Antimicrobial resistance (AMR) is a growing societal problem, and without new anti-infective drugs, the UK government-commissioned O'Neil report has predicted that infectious disease will claim the lives of an additional 10 million people a year worldwide by 2050. Almost all the antibiotics currently in clinical use are derived from the secondary metabolites of a group of filamentous soil bacteria called actinomycetes, most notably in the genus Streptomyces. Unfortunately, the discovery of these strains and their natural products (NPs) peaked in the 1950s and was then largely abandoned, partly due to the repeated rediscovery of known strains and compounds. Attention turned instead to rational target-based drug design, but this was largely unsuccessful and few new antibiotics have made it to clinic in the last 60 years. In the early 2000s, however, genome sequencing of the first Streptomyces species reinvigorated interest in NP discovery because it revealed the presence of numerous cryptic NP biosynthetic gene clusters that are not expressed in the laboratory. Here, we describe how the use of new technologies, including improved culture-dependent and -independent techniques, combined with searching underexplored environments, promises to identify a new generation of NP antibiotics from actinomycete bacteria.
More than half the clinically important antibiotics are based on the natural products (NPs) of filamentous actinomycetes like Streptomyces species, and most of these were isolated between 1940 and 1960 in the so-called Golden Age of antibiotic discovery. Repeated rediscovery of known strains and compounds in the 1960s, however, led most pharmaceutical companies to abandon NP discovery in the belief that all the useful compounds had been discovered from soil microorganisms. Genome sequencing in the 21st century has reinvigorated NP discovery because it revealed that most filamentous actinomycetes encode many more NPs than previously identified, most of which are not produced under standard laboratory conditions. For example, the actinomycete Streptomyces coelicolor has been studied since the 1960s but until its genome was published in 2002 only five secondary metabolites had been identified, the antibiotics actinorhodin, undecylprodigiosin, methylenomycin and calcium-dependent antibiotic (CDA) and the WhiE polyketide spore pigment . In depth, bioinformatic and biochemical analyses have since revealed that S. coelicolor encodes at least 32 secondary metabolites, including at least 15 distinct families [1–3]. Individual Streptomyces genomes contain between 20 and 60 biosynthetic gene clusters (BGCs), most of which are cryptic (i.e. not expressed) in the laboratory, and this represents a huge pool of untapped biosynthetic potential. In addition to unlocking cryptic BGCs in known strains, it is also now clear that terrestrial and marine environments have been vastly under sampled in terms of bacterial strains and NPs. This review will focus on our current understanding of how to activate antibiotic biosynthesis and the tools and techniques that can be used to switch on expression of cryptic BGCs. We also provide examples of new ways to isolate NPs from difficult to culture bacteria and some unusual environmental niches that are being explored to find new bacterial strains making novel NPs. We focus on filamentous actinomycete bacteria but most microorganisms make bioactive NPs, and the techniques we describe are widely applicable. We also predict that high-throughput genome and metagenome sequencing combined with recent advances in synthetic and chemical biology techniques will eventually enable scientists to synthetically rewire and express novel BGCs from cultured and uncultured microorganisms and that in the long term, this offers the best hope of unlocking all the biosynthetic potential in nature (Figure 1).
Discovery and isolation of novel NPs.
Control of antibiotic production in actinomycetes
Antibiotic production is linked to sporulation and triggered by stresses, including nutrient starvation and competition [4,5]. The simplest way to induce the production of bioactive NPs is to change the growth conditions to mimic environmental pressures and induce a starvation response. Although it obviously does not work for all BGCs, this can involve trialling different culture media or cultivating with extracts of soil. Strains of interest can also be co-cultivated with other microbes to encourage competition and the production of antimicrobial compounds. For Streptomyces, this approach has been shown to be particularly effective when they are co-cultured with strains that contain mycolic acids in their cell membrane. For example, when co-cultured with the pathogenic Tsukamurella pulmonis, Streptomyces endus produces the novel antibiotic alchivemycin and Streptomyces lividans produces an uncharacterised red antibiotic . Induction of the stringent response has also been shown to activate antibiotic production. In this nutrient limitation response, transcription is modulated to divert resources away from growth and division and towards amino acid biosynthesis. The stringent response is regulated by the ribosome-associated RelA protein which senses uncharged tRNA molecules in the A site and synthesises the nucleotide ppGpp (guanosine tetraphosphate) [7,8]. Manipulating ppGpp levels has been shown to affect antibiotic production in a range of Actinobacteria, for example, it switches on microbisporicin production in Microbispora corallina by inducing expression of its pathway-specific regulator MibR . Increasing ppGpp levels in S. coelicolor also directly activates the production of actinorhodin, by up-regulating expression of the activator gene actII-4, and increases expression of the BGC encoding CDA [10,11].
While the stringent response is a natural way to alter the global pattern of transcription and translation, the same can be achieved using small molecules or through targeted mutagenesis. Histone deacetylase (HDAC) inhibitors have been used to activate the production of secondary metabolites in fungi by partially unwinding DNA from histones and inducing the expression of genes which are usually cryptic [12,13]. HDAC inhibitors also activate cryptic antibiotic BGCs in actinomycetes and have been used to switch on antibiotic production in Streptomyces, Saccharopolyspora and Pseudonocardia species, although the molecular mechanism of action is unknown (Figure 2) . Transcription can also be altered by selecting for spontaneous resistance to the antibiotic rifampicin that induces spontaneous mutations in the RNA polymerase β-subunit gene (rpoB) and can activate the expression of cryptic BGCs . Translation can also be altered by inducing mutations in the ribosomal S12 protein gene (rpsL) and the 16S rRNA methyltransferase gene (rsmG) through exposure to low or high concentrations of streptomycin, respectively . These mutations also result in the overproduction of antibiotics in a relA background, suggesting that they mimic ppGpp activation . Sequentially introducing different antibiotic resistant mutations into the same strain of S. coelicolor increased antibiotic production up to 180-fold , and they have been introduced into so-called S. coelicolor ‘superhost’ expression strains . Alternative antibiotics have also been used to select for enhanced production of NPs .
Use of HDAC inhibitors to unlock cryptic BGCs.
Streptomyces species encode numerous transcriptional regulators and signal transduction pathways, which enable them to respond to changes in their environment. Nutrient availability is a key signal that controls antibiotic production, and this was reviewed recently by Urem et al. . Many of the regulators that have been characterised in streptomycetes are implicated in the regulation of NPs, presumably because most of these were identified through screens for mutants that affect antibiotic production or differentiation . Nevertheless, the complexity of regulation at most BGCs is breathtaking and still quite poorly understood . Refactoring of individual BGCs can be effective if the pathway-specific regulatory mechanisms are well understood, e.g. by removing pathway-specific repressors or constitutively expressing activators (for a recent review, see Aigle and Corre ). This was effectively demonstrated using a ‘Plug and Play’ synthetic biology approach to refactor and activate a cryptic gene cluster in Streptomyces griseus and identify three new polycyclic tetramate macrolactams . However, without prior knowledge of the regulatory mechanisms at work, this is a high-risk approach .
The life cycle of Streptomyces bacteria is temporally and genetically co-ordinated with the production of antibiotics, and deletion of key developmental genes can result in altered or enhanced production. An alternative strategy therefore is to understand the signal transduction pathways acting at a global, developmental level since they often control pathway-specific regulators. The MtrAB two-component system is a good example and is conserved in all actinomycetes, including unicellular mycobacteria and corynebacteria . In Mycobacterium tuberculosis, the sensor kinase MtrB is activated when it interacts with the cell division machinery and it then phosphorylates and activates its partner response regulator, MtrA, which modulates target gene expression to block DNA replication and accelerate cell division. When the signal abates, MtrB switches activity to dephosphorylate MtrA and switch off the response [28,29]. In S. coelicolor and S. venezuelae, MtrA binds to the promoter regions of genes in ∼85% of their secondary metabolite BGCs. Deletion of mtrB constitutively activates MtrA and results in increased production of antibiotics, including cryptic compounds. Constitutively active ‘gain-of-function’ MtrA proteins also activate antibiotic production in Streptomyces species and might be useful tools for activating cryptic BGCs in all actinomycetes . An additional example of a global regulator that controls developmental progression is the BldD transcriptional regulator where bldD mutants lack aerial hyphae . However, in Saccharopolyspora erythraea, the commercial producer of erythromycin, BldD, binds directly to five promoters in the erythromycin gene cluster and is essential for activation of erythromycin production .
Genome mining for novel BGCs
The explosion in bacterial genome sequencing combined with easy to use computing tools like antiSMASH  enables scientists to ‘mine’ genomes and metagenomes for novel BGCs . Genome mining means the identification and activation of interesting and novel BGCs from genome or metagenome sequence data. (for a recent review of the methods available to activate cryptic BGCs, see ref. ) Non-ribosomal peptide synthetase (NRPS) and polyketide synthase (PKS) BGCs are widespread and easy to identify in actinomycetes and they are well known to encode bioactive NPs. The real challenge is in identifying NP classes that have not previously been well characterised because they are more likely to encode novel chemical scaffolds. A good example are the RiPPs (ribosomally synthesised and post-translationally modified peptides) which represent an underexplored source of bioactive NPs . Purification and heterologous expression of RiPPs is more straightforward than PKS and NRPS complexes, because the latter are typically large, multisubunit enzymes. The wide range of potential modifications to RiPPs also provides scope for modifying existing compounds to make new drugs. The first RiPPs to be identified from bacteria were the bacteriocins, which are antimicrobial peptides produced by almost all bacteria during normal environmental competition. Bacteriocins range from simple peptides to complex high molecular mass proteins that exert their effects via specific receptors on the surface of susceptible bacteria, and many have been repurposed to treat infections, aid cancer treatment and extend the shelf-life of food products . More than 20 novel subgroups of RiPPs have since been identified , starting with the antibiotic thiostrepton which was presumed to be a non-ribosomal peptide but could not be attributed to an NRPS gene cluster . In fact, the thiostrepton peptide is encoded by a gene and represents a new family of RiPPs, which is distinct from the bacteriocins.
Most RiPPs are encoded as relatively simple, long precursor peptides with a leader peptide usually appended to the N-terminus. The leader peptides are recognised by a variety of post-translational modification enzymes that cleave, alter and sometimes cyclise the precursor peptide to produce a vast range of biosynthetic variability that would not be possible by peptide synthesis alone. For example, the C-terminal tail of lassopeptides is threaded through a macrolactam ring closed by an isopeptide bond between the N-terminal and a carboxylic acid side group to form a unique ‘threaded loop’ . Thiopeptides all contain a macrocyclic core with a six-membered nitrogen-containing ring, but hugely complex and diverse structures are achieved through modifications of the various side chains. They are so complex that the structure of thiostrepton from Streptomyces azureus was not solved until nearly 20 years after its initial discovery . Perhaps, the most complex of the RiPPs are the bottromycins, which represent a promising new class of antibiotics against drug-resistant pathogens such as methicillin-resistant Staphylococcus aureous (MRSA) and VRE (vancomycin-resistant enterococci) . Instead of an N-terminal leader peptide, the precursor has a C-terminal ‘follower’ peptide which is required for post-translational processing . There is much interest in the bottromycins, but they are unstable in blood plasma and efforts to synthetically increase stability need to be explored before they can be introduced into the clinic.
Searching for new strains in underexplored environments
In addition to genome mining old strains for novel BGCs, there are also great efforts underway to isolate new bioactive strains from underexplored environments. Although BGCs appear to be shared widely between actinomycetes, often these are relatively recent events so identifying new phyla of actinomycetes like Salinospora can lead to the discovery of novel chemistry . In addition to this, our inability to culture most bacterial phyla means that even terrestrial environments are vastly under sampled. For example, metagenetic comparison of PKS and NRPS gene clusters in soils taken from a range of environments on five continents revealed that the DNA sequences of these genes varied widely among the soil samples, except for samples that came from similar environments that were near to each other. The authors concluded that most soil microbial communities will encode largely distinct collections of bacterial secondary metabolites and such studies can be used to guide discovery efforts for novel NPs . As described above, half of all known antibiotics come from soil-dwelling Streptomyces species, but this is at least in part because these bacteria are easy to culture from soil and grow faster than other filamentous actinomycetes. Some useful NPs have been discovered by cloning BGCs from environmental DNA (eDNA), and this includes the type 2 polyketide fasamycins (Figure 3A) which target fatty acid biosynthesis and have activity against MRSA . Techniques have also been developed to isolate hard to culture bacteria from soil, and the best known of these is the iChip which allows bacteria to grow in their natural soil environment . The iChip was used to isolate teixobactin (Figure 3A) from an ‘unculturable’ strain of soil Proteobacteria. Teixobactin represents a new class of NPs and has activity against MRSA and M. tuberculosis. However, scaling up to industrial production of NPs from hard to culture bacteria is likely to be problematic, if not impossible .
Novel soil NPs.
There are a few terrestrial environments which were neglected in the Golden Age, including the Atacama Desert in Chile. Atacama is the driest place on Earth and was thought to be devoid of life due to extreme exposure to UV radiation, inorganic oxidants, high salinity and scarcity of organic carbon . However, despite having a low diversity of bacteria, the Atacama soil is populated by filamentous actinomycetes which are taxonomically novel. For example, Streptomyces leeuwenhoekii was isolated from these soils and shown to make many novel NPs including chaxamycins and chaxalactins (Figure 3B) [49,50]. Another Atacama strain, Streptomyces DB634, produces novel NPs called abenquines which have activity against bacteria and fungi . The abenquines have a novel structure consisting of an amino acid attached to an N-acetyl-aminobenzoquinone moiety (Figure 3B). Novel actinomycetes have also been isolated from Saharan Desert soils, including a Saccharothrix strain that produces two novel polyketides, called antibiotics A4 (Figure 3B) and A5, that have broad-spectrum antifungal and antibacterial activities [52,53]. Thus, even extreme environments where life cannot thrive can still be home to actinomycete bacteria and this reflects their adaptation to life in hostile and variable environments. This point is further demonstrated by the fact that marine, salt-dependent, Streptomycetes and other filamentous actinomycetes have been isolated in large numbers over the last few decades. These strains were largely ignored in the 20th century because they were thought to be terrestrial strains that entered marine environments through surface run off.
In fact, the oceans are a rich source of novel NPs  and new genera of actinomycetes such as Salinospora have been detected in marine sediments at depths of up to 5669 m . The lomaiviticins (Figure 4A) were the first compounds of significance to be discovered from this genus due to their potent antimicrobial and anticancer properties, followed by the salinosporamides (Figure 4A) which are interesting due to the presence of a rare γ-lactam β-lactone bicyclic ring fusion and their anticancer properties . Anthracimycin was identified from a marine Streptomyces strain and shows potent activity against both B. anthracis and MRSA as well as some activity against Gram-negative bacteria . Anthracimycin is a macrolide antibiotic with an unusual structure, consisting of tricyclic rings with an incorporated decalin moiety . Another important set of compounds are the abyssomicins (Figure 4A), first purified from a marine Verrucosispora strain isolated from sediment in the Japanese sea . Abyssomicins inhibit biosynthesis of p-aminobenzoate, an intermediate in the biosynthesis of folic acid. Bacterial symbionts of marine animals are also a rich source of new bioactive NPs , for example, Salinospora and Streptomyces species have been discovered in marine sponges from the phyla Porifera and Haliclona, which are thought to select for antibiotic-producing bacteria to defend themselves [60,61]. These strains contain a wealth of novel BGCs such as rifamycin-like PKS genes and novel NRPS scaffolds [60,62–64]. Prodigine compounds have been isolated from a sponge-derived Saccharopolyspora strain, including the known anticancer compounds metacycloprodigiosin and undecylprodigiosin . Streptomycindole is a novel indole alkaloid from a sponge-associated Streptomyces showing antibacterial activity and manzamine A (Figure 4B); another alkaloid was isolated from a sponge-associated Micromonospora with antitumour and antimalarial activity [66,67]. The lobophorins are polyketides containing a tetronic acid y-lactone ring (Figure 4B) made by marine actinomycetes including Streptomyces strains associated with sea water, multicellular algae, molluscs or sea sponges [68–71]. Urauchimycins A and B are novel antimycin-like compounds isolated from a Streptomyces species of sponge origin. They contain a novel branched side chain moiety (Figure 4B) . They show antifungal activity, which is typical of antimycins, and these compounds have also been identified in fungus-growing attine systems [73,74].
Novel marine NPs.
The attine ants are endemic to Central and South America and developed fungiculture 50–60 million years ago . They use coevolved actinomycete symbionts to protect their fungal cultivar from invading pathogens [76,77]. Their fungus, Leucoagaricus, is extremely susceptible to coevolved fungal pathogens in the genus Escovopsis, which can cause whole colonies to collapse [75,78,79]. To prevent infections, the ants groom and weed their fungus gardens and host an antifungal-producing Pseudonocardia strain on their cuticles which is vertically transmitted by the queens, along with the fungal cultivar . Foraging worker ants also acquire antimicrobial-producing Streptomyces bacteria from the soil, and these may offer further protection to their food fungus [73,77,80–82]. The coevolved Pseudonocardia bacteria encode some intriguing and novel antifungal compounds, including the depsipeptide gerumycins, the first of which to be discovered was dentigerumycin (Figure 5) , and it was subsequently shown that other Pseudonocardia mutualists of attine ants make variations of dentigerumycin named gerumycins A–C . Mutualists associated with the more highly derived leafcutter ant genus Acromyrmex, however, produce a novel nystatin-like antifungal called nystatin P1 (Figure 2), which has an additional hexose sugar attached to the mycosamine of nystatin A1 . A similar molecule, called NPP, is produced by Pseudonocardia autotrophica, a strain of unknown origin , and is 300× more soluble than the clinically relevant nystatin A1, presumably due to the additional sugar . This represents a marked natural improvement of an existing drug. There are further variations in nystatin-like polyenes in the attine ant system, including a shorter molecule called selvamicin (Figure 2) made by a Pseudonocardia mutualist of Apterostigma ants . Pseudonocardia mutualists of Acromyrmex echinatior also encode a nystatin-like PKS which has a reduced number of modules and is predicted to make a shorter polyene with novel modifications of the backbone . Many plant-ants also cultivate fungi as a source of food in specialised host plant structures called domatia. The domatia and worker ants are associated with actinomyete bacteria, but it is not yet clear if they benefit the ants and their fungiculture [89–91]. Nevertheless, Streptomyces strains isolated from Kenyan Tetraponera penzigi and South American Allomerus plant-ants have potent antibacterial and antifungal activity [90–93].
Novel NPs from insect–microbe interactions.
Other insect systems are known or suspected to use actinomycete-produced antibiotics [94,95]. Notable examples include digger wasps (genera Philanthus, Trachypus and Philanthinus) which grow endosymbiotic Streptomyces bacteria in their antennal glands . The bacteria are smeared onto the ceiling of each brood cell and the larvae incorporate the bacteria into their cocoon walls where they produce antibiotics that serve a protective role, preventing infection of the developing larva. The antibiotics they make are a cocktail of known NPs, including streptochlorin and a mixture of eight piericidins . Bark beetles that attack and kill pine trees also have a tight association with different fungi including the mutualists Entomocorticium sp. and Ceratocystiopsis brevicomi sp. [98,99]. Entomocorticium is grown in specialised mycangia compartments and serves as an essential food source, but it is susceptible to infection by a specialised pathogenic fungus. To protect its fungal cultivar, the beetle Dendroctonus frontalis carries Streptomyces thermosacchari which makes mycangimycin (Figure 5), a polyene peroxide with antifungal activity [100,101], and Streptomyces sp. SPB74 which makes frontalamides A (Figure 5) and B, which also have antifungal activity .
Four different arthropods including two wood-lice, a leaf beetle and a millipede species are known to have endosymbiotic Streptomyces anulatus strains that make endophenazines A–D and tubermycin B (phenazine-1-carboxylic acid) (Figure 5) [103–105]. When the endophenazine gene cluster was introduced into an S. coelicolor host strain, a new prenylated compound termed endophenazine E was produced . Endophenazine compounds exhibit antibacterial, antifungal and herbicidal activities [103,107]. As a final example of the isolation of novel compounds from actinomycetes, the Streptomyces species associated with the mud-dauber wasp make a novel compound sceliphrolactam [108,109] which is a polyene with a lactam incorporated into the macrocycle (Figure 5) and has antifungal activity, including against amphotericin-resistant Candida albicans. These examples show that symbiotic relationships between actinomycetes and eukaryotic hosts are a rich source of novel chemistry, and it appears that we have barely touched the surface in terms of understanding and exploring these mutualisms. Other niches which remain underexplored include plant endophytic actinomycetes, which represent a significant proportion of the root and rhizosphere microbiota, but this has been reviewed recently elsewhere [5,110].
Recent advances in genomic technologies has made the discovery of novel NPs easier than ever before, and we now have huge new realms of biosynthetic potential to explore as well as many underexplored environmental niches in which to search for new strains. However, it will take more than novel NP discovery to lift us from the current antimicrobial resistance crisis. Pathogenic bacteria and fungi will evolve resistance much faster than we can generate new antibiotics, so it is important that we implement more careful stewardship of their use. By prescribing multidrug therapies and using new and existing antibiotics in cyclical patterns, we may reduce the occurrence of resistance. Investing in quicker and more accurate detection methods for drug-resistant infections will alleviate the need for broad-spectrum antibiotics and allow more targeted therapies that do not damage the natural host microbiota. Simplifying the clinical trials process and extending patent lifetimes may also make antibiotic discovery more financially attractive to big pharmaceutical companies. Above all, there is an urgent need to address the question of who will pay for the next generation of antibiotics.
Most antibiotics are derived from the natural products (NPs) of filamentous actinomycete bacteria that were discovered in soil during a golden age of antibiotic discovery that lasted from 1940 to 1960.
In the 21st century, genome sequencing has revealed that actinomycetes only produce around 10% of the NPs they encode when they are grown in the laboratory. The rest are encoded by cryptic gene clusters that are not expressed under laboratory conditions.
Understanding and rewiring the regulation of cryptic biosynthetic gene clusters will therefore yield thousands of new natural products from known strains.
Using new techniques to culture novel actinomycete strains or clone gene clusters from terrestrial and marine environments and co-evolved symbiotic niches, in combination with tools to activate cryptic gene clusters, is likely to yield the next generation of antibiotics.
biosynthetic gene clusters
methicillin-resistant Staphylococcus aureous
non-ribosomal peptide synthetase
ribosomally synthesised and post-translationally modified peptides
The Authors declare that there are no competing interests associated with the manuscript.