosynthetically rich environments. These same methods can then be used to identify and guide the isolation of specific clones from eDNA libraries. Clones recovered from metagenomic libraries in this manner have been heterologously expressed to yield new bioactive Type II polyketide antibiotics (25, 27); new tryptophan-based cytotoxins (28-30); the marine-derived siderophores bisucaberin and vibrioferrin (31, 32); modified versions of the antitumor compound pederin (33); cyanobacterial–derived cyclic peptides (34); and new members of the microviridin family of ribosomally synthesized peptides.(35, 36)

Future Application of Sequencing Technology to Metagenomes

Several promising technologies may extend the power of whole genome sequencing to metagenomes. Nano-pore based sequencing boasts long read lengths that alleviate the problem of assembly of repetitive regions within the genome and is quickly becoming the method of choice for bacterial genome sequencing.(37, 38) Long read sequencing technologies are of particular interest when sequencing natural product gene clusters due to the highly repetitive nature of some biosynthetic gene clusters. Additionally, single-cell and microdroplet-based methods can now obtain sequence data from single cells, without requiring the generation of an eDNA library, and can facilitate the sequencing of the rare biosphere.(39, 40) These sequencing methods will expedite the in silico characterization of naturally occurring biosynthetic gene clusters and will push the molecule-discovery bottleneck downstream to the activation of gene cluster expression.

Function-Based Metagenomics

Sequence-based metagenomics takes full advantage of the information gained through advances in DNA sequencing. Unfortunately, pathways recovered by sequence-based methods often require genetic refactoring in order to be active in a heterologous host. Functional metagenomics provides a complementary approach that bypasses the refactoring steps by screening for, and isolating, clones that are already active in the heterologous host strain. A variety of functional screens have been developed to date that rely on phenotypic detection using: pigmentation, enzymatic or antibiotic activity (41-43); selection of complementation-dependent reporters (44, 45); and substrate induction (SIGEX and METREX).(46, 47) Direct screening of fermentation broths has also been used, although this approach becomes impractical with increasing library size.(48) While functional metagenomics presents a powerful set of tools for identifying bioactive compounds encoded by environmental microbiomes, the size and the heterogeneous nature of eDNA libraries pose a number of challenges that are currently being addressed through a combination of technical advances and new screening methods.

Library Creation and Maintenance

Recent advances in eDNA cloning and in broad-host range vector design can facilitate the creation of eDNA libraries and allow a single library to be screened in a variety of hosts. In contrast to genome-based cosmid libraries, creation of eDNA libraries can be challenging due to difficulties associated with obtaining sufficient quantities of high molecular weight (HMW) DNA free of environmental inhibitors that can interfere with cloning. A newly developed technology, synchronous coefficient of drag alteration (SCODA), has enabled recovery of HMW eDNA from virtually any source by removing interfering contaminants while concentrating dilute samples.(49, 50). The cloning of eDNA into a broad-host range shuttle vector facilitates transfer of eDNA libraries into different hosts, where orthogonal collections of genes are likely to be expressed. Vectors based on ΦC31 and ΦBT1 phage integrase systems for transfer of libraries into diverse Streptomyces spp. have existed for quite some time.(51, 52) More recently, RK2-derived vectors have been constructed to allow movement of libraries into diverse alpha, beta, and gamma proteobacterial species.(53-55) Creating libraries in shuttle vectors allows for conjugative transfer of eDNA into a wide range of hosts, including bacteria belonging to biosynthetically-rich phyla like Streptomyces and beta-proteobacteria.(56)

Host Improvement and New Detection Methods

It is thought that potential transcriptional, translational and biochemical blocks can hinder the expression of an exogenous gene cluster in an individual host. Using multiple heterologous hosts for functional screening maximizes the chance of identifying bioactive molecules by matching eDNA-derived clusters with native host biochemistries. It is also possible to imagine adapting hosts to more permissively express eDNA-derived biosynthetic gene clusters. Streptomyces appear to be one of the most prolific secondary metabolite producers and, as such, a number of strain improvement efforts have focused on this genus.(56) Ribosome engineering and use of mutant RNA polymerases have been employed in Streptomyces spp., as well as myxobacteria and fungi, in an effort to facilitate activation of cryptic pathways and improve expression in general.(57) Other efforts to activate silent gene clusters have focused on deleting global negative regulators of biosynthesis, such as DasR, or have used overexpression of positive regulators such as LAL.(58, 59) Similarly, overexpression of the alternative sigma factor σ⁵⁴ has been used to facilitate expression of a Type II polyketide synthase (PKS) gene cluster in Escherichia coli.(60) In general these host-manipulations result in more promiscuous transcription and may therefore allow for a higher percentage of cloned eDNA gene clusters to be expressed in high throughput functional screens.

The coupling of more sensitive natural product detection tools with libraries hosted in improved strains should lead to a greater number of eDNA-encoded compounds being identified in functional screens. In recent years, mass-spectrometry has emerged as a powerful and sensitive tool for natural product screening. Improved selective detection techniques, such as those for phosphonic acid and phosphonate group containing compounds, are making it possible to detect previously undetectable small molecules.(61) Similarly, a suite of tools termed peptido- and glyco-genomics, are able identify NRPS-derived peptides and O-/N-glycosyl containing sugar monomers directly from bacterial colonies on agar plates. Mass-spec data is used in combination with bioinformatic predictions of biosynthetic systems, providing a powerful cross-referencing system that can be used to predict the structure of potential metabolites and then to identify and validate the structure predictions.(62, 63) So far, these techniques have been applied only to cultured bacteria; however, it is easy to imagine how such approaches could be applied to direct functional screening of metagenomic libraries.

Library Enrichment

Analysis of sequenced genomes reveals that less than 2% of the genome is devoted to secondary metabolism;(64) consequently, metagenomic DNA libraries are sparsely populated with biosynthetic genes of interest. Gene cluster enrichment strategies can be used to simultaneously reduce the size and increase the biosynthetic density of eDNA libraries, thereby increasing the efficiency of functional metagenomic applications.(46, 65-68) Complementation of 4’-phosphopanthetheinyl transferase (PPTase) activity has been recognized as a valuable tool for gene cluster mining applications. PPTases are required to generate holo-nonribosomal peptide synthetase (NRPS) and polyketide synthase (PKS) enzymes and have been used in a phage display strategy designed to recover NRPS/PKS sequences.(69) More recently, PPTase gene complementation has been successfully used for enriching Escherichia coli and Pseudomonas aeruginosa hosted libraries for NRPS/PKS containing clones by linking it to production of either a colored indicator molecule or a siderophore for selective growth on low iron media.(70, 71) Using low-iron selection strategy in E. coli, 50-fold enrichment in NRPS/PKS gene content was achieved in just two rounds of selection. (71)

Future directions in functional metagenomics

While sequence-based metagenomic screening approaches are now quite robust and capable of targeting the discovery of diverse novel molecules from many different environments, functional metagenomic screening methods have not yet advanced to the extent that enables this approach to rapidly harvest molecules from the environment. We believe that three key advances are necessary to bring functional metagenomics to maturity. First, new model heterologous hosts must be identified and engineered that are able to promiscuously activate a more diverse set of biosynthetic gene clusters. Second, we need improved DNA cloning methods that enable capture of complete gene clusters on individual eDNA clones. Finally, new methods are needed for selectively enriching eDNA libraries for clones containing a variety of secondary metabolite genes. Together, these advances will facilitate a substantial increase in the frequency and diversity of small molecules with novel bioactivities harvested from the environment using functional metagenomic approaches.

Conclusions

By taking a gene-based approach, metagenomics can exploit the sequencing revolution and bypass many of the traditional hurdles to drug discovery. While cultured organisms have yielded many of our most important antimicrobial agents, these organisms represent only a small fraction of total microbial diversity. Metagenomic methods provide a means to evaluate the biosynthetic potential of the bacterial majority, thereby providing an opportunity to find truly novel antimicrobials. While there remain bottlenecks in the metagenomic drug-discovery platform, such as the heterologous expression of metagenomic pathways, these problems are not unique to metagenomics and are also being tackled by the broader micro- and synthetic biology communities. As the development of metagenomics-specific tools progresses and the most promising, high-throughput, genome based approaches are adapted into the field, metagenomics should play an increasingly important role in the future of antibiotic drug discovery.

Highlights.

• DNA sequencing finds tremendous unknown biosynthetic diversity in the environment.

• Advances in sequencing methods permit the targeted recovery of gene clusters.

• New hosts, vectors and methods speed the identification of new natural products.

Figure.

A brief overview of metagenomics-based small molecule drug discovery process (A) and key technical advances that are driving its success (B).